Introduction

Furfural, as an excellent organic solvent, is widely used in the oil and petroleum refining, pulp and paper, pharmaceutical, petrochemical, and food industries (He et al. 2014). Therefore, the effluent of these industries contains various concentrations of furfural. For instance, the furfural concentration in the Pars oil refinery wastewater is reported as 100–2000 ppm (Jafarinejad 2017). Also, the furfural concentration in raw wastewater of furfural manufacturing industry can reach to at least 600 mg/L (Wirtz and Dague 1993). Furfural, known as a toxic and non-biodegradable substance in aquatic ecosystems, causes health concerns for human life (Borghei and Hosseini 2008). For example, if furfural is accidentally ingested, it will cause death, as well as its dermal absorption will cause acute damage to the nervous system and the lungs (Cuevas et al. 2014). Liver enlargement, the feeling of weakness, skin rash, tremor, and nose bleeding are also the long-term (chronic) effects of exposure to furfural (Nezamzadeh-Ejhieh and Moeinirad 2011). Today, a variety of different physicochemical methods, including adsorption (Sahu et al. 2008; Singh et al. 2009), photocatalytic oxidation (Faramarzpour et al. 2009), membrane technology (Wang et al. 2018a), and biological treatment (Zhang et al. 2013), have been investigated for furfural removal. However, due to high temperature (98 °C) of the raw wastewater of furfural manufacturing industry (Wirtz and Dague 1993), commercialization of conventional methods will not be easy (Veisi et al. 2016). Generally, the biological process is a time-consuming and costly approach (He et al. 2014), and activated carbon as the most common adsorbent only transfers the pollutant from the liquid to the solid phase and its regeneration cost is very high (Purkait et al. 2007). Recently, sulfate radical-based advanced oxidation processes (\( {\mathrm{SO}}_4^{\bullet -} \)-AOPs) have been introduced as a promising method for degradation of resistant organic contaminants (Hu and Long 2016).\( {\mathrm{SO}}_4^{\bullet -} \) radicals can be generated through activation process of the persulfate anion (\( {\mathrm{S}}_2{\mathrm{O}}_8^{2-} \)-PS) (2.01 V). The homogeneous transition metal ions (Me2+), US radiation (Seid-Mohammadi et al. 2017), UV (Lin et al. 2011), heat (Frontistis et al. 2017), and electrochemical process (Rahmani et al. 2016; Song et al. 2018) are often used for this purpose. Among these methods, the use of heat-activated PS or thermally activated persulfate (TAP) oxidation process has attracted considerable attention as an attractive and clean method for the degradation of numerous organic contaminants (Nie et al. 2014).

According to the studies, \( {\mathrm{SO}}_4^{\bullet -} \)-AOPs have more advantages compared to OH●-AOPs, including (i) \( {\mathrm{SO}}_4^{\bullet -} \) (2.5–3.1 V) in contrast to OH● (acidic pH = 2.4–2.7 V, alkaline pH = 1.9–2.0 V) has a higher redox potential (Ferkous et al. 2017), (ii) the half-life period of \( {\mathrm{SO}}_4^{\bullet -} \) (3–4 × 10 s) is generally more than OH● (2 × 10 s) (Olmez-Hanci and Arslan-Alaton 2013), (iii) the oxidation potential of the OH● decreases significantly at alkaline pH, whereas \( {\mathrm{SO}}_4^{\bullet -} \) can oxidize the pollutants effectively over a wide pH range (Hu and Long 2016), and (iv) \( {\mathrm{SO}}_4^{\bullet -} \) is more selective than OH● for degradation of contaminants (Deng et al. 2013). The use of binary or ternary PS activation techniques due to synergistic effect can reduce the energy consumption and increase the process efficiency (Chakma et al. 2017). In this regard, nZVI can be a good permanent source of Fe2+ to continuously release iron for activation of PS (Dong et al. 2017). Eq. (1) shows the TAP oxidation process (Frontistis et al. 2017; Zarei et al. 2015), Eqs. (2) and (3) describe the reaction between the PS and nZVI particles (Wang et al. 2014), and Eq. (4) demonstrate the activation of PS by Fe2+ (Rahmani et al. 2017).

