Sustainable water reuse in textile processing: A hybrid nanofiltration–reverse osmosis–persulfate oxidation treatment approach

Abstract

Water reuse is becoming a key strategy for improving the environmental sustainability of textile processing, particularly in water-intensive operations such as dyeing and finishing, where membrane processes such as nanofiltration (NF) and reverse osmosis (RO) are increasingly applied. However, their implementation is often constrained by the management of highly concentrated residual streams. This work proposes a hybrid membrane–oxidation strategy to enable practical minimal liquid discharge (MLD) by combining selective separation with targeted post-treatment of the concentrate. Biologically treated textile wastewater was processed by NF and RO under different operating pressures to assess permeate quality, energy demand, and retentate composition. The NF–RO configuration produced very low-salinity permeate with specific energy consumption as low as 0.38–1.2 kWh m⁻³, with NF at 6 bar identified as the most energy-efficient condition. At this pressure, NF selectively concentrated sulfate-rich salts and refractory organics in the retentate, reducing scaling propensity in the downstream RO stage. The NF retentate was subsequently treated by heat-activated persulfate oxidation, optimized using a Box–Behnken design. Temperature was the dominant factor, and optimal conditions (90°C, 100% persulfate stoichiometric dosage, 10 mg L⁻¹ Fe(II)) achieved a total organic carbon removal of ~64%. Integrating selective membrane fractionation with thermally driven oxidation offers a robust, energy-efficient route for practical MLD implementation in saline textile wastewater treatment.

Graphical Abstract

This is a visual representation of the abstract.

Water reuse has become a strategic priority in water-intensive industries, driven by increasing regulatory pressure on effluent discharge and the growing scarcity of freshwater resources. Biological treatments, including anaerobic configurations such as upflow anaerobic sludge blanket (UASB) reactors, can effectively remove part of the biodegradable organic load and color from textile wastewaters, thereby playing an important role as pretreatment steps in integrated treatment trains. ¹ Nevertheless, even after biological treatment, the effluent still contains a complex mixture of refractory organics and dissolved salts. These residual components limit the feasibility of water reuse by compromising permeate quality and constraining the trade-offs between membrane selectivity, permeability, and energy efficiency in advanced treatment systems.^2,3

Water reuse strategies increasingly rely on pressure-driven membrane processes, particularly nanofiltration (NF) and reverse osmosis (RO), whose performance continues to benefit from advances in membrane materials and design aimed at improving selectivity, fouling resistance, and overall treatment efficiency. ⁴ NF selectively removes divalent ions and low-molecular-weight organic fractions through a combination of size exclusion and electrostatic interactions, whereas RO achieves near-complete rejection of dissolved species. In practice, membrane performance is highly sensitive to operating pressure. While increasing pressure can improve solute rejection, it also raises specific energy consumption, promotes fouling, and intensifies concentration polarization, especially in complex industrial matrices. In textile wastewaters, fouling is frequently driven by organic–inorganic interactions enhanced at high transmembrane pressures, resulting in irreversible permeability losses and a reduced operational window for NF and RO systems.^5,6

One of the main challenges in high-recovery water-reuse schemes remains the generation and management of membrane retentates (also referred to as “reject” stream). Even at high recoveries, most dissolved salts and refractory organic matter are confined to a relatively small fraction of the treated volume, disproportionately increasing the risks of scaling, energy demand, and environmental impact. From a process-engineering perspective, the feasibility of membrane-based reuse systems therefore depends not only on permeate quality, but also on the volume, composition, and downstream treatability of the residual concentrate.^7,8 In minimal liquid discharge (MLD) frameworks, process performance is typically benchmarked against high water recovery targets, generally exceeding 80–90%, while minimizing the volume of residual concentrate requiring further treatment or disposal.^7,8 In membrane-based systems treating industrial effluents, these targets are often pursued through staged separation processes, in which selective removal of multivalent ions and organic matter enables improved recovery and reduced scaling propensity. However, attaining such recovery levels while maintaining acceptable energy consumption and operational stability remains challenging, particularly in saline textile wastewaters, where fouling and scaling phenomena intensify at high concentration factors.^9,10 As a result, the design of MLD-oriented systems requires not only efficient separation, but also effective strategies for managing and conditioning the resulting retentates. To address these limitations, hybrid configurations combining NF and RO have been proposed. By selectively removing divalent ions and part of the organic load upstream, NF lowers osmotic pressure and scaling propensity in the downstream RO stage, enabling more favorable operating conditions and improved permeate quality.^11,12 Importantly, this upstream fractionation modifies both the ionic composition and the molecular-weight distribution of the solutes entering the RO unit. As a result, the retentate generated in a hybrid NF–RO configuration is chemically distinct from that produced by RO alone, with implications for its downstream management and treatability. Despite its practical importance, this link between membrane operating conditions, solute fractionation, and retentate treatability has received limited attention.

