Potential of Luffa cylindrica seed as coagulation-flocculation (CF) agent for the treatment of dye wastewater: Kinetic, mass transfer, optimization and CF adsorption studies

2.1.1 Luffa cylindrica seed (LCS)

Luffa cylindrica sponges were collected around farm settlements in Amawom, Ikwuano Local Government Area, Abia State, Nigeria, between August and September, and dried in sun. The Luffa cylindrica seeds were isolated from the matured dried sponges. Separated seeds were washed, dried and homogenized to obtain LCS and placed in an airtight container for further use (see Fig. 1). The proximate analysis of the LCS was based on standard methods (AOAC, 2005). Coagulant (LCS) and post treatment settled sludge (PTSS) were characterized using the following instruments: an infrared spectrometer (Agilent Technologies) with a resolution range of 4000–650 cm⁻¹ and 30 8 cm⁻¹ scans with 16 background scans to obtain FT-IR spectra. A scanning electron microscope (Phenon-World, MVE 016477830) for morphological characteristics of the sample. An X-Ray diffractometer (XRD) spectrum was obtained using PHILIPS – ASCII (UDF) with XPERT diffractometer to determine the crystalline nature of the materials.

Close

Fig. 1

Schematic diagram of LCS preparation up to application as coagulant.

Export to PPT

2.1.2 Dye-based wastewater

The characterization of dye-based wastewater was determined by standard methods (AWWA, 2012). The COD and the Cr(VI) were determined using dichromate method (5220A) and flame adsorption spectrophotometry (3111B), respectively (AWWA, 2012).The characterization was carried out at the National Soil, Plant, Fertilizer, and Water Laboratory, Umudike, Nigeria. Mettler Toledo Delta 320 pH meter, Conductivity meter, DDS 307 and UNICO 1100 Spectrophotometer were used to evaluate the pH, electrical conductivity and CTSS.

2.5.1 Langmuir isotherm

The Langmuir isotherm is based on the fact that the maximal adsorption on the adsorbent surface corresponds to a saturated monolayer of solvent molecules, the adsorption energy is uniform, and no adsorbate transmigration on the surface plane ((Ezemagu et al., 2016). The Langmuir equation is written as:

Where $K_L$ _, is the Langmuir constant (Lg⁻¹); $q_e$ is the equilibrium dye concentration on the adsorbent (gg⁻¹); $C_e$ is the equilibrium dye concentration in solution (gL^-1); qm is the concentration of dye when the adsorbent forms a monolayer (gg⁻¹). $K_L$ and $q_m$ can be evaluated from the slope and intercept of plot $\frac{1}{q_e}vs\frac{1}{C_e}$ .

2.5.2 Freundlich isotherm

The Freundlich isotherm is based on the premise that bio-sorption heat has a heterogeneous surface with a non-uniform distribution over the surface. The linearized empirical equation is expressed as:

Where the Freundlich constant is $K_f$ (gg^-1L^1/n g^−1/n) and 1/n is the heterogeneity factor associated with the bio-sorption potential and strength, respectively.

2.5.3 Tempkin isotherm

The Tempkin isotherm model considered the impact on adsorption isotherms of certain indirect adsorbate/adsorbate interactions and proposed that the adsorption heat of all the molecules in the layer would decrease linearly with coverage due to these interactions (Menkiti et al., 2018). The Tempkin isotherm linearized equation is given as:

2.5.4 Adsorption kinetics

To determine the mechanism of CF adsorption performance on the biomass during the CF process, the Lagergren pseudo first order model, Ho's pseudo second order model, was analyzed. The expression for pseudo first order rate is given as:

The rate constant of first order adsorption (min⁻¹) is $K_1$ _, where $q_e$ and $q_t$ are the amount of dye adsorption on biomass at equilibrium and at time t, (gg⁻¹), respectively. The rate of pseudo second order is expressed by (Menkiti et al., 2018):

Where the rate constant $K_2$ is for pseudo second order adsorption (gg^-1min⁻¹).