$$ {\mathrm{S}}_2{\mathrm{O}}_8^{2-}\overset{\mathrm{heat}}{\to }2{\mathrm{S}\mathrm{O}}_4^{\cdotp -}\kern0.5em \left({30}^{{}^{\circ}}\mathrm{C}<\mathrm{T}<{99}^{{}^{\circ}}\mathrm{C}\right) $$
(1)
$$ {\mathrm{Fe}}^0+{\mathrm{H}}^{+}\to {\mathrm{Fe}}^{2+}+{\mathrm{H}}_2\kern0.5em \left(\mathrm{Acidic}\ \mathrm{conditions}\right) $$
(2)
$$ {\mathrm{Fe}}^0+{\mathrm{S}}_2{\mathrm{O}}_8^{2-}\to {\mathrm{Fe}}^{2+}+2{\mathrm{S}\mathrm{O}}_4^{2-} $$
(3)
$$ {\mathrm{Fe}}^{2+}+{\mathrm{S}}_2{\mathrm{O}}_8^{2-}\to {\mathrm{Fe}}^{3+}+{\mathrm{S}\mathrm{O}}_4^{\bullet -}+{\mathrm{S}\mathrm{O}}_4^{2-} $$
(4)

Based on the literatures review and authors’ knowledge, no study has been done to optimize and identify the synergy effect of thermochemical degradation of furfural. Therefore, in this study, \( {\mathrm{SO}}_4^{\bullet -} \) radicals were generated using TAP/nZVI (binary PS activation) process, then the most effective operational parameters were optimized by RSM based on the CCD. After optimizing the process, the synergistic effect among PS, nZVI, and heat was investigated.

Materials and methods

Chemicals

Furfural (99%), sodium persulfate (≥ 99%), ferrous heptahydrate sulfate (≥ 98%), and sodium borohydride (≥ 98%) were procured from Sigma Company. Sodium hydroxide pellet (NaOH), sulfuric acid (98%), starch, and ethanol were procured from Merck company. A number of the most important physicochemical characteristics of furfural are given in Table 1.

Table 1 Physicochemical characteristics of furfural
Full size table

Experimental setup and procedure

The experiments were carried out in a batch rectangular cube Plexiglass reactor with the working volume of 4500 mL. In order to prevent evaporation of furfural from the samples, the reactor was equipped to a reflux condenser. Adjustment of the solution temperature at desired levels was conducted by a water heater element (Jupiter, 2000 W–220 V, number and length of legs were 2 and 30 cm, respectively) equipped with a thermostat. An air pump (7 W–220 V, air flow and pressure were 3000 mL/min, and 0.015 MPa, respectively) was used to ensure sufficient mixing and uniformity of the reactor temperature. The schematic of the experimental setup is shown in Fig. 1. About 0.44 mL of furfural solution was dissolved in 1000 mL of deionized water to prepare the stock solution of 500 mg/L concentration, then other studied concentrations were provided by dilution of the stock solution. After reaching to the intended temperature, a known amount of nZVI and PS was introduced into the sample solution, based on the experimental design. Then pH of the resulting solution was adjusted at the designed values using H2SO4 and NaOH (pH meter; SensION, HACH). At the end of 60-min thermochemical reaction, 10 mL of the solution was taken and the furfural concentration was measured at the maximum wavelength (λmax = 275 nm) using UV–Vis spectrophotometer (DR5000, HACH) (Lucas et al. 2004; Sahu et al. 2007). The mineralization of furfural was determined by vario TOC cube (Elementar Analysensysteme GmbH, Germany) TOC analyzer. The nZVI particles were prepared by sodium borohydride (NaBH4) reduction method (Babuponnusami and Muthukumar 2012) and characterized using transmission electron microscopy (TEM) analysis (EM208S, Philips).

Fig. 1
figure 1

The schematic of TAP reactor

Full size image

Experimental design

Different studies have been done to optimize and investigate the effect of independent variables and their interactions using RSM based on CCD (Cho and Zoh 2007; Hazime et al. 2013; Zarei et al. 2015). In this study, the RSM-CCD was applied to optimize of thermochemical degradation of furfural. Therefore, the initial pH of the solution, PS and furfural concentration, nZVI dosage, and solution temperature were investigated as independent variables at five levels (− α, − 1, 0, + 1, + α), while furfural removal efficiency was considered as response factor. The five independent variables and their coded levels are shown in Table 2. The ranges of variables were chosen based on pre-test and literatures review. Finally, a total of 50 experiment runs were designed based on CCD, as illustrated in Table 3. In this study, the number of independent variables is five (i = 5), so the CCD consists of 32 factorial runs, 10 axial runs, and 8 replicates at the central points (C), as calculated from Eq. (5) (Anupam et al. 2011; Cho and Zoh 2007; Eslami et al. 2016).

