Introduction

Phosphorus (P) exists in water in both particulate and dissolved forms. The usual forms of P in aqueous solutions are orthophosphates, polyphosphates, and organic phosphates (Mezenner and Bensmaili 2009). Phosphorus is necessary for the growth of organisms and plants and is an indicator of surface water quality. Excessive P present in natural waters is known to cause eutrophication (Penn and Warren 2009). Eutrophication results in the depletion of oxygen that leads to fish death and affects other aquatic life forms adversely. The major point sources that contribute to P built up in aquatic environment include municipal and industrial wastewaters. Run-offs from agriculture, including animal agriculture, are the major non-point sources. The amount of P compounds in these sources should be controlled to prevent eutrophication in lakes and other surface waters. Accelerated eutrophication not only affects the aquatic life but indirectly hinders the economic progress of communities that depend on aquatic food and other resources (Cleary et al. 2009). Dissolved phosphate of ~0.02 mg/L is considered to have potential that lead to profuse algal growth in waters (USEPA 1995).

On the other hand, phosphorus is one of the required nutrients for plants. P-based fertilizers are extensively used in food crops and it is intricately related to global food security. Phosphorous for fertilizer production comes predominantly from select mines from Morocco, Western Saharan region, and China (Cordell et al. 2009). Phosphorus is a nonrenewable resource. While an assessment of future consumption of phosphorus fertilizers indicates that natural phosphate (PO4 3−) deposits will last for approximately 60–240 years (Cornel and Schaum 2009), P production rate is predicted to decline sometime in year 2035 while the demand for P-based fertilizers is on the rise (Cordell et al. 2011). Short supply of P-fertilizer is a major concern in food security area. It is, therefore, essential recover P from ‘wastes’ for possible reuse in agriculture.

Chemical precipitation (de Bashan and Bashan 2004), physico-chemical processes (Mishra et al. 2010), and enhanced biological phosphate removal (Gouider et al. 2011) are the frequently used techniques to remove aqueous phosphate. Among them chemical treatment methods for aqueous phosphate removal are widely practiced using chemicals like lime (Ahn and Speece 2006), alum (Babatunde and Zhao 2010), and ferric chloride (Caravelli et al. 2010). However, the high cost of chemicals and problems associated with sludge management make these methods unattractive for waters containing high amounts of phosphate (for example, wastewater with a typical total P of 4–14 mg/L, Tchobanoglous et al. 2003).

Sorption has emerged as a viable option for phosphate removal from aqueous media. In the recent years considerable amount of emphasis has been put on the use of low cost (ad)sorbents. Cost effectiveness is identified as the prime criterion in the selection of a sorption technology whether it uses synthetic or natural sorbents (Mishra et al. 2010). Phosphate can be removed from water using sorbents such as oxides of iron, natural ores like calcite, and goethite (FeOOH), active red mud, and activated carbon. One of the problems encountered with these sorbents is that they have very low sorption capacities. For example, sorption capacities of iron oxides are 11.2 mg PO4 3−/g (Yan et al. 2010a) and 19.02 mg PO4 3−-P/g (Cordray 2008). Similarly, natural ores like calcite were reported to have a sorption capacity of 0.1 mM PO4 3−/g (i.e., 3.1 mg PO4 3−-P/g, Karageorgiou et al. 2007). Goethite (FeOOH, 17.3 mg PO4 3−-P/g, Chitrakar et al. 2006), active red mud (9.8 PO4 3−-P/g, Yue et al. 2010), and activated carbon (3.02 mg PO4 3−-P/g, Hussain et al. 2011) are so far tried for P removal.

In the last two decades nanoscale zero-valent iron (NZVI) particles have received a lot of attention because of their unique reactive and sorbtive characteristics (Bezbaruah et al. 2009, 2011; Li et al. 2006). NZVI particles show good sorbtive characteristics owing to their high surface to volume ratio (Yan et al. 2010b). However, most of the reported work on sorption by NZVI has been on metalloids and heavy metals including some actinides (Giasuddin et al. 2007; Kanel et al. 2005; Klimkova et al. 2011; Scott et al. 2011) and, to the best of the authors’ knowledge, there is no literature on phosphate removal using NZVI and subsequent recovery of phosphate.