2.6.1 Numerical solution technique (finite difference technique)

With a view to solving the Eq. (22), the technique of approximation was used. The explicit finite difference method (FDM) was applied and the equation was transformed into a differential equation by dividing the solution domain into a grid of points represented along with each mesh point and referred to as nodes in the form of mesh and derivatives. The solution was obtained when at each node the dependent variable was first defined, and then approximated until the final step for the next step. Therefore, the finite display of the mesh points is shown as follows (Jimoda et al., 2013):

Where δx and δt indicate grid sizes and subscripts denote the location of the dependent variable being considered in the × and t directions, respectively. Let the process of jar CF occupy the $0\le x\le L$ space and divide the gap in the space of the unit into equal δx intervals and time into δt intervals. The finite difference representations of various derivatives appearing in the governing equation are obtained from expanding the Taylor series (Stroud, 1987). Then, applying the expansion in the t-direction while preserving the constant and truncated second term of the series for the left side of Eq. (22) then gives,

Therefore,

Similarly, expanding in the x-direction, keeping t, constant, i.e., for the right-hand side of Eq. (22), becomes:

Again, applying Taylor’s series expansion in × direction (backward difference) keeping t constant:

Adding Eqs. (29) and (30) gives:

Equating Eq. (28) to Eq. (31) and introducing $D_{\mathit{eff}}$ as in Eq. (22) gives the governing equations:

Where r = $\frac{D_{\mathit{eff}}\delta t}{{(\delta x)}^2}$

At

Then, calculate $C_{-1,j}$ (pseudo concentration at external mesh point) and depicts the relevant grid points at x = 0 and x = L. The simplest way to represent the condition at x = 0, for forward difference, in finite difference is given by: $$\frac{C_{1,j}-C_{0,j}}{\delta x}=C_{0,j}$$

This gives an additional $C_{0,j}$ equation at any time stage to be used instead of a given $C_{0,j}$ value in an explicit finite difference. Now, $C_{-1,j}$ must be inserted at the external mesh point to represent a pseudo concentration more accurately at ×  = 0 by the central difference equation by imagining that the tissue sheet is very slightly extended. Accordingly, in terms of central difference representation, the finite difference approximation at boundary x = 0 is:

To eliminate $C_{-1,j}$ between Eqs. (33) and (34), we have that Eq. (35) as:

Substitute Eq. (36) into Eq. (34) and expanding gives $$C_{0,j+1}=C_{0,j}+r\left[C_{1,j}-2\delta x{C}_{0j}-2C_{0,j}+C_{1,j}\right]$$

Similarly, at ×  = L (for other side of the tissue) the derivative boundary condition becomes:

In terms of central difference representation, we have,

Similarly, if the value ( $C_{11}$ ) is eliminated between Eq. (38) and Eq. (39), we have:

2.6.2 Computer programming and simulation

Codes for Eq. (40) implementation for the rate of mass transfer during wastewater coagulation using LSC was developed in the MATLAB environment and analyzed under various process operating conditions. The criterion for the stability and convergence of the solution was tested and met ( $\frac{D_{\mathit{eff}}\delta t}{\delta x}\le \frac{1}{2}$ ) for each simulation run of the CF process. The following statistical parameters were used: mean square error (MSE), root mean square error (RMSE), mean absolute percentage error (MAPE), and mean absolute deviation (MAD). Using Eqs (41) – (44)., the statistical parameters above were calculated.

3.1.1 Luffa cylindrica seed (LCS)

The proximate composition of LCS previously documented (Nnaji et al., 2020b; Onukwuli et al., 2021), revealed 21.88% crude protein, which is high and comparable to published work as one of the active component for CF process. The relative high protein could be the reason for good performance of LCS as a coagulant for the treatment of dye-based wastewater. Documented literature has defatted LCS crude protein ranging from 42.17 to 70.65%, 45.06– 50.06%, whereas non-defatted LCS crude protein is 28.45% (Nnaji et al., 2020b).