$$ \mathrm{Number}\ \mathrm{of}\ \mathrm{runs}={2}^{\mathrm{i}}+2\mathrm{i}+\mathrm{c}={2}^5+\left(2\times 5\right)+8=50 $$
(5)
Table 2 The five independent variables and their levels (α = 2)
Full size table
Table 3 Experimental design and the value of actual and predicted furfural removal efficiency
Full size table

The experimental design and data analysis were conducted by the statistical software of design expert, version 11.0.0. The validation of the predicted model, as well as the relationship between independent variables and response factor, was analyzed through the analysis of variance (ANOVA). Also, the interaction of independent variables was investigated using 3D plots.

Synergy index calculation

In optimal condition, the synergetic effect among PS, nZVI, and heat was calculated using the pseudo-first-order degradation rate constants (Kobs), according to Eq. (6) (Durán et al. 2016). In this equation, SI values smaller or greater than 0 illustrate an antagonistic or synergistic effect between the components, respectively. While SI values equal to 0 demonstrate an additive effect.

$$ \mathrm{SI}=\frac{K_{\mathrm{T}}-\left({K}_1+{K}_2+\cdots +{K}_{\mathrm{n}}\right)}{K_{\mathrm{T}}} $$
(6)

where KT is the pseudo-first-order degradation rate constant of hybrid processes (TAP/nZVI, TAP, and PS/nZVI) and K1, K2, and Kn are the constants of separate processes (PS, heat, and nZVI).

Results and discussion

TEM analysis

Figure 2 depicted the TEM image of the synthesized nZVI particles. As can be seen, the synthesized particles are at the nano-scale with the average size about 20–100 nm (Yehia et al. 2015). According to the result of TEM, the nZVI particles were almost in a smooth spherical shape with a core-shell structure (Liu et al. 2015), so that the Fe0 nanoparticles were encapsulated by the iron oxide shell (Vilardi et al. 2018).

Fig. 2
figure 2

TEM image of synthesized nZVI

Full size image

The experiments design and data analysis

The results of the actual and predicted furfural removal efficiency are shown in Table 3. According to the fit summary analysis, the quadratic model was suggested as the most robust model to predict the response. The validity evaluation of the proposed model to predict the response factor was performed by analysis of variance (ANOVA). The results of ANOVA are presented in Table 4.

Table 4 ANOVA results of the reduced quadratic model for the furfural removal efficiency
Full size table

The values of P value lees than 0.0500 indicate that the model terms are significant, while the terms with the greater values of that were removed from the model. Accordingly, in this study A, B, C, D, E, CE, A2, B2 were the significant model terms. The reduced quadratic model to predict the furfural removal efficiency based on the actual factors has been illustrated in Eq. (7).

$$ \mathrm{X}=-93.21+3.58\;\mathrm{A}+1.93\;\mathrm{B}+0.37\;\mathrm{C}+2.77\;\mathrm{D}+2.42\;\mathrm{E}-0.0091\;\mathrm{C}\mathrm{E}-0.327{A}^2-0.05{B}^2 $$
(7)

In this equation, X is the predicted furfural removal efficiency as well as A, B, C, D, and E are independent variables, as defined in Table 2. As well as the Pareto chart was used to examine the contribution of each independent variable in the removal of furfural (Abdessalem et al. 2008). According to Eq. (8), the Pareto effect (Pi) was calculated based on the coded factors as displayed in the Fig. 3.

$$ {P}_{\mathrm{i}}\ \left(\%\right)=\left[\frac{\left({\mathrm{b}}_{\mathrm{i}}^2\right)}{\sum {\mathrm{b}}_{\mathrm{i}}^2}\right]\times 100\kern0.5em \left(\mathrm{i}\ne 0\right) $$
(8)

where bi represents the regression coefficient of each term in accordance with coded values. As seen from Fig. 3, the contribution of each term is different in the process, where the highest effect in first-order terms has been assigned to the heat. Also, the blue and red colors show the positive and negative effects of independent variables on the furfural removal efficiency, respectively (Moghaddam et al. 2010).

