Modified iron phosphate/polyvinyl alcohol composite film for controlled-release fertilisers

Traditional soluble phosphorous (P) fertilisers can be easily leached to pollute water systems, resulting in water eutrophication, a major environmental problem from the oversupply of unused nutrients. One innovative solution is to control the release of P upon demands of the plants. This study established a new concept of controlled-release P fertiliser via incorporation of ferric phosphate (FePO4) as a P source in polyvinyl alcohol (PVA) films, which can immobilise the FePO4 particles and stimuli-responsively accelerate their release rate in the presence of citric acid. More importantly, FePO4 used in this work originated from steelmaking slag as a potential waste reuse. Due to the low solubility of FePO4, diethylamine was introduced to modify FePO4 particles to facilitate the release of P before incorporating with PVA. The effects of diethylamine modification and the properties of FePO4/PVA films were systematically investigated through microscopic and spectroscopic methods. The release of P from particles and films was examined in both deionised water and citric acid solution for 30 days. The results showed a tenfold increase of the release rate of modified FePO4/PVA in citric acid solution compared with that in deionised water, and also a doubled release rate of the modified FePO4/PVA compared to that of FePO4/PVA in citric acid. The improved performance suggests that PVA can maintain the phosphorous content with exposure to water and expedite release in citric acid upon the demand of plants. This composite film offers a new opportunity for the application of insoluble phosphate as a phosphorous fertiliser.


Introduction
Phosphorous (P) as one of the most commonly used nutrients in agriculture can improve the nourishment of plants. Conventional soluble P fertilisers are made by dissolving phosphate rock (insoluble) in sulfuric acid 1-3 to convert P into a soluble form (hydrogen phosphate and dihydrogen phosphate) 4 which can be dissolved rapidly in water. However, soluble phosphate fertiliser (P-fertiliser) can be easily leached to pollute rivers and lakes, leading to water eutrophication and subsequent environmental issues. 5 The overdose of P-fertiliser may also cause fertiliser burn to plants. Therefore, a controlled-release formulation of P-fertiliser is required to provide this nutrient in a controllable fashion to minimize environmental and economic risks.
Due to limitations of soluble phosphate fertilisers, the use of controlled-release fertiliser has become a promising direction to improve and manage P release. 6,7 A traditional controlledrelease formulation contains a soluble phosphate core as the source of nutrient and a relatively insoluble shell as a release controlling buffer. The shell is commonly fabricated using sulphur 8 and polymer (e.g., polyolen, polyvinyl chloride (PVC), rubber, etc.). There are several problems with a sulphur-based shell. First, using sulphur as a shell may increase the acidity of soil, leading to calcium deciency. 9 Second, the mechanical property of a sulphur shell is poor for withstanding mechanical pressure during the fabricating or transporting processes. Third, once the sulphur shell is cracked by high osmotic pressure resulting from water absorption, all the loaded soluble fertiliser may be released instantly (catastrophic-type release) and the rate cannot be controlled. 10 Although polymer coatings show an outstanding swelling property, degradation of a polymer, such as PVC and rubber, is a major disadvantage in applications of this type of controlled release system. 11 In addition to the soluble fertiliser system, using insoluble salts like phosphate rock can offer a more controllable release rate and protection of phosphorous than soluble salts, as the release of phosphate is highly related to the solubility of a specic salt. Recently, ferric phosphate (FePO 4 ) has been widely used to replace scarce mineral sources such as phosphate rock, as it can be recycled from the by-products from steelmaking slag in the steel making industry, indicating a great environmental sustainability. 12 Besides the nutrient providing function, FePO 4 can also eliminate snails and slugs via interfering their metabolic systems. However, the low solubility of FePO 4 (0.6 g/100 mL at 100 C) 13 may limit its direct application in agriculture. Approaches have been developed to increase the availability of P and facilitate its release. Zhong et al. 14 embedded phosphate rock in sulfonated corn starch to achieve a release rate of 4.5 times more than the raw phosphate rock, but the sulfonate modication method may lead to soil acidulation. In addition, increasing the surface area may also improve the dissolution of insoluble salts as the result of an increase in available area for solvation. 15 Therefore, an agent that can modify the surface of the insoluble particles leading to a larger surface area is desired. Diethylamine (DEA) is one of the modication agents that have been used in catalysis to increase their surface areas, 16 and thus can be a potential modier for FePO 4 . However, the application of DEA on FePO 4 particles was merely investigated in the literature. Also, DEA can be toxic and corrosive if it is not well administered and its amount of use should be controlled within safety limits.
Washing off FePO 4 powder as a fertiliser is still an issue to be dealt with. A lm-based fertiliser system could potentially immobilise FePO 4 powders. In this case, polyvinyl alcohol (PVA) can be an ideal candidate as the polymer matrix to be incorporated with the phosphate salt, because of the nontoxic, biodegradable 17 and water-soluble 18 properties of PVA. Due to its excellent lm forming and high mechanical strength, 19,20 PVA can be applied as a sizing, coating, and thickening agent to package powders away from running water invasion. PVA-composited lm fertiliser can also be made into any shape and be easily sustained in the soil. Han et al. 21 prepared a controlled-release nitrogen fertiliser by coating a starch/PVA onto the soluble granules, and demonstrated that permeability of the NH 4 + is related to the thickness of the PVA coating package. Also, Noppakundilograt et al. 22 utilized PVA as a lm former and as the rst coating layer to form a 3-layer controlled-release fertiliser; the PVA-coated fertiliser displayed a 4.5 times slower rate than the uncoated one. Most recently, Sarkar et al. 23 coated a partially acidulated phosphate rock with PVA. The above studies used PVA as a coating lm to delay the dissolving of the soluble core; however, boosting PVA on release of an insoluble salt like FePO 4 has yet to be investigated.
Herein, an approach to synthesise a new lm-type phosphate fertiliser is proposed. FePO 4 is employed as phosphate source while the PVA acts as an auxiliary matrix. The modication of FePO 4 is examined by using different concentrations of the modier. The morphology and structures of modied FePO 4 (m-FePO 4 ) and PVA composited lm were systematically investigated. Release kinetics of m-FePO 4 /PVA fertiliser was carried out in deionised water as a control group and in solution of citric acid that mimics the environment around plant roots. This study will deliver a totally new platform for slag from industrial waste to be potentially used as P-fertiliser.

