**Evaluating the E**ff**ect of pH, Temperature, and Hydraulic Retention Time on Biological Sulphate Reduction Using Response Surface Methodology**

#### **Mukhethwa Judy Mukwevho 1,2,\*, Dheepak Maharajh <sup>3</sup> and Evans M. Nkhalambayausi Chirwa <sup>1</sup>**


Received: 27 June 2020; Accepted: 21 September 2020; Published: 23 September 2020

**Abstract:** Biological sulphate reduction (BSR) has been identified as a promising alternative for treating acid mine drainage. In this study, the effect of pH, temperature, and hydraulic retention time (HRT) on BSR was investigated. The Box–Behnken design was used to matrix independent variables, namely pH (4–6), temperature (10–30 ◦C), and HRT (2–7 days) with the sulphate reduction efficiency and sulphate reduction rate as response variables. Experiments were conducted in packed bed reactors operating in a downflow mode. Response surface methodology was used to statistically analyse the data and to develop statistical models that can be used to fully understand the individual effects and the interactions between the independent variables. The analysis of variance results showed that the data fitted the quadratic models well as confirmed by a non-significant lack of fit. The temperature and HRT effect were significant (*p* < 0.0001), and these two variables had a strong interaction. However, the influence of pH was insignificant (*p* > 0.05).

**Keywords:** acid mine drainage; sulphate reduction; sulphate reducing bacteria; response surface methodology

#### **1. Introduction**

Acid mine drainage (AMD) is a widespread problem that is considered the most important pollution problem caused by mining industries worldwide. AMD is formed when a sulphide-bearing mineral comes into contact with oxygen and water during or after the closure of mining operations. This oxidation process leads to the formation of sulphuric acid, which further reacts with the sulphide mineral and other exposed minerals and leaches out toxic heavy metals such as lead, cadmium, and arsenic [1–3]. Pyrite is the most common pathway for AMD formation, and its oxidation is shown in Equation (1) [4].

$$4\text{FeS}\_2 + 15\text{O}\_2 + 14\text{H}\_2\text{O} \to 4\text{Fe(OH)}\_3 + 8\text{SO}\_4^{2-} + 16\text{H}^+\tag{1}$$

AMD is characterised by low pH, a high concentration of sulphate, and high concentrations of heavy metals such as iron, manganese, arsenic, zinc, copper, and aluminum. In South Africa, sulphate is considered one of the major contributors to water quality issues for mining operations as it is typically above 2000 mg/L in AMD. As a result, the maximum sulphate discharge levels should be less than 600 mg/L [5].

Conventionally, AMD is treated by neutralisation using lime or calcium carbonate, which precipitate metals and increase the pH but do not effectively reduce sulphate concentration

in the mining effluent to levels suitable for discharge. This method is costly and produces large sludge volumes that are difficult to dispose of. Due to this, more research on AMD treatment has been done over the years, and biological sulphate reduction (BSR) has been identified as a promising alternative treatment for AMD. BSR is a process where sulphate is metabolically converted to sulphide by sulphate-reducing bacteria (SRB), and it simultaneously increases pH and precipitate metals under anaerobic conditions [6,7]. This process requires an electron donor and uses sulphate as a terminal electron acceptor. Electron donors, also known as substrates, that have been used include simple organic compounds such as ethanol, methanol, and butyrate [8–10], and complex organic compounds such as manure, food waste, woodchips, sludge, and hay [11–15]. Simple organic compounds are preferred as they are readily available; however, they are expensive [16]. Most studies have been leaning towards using complex compounds for BSR as they are considered to be cost-effective. SRB oxidises the organic matter, denoted by CH2O in Equation (2), to produce alkalinity and hydrogen sulphide, which binds to the metals and precipitates them as metal sulphides, as shown in Equation (3).

$$2\text{CH}\_2\text{O} + \text{SO}\_4^{2-} \xrightarrow{\text{SRB}} 2\text{HCO}\_3^- + \text{H}\_2\text{S} \tag{2}$$

$$\rm H\_2S + M^{2+} \rightarrow MS + 2H^+ \tag{3}$$

where M represents metals.

The performance of BSR is highly dependent on the availability of organic substrate, pH, temperature, and hydraulic retention time (HRT). Most known SRBs are mesophilic, and they perform optimally at neutral pH. Low pH and high pH suppress and inhibit SRBs, respectively [17–19], whereas low temperatures slow down the metabolic activity of SRBs [20]. Due to the sensitivity of SRB to temperature and pH, most research was done at neutral pH and temperatures greater than 20 ◦C [17,21–23]. HRT affects the rate at which sulphate is removed from AMD. Short retention times are known to washout biomass, whereas long retention times may lead to the depletion of organic matter if complex organic compounds are used [10,11].

Response surface methodology (RSM) is a collection of mathematical and statistical techniques based on the fit of a polynomial equation to the experimental data with the objective of statistically predicting and understanding the system's behaviour [24]. RSM was developed for the simplification of multivariable experimental design enabling the reduction of the number of experiments that are required to identify ideal variables for a process. An advantage of using RSM includes less time required for experimentation due to reduced experimental runs and therefore a cost reduction of materials and reagents [24,25].

In the literature, studies have been focused on increasing the pH of mining influent while simultaneously precipitating dissolved heavy metals [15,26–30]. However, the main focus of this study was primarily on the reduction of sulphate in AMD. Although previous studies have used response surface methodology to investigate how different factors affect biological sulphate reduction [25,31], as far as the authors are aware, there is no published work on how pH, HRT, and temperature affects biological sulphate reduction using response surface methodology. This study forms part of the ongoing BSR work at Mintek, which has a pilot plant running at a coal mine in Mpumalanga, South Africa. This study aimed to investigate the effects of operational factors—namely pH, temperature and HRT—on sulphate reduction efficiency and sulphate reduction rate using response surface methodology. Additionally, the purpose of this study was to develop mathematical models that can be used to predict how the pilot plant would behave when factors are changed within the investigated range. This will help operate the pilot plant such that the sulphate discharge standard is met. The lab-scale reactors were set up to mimic the pilot plant.

#### **2. Materials and Methods**

#### *2.1. Reactor Set-Up*

In the current study, three laboratory-scale water-jacketed reactors were operated in parallel in a downflow mode. The schematic diagram of the lab-scale reactors is shown in Figure 1. The lab-scale packed bed reactors contained the following components: feed and effluent buckets, peristaltic pumps (Watson-Marlow Fluid Technology Group, Johannesburg, South Africa), water-jacketed columns, and a PolyScience Whispercool® (PolyScience, Niles, IL, USA) heater/chiller for temperature control. Each reactor consisted of a base section that functioned as a stand and housed a conical section with an outlet at the bottom of the cone. Above the conical section was a perforated plate with 1 cm holes. The reactors were 1 m in height and 0.15 m in internal diameter, and they had a total working volume of 8 L. A piece of silicone tubing, with an outer and inner diameter of 1 cm and 0.7 cm respectively, was connected at the bottom of the cone. The tube was routed up the column to near the top edge. A T-piece was fitted at the top of the tube to assist in fluid level adjustment. The water jacket of the reactor was connected in closed circuit with a heater/chiller unit with a built-in pump that recirculated water through the water jacket of the column. The feed water for each reactor was stored in 10 L or 25 L plastic buckets, where it was pumped to the top of the column using a variable flowrate Watson Marlow 120 series peristaltic pump. The reactor overflow was collected in a 15 L bucket for each column.

**Figure 1.** Schematic diagram of the reactors.

#### *2.2. Substrates*

Initially, the three lab-scale reactors were packed with 30% woodchips, 30% wood shavings, 20% hay, 10% lucerne straw, and 10% cow manure measured by volume. A mixture of the above-mentioned substrates was blended and loaded into the reactors. Woodchips, wood shavings, hay, and lucerne straw are cellulosic compounds that can contribute as substrates, although their contribution is small [32,33]; hence, they were used as a support for the biofilm. Above the perforated plate, a 2–3 cm layer of woodchips was evenly spread to prevent the holes on the perforated plate

from blocking when the substrates migrated downwards. Cow manure purchased from Lifestyle, Johannesburg, South Africa and lucerne pellets purchased from Milmac Feeds, Fourways, South Africa were used as the main substrates. Then, 186 mL (128.07 g) of cow manure and 186 mL (63.69 g) of lucerne pellets were added on top of the reactor packing and replenished once every week.

#### *2.3. AMD and Inoculum*

The AMD used for all experiments was collected from a coal mine in eMalahleni, Mpumalanga province, South Africa. The raw AMD was characterised by pH less than 3 and a sulphate concentration ranging between 2500 mg/L and 5200 mg/L. The anaerobic mixed culture used was collected from one of the reactors that were operating at the Mintek's pilot plant at the coal mine. The pilot plant had been operating for 10 months at HRT varying between 5 days and 7 days, influent pH > 5 with sulphate reduction efficiency above 90%, sulphide concentration varying between 200 mg/L and 700 mg/L, and it was packed with the same mixture as that used in lab-scale reactors. The lab-scale reactors were inoculated with a mixture of mine water adjusted using hydrated lime to pH approximately 6.5 (70% *v*/*v*) and the inoculum (30% *v*/*v*). For the duration of the study, the flow rate varied between 1.14 L/day and 4 L/day with a sulphate loading rate between 0.36 g/L/day and 2.6 g/L/day. Hydrated lime was used for all pH adjustments.

#### *2.4. Sampling and Analysis*

The effluent pH was measured immediately after the samples were taken. Samples were collected using a beaker; then, the pH meter was immersed into the sample, and the reading was recorded once stabilised. A Metrohm pH sensor (Metrohm, Herisau, Switzerland) was used for pH measurements, and it was calibrated for pH 4 and pH 7 buffer solutions before analysis.

For sulphate analysis, a turbidimetric method was used to measure the influent and effluent sulphate concentration. This was achieved by using a Merck Spectroquant® Prove 300 (Merck, Darmstadt, Germany). All the samples were filtered using 0.22 μm membrane syringe filters before analysis to prevent interferences from suspended solids. Samples were analyzed immediately after collection.

The potentiometric determination of hydrogen sulphide using 0.1 M AgNO3 was used to determine the total sulphide concentration in the effluent. A Metrohm Titrando (Metrohm, Herisau, Switzerland) was used for sulphide titrations using AgNO3.

#### *2.5. Experimental Design*

Design-Expert® (version 11.1.2.0, Stat Ease Inc., Minneapolis, MN, USA), a statistical tool that helps with the design of experiments (DoE), was used to design the experiments. The Box–Behnken model with 3 centre points was used for the design. A total of 16 experiments were conducted. The effect of three factors—namely, pH ranging from 4 to 6 in 1 unit steps, temperature ranging from 10 ◦C to 30 ◦C in steps of 10 ◦C, and HRT ranging from 2 days to 7 days in steps of 2.5 days—was studied. The pH range was selected after it was found in the literature that most SRB perform better near neutral pH, and that at pH less than 4, SRB are suppressed and therefore affecting their performance [17,34]. The temperature range was selected after considering the average temperatures at eMalahleni throughout the year, and HRT was selected based on previous studies done at Mintek. The levels of the chosen variables in the design of experiments are shown in Table 1. The sulphate reduction efficiency and sulphate reduction rate were the corresponding response variables, as shown in Table 2.


**Table 1.** Box–Behnken design for three factors in experimental design.


**Table 2.** Experimental runs and obtained responses.

#### *2.6. Statistical Analysis*

Response surface methodology [24] was used to understand the interactions between the independent variables. This was achieved by fitting the experimental data into a polynomial quadratic equation to obtain regression coefficients, as shown in Equation (4).

$$\mathbf{Y} = \mathbf{b}\_0 + \sum \mathbf{b}\_i \mathbf{x}\_i + \sum \mathbf{b}\_{\text{ii}} \mathbf{x}\_i^2 + \sum \mathbf{b}\_{\text{ii}} \mathbf{x}\_i \mathbf{x}\_{\text{j}} \tag{4}$$

where Y is the response variable, b0 is the constant term, bi is the linear coefficient, bii is the quadratic coefficient, bij is the interaction coefficient, and xi and xj are the values of the coded variables. In this study, the sulphate reduction efficiency (%) and sulphate reduction rate (mol/m3/day) were chosen as response variables and therefore were fitted into Equation (4). Analysis of variance (ANOVA) was used to evaluate the validity and significance of the fitted model. The coefficient of determination R2, adjusted R2, predicted R2, lack of fit, adequate precision, F-value, and *p*-value were used to further evaluate the quality and accuracy of the model. In the present study, the significance level was set at 0.05.

#### **3. Results and Discussion**

#### *3.1. Statistical Analysis*

The data obtained from the 16 experiments that were conducted were fitted into polynomial quadratic equations as shown in Equations (5) and (6) in terms of coded factors.