The objective of this study is to investigate the efficacy of NZVI particles for phosphate removal and recovery from aqueous solutions. Phosphate removal was tried under different environmental conditions (temperature, ionic strength), and in the presence of interfering ions and organic compounds. The effect of NZVI particle size on phosphate removal was studied. Batch experiments were conducted under different pH conditions to investigate the optimal pH conditions for phosphate recovery from NZVI.

Materials and methods

Chemicals and reagents

Iron(III) chloride hexahydrate (FeCl3·6H2O, 98 %, Alfa Aesar), sodium borohydride (NaBH4, 98 %, Aldrich), methanol (production grade, BDH), calcium chloride (CaCl2, ACS grade, BDH), monopotassium phosphate (KH2PO4, 99 %, EMD), potassium nitrate (KNO3, 99 %, Alfa Aesar), sodium hydroxide (5 N NaOH, Alfa Aesar), potassium sulfate (K2SO4, ACS grade, HACH), natural organic matter (Suwannee River NOM, RO isolation, IHSS), and humic acid (H1452, Spectrum) were used as received unless and otherwise specified.

Synthesis of NZVI

NZVI particles were synthesized using sodium borohydride reduction method (Eq. 1, Huang and Ehrman 2007).

$$ 2 {\text{FeCl}}_{ 3} + {\text{ 6NaBH}}_{ 4} + {\text{ 18H}}_{ 2} {\text{O }} \to {\text{ 2Fe}}^{0} + {\text{ 21H}}_{ 2} + {\text{ 6B}}\left( {\text{OH}} \right)_{ 3} + {\text{ 6NaCl}} $$
(1)

Ferric chloride hydrate (1.35 g) was dissolved in 40 mL of deoxygenated de-ionized (DI) water (solution A), and 0.95 g of sodium borohydride was dissolved in 10 mL of deoxygenated DI water in a separate beaker (solution B). Then solution A was added drop wise to solution B under vigorous stirring conditions (using a magnetic stirrer). The resultant black precipitates (NZVI) were centrifuged and washed with copious amount of deoxygenated DI water. The NZVI in slurry form was then stored in 20 mL vials in methanol to prevent oxidation, and used for experiments later. NZVI slurry in the vials was withdrawn using a pipette after vigorous stirring. The average weight of dry NZVI present in 1 mL well stirred slurry was measured to be 20 mg ± 0.6 mg (n = 25).

Phosphate removal batch studies

Batch experiments were conducted using (a) NZVI and (b) microscale zero-valent iron (MZVI) particles. Phosphate solution (50 mL of 1, 5, 10 mg PO4 3−-P/L) with 20 mg of NZVI (i.e., 400 mg/L) in multiple 50 mL polypropylene plastic vials fitted with plastic caps (reactors). The reactors were rotated end-over-end at 28 rpm in a custom-made shaker to reduce mass transfer resistance. One of the reactors was withdrawn at specific time interval (0, 10, 20, 30, 60 min) and the content was centrifuged at 4,000 rpm. Bulk solution from this reactor was collected for phosphate analysis and reactor was sacrificed or the spent NZVI was used for phosphate recovery study (see “Phosphate recovery batch studies”). Ascorbic acid method (Eaton et al. 2005) was used for phosphate analysis. This method depends on the formation phosphomolybdic acid during the reaction between orthophosphate and molybdate. Ascorbic acid reduces phosphomolybdic to form a blue complex. The color was measured in a UV–vis spectrophotometer (HACH, DR 5000) at wavelength of 880 nm. A five-point calibration was done routinely.

Effect of initial NZVI concentration

Batch studies were conducted with seven different NZVI concentrations (80, 160, 240, 320, 400, 480, 560 mg/L) for an initial bulk PO4 3−-P concentration of 5 mg/L. The experimental procedure described earlier (see “Phosphate removal batch studies”) was followed. Samples were withdrawn for phosphate analysis at 30 min.

Interference studies

The effects ionic strength, presence of selected anions and cations, and humic substances were examined. Batch studies were conducted in room temperature (22 ± 2 °C) using 400 mg NZVI/L and 40 mL of solution with an initial bulk phosphate concentration of 5 mg PO4 3−-P/L. Sampling frequency was maintained as described earlier (see “Phosphate removal batch studies”).

