Phosphate Sorption Speciation and Precipitation Mechanisms on Amorphous Aluminum Hydroxide

Xiaoming Wang 1,2 , Brian L. Phillips 3, Jean-François Boily 4, Yongfeng Hu 5, Zhen Hu 1, Peng Yang 2, Xionghan Feng 1, Wenqian Xu 6 and Mengqiang Zhu 2,* 1 Key Laboratory of Arable Land Conservation (Middle and Lower Reaches of Yangtze River), Ministry of Agriculture and Rural Affairs, College of Resources and Environment, Huazhong Agricultural University, Wuhan 430070, China; wangxm338@mail.hzau.edu.cn (X.W.); huzhen09@yeah.net (Z.H.); fxh73@mail.hzau.edu.cn (X.F.) 2 Department of Ecosystem Science and Management, University of Wyoming, Laramie, WY 82071, USA; pyang@uwyo.edu 3 Department of Geosciences, Stony Brook University, Stony Brook, New York, NY 11794-2100, USA; brian.phillips@stonybrook.edu 4 Department of Chemistry, Umeå University, SE-901 87 Umeå, Sweden; jean-francois.boily@umu.se 5 Canadian Light Source, University of Saskatchewan, Saskatoon, SK S7N 2V3, Canada; Yongfeng.Hu@lightsource.ca 6 X-ray Science Division, Advanced Photon Source, Argonne National Laboratory, Argonne, IL60439, USA; wenqianxu@aps.anl.gov * Correspondence: mzhu6@uwyo.edu; Tel.: +1-307-766-5523


Introduction
Phosphorus (P) is one of the most limiting soil nutrients for agricultural and ecosystem productivity.To maintain or increase the crop productivity, massive phosphorus (P) fertilizers are applied into soils, while the utilization efficiency of P is low, usually less than 20% [1], since the majority of P was strongly sorbed by soil minerals, e.g., clay minerals, aluminum (Al) and iron (Fe) oxides [2][3][4][5].This excess P could be a potential risk to the surface water quality due to transportation of P-sorbed colloidal particles by surface runoff and soil erosion [6].The interactions between P and soil minerals are controlled by various physiochemical factors, such as pH, P loadings, and mineral phases.The prediction and assessment of P fate and bioavailability require the detailed knowledge of P speciation on mineral surface, which still remains poorly understood for phosphate sorption on Al oxides, a common mineral in acidic soils.Phosphate mainly forms inner-sphere complexes on Fe and Al oxide surfaces via ligand exchange reaction [7][8][9][10][11][12][13][14][15].That phosphate formed bidentate-binuclear (BB) complexes on ferrihydrite surfaces is based on the P-Fe interatomic distance (3.23 Å) of the complexes determined by differential atomic pair distribution function (d-PDF) analysis [15,16].Additionally, Li et al. (2010 and2013) investigated the adsorption mechanisms of phosphate on various Al minerals including gibbsite, boehmite (γ -AlOOH), corundum (α-Al 2 O 3 ), and γ -Al 2 O 3 by combining 31 P solid state NMR and ATR-FTIR spectroscopy and quantum chemical calculations [17][18][19], and proposed a BB surface complex under various solution conditions, probably with minor monodentate-mononuclear (MM) complexes when samples were wet.However, the binding geometries of phosphate on Al minerals in those studies were inferred via indirect analytical approaches, and need to be confirmed.In addition, it remains unclear how the binding geometries of phosphate on Al minerals change with solution conditions.
Besides surface adsorption, phosphate can also form surface precipitates on Fe and Al oxide surfaces [16,17,[20][21][22][23][24][25].High-resolution solid-state NMR spectroscopy can be used to distinguish surface adsorption and precipitation of phosphate on Al oxides based on chemical shifts.By using 27 Al and 31 P NMR spectroscopy, a few studies have identified the formation of surface precipitates when phosphate adsorbs on various Al (hydr)oxides, including AAH, gibbsite, boehmite (γ-AlOOH), corundum (α-Al 2 O 3 ), and γ-Al 2 O 3 .The precipitation is favored at higher P loadings, longer reaction time, and lower pH [17,[22][23][24][25]. Phytate, an organic phosphate, also can form precipitates on AAH surfaces, probably involving ternary complexes as intermediate products [21].However, the previous studies only describe the surface precipitates qualitatively, while the quantitative information about the precipitation with various environmental conditions remains unknown.
As a common Al mineral in soils and sediments, AAH can form at the earlier weathering process of K-feldspar and is a superior sorbent compared to other Al minerals because of its higher surface site density and reactivity [13,26,27].Thus, we aim to determine the effects of pH and P loading on the structure of phosphate complexes and the amount of surface precipitates on AAH surfaces, and the nature and the formation mechanisms of surface precipitates.To reach these aims, interactions between phosphate and AAH were studied over wide pH (pH 3-8) and P concentration ranges (0.03-15 mM).The experimental conditions bracket variations in P concentration and soil pH in natural and agricultural soils.For example, the high P concentration (15 mM) used in the experiment can occur in the proximity of P fertilizer grains right after P fertilizer application [28].The surface species, i.e., surface complexes and precipitates, were thoroughly characterized by synchrotron-based XRD, (d)-PDF, 31 P and 27 Al NMR, P K-edge XANES, and ATR-FTIR spectroscopy.

Synthesis of Amorphous Al(OH) 3 and Amorphous AlPO 4
Amorphous Al(OH) 3 (AAH) was synthesized by dropwise adding of 0.5 M NaOH into 300 mL of 0.167 M AlCl 3 solution until pH 6 under stirring conditions at room temperature (21 ± 0.5 • C).After the suspension pH stabilized, the suspension was further stirred for 24 h.Amorphous AlPO 4 (AAP) was prepared at room temperature by adding dropwise 0.5 M NaOH into 1 L deionized water containing 5 g AICI 3 •6H 2 O and 9 g KH 2 PO 4 until pH 3.8 was reached.The suspension was then further stirred at room temperature overnight [22].The obtained AAH and AAP solids were washed with deionized water until the electric conductivity of the supernatant dropped below 20 µS/cm and then air dried and stored at 4 • C in a refrigerator.

