Production of martite nanoparticles with high energy planetary ball milling for heterogeneous Fenton-like process

Abstract

Natural martite microparticles (NMMs) were prepared with a high energy planetary ball mill to form a nanocatalyst for a Fenton-like process. Martite nanoparticles (MNs) of different scales are formed when the milling time ranges from 1 to 5 h at the milling speed of 300 rpm. The catalytic performances of MNs are higher than the NMMs for the degradation of acid blue 5 (AB5) in a heterogeneous Fenton-like process. The NMMs and the MNs were characterized by SEM, EDX, BET, XRD and FT-IR analyses. The size distribution of the 5 h milled martite nanoparticles (MN₃) is in the range of 20 nm to 100 nm, and these have the highest surface area (19.23 m² g⁻¹). The influence of the main operational parameters, including initial pH, MN₃ dosage, H₂O₂ and initial dye concentration, were investigated on the AB5 degradation. The treatment process obeys pseudo first order kinetics and some of the degradation intermediates were recognized by the GC-MS method. The environmentally-friendly production of the MNs, low amount of leached iron and repeated catalyst usage are the significant advantages of this research. Finally, an artificial neural network (ANN) is expanded to estimate the degradation efficiency of AB5 on the basis of the experimental results, which indicates the appropriate performance (R² = 0.955).

---

1. Introduction

In a variety of industries such as textile, leather, paper, cosmetic and food, large amounts of dyes are utilized annually. Discharge of these organic contaminants to the environment via industrial wastewater can lead to pollution of water sources.^1–4 For instance, triarylmethane dyes like acid blue 5 (AB5) are widely used in the textile industry for wool, nylon, cotton and silk as well as for coloring fats, oils, varnishes, plastics and waxes; serious problems can be occur in aquatic ecosystems and human health owing to their toxicity and carcinogenicity.^5–7

The development of novel and highly efficient methods like advanced oxidation processes (AOPs) for the treatment of wastewaters, which can degrade and mineralize the pollutants whilst producing no secondary waste, is vital from an environmental viewpoint.^8–10 Among AOPs as green treatment technologies, the homogenous Fenton process has been extensively applied for the degradation of different contaminants in aquatic media.^11–14 However, this process has some drawbacks including the formation of iron-containing sludge and the deactivation of catalyst by intermediates generated during oxidation.^15–18

Usage of heterogeneous Fenton process without need for the catalyst segregation and low released iron to the solution is the practical method to decline the mentioned problems.¹⁹ Reactive species, particularly hydroxyl radicals (˙OH), have the main function in the AOPs owing to unselective oxidation of contaminants into inorganic mineral salts, water and carbon dioxide.²⁰ Superficial solid Fe ions in the heterogeneous Fenton-type catalysts such as iron-substituted zeolites,²¹ LAPONITE®,²² pyrite,²³ magnetite,^15,24,25 goethite²⁶ and pillared clays²⁷ catalyze the formation of ˙OH radicals. Martite (Fe₂O₃) which yielded upon conversion of magnetite (Fe₃O₄) into hematite (Fe₂O₃) indicates the physical properties of the latter.^28,29 It is a mineral, which can also be used as the potential Fenton-like catalyst. Since, generation of hydroxyl radicals from the decomposition of hydrogen peroxide is initiated in the presence of ferric ions from martite, it is commonly referred as “Fenton-like” although ferrous iron and ferric ions are both present in the cycle of reactions.³⁰

However, the heterogeneous Fenton process has also some restrictions including limited active reaction sites and low mass transfer. The heterogeneous catalytic performance can be enhanced by applying of nano-sized particles due to providing of a short diffusion path and more active sites for organic pollutants.^31–33 Production of nanoparticles by chemical syntheses require expensive and toxic reactants.³⁴ Recently, mechanical ball milling (MBM) via high energy planetary ball mill has been used for formation of nanoparticles owing to its cost effectiveness and simplicity.^27–29 Moreover, the ball milling method is a suitable method for large scale industrial productions from an economical viewpoint since the process takes place at ambient temperature in a reasonable processing time.^31,32 For instance, diverse nanomaterials such as ZnS,³⁴ LiF,³³ FePt³⁵ and Cu₂O³⁶ have been produced using the high energy MBM process.

Since the heterogeneous Fenton process is influenced by various operational parameters, it is hard to simulate and model this process applying conventional mathematical models. The Artificial Neural Networks (ANNs) are recently employed in various science and engineering fields, which are effective method owing to their logical prediction, modelling and simulation.^35–37 Furthermore, the main advantages of this modelling over other techniques are no requirement for customary mathematical training and ability to predict the results with less number of tests resulted in saving money and time.³⁸

In this study, first, the MBM is applied for production of martite nanoparticles by varying the milling time from 1 h to 5 h. Then, the properties of the NMMs and as-prepared samples were characterized by scanning electron microscopy (SEM), energy-dispersive X-ray spectroscopy (EDX), Brunauer, Emmett and Teller (BET), X-ray diffraction (XRD) and Fourier transform infrared spectroscopy (FT-IR). Next, the catalytic performance of the NMMs and MNs was determined in the heterogeneous Fenton-like process for degradation of AB5 as a model triarylmethane pollutant. The influence of experimental parameters including the solution pH, catalyst dosage, H₂O₂ concentration, initial AB5 concentration, and hydroxyl radical scavengers were investigated on the degradation efficiency (DE%) of the dye. Gas chromatography-mass spectrometry (GC-MS) was employed for the identification of the generated byproducts through the AB5 degradation. Finally, the artificial neural network modelling was applied as a helpful method to estimate the degradation efficiency of AB5 in heterogeneous Fenton-like process.

2. Experimental procedure

2.1. Materials

The natural martite (hematite pseudomorphs after magnetite) was supplied from Sangan mine (Mashhad, Iran). The triarylmethane dye, acid blue 5 was purchased from Shimi Boyakhsaz Co. (Iran) applied as a textile dye for wool and silk. The characteristics of AB5 involve molecular formula of C₃₇H₃₆N₂O₇S₂, colour index number of 45052, λ_max of 630 nm and molecule weight of 684.82 g mol⁻¹. Its molecular configuration is presented in Fig. 1. All of the other chemicals were purchased from Merck (Germany) and used without further purification. Distilled water was applied all over the experiments.

Fig. 1 Molecular configuration of AB5.