The ionic strength was varied from 0 to 10 mM by adding specific amounts of CaCl2 to the phosphate solution. The range of ionic strength was selected to represent groundwater conditions. The possible interference due to the presence of other important ions was also studied using two important anions (sulfate and nitrate). Potassium sulfate was used as the source of SO4 2− (0, 100, 500, 900 mg/L). The effect of NO3 − (0, 1, 5, 10 mg NO3 −-N /L) was studied by adding KNO3. Humic substances present in water may affect phosphate removal by NZVI, and to evaluate such impacts Suwannee River (USA) natural organic matter (0, 1, 10, 50 mg/L) and humic acids (0, 1, 10, 50 mg/L) were used in separate batch experiments. The batch experiments were conducted as described earlier (see “Phosphate removal batch studies”).

Effect of temperature

Experiments were conducted under different temperatures conditions (4, 22, 60 °C) to find out the effect of temperature change on phosphate removal by NZVI. The temperature of phosphate solution was first adjusted to the desired temperature by keeping it in the specific environment for long enough periods (~24 h). NZVI particles (400 mg/L) were added to phosphate solution (40 mL, 5 mg/L) once the specific temperature was reached. Samples were shaken at 100 rpm under temperature-controlled environment using an incubator-cum-orbital shaker (Thermo Scientific, MaxQ4000).

Effect of particle size

Effect of zero-valent iron (ZVI) particle size on phosphate removal was evaluated using NZVI particles synthesized within this research and MZVI particles purchased from a supplier (Aldrich, 99.9 % purity, used as received). ZVI reactions are known to be surface mediated (Thompson et al. 2010), and as such it was ensured that the same surface area concentrations were used in the experiments conducted with NZVI and MZVI. The NZVI particles used in this experiment had a surface area of ~25 m2/g (Bezbaruah et al. 2009) and MZVI had a surface area of ~2 m2/g (reported by the manufacturer). NZVI and MZVI surface area concentration of 10 m2/L (400 mg/L NZVI and 5 g/L MZVI) was used in the study.

Phosphate recovery batch studies

An initial batch study was conducted to find out the pattern of desorption (recovery) of phosphate into water from NZVI used for phosphate removal. Batch experiments were run first in 50 mL plastic vials fitted with plastic caps (reactors) with 400 mg/L NZVI and 50 mL of 5 mg PO4 3−-P/L to get the phosphate sorbed onto NZVI. The batch reactors were withdrawn after 60 min and centrifuged to separate the spent NZVI (i.e., NZVI particles with phosphate sorbed onto them). The bulk solution was decanted out and phosphate concentration was measured. A 50 mL volume of DI water was added to the spent NZVI and the pH was manipulated (2–12) with either 0.1 N HCl or NaOH. The reactors were closed and rotated end-over-end for 30 min. The reactors with the samples were then centrifuged and concentration of phosphate was measured in the bulk solution. The optimal pH for phosphate recovery (i.e., when maximum phosphate recovery) was determined based on this initial batch study results, and the rest of the phosphate recovery experiments were conducted at that particular (optimal) pH.

Additional batch studies were conducted in sacrificial reactors at the optimal pH, and phosphate recovery was monitored over time (0, 10, 20, 30 min). The data obtained from the removal experiments were normalized with respect to the original bulk phosphate concentration. For the data sets from the recovery studies, the initial phosphate concentration was calculated based on the mass of phosphate sorbed onto the NZVI and the data were normalized with respect to that.

NZVI characterization

X-ray diffraction (XRD) was done to find out NZVI composition. The samples were placed in stainless steel sample holders and XRD patterns were recorded using Cu Kα radiation (λ = 1.5418 Å) on a Philips X’Pert diffractometer operating at 40 kV and 40 mA between 5° and 90° (2θ) at a step size of 0.0167° (Xi et al. 2010).

High-resolution transmission electron microscopy (HRTEM, JEOL JEM-2100-LaB6 TEM) was used to observe the shape of NZVI particles and determine their particle size. NZVI particles were vacuum dried and the dry particles were placed in ethanol and sonicated for 5 min to achieve proper dispersion. Drops of the resulting solution were placed onto lacey carbon grids (Electron Microscopy Sciences, USA) and allowed to dry. Images were taken using a Gatan ORIUS large format CCD camera.