Sulphate reductionefficiency

$$\begin{array}{l} = +82.86 + 2.65 \times \text{A} + 33.79 \times \text{B} + 21.31 \times \text{C} - 0.7714 \times \text{AB} \\ -0.6874 \times \text{AC} - 5.67 \times \text{BC} + 1.08 \times \text{A}^2 - 23.64 \times \text{B}^2 + 10.51 \times \text{C}^2 \end{array} \tag{5}$$

Sulphate reductionrate

$$\begin{array}{l} = +4.92 + 0.2257 \times \text{A} + 2.11 \times \text{B} \, 1.97 \times \text{C} + 0.05 \times \text{AB} \\ - 0.1838 \times \text{AC} + 0.8878 \times \text{BC} - 0.1347 \times \text{A}^2 - 0.8936 \times \text{B}^2 \\ + 2.63 \times \text{C}^2 \end{array} \tag{6}$$

The reliability, quality, and accuracy of the fitted quadratic models were evaluated using analysis of variance (ANOVA), as shown in Table 3. The significance of the models is confirmed by high F-values and low *p*-values [35]. The models were significant as confirmed by low probability values of less than 0.0001 and high F-values of 101.70 for sulphate reduction efficiency and 221.37 for sulphate reduction rate. The reported F-values imply that there is only a 0.01% chance that the differences could be due to noise. For this study, the lack of fit for both models was insignificant, which shows that the data fitted the models well.


**Table 3.** ANOVA results for the fitted quadratic model.

df—degree of freedom; <sup>1</sup> Significant; <sup>2</sup> Not significant.

Fit statistics are shown in Table 4. The coefficient of determination R<sup>2</sup> is a statistical parameter that measures how well the data fits the line. Adjusted R2 is a version of R2 that is always smaller than R2, and predicted R2 measures the predictive accuracy of the model [25]. A model is considered well fitted when the R<sup>2</sup> value is greater than 0.8 [25]. R2, adjusted R2, and predicted R2 were found to be 0.9935, 0.9837, and 0.9716 for sulphate reduction efficiency and 0.9970, 0.9925, and 0.9853 for sulphate reduction rate. The difference between the predicted and adjusted R<sup>2</sup> should be less than 0.2 for the model to be considered well fitted and able to make satisfactory predictions. For this study, predicted and adjusted R<sup>2</sup> were in agreement with this. Adequate precision measures the signal-to-noise ratio, and a ratio greater than 4 is desirable. The values of adequate precision were 29.36 and 46.21 for sulphate reduction efficiency and sulphate reduction rate, respectively, which indicates an adequate signal.


**Table 4.** Fit statistics.

The diagnostic section provides plots that can be used to further validate the accuracy of the model. The normal probability plots illustrated in Figure 2 show that the residuals are normally distributed as the points are closer to the line. Residuals vs. predicted, as shown in Figure 3, proved the models' quality by having random scatters that are evenly distributed above and below the horizontal axis [25]. The correlation between predicted and actual values is shown in Figure 4. The clustering of the values along the diagonal line confirms that the model is accurate and robust [25,36].

**Figure 2.** Normal plot of residuals for (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate.

**Figure 3.** Plot of residuals vs. predicted for (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate.

**Figure 4.** Correlation between predicted and actual values for (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate.

#### *3.2. E*ff*ect of Individual Factors*

The effect of individual factors on the responses is shown in this section. One factor is changed at a time while keeping other factors at the centre point. The steeper the slope, the more sensitive a response is to the factor.

#### 3.2.1. pH

Figure 5a,b shows the effect of pH on sulphate reduction efficiency and sulphate reduction rate, respectively. From the graphs, both responses slightly increase with an increase in pH from pH 4 to pH 6; however, the increase is not significant. This is evident in Table 3, where pH was insignificant for both responses. This shows that SRBs were not suppressed at an initial pH of 4, which is considered too low for SRB to grow [37]. This could imply that lower costs need to be expended for pH adjustment, and this could have a positive impact on the process's operating expenses. Sulphate reduction at pH approximately 4 was highly impacted in some studies [38–40], which may be because the reactor pH was controlled [39]. However, Jong and Parry [34] found a sulphate reduction efficiency of above 80% when the reactor's influent pH was 4.07. For this study, only the influent pH was controlled, and the average effluent pH, for experiments that had an influent pH of 4, was above 6, as shown in Figure 6. Jong and Parry [34] also observed an effluent pH of above 6. As a result, there are uncertainties about the pH at which sulphate reduction occurs if only the influent pH is controlled, as the effluent pH is always higher [41].

**Figure 5.** pH effect on (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate at 20 ◦C and hydraulic retention time (HRT) of 4.5 days.

**Figure 6.** Average effluent pH at different conditions with an influent pH of 4.

#### 3.2.2. Temperature

The effect of temperature is depicted in Figure 7a,b. Decreasing temperature from 30 to 20 ◦C with HRT and pH at the centre point had a minimal impact on SRB activity, which was expected, as 20 ◦C is in a range that supports the growth and activity of SRB, as also observed in other studies [42–44]. A further decrease in temperature from 20 to 10 ◦C slowed down the metabolic activity significantly. Sheoran et al. [10] suggested that sulphate reduction is likely to decrease by 50% at temperatures lower than 10 ◦C compared to sulphate reduction at 20 ◦C. Similarly, the same effect can be observed from the graphs. The sulphate reduction efficiency and sulphate reduction rate decrease by more than 50% with a decrease in temperature from 20 to 10 ◦C. According to the graphs, the sulphate reduction efficiency decreases from approximately 80% to about 25%, and the sulphate reduction rate decreases from approximately 5 to 2 mol/m3/day when the temperature is decreased from 20 to 10 ◦C with HRT and pH kept constant at the centre point.

**Figure 7.** Temperature effect on (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate at pH 5 and HRT of 4.5 days.

#### 3.2.3. HRT

Figure 8a shows that the sulphate reduction efficiency response increases as the HRT increases from 2 to 7 days. Conversely, the sulphate reduction rate decreases with increasing HRT from 2 to approximately 5 days, followed by a slight increase with a further increase in HRT to 7 days, as shown in Figure 8b. Although sulphate reduction efficiency was observed to decrease with decreasing HRT, the sulphate reduction rate increased with a decrease in HRT due to higher feed rates. A longer HRT resulted in higher sulphate reduction efficiency but lower sulphate reduction rates. Similar observations were made in earlier studies [45–48].

Figure 9a,b shows the effect of HRT over time at 30 ◦C and pH 5. When HRT was decreased from 7 to 4.5 days, there was only a slight decrease in both the sulphate reduction efficiency and sulphate reduction rate. An interesting observation was made when the HRT was decreased from 4.5 days to 2 days. At a HRT of 2 days, both the sulphate reduction efficiency and sulphate reduction rate increased rapidly upon replenishing the substrates. This was followed by a decrease after a maximum was reached. Although some studies show that HRT leads to a decrease in sulphate reduction due to SRB washout [49,50], it is presumed in this study that the decrease in sulphate reduction efficiency and sulphate reduction rate was a result of substrates washout [51]. This was evident when both responses improved upon the replenishing of substrates. A study done by Poinapen et al. [52] using primary sewage sludge (PSS) as a substrate demonstrated that a decrease in HRT did not have an impact on sulphate reduction. This was because the PSS was fed into the reactor together with the synthetic AMD; therefore, a decrease in HRT implied that the PSS loading was increasing. In other words,

the substrate loading was increased with a decrease in HRT, which was not the case in this study. Hence, it is presumed that the substrates were washed out quicker.

**Figure 8.** HRT effect on (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate at 20 ◦C and pH 5.

**Figure 9.** Effect of HRT over time at 30 ◦C and pH 5 for (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate.

#### *3.3. Interactions between Factors*

The interactive effects between pH, temperature, and HRT on sulphate reduction efficiency and sulphate reduction rate are shown in this section. The interaction between temperature and pH illustrated in Figure 10a,b shows that both the sulphate reduction efficiency and sulphate reduction rate decrease with a decrease in temperature. Furthermore, it is clear from the graphs that pH does not affect both responses, especially at low temperatures. At maximum temperature, the responses slightly increase with an increase in pH.

**Figure 10.** Interactive effects between temperature and pH on (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate.

Figure 11a,b shows the interactive effects between HRT and pH on sulphate reduction efficiency and sulphate reduction rate. From the graphs, both responses are not impacted by pH at all HRTs. However, HRT is shown to have different effects on the responses. The sulphate reduction efficiency increases with an increase in HRT, whereas the overall trend for the sulphate reduction rate is that it increases with decreasing HRT. The graphs in Figure 10a,b and Figure 11a,b show that the interactions between pH and temperature (AB) and between pH and HRT (AC) were not strong, which is proven by ANOVA analysis in Table 3.

**Figure 11.** Interactive effects between HRT and pH on (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate.

Strong interactions were observed between HRT and temperature, as depicted in Figure 12a,b. In Figure 12a, the sulphate reduction efficiency increases with a simultaneous increase in temperature and HRT. However, temperature had more impact on the sulphate reduction efficiency compared to HRT. For example, at 30 ◦C, the sulphate reduction efficiency decreased from almost 100% at an HRT of 7 days to just above 60% at an HRT of 2 days. On the other hand, at an HRT of 7 days, it decreased to approximately 40% with a decrease in temperature from 30 to 10 ◦C. Conversely, the increase in sulphate reduction rate caused by a decrease in the HRT was greater at all temperatures than that caused by an increase in temperature at all HRTs, as shown in Figure 12b. The highest sulphate reduction rate was approximately 10 mol/m3/day, which was observed at 30 ◦C and an HRT of 2 days

**Figure 12.** Interactive effects between HRT and temperature on (**a**) sulphate reduction efficiency and (**b**) sulphate reduction rate.

In this section, it was shown that pH had an insignificant interaction with both temperature and HRT. However, HRT and temperature were shown to have strong interactions. The information obtained and the mathematical models developed will be used to evaluate the performance of the pilot plant. For example, due to the difficulty in controlling temperature in an open system, the models developed will be used to determine at what HRT and pH the pilot plant should operate during winter and summer seasons to compensate for drops in sulphate reduction associated with temperature changes. Considering that pH had no effect within the range investigated, it presumed that the only parameter that will be controlled is the HRT. However, further tests will be done on the pilot plant to confirm this.

#### *3.4. Optimisation*

RSM does not only help with understanding the behaviour of systems, but it is also used as a decision-making tool by evaluating the consequences of different scenarios [53]. The numerical optimisation section in design expert allows one to maximise the desirability function. The desirability level varies from 0 to 1, with level 0 indicating that one of the responses is outside the specified limit and a level closer to 1 indicating that the corresponding factor combination is closer to optimal [53]. The optimisation process was carried out with the goal to maximise the sulphate reduction efficiency simultaneously with the sulphate reduction rate by minimising the pH and setting the temperature at 10 ◦C, 20 ◦C, and 30 ◦C. The HRT was set to be in range between 2 days and 7 days. At 10 ◦C, a 41.89% sulphate reduction efficiency and 1.67 mol/m3/day sulphate reduction rate can be achieved at pH 5 and HRT of 7 days. However, the desirability was low at 0.068. This could mean that more retention time, beyond that which was investigated in this study, will be required to achieve a higher sulphate reduction efficiency and sulphate reduction rate. At 20 ◦C, a sulphate reduction efficiency of 92.56% and sulphate reduction rate of 5.06 mol/m3/day can be achieved at pH 4 and HRT 6.7 days, and the desirability was 0.756. A sulphate reduction efficiency of 94.46% and sulphate reduction rate of 5.65 mol/m3/day can be achieved at 30 ◦C, pH 4, and HRT 4.8 days, and the desirability was 0.8. This shows that at higher temperatures, the pilot plant can operate at lower HRTs.

#### *3.5. Sulphide*

Hydrogen sulphide is produced during the reduction of sulphate, as shown in Equation (2), and it is known for its toxicity. Hydrogen sulphide causes problems such as odour, corrosion, and sulphate reduction inhibition [54,55]. The sulphide concentration for all experiments in this study ranged between 114 and 798 mg/L. Due to the high sulphide concentrations observed, there are further tests that are currently being done at Mintek for a downstream process that will use sulphide oxidising bacteria to oxidise sulphide to elemental sulphur (S0), as shown in Equation (7). The oxidation of sulphide to elemental sulphur is a result of incomplete oxidation. The complete oxidation of sulphide results in the formation of sulphate, as shown in Equation (8). Therefore, it is recommended that the ratio of sulphide to oxygen be kept at 2:1 to prevent complete oxidation to sulphate [56].

$$\rm{HS^{-}} + 1/2O\_{2} \rightarrow \rm{S^{0}} + \rm{OH^{-}} \tag{7}$$

$$\rm{HS}^{-} + 2\rm{O}\_{2} \rightarrow \rm{SO}\_{4}^{2-} + \rm{H}^{+} \tag{8}$$

#### **4. Conclusions**

In the current study, RSM was used to statistically analyse the data. ANOVA results showed that the sulphate reduction efficiency and sulphate reduction rate models were significant and adequate, as proven by statistical indexes including lack of fit, coefficient of variation, and adequate precision. Individually, the pH effect was insignificant for both responses, and therefore, its interaction with other independent variables was also not significant. However, there was a strong interaction between HRT and temperature. Additionally, a decrease in HRT impacted the sulphate reduction rate positively due to increased flow rates. Conversely, it had a negative impact on the sulphate reduction efficiency, which was likely due to substrates washout. This study developed mathematical models that were found to be statistically significant. These models can be used as decision-making tools by using them to predict how the process will react to different conditions within the investigated range. This will help adjust controllable factors such as pH and HRT when the temperature fluctuates.