Advanced oxidation processes (AOPs) provide a complementary route for conditioning concentrated membrane retentates. Among these, persulfate-based oxidation has attracted increasing attention due to the high oxidative potential and relative selectivity of sulfate radicals, particularly in saline environments where hydroxyl-radical-based processes may be inhibited by chloride and carbonate species. ¹³ In such matrices, chloride ions can modify radical pathways by interacting with sulfate radicals and generating chlorine-derived oxidizing species, which may alter oxidation mechanisms and influence overall process efficiency. These effects highlight the need for robust activation strategies that remain effective in high-salinity conditions. In this context, thermal activation of persulfate (TAP) is especially attractive for integration with membrane systems, as it avoids additional chemical activators or irradiation units and can be applied to small, highly concentrated volumes. In industrial settings, the availability of low-grade waste heat further enables thermal integration, thereby improving the overall energetic profile of the treatment train.

Rather than being treated solely as a water purification step, membrane separation plays a decisive role in defining the physicochemical characteristics and downstream manageability of the residual concentrate. The present work establishes a process-level link between membrane operating pressure and the efficiency of subsequent TAP oxidation, moving beyond conventional studies that consider membrane separation and concentrate treatment as independent units. A hybrid NF–RO–TAP approach is investigated using biologically treated textile wastewater as a representative industrial case in a semi-pilot-scale configuration. NF and RO were operated under different pressures to evaluate how operating conditions govern permeate quality, water recovery, energy demand, and retentate composition. The most concentrated residual stream was then subjected to TAP oxidation as a targeted post-treatment step, enabling an integrated assessment of membrane performance and retentate conditioning within a unified, process-oriented framework for enhanced water reuse and concentrate minimization.

Materials and methods

Chemicals

All chemicals used in this study were of analytical grade and were used without further purification. Sodium persulfate (Na₂S₂O₈) was supplied by Panreac (Spain). Iron(II) chloride tetrahydrate (FeCl₂·4H₂O), sodium chloride (NaCl), sodium carbonate (Na₂CO₃), potassium iodide (KI), sulphuric acid (H₂SO₄), and sodium bicarbonate (NaHCO₃) were purchased from Sigma-Aldrich (USA).

Wastewater characterization

The wastewater was collected from a textile manufacturing facility located in central Portugal, downstream of the biological treatment stage. Upon arrival at the laboratory, the samples were analyzed for pH, carbon and nitrogen fractions, chloride, sulfate, and electrical conductivity (EC). The wastewater was stored at 4°C and used within 5 days to ensure matrix stability before experimentation. The main physicochemical characteristics of the biologically pretreated effluent are summarized in Table 1.

Table 1.

Physicochemical characteristics of the biologically pretreated textile wastewater.

Parameter	Mean value (± SD)
Chemical oxygen demand (mg L−1)	116 ± 4
Total carbon (mg L−1)	77 ± 2
Total organic carbon (mg L−1)	23 ± 1
Inorganic carbon (mg L−1)	54 ± 2
Total nitrogen (mg L−1)	4 ± 1
Cl− (mg L−1)	14 ± 1
SO42− (mg L−1)	132 ± 3
pH	7.8 ± 0.2
EC (µS cm−1)	830 ± 8

Experimental setup of membrane modules

The membrane experiments were performed in a semi-pilot-scale filtration system manufactured by Apria Systems S.L. (Spain), operated in cross-flow mode. The system was equipped with NF and RO modules housing commercial spiral-wound polyamide membranes. The NF stage employed a Hydranautics ESNA1-LF2-LD4-4040 membrane with an active surface area of 7.41 m², while the RO stage used a KeenSen ULP-2540 polyamide membrane with an active area of 2.5 m². All experiments were carried out using 100 liters of biologically pretreated textile wastewater as feed. Filtration runs were performed in batch mode with retentate recirculation, and the transmembrane pressure was adjusted to 4−10 bar for both separation stages. It should be noted that this pressure range was selected to enable a direct comparison between NF and RO under similar operating conditions and does not necessarily reflect typical industrial RO pressures.

