Introduction

Iron(III) oxides are known as ubiquitous, chemically stable, cheap, and nontoxic compounds. Their low bandgap energy (~2.2 eV) enables them to collect up to 40% of whole solar spectrum, and they are thus used, among other applications, as photocatalysts in water splitting and water treatment (Mishra and Chun 2015). Many studies reported that α-Fe2O3 thin films were successfully applied as semiconducting electrodes for photocatalytic water splitting and hydrogen production (Le Formal et al., 2010; Sivula et al. 2010; Frydrych et al. 2012; Le Formal et al. 2012; Lopes et al. 2014) and α-Fe2O3 nanoparticles or nanotubes effectively decomposed water contaminants, including dyes and 4-chlorophenol under visible light illumination (Khedr et al. 2009; Zhou et al. 2010). The application of α-Fe2O3 in disinfection of environment using visible light irradiation was reported as well (Akhavan et al. 2009; Akhavan 2010).

Both α-Fe2O3 and γ-Fe2O3 have been widely used as photocatalysts to degrade organic and inorganic pollutants in the environment. In particular, α-Fe2O3 has been commonly used in many applications due to its higher stability than other iron oxides under ambient conditions (Zhou et al. 2010; Kato and Komatsu 2013). Moreover, it is of great importance to recover used catalysts from water treatment processes for saving resources as well as avoiding additional water filtration causing extra expenses during water treatment. Ferrimagnetic behavior of γ-Fe2O3 or Fe3O4 gives a possibility to separate the catalysts by applying external magnetic field, e.g., using a permanent magnet. On the other hand, the photocatalytic performance of iron oxides can be limited by low diffusion lengths of holes, poor conductivity, and high rates of electrons (e−) and hole (h+) recombination. These difficulties can be significantly overcome by forming nanostructured iron oxides. Therefore, it is required to develop nanostructured magnetic iron oxides for water treatment.

Microcystins (MCs) are the most common toxins produced by harmful cyanobacteria, which can be present in natural water. Microcystin-LR (MC-LR) is mostly found in different water bodies among MCs. Up to now, several approaches of photocatalytic degradation of MC-LR have been investigated mostly based on the application of TiO2-based nanostructures, such as highly ordered TiO2 nanotubes (Su et al. 2013), Ag/AgCl/TiO2 nanotube arrays (Liao et al. 2013), nanostructured nitrogen-doped TiO2 (Trantis et al. 2012), C-N-codoped TiO2 films (Liu et al. 2013), Bi-doped TiO2 (Yang et al. 2011), N-F-codoped TiO2 films (Pelaez et al. 2010), and sulfur-doped TiO2 films (Han et al. 2011). Generally, a shift of the bandgap toward visible light was achieved by various kinds of doping of TiO2. Besides TiO2-based nanomaterials, AgCl/iron oxide composites were tested for MC-LR photocatalytic degradation (Zhang et al. 2013).

In our study, we used pure Fe2O3 nanopowder samples as visible light-active photocatalysts, which have been easily synthesized by thermal decomposition of ferrous oxalate dihydrate without any further processing, for MC-LR degradation under visible light irradiation. It is worth mentioning that such kind of Fe2O3 nanopowder consisting of very small nanoparticles (~5 nm) has been also tested in the past as a catalyst for heterogeneous catalysis of decomposition of hydrogen peroxide (Hermanek et al. 2007; Gregor et al. 2010; Hermanek et al. 2010) or degradation of phenol (Prucek et al. 2009).

Material and methods

Ferrous oxalate dihydrate (FeC2O4·2H2O, Sigma-Aldrich) was used as an iron precursor for thermal synthesis of ferric oxide nanopowders. MC-LR standard (Calbiochem) was purchased, and a MC-LR stock solution of 500 mg L−1 was prepared using autoclaved Milli-Q-grade water. For photocatalytic degradation experiments, the stock solution was diluted to 500 μg L−1.