3.1.2 Characterization of dye-based wastewater

The characteristics of dye-based wastewater revealed high dissolved solid particles with obvious high conductivity and the presence of heavy metals. The results indicated pH of 5.55, turbidity 340.53FAU, BOD 316.35 mg/L, COD 1930.4 mg/L, Lead 2.27 mg/L, Nickel 8.54 mg/L and Chromium 2.315 mg/L. The value of turbidity shows the presence of total dissolved and suspended solid in the wastewater. The high COD value and the presence of heavy metals such as Lead, Nickel and chromium are indicators of pollution.

3.1.3 FT-IR spectra characterization of LCS and PTSS

It is important to understand the essence and properties of a material in order to make the best use of it. Therefore, FT-IR was used to investigate the LCS and PTSS surface characteristics. The LCS and PTSS spectra are shown in Fig. 3. Fig. 3 showed the existence for LCS and PTSS of a polymer –OH, functional group extending between 3276.3 cm⁻¹ and 3280.1 cm⁻¹ (Jiang et al., 2020). Methyl C-H and methylene C = H asymmetric stretching is applied to spectra stretch 2951.1–2914.8 cm⁻¹ for LCS and PTSS, while 1707.1 cm⁻¹ could be carboxylic ketone stretching. For LCS and PTSS, 1636.3 cm⁻¹, 1543.1 cm⁻¹, and 1408.8 cm⁻¹ could be due to olefin, aromatic ring stretches and asymmetric alcohol bends, respectively. Similarly, the stretch for LCS and PTSS at 1226.3 cm⁻¹ and 1036.2 cm⁻¹ may be attributed respectively to skeletal C = C vibration and primary amine C-N The medium peak for PTSS at 1770.5 cm⁻¹ and 1457.4 cm⁻¹ could be due respectively to carboxylic and skeletal C-C stretching of vibration. Also present in PTSS are 1080.9–1028.7 cm⁻¹ stretching and 752.9 cm⁻¹ stretching, respectively, which may be cyclohexane vibration and C-H alkyne bend. Protein and fatty acid structures commonly contain these groups, especially the hydroxyl and carboxyl groups (Dalvand et al., 2016). The spectra are comparable to the spectra obtained for organic compounds like protein(Nandiyanto et al., 2019). The majority of the functional groups found in LCS and PTSS serves as an active site for color/colloidal particle attachment (de Souza et al., 2016; Onukwuli et al., 2021). Another indicator of LCS ability to pick up metal ions in solution is the presence of – OH (which could be proof for possible pore hydration) and carboxylic stretching. The changes in FTIR spectra in the two samples indicate that functional groups are involved in dye and colloidal adsorption during coagulation and flocculation (Joshi et al., 2020). The shift in peaks could be attributed to changes in functional groups’ chemical environment, while, bandwidths appearance and disappearance as seen in PTSS points at reactions that involves the relevant functional groups(Nnaji et al., 2020b).

Close

Fig. 3

FTIR spectra of (LCS) (a); Post treatment settled sludge (PTSS) (b).

Export to PPT

3.1.4 SEM/elemental results of LCS and SS

LCS and PTSS (post CF) surface morphology is provided in 1500 × magnification (see Fig. 4). Fig. 4a for LCS indicates fibrous nature and irregular structure, some fissures and pores are present. This is an example of a macroporous structure providing active sites for coagulation-flocculation and adsorption (Ahmad & Haseeb, 2015). The PTSS surface morphology displayed in Fig. 4b is in 1500 × magnification. The figure showed expansion of LCS molecules and aggregated particles of LCS and dye, attributed to the CF process involving certain chemical reactions, which led to the breakdown of the existing structure to allow floc formation. Table 2 describes the elemental compositions of LCS, CBD, DRD and PTSS. Clearly, the presence of elements, which were originally not seen in LCS, but subsequently appeared in the PTSS matrix was an indicator of the transfer of elemental particles from dye-based wastewater. The presence or absence of such particles could not be unconnected to a chemical reaction that contributed to depositing or adding the elements to the PTSS matrix, or dissolving them into decontaminated wastewater solution.

Close

Fig. 4

SEM micrograph LCS at 1500x magnification (a) and PTSS at 1500 × magnification (b).