Preparation of m-FePO 4 /PVA composite fertiliser
The m-FePO 4 derived from 0.03 : 1 wt of DEA to FePO 4 ratio was selected as the best condition and a series of samples with different ratios between PVA and m-FePO 4 were prepared. Specically, 10.0 g of PVA was dissolved in 90.0 g of deionised water at 90 C for 3 h under stirring. The suspension of m-FePO 4 and PVA aqueous solution was mixed at m-FePO 4 /PVA (v/v) volume ratio of 3 : 1, 1 : 1, and 1 : 3, respectively (corresponding weight ratio of 75 : 1 wt, 25 : 1 wt, and 0.83 : 1 wt). The viscous mixture solution was further mixed on a Vortex Mixer (Ratek, Australia). The total volume of all the samples was kept at 4 mL. The m-FePO 4 /PVA mixtures were dropped (each droplet 0.2 mL) onto the surface of Paralm® (Bemis, USA) which provided a hydrophobic surface for the formation of composite lms. Aer drying for 24 h at room temperature, lms with a diameter of about 0.5 cm were removed from the Paralm® for further use. The procedure is shown in Fig. 1.

Characterization
Scanning electron microscopy (SEM) samples were coated with a thin layer of carbon (powder samples) or gold (lm samples) using an EM ACE600 sputter coater (Leica Microsystems, Wetzlar, Germany). SEM images were captured on a Supra 55VP (Zeiss, Oberkochen, Germany). The surface area of the samples was measured by a nitrogen adsorption instrument (ASAP 2020, Micromeritics, USA) at 77 K and calculated using the Brunauer-Emmett-Teller (BET) method. The cross-section of FePO 4 /PVA lm was obtained by fracturing the samples in liquid nitrogen. Modied FePO 4 powder was characterized by X-ray diffraction (XRD) (PANalytical, Netherlands) and the particle size was measured by Mastersizer (Malvern, UK). Attenuated total reectance-Fourier-transform infrared spectroscopy (ATR-FTIR) technique was used to characterize the chemical structure of the modied FePO 4 powder and FePO 4 /PVA lms via a Bruker Vertex 70 spectrometer (Massachusetts, USA). The spectra were obtained at a resolution of 4 cm À1 in a range from 600 to 4000 cm À1 . The percentage of FePO 4 in FePO 4 /PVA lm was conrmed by measuring absorbency with a USB-2000 UV-Vis (Ultraviolet and Visible Absorption Spectroscopy) spectrophotometer (Ocean Optics, Dunedin, FL, USA) aer dissolving the samples in hydrochloric acid (35 wt%). A calibration curve of FePO 4 was obtained through dissolving pure FePO 4 (>99%) in hydrochloric acid and taking 334.49 nm as the absorption peak.