**Author Contributions:** Conceptualisation, E.M.N.C. and D.M.; methodology, M.J.M., E.M.N.C. and D.M.; formal analysis, M.J.M.; investigation, M.J.M.; resources, E.M.N.C. and D.M.; data curation, M.J.M.; writing—Original draft preparation, M.J.M.; writing—Review and editing, E.M.N.C. and D.M.; supervision, E.M.N.C. and D.M.; project administration, M.J.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors gratefully acknowledge Mintek for providing facilities to conduct the research.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **The Fate of Anthropogenic Nanoparticles, nTiO2 and nCeO2, in Waste Water Treatment**

#### **Thomas Lange 1, Petra Schneider 2,\*, Stefan Schymura <sup>3</sup> and Karsten Franke <sup>3</sup>**


Received: 13 July 2020; Accepted: 3 September 2020; Published: 9 September 2020

**Abstract:** Wastewater treatment is one of the main end-of-life scenarios, as well as a possible reentry point into the environment, for anthropogenic nanoparticles (NP). These can be released from consumer products such as sunscreen or antibacterial clothing, from health-related applications or from manufacturing processes such as the use of polishing materials (nCeO2) or paints (nTiO2). The use of NP has dramatically increased over recent years and initial studies have examined the possibility of toxic or environmentally hazardous effects of these particles, as well as their behavior when released. This study focuses on the fate of nTiO2 and nCeO2 during the wastewater treatment process using lab scale wastewater treatment systems to simulate the NP mass flow in the wastewater treatment process. The feasibility of single particle mass spectroscopy (sp-ICP-MS) was tested to determine the NP load. The results show that nTiO2 and nCeO2 are adsorbed to at least 90 percent of the sludge. Furthermore, the results indicate that there are processes during the passage of the treatment system that lead to a modification of the NP shape in the effluent, as NP are observed to be partially smaller in effluent than in the added solution. This observation was made particularly for nCeO2 and might be due to dissolution processes or sedimentation of larger particles during the passage of the treatment system.

**Keywords:** synthetic nanoparticles; nTiO2 and nCeO2; waste water treatment; sp-ICP-MS nanoparticle tracking

#### **1. Introduction**

Recent global challenges, such as climate change adaptation and the transition to green energy, can only be overcome using innovative new technologies such as nanotechnology. At the same time, research on risks of these new materials for humans and the environment must be promoted. The advancing development of synthetic nanoparticles (NP) continuously produces new types of materials. NP are materials in which 50 percent or more of the particles have one or more dimensions between 1 nm and 100 nm, often exhibiting vastly different properties than bulk materials. These NPs can be divided into various groups according to their origin and respective properties. For example, there are carbon NPs (fullerenes, carbon nanotubes), metal NPs (Ag, Au), metal oxide NPs (TiO2, CeO2), and polymeric NPs [1]. Because of their special antibacterial, photocatalytic, mechanical, electronic and biological properties, NPs made of silver (nAg), titanium dioxide (nTiO2), zinc oxide (nZnO) and cerium oxide (nCeO2) are used as substantial components of personal care products, pharmaceuticals, paints, electronic devices, energy storage, coatings and new environmental engineering technologies [2–4]. The widespread use of such NPs leads to their inevitable arrival in domestic sewage treatment plants [5]. It is predicted that wastewater treatment (WWT) serves as an important sink for NPs released from consumer products. Therefore, wastewater treatment plants (WWTPs) are key factors controlling entry paths of NPs into the environment and the food chain. The distribution of NPs between the WWTP effluents, sludge and cleared water, determines the NP mass flow and thus controls the expected dose to the environment and humans. The need for the investigation of NP behavior and properties in sewage sludge lead to various research efforts on the topic. For example, Brar et al. [6] investigated NP behavior in different WWTP process stages and ultimately also in sewage sludge, but at the same time, found that no methodological studies were carried out to determine the NP presence and removal during various WWT processes or their presence in wastewater at all. Furthermore, DiSalvo et al. [7] addressed the knowledge gap on NP behavior in the sewer network and in wastewater treatment. Brar et al. [6] and Yamaguchi [8] gave a general overview of the origin of various NP types and their source products. More recent studies went beyond just looking at the source products and were related to the NM life cycle [5]. Some authors go further and, in addition to the manufacturing mechanism, consider the regeneration and reusability of NM in wastewater treatment [9,10] to exploit their positive effects during treatment [11,12]. However, other authors also describe cumulative and/or combination effects of NP in the wastewater-sewage sludge pathway [13]. The impairment of COD and ammonium degradation by stabilized silver NP in sewage treatment plants was investigated by Hou et al. [14] and Jarvie et al. [15], for example, with silicon NP in waste water. Kim et al. [16,17] and Gartiser et al. [18] considered nTiO2. Limbach et al. [19] and Yang et al. [20] dealt with the degradation of oxidic NP in model sewage treatment plants. Oleszczuk et al. [21] focused on the effects of applied sewage sludge on different types of plants. Soils and sediments act as a NM sink in the environment [22,23], whereby an essential NM entry path consists of the agricultural use of sewage sludge [18,24–28]. Fundamental for the understanding of the interaction of NM are reactivity and bioavailability [29–34]. Various experimental approaches attempt to determine key parameters such as surface charge, degradation, geochemical milieu, interaction with natural colloids, parameters influencing hydrodynamic conditions (e.g., pore size, roughness, flow velocity), and to derive statements on mobility and bioavailability [23,35–45].

Besides the emerging risks of NPs released through WWTP effluents, the effects of NPs on the WWT itself are reason for concern. Most municipal WWTPs depend on an activated sludge process that degrades waste water components with the help of microorganisms [46]. Activated sludge contains microbial eukaryotes, including protozoa, fungi and metazoans, and various types of bacteria that are responsible for metabolic functions, e.g., the oxidation of organic compounds, the removal of nitrogenous pollutants and phosphates. Several authors refer to toxic effects of different NPs on different microorganisms which may pose potential environmental hazards and effect the effectiveness of WWT. It was shown that Ag-NPs penetrate the membrane wall of Escherichia coli and other gram-negative bacteria. In addition, growth experiments with nitrifying bacteria showed a strong inhibition caused by Ag-NPs [47–49]. NPs show toxicity to many species, including bacteria, algae, invertebrates and vertebrates [22,50–55]. While studies have shown that NPs are toxic to single species, the complexity of an activated sludge community might not respond to NPs in the same way as single species systems. Therefore, little is known about the impact of NPs on complex microbial communities that are effective in degrading waste in the activated sludge, or if microorganisms are able to remove them [56]. A sudden increase in the NP concentration in the feed water (shock load) poses a toxicity risk even for the beneficial microorganisms contained in the sludge, and it might take months to recover the performance of the treatment plant [56,57]. Furthermore, the availability of the organic substances to microorganisms can decrease due to competitive adsorption, and thus organic substances might remain untreated by the microorganisms [58,59]. This might lead to a reduced treatment efficiency of the sewage treatment plant, which consequently poses the risk of potentially pathogenic microbes remaining in the treated water.

The ecological risks that are expected with increasing NP use cannot yet be adequately assessed in many areas [60,61]. Although previous research shows that initial fears that NM are an inherent risk to humans and the environment have not been confirmed, long-term, low-dose and interaction effects have not yet been adequately investigated. The low expected environmental NP concentrations ranging from ng/L to μg/L (water) or μg/kg (soils) [26] make the investigation of the consequences of an environmental impact considerably more difficult, because of the challenge of sensitive detection. Furthermore, the complexity of the matrices involved leads to a considerable experimental effort [62], leading, for example, to a considerable uncertainty regarding NP behavior in soils, although soil is one of the main sinks for released NP [63]. A resulting trend towards studies with high concentrations and a lack of characterization can actually lead to a significant reduction in the informative value of this work up to the complete meaninglessness of the results [64]. The main entry path for the control of the entry of synthetic NM in soils is the waste water purification process, which is considered a preliminary NM sink, and at the same time represents their release into the environment via sewage sludge from sewage treatment plants [26]. While the water clarification process is the main entry path for such NM into the environment, the subsequent main exposure path for humans is the potential NP absorption introduced into the soil through their absorption in plants and thus their entry into the food chain [60]. So far, there is no country yet that has comprehensive legislation to deal with NM, particularly in wastewater treatment processes.

Here, we report on our studies concerning the fate nTiO2 and nCeO2, both widely used in consumer products, in WWT. In light of the studies referenced above, the main goal was to investigate the influence of the NP on the effectiveness of the WWT process, to gain valuable information on the distribution of the NP between the WWTP effluents, and to establish single-particle mass spectroscopy (ICP-MS) [65] as a measurement modality to gain a detailed insight in the fate of the NP. Only few studies exist that applied sp-ICP-MS as determination approach for NP in water or wastewater treatment samples, particularly for nCeO2 [66], nTiO2 [67,68], and nAg NP, that are indicators for medical NP residues in wastewater samples [67–70]. Thus, the scope of the present study was to further investigate the feasibility of the sp-ICP-MS for the determination of nCe and nTi in wastewater, and to suggest recommendations for the monitoring of municipal wastewater samples.

#### **2. Materials and Methods**

#### *2.1. NM Measuring Technique*

Conventional ICP-MS can be used to determine element concentrations with very high sensitivity. In general, samples are digested and transformed into a homogenous solution. This solution is injected into an inductively coupled plasma that ionizes the dissolved species which are then analyzed by the mass spectrometry. The uniform distribution of ions in the analyzed solution results in a constant MS signal, which is recorded at specified intervals ("dwell times") at a specified frequency. In contrast to that, in sp-ICP-MS, the solution is diluted to such a degree and the dwell time of the measurements is reduced, such that only a single particle is measured at each step. NP enter the MS individually as a cluster of ions from the plasma. This results in a time-resolved accumulation of detected ions giving rise to peaks, which provide information about the concentration (peak frequency), size (peak height) and the type of nanoparticles (mass-charge ratio). Ions which originated from non-NP-species still reach the detector when these transient signals are recorded and contribute to the background signal. The height of this element characteristic background signal has a significant influence on the detection limit for the nanoparticles. The relevant factor for detecting single particles is the speed: for sp-ICP-MS analysis, continuous data acquisition at a dwell time is substantially smaller than for conventional ICP-MC, and the fundamental instrumental requirement for NP counting and sizing. For the NP measurements in the present investigation, we used an Elan DRCII ICP-MS (PerkinElmer, Waltham, MA, USA). The assessment of the ICP-MS results was done through the Single Particle Calculation tool (SPC) that was developed by the Food Safety Research group from Wageningen, The Netherlands [71,72].

#### *2.2. Experimental Setup*

The investigation workflow is illustrated in Figure 1. Two behrotest® KLD4N/SR (behr Labor-Technik GmbH, Düsseldorf, Germany) two-stage lab-scale wastewater treatment systems (three tanks: denitrification, activation and post-treatment, see Figure 2) were operated according to OECD TG 312 [23]. Both were inoculated with activated sludge from a municipal sewage treatment plant in Chemnitz-Heinersdorf (Chemnitz, Germany) and fed with standardized, synthetic wastewater. Synthetic wastewater was generated according to OECD Test 303A. The system has a total volume of 10.5 L. The operating parameters were set to hydraulic retention time between 4 to 5 h and a sludge age of around 12 days. The oxygen content in the aeration tank was controlled in real time between 1.5 and 2.0 mg/L. The feed was pumped from the feed tank (1) into the denitrification vessel (3) by a peristaltic pump. From there, the sludge reached the activation tank (2) via an overflow, which was aerated by an air pump and a frit at the bottom of the tank (4). The activated sludge reached the clarifier via a settling pipe (5). The sludge recirculation pumped the sludge back into the denitrification vessel. Clarified water flowed into the drain tank via an overflow. Fresh sludge from the Chemnitz sewage treatment plant was used for each inoculation, and background concentrations of the relevant elements, and the media used (activated sludge, wastewater, cleared water) were determined. The mass-charge ratio *m*/*z* = 48 was used for titanium dioxide and *m*/*z* = 139 for cerium dioxide.