Quality control

All experiments were done in triplicates during this research and the average values are reported along with the standard deviations. Blanks with only phosphate (without NZVI/MZVI) were run along with the NZVI and MZVI experiments. The analytical instruments and tools were calibrated before the day’s measurements. One-way analysis of variance (ANOVA) tests were performed to compare the variance between data sets as needed. Additionally, Dunnett Method was used to compare control with rest of the treatment data. Minitab 16 software (Minitab, USA) was used for all statistical analyses.

Results and discussion

NZVI synthesis and characterization

NZVI synthesized (Fig. 1a) during this research were mostly spherical in shape and had particle size distribution from 10 to 30 nm with an average size of 16.24 ± 4.05 nm (n = 109, Fig. 1b). Huang and Ehrman (2007) reported particle size of 20 nm using the same method. The XRD spectrum (Fig. 1c) for the particles synthesized during this study shows three peaks of zero-valent iron (Fe0). A couple of iron oxide peaks were also observed which might be because of exposure of the particles to air during the XRD experiment. During the synthesis of NZVI, the particles were not bleed with air (as in Bezbaruah et al. 2009, 2011) but there is still a possibility that a thin oxide layer around the particles was formed due to reaction with atmospheric oxygen. A peak for iron chloride was also observed which might be from the left over reactants used for the synthesis of NZVI (see Eq. 1).

Fig. 1
figure 1

a High-resolution transmission electron microscopy (HRTEM) image of NZVI. b Particles size distribution of the nanoparticles synthesized was 10–30 nm with an average size of 16.24 ± 4.05 nm (n = 109). c X-ray diffraction (XRD) spectrum of NZVI with prominent peaks for Fe0. Peaks for oxides are from Fe-oxide layer on the NZVI, and the FeCl3 peak is from residuals of raw materials used in NZVI synthesis

Full size image

Phosphate removal

Batch experiments were conducted for phosphate removal using 400 mg/L NZVI and different phosphate concentrations (1, 5, 10 mg PO4 3−-P/L). Rapid phosphate removal was observed in the first a few minutes of the experiment for all three concentrations. About 88–95 % of phosphate was removed within the first 10 min and only minimal removal was observed beyond that (Fig. 2). Blanks did not show any removal of phosphate. Three consecutive data points (20, 30, and 60 min) showed no major change (maximum 2.7 % variation) in phosphate removal for the two higher concentrations (5 and 10 mg PO4 3−-P/L) while a much larger variation (~7.8 % from 20 to 60 min) was observed for 1 mg PO4 3−-P/L. While complete (100 %) phosphate removal was observed for 1 mg PO4 3−-P/L solution, 96.29 ± 0.13 and 97.53 ± 0.16 percent removals were observed for 5 and 10 mg PO4 3−-P/L, respectively. The sorption capacities at 60 min were found to be 2.27 ± 0.00, 12.00 ± 0.02, and 24.38 ± 0.04 mg/g for 1, 5, and 10 mg PO4 3−-P/L, respectively. The sorption capacity increased linearly (R 2 = 0.9999) with the increase in phosphate concentration.

Fig. 2
figure 2

Phosphate removal by NZVI/L from bulk solutions with different initial phosphate concentrations (triangle 10 mg PO4 3−-P/L, diamond 5 mg PO4 3−-P/L, square 1 mg PO4 3−-P/L, circle Blank). NZVI = 400 mg/L