Phosphate Sorption Isotherms and Kinetics
The sorption isotherms were obtained by reacting 2 g/L AAH with 0.03-15 mM phosphate in 10 mM NaNO 3 solution at pH 3-8 and room temperature.The P solution was taken from a stock solution of 25 mM Na 3 PO 4 that was stored at 4 • C and used within one day.The total suspension volume was 40 mL.Each P-loaded suspension was shaken for 24 h, during which the pH was maintained at the target value by adding 0.1 or 1 M HNO 3 or NaOH solution.After that, each suspension was vacuum filtered through 0.22 µm membrane to minimize the residual P solution.The obtained wet solids on the membrane were air-dried, ground, and stored at 4 • C for the following characterizations.The sorption samples were air dried because wet samples cannot be used for the PDF analysis.Vacuum filtration was used to minimize the amount of the residual solution to mitigate the potential precipitation of P from the solution during drying.Total dissolved P and Al concentration in the filtrates was measured by inductively coupled plasma optical emission spectrometer (ICP-AES).
The sorption kinetics were evaluated by reacting 2 g•L −1 AAH with 8 mM P in 10 mM NaNO 3 solution at pH 5 and room temperature.The total volume of the suspension was 200 mL and the reaction lasted for 96 h under stirring conditions.At pre-determined time intervals, a 3 mL suspension was sampled and immediately filtered through a 0.22 µm membrane syringe filter, and the dissolved P and Al in the filtrate were measured by ICP-AES.Each isotherm and kinetic run was made in duplicates.At selected time points, a 20 L suspension was withdrawn and vacuum filtered through 0.22 µm membrane, and the membrane-loaded wet solids were treated as the isotherm samples for subsequent P speciation and solid characterizations.

XRD, PDF and Infrared Spectroscopy
Total X-ray scattering data for the air-dried isotherm samples were collected using X-rays of 58.66 keV (λ = 0.2114 Å) for both XRD and PDF analyses at beamline 11-ID-B at the Advanced Photon Source (APS), Argonne National Laboratory.The procedures for data collection and processing were described previously [29].d-PDFs were obtained by subtracting the PDF of the unreacted AAH from the PDF of each sorption sample by minimizing the intensity of edge-sharing Al-Al peak at 2.85 Å.We did not minimize the Al-O peak intensity at 1.89 Å to get d-PDFs because precipitates and the ternary complexes, if present, also contain the Al-O peak; otherwise, these species cannot be identified [16].The XRD patterns of the dried and wet kinetic samples were collected using X-rays of 28.84 keV (λ = 0.4521 Å) at beamline 17-BM-B at APS.The detailed methods and procedures are described in the Supporting Information (SI-1).
Attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectra of the dried isotherm samples were collected with a PerkinElmer spectrometer using the single bounce model with a diamond internal reflection element (IRE) at ambient temperature.The samples were coated directly on the diamond crystal surface and pressed tightly by a pressure head.Thirty-two scans (~1.5 min) were collected against air and the diamond background for each sample with a resolution of 4 cm −1 .
Multivariate curve resolution (MCR) [30] analyses of both PDFs and ATR-FTIR spectra were carried out to assist in our further interpretation of the experimental data by extracting linearly-independent components explaining the variance in the data.All calculations were made using the program MCR-ALS [30] in the computational environment of MATLAB 2016 (The Mathworks, Inc., Natick, MA, USA).The detailed methods and procedures are similar as those described in our recent study [31].

Phosphorus K-Edge XANES Spectroscopy
Phosphorus K-edge X-ray absorption spectra of the dried isotherm samples and AAP were collected at the soft X-ray microcharacterization beamline (SXRMB) equipped with Si (111) and InSb (111) double-crystal monochromators at the Canadian Light Source (CLS), Saskatoon, Canada.The operating condition of the storage ring in CLS was 2.9 GeV with a maximum current of 250 mA.The AAP and phosphate sorption samples were spread on P free carbon tape adhered to a copper stick and measured in vacuum.The spectra were collected in fluorescence mode using a four-element Si-drift fluorescence detector.Multiple scans were collected for each sample.The program SixPack [32] was used to average the XAS spectra, whereas background removal and post-edge normalization of the averaged spectra, as well as XANES linear combination fitting (LCF) analysis in the energy range of 2132-2182 eV, were performed with the program Athena [33].2.5. 27Al and 31 P NMR Spectroscopy Solid-state 27 Al and 31 P single-pulse MAS (SP/MAS) NMR spectra of the dried isotherm samples and references were collected on a 500 MHz (11.7 T) Varian Infinity-plus spectrometer.The 31 P and 27 Al SP/MAS NMR spectra were collected at operating frequencies of 202.32 MHz and 130.23 MHz respectively using Varian/Chemagnetics T3-type sample probe assemblies, with samples contained in 4 mm (o.d.) ZrO 2 rotors at a spinning rate of 15 kHz for the 27 Al experiments and 10 kHz for 31 P.The 31 P chemical shifts (δ P ) are reported relative to an external 85% H 3 PO 4 solution, and the 27 Al chemical shifts (δ Al ) relative to Al(OH 2 ) 6 3+ (aq) in the form of a 0.1 M AlCl 3 solution.The 27 Al SP-MAS spectra were acquired under quantitative acquisition conditions, using 0.5 µs pulses (B1 field corresponding to 4.5 µs non-selective 90 • pulse) and 0.5 s relaxation delays.Acquired with these parameters, the spectra were fully relaxed.Complete relaxation under these conditions was verified by acquiring a series of spectra with longer pulses (1 µs) and relaxation delays varying from 0.5 to 20 s, which showed no significant change in intensity.The 31 P spectra were acquired using 5 µs 90 • pulse and relaxation delays of 80-120 s.The 31 P relaxation was estimated from saturation-recovery experiments which yielded T 1 values that varied from 17 s to 34 s among the samples.These acquisition conditions correspond to approximately 95% recovery, but no differential relaxation effects were observed in the spectra profiles.

Sorption Kinetics and Isotherms
For the sorption kinetics at pH 5, phosphate sorption increases sharply within the first 10 h but slowly afterwards (Figure 1a).The phosphate 24 h isotherms at various pH values indicate that phosphate sorption is favored at lower pH.The 24-h sorption capacities (Q max ) obtained from the Langmuir model are 4.63, 3.97, and 3.80 mmol/g at pH 3, 5, and 7, respectively (Figure 1b).
In the absence of phosphate, the dissolved Al III in the bulk solution decreases with increasing pH and is not detectable at pH 7 (Figure 1c, see the first point of each condition).In the presence of phosphate, the dissolved Al III versus initial P concentration differs remarkably at different pH (Figure 1c).At pH 7, dissolved Al III is not detected in the presence of phosphate.At pH 5, the dissolved Al III decreases with increasing initial P concentration and is undetectable when the initial P concentration is higher than 1 mM P, suggesting that phosphate inhibits Al III release into the bulk solution.At pH 3, the dissolved Al III firstly decreases and then increases with increasing initial P concentration, indicating that low P loading inhibits the release of Al III into the solution, while high P loading promotes it.The increase of dissolved Al III  normalization of the averaged spectra, as well as XANES linear combination fitting (LCF) analysis in the energy range of 2132-2182 eV, were performed with the program Athena [33].

