Rationalization of the X-ray photoelectron spectroscopy of aluminium phosphates synthesized from different precursors

The aim of this paper is to clarify the assignments of X-ray photoelectron spectra of aluminium phosphate materials prepared from the reaction of phosphoric acid with three different aluminium precursors [Al(OH)3, Al(NO3)3 and AlCl3] at different annealing temperatures. The materials prepared have been studied by X-ray photoelectron spectroscopy (XPS), powder X-ray diffraction (XRD), infrared spectroscopy and high-resolution solid-state 31P NMR spectroscopy. A progressive polymerization from orthophosphate to metaphosphates is observed by XRD, ATR-FTIR and solid state 31P NMR, and on this basis the oxygen states observed in the XP spectra at 532.3 eV and 533.7 eV are assigned to P–O–Al and P–O–P environments, respectively. The presence of cyclic polyphosphates at the surface of the samples is also evident.


Introduction
There are a wide array of applications of phosphate based materials, including biomedical (due to the natural occurrence of phosphates in physiology and their biocompatibility 1-4 ), ceramics (due to their high strength and thermal stability) and refractory materials (in which phosphates are frequently used as binders 5 ). Curing temperatures of phosphates are critical to their performance in the latter applications and as a result there has been a concerted effort to understand the thermal evolution of phosphate systems. Heating induces a progressive polymerization starting from orthophosphates (PO 4 3À ), which polymerize to form a series of polyphosphates such as the pyrophosphates (P 2 O 7 4À ) and nally metaphosphates, which are long range cross-linked networks (PO 3 À ) n . 6 This paper is focused on aluminium phosphates generated from a mixture of phosphoric acid and three different aluminium based precursors: Al(OH) 3 , Al(NO 3 ) 3 and AlCl 3 . The majority of studies on aluminium phosphates have focused on bulk analysis techniques, 7 such as powder X-ray diffraction (XRD), solid-state 31 P NMR, FTIR and thermogravimetric methods, but the surface properties of these materials are also of interest and one of the most commonly used techniques to investigate this aspect is X-ray photoelectron spectroscopy (XPS). A seminal study by Gresch et al. 8 in 1979 on XPS of sodium phosphates provided well substantiated peak assignments for the oxygen region. This paper has since been extensively cited and used as a benchmark for XPS studies of phosphates. Gresch et al. proposed that the different states of oxygen created by crosslinking between phosphate units could be distinguished by their XP spectra, with the "bridging" oxygens appearing at higher binding energy (533.1-533.6 eV) than "non-bridging" oxygens (530.5-531.7 eV). This assignment was based on electronegativity arguments and spectra of model compounds. Concomitant with the shi in O(1s) binding energy, Gresch et al. reported that the P(2p) binding energy shis from 132.5 eV to 134.5 eV as the degree of P-O-P bridging bonds increases.
More recently, Crobu et al. 9 have also used the shi in O(1s) binding energy to assess the ratio of bridging to non-bridging oxygen in zinc polyphosphate glasses, correlating their results with secondary ion mass spectrometry measurements. They reported a shi from 532.2 eV to 534.0 eV as the reference material changed from an orthophosphate to a metaphosphate, with the P(2p) peak shiing from 134.0 eV to 134.8 eV.
Rotole and Sherwood, on the other hand, studied electrochemically deposited phosphate lms on aluminium substrates. 10 The XP spectra were referenced against data on aluminium orthophosphate 11 and metaphosphate 12 samples. They reported a constant P(2p) binding energy of 134.5 eV and a shi in the O(1s) peak from 531.4 eV for the orthophosphate to a broader peak at 531.8 eV for the metaphosphate. The latter peak clearly showed evidence for a second component at about 533.5 eV. However, the O(1s) data from the electrochemically treated surface were not so conclusive; the O(1s) spectrum of the "metaphosphate" was signicantly broader than that of the orthophosphate and seemed to consist of two components. It is not clear whether the component due to bridging oxygens is at higher or lower binding energy than the component due to terminal oxygens.
In the present study, we have used solid-state 31 P NMR, powder XRD and FTIR data to explore the structural changes that occur in aluminium phosphate materials synthesized from 3 : 1 mixtures of phosphoric acid and an aluminium precursor (either aluminium hydroxide, aluminium nitrate or aluminium chloride), which throw light on the information available from the surface specic XPS technique. Our results show that the expected polymerization occurs in materials prepared from all three precursors, and largely conrm the assignments of XP spectra based on existing literature. However, we also nd evidence for the presence, at the surface, of polyphosphate species that do not contain aluminium, which distorts the Al : O : P ratios established from the XPS spectra.