Export to PPT

3.1.5 XRD results of LCS and PTSS

X-ray diffraction is a valuable method for the study of the material's crystalline structure. Fig. 5 displays the X-ray diffraction spectra for both LCS and PTSS. Fig. 4a for LCS shows 14 clear and high peaks and numerous halos with amorphous humps and lots of noise, while, Fig. 5b indicated 13 clear peaks, between 10 < 2 $\theta$  < 70° strip. This indicates characteristic patterns of a partially crystalline nature where scattered bands were formed by poorly organized molecules. For PTSS, Fig. 5b showed similarly organized molecules with peak changes. Also, the crystal peaks in Fig. 5 a, b, when compared with the regular crystal peaks shifted from left to right axes, which could be attributed to the expansion or contraction of LCS (Menkiti et al., 2016). Fig. 5a, b indicated crude lattice structure for LCS and PTSS, respectively. The peak changes may be due to reactions that lead to molecular particle attachment. It could be formation of magnesium compound which dissolved into the solution of the treated wastewater. If material studied indicates well defined peaks, it is said to be crystalline, whereas non-crystalline or amorphous material displays halos or humps(Zhao et al., 2017).

Close

Fig. 5

The XRD spectra of (a) LCS (b) post treatment settled sludge (PTSS).

Export to PPT

3.2.1 Effect of LCS on CTSS, COD and chromium (VI) removal

Fig. 6 shows the effect of varying dosages of LCS between 1.0 and 1.8gL^-1 for the efficiency of CTSS, COD and Cr (VI) removal at varying pH after 300 min settling. Fig. 6a clearly showed better coagulant performance in removing CTSS at acidic medium, especially pH 2, recording>94% between 1.0 and 1.6gL^-1. The optimum dosage, however, which provided 98.4% CTSS removal was 1.2gL^-1 at pH 2. The high efficiency of removal of CTSS could be attributed to the strong interaction between LCS and wastewater molecules that led to the destabilization/neutralization of the negatively charged CTSS particles in dye-based wastewater (Okolo et al., 2016). The lower efficiency of removal recorded at other pH, in particular, pH 4 to 8 could be linked to reversal of the charges on particles resulting in re-stabilization and a decline in CF. LCS was believed to have functional groups which serve as an active site for CTSS attachment(Ani et al., 2012). Fig. 6b shows the COD removal efficiency at varying pH and biomass dosage after 300 min of settling. Fig. 6b clearly highlighted better biomass performance at pH 2, which recorded between 86.53 and 88% for all dosages. This can be due to a reduction of the wastewater organic portion (CTSS) prior to the COD test. Acidic medium performance is compatible with other literatures, as it favors the oxidation reaction that resulted in the removal of organic wastewater components (Jorfi et al., 2018). The optimum dosage of 1.4gL^-1 at pH 2 gives the best COD removal efficiency of 88%. Fig. 6c shows the effect of varying dosage of coagulant between 1.0 and 1.8gL^-1 and pH between 2 and 10 for removal efficiency of chromium (VI) after 300 min settling time. Fig. 6c shows a different impact on coagulant efficiency when removing chromium (VI) from dye-based wastewater compared with CTSS and removal of COD. The figure indicated better output at an aqueous pH 6–10 for all dosages applied. As the dosage increases, the removal of chromium (VI) increased, but undeniably the lowest was reported at pH 2. The best coagulant performance for removal of chromium (VI) was reported at pH 6 with 1.8gL^-1 providing a removal efficiency of 97.5%. The active attachment site established in LCS could be attributed to this performance.

Close

Fig. 6

The plot of dosage and pH effect on the removal efficiency of CTSS(a), COD(b) and Cr (VI)(c) after 300 min settling time.