Two iron-containing nanopowder samples were prepared by thermal decomposition of 4 g of ferrous oxalate dihydrate (FeC2O4·2H2O) in a muffle furnace in air. The calcination at 200 °C for 5 h resulted in sample labeled MG-9, whereas the calcination at 180 °C for 16 h led to sample labeled MG-11.

The as-prepared ferric oxide nanopowders were characterized using X-ray powder diffraction (XRD), 57Fe transmission Mössbauer spectroscopy, transmission electron microscopy (TEM), magnetrometry, UV–visible diffuse reflectance spectroscopy (DRS), and Brunauer–Emmett–Teller (BET) physical adsorption technique.

TEM (JEOL JEM-2010) was performed to investigate the detailed morphology of synthesized catalysts at 160 kV with 1.9-Å point-to-point resolution. For TEM analysis, diluted dispersion was fixed on holey carbon film copper grids (from Emsdiasum Company) and then vacuum dried at room temperature.

XRD patterns of the prepared samples were obtained using PANalytical X’Pert PRO diffractometer in Bragg–Brentano geometry with iron-filtered CoKα radiation source, an X’Celerator detector, programmable divergence, and diffracted beam anti-scatter slits. The samples were prepared on a zero-background single crystal silicon slide and measured in continuous mode with a resolution of 0.017° of 2θ. The commercial standards SRM640 for Si and SRM660 for LaB6 from NIST were used to evaluate line positions and instrumental line broadening, respectively. The X’Pert HighScore Plus software (PANalytical, The Netherlands), combined with PDF-4+ and ICSD databases, was used to analyze the obtained patterns.

A transmission MS96 Mössbauer spectrometer (Pechoušek et al. 2012) with a 57Co(Rh) source of γ-rays was employed for the zero-field Mössbauer experiment at 300 K. The recorded spectra were fitted by the Lorentzian line shapes with the MossWinn program (Klencsár et al. 1996) based on the least-square method. The values of isomer shift were referred to α-Fe foil samples at room temperature.

UV–vis absorption spectra of the iron oxide nanoparticle dispersions were acquired in a quartz cuvette using a Specord S 600 spectrophotometer (Analytic Jena AG, Germany) in a range from 200 to 800 nm.

Surface areas of the prepared iron oxide nanopowders were characterized with nitrogen adsorption–desorption measurements. Adsorption–desorption isotherms were measured at temperature of liquid nitrogen (77.4 K). Experimental points acquired by the statistic volumetric technique were obtained by a Thermo Finnigan Sorptomatic 1990 analyzer. Before the measurement, the surface of the nanopowders was degassed at room temperature for at least 20 h. BET surface areas were determined with the multipoint BET(3) model. The analysis was performed using an ADP (Advanced Data Processing) 4.0 software package (CE Instruments). Samples were measured with identical settings of number adsorption and desorption points under the same analysis conditions.

Magnetic data were obtained with a Quantum Design Physical Property Measurement System (PPMS DynaCool system, Quantum Design) with the VSM option. Diamagnetism of samples and a signal of the sample holder were collected. The hysteresis loop was recorded at 300 K with external magnetic fields in a range of −5 to +5 T.

Concentrations of metal ions within leaching tests were measured by an atomic absorption spectroscopy (AAS) using a ContrAA 300 flame ionizer (Analytik Jena AG, Germany) with a high-resolution Echelle double monochromator (spectral bandwidth of 2 pm at 200 nm) and a continuum radiation source (xenon lamp). The absorption line used for the analyses of Fe content was 248.3270 nm. Calibration standards of Fe for performing AAS were of TraceCERT (1 g L−1) type, purchased from Fluka. Solutions were prepared using deionized water (18 MΩ cm−1, Millipore).

For the evaluation of the photocatalytic activity of the prepared iron oxide nanopowders under visible light irradiation, a reactor (glass petri dish, dia. 60 mm) was used. For visible light irradiation (Zhao et al. 2013), two Cole-Parmer 15-W fluorescent lamps were employed with an actual light intensity of 3.99 × 10−4 W cm−2 and a UV420 Opticology filter was employed to eliminate UV component from the lamp spectra. The spectra of visible light are available in Additional file 1: Fig. S1 in a previous study (Han et al. 2014). A reactor containing 10 mL total volume of the solution was used to evaluate the visible light-induced photocatalytic activity of the two catalysts for MC-LR degradation.