2.2. Martite nanoparticles preparation

First, the martite ore was crushed applying the jaw and cone crushers (Geological survey of East Azerbaijan) to the particle size of 0.3–1.5 cm. Next, the sample was more crushed by rod and ball milling (Institute of Materials and Energy (MERC), Karaj, Iran) to produce micro-grained martite particles with diameter between 35 and 55 μm. Eventually, the interrupted milling in a high energy planetary ball mill (Retsch, Model PM 400, Germany) was carried for 1 h (MN₁), 3 h (MN₂) and 5 h (MN₃) to form the martite nanoparticles. The mass ratio of balls to the particles was 10 : 1 and the rotation speed was selected as 300 rpm.

2.3. The martite nanoparticles characterization

In order to characterize the NMMs and MNs, the following analyses were carried out: (1) the SEM equipped with an EDX microanalysis (Mira3 FEG-SEM Tescan, Czech) were used after gold plating of the particles to observe dimensions, surface morphology and chemical composition of them; (2) the nanoparticles size distribution was calculated using microstructure distance measurement software (Nahamin Pardazan Asia Co., Iran); (3) nitrogen sorption analyses were carried out with a sorptometer (Autosorb-6B, Japan) using standard procedure at 77 K on calcined particles that had been degassed at 250 °C for 3 h; the surface area was determined according to the BET model over a relative pressure range of 0.048–0.98; (4) crystal structure identification was implemented via the XRD method applying a Siemens X-ray diffractometer (D8 Advance, Bruker, Germany) with Cu-kα radiation (40 kV, 30 mA, 0.15406 nm); (5) the FT-IR spectra of the samples were recorded over the range of 4000–400 cm⁻¹ wavelength by Tensor 27, Bruker spectrometer (Germany) using KBr pellet technique.

2.4. Heterogeneous Fenton-like procedure

Experiments were performed in a 500 mL beaker; in each run, a specified concentration of AB5 solution (250 mL) with different amounts of the catalyst and hydrogen peroxide (H₂O₂) was employed. The NaOH (0.1 M) and H₂SO₄ (0.1 M) were added to adjust the solution pH determined by a Metrohm pH meter (Switzerland). During the Fenton-like process, a sample (3 mL) was withdrawn, then, 1 mL of ethanol (96% w/w) was added to the sample for quenching the Fenton reaction. The particles were separated from the sample by centrifugation. The AB5 solution absorbance was measured at its maximum absorbance wavelength (λ_max = 630 nm) by an UV-Vis spectrophotometer (WPA lightwave S2000, England) to calculate the degradation efficiency (DE%).^39,40 An atomic absorption spectroscopy (AAS) (Novaa 400, Analytikjena, Germany) was applied for measurement of the leaked iron concentration in the solution. The GC-MS (Agilent 6890 gas chromatography coupled with 5973 mass spectrometer, PaloAlto, Canada) analysis was performed for identification of the generated degradation intermediates of AB5 employing a method explained in our previous study.⁴¹

3. Results and discussion

3.1. Characterization of the NMMs and MNs

The SEM images of NMMs and MNs samples are shown in Fig. 2. The bulky structure of the NMMs is observed in Fig. 2a. However, Fig. 2b–d reveal that the NMMs is gradually converted to the MNs using high energy planetary ball milling process after 5 h. The EDX spectrum of the mentioned nano-sized particles (Fig. 2e) proves the presence of Fe and O elements. There are other elements including C and Si in the structure of martite. But, C is the inert element and doesn't participate in the reaction. Also, the weight percentage of Si is less than 3% and it can be negligible in comparison to the other elements.

Fig. 2 SEM micrographs of (a) unmilled martite sample (NMMs), (b) 1 h milled (MN₁), (c) 3 h milled (MN₂) and (d) 5 h milled (MN₃) nanosized martite with different magnifications. (e) EDX spectrum of martite nanoparticles after 5 h ball milling.

Table 1 demonstrates the size distribution of the NMMs, MN₁, MN₂ and MN₃. 46.23% of size distribution of the MN₃ is in the range of 40–60 nm. In other words, it proves that the particle size of the martite has been increased by passing the ball milling time.

Table 1 Size distribution of NMM, MN₁, MN₂ and MN₃

Frequency (%)
Microparticles size distribution (μm)	30–50	50–70	70–90	90–140
NMM	21.43	38.57	25.71	14.29

---

Nanoparticles size distribution (nm)	20–40	40–60	60–80	80–100
MN1	13.11	36.24	17.79	32.86
MN2	20.00	41.50	21.50	17.00
MN3	29.50	46.23	15.61	8.66

---

Fig. 3 displays the nitrogen adsorption–desorption isotherms of catalysts samples observed at 77.38 K.^42,43 According to the IUPAC classification, a simple review indicates that the experimental isotherms are of type IV. However, a more detailed observation of the figure reveals that it is a combination of type IV and V isotherms which characterize mesoporous structure for the martite sample. Therefore, the isotherms can be appropriately defined by the BET surface area method.⁴⁴ Besides, the hysteresis shaping started from nearly 0.55 and extended almost to 0.95 demonstrating their large porosity.⁴⁵ Table 2 presents the specific surface area, pore diameter and pore volume of the catalysts. As seen in the table, after 5 hours of MBM process, specific surface area, pore diameter and pore volume of the catalysts have been increased from 5.675 to 19.23 m² g⁻¹, from 17.20 to 32.24 Å and from 7.961 × 10⁻³ to 2.984 × 10⁻² cm³ g⁻¹, respectively. It is because of increase in collisions between balls and martite particles in the planetary ball milling process by passing time. Therefore, under effect of this collision, pore diameter, pore volume and surface area of martite nanoparticles increase.^46,47

Fig. 3 N₂ adsorption–desorption isotherm of (a) NMMs, (b) MN₁, (c) MN₂ and (d) MN₃ samples achieved from the BET test.