Full size image

For 5 mg PO4 3−-P/L, 30 min was found to be long enough time to achieve equilibrium with 400 mg NZVI/L. As such all experimental data for 5 mg PO4 3−-P/L and 400 mg NZVI/L were collected up to 30 min. Iron-based removal techniques are reported by others where 15–100 % phosphate removal has been achieved (Table 1). Hydroxides of iron were found to be most effective in the removal process but a wide range of efficiency (15–100 %) has been reported (Chitrakar et al. 2006; Cordray 2008; Mezenner and Bensmaili 2009; Yan et al. 2010a). Synthetic goethite (α-FeOOH) was found to remove up to 1 mg P/L completely (100 %) from NaH2PO4 solution (Chitrakar et al. 2006). Again 100 % phosphate removal was observed with akaganeite (β-FeOOH) up to 0.3 mg P/L. It took 2–8 h to reach equilibrium in most of the reported phosphate removal experiments done with DI/distilled/wastewater (Mezenner and Bensmaili 2009; Chitrakar et al. 2006; Xiong et al. 2008; Yan et al. 2010a) but took 24 h to reach equilibrium in seawater (Chitrakar et al. 2006). It is pertinent here to discuss treatment time in other sorption systems for comparison purposes. Hussain et al. (2011) reported 95 % removal of phosphate with granular activated carbon over a 150-min period. Sorption of ~95 % of phosphate on calcite in 45 min was reported by Karageorgiou et al. (2007).

Table 1 Different iron-based adsorbents used for phosphate removal and their performance data
Full size table

In the present study, very fast removal of phosphate (88–95 % in 10 min) was achieved, and that makes this research very relevant for continuously flowing (pumped) water (i.e., required contact time will be short). The sorption capacities of NZVI after 10-min interaction with the aqueous solution containing phosphate were found to be 2.20 ± 0.06, 11.87 ± 1.20, and 23.62 ± 0.11 mg/g for 1, 5, and 10 mg PO4 3−-P/L solutions, respectively. The sorption capacities of 3.02–19.02 mg PO4 3−-P/g reported by others (Chitrakar et al. 2006; Cordray 2008; Hussain et al. 2011; Karageorgiou et al. 2007; Yan et al. 2010a, b) are comparable to the sorption capacities achieved for the NZVI in this study. However, the reaction time is much shorter with NZVI.

The mechanism of phosphate removal by NZVI in the present study can be explained based on point of zero charge (PZC) and ligand exchange (Eq. 2, Karageorgiou et al. 2007, and Fig. 3). PZC for NZVI is around 7.7 (Giasuddin et al. 2007), and when pH is less than PZC the surface of NZVI is positively charged which makes the surface suitable for anion (PO4 3−) sorption. The initial pH of the test solutions used in this study was ~4.0 and final pH after 60 min reaction was ~7.5 which was still lower than the PZC of NZVI. The pH environment maintained in the reactor was ideal for PO4 3− sorption and that is why 97.53–100 % removal was achieved in this study.

(2)
Fig. 3
figure 3

Phosphate sorption by NZVI under various pH conditions (after Cordray 2008). Lower pH is more conducive for phosphate adsorption while desorption is the dominant phenomenon at higher pH

Full size image

Effect of initial NZVI concentration

The removal of phosphate (C 0 = 5 mg/L) was found to increase with increase in the initial NZVI concentration (Fig. 4) and followed a linear trend (R 2 = 0.9539) as NZVI concentration increased from 0 to 560 mg/L. NZVI concentration beyond 400 mg/L did not improve PO4 3− removal significantly. Phosphate removal of 100 % was obtained with 560 mg NZVI/L. When the initial NZVI concentration was increased from 80 to 560 mg/L, the removal of phosphate increased by ~78 %. The increase in phosphate removal efficiency with the increase in NZVI concentration was expected as the contaminant removal by NZVI is a surface area mediated process. When NZVI concentration increased from 0 to 560 mg/L the reactive iron surface area in solution increased from 0 to 14 m2/L (NZVI surface area = 25 m2/g). The observations are consistent with findings by others with sorption media where surface area controls the sorption of phosphate (Mezenner and Bensmaili 2009).