The initial MC-LR concentration and catalyst loading were 500 μg L−1 and 0.5 g L−1, respectively. To investigate the effects of catalyst loading on the MC-LR degradation, three catalyst loadings (0.25, 0.5, and 1.0 g L−1) were tested. To explore the effects of the initial pH of solution on the decomposition of MC-LR, experiments were performed at four different pH values (3.0, 5.8, 7.0, and 10.0). To study MC-LR degradation under visible light, the reactor was placed in the dark for the initial 3 h to reach adsorption equilibrium since the initial MC-LR concentration was very low, and then exposed to visible light for 5 h. Aliquot samples of 200 μL were taken at reaction times of 0, 1, and 3 h under dark condition and 1, 2, 3, and 5 h under visible light illumination. A high-performance liquid chromatograph (HPLC, Agilent Series 110) was used with the previously reported analytical method to measure the concentration of MC-LR in the samples. A Discovery HS C18 column (150 mm × 2.1 mm, 5-μm particle size, Supelco, Bellefonte, PA, USA) was equipped in HPLC. The flow rate of mobile phase was 0.2 mL min−1, the sample injection volume was 20 μL, and the column temperature was 45 °C. For safety issue, all experiments were performed in an advanced SterilchemGARD III Class II biological safety cabinet (Baker Company, Sanford, ME) (Pelaez et al. 2009; Han et al. 2011). All experiments were performed three times to confirm the reproducibility.

In order to confirm the stability of the Fe2O3 catalysts in water under lower pH values, leaching tests were performed. Catalyst concentrations equal to 0.5 g L−1 and pH values equal to 3.0 and 5.8. The 50 mL of catalyst solution was shaken on a conventional multi-rotator shaker (Multi RS-60, Biosan) in 60-mL vials. Aliquot samples of 5 mL were taken at reaction times of 3, 8, and 24 h and subsequently three times filtered through 0.2-μm syringe filters before Fe content in the filtered solution was analyzed by AAS.

Results and discussions

Synthesis and characterization of ferric oxide nanopowders

XRD patterns of both MG-9 and MG-11 samples (Fig. 1) show the presence of two crystalline ferric oxide phases: γ-Fe2O3 and α-Fe2O3. Relative weight contents, mean X-ray coherence lengths, and cell parameters are summarized in Table 1. In addition to crystalline γ-Fe2O3 and α-Fe2O3 with the small X-ray coherence lengths 8 and 6 nm, respectively, the XRD patterns indicate the presence of amorphous phase as well.

Fig. 1
figure 1

X-ray diffraction patterns of MG-9 (top) and MG-11 (bottom) samples. The positions of Bragg reflection of each identified phase are shown under the pattern. The quantification of identified phases is presented on the right side

Full size image
Table 1 The phases identified in the XRD patterns with specified parameters derived by Rietveld refinement
Full size table

Room-temperature Mössbauer spectra of both samples (Fig. 2) were fitted by doublets corresponding to (super) paramagnetic Fe(III) in an octahedral coordination and distributions of hyperfine magnetic fields originating from particles with sizes, which are slightly above the limit for superparamagnetism. The hyperfine parameters are summarized in Table 2. Maghemite and hematite particles of ferric oxide contribute to both superparamagnetic doublet and sextet components in case of both studied samples. This overlap is well demonstrated by the value of quadrupole shift of −0.16 mm s−1 for the sextet of sample MG-9, which lies between the value of −0.21 mm s−1 for hematite and approximately zero for maghemite. Therefore, a quantification of the phase composition was not possible to perform like in the case of XRD analysis.