Table 2 Surface area, pore diameter and volume characteristics of NMMs, MN₁, MN₂ and MN₃ samples

Sample	Primary martite	1 h-milled martite	3 h-milled martite	5 h-milled martite
Specific surface area (m^2 g^−1)	5.675	15.28	16.47	19.23
Pore diameter (Å)	17.20	21.36	25.83	32.24
Pore volume (cm^3 g^−1)	7.961 × 10^−3	1.425 × 10^−2	2.201 × 10^−2	2.984 × 10^−2

---

Fig. 4 shows XRD patterns of the NMMs, the 1 h (MN₁), 3 h (MN₂) and 5 h (MN₃) milled martite nanoparticles. The obtained peaks at 2θ values of 24.13, 33.2, 35.44, 40.96, 49.46, 53.87, 57.19, 62.23, and 64.04° are related to the characteristic (0 1 2), (1 0 4), (1 1 0), (1 1 3), (1 2 4), (1 1 6), (1 2 2), (2 1 4), and (3 0 0) planes. The observed XRD patterns are very close to the hematite (JCPDS no. 33-664). This is a rational finding because the martite is hematite pseudomorphs after magnetite and its physical properties is similar to the hematite.^48,49Also, observation of identical peaks indicate that the primary martite sample is maintained after the MBM process. In addition, the peaks at 2θ values of 43.10 and 57.08° are relevant to the (4 0 0) and (5 1 1) reflection planes of the magnetite (JCPDS no. 19-0629), respectively.⁵⁰ The reduction in XRD peak intensity takes place with increase in ball milling time. It is owing to the crystallinity decrease of the as-obtained nanosized martite.⁵¹

Fig. 4 XRD pattern of unmilled and ball-milled martite nanoparticles.

The FT-IR analyses indicated in Fig. 5 were carried out to find the existing bonds and functional groups of the NMMs (curve (a)) and the milled samples (curves (b–d)). The peaks at 2854.69 and 2923.42 cm⁻¹ are attributed to the symmetric and asymmetric C–H bonds, respectively.^52,53 The O–H vibration peak at 3436.58 cm⁻¹ indicates the presence of O–H group.⁵⁴ The peaks of the –COO– vibration (1386.19 and 1598.89 cm⁻¹), the Si–O vibration (1094.96 cm⁻¹) and the Fe–O vibration (467.42 and 552.64 cm⁻¹) are also found in all samples.^55–57 Thus, the functional groups on the martite nanoparticles are identical to the ones in unmilled sample. Moreover, the vibration peak intensity of the Fe increases after passing time of ball milling process which is due to the surface oxidation of the martite nanostructures.⁵¹ On the other hand, the –COO– vibration peak, symmetric and asymmetric C–H bonds intensity reduces which is owing to the partial oxidation of the compounds.⁵⁸

Fig. 5 FT-IR spectra of (a) NMMs (b) MN₁, (c) MN₂ and (d) MN₃ samples.

3.2. Comparison of different catalysts in heterogeneous Fenton-like process

The Fenton-like process was used for the treatment of a model triarylmethane pollutant in the aqueous solution using the NMMs and diverse ball milled samples as catalyst. The adsorption of AB5 by all the catalysts is lower than 7% in the dark after 70 min. In the heterogeneous Fenton-like process in which Fe³⁺ ions are applied, the following reactions are involved to form the reactive oxygen species:^7,15,58,59

Fe³⁺–martite + H₂O₂ → Fe–martite–OOH²⁺ + H⁺ (1)

Fe–martite–OOH²⁺ → Fe²⁺–martite + HO₂˙ (2)

Fe²⁺–martite + H₂O₂ → Fe³⁺–martite + HO⁻ + ˙OH (3)

The DE% for the Fenton-like process during 70 min of contact time are shown in Fig. 6. The experiments containing martite catalyst shows higher degradation efficiency compared to the ones without the catalyst. The highest degradation efficiency is observed when MN₃ is used as a catalyst.

Fig. 6 Comparison of the degradation efficiency of AB5 with different processes (curve a) adsorption, (curve b) H₂O₂ oxidation, (curve c) NMMs, (curve d) MN₁, (curve e) MN₂, (curve f) MN₃ in heterogeneous Fenton-like process. Experimental conditions: [AB5]₀ = 10 mg L⁻¹, [Martite] = 2 g L⁻¹, [H₂O₂]₀ = 3 mM, and pH = 3. The inset depicts apparent pseudo-first order reaction kinetic.

The inset of Fig. 6 shows that the heterogeneous Fenton-like process obeys the pseudo-first order kinetic, which is comparable to the other Fenton studies.⁶⁰ The straight lines with appropriate coefficient of determinations (R²), which are more than 0.99, verify the validity of suggested kinetic. The apparent pseudo-first order rate constants (k_app) for the AB5 degradation are given in Table 3. The k_app increases by using the martite catalyst and increasing ball milling time.

Table 3 Effect of the different processes on the apparent pseudo-first order constants of degradation of AB5

No.	Process	kapp (min^−1)	Normalized kapp (min m^2 g^−1)^−1	Coefficient of determination (R^2)
1	Adsorption	0.0010	—	0.9988
2	H2O2	0.0030	—	0.9950
3	NMMs	0.0120	2.11 × 10^−3	0.9989
4	MN1	0.0269	1.76 × 10^−3	0.9973
5	MN2	0.0344	2.09 × 10^−3	0.9982
6	MN3	0.0415	2.16 × 10^−3	0.9991

---

In order to consider the degradation kinetic of the AB5 through different processes, experimental data are fitted with the pseudo-first order kinetic equation (eqn (4)):⁶¹

where A indicates the absorbance of AB5 at a reaction time of t min, A₀ is the initial AB5 absorbance at 0 min, and k_app is the pseudo-first order rate constant (min⁻¹). The amount of k_app is gained from the linear relationship of ln(A₀/A_t) versus time.

The apparent pseudo-first order constants are directly correlated with the specific surface area and pore volume. The highest kinetic rate is observed with the catalysts that has the highest specific surface area and pore volume. Also as seen in Table 3, when the apparent pseudo-first order reaction constant is normalized with the specific surface area of catalysts, the rate constants are very close to each other, indicating the importance of specific surface area of the catalyst on the AB5 oxidation. The catalyst MN₃ has the highest apparent pseudo-first order reaction constant (Table 3), specific surface area and pore volume (Table 2) among unmilled catalyst and catalyst milled for a shorter period of time. Also, at the end of 70 minutes contract time, the highest degradation ratio of 94.5% was observed in the reactor containing MN₃. As a consequence, the MN₃ is chosen as the desired nanocatalyst for the rest of experiments in the degradation process. The modified NMMs by the mechanical ball milling leads to more surface area and active sites owing to the production of nano-sized particles and thus its enhanced performance.^38,54,57

3.3. Effect of the operational parameters on the heterogeneous Fenton-like process

The effect of main operational parameters including pH, MN₃ dosage, and initial H₂O₂ and AB5 concentration, on the degradation efficiency of the organic pollutant was studied.