Release of m-FePO 4 /PVA composite fertiliser in deionised water and in citric acid solution
The release behaviours of m-FePO 4 /PVA in deionised water and citric acid solution were examined. The content of FePO 4 in all the samples were kept identical. The m-FePO 4 /PVA lms were added into 15 mL centrifuge tubes with 2.5 mL of deionised water and 50 mmol mL À1 of citric acid solution, respectively. The release of P was monitored at room temperature and recorded for 30 days. On each day, 2 mL of the solution was removed to measure the concentration of phosphorous. Aer that, 2 mL of fresh deionised water or citric acid solution was added to maintain a constant volume. Phosphorus contents were estimated using colorimetric analysis. 25,26 Results and discussion

Morphology of the m-FePO 4
The morphology of the raw FePO 4 particles is shown in Fig. 2a, while Fig. 2b and S1 † illustrate the appearance of akestructures on the surface of the m-FePO 4 . With increasing of DEA content, the akes on the m-FePO 4 particles become more noticeable (Fig. S1 †). The surface areas of m-FePO 4 treated with different amounts of DEA were further measured. It can be seen from Fig. S2a † that the surface area increases with the amount of DEA. This could be attributed to the ake structure as observed under SEM. An increased surface area could offer more contacts with water, leading to an expedited dissolution of phosphate during the release process.
The distribution of particle size with different DEA contents is shown in Fig. S3. † The majority of m-FePO 4 particles have a diameter of around 50 mm. The volume fraction of small particles (diameters ranged from 0 to 25 mm) resulting from a small portion of DEA (from 0.01 : 1 wt to 0.05 : 1 wt of DEA to FePO 4 ratio) does not show a noticeable change, while a remarkable increase was observed at a higher DEA content (the DEA to FePO 4 ratio ranged from 0.05 : 1 wt to 0.1 : 1 wt). The increase in volume fraction of small particles could be due to the increased reaction of FePO 4 with DEA, which will be further discussed in the following sections. In addition, FTIR was employed to examine changes of m-FePO 4 with different contents of DEA. Fig. 3b shows the FTIR spectra of m-FePO 4 , which reveal bending and stretching vibrations of the phosphate group (PO 4 3À ) at 800 to 1200 cm À1 . 27 The intensity of the peak at 1000 cm À1 declines when the content of DEA increases, thereby indicating the decrease of phosphate content remaining in the FePO 4 particles with the addition of DEA. This is further conrmed by Fig. S2b. †