Export to PPT

3.2.2 Effect of settling time on CTSS removal using LCS

The effect of settling time on removing CTSS using LCS was assessed at varying wastewater pH and dosage of 2 – 10 and 1.0 – 1.8gL^-1, respectively, in the range 5 – 300mins. Fig. 7a showed the effects of assessing the efficiency of removal of CTSS at constant optimum pH of 2 and varying dosages ranging from 1.0 to 1.8gL^-1. The efficiency of removal of the CTSS improved with increased settling time. Throughout the indicated time, the first three dosages (1.0, 1.2, 1.4gL^-1) maintained a steady increase in CTSS removal. The dosage of 1.2gL^-1 with removal efficiency>87% after 30mins of settling and reported 98.4% removal at the end of 300mins was considered to be optimal dosage. The results of the assessment of CTSS removal at the constant optimum dosage of 1.2gL^-1 and the varying aqueous solution pH range 2–10 were shown in Fig. 7b. The superiority of the pH value was clearly observed in the CF process. In this investigation the bio-coagulant better performance recorded at acidic pH (especially pH 2) was evident as indicated in Fig. 7b. The figure showed that fair removal efficiency was observed at an alkaline pH of 10 as settling time increased. This is consistent with the literature available (Patrick et al., 2015; Onukwuli et al., 2021).

Close

Fig. 7

The plot of settling time effect on the removal efficiency of CTSS at constant pH and varying dosage (a), constant dosage and varying pH (b).

Export to PPT

3.4.1 Adsorption isotherm

Summaries of isotherm parameters, R² and adjR², SSE and RMSE, were shown in Table 3. The maximum monolayer adsorption capacity (qm) of dye-based wastewater was found to be 533.0gg⁻¹ by Langmuir, while R² was 0.8040. R_L > 0 but<1 is an indicator of favorable dye mixture adsorption on LCS. The value of Freundlich's K_F parameter is 1271.6gg⁻¹(g⁻¹)^{1 / n}. The value of n is 4.819. n > 1 showed a favorable and heterogeneous adsorption value. Also, 4.558e-4 and 148.7 were found to be the values of the Tempkin parameters A and B. The experimental data of these isotherms summarized in Tables 3, shows that compared to the Langmuir and Tempkin models, the Freundlich model provided a better fit with the highest R² value. The SSE and RMSE, were also found to be relatively small.

3.4.2 Adsorption kinetic results

Coagulation-flocculation data were applied to the Lagergren's pseudo first and Ho's pseudo second order kinetic models for the adsorption of dye-based wastewater onto biomass and analyzed. The model plots of the adsorption kinetic have been presented, see Fig. 8. Lagergren adsorption constant, K₁ of 0.02439 min⁻¹, adsorption capacity at equilibrium, q_e.cal 140gg⁻¹, R² 0.9449, adjR² 0.9394, SSE 0.6693 and RMSE 0.2587 were obtained from the log plot (q_e-q_t) against t and the fit tables using linear polynomial in the MATLAB setting, while the q_e._exp, was 712.1gg⁻¹. The adsorption capacity determined from the fit data at equilibrium, q_{e cal}, was found not to be similar to the experimental adsorption capacity, q_{e exp}. The R² and adjR² were found to be > 0.9, while the SSE and RMSE were moderately low.

Close

Fig. 8

Adsorption kinetic model plots (a) Lagergren's pseudo first, (b) Ho's pseudo second order.

Export to PPT

Likewise, the pseudo second order model of Ho fitted in Fig. 8b and based on Eq. (14) revealed K₂ 1.404e-1 min⁻¹, q_{e cal} 821.7gg⁻¹, R² 0.9771, adjR² 0.9748, SSE 0.003557 and RMSE 0.01886 were obtained from the t/qt versus t plot and the fit table of the linear polynomial of MATLAB curve fitting tools. The experimental adsorption capacity, q_{e exp}, was 712.1gg⁻¹. From the results, the equilibrium adsorption capacity, q_{e cal}, calculated from the fit data, was found to be relatively close to the experimental q_e. It was found that the R² and modified R² were > 0.9, while the SSE and RMSE were very low. The kinetic data showed that when compared to the first order model, Ho 's second order model was best suited due to high R² and its agreement with the experimental values. This is a strong indication that the second order kinetics model was followed by the method. Moreover, relative to the first order model, the SSE and RMSE for the second order were very low. The above confirmed that chemisorption, which involves valence forces through electron sharing or exchange, is the rate limiting step.