One of the most significant operational parameters in the Fenton-like process is pH. Hence, its effect on the degradation of AB5 was studied, and the results are displayed in Fig. 7. The inset plot shows the pseudo-first order kinetic (k_app) for different pHs. As it is seen, the amount of k_app increases as pH decreases. At low pHs, higher degradation efficiency is achieved, and the DE% diminishes with the enhancement of pH.⁶² This can be attributed to more oxidation potential of the ˙OH radicals in acidic condition;⁶³ moreover, the dissolved iron concentration increases in lower pHs determined by AAS (Fig. 8). The highest DE% is achieved when the pH is adjusted to 3 after 70 min of treatment. It should be noticed that, the total leached iron concentration is lower than 1 mg L⁻¹ after heterogeneous Fenton-like process in the presence of MN₃ for the pH range of 3 to 9. As mentioned in the other similar studies, at this low level of iron concentration, the ˙OH radicals generation is almost owing to the heterogeneous process on the catalyst surface compared to the homogeneous one in the bulk solution for the AB5 degradation.^60,64

Fig. 7 The effect of suspension pH on the degradation of AB5 in the heterogeneous Fenton process. Experimental conditions: [AB5]₀ = 10 mg L⁻¹, [MN₃] = 2.5 g L⁻¹, [H₂O₂]₀ = 3 mM. The inset depicts apparent pseudo-first order reaction kinetic.

Fig. 8 Dissolved iron concentration in solution phase after 70 min; [MN₃] = 2.5 g L⁻¹.

Fig. 9 shows the DE% as a function of process time for the different MN₃ dosages. The DE% increases over the initial dosage range of 0 to 2.5 g L⁻¹; this is due to the further available sites for radicals formation and AB5 adsorption.⁶⁵ Afterwards, an inverse trend is seen, and the DE% reduces. This observation is due to the scavenging effect of ferrous ions on the hydroxyl radicals (eqn (5)).^66–68 Also, the inset plot demonstrates the degradation of AB5 in the presence of the martite nanocatalysts following pseudo-first order kinetic. As the MN₃ dosage increases, k_app increases. However, it reduces beyond 2.5 g L⁻¹ of the catalyst.

˙OH + Fe²⁺ → OH⁻ + Fe³⁺ (5)

Fig. 9 The effect of martite dosage (MN₃) on the degradation of AB5 in heterogeneous Fenton process. Experimental conditions: [AB5]₀ = 10 mg L⁻¹, [H₂O₂]₀ = 3 mM, and pH = 3. The inset depicts apparent pseudo-first order reaction kinetic.

It is essential to optimize the H₂O₂ concentration because a main expense of the Fenton-like process is the cost of H₂O₂. In addition, an excessive oxidant amount decreases the process efficiency.⁶⁹ Fig. 10 shows that the DE% increases by H₂O₂ addition up to 2 mM. This observation is due to the extra reactive radicals generation from the H₂O₂ decomposition in the presence of catalyst (eqn (1)–(3)). But, more oxidant addition leads to decline in the DE%;⁶¹ hydrogen peroxide scavenging effect on hydroxyl radicals declines the dye degradation (eqn (6)).^70–74 The inset plot presents that k_app increases up to 2 mM, but over 2 mM it diminishes.

H₂O₂ + ˙OH → H₂O + HO₂˙ (6)

Fig. 10 The effect of H₂O₂ concentration on the degradation of AB5 in heterogeneous Fenton-like process. Experimental conditions: [AB5]₀ = 10 mg L⁻¹, [MN₃] = 2.5 g L⁻¹ and pH = 3. The inset depicts apparent pseudo-first order reaction kinetic.

Increasing the AB5 concentration reduces the DE% (Fig. 11). In the same operational conditions, the generated active radicals amount and the MN₃ accessible surface area are constant even with enhancement of the AB5. Besides, because of the competitive consumption of the ˙OH radicals through the formed by-products at higher dye concentration, the DE% diminishes.^67,75–77 Even though the DE% of AB5 reduced with increasing initial dye concentration, the absolute dye removal has increased. For instance, as the dye concentration is increased from 10 to 40 mg L⁻¹, the degradation of AB5 has been increased after 50 min from 9.91 to 17.48 mg L⁻¹, respectively.⁷⁷ Also, as observed from the inset plot in Fig. 11, the pseudo-first order kinetic (k_app) is reduced with increasing initial AB5 concentration from 10 to 40 mg L⁻¹.

Fig. 11 The effect of initial dye concentration on the degradation of AB5 in heterogeneous Fenton-like process. Experimental conditions: [MN₃] = 2.5 g L⁻¹, [H₂O₂]₀ = 2 mM and pH = 3. The inset depicts apparent pseudo-first order reaction kinetic.

The influence of inorganic salts and organic compounds such as NaCl, Na₂SO₄, Na₂CO₃, ethanol and chloroform were investigated on the AB5 degradation. Chloride, sulfate and carbonate are common observed anions in the textile wastewaters. They affect the efficiency of the treatment, which is named the salting-out effect.⁶⁰ As can be observed in Fig. 12, the DE% reduces by the addition of these scavengers to the AB5 solution. The ions with negative charges are adsorbed on the martite nanocatalysts surface and active superficial sites of the catalyst are occupied by them inhibiting the Fenton reaction.^41,78 Therefore, this reduces the amount of hydroxyl radicals generation. The impact of NaCl over the degradation of AB5 is higher than that of Na₂CO₃ and Na₂SO₄. The reason can be ascribed to the scavenging of ˙OH by chloride and the generation of reactive radicals with low oxidation potential instead of ˙OH only given by the eqn (7)–(9).^41,60,79 Also, ˙Cl produced in eqn (9) has a high tendency to react with H₂O₂, which decreases ˙OH concentration in the process (eqn (10)).⁸⁰

Cl⁻ + ˙OH → ˙ClOH⁻ (7)

˙ClOH⁻ + H⁺ → H₂O + ˙Cl (8)

Cl⁻ + ˙Cl → ˙Cl₂⁻ (9)

˙Cl + H₂O₂ → H⁺ + Cl⁻ + HO₂˙ (10)

Fig. 12 Effect of scavengers on the degradation of AB5 in heterogeneous Fenton-like process. (a) Without scavenger, (b) sodium sulfate, (c) sodium chloride, (d) chloroform, (e) sodium carbonate and (f) ethanol. Experimental conditions: [AB5]₀ = 1.46 × 10⁻⁵ N, [MN₃] = 2.5 g L⁻¹, [H₂O₂]₀ = 2 mM, pH = 3, [NaCl]₀ = 1.71 × 10⁻⁴ N, [Na₂CO₃]₀ = 1.88 × 10⁻⁴ N, [Na₂SO₄]₀ = 1.41 × 10⁻⁴ N, [CH₃CH₂OH]₀ = 2.17 × 10⁻⁴ N and [CHCl₃]₀ = 8.38 × 10⁻⁵ N. The inset depicts apparent pseudo-first order reaction kinetic.