Synthesis of aluminium phosphate powders
The aluminium precursor AlX 3 , where X ¼ OH (Sigma-Aldrich, Reagent Grade), NO 3 (Sigma-Aldrich, ACS Reagent, >98%) or Cl (Sigma-Aldrich, 99.99%), was added to water (50 ml) at 298 K and mixed until dissolved. H 3 PO 4 (8.75 ml, Sigma-Aldrich, 85% wt, 99.99% purity) was added to the AlX 3 solution to give a P : Al ratio of 3 : 1 and the solution was allowed to mix for 30 min. The reaction mixture was then heated on a hotplate to evaporate the water, resulting in the formation of a viscous gel. Separate samples of the gel were heated in air in a furnace for a period of one hour at three different temperatures (300, 500 and 800 C), and then allowed to cool to room temperature before analysis. In the case of the materials prepared from Al(OH) 3 and Al(NO 3 ) 3 , a slightly wet white powder was obtained on heating to 300 C which was observed (by eye) to dry completely on heating at the higher temperatures. However, the samples prepared from AlCl 3 remained gel-like at 300 C. These samples were stable under vacuum and so could be analysed by XPS, but they had to be heated to 400 C to obtain a solid that was suitable for analysis by solid-state NMR and ATR could be obtained. In the discussion below, the materials prepared from the Al(OH) 3 , Al(Cl) 3 and Al(NO 3 ) 3 precursors are labelled AlP OH , AlP Cl and AlP NO 3 , respectively.

Materials characterization
XP spectra were recorded at room temperature on the powder samples using a Kratos Axis Ultra-DLD photoelectron spectrometer with a monochromatic Al Ka X-ray source in the "hybrid spectroscopy" mode resulting in an analysis area of 700 Â 300 mm 2 at a pass-energy of 40 eV for high-resolution scans and 160 eV for survey scans. The XPS data were analysed using CasaXPS 13 with all binding energies referenced to the C(1s) peak at 284.7 eV with an uncertainty of $0.2 eV. Since intensities for powder samples are dependent on the surface area analysed, which is poorly reproducible between different powder samples, the XP spectra shown in the gures are normalized to the point of maximum intensity. Curve ts were made using Gaussian-Lorentzian (GL (30)) line-shapes.
Powder X-ray diffraction (XRD) data were recorded at room temperature using a PANalytical X'Pert Pro diffractometer with a monochromatic Cu Ka source (l ¼ 0.154 nm) operating at 40 kV and 40 mA. The data were recorded over the 2q range 10-80 with a step size of 0.016 .
High-resolution solid-state 31 P NMR spectra were acquired at room temperature on a Chemagnetics Innity Plus spectrometer ( 31 P Larmor frequency, 121.50 MHz). The samples were contained in a 4 mm rotor with magic-angle spinning at 12 kHz. Methyldiphenylphosphine oxide (MDPPO) was used as a reference, with 31 P chemical shi at 30.8 ppm.
FTIR spectra were recorded using a germanium crystal ATR on a Varian 3100 Excalibur system with Varian Resolutions Pro soware.