Fig. 3
figure 3

Pareto chart (red, negative effect; blue, positive effect)

Full size image

According to the ANOVA results, the model F value of 199.83 and P > F value less than 0.0001 indicate that the proposed quadratic model is significant (Priya et al. 2018). For the Lack of fit, the F value was obtained to be 2.54, which means that it is not significant relative to the experimental error, so the model is fitted as well (Moradi et al. 2016). The validity and accuracy of the obtained quadratic model can be guaranteed by a suitable R2 value. In this way, the closer coefficients to 1 indicate more favorable conditions (Ghafari et al. 2009). Also, if R2 and adjusted R2\( \left({R}_{\mathrm{adj}}^2\right) \)are close together, it means that there are no unnecessary variables in the model (Asfaram et al. 2016). According to Table 4, in this study, the values of R2 and \( {R}_{\mathrm{adj}}^2 \) are 0.9750 and 0.9701, respectively. The Adeq Precision term is defined as the signal to noise ratio and the value greater than 4 will be desirable for the model. In this study, the signal to noise ratio is 56.417, so the signal of the model is adequate (Sharma et al. 2017). The ratio of standard error of estimate to the mean value of observed response defines as coefficient of variance (CV). This coefficient shows the repeatability power of the proposed model. In general, if the CV is less than 10%, the model is repeatable (Ghafari et al. 2009). Considering the value of CV (4.27%), it can be concluded that the proposed model has a desirable repeatability.

To determine the adequacy of the model, diagnostic plots were used. The difference between the experimental values from the predicted values by the model is called the residual. Accordingly, for diagnosis of data normality, the normal probability plot of the studentized residuals was used.

As shown in Fig. 4 a, all the experimental values are located approximately on the central line, indicating that the data is normalized in the model prediction (Cho and Zoh 2007; Eslami et al. 2016).

Fig. 4
figure 4

Normal plot of residuals (a), predicted vs actual plots (b)

Full size image

Figure 4 b shows the actual values versus the predicted response by model. It is observed that the experimental results were in excellent correlation with the values predicted, which indicates the suitability of the model (Moradi and Ghanbari 2014).

The effect of independent variables on furfural removal efficiency

The effect of PS concentration

Based on the obtained quadratic model, the effect of independent variables as well as their interaction on the prediction of furfural removal efficiency was investigated using 3D response surface plots, as shown in Fig. 5.

Fig. 5
figure 5

3D plots for prediction of furfural removal efficiency (%) at optimal conditions (α = ± 2): The interaction of PS-pH (a), the interaction of heat-furfural (b), and the interaction of heat-nZVI (c)

Full size image

Accordingly, the interaction of PS concentration and pH on furfural removal efficiency is shown in Fig. 5 a. As can be seen, PS concentration variations can affect the removal efficiency in this way that the removal efficiency continuously increased with increasing of PS concentration from 5 to 20 mM. However, with increasing PS concentrations from 20 to 25 mM, the removal efficiency decreased. Increasing the PS concentration as the only source of \( {\mathrm{SO}}_4^{\bullet -} \) radicals (Eq. 1) can enhance the removal efficiency (Liu et al. 2018; Nie et al. 2014). However, it has been reported that with increasing PS from a certain concentration, the removal efficiency will be reduced due to self-scavenging of \( {\mathrm{SO}}_4^{\bullet -} \) radicals according to Eq. (9). This scavenging reaction can also occur between \( {\mathrm{SO}}_4^{\bullet -} \) radicals and high concentrations of PS, as shown in Eq. (10). These undesirable reactions will consume both PS and radicals and subsequently, the removal efficiency will be reduced. In this regard, similar results have been reported in degradation of penicillin G by heat-activated persulfate (Norzaee et al. 2018), and the treatment of dinitrodiazophenol industrial wastewater in heat-activated persulfate system (Wei et al. 2018).

$$ {\mathrm{S}\mathrm{O}}_4^{\bullet -}+{\mathrm{S}\mathrm{O}}_4^{\bullet -}\to {\mathrm{S}}_2{\mathrm{O}}_8^{2-} $$
(9)
$$ {\mathrm{S}}_2{\mathrm{O}}_8^2+{\mathrm{S}\mathrm{O}}_4^{\bullet -}\to {\mathrm{S}}_2{\mathrm{O}}_8^{\bullet -}+{\mathrm{S}\mathrm{O}}_4^{2-} $$
(10)