Carbonate ion reacts with ˙OH through eqn (11) and results in the production of CO₃˙⁻. It has lower oxidation potential than that of ˙OH; so, the degradation efficiency is reduced.

CO₃²⁻ + ˙OH → CO₃˙⁻ + OH (11)

Sulfate ion can also react with ˙OH (eqn (12)) and decreases its concentration in the solution. But, according to Fig. 12, the scavenging effect of sulfate ion is less than the others. It is because of peroxydisulfate anion (S₂O₈²⁻) produced by eqn (13). Its oxidation potential is 2.01 eV.⁸¹ which is less than the oxidation potential of ˙OH.

SO₄²⁻ + ˙OH → SO₄˙⁻ + OH⁻ (12)

SO₄˙⁻ + SO₄˙⁻ → S₂O₈²⁻ (13)

Furthermore, ethanol and chloroform as organic compounds can scavenge ˙OH in the solution. Due to eqn (14) and (15), the reduction in the degradation efficiency of AB5 occurs.^82,83

˙OH + CH₃CH₂OH → CH₃˙CHOH + H₂O (14)

˙OH + CHCl₃ → ˙CCl₃ + H₂O (15)

The inset plot shows the gradient of k_app in the presence of different scavengers. As it can be seen, the k_app decreases in the presence of ethanol more than the other scavengers. On the other hand, it reduces in the presence of sodium sulfate less than the others. The reaction rate constants for eqn (12) and (14) are (0.95 ± 0.08) × 10¹⁰ M⁻¹ s⁻¹ and (0.16 ± 0.02) × 10¹⁰ M⁻¹ s⁻¹, respectively.^84,85

In all runs, the degradation rate also obey the pseudo-first order kinetic and the apparent reaction rate constants and the corresponding coefficient of determination are estimated from the inset plots of Fig. 7–12 and presented in Table 4.

Table 4 Effect of the operational parameters on the apparent pseudo-first order constant of degradation for Fenton-like process

Operational parameters and amounts	kapp (min^−1)	Coefficient of determination (R^2)
a Experimental conditions: [AB5]0 = 10 mg L^−1, [MN3] = 2.5 g L^−1 and [H2O2]0 = 3 mM.b Experimental conditions: [AB5]0 = 10 mg L^−1, pH = 3 and [H2O2]0 = 3 mM.c Experimental conditions: [AB5]0 = 10 mg L^−1 [MN3] = 2.5 g L^−1 and pH = 3.d Experimental conditions: [MN3] = 2.5 g L^−1, pH = 3 and [H2O2]0 = 2 mM.e Experimental conditions: [AB5]0 = 1.46 × 10^−5 N, [MN3] = 2.5 g L^−1, [H2O2]0 = 2 mM and pH = 3.
pH^a
3	0.0490	0.9996
4	0.0197	0.9988
5	0.0126	0.9987
7	0.0093	0.9990
9	0.0058	0.9982
Catalyst dosage^b (g L^−1)
1	0.0167	0.9995
1.5	0.0191	0.9996
2	0.0415	0.9991
2.5	0.0490	0.9996
3	0.0281	0.9985
H2O2 concentration^c (mM)
0	0.0010	0.9988
1	0.0195	0.9966
2	0.0960	0.9975
3	0.0490	0.9996
4	0.0263	0.9981
5	0.0215	0.9983
AB5 concentration^d (mg L^−1)
10	0.0960	0.9975
20	0.0317	0.9988
30	0.0180	0.9989
40	0.0118	0.9978

---

Scavengers^e	Concentration (N)	kapp (min^−1)	Coefficient of determination (R^2)
NaCl	1.71 × 10^−4	0.0327	0.9987
Na2CO3	1.88 × 10^−4	0.0216	0.9962
Na2SO4	1.41 × 10^−4	0.0394	0.9993
CH3CH2OH	2.17 × 10^−4	0.0102	0.9965
CHCl3	8.38 × 10^−5	0.0272	0.9990

---

3.4. Reusability of the nanocatalyst and the dye degradation intermediates

One of the most significant properties of an ideal catalyst from practical view point is its reusability in successive applications. Thus, the MN₃ was used in five repeated processes. After each experiment, the nanoparticles are separated from the solution and used for the next experiment. The degradation efficiency for these five consecutive runs is indicated in Fig. 13. The results reveal that the DE% is not significantly altered after the successive usage of the catalyst. Besides, the leached iron amounts to the solution (Fig. 8) demonstrates that at pH 3, the dissolved iron concentration is 0.98 mg L⁻¹. Considering that in each run, the MN₃ dosage was 2.5 g L⁻¹, the leaked iron mass was less than 0.1% of the total catalyst mass and major part of iron remained in the martite structure. Therefore, the catalyst somewhat showed stability but still acceptable.

Fig. 13 Reusability of the martite nanocatalyst for five successive cycles. Experimental conditions: [AB5]₀ = 10 mg L⁻¹, [MN₃] = 2.5 g L⁻¹, [H₂O₂]₀ = 2 mM and pH = 3.

The UV-vis spectra along with GC-MS analysis were performed to specify the gradient in the AB5 amount over the Fenton-like process under optimum achieved condition. Fig. 14 shows the dye absorption peaks decline with time passing owing to the dissociation of AB5 molecular structure in the heterogeneous Fenton-like process. Besides, the identified intermediates of the dye degradation by the Fenton-like process are given in Table 5. The AB5 conversion into the smaller compounds occurred by breaking of chemical bonds including C–C, C–N, or N=N.⁸⁶ However, all intermediates identification is not conceivable due to their slight accumulation and limitations of the GC-MS analysis.⁸⁷

Fig. 14 UV-vis spectra changes of AB5 solution during the Fenton-like process under the optimized conditions.