**Figure 1.** Investigation strategy for nCeO2 and nTiO2 on the pathway wastewater-sewage sludge.

**Figure 2.** Lab-scale wastewater treatment systems: 1—storage tank for feed water, 2—storage tank for clarified drain water, 3—denitrification tank, 4—activation tank, 5—secondary clarification tank, 6—oxygen meter and control panel oxygen entry control, 7—control panel sludge return, 8—agitators, 9—sludge recirculation, 10—sludge recirculation, 11—volume flow measurement.

The monitoring of the performance included the daily measurement of the pH value in the aeration tank and the acid capacity in the secondary clarifier. With each change of feed, the parameters electrical conductivity, pH, carbon, nitrogen and phosphorus content in the inlet and outlet were determined. The dry matter content in the aeration and secondary clarification was monitored every working day. Two NP experiments were carried out with different concentrations using a single NP addition and one experiment using continuous particle addition, was performed. Cerium dioxide was added at 3 mg or 380 μg once and 3.95 μg/day continuously. Titanium dioxide was added at 50 mg or 63 mg once and 1 mg/day continuously. The sludge and drainage samples were taken on a weekly basis, digested using acid-assisted microwaves, and then measured using ICP-MS on the *m*/*z* 48 for titanium and 139 for cerium. The measurement results were converted into daily loads and summed up. In addition, selected run-off samples were analyzed without digestion using sp-ICP-MS to compare the particle size and number with the initial suspension.

#### *2.3. Operational Setting of the Lab-Scale Wastewater Treatment System*

To establish the optimal operational setting and to enhance the treatment efficiency of the lab-scale wastewater treatment system, the first test runs were performed using the wastewater of the Chemnitz-Heinersdorf municipal WWTP. Chemnitz has approx. 246,000 inhabitants. The sewage treatment plant has a total size of 400,000 population equivalents. The daily wastewater volume averages 71,400 m3/day. The canal system has a length of around 1000 km; more than 600 km are mixed wastewater and rain water, more than 190 km are wastewater and approximately 160 km are rainwater. The connection to the sewage treatment plant is 96.6 percent. Beside residential areas, commercial centers are also connected to the treatment plant. The plant is operated as anaerobic sludge stabilization and the biological cleaning as denitrification with subsequent nitrification. The inlet to the sewage treatment plant is characterized by an average COD content of around 398 mg/L, an average BOD5 of 196 mg/L, and a total Kjeldahl nitrogen of 38 mg/L. In the outlet, there is an average of only 21 mg/L COD. Fresh sludge from the Chemnitz sewage treatment plant was used for the inoculation of the lab-scale wastewater treatment system. On the 9th (system 1) and 3rd trial day (system 2), the performance of both recirculation pumps was increased to 100 percent, in order to improve the denitrification performance and thus raise the pH in the activation vessel. The final operating settings and sludge retention times are shown in Table 1.


**Table 1.** Operational parameters of the lab-scale sewage treatment systems.

The return pumps (7) were operated in batch mode (8 min on, 32 min off). This setting was made together with the increase in performance of the pumps (previously 8 min on, 22 min off). Less sludge with a low pH should be pumped back in order to prevent a further drop in the pH in the aeration. A sludge age of 12 days was targeted for both systems. For this purpose, excess sludge was removed daily via the return pump. In both systems, an oxygen content of approx. 1.5–2 mg/L was set up in the aeration tank. The lower limit of the automatic oxygen control (6) was set to 1.3 mg/L, the upper limit to 1.4 mg/L. The diaphragm pump flow rate was set to approx. 20 L/h (11). The performance of the stirrer motors (8) was reduced to 10 percent in the denitrification and in the activation vessel. With this setting, the targeted carbon removal rate on average of minimum 90 percent for dissolved organic carbon (DOC) and total organic carbon (TOC) was achieved in both lab-scale wastewater treatment systems. The obtained optimized operational setting was used for the further operation of the lab-scale wastewater treatment system with synthetic wastewater, based on the OECD guideline 303.

#### *2.4. Input: Synthetic Wastewater*

Synthetic wastewater based on the OECD guideline 303 served as the inlet into the lab-scale systems. Moreover, 8 g peptone from casein, 5.5 g meat extract, 1.5 g urea, 1.4 g dipotassium hydrogen phosphate, 0.35 g sodium chloride, 2 g CaCl2·2 H2O and 0.1 g MgSO4·7 H2O were used. The components were dissolved in 1 L of ultrapure water (≥18.2 MΩ/cm, Evoqua Water Technologies). This concentrate was divided into two fillings, since 500 mL of concentrate were diluted with 25 L of deionized water in the feed tank. 10 mL of methanol were added to each replenishment. The feed tank was emptied, cleaned and refilled after two days. The dry matter was determined daily from activated sludge and excess sludge from day 1. The sample of the old feed (rest of the two-day-old feed) and the sample of the new feed were withdrawn via the feed hose. The drain sample was withdrawn using a pipette on the water surface of the secondary clarifier. Furthermore, the pH value in the aeration tank was determined every day according to DIN EN 38404-5:2009-07, and the acid capacity in the drain according to DIN 38409-7:2005-12. 50 mL of sample were used to measure the acid capacity. With each change of inlet, the parameters pH value, TOC/DOC/TNb according to DIN EN 1484:1997-08 and total phosphorus according to DIN EN ISO 6878:2004-09 were determined for the inlet and outlet. These series of measurements started with the first change of inflow on the second day.

#### *2.5. NM Investigation Methodology*

Endurance tests were then carried out with nTiO2 and nCeO2 from ready-to-use suspensions (US Research Nanoparticles Inc., Houston, TX, USA, each 30–50 nm). Two types of particle addition were used:

*Single NP addition:* To determine the blank value levels, the inoculum was sampled twice and the system (sludge and drain) 7 times in 4 days and averaged. On the fifth day after test start, 62.7 mg of titanium dioxide (anatase, 15 wt%, particle size 30 to 50 nm; US Research Nanomaterials Inc., Houston, TX, USA) and 380 μg of the cerium dioxide NP (NM-212, particle size 28.4 ± 10.4 nm, Joint Research Center, Institute for Health and Consumer Protection, Ispra, Italy) were added as a pure water suspension in the denitrification basin. The suspensions were ultrasonicated with ice cooling for 30 min and shaken vigorously before the addition in order to destroy agglomerates.

*Continuous NP addition: To* determine the blank value level, the inoculum was sampled twice and the system (sludge and drain) 11 times in 10 days. From day 13 after commissioning the system, approx. 1 mg/day nTiO2 was added to the denitrification basin as a pure water suspension. This was achieved by adding approximately 1.9 mg of nTiO2 suspension to each new feed. The suspension was ultrasonicated with ice cooling before the addition, in order to destroy agglomerates. A sample of the stock solution was taken once a week immediately after the addition to check the concentration. At the same time, two solutions for two future inlets were taken.

#### *2.6. Sampling and Analysis*

To determine the titanium load in the sludge, 10 mL of excess sludge was removed every week via return pump. The sludge was dried in a ceramic dish at 105 ◦C and then acidified with 5 mL of nitric acid (67 percent, suprapur; Merck KGaA, Darmstadt, Germany). This solution was transferred to a microwave vessel and mixed with 5 mL of nitric acid and 2 mL of hydrochloric acid. For sampling, 50 mL of sample were removed from the water surface of the secondary clarifier every week. From the digestion, 10 mL of sample with 3 mL of nitric acid and 2 mL of hydrogen peroxide (30 percent, e.g., Merck KGaA) were subjected to a microwave treatment. The isotope Titan 48 (47.948 amu) was examined. Rhodium (10 μg/L; 102.905 amu) was used as the internal standard. The device settings of the Elan DRC II (PerkinElmer) used can be found in Table 2.

Data evaluation: the measured values were converted into a solid concentration through the dry substance weight. The activated sludge mass was calculated using the dry substance and the total volume of the activated sludge and denitrification vessel. The secondary clarifier sludge mass was

calculated using the dry matter of the excess sludge and the sludge volume of the secondary clarifier. The titanium load was calculated using those figures and the solids concentration. The accumulated masses of NP in aeration, recirculation, clarifier and sludge return resulted in the daily NP load (mg).


**Table 2.** Mass spectroscopy (ICP-MS) settings.

#### **3. Results**

#### *3.1. Operational Monitoring of the Lab-Scale Sewage Treatment Systems*

The structure of the sludge changed from coarse-flaky dark brown to fine-flaky ocher in the course of the experiments. As this change was observed with every system run, it can be assumed that the sludge adapted to the changed composition of the standardized wastewater. Organic carbon degradation was always higher than 90 percent. The removal of nitrogen fluctuated in the range of 40 to 80 percent. Since the experimental set-up did not contain phosphorous removal processes, phosphorus degradation was only possible biochemically.

#### *3.2. Discharge Loads of Titanium NP-Spike Particle Addition*

The series of measurements of the singe particle addition of nTiO2 began five days after the system was commissioned. The mean blank level before addition was 635.8 mg/kg in the sludge and 36.6 μg/L in the aqueous phase. The blank value of the deionized water was 0.6 μg/L. The experiment was terminated after 37 days because the titanium concentration in the excess sludge fell below the blank value level (618.0 mg/kg). In Figure 3, the titanium loads are plotted against the test days. The first vertical line marks the day of the NP addition; the second vertical line marks the day of the undershoot. Until the NP addition, the titanium load in the sludge steadily dropped to 12.9 mg, while the load in the aqueous phase remained almost constant. After the particle addition, the load in the excess sludge increased to a maximum of 71.6 mg on the 7th day of the experiment. An increase in the titanium load in the aqueous phase was also observed, but a maximum of 4.5 mg was reached on the 6th day of the experiment. In the following test days, the titanium load in the excess sludge fell continuously until day 14 and then fluctuated between 40.5 mg and 17.9 mg by the 28th test day. Until the test end, the values decreased continuously to 6.8 mg. After the maximum load in the aqueous phase, the values decreased to 0.4 mg by the 19th day of the experiment, and fluctuated between 0.2 mg and 1.3 mg by the last day of the experiment. The titanium load in the aqueous phase was almost constant. From the first to the 33rd day of the trial, 74.0 mg of titanium left the system via excess sludge and drainage. As expected, the load in the sludge and in the aqueous phase reached its maximum after the addition. The load in the sludge sank as the trial days progressed, as titanium was removed from the system by removing excess sludge. The load in the aqueous phase decreased faster when the water retention time was shorter than that of the sludge. This load consequently left the system faster. The small fluctuations in the aqueous phase are due to an almost homogeneous particle distribution in the water. The NPs are not homogeneously distributed in the sludge.

**Figure 3.** Discharge loads of titanium nanoparticles (NP) after single particle addition on test day 5. (**a**) Titanium load over experiment duration, (**b**) correlation between titanium load in water and sludge.

Compared to Gartiser et al. [18] with R2 = 0.84, there was a lower correlation of R<sup>2</sup> = 0.68. The adsorption was determined with 90 percent in the sludge and 9 to 10 percent in the aqueous phase, compared to 95 percent in the sludge and 3 to 4 percent in the aqueous phase in Gartiser et al. [18]. The reason for the lower correlation were the different test procedures. Gartiser et al. [18] metered constant titanium dioxide into the system (similar to system 2), and the loads increased in the course of the experiment. With the single particle addition, the loads decreased steadily due to washing, which brought the sludge load closer to the water load. Calculating the adsorption ratio from the absolute mass gave a ratio of 96 percent in the sludge and 4 percent in the aqueous phase. In principle, this experiment confirms the fractionation processes observed in the literature.

#### *3.3. Discharge Loads of Titanium NP-Continuous Particle Addition*

The titanium load in the excess sludge and in the aqueous phase are shown in Figure 4, as well as the cumulative NP addition. The vertical line marks the beginning of the continuous NP addition. Before the particle addition, the sludge loads fluctuated between 8.9 mg and 24.8 mg (σ = 5.9 mg), while the loads in the aqueous phase remained almost constant. After the addition, there were minor fluctuations in the sludge, from 7.3 mg to 17.1 mg (σ = 2.4 mg). The constancy in the aqueous phase remained. Up to the last day of the test, approximately 26 mg of nTiO2 was added to the system.

**Figure 4.** Discharge loads of titanium NP after continuous particle addition.

#### *3.4. nTiO2 Characteristics of Added NP Suspension and E*ffl*uent*

Furthermore, we analyzed the nTiO2 scan of the added NP suspension and the effluent. The height of the peaks corresponds to the size of the particles and the frequency to the quantity. Even if a direct comparison is not possible due to different dilutions, it can be concluded that the particles are both smaller and fewer in the effluent. The background level of the ionic non-NP-species value is in the range from 0 to approx. 30,000 cps. Most of the individual peaks end in a range from 100,000 cps to 280,000 cps. Two peaks exceeded this range. This observation is an indication of processes during wastewater treatment that contribute to the partial or complete dissolution of the NP.