27 Al and 31 P NMR Spectroscopy
Solid-state 27 Al and 31 P single-pulse MAS (SP/MAS) NMR spectra of the dried isotherm samples and references were collected on a 500 MHz (11.7 T) Varian Infinity-plus spectrometer.The 31 P and 27 Al SP/MAS NMR spectra were collected at operating frequencies of 202.32 MHz and 130.23 MHz respectively using Varian/Chemagnetics T3-type sample probe assemblies, with samples contained in 4 mm (o.d.) ZrO2 rotors at a spinning rate of 15 kHz for the 27 Al experiments and 10 kHz for 31 P.The 31 P chemical shifts (δP) are reported relative to an external 85% H3PO4 solution, and the 27 Al chemical shifts (δAl) relative to Al(OH2)6 3+ (aq) in the form of a 0.1 M AlCl3 solution.The 27 Al SP-MAS spectra were acquired under quantitative acquisition conditions, using 0.5 μs pulses (B1 field corresponding to 4.5 μs non-selective 90° pulse) and 0.5 s relaxation delays.Acquired with these parameters, the spectra were fully relaxed.Complete relaxation under these conditions was verified by acquiring a series of spectra with longer pulses (1 μs) and relaxation delays varying from 0.5 to 20 s, which showed no significant change in intensity.The 31 P spectra were acquired using 5 μs 90° pulse and relaxation delays of 80-120 s.The 31 P relaxation was estimated from saturation-recovery experiments which yielded T1 values that varied from 17 s to 34 s among the samples.These acquisition conditions correspond to approximately 95% recovery, but no differential relaxation effects were observed in the spectra profiles.

Sorption Kinetics and Isotherms
For the sorption kinetics at pH 5, phosphate sorption increases sharply within the first 10 h but slowly afterwards (Figure 1a).The phosphate 24 h isotherms at various pH values indicate that phosphate sorption is favored at lower pH.The 24-h sorption capacities (Qmax) obtained from the Langmuir model are 4.63, 3.97, and 3.80 mmol/g at pH 3, 5, and 7, respectively (Figure 1b).
In the absence of phosphate, the dissolved Al III in the bulk solution decreases with increasing pH and is not detectable at pH 7 (Figure 1c, see the first point of each condition).In the presence of phosphate, the dissolved Al III versus initial P concentration differs remarkably at different pH (Figure 1c).At pH 7, dissolved Al III is not detected in the presence of phosphate.At pH 5, the dissolved Al III decreases with increasing initial P concentration and is undetectable when the initial P concentration is higher than 1 mM P, suggesting that phosphate inhibits Al III release into the bulk solution.At pH 3, the dissolved Al III firstly decreases and then increases with increasing initial P concentration, indicating that low P loading inhibits the release of Al III into the solution, while high P loading promotes it.The increase of dissolved Al III at high P loading and pH 3 might indicate the formation of more soluble Al-phosphate complexes.

X-Ray Diffraction
The synchrotron-based XRD patterns of the air dried sorption isotherm samples and the references are shown in Figure 2a-c.With increasing P sorption at pH 3, 5, and 7, a new broad diffraction peak appears and grows, suggesting formation of increasingly more surface precipitates.The peak position is close to that of the characteristic peak for AAP (~3.17 Å ) but not the same value, suggesting that the structure of the precipitates differs slightly from that of AAP.The peak position of the precipitate at lower pH is more similar to that of AAP, indicating that the precipitate formed at lower pH is more similar to AAP.Additionally, the diffraction peaks of AAH at ~2.16 Å and ~1.44 Å gradually decrease with increasing P loading, which can be ascribed to the formation of precipitates by the consumption of AAH.

X-ray Diffraction
The synchrotron-based XRD patterns of the air dried sorption isotherm samples and the references are shown in Figure 2a-c.With increasing P sorption at pH 3, 5, and 7, a new broad diffraction peak appears and grows, suggesting formation of increasingly more surface precipitates.The peak position is close to that of the characteristic peak for AAP (~3.17 Å) but not the same value, suggesting that the structure of the precipitates differs slightly from that of AAP.The peak position of the precipitate at lower pH is more similar to that of AAP, indicating that the precipitate formed at lower pH is more similar to AAP.Additionally, the diffraction peaks of AAH at ~2.16 Å and ~1.44 Å gradually decrease with increasing P loading, which can be ascribed to the formation of precipitates by the consumption of AAH.
The XRD patterns of selected kinetic samples are given in Figure 2d.With increasing sorption time, the XRD peak at ~3.17 Å becomes increasingly pronounced, suggesting formation of more precipitates.Additionally, this broad peak is also present for the wet sample at 96 h (Figure 2d), indicating that the precipitates already formed under the wet conditions prior to air drying.The XRD patterns of selected kinetic samples are given in Figure 2d.With increasing sorption time, the XRD peak at ~3.17 Å becomes increasingly pronounced, suggesting formation of more precipitates.Additionally, this broad peak is also present for the wet sample at 96 h (Figure 2d), indicating that the precipitates already formed under the wet conditions prior to air drying.