Table 5 Identified by-products during removal of acid blue 5 by Fenton-like process

No.	Compound	Structure	tr (min)	Main fragments (m/z)
a Value corresponding to the trimethylsilyl derivative.b Value corresponding to the bis-trimethylsilyl derivative.
1^a	2-Hydroxy-succinic acid		9.41	147 (100%), 73 (53.46%), 148 (15.42%)
2^a^,^b	1-Aminoethanol		9.81	73 (100%), 147 (59.71%), 261.1 (21.12%), 113 (19.83%)
3^a	1,4-Benzenediol,2,5-bis(1,1-dimethylethyl)-		21.93	207.10 (100%), 221.1 (59.43%), 193 (18.66%)
4^a	2-Hydroxyacrylic acid		6.05	147 (100%), 73 (74.07%), 217.1 (23.885), 148 (15.73%)
5^a	1,5-Naphthalenediol		18.21	304 (100%), 73 (48.67%), 305.1 (26.92%)
6^a	2-[(Trimethylsilyl)oxy]-5-methylacetophenone		36.18	273 (100%), 73 (75.63%), 147 (58.62%), 363.1 (23.57%), 347.1 (21.83%)
7^a	4-Hydroxybutyric acid		13.99	147 (100%), 73 (49.45%), 75 (33.92%), 117 (33.28%)

---

4. Neural network modelling of Fenton-like process

The optimization of a topology functions a significant role in the development of an ANN model. The important parameters in an ANN topology are the number of layers, number of neurons in each individual layer and the transfer functions.^1,35,88 Here, four-layered feed forward back propagation neural network neural network (5 : 7 : 9 : 1) was utilized for modelling of AB5 degradation by heterogeneous Fenton-like process. Fig. 15 shows the architecture of performed network in which five input layer neurons corresponding to the five network input variables, seven and nine neurons in the hidden layers 1 and 2, respectively and the output layer with a neuron related to output variable. In this study, input variables to feed forward neural network are reaction time (min), initial pH, initial catalyst dosage (g L⁻¹), initial dye concentration (mg L⁻¹) and initial hydrogen peroxide concentration (mM). The degradation efficiency (%) is established as the output variable or experimental reply. The input variables ranges are depicted in Table 6. The optimum topology of the ANN model, weights and biases are specified based upon the smallest amount of the mean square error (MSE) of the training and validation series.^89,90

Fig. 15 The artificial neural network optimized structure.

Table 6 Variables of the ANN model and their ranges

Variable	Range
Input layer
Time (min)	0–70
Initial pH	3–9
Catalyst dosage (g L^−1)	0–3
Dye concentration (mg L^−1)	10–40
H2O2 concentration (mM)	0–5
Output layer
Removal efficiency (%)	0–97

---

Various topologies were used to obtain the optimal number of neurons in hidden layers 1 and 2. The number of neurons were ranged from 3 to 11 for the first hidden layer and 0 to 9 for the second one. Fig. 16 illustrates the MSE achieved from different topological runs, where h₂ is the number of neurons in the second layer. Each topology was reiterated four times to prevent random correlation owing to random initialization of the weights. The MSE demonstrates a network performance according to eqn (16):

Fig. 16 Effect of number of neurons in the hidden layers on the performance of the ANN model; h₂ demonstrates the number of neurons in the second hidden layer.

Q represents the number of test data point, X_i,pred and X_i,exp are relevant to the predicted and experimental responses, respectively.^89,91,92 The lowest value of MSE is obtained as 3.3249 × 10⁻⁵ through 7 neurons in first hidden layer and 9 neurons in the second one.

The training function used was trainscg, and the tan-sigmoid transfer function was employed as a transfer function in first and second hidden layers. This function is commonly applied and it is given in eqn (17):⁹³

Among several data point, 165 experimental data were applied to feed the ANN model. The samples were separated into training, validation and test subsets included 99, 36 and 30 sets, respectively. In order to evaluate the ability of validation and modelling of the developed ANN, the validation and test data were chosen by chance. In view of the fact that the transfer function applied in the hidden layers was sigmoid, all of the data should be in the range of 0.2–0.8. Therefore, each data (X_i) including the training, validation and test series was converted into a new value X_norm as follows:

X_i,min and X_i,max are minimum and maximum amount of X_i variable.⁹⁴ After running, all outputs were turned back to their original value in order to figure out training, validation and test errors according to the eqn (18). Then, the comparison between the experimental replies and predicted outputs were conducted (Fig. 17). As it is seen in this figure, the coefficient of determination is 0.955 which exhibits the credibility of the obtained model. Also, the results prove that the ANN model regenerated the degradation efficiency in this process, within experimental ranges adopted in the fitting model.^89,95

Fig. 17 Comparison between experimental and predicted values via the ANN modelling for the test data.

5. Conclusions

In this study, environmentally friendly high energy planetary ball milling process was used for production of the martite nanoparticles, which confirmed by SEM, EDX, BET, XRD and FT-IR, techniques. The promoted catalytic activity was obtained for the degradation of the triarylmethane dye in the heterogeneous Fenton-like process using the ball-milled nanocatalyst. The most effective modified nanocatalyst was generated by the MBM process after 5 h owing to the highest generated surface area and active sites, which verified by the SEM and BET analyses. The desired experimental conditions for the AB5 degradation were found to be pH 3, MN₃ dosage of 2.5 g L⁻¹, initial dye concentration of 10 mg L⁻¹, and H₂O₂ concentration of 2 mM after 70 min of process. The treatment process follows pseudo-first order kinetic. The hydroxyl radical scavengers diminished the degradation efficiency of the dye confirming the active radicals role for the AB5 degradation. The main advantages of the MN₃ were its potential for application in the successive runs and low level of iron leached into the solution. GC-MS analysis identified formation of seven different intermediates from degradation of AB5. Eventually, the neural network modelling was successfully employed to predict the process performance for the AB5 degradation by using a four-layered neural network with 7 and 9 neurons in first and second hidden layer, respectively.

Acknowledgements

The authors thank the University of Tabriz and Middle East Technical University for the support provided.