#### *3.5. Discharge Loads of Cerium NP-Spike Particle Addition*

Figure 5 shows the cerium load in the sludge and the outlet within the test period. The cerium load of sludge increases at the 5th day of the experiment. A short-term increase in cerium load to over 300 μg was found at day 6 and 7. The loads then fell until day 12. The loads fluctuated between 90 and 220 μg between day 13 and day 27. From trial day 28, the loads dropped continuously to a value of 30 μg. The discharge loads are approximately constant between 1 and 35 μg over the test period. All loads are well below 50 μg. The discharge loads were always lower than the mud loads. The concentration of the sludge blank value is 1.79 mg/kg; the blank value of the drain was 0.15 μg/L.

**Figure 5.** Discharge loads of cerium NP after single particle addition on test day 5. (**a**) Cerium load over experiment duration, (**b**) correlation between cerium load in water and sludge.

The maximum concentration was reached immediately after the NP addition, since at this point, all the particles were in the system and hardly any sludge has been removed. In the following test days, the nCeO2 load in the sludge decreased more and more. This was done by taking daily samples to determine the dry matter of the excess sludge and by withdrawing excess sludge to maintain the sludge age. The significant drop in sludge loads from test day 7 to 12 could be due to the fact that excess sludge was withdrawn on test days 7, 8 and 9. This relationship was also observed from day 20 to 22. The blank level was reached both in the sludge and in the effluent at day 34. It can be concluded that the cerium dioxide NP were removed from the system by daily sampling, the removal of excess sludge and by the water running off. As a result, it was found that 5 percent of the nCeO2 was in the effluent and thus 95 percent was accumulated in the sludge. When comparing the results with other studies carried out, it was found that similar results were achieved.

#### *3.6. Discharge Loads of Cerium NP-Continuous Particle Addition*

Figure 6 shows both the cerium load in the sludge and in the effluent. It is noticeable that the cerium load in sludge fluctuated between 280 and 410 μg by the seventh day of the experiment. Then, there was a drop to 125 μg. No trend can be seen between days 8 and 34; the loads fluctuated between 45 and 200 μg. An increase in cerium load can be observed after day 35 with one exception. The loads in the effluent fluctuated between 1 and 30 μg up to day 17. The discharge loads were almost constant between 0.5 and 7 μg after day 20.

**Figure 6.** Discharge loads of cerium NP after continuous particle addition.

The concentration of the sludge blank value was 8.81 mg/kg, the blank value of the drain is 1.02 μg/L. The sludge loads decreased despite the addition of the nanoparticles by the 35th day of the experiment. The reason for this is that the cerium load that leaves the lab-scale wastewater treatment system via the effluent, by removing excess sludge and taking daily samples, is higher than the 3.95 μg added per day. The fluctuations between test days 0 and 13 are more pronounced than the fluctuations between test days 14 and 35. As a result, it was found that 4 percent of the nCeO2 remained in the effluent and thus 96 percent in the sludge.

#### *3.7. nCeO2 Characteristics of Added NP Suspension and E*ffl*uent, Particle Size Distribution*

The NP characteristic is shown in Figure 7 as the nCeO2 scan of the added NP suspension and the effluent. Figure 7a shows the intensity of the stock solution of the nCeO2 before they are added to the lab-scale sewage treatment systems. The time used is given in seconds and the intensity in cps. The background contributes to the intensity range 0 to 80,000 cps. The majority of the particles are detected between 100,000 and 300,000 cps. Four peaks have an intensity in the range of 350,000 to 600,000 cps. Figure 7b shows the intensity of the nCeO2 in the course of experiment day 33. The time is again given in seconds and the intensity in cps. Signals in the range 0 to 2000 cps are formed by the non-NP-background. The majority of the particles are between 3000 and 10,000 cps. Six peaks are above an intensity of 20,000 cps. The nCeO2 used for the test can be detected in the effluent after addition. As is visible from Figure 7, from left (stock solution) to right (effluent), the particle number concentration decreases (expressed through the decreased number of peaks), and the particle size decreases (expressed through a significantly smaller scale of intensity). This leads to the conclusion that after the addition to the system, the NPs are observed to be significantly smaller than in the stock solution (before the addition of Figure 7). The implications for dissolution processes in the course of wastewater treatment are even clearer than for nTiO2.

**Figure 7.** nCeO2 scan of the added NP stock solution (**a**) and the effluent (**b**).

Since the background level of cerium is negligible in contrast to that of titanium, this comparison points to dissolution processes during the passage through the system leading to an increase in the baseline.

#### **4. Discussion and Conclusions**

Scope of the investigation was to investigate the fate of synthetic NP on the pathway wastewater-sewage sludge exemplified by nTiO2 and nCeO2. Furthermore, the feasibility of sp-ICP-MS as standard operation procedure for NP tracking in wastewater was investigated. Even though there is still room for further development, as was stated by Mozhayeva and Engelhardt [73], the present study confirmed the feasibility of the procedure. In this context, the investigation approach is of fundamental and application-oriented importance for research and industry, since the results provide fundamental knowledge for consulting in wastewater disposal. Particularly in the area of municipal wastewater treatment, the topic is met with great interest. This applies to both information on the operating modes of the wastewater treatment systems that contain NPs and the possibilities of verifying NPs in wastewater. In addition, the results obtained can be used to draw conclusions about expected procedural adjustments to WWTPs.

The investigation results show that both nTiO2 and nCeO2, are adsorbed, to a substantial extent, in the sludge. For both NM, the experiments resulted in an adsorption ratio of at least 90 percent in the sludge. The results for nTiO2 show that the adsorption ratio in the sludge during a single particle addition is even higher; up to 96 percent. Thus, 4 percent of the NP remained in the aqueous phase. In case of continuous particle additions, 90 percent of the added nTiO2 were adsorbed to the sludge, while 9 to 10 percent remained in the aqueous phase. The adsorption process mechanism of the different application procedures is subject to future investigations. For example, it might be assumed that the adsorption efficiency is higher during a single particle application. Thus, the available adsorption capacity might be used more comprehensively. The results also indicate that during a continuous particle addition, the maximum adsorption capacity is reached sooner. For the practical wastewater treatment, it can be concluded that a permanent nTiO2 presence in the wastewater might reduce the NP adsorption capacity over time. In such a case, the remaining nTiO2 could pass the wastewater treatment process. However, for nCeO2 NP, a difference in the adsorption capacity during single or continuous application was not observed. This may be due to the much lower concentration used in the CeO2 experiments. Independently from the type of nCeO2 addition, 4 to 5 percent of the nCeO2 remained mobile in the effluent, while 96 percent were accumulated in the sludge. In case of the application of such sludge in agriculture, agglomerated metal concentrations might accumulate

in crops. Investigations by the Bavarian State Office for the Environment (LfU) [74] and by the ETH Zurich [75] gave comparable results using a Hitachi S300N scanning electron microscope and X-ray diffraction and IR diffuse reflectance spectrum as NP detection methodology, respectively. Both studies can be considered as validation for the applicability of the sp-ICP-MS approach for NP detection in wastewater. Even though the scope of the respective studies was similar, the investigation approach for NP detection was different. Generally speaking, if sp-ICP-MS is further developed and comprehensively applied, it could become a standard method for NP monitoring in communal wastewater systems, particularly as the majority of monitoring labs are usually already equipped with ICP-MS technology. The further development of the sp-ICP-MS methodology might enable them to perform NP monitoring in wastewater with an acceptable affordability, once real wastewater samples can be determined in a reliable way. The lab-scale investigations with synthetic wastewater are a first step in that direction. Future investigations will have to look also on radio-labelled NPs to identify the process behavior of NPs in real wastewater. A further question in the beginning of the investigations was if nTiO2 and nCeO2 might have a negative impact on the microbial community and lead to disturbances of the wastewater treatment performance. Even though microbiological measurements were not performed during the lab-scale tests, it can be implicitly concluded that this is not the case, as a decrease of the degradation of organic carbon or the removal of nitrogen was not observed during the experiments. No negative effects on the microbial sludge degradation behavior due to nTiO2 and nCeO2 presence were observed. This confirms literature observations [64,74], which refers to the photocatalytic activity of nTiO2 only to UV light [76]. They are not toxic in the opaque sludge dispersion of the lab-scale sewage treatment systems. However, due to the lower pH value, the constantly higher temperature and the synthetic wastewater, there are different conditions in the lab-scale sewage treatment systems than in the Chemnitz WWTP. The study of the Bavarian State Office for the Environment (LfU) [74] also showed that NP (nAg, nZnO, nCuO, nTiO2) have no significant influence on the performance of microbiology in activated sludge.

Furthermore, the results indicate that there are processes that lead to a modification of the NP shape in the effluent, as NP are observed to be significantly smaller than in the stock solution before the addition. Two mechanisms can be assumed to play a role in the shift towards smaller particle size distributions during WWT. First, a slow dissolution of the particles, which should be more pronounced for CeO2 than for the relatively inert TiO2, and second, a preferential sedimentation of larger particles in the final clarifier. This has further implications for the risk assessment of consequent NP release, considering that smaller size NP (which pose the highest risks) might not be cleared from waste water to the same extent as larger particles.

**Author Contributions:** Conceptualization, P.S. and T.L.; methodology, P.S. and T.L.; validation, T.L., formal analysis, P.S., S.S. and K.F.; investigation, T.L.; writing—original draft preparation, P.S.; writing—review and editing, T.L., S.S. and K.F.; visualization, T.L.; supervision, K.F.; project administration, T.L.; funding acquisition, P.S. and T.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Ministry of Education and Research of Germany under the project "NanoSuppe-Behavior of engineered nanoparticles in the pathway wastewater-sewage sludge-plant, using the examples TiO2, CeO2, MWCNT and quantum dots", grant number 03X0144C.

**Acknowledgments:** Thanks to Marcel Neugebauer for conducting the sp-ICP-MS lab experiments.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article Lemna minor* **Bioassay Evaluation Using Computer Image Analysis**

**Oto Ha**ff**ner 1,\*, Erik Kuˇcera 1, Peter Drahoš 1, Ján Cigánek 1, Alena Kozáková <sup>1</sup> and Barbora Urminská <sup>2</sup>**


Received: 17 June 2020; Accepted: 3 August 2020; Published: 5 August 2020

**Abstract:** This article deals with using computer vision in the evaluation of the *Lemna minor* bioassay. According to the conventional method, the growth of *Lemna minor* mass is determined from the number of leaves grown. In this work, instead of counting individual leaves, we propose measuring the area occupied by the leaves using computer vision and compare the new approach with the conventional one. The bioassay is performed according to the ISO 20079 standard as a 168 h growth inhibition test; the aim of the experiment was to quantify the negative effects on the vegetative growth using two parameters—the number of leaves and the area occupied by the leaves. The method based on image processing was faster and also more precise since it enabled us to detect the negative effect of the tested substance on leave size, not only on their number. It can be concluded that the toxic effect has shown to be more significant when considering the leaves area rather than the number of leaves. Moreover, mistakes caused by human factor during leaves counting are eliminated using the computer vision based method.

**Keywords:** *Lemna minor* bioassay; visual system; computer vision; water pollution assessment; bioindicators

#### **1. Introduction**

It is necessary to develop more and more advanced wastewater treatment methods. Novel possibilities of environment monitoring are constantly advancing, which leads to the improvement of toxicant monitoring methods in threatened water. As a relevant approach for the evaluation of the impact of this water on ecosystems, ecotoxicological tests are used. Pollution in wastewater treatment plants is not always removed perfectly, it is only lowered to acceptable levels. It must be judged thoroughly whether treated water has a sufficiently low level of ecotoxicity and if there will be any effect on the organisms in it [1,2].

The *Lemna minor* (lesser duckweed—Angiospermae, Lemnaceae) is one of the water organisms used to measure ecotoxicity. It is a freshwater plant which can be found in most countries all over the world, mainly in lowlands and foothill areas in stagnant and slow-flowing waters. The plant's body consists of a relatively long root and 2 to 5 mm oval shape leaves which can float on the water surface. Mostly, the plant occurs in colonies of two to five leaves. A small single plant can itself reproduce every three days under ideal conditions in nutrient-rich waters [3]. *Lemna minor* bioassays are quite common and easy to perform. The test takes 7 days, where the plant is in the tested samples and by evaluation of the growth inhibition, the effects of substances on vegetation growth are quantified.

During the test of ecotoxicity, the biostimulation or the inhibition of plant growth in the tested samples is monitored for seven days. The objective of the test is to measure the effects of the tested substance on the growth of the *Lemna minor* plant. The basic measured indicator is the number of insoles (leaves). In addition, a second indicator must be measured—either the area overgrown with insoles, dry matter, or chlorophyll. For the legitimacy of the tests, the average increase of the number of leaves has to increase seven times in reference samples. For a standard substance, the level of EC50 (effective concentration causing a 50% inhibition) must be from 5.5 to 10 g·L−<sup>1</sup> and the pH should not change by more than 1.5 unit [4].