Pair Distribution Function Analyses
The PDFs of the dried sorption isotherm samples at pH 3, 5, and 7 and the references are shown in Figures 3a and S1a,c.With increasing P sorption, the P-O peak at 1.53 Å and O-O peak at 2.56 Å grow, while the Al-O peak at 1.89 Å and edge-sharing Al-Al peak at 2.85 Å and corner-sharing peak Al-Al at 3.45 Å attenuate, which are caused by more phosphate sorbed on AAH.The following d-PDF analysis is conducted to obtain the structure of the sorbed P species.
The d-PDFs of the sorption samples (Figures 3b and S1b,d) indicate that, with increasing P loading, the intensities of the P-O peak, Al-O peak, and the O-O peak all increase with relatively constant peak positions, and the P-Al peak grows but its peak position shifts from ~3.12 Å to ~3.17 Å , i.e., towards the P-Al peak at 3.18 Å of AAP.In addition, the whole profile increasingly resembles that of AAP.Therefore, these results suggest a transition from phosphate surface complexes, with a P-Al distance of ~3.12 Å , to surface precipitates with a P-Al distance similar to that of AAP with increasing P sorption, consistent with the XRD analysis.Additionally, the MCR analysis of PDF data for all sorption samples extracts three components, i.e., PDF-MCR-I, -II, and -III, corresponding to AAH, phosphate adsorption on AAH, and AAP precipitation, respectively (Figure 3c).The P-Al distance, obtained from the difference PDF profile of PDF-MCR-II and PDF-MCR-I (i.e., PDF-MCR-II-PDF-MCR-I) (Figure 3d), is similar to those of the sorption samples prepared at low P

Pair Distribution Function Analyses
The PDFs of the dried sorption isotherm samples at pH 3, 5, and 7 and the references are shown in Figure 3a and Figure S1a,c.With increasing P sorption, the P-O peak at 1.53 Å and O-O peak at 2.56 Å grow, while the Al-O peak at 1.89 Å and edge-sharing Al-Al peak at 2.85 Å and corner-sharing peak Al-Al at 3.45 Å attenuate, which are caused by more phosphate sorbed on AAH.The following d-PDF analysis is conducted to obtain the structure of the sorbed P species.
The d-PDFs of the sorption samples (Figure 3b and Figure S1b,d) indicate that, with increasing P loading, the intensities of the P-O peak, Al-O peak, and the O-O peak all increase with relatively constant peak positions, and the P-Al peak grows but its peak position shifts from ~3.12 Å to ~3.17 Å, i.e., towards the P-Al peak at 3.18 Å of AAP.In addition, the whole profile increasingly resembles that of AAP.Therefore, these results suggest a transition from phosphate surface complexes, with a P-Al distance of ~3.12 Å, to surface precipitates with a P-Al distance similar to that of AAP with increasing P sorption, consistent with the XRD analysis.Additionally, the MCR analysis of PDF data for all sorption samples extracts three components, i.e., PDF-MCR-I, -II, and -III, corresponding to AAH, phosphate adsorption on AAH, and AAP precipitation, respectively (Figure 3c).The P-Al distance, obtained from the difference PDF profile of PDF-MCR-II and PDF-MCR-I (i.e., PDF-MCR-II-PDF-MCR-I) (Figure 3d), is similar to those of the sorption samples prepared at low P concentration or high pH.This similarity further confirms that the PDF-MCR-II component corresponds to phosphate adsorption complexes on AAH.

Phosphorus K-Edge XANES Spectroscopy
All sorption samples and AAP exhibit a very weak pre-edge feature at ~2151 eV (peak A) (Figures 4a and S1a) [34], and its intensity remains almost constant at different conditions.In contrast, phosphate sorption on ferrihydrite shows a pronounced pre-edge peak, the intensity of which positively correlates with the number of Fe atoms bound to each PO4 [35,36].With increasing P loading on AAH or decreasing pH, the white-line peak B slightly shifts to a higher energy, and a shoulder peak at ~2162 eV (peak C) gradually appears and grows, increasingly resembling that of AAP.In addition, the pH and initial P concentration both show remarkable effects on the intensity and energy shift of peak C, exhibiting an obvious trend with the change of pH value (Figure 4a,b) and initial P concentration (Figure S2a,b), suggesting that lower pH and higher initial P concentration favor the formation of surface precipitates.
The LCF analysis of P K-edge XANES spectroscopy is used to quantify the proportions of the adsorbed P and precipitated P with the sorption sample at pH 8 with 2 mM P and AAP as the end members, respectively (Figure 4c).The high goodness of fit (Figure S3) suggests that a physical mixture of the two P standards can account for the P speciation variation.The AAP proportion in P decreases from 84.3% to 5.5% as the condition changes from pH 3 with 15 mM P to pH 7 with 2 mM P (Figure 4d).

Phosphorus K-Edge XANES Spectroscopy
All sorption samples and AAP exhibit a very weak pre-edge feature at ~2151 eV (peak A) (Figure 4a and Figure S1a) [34], and its intensity remains almost constant at different conditions.In contrast, phosphate sorption on ferrihydrite shows a pronounced pre-edge peak, the intensity of which positively correlates with the number of Fe atoms bound to each PO 4 [35,36].With increasing P loading on AAH or decreasing pH, the white-line peak B slightly shifts to a higher energy, and a shoulder peak at ~2162 eV (peak C) gradually appears and grows, increasingly resembling that of AAP.In addition, the pH and initial P concentration both show remarkable effects on the intensity and energy shift of peak C, exhibiting an obvious trend with the change of pH value (Figure 4a,b) and initial P concentration (Figure S2a,b), suggesting that lower pH and higher initial P concentration favor the formation of surface precipitates.
The LCF analysis of P K-edge XANES spectroscopy is used to quantify the proportions of the adsorbed P and precipitated P with the sorption sample at pH 8 with 2 mM P and AAP as the end members, respectively (Figure 4c).The high goodness of fit (Figure S3) suggests that a physical mixture of the two P standards can account for the P speciation variation.The AAP proportion in P decreases from 84.3% to 5.5% as the condition changes from pH 3 with 15 mM P to pH 7 with 2 mM P (Figure 4d).with the spectra of AAP as the reference.P K-edge XANES spectra of the end members of adsorption and precipitation speciation used in the linear combination fitting (LCF) analysis (c), and the obtained molar percentages of AAP in P (d) as a function of pH and initial P concentration.

31 P and 27 Al NMR Spectra
The 31 P NMR spectra of phosphate sorption on AAH as a function of P loading at different pH are shown in Figure 5a-c.The unreacted phosphate species has a chemical shift (δ P ) of +4-+5.5 ppm [37], which decreases after phosphate sorption on AAH.For the sorption samples with a lower P loading or higher pH, a broad peak with δ P at −7.4 ppm appears and the δ P values are similar with P sorption loading increasing from 0.25 mmol/g (pH 5, 0.5 mM P) (Figure 5b) to 0.52 mmol/g (pH 8, 2 mM P) (Figure 5c).As the P sorption loading continues increasing to 4.05 mmol/g (pH 3, 15 mM P) (Figure 5a), the peak becomes broader, and δ P shifts from −7.5 ppm to −12.8 ppm, increasingly close to δ P (−15.5 ppm) of AAP, indicating formation of more surface precipitates.The precipitates have more negative δ P values than adsorption complexes likely due to the more P-O-Al linkages per PO 4 tetrahedron involved in the structure of the precipitates [22].Additionally, for the 2 mM P systems with different pH values (Figure 5a-c), δ P decreases with decreasing pH, suggesting lower pH favors the formation of surface precipitates.