The effect of pH

As seen in Fig. 5 a, the highest furfural removal efficiency was observed at pH 5–6. In PS-based oxidation process, the pH variations can have a significant effect on the contaminant removal efficiency (Cai et al. 2018). Liang and Su (2009) found that the pH can affect the dominant radical species at heat-activated PS system, and the results of their study are summarized in Fig. 6. According to Eqs. (11) and (12), \( {\mathrm{SO}}_4^{\bullet -} \) in acidic solution can be easily formed through the acid-catalyzation reaction. In this condition, the \( {\mathrm{SO}}_4^{\bullet -} \) has the highest oxidation potential. As a result, furfural removal efficiency will increase (Ji et al. 2016).

Fig. 6
figure 6

The effect of pH on the dominant radical species at TAP oxidation process

Full size image

However, in alkaline solution (pH > 8.5),\( {\mathrm{SO}}_4^{\bullet -} \)and OH− react and produce OH• radical as shown in Eq. (13) (Huang et al. 2002; Nie et al. 2014). Since the oxidation potential and lifetime of the OH• radical decreases significantly under alkaline conditions, the removal efficiency decreases even in the presence of this radical (George et al. 2001; Lau et al. 2007). With further increase in pH, especially pH > 12, the reaction between \( {\mathrm{SO}}_4^{\bullet -} \)and OH• radicals (Eq. 14), as well as scavenging of OH• radicals by OH− (Eq. 15) will lead to radical elimination from the oxidation system (Khan et al. 2017; Rani et al. 2009). For these reasons, the lowest furfural removal efficiency was obtained at pH 11.

$$ {\mathrm{S}}_2{\mathrm{O}}_8^{2-}+{\mathrm{H}}^{+}\to 2{\mathrm{H}\mathrm{S}}_2{\mathrm{O}}_8^{-} $$
(11)
$$ {\mathrm{H}\mathrm{S}}_2{\mathrm{O}}_8^{-}+\to {\mathrm{SO}}_4^{\bullet -}+{\mathrm{SO}}_4^{2-}+{\mathrm{H}}^{+} $$
(12)
$$ {\mathrm{SO}}_4^{\bullet -}+{\mathrm{OH}}^{-}\to {\mathrm{SO}}_4^{2-}+{\mathrm{OH}}^{\bullet } $$
(13)
$$ {\mathrm{SO}}_4^{\bullet -}+{\mathrm{O}\mathrm{H}}^{\bullet}\to {\mathrm{HSO}}_4^{-}+\frac{1}{2}{\mathrm{O}}_2 $$
(14)
$$ {\mathrm{O}\mathrm{H}}^{\bullet }+{\mathrm{O}\mathrm{H}}^{-}\to {\mathrm{O}}^{-}+{\mathrm{H}}_2\mathrm{O} $$
(15)

Ghauch et al. (2012) obtained the maximum ibuprofen removal efficiency at pH 4. Liu et al. (2018) also found that the best degradation of sulfachloropyridazine (SCP) achieved at the acidic condition of reaction solution.

The effect of temperature

Figure 5 b illustrates the effect of variations of heat and furfural concentration on the prediction of removal efficiency. As can be seen, with increasing heat, the removal efficiency is drastically increased. So that, at the optimum concentration of furfural (see Fig. 8), the predicted removal efficiency at temperatures of 40, 50, 60, 70, and 80 °C was obtained 33.5, 52,65, 82, and 100%, respectively. According to Eq. (1), higher temperatures provide more energy to cleavage the O–O bands of PS anions, as a result, the furfural removal efficiency will increase continuously with increasing temperature (Gao et al. 2015). Similar results have been reported in accordance with our findings, for example, degradation of p-nitrophenol in TAP process (Chen et al. 2016), oxidation of indomethacin in the thermo-activated peroxydisulfate system (Li et al. 2018), and oxidation of cefalexin by TAP (Qian et al. 2018).