Fig. 4
figure 4

Effect of initial NZVI concentration on phosphate removal. Initial PO4 3−-P = 5 mg/L

Full size image

Interference studies

The interferences of various ions and organic matters on phosphate removal were studied with an objective to understand how NZVI is going to behave during real field applications. Ionic strength (varied from 0 to 10 mM) did not have any statistically significant effect on phosphate (C 0 = 5 mg/L) removal by NZVI (Fig. 5a, α = 0.005, p = 0.225). However, ANOVA tests showed statistically significant differences in the treatment data for nitrate (α = 0.005, p = 0.001), sulfate (0–900 mg/L, α = 0.005, p = 0.00), humic acid (0–50 mg/L, α = 0.005, p = 0.00), and NOM (0–50 mg/L, α = 0.005, p = 0.00). Dunnett method was used to further compare the control with rest of the treatment data. All nitrate concentrations (1, 5, 10 mg NO3 −-N/L, Fig. 5b) were found to significantly interfere in phosphate removal from aqueous solution. However, this statistically significant increase in phosphate removal (1.40–2.77 %) may not have any practical significance bring very marginal. Again, the treatment data were significantly different from the control for all sulfate concentrations (100, 500, 900 mg/L). Phosphate removal by NZVI decreased by 5.16–6.27 % in the presence of sulfate in the solution (Fig. 5c). While the presence of NOM (1, 10, 50 mg/L, Fig. 5d) decreased phosphate removal by 6.01–11.03 % (all statistically significant), the presence of humic acid showed mixed results. The presence of 1 mg/L humic acid (Fig. 5e) significantly reduced (13.86 %) phosphate removal but interference was not statistically significant when humic acid concentration was increased (10 and 50 mg/L).

Fig. 5
figure 5

a Phosphate removal under different ionic strength conditions (filled circle 0 mM ionic strength, diamond 5 mM ionic strength, triangle 10 mM ionic strength). b Phosphate removal in the presence of nitrate (filled circle 0 mg NO3 −-N/L, square 1 mg NO3 −-N/L, diamond 5 mg NO3 −-N/L, triangle 10 mg NO3 −-N/L). c Phosphate removal in the presence of sulfate (filled circle 0 mg SO4 2−/L, square 100 mg SO 2-4 /L, diamond 500 mg SO 2-4 /L, triangle 900 mg SO 2-4 /L). d Phosphate removal in the presence of natural organic matter (filled circle 0 mg NOM/L, square 1 mg NOM/L, diamond 10 mg NOM/L, triangle 50 mg NOM/L). e Phosphate removal in the presence of humic acids (filled circle 0 mg/L, square 1 mg/L, diamond 10 mg/L, triangle 50 mg/L). For all figures: open circle Blank, NZVI = 400 mg/L, Initial PO4 3−-P = 5 mg/L

Full size image

Liu et al. (2011) have reported interference due to ionic strength during phosphate removal with lanthanum-doped activated carbon fiber. They increased ionic strength from 0 to 10 mM and observed an 8.1 % drop in phosphate removal (from 98.8 to 90.7 %). Even 10 mM ionic strength did not affect the phosphate removal efficiency in the present study, and a removal of 96.0–98.5 % was achieved in all cases (ionic strength varied from 0 to 10 mM). Introducing competing anions was expected to have negative effects on phosphate adsorption (Liu et al. 2011). Fe0 was successfully used by others to remove nitrate from aqueous solution (Bezbaruah et al. 2009; Hwang et al. 2011) and so it was expected that nitrate will compete with phosphate for reactive/sorption sites on NZVI. Nitrate was found to interfere in phosphate removal in layered double hydroxides (Das et al. 2006) and ~12 % reduction in phosphate removal in the presence of nitrate was reported. Xue et al. (2009), however, did not find any interference of NO3 − during phosphate removal using basic oxygen furnace slag. In the present study slight increase (1.40–2.77 %) in phosphate removal (though not environmentally significant) was observed. Further research would be needed to find out the possible reason for this increase but the authors would like to hypothesis that nitrate reacted with NZVI to produce iron (hydr)oxides with a higher affinity for phosphate. Sulfate was also reported to interfere with phosphate and reduced phosphate removal by 24.5 % in layered double hydroxides (Das et al. 2006). In this study, sulfate retarded the phosphate removal process and reduced removal efficiency as high 6.27 %. The adsorption of phosphate in presence of humic acid was studied by Antelo et al. (2007) and found that phosphate adsorption onto the surface of goethite decreased by 45 and 25 % in the presence of humic acid at pH of 4.5 and 7, respectively. This can be explained by the competition of the humic acids functional groups with phosphate for the sorption sites where the humic acid outcompeted the phosphate. Also, the sorption sites on the surface could be blocked by the relatively large size of humic acids (~15 Å in diameter, Simeoni et al. 2003), thus less sorption sites will be available for phosphate (~2.56 Å in diameter) (Antelo et al. 2007). Similar results were reported by others (Shuai and Zinati 2009). In the present study only low concentration (1 mg/L) of humic acid affected phosphate removal while higher concentration did not. This happened possibly because of increased sorption of phosphate onto NZVI due to lowering of solution pH (see Fig. 3) at higher humic acid concentrations. Additional experiments are needed to investigate why the presence of the humic substances did not adversely affect phosphate removal by NZVI, the authors feel that NZVI reacts very fast with phosphate in the first 10 min or so the possible inferring compounds are not competitive enough. In this study, phosphate removal in the presence of Suwannee River NOM (1, 10, 50 mg/L) was found to be significantly different from the control (without NOM). Phosphate removal efficiency of NZVI reduced by 9.01–11.03 % in the presence of NOM. This result was expected as NOM negatively impacts NZVI reactivity. Li et al. (2010) reported the minimum concentration of NZVI that inhibited E. coli growth after 24 h exposure as 5 mg NZVI/L. However, in the presence of NOM, the NZVI concentration had to be increased to 100 mg/L to achieve the same degree of inhibition. This happened possibly because the NZVI particles’ reactive surfaces were covered with NOM and, thus, reducing the overall reactivity of NZVI. Chen et al. (2011) also observed a 23 % reduction in trichloroethylene (TCE) degradation by NZVI in the presence of Suwannee River NOM.