The expansion of image analysis algorithms has allowed semi-automatic or fully automatic measurements. Computer systems can evaluate the changes and describe their nature, both qualitatively and quantitatively. It is important to develop a method to extract data encoded in a graphical form, usually created on the features of predefined characteristics. To acquire information from living organisms in the form of digital images, various progressive evaluation tools are used [5–7], including plant species classification using a deep convolutional neural network [8].

Computer vision includes methods for acquiring, analyzing, and understanding images. Applications of computer vision range from computers or robots that can comprehend the world around them through artificial intelligence up to industrial machine vision. Computer vision covers the core technology of automated image analysis which is used in many fields. Machine vision represents a process of combining automated image analysis with other methods and technologies to provide automated inspection or robot guidance in industrial environments and applications [9–12].

In classical bioassays using the *Lemna minor*, dry biomass of plants from the control and the tested samples is compared. However, this is an invasive approach that does not allow one to continue the experiment because the plants are dried for biomass measurements. A non-invasive approach is based on manually counting the number of leaves. The leaf size is not taken into account in this approach. A more accurate and reliable approach to defining the biomass growth rate can be comparison of the area occupied by the leaves in the experimental and control samples. This non-invasive approach allows one to continue the experiment. Computer image analysis, algorithms, and methods improve the objectivity of data collection and evaluation [13,14].

The visual evaluation of the number of leaves is done manually by laboratory workers. This process can be long and loaded with human error. There are devices with a partially automated process for *Lemna minor* experiment evaluation that are owned by big scientific institutions dealing with wastewater management. These devices are often not available for smaller institutions or scientific teams because of the high price.

In the work [15], authors deal with using computer vision in the *Lemna minor* bioassay. Calibration of their method showed a strong correlation between the frond area and the fresh biomass weight. Based on it, several experiments on the reaction of the *Lemna minor* to various detergent solutions (Brij 59 and Brij 38) were performed. Experiments have shown that non-ionic detergents changed the plant surfaces. However, in the ISO 20079:2005 standard [4], the dry biomass weight is considered, whereas in [15], the fresh biomass is used as an indicator for assay evaluation.

In this article, we focus on testing a low-cost method for a *Lemna minor* bioassay evaluation based on computer and machine vision, realizable by a research worker using reliable tools. The objective is to design a methodology for image analysis using a digital camera and a framework software tool. This method was based on the ISO 20079 standard as a 168 h growth inhibition test with the aim to quantify the effects on vegetative growth in terms of the number of leaves compared to leaves-occupied area. The number of leaves is considered as the basic indicator, however this is old and inaccurate according to current modern technological possibilities; growth reaction of the *Lemna minor* can be evaluated more accurately by the leaves-occupied area using computer vision algorithms.

#### **2. Experiment**

A test with the *Lemna minor* was performed according to the ISO 20079 standard (Water quality—Determination of the toxic effect of water constituents and waste water on duckweed (*Lemna minor*)—Duckweed growth inhibition test) as a 168 h growth inhibition test [4]. The observed endpoint was growth inhibition by the effect of the samples during a 7-day exposure (168 h). The aim of the test was to quantify the negative effects on vegetative growth in two parameters—the number of leaves and the leaves-occupied area. Then, it was possible to compare these two evaluation approaches.

The potassium dichromate K2Cr2O7 was used as the reference substance for experiments; the concentrations used were: 5, 10, 20, 40, 80, and 160 mg·L−1. Tests on reference substances serve to verify the quality and sensitivity of organisms and thus, their usability for the test. For K2Cr2O7, the EC50 reference value is in the range of 5.5 to 10.0 mg·L<sup>−</sup>1.

For the purpose of testing, the modified Steinberg medium was used. This medium provides all the necessary nutrients for the *Lemna minor* growth. The Steinberg medium composition was according to [4] and can be seen in Table 1.


**Table 1.** Ingredients of the modified Steinberg medium.

For 1 L of medium, 20 mL of each solution of macronutrients (I–III) and 1 mL of each solution of micronutrients (IV–VIII) were pipetted. The remaining volume was distilled water. At the beginning of the test, approximately 9 beads were placed into each beaker with 100 mL of sample (3 parallel assays for each K2Cr2O7 concentration and 6 assays with a clean modified Steinberg medium as a control) and the beakers were covered with a transparent foil and kept under constant light for 7 days. After a week, the test was evaluated and an EC50 value was determined from both the number of leaves and the area occupied by the leaves.

The growth rate was calculated as follows:

$$r = \frac{\ln|\mathbf{x}\_{tn} - \ln|\mathbf{x}\_{t1}}{t\_n} \times 100,\tag{1}$$

Where *xtn* is the final number of leaves (or the final area), *xt*<sup>1</sup> is the initial number of leaves (or the initial area), *tn* is the duration of the test.

The growth inhibition was calculated as follows:

$$I\_{\mathcal{I}} = \frac{(r\_{\mathcal{L}} - r\_{\mathcal{I}})}{r\_{\mathcal{L}}} \times 100,\tag{2}$$

where *rc* is the growth rate of the reference sample, *rt* is growth rate of tested sample.

In order to evaluate the assay and determine the EC, the dependence of percentual growth inhibition on the sample concentration logarithm was plotted. From the regression line equation, the EC50 value was calculated.

#### **3. Visual System for the** *Lemna minor* **Bioassay Evaluation**

The main idea for using a visual system was the low-cost solution. Our proposed visual system consists of a camera holder, a light holder, a camera, a light, a voltage source, and a computer with image acquisition software. The whole visual system can be seen in Figure 1.

**Figure 1.** Visual system for the *Lemna minor* bioassay evaluation.

The holders are built using the building kit Merkur (Merkurtoys Ltd., Hradec Králové, Czechia). The lights are cheap LED spot lights (Vakesun Industrial Co., Ltd., Shenzhen, China) with constant luminosity and light diffusors made of paper. The camera used in the experiment is Logitech HD Webcam C270 (Logitech Inc., California, CA, USA), with its own image acquisition software Logitech Webcam Software (Logitech Inc., California, CA, USA). The sensing parameters of the camera were set manually to receive focused and well-exposed images (neither under- nor over-exposed), see Figure 2.

**Figure 2.** Well exposed (**left**) and overexposed (**right**) images.

#### **4. Image Processing**

The experiment images have to be processed after image acquisition. In this article, we compare the conventional evaluation method based on leaves counting and the new method based on the leaves-occupied area measurement. To measure the leaves-occupied area, we used the number of pixels representing the leaves. At first, we had to check whether the leaves overlap. If so, it was necessary to detach them from each other. Then, we had to select the area occupied by the leaves. This can be done automatically using sophisticated computer vision algorithms.

In our previous work [16], we proposed the system for visual evaluation of the *Lemna minor* bioassays. In this paper, we will just describe the main algorithm of the proposed system. The block diagram of the algorithm is depicted in Figure 3.

**Figure 3.** The flow chart of the leaves-occupied area counting algorithm.

After the image acquisition, the image was converted from the RGB (red, green, blue) image color space to the HSL (hue, saturation, value) color space. In the HSL color space, it is easy to determine which pixels of the image have a green hue (the leaves), thus the HSL color segmentation can be done. After the color segmentation, we can see the white mask in the image, which determines the area occupied by the leaves (Figure 4).

**Figure 4.** Original image (**a**), image of segmented leaves (**b**), image of the mask of the leaves (**c**).

The result obtained from the selected area is a new binary image where the background is black and the leaves are white (Figure 4c). It is then very easy to count the number of pixels (the white ones) from the binary image using the ready-to-use tool in the image analysis framework. The algorithm was implemented in the NI Vision Assistant software which offers various pre-prepared machine vision algorithms. We used the simple area counting using a module called Particle Analysis.

#### **5. Results**

The numerical results of the tests after 168 h exposure are shown in Tables 2 and 3. Unfortunately, we were not able to maintain the recommended temperature of 24 ◦C during the test, which caused lower growth rate values in the control than required by the ISO standard. However, it was still possible to compare the two methods of evaluation. The method using image processing showed

significantly higher values of growth inhibition, especially at the lowest tested concentrations. A box and whisker plot representation of the result is in Figures 5–8. There are significant differences in growth rate and growth rate inhibition, especially at concentration values 5 and 10 mg·L<sup>−</sup>1. This could be attributed to the fact that low concentration of the toxic substance did not yet have much impact on the number of leaves grown but rather on their size. The differences in concentrations between 20–160 mg·L−<sup>1</sup> were not so significant. In both evaluation methods, higher K2Cr2O7 concentrations caused higher growth rate inhibition (number of leaves and pixels). In case of lower toxic concentrations, using the area counting method may lead to more accurate measurements.


**Table 2.** Results of the test on the *Lemna minor* obtained from the counted number of leaves.

<sup>1</sup> standard deviation.

**Table 3.** Results of the test on the *Lemna minor* obtained from the leaves-occupied area measurement (image analysis).


<sup>1</sup> standard deviation.

Figures 9 and 10 depict the evaluation of EC50 by both methods. Only the values of growth rate inhibition that are lower than 100% have to be used for the calculation, therefore the regression line consists only of these values (black points on the graph). According to the leaves counting method, the EC50 had a value of 8.50 mg·L−1, which lies within the expected range. According to the image processing method, the EC50 was 3.05 mg·L<sup>−</sup>1, which is slightly below the mentioned range. A lower value of EC50 means higher toxicity. It can be concluded that the toxic effect was more significant in the case of the leaves-occupied area than in the case of the number of leaves.

**Figure 5.** Box and whisker plot of the initial and end area of the *Lemna minor* leaves.

**Figure 6.** Box and whisker plot of the initial and end number of the *Lemna minor* leaves.

**Figure 7.** Box and whisker plot of the growth rate for pixels and the number of leaves.

**Figure 8.** Box and whisker plot of the growth rate inhibition for pixels and the number of leaves.

**Figure 9.** Graphical results of the test on the *Lemna minor* from the counted number of leaves.

**Figure 10.** Graphical results of the test on the *Lemna minor* from area measurement (image analyzing).

#### **6. Discussion**

In this study, the image processing-based method was used for the *Lemna minor* bioassay evaluation. According to the ISO 20079: "Water quality—Determination of the toxic effect of water constituents and waste water on duckweed (*Lemna minor*)—Duckweed growth inhibition test" [4], the main indicator of the assay evaluation is the number of leaves. However, our study shows that the number of leaves is not as precise an indicator as the leaves-occupied area and can cause significant differences mainly for lower toxic concentrations.

The image processing approach was faster and more precise since it enabled us to detect the negative effect of the tested substance on the size of the leaves, not just on their number. Also, the errors caused by the human factor during the leaves counting were eliminated when using the image processing. On the other hand, overlapping leaves can cause a false negative result (smaller occupied area) if overlooked. Therefore, it is necessary to carefully check and detach the overlapping leaves before processing. More details of our proposed and used low-cost, reliable method and algorithm can be found in previous works [16–18].

In the work [15], authors showed a strong correlation between frond area and fresh biomass weight of the *Lemna minor*. Based on this, computer image analysis was used to measure the plant surface area. It was confirmed that the *Lemna minor* reacts to a toxic environment (pollution by detergents) by changing the surface area, thus modern techniques such as computer image analysis can be used to

evaluate assays. A small disadvantage of the study [15] is that the weight of the fresh biomass is not considered as a second indicator, but the dry biomass weight as required in the ISO 20079 standard [15]. However, results of experiments confirm that computer image analysis can provide important support in the bioassays. Technical details of the proposed method are not described in depth.

Both our results and the results of [15] confirm that the application of modern technologies, such as computer image processing, and in combination with reliable low-cost hardware for bioassays evaluation, may lead to new standards for quantified and more objective evaluation of the negative influence of toxins on bioindicators. This also leads to the need to perform similar experiments and confirm the need for new, more accurate metrics. However, such experiments require the cooperation of seemingly different scientific disciplines. The synergy of such disciplines is rare, resulting in a low number of similar research works.

In the following research, our aim is to also include color evaluation (hue) in the image processing, which can further improve the sensitivity of the test since many toxic substances cause color changes of the leaves (e.g., a white color is caused by necrosis and a yellowish color by chlorosis). This effect is hard to quantify when using the standard method of counting leaves and can be easily overlooked, especially if the leaves growth has not yet been affected by the toxic substance.

#### **7. Patents**

The system for visual evaluation of the *Lemna minor* bioassays mentioned in Section 4 and in our previous work [16] is protected by the utility model number 8230 by the Industrial Property Office of the Slovak Republic [18].