References

1. A. R. Khataee and M. Zarei, Desalination, 2011, 273, 453–460.
2. M. Fathinia, A. R. Khataee, M. Zarei and S. Aber, J. Mol. Catal. A: Chem., 2010, 333, 73–84.
3. A. R. Khataee, M. Fathinia, S. Aber and M. Zarei, J. Hazard. Mater., 2010, 181, 886–897.
4. A. Petrella, M. Petrella, G. Boghetich, P. Mastrorilli, V. Petruzzelli, E. Ranieri and D. Petruzzelli, Ind. Eng. Chem. Res., 2013, 52, 2201–2208.
5. C.-C. Chen, H.-J. Fan and J.-L. Jan, J. Phys. Chem. C, 2008, 112, 11962–11972.
6. T. Gessner and U. Mayer, in Ullmann's Encyclopedia of Industrial Chemistry, Wiley-VCH Verlag GmbH & Co. KGaA, 2000,  DOI:10.1002/14356007.a27_179.
7. O. Iglesias, E. Rosales, M. Pazos and M. A. Sanromán, Environ. Sci. Pollut. Res., 2013, 20, 2252–2261.
8. C. A. Orge, J. J. M. Orfao, M. F. R. Pereira, A. M. Duarte de Farias and M. A. Fraga, Chem. Eng. J., 2012, 200–202, 499–505.
9. M. Neamtu, C. Zaharia, C. Catrinescu, A. Yediler, M. Macoveanu and A. Kettrup, Appl. Catal., B, 2004, 48, 287–294.
10. M. Dukkanci, G. Gunduz, S. Yılmaz and R. V. Prihoduko, J. Hazard. Mater., 2010, 181, 343–350.
11. M. Sheydaei, S. Aber and A. Khataee, J. Ind. Eng. Chem., 2013, 20, 1772–1778.
12. E. Neyens and J. Baeyens, J. Hazard. Mater., 2003, 98, 33–50.
13. A. R. Khataee, M. Safarpour, A. Naseri and M. Zarei, J. Electroanal. Chem., 2012, 672, 53–62.
14. C. n. T. Muranaka, C. Julcour, A.-M. Wilhelm, H. Delmas and C. A. O. Nascimento, Ind. Eng. Chem. Res., 2009, 49, 989–995.
15. P. V. Nidheesh, RSC Adv., 2015, 5, 40552–40577.
16. P. V. Nidheesh, R. Gandhimathi and S. T. Ramesh, Environ. Sci. Pollut. Res., 2013, 20, 2099–2132.
17. P. V. Nidheesh and R. Gandhimathi, RSC Adv., 2014, 4, 27946–27954.
18. P. V. Nidheesh and R. Rajan, RSC Adv., 2016, 6, 5330–5340.
19. M. Sheydaei, S. Aber and A. Khataee, J. Mol. Catal. A: Chem., 2014, 392, 229–234.
20. A. L. Petre, J. B. Carbajo, R. Rosal, E. Garcia-Calvo and J. A. Perdigón-Melón, Chem. Eng. J., 2013, 225, 164–173.
21. M. Tekbas, H. C. Yatmaz and N. Bektas, Microporous Mesoporous Mater., 2008, 115, 594–602.
22. B. Iurascu, I. Siminiceanu, D. Vione, M. A. Vicente and A. Gil, Water Res., 2009, 43, 1313–1322.
23. W. Wang, Y. Qu, B. Yang, X. Liu and W. Su, Chemosphere, 2012, 86, 376–382.
24. X. Liang, Y. Zhong, H. He, P. Yuan, J. Zhu, S. Zhu and Z. Jiang, Chem. Eng. J., 2012, 191, 177–184.
25. P. V. Nidheesh, R. Gandhimathi, S. Velmathi and N. S. Sanjini, RSC Adv., 2014, 4, 5698–5708.
26. Z.-R. Lin, L. Zhao and Y.-H. Dong, Chemosphere, 2015, 141, 7–12.
27. Q. Chen, P. Wu, Z. Dang, N. Zhu, P. Li, J. Wu and X. Wang, Sep. Purif. Technol., 2010, 71, 315–323.
28. A. K. Jouhari, P. S. R. Reddy, B. K. Mohapatra and V. N. Misra, Raw material preparation for metallurgical industries, Allied Publishers Private Limited, pp. 99-104, ISBN 81-7764-292-8.
29. A. Mücke and A. Raphael Cabral, Chem. Erde, 2005, 65, 271–278.
30. M. Munoz, Z. M. de Pedro, J. A. Casas and J. J. Rodriguez, Appl. Catal., B, 2015, 176–177, 249–265.
31. M. Sheydaei and A. Khataee, Ultrason. Sonochem., 2015, 27, 616–622.
32. C. Yin, Y. Wei, F. Wang and Y. Chen, Mater. Lett., 2013, 98, 194–196.
33. Z. Cheng, S. Han, W. Sun and Q. Qin, Mater. Lett., 2013, 95, 193–196.
34. H. Jung, J.-W. Kim, H. Choi, J.-H. Lee and H.-G. Hur, Appl. Catal., B, 2008, 83, 208–213.
35. A. R. Khataee, B. Vahid, B. Behjati and M. Safarpour, Environ. Prog. Sustainable Energy, 2013, 32, 557–563.
36. E. Dogan, A. Ates, E. C. Yilmaz and B. Eren, Environ. Prog., 2008, 27, 439–446.
37. N. Muhd Julkapli, S. Bagheri and S. Bee Abd Hamid, Sci. World J., 2014, 2014, 25.
38. J. V. Tu, J. Clin. Epidemiol., 1996, 49, 1225–1231.
39. A. Khataee, P. Gholami and M. Sheydaei, J. Taiwan Inst. Chem. Eng., 2016, 58, 366–373.
40. X. Wang, W. Yang, Y. Ji, X. Yin, Y. Liu, X. Liu, F. Zhang, B. Chen and N. Yang, RSC Adv., 2016, 6, 26155–26162.
41. A. Khataee, M. Sheydaei, A. Hassani, M. Taseidifar and S. Karaca, Ultrason. Sonochem., 2015, 22, 404–411.
42. M. D. Donohue and G. L. Aranovich, Adv. Colloid Interface Sci., 1998, 76–77, 137–152.
43. P. Zheng, B. Bai, W. Guan, H. Wang and Y. Suo, RSC Adv., 2016, 6, 4101–4107.
44. A. R. Khataee and S. G. Pakdehi, J. Taiwan Inst. Chem. Eng., 2014, 45, 2664–2672.
45. C. Zheng, X. Cheng, P. Chen, C. Yang, S. Bao, J. Xia, M. Guo and X. Sun, RSC Adv., 2015, 5, 40872–40883.
46. F. J. Gotor, M. Achimovicova, C. Real and P. Balaz, Powder Technol., 2013, 233, 1–7.
47. D. Chen, S. Ni and Z. Chen, China Particuol., 2007, 5, 357–358.
48. Z. Weiwei, T. Kaibin, Z. Suyuan and Q. Yunxia, Nanotechnology, 2008, 19, 065602.