Effects of temperature

Experiments were conducted at 4, 22, and 60 °C during this study. The removal of phosphate at 4 and 22 °C was relatively slower than the removal at 60 °C. However, after 30 min, the removal was the more or less same (91.4–95.3 %) for all temperatures (Fig. 6) and there was no significant differences between the values (one-way ANOVA: α = 0.005, p = 0.144). This is in contrast to findings by others. Increasing the temperature from 25 to 45 °C increased the phosphate adsorption capacity of granular ferric hydroxide from 3.6 to 5.1 M P/g (i.e., 0.11–0.16 mg/g, Saha et al. 2010). Liu et al. (2011) also reported increase adsorption capacity in lanthanum-doped activated carbon fibers from 8.54 to 9.41 mg/g of with the increase of temperature from 20 to 50 °C. Mezenner and Bensmaili (2009) reported ~60 % increase in phosphate adsorption onto iron hydroxide–eggshell waste when the temperature was increased from 20 to 45 °C. Fast phosphate removal that takes place within the first 10 min may be the reason why no distinction could be made between removal achieved in three different temperatures.

Fig. 6
figure 6

Effect of temperature on phosphate removal by NZVI, (square 4 °C, triangle 22 °C, diamond 60 °C, circle Blank with only PO4 3− solution). NZVI = 400 mg/L, Initial PO4 3−-P = 5 mg/L. Blank shown here is for 22 °C only. The blanks at other temperatures followed similar trends and are not shown here to maintain clarity

Full size image

Effect of particle size

Sorption is dependent on surface area and, hence, in this study the same surface area (10 m2/L) was used for the comparison of phosphate removal by NZVI (particle size ~16 nm) and MZVI (<10 μm). MZVI removed only 23 % of phosphate (C 0 = 5 mg PO4 3−-P/L) in 30 min whereas NZVI removed ~96 % of phosphate over the same time period (Fig. 7). NZVI was 13.9 times more efficient than MZVI in removing aqueous phosphate. Others reported similar observations with NZVI for other contaminants. Surface area normalized rate constant (k sa) of NZVI (surface area ~30–35 m2/g) for tetrachloromethane degradation was reported as over two orders of magnitude higher than that of MZVI (Li et al. 2006). Also, removal capacity of Cr(VI) using NZVI was more than 100 times that of the removal capacity using MZVI (Li et al. 2006). Kanel et al. (2005) reported that the k sa for As(III) removal by NZVI was 1–3 orders of magnitude higher than MZVI. While using NZVI, k sa of alachlor degradation was found to be ~10 times that with MZVI (k sa-NZVI = 38.5  ×  10−5 and k sa-MZVI = 3.8–7.7  ×  10−5 L h−1 m−2, Thompson et al. 2010).