Fig. 2
figure 2

Room-temperature Mössbauer spectra of MG-9 (a) and MG-11 (b) samples

Full size image
Table 2 Hyperfine parameters and relative spectral areas calculated from room-temperature spectra of samples MG-9 and MG-11
Full size table

TEM images of MG-9 and MG-11 samples (Fig. 3) show strongly agglomerated nanoparticles with size in the range from 5 to 20 nm. Specific surface areas (BET) of MG-9 and MG-11 samples were determined as 234 and 207 m2 g−1, respectively. Moreover, the light absorption of MG-9 and MG-11 was measured with DRS. Both MG-9 and MG-11 effectively absorbed a broad range of UV–visible light from 200 to 800 nm, and their light absorption properties were pretty much the same (Additional file 1: Fig. SI1). Zero point charges were measured for the both MG-9 and MG-11 samples. The results are summarized in Additional file 1: Table SI1.

Fig. 3
figure 3

TEM images of MG-9 (left) and MG-11 (right) samples

Full size image

Magnetic properties of sample MG-9 were investigated by measuring of hysteresis loop at room temperature (Fig. 4). The “S” shape of the curve is typical for superparamagnetic behavior of nanoparticles. There is no saturation of the curve at the highest magnetic fields, which is due to the presence of hematite in the sample. The maximum measured magnetization reached the value of 16.2 emu g−1, which is enough for attracting the nanopowder iron oxides to a conventional permanent magnet. Remanent magnetization as well as coercivity is almost zero. Similar magnetic behavior for sample MG-11 at room temperature is expected. In order to confirm possibility of magnetic separation of the samples in water, an ordinary permanent magnet was placed into the vials with MG-9 and MG-11 samples dispersed in water. Nanoparticles of both samples were separated in the proximity of the magnet during approx. 2 min. The experiment is illustrated by the photos in Fig. 5.

Fig. 4
figure 4

Hysteresis loop for sample MG-9 at room temperature. The inset shows that remanent magnetization as well as coercivity is almost zero

Full size image
Fig. 5
figure 5

Magnetic separation of the catalysts by using of a permanent magnet. Left: the samples before application of the magnet; right: the most of nanoparticles are attracted to the magnet during 2 min

Full size image

Degradation of MC-LR using the ferric oxide nanopowders

MG-9 ferric oxide sample showed higher photocatalytic activity to decompose MC-LR in water under visible light irradiation in comparison with MG-11 sample at the given conditions (i.e., pH 5.8, 500 μg L−1 of initial concentration of MC-LR and 0.5 g L−1 of the catalyst loading). As seen in Fig. 6, MG-9 and MG-11 adsorbed about 26 and 8% MC-LR, respectively, under the dark condition for 3 h. After 5 h of visible light illumination, total removal of MC-LR was 86% for MG-9 and 40% MC-LR for MG-11. Based on the results, 60 and 32% MC-LR was decomposed by visible light-induced MG-9 and MG-11 photocatalysis, respectively. The phase composition as well as morphology and degree of agglomeration of nanoparticles is almost the same for both tested samples as it is documented by practically the same shapes of the BET isotherms (Additional file 1: Fig. SI2). Also, as mentioned in “Synthesis and characterization of ferric oxide nanopowders” section, the light absorption property of both samples was similar in a range of wavelength from 200 to 800 nm. The nanopowder photocatalysts differ in specific surface area (234 vs. 207 m2 g−1) that evidently contributed to the difference in the extent of MC-LR removal. The MC-LR removal per unit surface area was 3.72 μg m−2 for MG-9 and 2.03 μg m−2 for MG-11.

Fig. 6
figure 6

The removal of MC-LR for MG-9 and MG-11 samples at pH 5.8 with a 500 μg L−1 of MC-LR and the catalyst loading of 0.5 g L−1

Full size image

Further, different loadings of MG-9 sample were tested in the decomposition of MC-LR under visible light irradiation (Fig. 7) since this sample demonstrated higher photocatalytic activity than MG-11 sample. The photocatalytic degradation of MC-LR with MG-9 under visible light followed pseudo first-order kinetics with the observed rate constant of 0.0359, 0.3289, and 0.0518 h−1 for 0.2, 0.5, and 1.0 g L−1 of MG-9, respectively (see Additional file 1: Fig. SI3). At 0.5 g L−1 of MG-9 loading, the highest MC-LR removal was observed. At lower loading of catalysts (<0.5 g L−1), lower number of active sites of the catalysts may be involved to decompose MC-LR. Moreover, the degradation of MC-LR decreased at higher loading (>0.5 g L−1) due to particle agglomeration and the increase of turbidity of solution, associated with light penetration. Therefore, the optimal catalyst loading (0.5 g L−1) was used to investigate the effect of initial pH of solution on the degradation of MC-LR under visible light irradiation.