49. M. C. Morris, S. United, S. National Bureau of, J. C.-I. C. f. D. Data and L. National Measurement, Standard X-ray diffraction powder patterns. Section 18. Section 18, U.S. Dept. of Commerce, National Bureau of Standards, Washington, D.C., 1981.
50. W. Cai and J. Wan, J. Colloid Interface Sci., 2007, 305, 366–370.
51. S. Fathinia, M. Fathinia, A. A. Rahmani and A. Khataee, Appl. Surf. Sci., 2015, 327, 190–200.
52. S. Aber, A. Khataee and M. Sheydaei, Bioresour. Technol., 2009, 100, 6586–6591.
53. S. Karthikeyan, C. J. Magthalin, A. B. Mandal and G. Sekaran, RSC Adv., 2014, 4, 19183–19195.
54. M. Ma, Y. Zhang, W. Yu, H.-y. Shen, H.-q. Zhang and N. Gu, Colloids Surf., A, 2003, 212, 219–226.
55. M. A. G. Soler, G. B. Alcantara, F. Q. Soares, W. R. Viali, P. P. C. Sartoratto, J. R. L. Fernandez, S. W. da Silva, V. K. Garg, A. C. Oliveira and P. C. Morais, Surf. Sci., 2007, 601, 3921–3925.
56. S. Akyuz, T. Akyuz, S. Basaran, C. Bolcal and A. Gulec, Vib. Spectrosc., 2008, 48, 276–280.
57. Z. Jing and S. Wu, Mater. Lett., 2004, 58, 3637–3640.
58. G. Sekaran, S. Karthikeyan, C. Evvie, R. Boopathy and P. Maharaja, Clean Technol. Environ. Policy, 2013, 15, 245–253.
59. S. Navalon, M. Alvaro and H. Garcia, Appl. Catal., B, 2010, 99, 1–26.
60. A. Khataee, P. Gholami and B. Vahid, Ultrason. Sonochem., 2016, 29, 213–225.
61. A. Babuponnusami and K. Muthukumar, Environ. Sci. Pollut. Res., 2013, 20, 1596–1605.
62. P. V. Nidheesh, R. Gandhimathi and N. S. Sanjini, Sep. Purif. Technol., 2014, 132, 568–576.
63. A. Babuponnusami and K. Muthukumar, J. Environ. Chem. Eng., 2014, 2, 557–572.
64. Z. Xu, J. Liang and L. Zhou, J. Alloys Compd., 2013, 546, 112–118.
65. X. Zhang, Y. Ding, H. Tang, X. Han, L. Zhu and N. Wang, Chem. Eng. J., 2014, 236, 251–262.
66. Y. Kuang, Q. Wang, Z. Chen, M. Megharaj and R. Naidu, J. Colloid Interface Sci., 2013, 410, 67–73.
67. S.-P. Sun and A. T. Lemley, J. Mol. Catal. A: Chem., 2011, 349, 71–79.
68. J. Zhang, X. Zhang and Y. Wang, RSC Adv., 2016, 6, 13168–13176.
69. H. Lee and M. Shoda, J. Hazard. Mater., 2008, 153, 1314–1319.
70. L. Xu and J. Wang, Appl. Catal., B, 2012, 123–124, 117–126.
71. S. Bae, D. Kim and W. Lee, Appl. Catal., B, 2013, 134–135, 93–102.
72. S. Tunç, T. Gürkan and O. Duman, Chem. Eng. J., 2012, 181–182, 431–442.
73. Y. Su, Z. Wu, Y. Wu, J. Yu, L. Sun and C. Lin, J. Mater. Chem. A, 2015, 3, 8537–8544.
74. W. Pan, G. Zhang, T. Zheng and P. Wang, RSC Adv., 2015, 5, 27043–27051.
75. M. Neam u, C. Zaharia, C. Catrinescu, A. Yediler, M. Macoveanu and A. Kettrup, Appl. Catal., B, 2004, 48, 287–294.
76. B. G. Kwon, D. S. Lee, N. Kang and J. Yoon, Water Res., 1999, 33, 2110–2118.
77. P. V. Nidheesh and R. Gandhimathi, Desalin. Water Treat.Desalin. Water Treat., 2014, 52, 1872–1877.
78. A. Khataee, R. D. C. Soltani, A. Karimi and S. W. Joo, Ultrason. Sonochem., 2015, 23, 219–230.
79. B. G. Kwon, D. S. Lee, N. Kang and J. Yoon, Water Res., 1999, 33, 2110–2118.
80. M. C. Ortega-Liébana, E. Sánchez-López, J. Hidalgo-Carrillo, A. Marinas, J. M. Marinas and F. J. Urbano, Appl. Catal., B, 2012, 127, 316–322.
81. R. D. C. Soltani, A. Rezaee, H. Godini, A. R. Khataee and A. Hasanbeiki, Chem. Ecol., 2013, 29, 72–85.
82. M. B. Yim, P. B. Chock and E. R. Stadtman, Proc. Natl. Acad. Sci. U. S. A., 1990, 87, 5006–5010.
83. Y. Liu, J. C. Crittenden, L. Wang and P. Liu, J. Hazard. Mater., 2016, 308, 84–90.
84. A. R. Ravishankara and D. D. Davis, J. Phys. Chem., 1978, 82, 2852–2853.
85. S. Khan, X. He, J. A. Khan, H. M. Khan, D. L. Boccelli and D. D. Dionysiou, Chem. Eng. J., 2016 DOI:10.1016/j.cej.2016.05.150.
86. C.-Y. Chen, Water, Air, Soil Pollut., 2009, 202, 335–342.
87. A. R. Khataee, M. Safarpour, M. Zarei and S. Aber, J. Mol. Catal. A: Chem., 2012, 363–364, 58–68.
88. M. Park, T. Anumol and S. A. Snyder, Environ. Sci.: Water Res. Technol., 2015, 1, 699–708.
89. D. Salari, A. Niaei, A. Khataee and M. Zarei, J. Electroanal. Chem., 2009, 629, 117–125.
90. A. R. Khataee, A. Movafeghi, S. Torbati, S. Y. Salehi Lisar and M. Zarei, Ecotoxicol. Environ. Saf., 2012, 80, 291–298.
91. A. Bos, M. Bos and W. E. van der Linden, Anal. Chim. Acta, 1993, 277, 289–295.
92. G. R. Shetty, H. Malki and S. Chellam, J. Membr. Sci., 2003, 212, 99–112.
93. A. Akbarpour, A. Khataee, M. Fathinia and B. Vahid, Electrochim. Acta, 2016, 187, 300–311.
94. M. Zarei, A. Niaei, D. Salari and A. R. Khataee, J. Electroanal. Chem., 2010, 639, 167–174.
95. K. V. Shetty, S. Nandennavar and G. Srinikethan, J. Chem. Technol. Biotechnol., 2008, 83, 1181–1189.

---