**Author Contributions:** Conceptualization, O.H. and B.U.; methodology, O.H.; software, E.K.; validation, E.K., J.C., and B.U.; formal analysis, B.U.; resources, B.U.; data curation, J.C.; writing—original draft preparation, B.U.; writing—review and editing, O.H. and A.K.; visualization, J.C.; supervision, P.D. and A.K.; project administration, E.K.; funding acquisition, P.D. and A.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work has been supported by the Cultural and Educational Grant Agency of the Ministry of Education, Science, Research and Sport of the Slovak Republic, KEGA 038STU-4/2018 and KEGA 016STU-4/2020, by the Slovak Research and Development Agency APVV-17-0190.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **pH-Dependent Degradation of Diclofenac by a Tunnel-Structured Manganese Oxide**

#### **Ching-Yao Hu 1, Yu-Jung Liu <sup>2</sup> and Wen-Hui Kuan 3,4,\***


Received: 25 June 2020; Accepted: 3 August 2020; Published: 5 August 2020

**Abstract:** The mechanism of diclofenac (DIC) degradation by tunnel-structured γ-MnO2, with superior oxidative and catalytic abilities, was determined in terms of solution pH. High-performance liquid chromatography with mass spectroscopy (HPLC–MS) was used to identify intermediates and final products of DIC degradation. DIC can be efficiently oxidized by γ-MnO2 in an acidic medium, and the removal rate decreased significantly under neutral and alkaline conditions. The developed model can successfully fit DIC degradation kinetics and demonstrates electron transfer control under acidic conditions and precursor complex formation control mechanism under neutral to alkaline conditions, in which the pH extent for two mechanisms exactly corresponds to the distribution percentage of ionized species of DIC. We also found surface reactive sites (Srxn), a key parameter in the kinetic model for mechanism determination, to be exactly a function of solution pH and MnO2 dosage. The main products of oxidation with a highly active hydroxylation pathway on the tunnel-structured Mn-oxide are 5-iminoquinone DIC, hydroxyl-DIC, and 2,6-dichloro-N-o-tolylbenzenamine.

**Keywords:** diclofenac (DIC); pH-dependent degradation mechanism; reactive site; tunnel-structured manganese oxide; γ-MnO2

#### **1. Introduction**

Diclofenac (DIC), one of the most commonly used nonsteroidal anti-inflammatory drugs (NSAIDs) worldwide, is discharged in large amounts from wastewater treatment plants because of its high hydrophilic nature [1] and low biodegradability [2]. Thus, DIC is widely found in the aquatic environment in a range from ng/L to μg/L and is one of the most frequently detected pharmaceutical and personal care products in water [3,4]. No evidence suggests that DIC is harmful to humans; however, it might be toxic to aquatic organisms and harmful to embryos, infants, children, and adults with low immunity and being sensitive to pharmaceuticals [5–8]. Most of the evidence was focused on its adverse effects on the aquatic and terrestrial organisms, which might cause ecological damage [9–11]. Besides, the transformation products of diclofenac might be more toxic than diclofenac [12,13], which needs to be investigated further.

Manganese oxides (Mn-oxides) are effective natural oxidants for organic pollutants including phenols [14], chlorophenol [14–17], aliphatic amines [18] and anilines [16,19] in soils and sediments. In past decades, Mn-oxides have been used to remove antibacterials and compounds with phenolic and fluoroquinolonic moieties [20,21], triazine, aromatic *N*-oxides [22], tetracyclines [23], and estrogenic compounds such as the synthetic hormone 17R-ethinylestradiol [24,25] and as an alternative

treatment for wastewater or groundwater containing DIC because Mn-oxides are cheap and operated-friendly [26,27]. These compounds may be endocrine disruptors (EDS) and precursors of harmful disinfection byproducts such as haloacetonitriles, haloacetamides, and nitrogenous heterocyclic [28].

In general, Mn-oxides are classified into a layer phase and tunnel structure with edge-sharing and corner-sharing octahedral MnO6, respectively. Studies have commonly used layer-structured birnessite because of its high sorption capacity for target pollutants to remove organic pollutants [15,20,24]. Recently, tunnel-structured Mn-oxides have received considerable attention because of their excellent catalytic oxidative capacity for organic pollutant degradation; however, their mechanisms and feasibility remain unclear and must be investigated [29,30]. γ-MnO2, a tunnel-structured Mn-oxide, normally contains a combination of pyrolusite (1 × 1 tunnel) and ramsdellite (1 × 2 tunnel) and has been confirmed to be environmentally friendly without apparent cell toxicity [31,32]. To the best of the authors' knowledge, this is the first study investigating the action of γ-MnO2 on DIC degradation.

The ionized and acid form of a weak acid such as DIC significantly alters its adsorptive behavior between solution and solid surface, i.e., pollutants adsorption onto oxides from water and analytes separation in high-performance liquid chromatography (HPLC) column. The ionization of a weak acid such as DIC for which forms of the acid exist under different pH values [33]. The extent of adsorption is, as with anions of a weak acid, strongly dependent on pH and favored by lower pH values [34]. In HPLC if the pH of the sample solution and the eluent is not well-matched with each other and around the pKa of the organic acid deformed peaks will appear and then mislead the HPLC analysis conclusions [35,36]. Therefore, the pH could influence the mechanism of the antibiotic interaction with the manganese oxide while those issues have not been systematically addressed.

This study investigated the oxidative degradation of DIC on γ-MnO2 suspensions by varying the key operating parameter pH, which highly influenced the surface features and redox potential of Mn-oxides [37] and the charge density of the chemical form of DIC [38]. Therefore, degradative mechanisms, intermediates, and final products were investigated in terms of pH and compared with the performances of other structures of Mn-oxides presented in the literature.

#### **2. Materials and Methods**

#### *2.1. Materials*

The sodium diclofenac (CAS 15307-86-5) with a purity of >98.5% was purchased from Sigma-A. All other chemicals used in this study were of analytical grade and purchased from Sigma A (St. Louis, MO, USA), J. T. Baker (Phillipsburg, NJ, USA) and Riedel-deHean (Seelze, Germany). Mn-oxide purchased from Tosoh was characterized as tunnel-structured (or molecular sieve) γ-MnO2 (JCPDS 14-0644, PANalytical X'Pert Pro MRD diffractometer) with the Brunauer–Emmett–Teller (BET) specific surface area of 45.6 m2 g−<sup>1</sup> (Micrometrics ASAP, USA, 2010) and pHzpc of 5.1 (Dispersion Technology, USA, DT-1200).

#### *2.2. Batch Experiments*

Experiments were conducted as a function of pH (4–9). For each batch system, various amounts of a γ-MnO2 suspension solution were added to 15-mL glass centrifuge tubes. In the solution, 0.005 M NaH2PO4 and NaH2BO3 were added as a buffer. Various proportions of 0.1 M HCl and NaOH were used to adjust pH to the designed value within a controlled range (±0.07). After each reaction course, the solution pH was remeasured to confirm that it remained within the controlled range. For simplification, the pH value is indicated as the designed value in the following paragraphs. The initial concentration of sodium diclofenac (CAS 15307-86-5) prepared was 100 μM, which was confirmed to be completely soluble, with a water solubility of 10<sup>−</sup>5.1–10−1.78 M [39,40]. The centrifuge tubes were covered with an aluminum foil to prevent light exposure. The suspensions were rotated at 25 ◦C through end-over-end rotations at 10 rpm for a specific time in kinetic trials and 24 h in

thermodynamic tests. All experiments were conducted in duplicate. Moreover, controls (no MnO2 powder) were established using a similar preparatory process to account for sorption on glass tubes and other reactions in the solution.

#### *2.3. Sample Preparation and Analysis*

After reactions, the suspensions were centrifuged (Pico 17, Thermo Scientific, Waltham, MA, USA) at 8000 rpm for 40 min, and then the supernatant was quantified using high-performance liquid chromatography (HPLC, L-7200, Hitachi, Japan) with a diode array (DAD) detector (L-7455, Hitachi) at 270 nm. Chromatographic separation was conducted using an RP-18 column (150 μm × 4.6 μm and an internal diameter of 5 μm) purchased from Mightysi with an eluent comprising 60% acetonitrile and 40% acidified water (25 mM phosphoric acid). The flow rate and injection volume were 1 mL min−<sup>1</sup> and 20 μL, respectively.

#### *2.4. Identification of Oxidation Products*

Major oxidation products were identified using HPLC with mass spectrometry (HPLC–MS). The HPLC–MS system comprised an Agilent 1100 Series liquid chromatography system (LC, Agilent, Palo Alto, CA, USA) with a CTC PAL auto-sampler (CTC Analytica, Carrboro, NC, USA) separation module interfaced with an API 4000 triple quadrupole mass spectrometer (Applied Biosystems AB/MDS Sciex, Foster City, CA, USA). The LC column was a Luna Polar RP (150 mm × 2.1 mm internal diameter) column purchased from Phenomenex (Torrance, CA, USA). The flow rate and injection volume were 0.5 mL min−<sup>1</sup> and 10 μL, respectively. An HPLC gradient was established by mixing two mobile phases: acetonitrile and deionized water, with 10 mM formic acid. Chromatographic separation was achieved with the following gradient: 0 to 1 min: 0% acetonitrile; 1–5 min: linear-gradient to 100% acetonitrile; 5–10 min: 100% acetonitrile; 10 to 10.1 min: 0% acetonitrile; and 10.1–15 min: 0% acetonitrile. Mass spectrometer parameters operated in a positive ion mode were as follows: curtain gas, 20 psi; ion source gas 1, 30 psi; ion source gas 2, 40 psi; source temperature, 500 ◦C; entrance potential, 10 V; and nebulizer current, 5 μA, and the interface heater was turned on. Positive ions in the range of 100–500 m/z were scanned at a cycle time of 1 s. The data obtained were processed with Analyst 1.4.2 software.

#### **3. Results and Discussions**

#### *3.1. DIC Degradation Kinetics on the Tunnel-Structured Mn-Oxide*

Figure 1 presents the kinetic data of DIC degradation on the tunnel-structured Mn-oxide denoted as dot symbols. Interfacial reactions between DIC and γ-MnO2 were highly pH-dependent and initially involved a rapid removal of DIC, followed by gradual slowdown and eventual approach to a plateau. In the acidic medium (pH 4–6), the gradual slowdown period was longer than that in neutral and alkaline conditions (pH 7–9) within the tested pH range. The complicated and multistep reactions between the organic micropollutant interface and Mn-oxides result in limitations of kinetics studies; therefore, only initial reaction rates have been explored in most studies and have been generally characterized with a pseudo-first-order degradation model [27,41]. However, the pseudo-first-order kinetics may not satisfy conditions for the later stage of the interfacial reaction. In general, the interfacial reaction can be initiated by the formation of a precursor complex between the Mn(IV) of oxide surface and target organic pollutants, subsequently followed by electron transfer within the precursor complex, redox product formation (Equations (1)–(5)). Either formation of the precursor complex (Equation (1)) or electron transfer within the precursor complex (Equation (2)) is likely to be the rate-limiting step [42]. The formation of redox products, including the surface Mn(III) and Mn(II) (Equation (4)), and further oxidization or combination of organic radicals to form products (Equation (5)), was rapid because of the unstable nature of intermediates.

$$\equiv Mn^{IV} + DIC \overset{k\_1}{\underset{k\_{-1}}{\rightleftharpoons}} \equiv Mn^{IV} \cdot \cdots \cdot DIC \tag{1}$$

$$\equiv Mn^{IV} \cdot \cdots \cdotDIC \stackrel{k\_2}{\longleftarrow} = Mn^{III} \cdot \cdots \cdotDIC \tag{2}$$

$$\stackrel{\text{g}}{=}Mn^{\text{III}} \cdots \cdotDIC \overset{\text{k}\_3}{\longleftarrow} \equiv Mn^{\text{III}} + DIC \tag{3}$$

$$\mathbf{h} \equiv \mathbf{M} \mathbf{n}^{\text{III}} \xrightarrow{\text{fast}} \mathbf{m} \mathbf{M} \mathbf{n}^{\text{II}} \tag{4}$$

$$DIC \stackrel{fast}{\rightarrow} products\tag{5}$$

An integrated kinetic model [42] was applied to examine the DIC reaction over γ-MnO2. The kinetic equation can be expressed as follows:

$$\frac{d\left[\equiv Mn^{IV}\cdots DM\right]}{dt} = k\_1 \left[\equiv Mn^{IV}\right][DIC] - k\_{-1} \left[\equiv Mn^{IV}\cdots DM\right] - k\_2 \left[\equiv Mn^{IV}\cdots DM\right] \tag{6}$$

Total reactive surface sites (S) for DIC degradation can be represented as follows:

$$S = \left[ \equiv Mn^{IV} \right] + \left[ \equiv Mn^{IV} \cdot \cdots \cdot DIC \right] + \left[ \equiv Mn^{III} \cdot \cdots \cdotDIC \right] + \left[ \equiv Mn^{III} \right] + \left[ \equiv Mn^{II} \right] \tag{7}$$

Both <sup>≡</sup>MnIII—DIC and <sup>≡</sup>MnIII are negligible in Equation (7) because of their high instability. The <sup>≡</sup>MnII concentration can be calculated with the concentration difference between parent DIC at initial (C0) and specific (C) times on account of two electrons transferred from parent DIC to Mn-oxide [43]. To verify the rate-limited step and degradation mechanism, *k1C* >> *k*-1 + *k*<sup>2</sup> and *k*−<sup>1</sup> + *k*<sup>2</sup> >> *k*1*C*, respectively, were assumed for electron transfer control and precursor complex formation control, and the analytical solution of Equation (6) for electron transfer control kinetic model is as follows:

$$\mathbf{C} = (\mathbf{C}\_0 - \mathbf{S}) + \mathbf{S}e^{-k\_{\mathrm{eff}}t},\tag{8}$$

where *ket* equals to *k*<sup>2</sup> and represents the rate constant of the electron transfer control mechanism.