Persulfate treatment of the NF retentate

The TAP process was applied to the NF retentate in batch mode to degrade refractory organic matter. Experiments were carried out in 100 mL borosilicate reactors placed in a thermostatic water bath under constant stirring at 250 rpm. A Box–Behnken design (BBD) was employed to evaluate the influence of temperature (T), persulfate (PS) dosage, and Fe(II) concentration on total organic carbon (TOC) removal. PS dosage was expressed as the percentage of the stoichiometric requirement, taken as 12 g PS g⁻¹ chemical oxygen demand (COD). Reaction time was fixed at 1 h for all experiments. The experimental matrix and corresponding results are presented in Table 2. The X_TOC value corresponds to the remaining TOC concentration after PS activation, relative to an initial TOC concentration of 412 ± 8 mg L⁻¹ in the retentate. Response surface methodology was used to evaluate variable interactions and identify the optimal operating conditions for retentate treatment.

Table 2.

Box–Behnken response surface design matrix and experimental results.

Run	A: T (°C)	B: PS (%)	C: Fe (II) (mg L−1)	XTOC (mg L−1)
1	60	100	10	177.69
2	30	100	40	77.71
3	60	20	10	113.95
4	60	20	70	126.36
5	60	60	40	106.91
6	60	100	70	105.71
7	30	20	40	106.09
8	90	60	10	145.11
9	30	60	10	16.65
10	90	60	70	208.84
11	90	100	40	32.92
12	60	60	40	230.76
13	90	20	40	282.01
14	60	60	40	207.48
15	30	60	70	65.43

Analytical methods and calculations

TOC and total nitrogen (TN) were quantified using a TOC-L analyzer equipped with a TNM module (Shimadzu). COD was determined using Hach LC1400 kits and measured with a Hach DR3900 spectrophotometer. Sulfate and chloride concentrations were analyzed by ion chromatography using a Metrohm 883 Basic IC Plus system equipped with a conductivity detector and a Metrosep A Supp 5 column (250 mm × 4 mm). The mobile phase consisted of an aqueous solution of 3.2 mM Na₂CO₃ and 1.0 mM NaHCO₃, operated at a flow rate of 0.7 mL min⁻¹. Calcium concentration was determined by atomic absorption spectrophotometry using a Shimadzu AA-7000 instrument.

Residual PS concentration after oxidation experiments was determined by a colorimetric method described by Liang et al. ¹⁴ Solution pH and EC were measured using a Crison GLP 21+ pH meter and a Hach sensION+ EC7 conductivity meter, respectively. UV-Vis absorbance measurements were performed with an Agilent Cary 60 spectrophotometer.

The specific energy consumption of the membrane operation (SEC, kWh m⁻³) was determined from plant measurements according to

$$SEC=\frac{\mathrm{EP}}{Q_p}$$

(1)

where EP is the steady-state electrical power of the NF or RO unit (kW), and Q_p is the permeate flow rate (m³ h⁻¹). ¹⁵

The permeate flux (J, L m⁻² h⁻¹) was calculated according to

$$J=\frac{Q_p}{A}$$

(2)

where Qp is the permeate flow rate (L h⁻¹) and A is the effective membrane surface area (m²).

The apparent hydraulic permeability (Lp, L m⁻² h⁻¹ bar⁻¹) was calculated from the ratio between permeate flux and the applied transmembrane pressure (ΔP, bar), according to ¹⁶

$$L_p=\frac{J}{\Delta P}$$

(3)

As membrane experiments were conducted in batch mode with recirculation of the retentate, water recovery was calculated on a volumetric basis rather than using instantaneous flow ratios. The cumulative water recovery (R, %) was calculated according to

$$R=\frac{V_p}{V_0}\times 100$$

(4)

where Vp is the cumulative permeate volume collected during the experiment (L) and V₀ is the initial feed volume in the tank (L).