Results
Bulk structural analysis using FTIR, XRD and solid-state 31 P NMR Throughout the discussion, we refer to each sample by the notation AlP OH (T), AlP Cl (T) or AlP NO 3 (T), where the subscript identies the precursor used in the synthesis and T denotes the annealing temperature. All the analysis was performed aer cooling to room temperature.
The crystalline phases present within each sample were investigated by powder XRD. Fig. 1, shows the XRD patterns for the AlP OH samples. For AlP OH (300) [i.e., the sample prepared from the Al(OH) 3 precursor and annealed at 300 C], only low intensity peaks are observed in the XRD data, and we have been unable to denitively match these peaks to a known structure. For the AlP OH (500) sample, the XRD data show sufficient crystallinity to allow Le Bail tting, although Rietveld renement was not possible. The Le Bail tting conrms that two distinct phases are present: a cubic aluminium metaphosphate d'Yvoire 14,15 also reported a pure cubic phase for a material prepared from Al(OH) 3 and annealed at 800 C, with a second phase present in the material annealed at 500 C. However, d'Yvoire assigned the second phase to a monoclinic structure rather than the aluminium hexacyclophosphate observed here.
XRD indicates that the AlP NO 3 (500) and AlP NO 3 (800) samples ( Fig. S3 and S4 †) are a mixture of the cubic aluminium metaphosphate and aluminium hexacyclophosphate phases [similar to AlP OH (500) but different from AlP OH (800)]. Among the samples prepared from the AlCl 3 precursor, the only crystalline product was AlP Cl (800), identied from XRD as pure cubic Al(PO 3 ) 3 (Fig. S5 †).
Further structural insights are obtained from highresolution solid-state 31 P NMR spectra (Fig. 2). For several of the samples, a peak at 0 ppm is present and assigned as the phosphoric acid starting material. As the annealing temperature increases, there is a trend towards increasingly negative 31 P chemical shis, attributed to polymerization. 16 For AlP OH (300), a broad set of overlapping peaks is observed between 5 ppm and À40 ppm, possibly suggesting an amorphous structure. The peaks at À21 ppm and À32 ppm (which represent ca. 20% of the total signal) are assigned 16 to aluminium tripolyphosphate (AlH 2 P 3 O 10 $H 2 O). These peaks are also observed for AlP NO 3 (300). For AlP OH (800), only one peak is observed (at 50.5 ppm) and is attributed unambiguously to cubic Al(PO 3 ) 3 , consistent with the presence of a single phosphorus environment in this structure (Fig. 3). This peak is also present for the AlP OH (500) sample, together with peaks at À36.5 ppm and À43.0 ppm; the area ratio for these two peaks is 2 : 1, consistent with the presence of three crystallographically distinct phosphorus environments in the aluminium hexacyclophosphate phase (Fig. 3). Monoclinic Al(PO 3 ) 3 , on the other hand, has 9 distinct phosphorus environments (Fig. 3).
For the AlP NO 3 (300) sample, three peaks are observed between 5 ppm and À35 ppm [in contrast to the overlapping set of peaks observed in this region for AlP OH (300)], including peaks at À21 ppm and À32 ppm assigned to aluminium tripolyphosphate (AlH 2 P 3 O 10 $H 2 O), which represents $50% of the signal. The peak at À3.4 ppm is attributed to some remaining aluminium orthophosphate. For the AlP NO 3 (500) sample, the 31 P NMR spectrum contains peaks characteristic of the hexacyclophosphate (43.3%) and cubic metaphosphate (52%). Annealing at a higher temperature does not complete the transformation to the cubic metaphosphate as the 31 P NMR spectrum for the AlP NO 3 (800) sample clearly contains peaks due to the hexacyclophosphate phase (ca. 36% of the signal intensity).
The 31 P NMR spectra for AlP Cl (400) and AlP Cl (500) are signicantly different from those observed for the AlP OH and AlP NO 3 materials. The major peaks are due to cubic metaphosphate (50.5 ppm), orthophosphate (1 ppm and 0 ppm), pyrophosphate (À12 ppm and À13.5 ppm) and polyphosphate (peaks in the range À20 ppm to À28 ppm), with only very weak peaks observed for hexacyclophosphate. In contrast, the AlP Cl (800) sample is a pure phase of cubic Al(PO 3 ) 3 . These observations suggest that the metaphosphate formed from the AlCl 3 precursor may be produced via a slightly different pathway than from the Al(OH) 3 and Al(NO 3 ) 3 precursors. We deduce that the stability of the Al-Cl bond hinders formation of aluminium phosphate from the Al(Cl) 3 precursor at lower temperatures, leaving the phosphoric acid to react mostly with itself to form varying degrees of polyphosphates. However, at the higher annealing temperature of 800 C, the phosphate transforms completely to cubic Al(PO 3 ) 3 .