Reaction between DEA and FePO 4
A higher concentration of DEA (0.5 : 1 wt and 1 : 1 wt of DEA to FePO 4 ratio) and a strong base (sodium hydroxide) was introduced to investigate the reaction between DEA and FePO 4 . It can be seen from Fig. S4 † that the colour of the samples with 0.01 : 1 to 0.1 : 1 wt of DEA to FePO 4 ratio gradually changed from white to brown, while the colour turned to dark red for the samples treated with high amounts of DEA (0.5 : 1 wt and 1 : 1 wt of DEA to FePO 4 ratio). Similarly, the colour of dark red is observed with the adding of identical amounts of sodium hydroxide. The dark red colour is a characteristic indicator of the presence of Fe(OH) 3 . Furthermore, the P content in the supernatant aer centrifugation was also measured and the results are shown in Fig. S2b, † which indicates that the dissolution of FePO 4 produces PO 4 3À in the solution and more likely precipitates Fe(OH) 3 due to the large amount of OH À in the solution. Therefore, the reaction between DEA and FePO 4 is proposed as follows: (1a) Proposed reaction between FePO 4 and DEA.
The formation of Fe(OH) 3 is not desirable, as the PO 4 3À was disposed of with the supernatant, leading to a decrease of the P content remaining in the particles. The obvious change of colour happened at the 0.04 : 1 of DEA to FePO 4 ratio (Fig. S4 †), therefore, the optimal ratio was selected at 0.03 : 1, which has a low phosphate loss and reasonable ake structure that provided a relative large contact area with water. In addition, the use of DEA at this ratio is 29 mg g À1 FePO 4 , which is much less than the toxicity limit of diethylamide (LD 50 ¼ 540 mg g À1 ). 28 The following experiment was conducted with m-FePO 4 resulting from this ratio. The lm properties and the release behaviour were examined.
Morphology and structure of m-FePO 4 /PVA lms Fig. 4 shows morphology of the cross-section of m-FePO 4 /PVA composite lms. The majority of the m-FePO 4 particles were distributed at the bottom of the PVA matrix as the particle size of m-FePO 4 was still at a micron scale and the precipitation rate was very high. The PVA matrix shows a uniform one phase without any pores. The sediment of m-FePO 4 on one surface is benecial for the dissolution process and the release of phosphorous, as a prior swelling and degradation of dense PVA lm is not required. PVA in this case was used as a matrix to hold all m-FePO 4 particles in the lm (Fig. 5), which can protect the nutrient from being washed away compared to a powder-or granules-based fertiliser system. The composition of m-FePO 4 /PVA lms was conrmed by FTIR (Fig. 6). The absorption peaks at around 800-1200 cm À1 Fig. 6 FTIR of m-FePO 4 /PVA with different mixture ratios.   are regarded as P-O bonds. 29,30 The peak at around 3640-3100 cm À1 represents O-H from inter-and intra-molecular hydrogen bonds, 31 while the peak at 1740 cm À1 corresponds to the C]O bonds. Characteristic peaks from phosphate and PVA are present in all samples, so it may be concluded that the chemical structures of PVA and FePO 4 were retained in the composite lm.