The tendency for gypsum (CaSO₄·2H₂O) scaling was evaluated through the gypsum saturation index (SI), calculated according to the classical thermodynamic definition

$$SI={\log }_{10}\left(\frac{\mathrm{IAP}}{K_{sp}}\right)$$

(5)

where IAP is the ion activity product, approximated in this study by the product of measured molar concentrations of Ca²⁺ and SO₄²⁻, and K_sp is the solubility product constant of gypsum at 25°C, taken from the literature. Negative SI values indicate undersaturation conditions, a value of zero equates to equilibrium, and positive values indicate supersaturation with potential precipitation.^17,18

Results and discussion

NF: hydraulic and energetic response

NF performance is strongly governed by the applied transmembrane pressure, which directly controls permeate flux, water recovery, and specific energy consumption. Understanding how these parameters evolve with pressure is essential for identifying operating conditions that maximize water productivity while minimizing energetic penalties in high-recovery reuse schemes. Figure 1 illustrates the evolution of permeate flux, recovery rate, and energy demand as a function of operating pressure, under the experimental conditions evaluated. Increasing pressure enhanced permeate flux and recovery, but this improvement was accompanied by a nonlinear increase in specific energy consumption, reflecting the growing hydraulic and osmotic resistance of the system. These opposing trends highlight an inherent trade-off between volumetric productivity and energetic efficiency. Electrical power demand rose from 0.38 kW at 4 bar to 1.32 kW at 10 bar (Figure 1a), reflecting the increasing hydraulic work required to sustain permeation at higher transmembrane pressures. These values are in the same range as those usually reported for NF treating industrial and saline wastewater, where pumping and recirculation use most of the energy consumed by the overall process.^19-21

Figure 1.

Hydraulic performance of the NF stage: (a) electrical power versus pressure; (b) permeate flow versus pressure; (c) electrical power versus permeate flow; and (d) permeate flux versus SEC.

Permeate flux increased quasi-linearly with pressure over the 4–10 bar range (Figure 1b), indicating that water transport was predominantly pressure-driven within the investigated operating window. Comparable pressure–flux relationships have been reported for NF treatment of textile effluents using commercial and commercial-equivalent polyamide membranes operated under similar pressure ranges.^2,7

The relationship between permeate production and power consumption (Figure 1c) demonstrates that a higher energy input results in a greater water throughput. However, normalization by specific energy consumption (Figure 1d) reveals a systematic decline in energetic efficiency with increasing pressure. Energy-normalized flux decreased from approximately 25 L m⁻² h⁻¹ at 4 bar to about 6 L m⁻² h⁻¹ at 10 bar, corresponding to a reduction exceeding 70% in energetic productivity. This behavior reflects diminishing marginal gains in permeate production as pressure increases. Such trends are consistent with experimental studies on NF applied to textile wastewater, which show that increasing transmembrane pressure primarily intensifies hydraulic energy demand without proportional improvements in effective water production, even when separation performance remains high. ⁸ Under these conditions, pressure becomes an increasingly inefficient control variable for enhancing process performance. A similar need to identify an optimal operating window rather than simply maximizing the driving force has also been reported in forward-osmosis studies, where temperature and concentration gradients strongly affected water flux and overall process efficiency.^22,23

The obtained results identify an operational window between 6 and 8 bar that provides a more favorable balance between permeate flux and energy consumption under the investigated conditions.

RO: hydraulic and energetic response

The hydraulic and energetic response of the RO stage under polishing conditions is summarized in Figure 2. Operating pressure was deliberately restricted to a maximum of 10 bar in the pilot-scale system to adhere to membrane specifications and the hydraulic constraints of the experimental setup. Under these conditions, permeate production reflects the combined influence of membrane hydraulic resistance and the increasing relevance of osmotic pressure as recovery progresses, a behavior commonly reported for RO applied to biologically treated textile effluents under reuse-oriented operation.^10,24

Figure 2.

Hydraulic performance of the RO stage: (a) electrical power versus pressure; (b) permeate flow versus pressure; (c) electrical power versus permeate flow; and (d) permeate flux versus SEC.