Fig. 7
figure 7

Effect of ZVI particles size on phosphate removal (diamond Micro-ZVI, square NZVI, circle Blank). MZVI = 5 g/L; NZVI = 400 mg/L; Equal ZVI surface area concentrations (10 m2/L) were used for both MZVI and NZVI, Initial PO4 3−-P = 5 mg/L

Full size image

Phosphate recovery

In the initial batch study ran to find out the optimal pH for phosphate recovery maximum phosphate recovery was achieved at pH 12, and the recovery was minimal at acidic pH (data not shown). In the follow up phosphate desorption (recovery) batch studies conducted at pH 12, 78.4 % phosphate recovery was obtained (Fig. 8). The 78.4 % recovery is based on the mass of phosphate sorbed onto NZVI during removal experiment. If the recovery is calculated based on the mass of the phosphate present in the original bulk solution from which removal was achieved than the phosphate removal is 74.5 %. In terms of practical applications, if 5 mg/L phosphate is present in bulk solution 4.80 mg/L (96 % removal, see “Phosphate removal” under “Results and discussion”) will be removed by NZVI and 3.73 mg/L (74.5 % recovery) can be recovered back from the NZVI. Better phosphate recovery at higher pH was achieved possibly because of the abundantly present hydroxide ions at a higher pH. The presence of these hydroxide ions would result in a net negative surface charge to which few phosphate anions would be bound. The opposite phenomena would occur at a lower pH which would result in more sorption. Poor recovery of phosphate at pH 4 and 6 supports the previously proposed mechanism where an electrostatic attraction between the phosphate ions and the surface of NZVI occur resulting in phosphate sorption on the surface of NZVI. Also, pH 12 is higher than PZC of NZVI and particles are negatively charged resulting in desorption of phosphate (Eq. 3, Karageorgiou et al. 2007 and Fig. 3). Research indicate that phosphate can be recovered from sorptive media under high pH conditions (Babatunde and Zhao 2010; Cordray 2008; Karageorgiou et al. 2007; Liu et al. 2011). Similar results were reported by others using other forms of iron oxides (Yan et al. 2010a; Zeng et al. 2004).

(3)
Fig. 8
figure 8

Phosphate removal and recovery using NZVI (square Removal, circle Blank in removal experiment (PO4 3− solution), diamond Recovery, triangle Control in recovery experiment (pH adjusted DI water + fresh NZVI)). NZVI = 400 mg/L, Initial PO4 3−-P = 5 mg/L. Control for the removal experiment was DI water with NZVI; no phosphate was detected in the sample, and the data points coincided with the control for the recovery experiment

Full size image

Conclusions

Results from the batch studies conducted during this research demonstrate the effectiveness of NZVI for phosphate removal and recovery with different initial phosphate concentrations (1, 5, 10 mg PO4 3−-P/L). Phosphate removal of 88–95 % was achieved in the first 10 min itself and 96–100 % removal was achieved after 30 min. Increase in phosphate removal efficiency improved with the increase in initial NZVI concentration use and followed a linear trend (R 2 = 0.9539). When the initial NZVI concentration was increased from 80 to 560 mg/L, the removal of phosphate increased by ~78 % (C 0 = 5 mg/L). Little interference was observed in phosphate removal due to ionic strength and temperature change. Sulfate and natural organic matters had statistically significant negative impacts but nitrate marginally improved phosphate removal. The phosphate removal efficiency was also not affected by high concentrations of humic acid. Phosphate sorbed onto NZVI was successfully recovered (~78 %). The phosphate recovery process was found to be pH dependent with maximum recovery achieved at pH 12.

Environmental significance

Phosphate removal using NZVI has potential applications in wastewater treatment plants, where phosphate removal is otherwise not very efficient. The speed of phosphate removal using NZVI (88–95 % removal in the first 10 min) gives the nanoparticles an advantage over other sorbents. The high speed of phosphate removal by NZVI can be used to engineer a commercially viable treatment process with low detention time and minimal infrastructure. More research is needed to optimize the recovery of phosphate from NZVI as pH 12 may not be a practical value from economic and hazard perspectives.