3.6.1 Model adequacy

To assess the adequacy of the models, model overview statistics and ANOVA were used. While the ANOVA was presented in Table 4, the model overview statistics revealed R² 0.8293 and adjR² 0.6099, R² 0.9584 and adjR² 0.9050, and R² 0.9935, adjR² 0.9851, for CTSS, COD and Cr (VI) removal responses, respectively. The p-value for CTSS removal from the ANOVA table was 0.0467, while that of COD and Cr (VI) removal were < 0.01 for the model chosen. For CTSS, COD and Cr (VI) removal, the model summary statistics showing R², of 0.8293, 0.9584 and 0.9935, respectively, suggested that the empirical model could not explain only 17.07%, 4.16% and 0.65% of the variance. The high R² values suggested that the model was well suited to the response (Okolo et al., 2018). It can also be seen from the findings that the experiment indicates a desirable and rational alignment with the R² 's proximity to the adjR². This closeness ensures that the quadratic models are satisfactorily adapted to the experimental data (Okolo et al., 2018).

The second order regression for the efficiency of CTSS, COD and Cr (VI) removal indicate that the models were significant because the f-values of 3.78, 17.6 and 118.56, respectively, were high. Similarly, for quadratic regression models, the p-value that gives an indicator of the importance of a model in relation to the f-value was<0.05. For a 95% confidence level, this showed that the models were statistically significant, indicating only a 5% risk that the f-value was due to noise. The model is not significant if the p-value is above 0.1(Okolo et al., 2018).

3.6.2 Process analysis

Table 4 also, explains the linear (X₁, X₂, X₃), quadratic (X₁², X₂², X₃²) and interaction (X₁X₂, X₂X₃, X₁X₃) effects of the parameters. From the table, the response, Y₁, revealed that X₁, X₂, X₁², X₂², and X₂X₃ are significant terms with p < 0.05, while X₃, X₃², X₂X₃ and X₁X₃ are not significant terms. Similarly, Y₂ response have significant terms of X₁, X₂, X₂², X₁X₂, while X₃, X₁², X₃², X₁X₃, X₂X₃ are not significant. Also, from the table, the response, Y₃, shows that X₂, X_3, X₁X_3, X₂X_3, X₁², X₂² and X₃² are all significant terms. Only X₁ and X₁X₂ are not significant. The lack of fit f-values of 1.62, 5.60 and 5.10 for Y_1, Y₂ and Y₃, respectively, imply that the lack of fit is not significant relative to the pure error. There are only 13.770%, 28.97% and 7.48% for Y_1, Y₂ and Y₃, respectively, chance that the lack of fit f-values this large could occur due to noise. Non-significant lack of fit is good, the model needs to fit.

3.6.3 Effect of variables on CTSS, Cr (VI) and COD removal efficiency

The 3-D response surface plots in Fig. 10 showed the individual and combined effect of LCS dosage, solution pH and stirring time on the percent CTSS, COD and Cr (VI) removal based on a total of 17 BB-design of the three variables. The surface graph which is a representative of the model used to visualize the relationship between the response and the experimental data showed curvilinear profiles and significant curvatures. The distinct curves for the relationship between pH and dosage, pH and stirring time, dosage and stirring time variables, clearly depict their interactions in the process This is consistent with the quadratic model and an indication that the independent variables have strong interactions(Imen et al., 2013). Optimum condition at 1.4gL^-1 dosage, pH 2 and 30mins stirring were indicated by the response surface plots, giving 99.20% CTSS removal, with 1.4gL^-1 dosage, pH 2 at 30mins stirring, giving 90.07% COD removal and at 1.8 g/L dosage, pH6 and 15mins giving 98.29% Cr (VI) removal, respectively.

Close

Fig. 10

3-D response surface plots for CTSS (a), COD (b) and Cr (VI) (c) removal efficiency.

Export to PPT

3.6.4 Validation of the optimized parameters of the coagulation process based on RSM

The optimized results were validated experimentally by carryout triplicate experiment and the average treatment efficiencies obtained are shown in Table 5. It can be shown in Table 5 that the optimum experimental validated % removal efficiency of CTSS, COD and Cr (VI) are close to the model predicted values at the same optimum conditions.