To calculate the PS activation energy at temperatures of 40–80 °C, the furfural removal efficiency was investigated at the reaction time of 0 to 60 min. The results of this section of the study are shown in Fig. 7. As illustrated in Figs. 7 a and b, the furfural degradation was well fitted with the pseudo-first-order kinetic model. The furfural degradation pseudo-first-order rate constant (kobs) can be calculated from Eq. (16) (Ji et al. 2015).

$$ \ln \left(\frac{{\left[\mathrm{furfural}\right]}_{\mathrm{t}}}{{\left[\mathrm{furfural}\right]}_0}\right)=-{k}_{\mathrm{obs}}\times \mathrm{time} $$
(16)
Fig. 7
figure 7

The effect of TAP temperatures on furfural degradation (a). The pseudo-first-order kinetic reaction equations derived from a (b). The pseudo-first-order rate constants under different temperatures derived from b (c). The Arrhenius plot for Ea estimation (d). (pH = 5.26, [furfural]0 = 84.3 mg/L, nZVI = 1.15 mg/L, PS = 20.52 mM)

Full size image

Figure 7 c shows that with increasing of reaction temperature, the Kobs greatly increased. Therefore, based on the temperature–Kobs dependence, the energy required for PS activation was calculated from the Arrhenius equation (Eqs. 17, 18, and 19) (Zhao et al. 2014).

$$ \ln {K}_{\mathrm{obs}}=\ln A-\left(\frac{{\mathrm{E}}_{\mathrm{a}}}{\mathrm{RT}}\right) $$
(17)
$$ \ln {K}_{\mathrm{obs}}=-\left(\frac{{\mathrm{E}}_{\mathrm{a}}}{R}\right)\left(\frac{1}{\mathrm{T}}\right)+\ln A $$
(18)
$$ {E}_{\mathrm{a}}=-\left(\mathrm{the}\ \mathrm{slope}\ \mathrm{of}\ \mathrm{Arrhenius}\ \mathrm{equation}\right)\times \mathrm{R} $$
(19)

where A is the pre-exponential factor, Ea is the apparent PS activation energy (J mol), R is the ideal gas constant (8.314 J mol K), and T is the reaction temperature (k). Accordingly, based on the obtained equation from Fig. 7 d (Eq. 20), the Ea was determined to be 40.84 kJ mol (Eq. 21), which is much less than the values reported in previous studies (Ji et al. 2015; Liu et al. 2018; Norzaee et al. 2018).

$$ \ln \left({K}_{\mathrm{obs}}\right)=-\left(-4911\bullet 7\right)\left(\frac{1}{T}\right)+10\bullet 832\kern0.75em \left(R=0.9837\right) $$
(20)
$$ {\mathrm{E}}_a=-\left(-4911\cdot 7\right)\times 8\cdot 314=40835\cdot 87\mathrm{J}\cdot {\mathrm{mol}}^{-1} $$
(21)

The effect of furfural concentration

The effect of different furfural concentrations on the TAP/nZVI oxidation process efficiency was also shown in Fig. 5b. As the results show, at 70 °C, by increasing the furfural concentration from 25 to 125 mg/L, the predicted furfural removal efficiency decreases from 100 to 71.5%. At a constant PS concentration (20.52 mM), the increase of furfural concentrations led to the decrease of \( \frac{\left[{\mathrm{SO}}_4^{\bullet -}\right]}{\left[\mathrm{furfural}\right]} \) ratios (Tan et al. 2015). Also by increasing furfural concentration, degradation of byproducts derived from that increases during the process. These intermediate compounds can consume both \( {\mathrm{SO}}_4^{\bullet -} \) and OH• radicals. Therefore, the \( \frac{\left[{\mathrm{SO}}_4^{\bullet -}\right]}{\left[\mathrm{furfural}\right]} \) ratio will decrease further and as a result, the process efficiency will be reduced (Wang et al. 2018b).

The effect of nZVI dosage

Figure 5 c illustrates the influence exerted by nZVI concentration on furfural removal efficiency. At a constant temperature of 40 °C, it is quite clear that as nZVI increased from 0.5 to 2.5 mg/L, the predicted furfural removal efficiency increase from 34.3 to 40%. According to Eqs. (2), (3), and (4), by increasing the nZVI dosage, the number of active sites for Fe release will increase. Consequently, more \( {\mathrm{SO}}_4^{\bullet -} \)radicals will be generated, improving the removal efficiency (Gao et al. 2018; Hussain et al. 2012; Oh et al. 2009).