31 P and 27 Al NMR Spectra
The 31 P NMR spectra of phosphate sorption on AAH as a function of P loading at different pH are shown in Figure 5a-c.The unreacted phosphate species has a chemical shift (δP) of +4-+5.5 ppm [37], which decreases after phosphate sorption on AAH.For the sorption samples with a lower P loading or higher pH, a broad peak with δP at −7.4 ppm appears and the δP values are similar with P sorption loading increasing from 0.25 mmol/g (pH 5, 0.5 mM P) (Figure 5b) to 0.52 mmol/g (pH 8, 2 mM P) (Figure 5c).As the P sorption loading continues increasing to 4.05 mmol/g (pH 3, 15 mM P) (Figure 5a), the peak becomes broader, and δP shifts from −7.5 ppm to −12.8 ppm, increasingly close to δP (−15.5 ppm) of AAP, indicating formation of more surface precipitates.The precipitates have more negative δP values than adsorption complexes likely due to the more P-O-Al linkages per PO4 tetrahedron involved in the structure of the precipitates [22].Additionally, for the 2 mM P systems with different pH values (Figure 5a-c), δP decreases with decreasing pH, suggesting lower pH favors the formation of surface precipitates.
The 27 Al NMR spectra in Figure 5d-f further confirms the above variation of P speciation with P loading and pH.The 27 Al NMR spectrum of the sample prepared at pH 8 and 2 mM is similar to that of the pristine AAH (Figure 6b), indicating no or very little surface precipitates at this condition.With increasing P loading and decreasing pH (Figure 5d-f), the peak at +6.9 ppm becomes broader and a shoulder at ~−2.3 ppm gradually appears and shifts to a lower value, and a new weak peak at ~+42.4 ppm gradually appears.The differential 27 Al NMR spectra (Figure 6a) show that with increasing P loading these two new peaks gradually grow and shift to lower values, increasingly approaching those of AAP, suggesting the formation of more surface precipitates.Additionally, for the 2 mM P systems with different pH values (Figure 6b), the spectrum shows increased intensity from 0 to −10 ppm at lower pH compared to that of AAH, consistent with presence of precipitates at lower pH.The pristine AAH shows a major peak at δAl = +6.9ppm and two small peaks at δAl = +35.0ppm and +62.3 ppm (bottom of Figure 5d-f), respectively, corresponding to AlO6, AlO5, and AlO4 coordination [38].In contrast, AAP shows two strong peaks at −11.9 and +42.4 ppm, respectively, corresponding to AlO6 octahedra and AlO4 tetrahedra bonded to phosphate [22].According to the relative peak intensities or peak areas, AAP shows a much higher proportion of AlO4 than AAH.Thus, the formation of surface precipitates from AAH in the presence of phosphate involves the transformation of AlO6 octahedra to AlO4 tetrahedra.The asymmetric peak for the AlO6 in AAH structure is due to a wide distribution of second-order quadrupolar effects [22].The 27 Al NMR spectra in Figure 5d-f further confirms the above variation of P speciation with P loading and pH.The 27 Al NMR spectrum of the sample prepared at pH 8 and 2 mM is similar to that of the pristine AAH (Figure 6b), indicating no or very little surface precipitates at this condition.With increasing P loading and decreasing pH (Figure 5d-f), the peak at +6.9 ppm becomes broader and a shoulder at ~−2.3 ppm gradually appears and shifts to a lower value, and a new weak peak at ~+42.4 ppm gradually appears.The differential 27 Al NMR spectra (Figure 6a) show that with increasing P loading these two new peaks gradually grow and shift to lower values, increasingly approaching those of AAP, suggesting the formation of more surface precipitates.Additionally, for the 2 mM P systems with different pH values (Figure 6b), the spectrum shows increased intensity from 0 to −10 ppm at lower pH compared to that of AAH, consistent with presence of precipitates at lower pH.
The pristine AAH shows a major peak at δ Al = +6.9ppm and two small peaks at δ Al = +35.0ppm and +62.3 ppm (bottom of Figure 5d-f), respectively, corresponding to AlO 6 , AlO 5 , and AlO 4 coordination [38].In contrast, AAP shows two strong peaks at −11.9 and +42.4 ppm, respectively, corresponding to AlO 6 octahedra and AlO 4 tetrahedra bonded to phosphate [22].According to the relative peak intensities or peak areas, AAP shows a much higher proportion of AlO 4 than AAH.Thus, the formation of surface precipitates from AAH in the presence of phosphate involves the transformation of AlO 6 octahedra to AlO 4 tetrahedra.The asymmetric peak for the AlO 6 in AAH structure is due to a wide distribution of second-order quadrupolar effects [22].) and 27 Al NMR ((d) pH 3, (e) pH 5, (f) pH 7 and 8) spectra of the dried isotherm samples prepared at different pH with the spectra of AAH and AAP as the references.The data in each panel are stacked in the order of increasing initial P concentration from the bottom to the top.
The pristine AAH shows a major peak at δAl = +6.9ppm and two small peaks at δAl = +35.0ppm and +62.3 ppm (bottom of Figure 5d-f), respectively, corresponding to AlO6, AlO5, and AlO4 coordination [38].In contrast, AAP shows two strong peaks at −11.9 and +42.4 ppm, respectively, corresponding to AlO6 octahedra and AlO4 tetrahedra bonded to phosphate [22].According to the relative peak intensities or peak areas, AAP shows a much higher proportion of AlO4 than AAH.Thus, the formation of surface precipitates from AAH in the presence of phosphate involves the transformation of AlO6 octahedra to AlO4 tetrahedra.The asymmetric peak for the AlO6 in AAH structure is due to a wide distribution of second-order quadrupolar effects [22].  2Al NMR obtained by subtracting the AAH spectra from the spectra of selected isotherm samples, with gradually increasing P sorption loading from the bottom to the top, by minimizing peak intensity at 6.9 ppm (a), and the 27 Al NMR spectra of the 2 mM phosphate sorption at different pHs with the spectra of AAH as the reference (b).