Surface analysis with XPS
The XP spectra in Fig. 4 show the O(1s) data for all samples. For samples prepared from each of the three precursors, there is a general shi in peak position towards lower binding energy as the annealing temperature is increased. Curve tting conrms that two distinguishable peaks are present at binding energies of ca. 533.7 eV and 532.3 eV in all spectra, with a transfer of intensity from the higher binding energy peak to the lower binding energy peak as the annealing temperature is increased (peak area ratios are given in Table 1). In particular, we note the close similarity between the spectra for the AlP OH (800), AlP NO 3 (800) and AlP Cl (800) samples, all of which show an approximately 2 : 1 intensity ratio (lower : higher binding energy peaks) and resemble the O(1s) spectrum of aluminium metaphosphate published by Rotole and Sherwood. 12 The Al(2p) XP spectra for all samples are shown in Fig. 5. For the AlP OH samples, a strong peak is present for all annealing temperatures at $75.3 eV for AlP OH (300) and shiing slightly to Fig. 1 Powder XRD patterns recorded for the AlP OH materials, with annealing at different temperatures. The AlP OH (300) sample is mostly non-crystalline, but both cubic metaphosphate and hexacyclophosphate are present in the AlP OH (500) sample. The AlP OH (800) sample is a pure phase of the cubic metaphosphate. Data for the AlP Cl and AlP NO 3 materials are given in ESI. † Fig. 2 High-resolution solid-state 31 P NMR spectra of the aluminium phosphate materials prepared from the Al(OH) 3 , Al(NO 3 ) 3 and AlCl 3 precursors at different annealing temperatures. All spectra were recorded at room temperature.  For all three precursors, the intensity of the peak at lower binding energy increases relative to the peak at higher binding energy as the annealing temperature is increased.  (800) is not. The XP spectra in the P(2p) region (Fig. 6) have a single peak at $134.8 eV for all samples with a small shi ($0.2 eV) to lower binding energy as the annealing temperature is increased to 800 C. The observed peak is consistent with the average peak position for metaphosphates in the NIST database 17 (134.8 eV; s ¼ 0.5 eV) and with results of Rotole and Sherwood 10,11 on aluminium phosphates. However, it is in marked contrast to sodium phosphates, 8 for which the P(2p) binding energy shis by 2 eV from the orthophosphate (132.5 eV) to the oxygenbridged metaphosphate (134.5 eV).
The atomic ratios calculated from the XPS spectra (Table 1) are informative. The P : O ratio is very close to 1 : 3 for all samples but the P : Al ratio is always higher than 3 : 1. The fact that the XPS survey scans for materials prepared at lower annealing temperatures (Fig. S6 †) do not contain any signal for aluminium or for any other cation suggests that the surface is dominated by hydrogen phosphates, with the consistent O : P ratio of 3 : 1 indicating extensive polymerization at the surface. We also note that no XPS signal due to chlorine is observed for any of the samples (Fig. S7 †). Following annealing at 800 C, the samples from all precursors are highly crystalline, and the presence of the Al(2p) peak in the XP spectra suggests that the surface is now dominated by aluminium metaphosphate. However, the P : Al ratio remains higher than the expected 3 : 1 ratio, particularly for AlCl 3 (800), suggesting that some hydrogen polyphosphates are present at the surface.

Surface analysis with ATR-FTIR
The sampling depth of ATR-FTIR spectroscopy, typically between 0.5-2 mm, is signicantly larger than that for XPS, which detects only the top 2-4 nm of the surface for the elements studied here. Nevertheless, there is excellent agreement between the IR results and the results from the more surface sensitive XPS method, as illustrated in Fig. 7, which shows FTIR data for several samples. For AlP OH (300), broad, weak bands are present in the range ca. 1100-1250 cm À1 , characteristic of an aluminium orthophosphate with some indication of P-O-P bond formation from the weak peak at 1025 cm À1 . 18 The spectra for AlP OH (500) and AlP OH (800) are dominated by strong bands assigned to metaphosphates. In particular, the peak at 738 cm À1 is assigned to Al-O-P, bands at 811, 1025, 1060 and 1070 cm À1 are assigned to P-O-P modes, and bands at 1282 and 1305 cm À1 are assigned to P]O bonds in the aluminium metaphosphate.