Fig. 7
figure 7

a The effect of catalyst loadings for MG-9 on the removal of MC-LR at pH 5.8 with 500 μg L−1 of MC-LR. b Total removal of MC-LR with different loadings of MG-9 after 5 h of visible light irradiation at pH 5.8 with 500 μg L−1 of MC-LR

Full size image

To investigate the effect of initial pH of solution on the degradation of MC-LR under visible light illumination, the pH of solution was adjusted using nitric acid or 10% sodium hydroxide solution. Higher concentration of MC-LR (750 μg L−1) was used for appropriate HPLC analysis. At pH 3, MG-9 showed the highest removal of MC-LR along with the highest adsorption of MC-LR due to high electrostatic attraction because MC-LR is negatively charged but iron oxides are positively charged (Mustafa et al. 2004; Gao and Chorover 2010; Pelaez et al. 2011). After 5 h of visible light irradiation, 70.5% MC-LR was removed (Fig. 8). The total removal of MC-LR was inversely proportional to the initial pH of solutions (Fig. 8b). These results may also be caused by Fenton reaction due to dissolution of iron at low pH (i.e., 3.0). Fe2O3 produces pairs of electrons (e−) and holes (h+) via light absorption (Eq. (1)) (Kiwi and Grätzel 1987; Shelton et al. 2016) and hydroxyl radical (·OH) and superoxide anions (O2·−) form according to Eqs. (2) and (3). Also, hydrogen peroxide (H2O2) could be produced according to Eqs. (4), (5), (6), and (7) (Zhao et al. 2014; Khan et al. 2016). Formed H2O2 reacts with dissolved iron (Eq. (8)) via Fenton reaction (Ruppert et al. 1993; Winterbourn 1995), which improves the efficiency of visible light-induced Fe2O3 photocatalysis to decompose MC-LR at pH 3.

$$ {\mathrm{Fe}}_2{\mathrm{O}}_3+ h\nu \to {\mathrm{h}}^{+} + {\mathrm{e}}^{-} $$
(1)
$$ {\mathrm{h}}^{+}+{\mathrm{H}}_2\mathrm{O}{\to}^{\bullet}\mathrm{OH} + {\mathrm{H}}^{+} $$
(2)
$$ {\mathrm{e}}^{\hbox{-} }{+\mathrm{O}}_2\to {{\mathrm{O}}_2}^{\bullet -} $$
(3)
$$ {\mathrm{e}}^{\hbox{-} }{{+\mathrm{O}}_2}^{\bullet \hbox{-} }{+2\mathrm{H}}^{+}\to {\ \mathrm{H}}_2{\mathrm{O}}_2 $$
(4)
$$ {{\mathrm{O}}_2}^{\bullet \hbox{-} }{{+\mathrm{O}}_2}^{\bullet \hbox{-} }{+2\mathrm{H}}^{+}\to {\ \mathrm{H}}_2{\mathrm{O}}_2{+\mathrm{O}}_2 $$
(5)
$$ {{\mathrm{O}}_2}^{-\bullet } + {\mathrm{H}}^{+}\to {{\ \mathrm{H}\mathrm{O}}_2}^{\bullet}\kern1.5em $$
(6)
$$ 2{{\mathrm{H}\mathrm{O}}_2}^{\bullet }+2{\mathrm{H}}^{+}\to 2{\mathrm{H}}_2{\mathrm{O}}_2{+\mathrm{O}}_2 $$
(7)
$$ {\mathrm{Fe}}^{2+} + {\mathrm{H}}_2{\mathrm{O}}_2\to {\ \mathrm{Fe}}^{3+}{\mathrm{O}\mathrm{H}}^{-}{+}^{\bullet}\mathrm{OH} $$
(8)
Fig. 8
figure 8

a The effect of initial pH for MG-9 on the removal of MC-LR at 750 μg L−1 of MC-LR with 0.5 g L−1 of MG-9. b The effect of initial pH for MG-9 on the removal of MC-LR after 5 h of visible light illumination at 750 μg L−1 of MC-LR with 0.5 g L−1 of MG-9