As illustrated in Figure 2(a), the electrical power demand experienced a significant increase as the operating pressure increased, increasing from approximately 0.5 kW at 5 bar to nearly 1.2 kW at 10 bar. Over the same pressure range, permeate flow increased from about 30 L h⁻¹ to approximately 85 L h⁻¹ (Figure 2b). As a result, the increase in permeate production was not directly proportional to the additional energy input, which implies that hydraulic returns are decreasing at higher pressures. In pilot-scale RO systems that treat biologically treated textile wastewater, similar pressure–energy trade-offs have been reported, particularly under high-recovery and recirculating conditions. ²⁵

This disparity is further demonstrated by the relationship between power consumption and permeate production (Figure 2c), where small power increases led to ever-tinier gains in permeate flow. When permeate flux is normalized by specific energy consumption (Figure 2d), energetic productivity decreased from approximately 11 L m⁻² h⁻¹ at lower energy demand to less than 5 L m⁻² h⁻¹ at the highest values tested. Studies on the reuse of textile wastewaters based on RO have shown comparable decreases in energy-normalized performance, which is indicative of the inherent energetic penalty of operating near the upper pressure limits in polishing configurations.^9,24

The obtained results indicate that within the investigated pressure range, increasing operating pressure in the RO stage leads to a disproportionate increase in energy demand relative to permeate production. This hydraulic–energetic response is consistent with previous pilot-scale studies on textile wastewater reuse, confirming that RO performance under polishing conditions is primarily constrained by energetic efficiency rather than by hydraulic productivity. ²⁶ Accordingly, process performance should be evaluated not based on flux improvement alone, but through the combined balance between solute rejection, water recovery, and specific energy consumption. ²⁷

Permeate quality

Figure 3 presents pressure-resolved permeate quality maps for NF and RO. Permeate quality was primarily governed by membrane selectivity, with operating pressure exerting distinct effects depending on ion valence. In NF, permeate EC increased from 46 µS cm⁻¹ at 4 bar to 56 µS cm⁻¹ at 6 bar, followed by a more pronounced increase at higher pressures (78–81 µS cm⁻¹ at 8–10 bar). This behavior reflects the selective nature of NF, where increasing transmembrane pressure enhances the convective transport of weakly rejected monovalent ions rather than uniformly improving salt rejection. In contrast, RO permeates exhibited consistently low EC with limited pressure sensitivity, decreasing from 26 µS cm⁻¹ at 5 bar to 8–16 µS cm⁻¹ at 8–10 bar, consistent with the strong and pressure-independent salt rejection characteristic of RO membranes treating textile effluents. ⁶

Figure 3.

Permeate quality maps for NF and RO at different applied pressures.

Chloride concentration followed the same trend. In NF, permeate Cl⁻ concentration decreased from 4.0 mg L⁻¹ at 4 bar to 2.9 mg L⁻¹ at 6 bar, but remained within the 2.3–2.9 mg L⁻¹ range at higher pressures, indicating that increasing pressure provides limited control over monovalent ion passage. In RO, chloride concentrations remained below the reporting threshold across all investigated pressures (⩽ 0.5 mg L⁻¹), confirming stable and pressure-insensitive rejection under polishing conditions.

Sulfate concentration demonstrated a significantly distinct response. As pressure increased in NF, SO₄²⁻ concentration dropped considerably, moving from 2.0–3.0 mg L⁻¹ at 4–6 bar to values below the reporting threshold (⩽ 0.5 mg L⁻¹ at 8–10 bar), corresponding to sulfate removals exceeding 80–90% at higher pressures. This trend is consistent with the dominant role of electrostatic exclusion and size-based hindrance in governing divalent ion retention in NF membranes. In RO, sulfate concentrations remained below the reporting threshold across all tested pressures, corresponding to removals above 95–99% within analytical resolution.

From a water reuse perspective, these results indicate that 6 bar represents a balanced operating condition for NF, combining moderate conductivity (56 µS cm⁻¹) with effective attenuation of divalent ions, while avoiding the progressive increase in salinity observed at higher pressures. Operation beyond this pressure improves sulfate rejection but does not enhance overall salinity control, as permeate composition becomes increasingly dominated by highly mobile monovalent ions. RO provides consistently low EC and near-complete removal of both monovalent and divalent ions, supporting reuse applications with more stringent water quality requirements. Overall, permeate quality trends confirm that membrane selection and pressure optimization should be guided by the targeted reuse category rather than by pressure increase alone, as also highlighted in recent textile water-reuse studies where membrane operation is aligned with both contaminant removal targets and the intended reuse pathway of the recovered water or process chemicals.^9,28,29

Combined NF–RO process

The combined NF–RO configuration resulted in an improved balance between hydraulic performance, salt rejection, and scaling control compared with standalone membrane operation. As shown in Figure 4(a), the integrated system operated at intermediate Q_p values relative to NF and RO, reflecting the hydraulic coupling of both stages and the reduction of osmotic constraints at the RO inlet following NF pretreatment. While NF exhibited the highest permeate flow rate, the NF–RO process achieved an overall recovery approximately 15–20% higher than standalone RO, in agreement with previous studies reporting improved operational stability in integrated membrane systems.^9,30

Figure 4.