ATR-FTIR Spectroscopy
With decreasing pH or increasing initial P loading (Figure 7a-c), the IR spectra of the dried isotherm samples evolve from that of pristine AAH towards spectra increasingly resembling those of AAP.Specifically, the intensities of the P-O stretching bands at ~1070 cm −1 , 925 cm −1 , and 627 cm −1 of AAP gradually increase, while the intensities of bulk OH deformation bands at 526 cm −1 and 952 cm −1 of AAH become lower.These trends reveal that more surface precipitates formed at lower pH or higher P loading.  2Al NMR obtained by subtracting the AAH spectra from the spectra of selected isotherm samples, with gradually increasing P sorption loading from the bottom to the top, by minimizing peak intensity at 6.9 ppm (a), and the 27 Al NMR spectra of the 2 mM phosphate sorption at different pHs with the spectra of AAH as the reference (b).

ATR-FTIR Spectroscopy
With decreasing pH or increasing initial P loading (Figure 7a-c), the IR spectra of the dried isotherm samples evolve from that of pristine AAH towards spectra increasingly resembling those of AAP.Specifically, the intensities of the P-O stretching bands at ~1070 cm −1 , 925 cm −1 , and 627 cm −1 of AAP gradually increase, while the intensities of bulk OH deformation bands at 526 cm −1 and 952 cm −1 of AAH become lower.These trends reveal that more surface precipitates formed at lower pH or higher P loading.The MCR analysis of the ATR-FTIR spectra of phosphate sorption on AAH extracted three components that best represent the pH and surface loading dependence of adsorbed and precipitated phosphate (Figure 7d), and their P loading-dependent fractions are shown in Figure S4.The IR-MCR-I component with the ~952 cm −1 and ~529 cm −1 band corresponds to the pure AAH, the fraction of which dominates at low P loadings.As the P sorption loading increases to ~1 mmol/g, it is gradually replaced by IR-MCR-II with bands centered at ~1040 cm −1 and 529 cm −1 , corresponding to phosphate surface complexes.With further increase of P sorption loadings, i.e., >1 mmol/g, the fraction of adsorbed species increases at pH 3, remains almost constant at pH 5, but decreases at pH 7. In contrast, the IR-MCR-III with bands centered at 1060 cm −1 , 915 cm −1 , 610 cm −1 , and 517 cm −1 , The MCR analysis of the ATR-FTIR spectra of phosphate sorption on AAH extracted three components that best represent the pH and surface loading dependence of adsorbed and precipitated phosphate (Figure 7d), and their P loading-dependent fractions are shown in Figure S4.The IR-MCR-I component with the ~952 cm −1 and ~529 cm −1 band corresponds to the pure AAH, the fraction of which dominates at low P loadings.As the P sorption loading increases to ~1 mmol/g, it is gradually replaced by IR-MCR-II with bands centered at ~1040 cm −1 and 529 cm −1 , corresponding to phosphate surface complexes.With further increase of P sorption loadings, i.e., >1 mmol/g, the fraction of adsorbed species increases at pH 3, remains almost constant at pH 5, but decreases at pH 7. In contrast, the IR-MCR-III with bands centered at 1060 cm −1 , 915 cm −1 , 610 cm −1 , and 517 cm −1 , corresponding to the precipitates, generally increase in prominence with increased P loading, and lower pH.

Structure of Phosphate Surface Complexes on AAH
The d-PDF analysis has been demonstrated to be an effective approach to determine element binding geometries on poorly-crystalline minerals [15,16,39,40].Li et al. [39] used d-PDF analysis for the first time in anion (arsenate) adsorption on an Al oxide to characterize structures of sorption complexes.In the present study, based on the d-PDF analysis, the P-Al distance of the phosphate complexes on AAH surface is determined to be ~3.12Å (Figure 3b and Figure S1b,d), in agreement with the calculated distance for BB complexes formed on Al oxides [18].Such complexes were also consistent with phosphate adsorption on other Al oxides [17][18][19].We do not observe a dependence of the binding geometry on either pH or P loading, consistent with previous studies [17,18], although a wet sample was found to possibly include minor (<3%) MM surface complexes in addition to the majority of BB complexes [18].
Moreover, the d-PDFs of the phosphate sorption samples prepared at low P loadings or high pH contain a strong peak at ~1.90 Å (Figure 3b and Figure S1d), likely corresponding to the Al-O pair of AlO 6 .Such a peak also occurs in the differential PDF of PDF-MCR-II minus PDF-MCR-I (Figure 3d).If phosphate only forms binary surface complexes, AlO 6 should be absent in the complexes.The presence of this peak suggests that PO 4 -Al III ternary surface complexes form besides the binary surface complexes.This peak is more significant for the adsorption sample prepared at pH 5 than at pH 8, consistent with lower pH favoring AAH dissolution, hence the formation of the ternary surface complexes.Similar ternary complexes also occur with phosphate sorption by ferrihydrite [16], but not for sulfate adsorption on ferrihydrite [41].

Quantitative Relationship between P Sorbed and P Precipitated
The phosphate exhibits a slowly increasing sorption process after 10 h (Figure 1a), which is probably due to micropore diffusion and/or the formation of precipitates [16,42].The Q max at different pHs (Figure 1b) are comparable with the reported values of phosphate sorption on AAH [43,44], but obviously higher than those of phosphate sorption on other Al minerals [17,18,26].
Interestingly, regardless of the sample preparation conditions, the percentage of precipitated P almost linearly correlates with the amount P sorbed for phosphate sorption on AAH or ferrihydrite surfaces [16] (Figure 8).At a given P sorption loading below 3.4 mmol/g, phosphate more readily precipitates on AAH surfaces than on ferrihydrite, which can be ascribed to the higher solubility product of AAH (Ksp = 1.3 × 10 −33 ) than that of ferrihydrite (Ksp = 2.79 × 10 −39 ) [43,45].However, at a given P sorption loading above 3.4 mmol/g, the result is contrary, i.e., phosphate more readily precipitates on ferrihydrite surfaces.The reason might be that the amorphous FePO 4 (AFP, Ksp = 9.92 × 10 −29 ) is easier to form due to its lower solubility product than AAP (Ksp = 9.83 × 10 −21 ) [46] at such high P loading.Thus, the solubility of both bulk material and the precipitated P-bearing phase control the amount of surface precipitates.In addition, the linear relationship indicates that the precipitate starts to form once the 24 h P sorption loading reaches ~0.248 mmol/g, which can be used to approximately estimate the threshold P loading for the formation of AAP on AAH surfaces, but precipitation is hard to detect at low P sorption loading.This threshold value of phosphate precipitation on AAH is lower than that (~0.945 mmol/g) for ferrihydrite [16], meaning that the P-Al precipitates preferentially form compared to the P-Fe phases in soils and sediments.The P precipitation percentage derived from P K-edge XANES LCF analysis, versus the amount of P sorbed (i.e., the total P removed from the solution) on AAH, fitted with linear equation (black line).Similar data for ferrihydrite from a previous study [16] is included for comparison (AFP: amorphous FePO4).