The analytical solution for the precursor complex formation control model is as follows:

$$\mathcal{C} = \frac{\mathcal{S} - \mathcal{C}\_0}{\frac{\mathcal{S}}{\mathcal{C}\_0} e^{k\_{pf}(\mathcal{S} - \mathcal{C}\_0)t} - 1},\tag{9}$$

where *kpf* equals to *<sup>k</sup>*1*k*<sup>2</sup> *<sup>k</sup>*−1+*k*<sup>2</sup> and denotes the rate constant of the precursor formation control mechanism.

**Figure 1.** Diclofenac (DIC) oxidation by Mn-oxide. The initial γ-MnO2 dosage ([MnO2]0) was (100 mg/L) (1150 μM) and the initial DIC concentration ([DIC]0) was 100 μM. Dotted symbols represent the experimental data, lines indicate the fitting model, solid lines denote the electron-transfer control mechanism model, dash lines indicate the precursor complex-formation control mechanism model.

#### *3.2. Correlation between pH and Oxidative Kinetic Constants*

The pH of the solution shifted the degradation kinetics from electron-transfer control to precursor complex-formation control (Figure 1). The electron-transfer control mechanism model successfully described degradation evolution with time under acidic conditions (pH 4–6), whereas under neutral-to-alkaline conditions (pH 7–9), the precursor complex-formation control mechanism was highly fitting to the experimental data (high r-value in Table 1). As Table 1 indicates, the pHs (7–9) with precursor complex-formation control mechanism exactly correspond to the DIC existing as 100% ionized species. Figure S1 (supplementary materials) displayed the DIC species distribution versus solution pH based on the calculation of DIC's pKa 4.15 [43].

When pH was higher than 7, the ionized species account for 100% of DIC in solution. Since the formation of precursor complex of DIC with the γ-MnO2 surface is coupled with a release of OH<sup>−</sup> ions [34], the anionic species of DIC confront with the competition of OH− ions for the surface sites under pH 7–9. Thus, adsorption was not favored by higher pH value (7–9) and the precursor complex formation becomes the control mechanism.

Moreover, the rate constant (ket or kpf) and surface reaction site (S) for DIC degradation decreased when pH increased (Table 1). The inverse relationship between k and pH for DIC could be partially attributed to a decrease in the reduction potential of MnO2 with an increase in pH (Equation (10)).

$$\mathrm{MnO\_2 + 4H^+ + 2e^- \leftrightarrow Mn^{2+} + 2H\_2O \to E\_H^0 + 0.0296 \log \frac{[MnO\_2][H^+]^4}{[Mn^{2+}]}.\tag{10}$$

**Table 1.** Fitting constants (k and S) of the kinetic model under various pH.


\* et: electron-transfer-limited mechanism; # pf: precursor complex-formation-limited mechanism; † calculated by the DIC's pKa 4.15.

In addition to reducing the potential, solution pH alters the amounts of surface reactive sites (S, Table 1). Under acidic conditions, a large amount of S was expected because of the relatively strong affinity of anionic DIC (pKa = 4.15 [44]) species on the surface of net positively charged MnO2. Consequently, electron transfer was limited against sufficient active reaction sites for DIC within the tested acidic pH (4–6). When pH increased, electrostatic attraction between the net negatively charged surface and anionic DIC species decreased. Furthermore, OH− strongly competed against DIC for surface-bound Mn(IV). A lower amount of S at higher pH represented insufficient active reaction sites for DIC attachment, leading to the removal mechanism to shift to precursor complex-formation control mechanism. In addition, each component was actually derived from the initial dosage of MnO2 (Equation (7)); therefore, surface reaction sites were presumed to be functions of initial dosage of MnO2 and solution pH:

$$S = \left[MnO\_2\right] \left[H^+\right]^n. \tag{11}$$

To present the H<sup>+</sup> concentration as pH, the log form can be written as follows:

$$
\log \text{S} = \log[\text{MnO}\_2] - \text{n } pH.\tag{12}
$$

To investigate the influence of pH on S and the kinetic mechanism shift, the log S values extracted from Table 1 and the log value of original MnO2 loading (log 1150 = 3.06) were plotted as a function of pH (Figure 2). The influencing order (n) of pH was determined as the slope of the straight line of correlation and was equal to 0.24. The correlation coefficient (r) of 0.95 corroborates the presumption that the amounts of surface reactive sites are the function of solid oxidant loadings and solution pH.

**Figure 2.** Surface active sites (S) relative to solution pH.

Compared with the Mn-oxide dosage yield (total mole of DIC removal per mole of MnO2 dosage), that for DIC removal by employing γ-MnO2 in this study was 0.07, which falls in relatively higher than 4.49 <sup>×</sup> 10–4–0.14 reported in studies on DIC degradation using other structured Mn-oxides [27,45,46], at 24 h under similar pH conditions. Despite varying the DIC concentration and Mn-oxide dosage from μM to mM in these studies, degradation efficiencies could be compared in a unified manner when the oxide dosage yield was introduced. The remarkable differences in dosage yields indicated that the structure of Mn-oxides substantially influenced their degradative capacity toward DIC, and striking differences were observed for their sorption, oxidative, catalytic, and electrochemical properties [47–49]. The higher oxide dosage yield of γ-MnO2 could be ascribed to the higher amounts of more flexible corner-shared MnO6 sites dominated in Mn-oxide bulk, which may facilitate oxidation for target pollutant degradation [49–51].

#### *3.3. Identification of Oxidation Products Using HPLC–MS*

HPLC–MS was used to determine the M/Z ratio of parent DIC, oxidation intermediates (reaction time of 2 h), and products (reaction time of 24 h), and Figures 3–5, respectively, present their MS chromatograms. Figure 3a has a peak with a very pronounced tailing and this phenomenon should be due to the pH mismatch effect mentioned in a previous study [35]. This fact should not affect the identification of the oxidation intermediates and products because the pronounced tail did not appear after reaction (because the concentration of DIC decreased significantly) and most of the compounds did not appear in this region. System peaks were observed in Figure 4 (Figure 4a,b) and Figure S2. This phenomenon reflected some compounds which are strongly absorbed to the stationary phase were generated after reaction [36]. These compounds cannot be identified using this effluent procedure.

Under neutral-to-alkaline conditions, no intermediates were detected at a reaction interval of 2 h, and compared with acidic conditions, fewer oxidative products were obtained at 24 h according to MS analysis results (Figure S2). Because of the relatively low degradation of DIC under neutral and alkaline conditions, the MS analyses of intermediates and final products were mainly conducted under the acidic condition.

Because of the ionic nature of DIC, two electrospray ionization (ESI) methods, ESI+ and ESI−, were employed to study degradation products, and Table 2 presents the results. Under acidic conditions, three intermediates (I1, I2, and I3) were formed after 2 h of the reaction, and four final products (F1, F2, F3, and F4) were obtained after a day of the reaction.

**Figure 3.** High-performance liquid chromatography (HPLC)– mass spectroscopy (MS) chromatographic patterns of DIC standard. (**a**) Total ion chromatogram (TIC) in the ESI+ mode, (**b**) TIC in the ESI− mode, (**c**) MS patterns in the ESI+ mode, and (**d**) MS patterns in the ESI− mode.

According to Monteagudo et al. [52,53], I1 (RT = 2.09, m/z = 346) correspond totri-hydroxyl-DIC (m/z = 346) or di-hydroxyl-DIC (m/z = 328). I2 (RT = 2.48, m/z = 298) should be a hydrolyzeddecarboxylated DIC (296 − 14 + 16 = 298). The molecular weight of I3 (RT = 2.92, m/z = 597) is considerably higher than that of DIC. Moreover, I3 exhibited numerous isotopic peaks, and its intensity ratio of (M + 1)/Z to (M + 3)/Z was approximately 3:4, which revealed that these compounds contained four chlorine atoms. Thus, I3 should be a dimmer of 5-iminoquinone DIC (m/z = 308) and another intermediate. This finding indicated that polymerization or dimerization, which was found during the reaction of other aromatic compounds with Mn-oxides, may occur during DIC degradation by γ-MnO2. Similar results were reported by Huguet et al. [26].

F1 (RT = 3.02, m/z = 503) is a new product, and its molecular weight is substantially higher than that of DIC. Therefore, it should be a transformation product of I3. F2 (RT = 3.58, m/z = 308) and F3 (RT = 3.93, m/z = 312) correspond to 5-iminoquinone DIC and hydroxyl-DIC, respectively, which have been reported in literature [26]. The peak of hydroxyl-DIC (F3) split into two and the m/z ratio (255) in negative mode was considerably lower than the m/z ratio (312) in the positive mode. The split of the peak could be attributed to the different sites of the hydroxyl group of the compound (structure isomers) leading to different hydrophilicity, and the observed difference of m/z ratio for F3 between in positive mode and negative mode should result from the carbon chain (–CH2COOH, M = 57) broke during ionization. F4 (RT = 2.59, m/z = 250) should be a decarboxylation product of DIC (2,6-dichloro-N-o-tolylbenzenamine), which was reported by Martínez et al. [54].

The intensity of F3 was much higher than that of F2, and multiple hydroxyl intermediates (I1) were found. According to studies, decarboxylation, hydroxylation, and dimerization are the three main pathways of DIC transformation by Mn-oxides [26]. The pathways of DIC transformation by γ-MnO2 are the same as those of birnessite or other natural manganese oxides [26], and compared with the layer-structured birnessite that is widely used in studies, hydroxylation of DIC by γ-MnO2 was more active than that through other pathways. This phenomenon could corroborate that the large amounts of highly flexible corner-shared MnO6 may provide abundant reactive hydroxyl groups and facilitate oxidation for target pollutant degradation [50,51]. Therefore, the dimerization products of DIC obtained through γ-MnO2 are highly hydrophilic and can be detected without extraction. Hydroxylation intermediates were not detected after 1 day because they were oxidized to smaller or hydrophilic compounds due to further hydroxylation.


**Table 2.** MS data of intermediate and final products.


**Table 2.** *Cont.*

RT: retention time; bold number: parent ion.

#### **4. Conclusions**

This study demonstrated that the pH of media highly influences DIC oxidative degradation on the tunnel-structured Mn-oxide (γ-MnO2). The reduction potential of Mn-oxide, the number of surface reactive sites (S), and electrostatic affinity between DIC and γ-MnO2 increase with a decrease in pH value. Consequently, the electron-transfer control mechanism model successfully described degradation evolution with time under acidic conditions (pH = 4–6). While under neutral-to-alkaline conditions (pH = 7–9), the precursor complex-formation control mechanism was highly fitting to the experimental data. At pH 7–9 the anionic species account for 100% DIC in solution and hence confront with the competition of OH<sup>−</sup> ions for the complex formation on the γ-MnO2 surface. In contrast, the acid form of DIC with a substantial ratio under pH 4–6 is favored for the surface complex formation with less competition. The results of the analysis of oxidative intermediates and products by using HPLC–MS revealed decarboxylation, hydroxylation, and dimerization as the three main pathways of DIC transformation by γ-MnO2. Although the oxidation products obtained by γ-MnO2 are similar to those obtained by other Mn-oxides, hydroxylation of DIC by γ-MnO2 is more active than other pathways because of an abundance of flexible corner-shared MnO6 for target pollutant degradation.

**Supplementary Materials:** The following are available online: http://www.mdpi.com/2073-4441/12/8/2203/s1, Figure S1: DIC species distribution versus solution. HA and A− represent the acid and ionized form of DIC, respectively. The black and gray line were calculated based on pKa = 4.15 of DIC [44]. When pH higher than 7, the ionized species (A−) accounts for 100% of DIC in solution pH. Figure S2: LC–MS chromatographic patterns of degradation intermediates (a) TIC in ESI<sup>+</sup> mode (b) MS patterns in ESI<sup>+</sup> mode under pH 7.0 for reaction time = 24 h with initial MnO2 loading 200 mg.

**Author Contributions:** C.-Y.H. conceived and designed the experiments. Y.-J.L. performed the experiments. C.-Y.H. and W.-H.K. analyzed the data and wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received the funding from Ministry of Science and Technology of the Republic of China, Contract No. NSC 98-2221-E-038 -004, MOST 105-2221-E-131-001-MY3 and MOST 108-2221-E-131-024-MY3.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