Hydraulic and separation performance of NF, RO, and combined NF–RO processes: (a) permeate flow rate and cumulative water recovery; (b) apparent hydraulic permeability; (c) retentate EC and ionic composition; and (d) gypsum saturation index as a function of Ca²⁺ concentration.

Despite the lower Q_p compared with NF alone, the overall recovery of the NF–RO process exceeded that of standalone RO, enabling operation at higher effective recovery without loss of stability. This behavior is consistent with previous studies showing that selective removal of multivalent ions in pretreatment stages allows RO systems to operate at higher recovery before reaching hydraulic or osmotic constraints. ³¹

The effect of NF pretreatment on membrane transport properties is reflected in the permeability data (Figure 4b). The NF–RO configuration exhibited higher apparent hydraulic permeability than standalone RO, indicating improved water transport per unit pressure. This enhancement has been widely associated with reduced concentration polarization and lower fouling propensity when RO is preceded by a selective separation step, as reported for integrated pretreatment–RO systems treating textile and industrial effluents. ²⁶

The effect of the combined process on salt distribution is evidenced by the retentate composition (Figure 4c). EC in the NF–RO retentate decreased by approximately 40–45% relative to NF alone, driven primarily by a marked reduction in sulfate concentration (around 55–60%), while chloride exhibited limited removal. This behavior is consistent with the intrinsic selectivity of NF membranes toward divalent anions and highlights the preferential rejection of sulfate over chloride as a key advantage of NF pretreatment. ³²

The implications of sulfate removal are confirmed by the gypsum saturation analysis (Figure 4d). In the standalone RO, the saturation index increased with calcium concentration and reached supersaturation conditions, indicating a high risk of gypsum precipitation. In contrast, the NF–RO configuration maintained negative saturation index values across the full calcium concentration range evaluated, demonstrating effective suppression of gypsum formation. This behavior is directly linked to sulfate depletion in the NF step, which decouples calcium concentration from sulfate availability and has been identified as a primary strategy for gypsum scaling control in membrane systems.^17,18

The combined NF–RO process offers a technically sound approach to improve RO performance, facilitating higher recovery, improved permeability, selective sulfate removal, and effective control of scaling phenomena in sulfate-rich wastewaters. The treatment strategy transitions from selective separation to chemical transformation under these circumstances, with the NF retentate serving as the appropriate stream for advanced oxidation. From an MLD perspective, these results indicate that the combined NF–RO configuration enables high water recovery while maintaining acceptable energy demand and controlling scaling risks through selective sulfate removal. At the same time, the concentration of refractory organics and salts into a reduced retentate volume creates a suitable stream for targeted post-treatment, highlighting the relevance of process integration for effective concentrate management. This integrated perspective is consistent with recent textile wastewater studies in which membrane separation is coupled with biological or osmotic processes, not only to improve permeate quality but also to better control the fate of retained contaminants and dissolved salts.^22,33 Although the present study was conducted under controlled batch conditions, the observed trends in permeability and scaling propensity provide useful insight into the expected fouling behavior under continuous operation. Nevertheless, long-term fouling dynamics and the effectiveness of cleaning strategies require further investigation under steady-state conditions representative of industrial practice.