Full size image

In order to confirm a stability of the Fe2O3 catalysts in water under acidic solution conditions (i.e., pH 5.8 and pH 3.0), leaching tests were performed since iron oxides (i.e., lepidocrocite, ferrihydrite, hematite, goethite, ferrihydrite, and “soil-Fe”) show the minimal solubility at pH 7–8 (experimental details are mentioned in the “Material and methods” section) (Cornell and Schwertmann 2003). The results showed no leaching of Fe2O3 at pH = 5.8 even after 24 h of the experiment within the detection limit of the method (0.2 mg L−1 of Fe, which was determined as the end of the calibration curve). Addition of Fe2O3 nanopowder to water with pH = 3.0 led to dissolution of 0.065 wt% of Fe2O3 after 8 h of the experiment, and this value remained almost the same after the subsequent 16 h. Iron (Fe2+) dissolved at pH 3.0 reacted with formed H2O2 (Eq. (8)) during visible light-induced Fe2O3 photocatalysis, resulting in the enhanced removal of MC-LR. However, the observed rate constant of pseudo first-order kinetics for the photocatalytic degradation of MC-LR at different pH values was 0.0648, 0.0811, 0.0325, and 0.0123 h−1 at pH 3.0, 5.8, 7.0, and 10.0, respectively (see Additional file 1: Fig. SI4). It seems that the photocatalytic degradation of MC-LR at pH 5.8 was higher than at pH 3.0, but the actual removal of MC-LR at pH 3.0 is still high. At that pH value, MC-LR of ~35% was adsorbed on MG-9 and it significantly contributed to the overall MC-LR removal. Although the highest removal of MC-LR was obtained at pH 3.0, pH 5.8 could be an appropriate operating condition for the catalyst to avoid any iron contamination problems (Kohgo et al. 2008).

Concerning the issue of using of a mixture of maghemite and hematite as photocatalyst, it is expected that the mixture can lead to the higher photocatalytic activity than maghemite or hematite alone. The bandgaps for maghemite and hematite are almost the same (2.02 eV), but they have different positions. Hematite has a conduction band level at −0.62 V and a valence band level at +1.40 V; maghemite has the two band levels at −0.08 and +1.94 V, respectively (Leland and Bard 1987). In the mixture, the excited electrons at the conduction band of hematite would transfer to the conduction band of maghemite easily. This leads to an effective electron separation and transformation within the iron oxides.

Conclusion

Pure α-Fe2O3/γ-Fe2O3 nanopowders were proved to be promising as photocatalysts in the removal of the cyanotoxin MC-LR. Higher specific surface area of the photocatalysts led to a significant increase of the degradation efficiency. The catalyst loading of 0.5 g L−1 was shown to be optimal; lower or higher loadings resulted in a decrease of the efficiency under the experimental conditions in this study. Solution pH proved to be an important parameter for Fe2O3 photocatalysis for the removal of MC-LR due to iron dissolution. The dissolved iron could react with formed H2O2 during the photocatalysis process, improving the efficiency of the photocatalysis process via Fenton reaction. Even though the highest photocatalytic degradation of MC-LR was achieved at pH 3, solution pH needs to be carefully controlled to avoid iron contamination problems by iron dissolution. These investigations could be continued by a test of other Fe2O3 nanoparticulate systems, like amorphous Fe2O3 or beta-Fe2O3, to improve the efficiency of iron oxide photocatalysis for the degradation of contaminants in water. Also, in future, it would be interesting to confirm whether the Fe2O3 catalysts could be regenerated and reused for more than one photocatalytic test/cycle, e.g., by employing of magnetic separation.