Discussion
The structures of aluminium orthophosphate, aluminium hexacyclophosphate and two aluminium metaphosphates are shown in Fig. 3, and can be used to rationalize the changes in the XP spectra observed for different annealing temperatures. Small shis to lower binding energy are observed for the P(2p) and Al(2p) peaks as the annealing temperature is increased, but the most signicant changes arise in the O(1s) spectra, which show a transfer of intensity from a higher binding energy peak at 533.7 eV to a lower binding energy peak at 532.3 eV. The best starting point to understand these changes is the O(1s) spectra for all samples annealed at 800 C, as these spectra are all very similar to those reported by several other authors, including Gresch et al. for sodium metaphosphates, 8 Crobu et al. for zinc phosphates, 9 and Rotole and Sherwood for model aluminium phosphates. 11,12 The FTIR, XRD and solid-state 31 P NMR data indicate the presence of mainly cubic aluminium metaphosphate aer annealing at 800 C and we can therefore denitively assign the peaks in the O(1s) region at 533.7 eV and 532.3 eV, respectively, to oxygen atoms bridging between phosphorus atoms (P-O-P) and oxygen atoms bridging between phosphorus and aluminium atoms (P-O-Al), in agreement with Gresch et al.
In the metaphosphate, these two bonding environments are expected to be present in a 2 : 1 ratio of P-O-Al to P-O-P. However, as shown in Table 1, quantication of the XPS data for the samples annealed at 800 C gives a peak area ratio (532.3 eV : 533.7 eV) of ca. 1.6 : 1, whereas the expected ratio is 2 : 1. Thus, the P : Al ratio at the surface of these materials is higher than the expected 3 : 1 ratio. To understand these differences, we now consider the XP spectra recorded for samples annealed to lower temperatures.
A key observation is that the AlP NO 3 (300) and AlP Cl (300) samples show no evidence, in the XP spectra, for the presence of aluminium. The AlP NO 3 (500) sample does show evidence for aluminium, but the AlP Cl (500) sample does not. The absence of aluminium indicates a purely hydrogen terminated phosphate material at the surface. Unreacted phosphoric acid can be ruled out based on the O : P ratio of 3 : 1, but there is evidence from the solid-state 31 P NMR results for hydrogen terminated or cyclic polyphosphates (giving peaks at À28 ppm and À32 ppm) which would have a 3 : 1 ratio. A cyclic polyphosphate such as P 4 O 10 has a P-O-P to P]O bond ratio of 1.5 : 1, which could account for the XPS ratios if the oxygen in P]O has a binding energy of $532 eV, overlapping with the XPS peak for the oxygen in P-O-Al. This assignment would be in agreement with Gresch et al. 8 Finally, the "wet" physical appearance of samples annealed at lower temperatures is also consistent with the presence of hydrogen polyphosphates which would be poorly crystalline.
From the data presently available, we cannot determine whether annealing ultimately leads to sublimation or decomposition of the polymeric phosphates, or whether further reaction with unreacted aluminium precursor occurs. However, for samples annealed at 800 C, the XP spectra are consistent with the presence of aluminium metaphosphates although the slightly higher P : Al ratio in the case of the material prepared from AlCl 3 suggests the surface contains some hydrogen terminated polyphosphates.

Conclusions
Aluminium metaphosphate is formed from the reaction of phosphoric acid with three different aluminium compounds (Al(OH) 3 , Al(NO 3 ) 3 and AlCl 3 ) followed by annealing in air. XRD, XPS and FTIR measurements of the resulting materials show almost identical behaviour from all three precursors, but the solid-state 31 P NMR spectra are signicantly different at the lower annealing temperatures (300 C and 500 C). The unique solid-state 31 P NMR spectra of the materials annealed at lower temperatures indicates the presence of amorphous materials which would not be identied by XRD, but explains the lack of an Al(2p) signal in XP spectra of the materials prepared from the Al(NO 3 ) 3 and AlCl 3 precursors at lower annealing temperatures. For all three precursors, a cubic metaphosphate is produced on annealing at 800 C, with a hexacyclophosphate present at lower annealing temperatures in the case of the Al(NO 3 ) 3 and Al(OH) 3 precursors. The XP spectra in the O(1s) region of the aluminium phosphate materials show two components at 532.3 eV and 533.7 eV, which are denitively assigned to the P-O-Al and P-O-P bonding environments, respectively. However, samples annealed at lower temperatures also exhibit surface species assigned as cyclic polyphosphates, with binding energies of 532.3 eV and 533.7 eV for the P]O and P-O-P bonding environments, respectively.

Conflicts of interest
There are no conicts to declare. was funded by EPSRC and BBSRC (contract reference PR140003), as well as the University of Warwick including via part funding through Birmingham Science City Advanced Materials Projects 1 and 2 supported by Advantage West Midlands (AWM) and the European Regional Development Fund (ERDF).