Al III Release during Phosphate Sorption and Formation Mechanisms of AAP
Besides the dissolved Al III detected in the bulk solution (Figure 1c), a large amount of Al III released from AAH occurs in the form of surface precipitates analogous to AAP (Figure 9a).Therefore, the total Al III released from AAH should include both dissolved Al III in the bulk solution and precipitated Al III .The released Al III increases with increasing initial P concentration at both pH 5 and pH 7 (Figure 9b), suggesting that phosphate adsorption promotes the Al III release.In contrast, the total released Al III at pH 3 firstly decreases and then increases with increasing initial P concentration (Figure 9b), similar to the variation of dissolved Al III (Figure 1c).These observations might be related to two processes.On one hand, phosphate surface complexes could block the dissolution of active surface sites [11,[47][48][49], thus inhibiting Al III release at a low P concentration and pH 3. On the other hand, the formation of AAP could consume the released Al III , promoting further Al III release at a high P concentration.Figure 8.The P precipitation percentage derived from P K-edge XANES LCF analysis, versus the amount of P sorbed (i.e., the total P removed from the solution) on AAH, fitted with linear equation (black line).Similar data for ferrihydrite from a previous study [16] is included for comparison (AFP: amorphous FePO 4 ).

Al III Release during Phosphate Sorption and Formation Mechanisms of AAP
Besides the dissolved Al III detected in the bulk solution (Figure 1c), a large amount of Al III released from AAH occurs in the form of surface precipitates analogous to AAP (Figure 9a).Therefore, the total Al III released from AAH should include both dissolved Al III in the bulk solution and precipitated Al III .The released Al III increases with increasing initial P concentration at both pH 5 and pH 7 (Figure 9b), suggesting that phosphate adsorption promotes the Al III release.In contrast, the total released Al III at pH 3 firstly decreases and then increases with increasing initial P concentration (Figure 9b), similar to the variation of dissolved Al III (Figure 1c).These observations might be related to two processes.On one hand, phosphate surface complexes could block the dissolution of active surface sites [11,[47][48][49], thus inhibiting Al III release at a low P concentration and pH 3. On the other hand, the formation of AAP could consume the released Al III , promoting further Al III release at a high P concentration.
The formation of the surface precipitates of analogous AAP could involve two pathways, including precipitation of dissolved Al III with soluble phosphate, like HPO 4 2− and H 2 PO 4 − , in the bulk solution and precipitation of the released Al III with sorbed P on AAH surfaces.Both pathways should occur at pH 3 and pH 5 where AAH solubility is relatively high, while the later pathway dominates at pH 7 because of the very low AAH dissolution.For the conditions under which precipitates occur at the mineral surfaces, the surfaces are likely covered by a mixture of bidentate-binuclear complexes, ternary surface complexes, probably pre-nucleation clusters [50], and surface precipitates with structure similar to that of AAP.The formed AAP might be more stable at ~pH 4 than other pH values because it was usually synthesized at this pH [22].The co-existence of various P and Al species can explain similar but different 31 P and 27 Al NMR spectra of the sorption products compared to those of AAP (Figures 5 and 6).These species of increasing condensation degree likely form sequentially with increasing reaction time and sorption loading [21,35].Within the mineral-water interface, P firstly forms surface complexes, inducing AAH dissolution.The sorbed P then provides sorption sites for the released Al III , resulting in the formation of (i) ternary complexes (e.g., (≡Al-O) 2 -PO 2 -Al), as well as (ii) pre-nucleation clusters that further condense to precipitates [16,21], a mechanism similar to those proposed for the formation of aluminum phytate by phytate sorption on AAH [21], and the formation of AFP during phosphate sorption on ferrihydrite.The formation rate and amount of AAP depends on the saturation state of the interfacial solution with respect to the solid phases [20,51], which favors at a higher initial P loading and a lower reaction pH.Such a coupled dissolution and precipitation mechanism is commonly used to explain surface precipitation at mineral-water interfaces, including of phosphate and other ions sorbed by various natural and synthetic Fe and Al oxides [20,21,[51][52][53][54].
Besides the dissolved Al III detected in the bulk solution (Figure 1c), a large amount of Al III released from AAH occurs in the form of surface precipitates analogous to AAP (Figure 9a).Therefore, the total Al III released from AAH should include both dissolved Al III in the bulk solution and precipitated Al III .The released Al III increases with increasing initial P concentration at both pH 5 and pH 7 (Figure 9b), suggesting that phosphate adsorption promotes the Al III release.In contrast, the total released Al III at pH 3 firstly decreases and then increases with increasing initial P concentration (Figure 9b), similar to the variation of dissolved Al III (Figure 1c).These observations might be related to two processes.On one hand, phosphate surface complexes could block the dissolution of active surface sites [11,[47][48][49], thus inhibiting Al release at a low P concentration and pH 3. On the other hand, the formation of AAP could consume the released Al III , promoting further Al III release at a high P concentration.Note that the molar amount of precipitated Al III is equal to that of precipitated PO 4 which was calculated by the ratio of P in the precipitates, obtained from the LCF analysis of P XANES spectra, multiplied by the P sorption loading.

Conclusions
Compared to other Al oxides and most synthesized or natural minerals, AAH shows a much higher phosphate sorption capacity, suggesting that AAH would be a highly efficient sorbent for immobilizing P after P fertilizer application to the soils.This partly accounts for the low utilization efficiency of P in the paddy soils with a large amount of amorphous Fe and Al oxides [55].Our study clarifies the structure of phosphate surface complexes and the identity of the surface precipitates, and suggests the existence of ternary complexes in a wide range of pH and P sorption loadings.Identification of theses surface species is useful to constructing physically meaningful models, such as surface complexation models and reactive-transport models [56,57].Moreover, the quantitative analysis of surface complex and surface precipitation of P indicate that, compared to ferrihydrite, the precipitates on AAH starts to form at a lower P sorption loading, suggesting that the AAP precipitates more readily than AFP near the phosphate fertilizer grains where dissolved P concentration could be high temporarily after fertilizer application.These new insights into the phosphate sorption on Al oxides improve our understanding of P mobility and bioavailability in Al oxide enriched soils and sediments.Further studies are required to assess the impacts of other environmental factors, such as the coexistence of natural organic matter or heavy metals, on the phosphate sorption on Al oxides, and to explore the effects of pH on the stability of AAP.