Release behaviours of m-FePO 4 /PVA in water and citric acid
When plants lack phosphorous they release organic acids, such as oxalic acid, malic acid, and citric acid. 32 In this study, citric acid was selected as a typical organic acid generated by a plant to investigate the release behaviour of powder samples and our lm-type fertiliser. The concentration of citric acid used was 50 mmol mL À1 which is the concentration normally produced by plant roots when phosphorus deciency happens. 33 The release behaviour of lm-type fertilisers with different ratios of m-FePO 4 and PVA was investigated in citric acid and in water and the results are shown in Fig. 7a and b. All samples displayed linear release kinetics which was tted with a linear release model. The equation is expressed as follows: where Y is the amount released in percentage, x is the time, K is the kinetic constant, and b is the intercept. The calculated parameters are listed in Table 1. R 2 represents the accuracy of the ttings, which are all higher than 0.99, indicating a good linear regression. By comparing K values for all the samples, all samples showed a faster release rate in citric acid solution than in water. The increase of m-FePO 4 content did not signicantly affect the release rate in water and citric acid solution. The m-FePO 4 to PVA ratios of 0.83 : 1 and 25 : 1 possess the lowest release rates in deionised water, while having the highest release rate in citric acid solution (90% faster than in water). This may be attributed to the increase of free PVA area that expedites the P release in the citric acid but hinders the release rate in water. These results indicate that the current lm-type fertiliser can maintain P content in not only a water environment but stimuli-responsively accelerates the release amount of P in the presence of citric acid.
Based on the kinetic model proposed, the life cycle of the current m-FePO 4 /PVA can be predicted. At 100% released phosphate, the calculated life cycle of the fertiliser is listed in Table 1. Assuming a constant presence of the citric acid, the estimated life cycle of lm composited at three different ratios of m-FePO 4 to the PVA contents, 0.83 : 1, 25 : 1, and 75 : 1, is 2.0, 2.0, and 2.4 years, respectively, which all are less than the raw FePO 4 powder (3.8 years). In contrast, the life cycle of the lm in water only is around 17 years, which is more than that of raw FePO 4 powder (13.3 years). These results further suggest that the incorporation of PVA can control the release of phosphate, namely maintaining P content in the absence of organic acid but also facilitating the release of phosphate in the presence of an organic acid.
In general, the total amount of P required to grow a plant depends on the total weight of it, as described by grams of P per fresh weight of plant (g kg À1 ). 34 The P requirements of different plant species and different plant growth cycles usually are varied, which makes it difficult to determine the appropriate P weight for plants. However, a standard medium can provide a reference to roughly determine the appropriate P content range. The safe concentration range of P in the Murashige and Skoog (MS) medium, as a common medium for the in planta nutrient experiments, is between 6 mmol L À1 (lower bound) and 1 mol L À1 (upper bound). 35 When comparing this with the daily release of the current lm-type fertiliser, the released phosphate ranges from 0.01 to 0.011 mol L À1 in water and 0.079 to 0.11 mol L À1 in citric acid solution, which is much more than the lower bound of P concentration in the MS medium. The release amount of P in both water and citric acid solution can meet the requirement for plant growth to some extent. Also, the release rate can be nely tuned by tailoring the ratio between m-FePO 4 and PVA for specic plant species.
The PVA lms aer a 30 day release experiment were washed and dried for FTIR measurements (Fig. 8). The results show that there is a new peak present at around 1708 cm À1 in the sample  immersed in citric acid solution, which corresponds to the carboxylic groups. Citric acid containing both hydroxyl and carboxyl groups hydrogen bonding can occur even at a low temperature (5 C 36 ); thus, combined with FTIR results the citric acid is more likely to be adsorbed on PVA, which has indirectly increased the density of acidity just around the surface of the lm, facilitating the release of phosphate. Fig. 9a presents the results of P release from four types of fertiliser in citric acid solution for 30 days. The sample of m-FePO 4 powders has the highest release rate while the raw FePO 4 powders release at the lowest rate. This may be due to an increased surface area that benets from the contact between particles and solution, leading to a faster release of P. It is interesting to note that PVA moderates the release of m-FePO 4 while accelerating P release in the raw FePO 4 . This can be explained as that the m-FePO 4 has a higher surface area to be integrated with PVA than the raw one, so the area of exposed PVA that contributes to the adsorption of citric acid is much less than raw one. Therefore, the PVA covers one side of the m-FePO 4 particles to reduce the valid surface area while acting as a citric acid adsorption container to facilitate P release. Fig. 9b shows the nal release amount of P aer 30 days. The P release in citric acid solution is greater than that in the water counterpart for both the particle-type fertiliser and the lm-type fertiliser. A tenfold increase of the release rate of m-FePO 4 /PVA was found in citric acid solution than in the deionised water. The release rate of the m-FePO 4 is twofold when compared to the raw one in the citric acid. Even though m-FePO 4 powder possesses a higher release amount and rate than FePO 4 /PVA, powder fertilizer is more easily washed away than the lm-type during irrigation, which indicates the advantage of the lm-type fertilizer developed in this paper.

Conclusions
A formulation of controlled-release phosphorus fertiliser was fabricated using recycled FePO 4 as the P source and PVA as the polymeric matrix to form a freestanding lm. The modication of FePO 4 , aiming at increasing the surface area of the particles, was examined in terms of the effect of DEA concentration, structure, and reaction mechanism. PVA acts as matrix to keep the particles from water washing away. It was found that this new lm-type fertiliser is responsive to an organic acid, such as citric acid. The release rate of m-FePO 4 /PVA lm is slower than that of raw FePO 4 powder in deionised water, while it is faster in the presence of citric acid. This suggests that the phosphorous content can be maintained in water and released in citric acid solution such as that generated by plants when they have phosphorous deciency. In addition, through manipulating the ratio between m-FePO 4 and PVA, the release rate can be accurately tuned, which offers an opportunity to adapt for a specic condition of plants in agriculture. More importantly, FePO 4 was recycled from steelmaking slag so the cost of other materials used in this work is low. This fabrication process can be easily scaled up at low cost, making it potentially suitable for large eld use in agriculture. Except for the citric acid, the organisms in soil will also release other types of organic acid, 37 on which the response of the lm fertiliser will be further investigated in future work.

Conflicts of interest
There are no conicts to declare.