Process optimization

In order to obtain the highest furfural removal efficiency, the effective operating parameters must be optimized. For this purpose, optimization was performed based on the proposed model (Eq. 7). In this way, the desired goal for independent variables was adjusted at “in range” mode with the importance of 3, while furfural removal efficiency as the response factor was adjusted at the “maximize” mode (equal to 100%) with the importance of 5. Finally, one solution (desirability = 1.000) was selected as the optimal condition. The output of optimal conditions for the TAP/nZVI process is shown in Fig. 8. To evaluate the accuracy of the model prediction, a supernumerary experiment was performed at three replications in the optimized values. Subsequently, the experimental removal efficiency was obtained to be 98.4%, which was nearly consistent with the model’s predicted response (100%). Therefore, the accuracy of the proposed model can be proved.

Fig. 8
figure 8

The optimal conditions for the furfural removal in the TAP/nZVI process (α = ± 2)

Full size image

The identification of synergistic effect

In order to obtain the furfural removal efficiency under different reaction system as a function of time, the performance of non-activated PS, nZVI, heat, PS/nZVI, PS/heat (TAP), and TAP/nZVI systems were investigated at optimal condition. Figure 9 a shows the removal efficiency of furfural under different systems. As can be seen, after 60 min of reaction, the furfural removal efficiency for PS and nZVI alone was between 2 and 3%, as well as sole heat had a negligible removal efficiency. In such circumstances, the removal efficiency for PS/nZVI, PS/Heat, and PS/Heat/nZVI systems was observed to be about 9, 94.5, and 98.4%, respectively. A number of similar results have been reported, which confirm our results (Cai et al. 2018; Zhou et al. 2018). The synergism among the PS, heat, and nZVI was determined using the pseudo-first-order degradation rate constants (K). Figure 9 b shows the coefficient (K) values for combined and separate processes. Based on Eq. (6), the synergy index (SI) was calculated by coefficient (K) values. As it is seen from Fig. 9 c, the SI values higher than zero indicate a synergistic effect between PS, heat, and nZVI.

Fig. 9
figure 9

Removal efficiency of furfural under different systems (a). The pseudo-first-order rate constants under different reaction system (b). The SI for hybrid processes (pH = 5.26, PS = 20.52 mM, furfural = 84.32 mg/L, nZVI = 1.15 mg/L, heat = 79 °C)

Full size image

Furfural mineralization during TAP/nZVI oxidation process

Under the optimum conditions to confirm the furfural degradation, one sample was evaluated by monitoring the TOC removal rate, as shown in Fig. 10. The results showed that the TOC removal efficiency increased with reaction time increasing. So that, after 30- and 120-min operating time of the TAP/nZVI oxidation process, the TOC removal efficiency was observed to be 12.5 and 86.4%, respectively. These results showed furfural was degraded into its intermediate components that are not CO2 and H2O (Veisi et al. 2016). It is worth mentioning the decrease of TOC removal rate within 90 to 120 min indicates that the produced intermediates are resistant to degradation at present condition (Nezamzadeh-Ejhieh and Moeinirad 2011).

Fig. 10
figure 10

Furfural mineralization during TAP/nZVI oxidation process. (pH = 5.26, PS = 20.52 mM, pH = 4.84, furfural = 84.32 mg/L, nZVI = 1.15 mg/L, heat = 79 °C)

Full size image

Conclusion

In this study, thermochemical degradation of furfural was investigated using TAP/nZVI oxidation process, and the five independent variables were optimized by RSM based on CCD. The validity and accuracy of the obtained quadratic model was confirmed by a high R2 coefficient (R2 = 0.975). The Pareto analysis confirmed that the heat was the most effective operational parameter with 89.2% effect. The furfural removal efficiency at optimum conditions for the TAP/nZVI, TAP, PS/nZVI, non-activated PS, and nZVI was obtained as 98.4, 94.5, 9, 3, and 2%, respectively, which confirms the synergistic effect for the combined process. This study illustrated that the TAP process is a promising method for the removal of furfural in aqueous solution, as well as RSM-CCD, can be a beneficial tool to identify the most effective operating parameter and to optimize of the process. However, real applications of this procedure under typical treatment conditions should be further investigated in various aspects, for instance, technical and engineering studies for the feasibility of implementing this process on a practical scale, and further studies to investigate the effect of inhibitors and the application of the process in a complex natural matrix, economic studies to estimate the cost of construction and operation, and ecotoxicological analysis to assess the toxicity of products derived from furfural degradation.