Figure 1 .
Figure 1.Phosphate sorption kinetic on AAH at pH 5 with 8 mM P (a) and sorption isotherms (24 h) at pH 3, 5, and 7, and their fits with the Langmuir equation [Q e = Q max K L C/(1 + K L C), where Q e is the amount of phosphate sorbed, Q max is the maximum sorption capacity, C is the equilibrium concentration, and K L is a constant related to the binding strength]; (b) and the corresponding dissolved Al III in the bulk solution (c) at different pH.Each value represents the mean of two replicates with standard deviation shown by error bars.

Figure 2 .
Figure 2. X-ray diffraction (XRD) patterns of the air dried phosphate sorption isotherm samples at pH 3 (a), pH 5 (b), and pH 7 (c), and the selected kinetic samples prepared at pH 5 with 8 mM P (d).AAH and AAP are included as references.

Figure 2 .
Figure 2. X-ray diffraction (XRD) patterns of the air dried phosphate sorption isotherm samples at pH 3 (a), pH 5 (b), and pH 7 (c), and the selected kinetic samples prepared at pH 5 with 8 mM P (d).AAH and AAP are included as references.

Soil
Syst. 2019, 3,  x FOR PEER REVIEW 7 of 17 concentration or high pH.This similarity further confirms that the PDF-MCR-II component corresponds to phosphate adsorption complexes on AAH.

Figure 3 .
Figure 3.The PDFs of the air-dried phosphate sorption isotherm samples at pH 5 (a) and the corresponding d-PDFs (b) obtained by minimizing the Al-Al peak at 2.85 Å ; the (d)-PDF data at pH 3 and pH 7 are in the supporting information.The MCR analysis of PDF data for all sorption samples identifies three components (c), i.e., PDF-MCR-I, -II, and -III.The difference PDF profile of PDF-MCR-II and PDF-MCR-I is shown in (d).The PDF data of AAH and AAP are included as the references.

Figure 3 .
Figure 3.The PDFs of the air-dried phosphate sorption isotherm samples at pH 5 (a) and the corresponding d-PDFs (b) obtained by minimizing the Al-Al peak at 2.85 Å; the (d)-PDF data at pH 3 and pH 7 are in the supporting information.The MCR analysis of PDF data for all sorption samples identifies three components (c), i.e., PDF-MCR-I, -II, and -III.The difference PDF profile of PDF-MCR-II and PDF-MCR-I is shown in (d).The PDF data of AAH and AAP are included as the references.

Figure 4 .
Figure 4. P K-edge XANES spectra in the energy range of 2145-2180 eV (a) and 2157-2180 eV (b) for the phosphate sorption isotherm samples prepared at different pH values and initial P concentrations

Figure 5 .
Figure 5. Solid-state 31 P NMR ((a) pH 3, (b) pH 5, (c) pH 7 and 8) and27 Al NMR ((d) pH 3, (e) pH 5, (f) pH 7 and 8) spectra of the dried isotherm samples prepared at different pH with the spectra of AAH and AAP as the references.The data in each panel are stacked in the order of increasing initial P concentration from the bottom to the top.

Figure 5 .
Figure 5. Solid-state 31 P NMR ((a) pH 3, (b) pH 5, (c) pH 7 and 8) and27 Al NMR ((d) pH 3, (e) pH 5, (f) pH 7 and 8) spectra of the dried isotherm samples prepared at different pH with the spectra of AAH and AAP as the references.The data in each panel are stacked in the order of increasing initial P concentration from the bottom to the top.

Figure 5 .
Figure 5. Solid-state 31 P NMR ((a) pH 3, (b) pH 5, (c) pH 7 and 8) and27 Al NMR ((d) pH 3, (e) pH 5, (f) pH 7 and 8) spectra of the dried isotherm samples prepared at different pH with the spectra of AAH and AAP as the references.The data in each panel are stacked in the order of increasing initial P concentration from the bottom to the top.

Figure 6 .
Figure 6.The differential27 Al NMR obtained by subtracting the AAH spectra from the spectra of selected isotherm samples, with gradually increasing P sorption loading from the bottom to the top, by minimizing peak intensity at 6.9 ppm (a), and the27 Al NMR spectra of the 2 mM phosphate sorption at different pHs with the spectra of AAH as the reference (b).

Figure 6 .
Figure 6.The differential27 Al NMR obtained by subtracting the AAH spectra from the spectra of selected isotherm samples, with gradually increasing P sorption loading from the bottom to the top, by minimizing peak intensity at 6.9 ppm (a), and the27 Al NMR spectra of the 2 mM phosphate sorption at different pHs with the spectra of AAH as the reference (b).

Figure 7 .
Figure 7. ATR-FTIR spectra for the dried isotherm samples at pH 3 (a), pH 5 (b), and pH 7 and 8 (c) with the spectra of AAH and AAP as the references.The MCR analysis of all ATR-FTIR spectra identifies three components, IR-MCR-I, -II, and -III, respectively (d).

Figure 7 .
Figure 7. ATR-FTIR spectra for the dried isotherm samples at pH 3 (a), pH 5 (b), and pH 7 and 8 (c) with the spectra of AAH and AAP as the references.The MCR analysis of all ATR-FTIR spectra identifies three components, IR-MCR-I, -II, and -III, respectively (d).

Figure 8 .
Figure8.The P precipitation percentage derived from P K-edge XANES LCF analysis, versus the amount of P sorbed (i.e., the total P removed from the solution) on AAH, fitted with linear equation (black line).Similar data for ferrihydrite from a previous study[16] is included for comparison (AFP: amorphous FePO4).

Figure 9 .
Figure9.The released Al III in the form of precipitates (a) and the total released Al III including both dissolved Al III and precipitated Al III (b) at different pH and initial P concentrations.Note that the molar amount of precipitated Al III is equal to that of precipitated PO 4 which was calculated by the ratio of P in the precipitates, obtained from the LCF analysis of P XANES spectra, multiplied by the P sorption loading.