Post-treatment of NF retentate by TAP oxidation

The NF retentate, representing the most concentrated and recalcitrant fraction of the membrane treatment, was subjected to post-treatment by TAP oxidation. This step aimed to degrade persistent dissolved organic matter and improve the overall manageability of the concentrate within an MLD framework. To evaluate the influence of operating conditions on oxidation performance, the effects of temperature, persulfate stoichiometric dosage (PSD), and Fe(II) concentration were investigated using BBD. The resulting experimental data (Table 2) were fitted to a second-order polynomial model describing the relationship between the selected variables and the observed response:

$$\begin{array}{l}\mathrm{TOC}\mathrm{removal}=16.95142+0.788493\times T-1.38055\\ \times \mathrm{PSD}-0.686521\times \mathrm{Fe}+0.002806\\ \times T\times \mathrm{PSD}-0.002942\times T\times \mathrm{Fe}+0.004202\\ \times \mathrm{PSD}\times \mathrm{Fe}-0.018575\times T^2-0.012818\times {\mathrm{PSD}}^2\\ +0.012818\times {\mathrm{Fe}}^2\end{array}$$

(6)

The analysis of variance (ANOVA) analysis (Table 3) showed a high adjusted coefficient of determination (adjusted R² = 0.9742) and a statistically significant model (F = 59.73, p = 0.0001), demonstrating good agreement between experimental and predicted values and consistency with results reported in the literature.^34,35 Although the lack-of-fit test was statistically significant (p = 0.0015), this result should be interpreted with caution due to the very low pure error associated with the limited number of replicate runs. Under such conditions, the lack-of-fit test becomes highly sensitive to small deviations between predicted and experimental values.

Table 3.

ANOVA analysis for TOC removal.

	Sum of squares	dF	Mean	F-values	p-values
Model	78,208.48	9	8689.83	59.73	0.0001
A	67,787.24	1	67,787.24	465.94	< 0.0001
B	5944.40	1	5944.40	40.86	0.0014
C	1235.43	1	1235.43	8.49	0.0332
AB	45.38	1	45.38	0.3119	0.6006
AC	28.03	1	28.03	0.1927	0.6790
BC	101.89	1	101.89	0.7003	0.4408
A2	1031.41	1	1031.41	7.09	0.0447
B2	1954.38	1	1954.38	13.43	0.0145
C2	491.08	1	491.08	3.38	0.1256
Residual	727.43	5	145.49
Lack of fit	726.68	3	242.23	647.66	0.0015
Pure error	0.7480	2	0.3740
Corr Total	78,935.91	14

The individual contributions of each variable dominate the system response, as evidenced by the statistical significance of the linear effects of temperature, persulfate dosage, and Fe(II) concentration, while the interaction terms did not. Similar to the conclusions reached by Yabalak, ³⁶ it was found that the nonlinear behavior within the experimental domain is illustrated by the significance of the quadratic terms associated with temperature and persulfate dosage.

The response surfaces presented in Figure 5 illustrate the effect of the operating variables on TOC removal. Residual TOC values are consistently high at 30°C throughout the entire temperature–persulfate plane (Figure 5a), which corresponds to removals that are typically less than 100 mg L⁻¹. Increasing PS dosage under these conditions produces only marginal changes in residual TOC. At higher temperatures, a marked decrease in residual TOC is observed, particularly at 90°C, where increasing PS dosage leads to removals exceeding 250 mg L⁻¹ under the most favorable conditions.

Figure 5.

Response surface plots (left) and contour plots (right) for the optimal conditions for the following variables: (a) temperature and PSD, with variation of Fe(II) concentration; (b) temperature and Fe(II) concentration, with variation of PSD; (c) Fe(II) concentration and PSD, with variation of temperature.

The temperature–Fe(II) response surface (Figure 5b) exhibits a similarly strong temperature-dependent behavior. At low temperatures, variations in Fe(II) concentration have little effect on TOC removal, indicating limited catalytic contribution under weak thermal activation. This behavior can be attributed to the Fe(II)-mediated activation mechanism: while Fe(II) promotes the generation of SO₄^–• radicals, these radicals are simultaneously consumed during the oxidation of Fe²⁺ to Fe³⁺, resulting in a reduced availability of reactive species for pollutant degradation. ³⁷ At elevated temperatures, however, increasing Fe(II) concentration leads to moderate yet consistent additional TOC removal, typically amounting to several tens of milligrams per liter. This trend indicates that the contribution of Fe(II) becomes relevant only when sufficient thermal activation ensures sustained radical generation. In addition, the persulfate–Fe(II) surface (Figure 5c) suggests that variations in chemical inputs are insufficient to mitigate unfavorable thermal conditions. Even at high PS dosages and Fe(II) concentrations, residual TOC remains elevated at low temperatures. Significant reductions in TOC are only observed in the high-temperature regime, which is consistent with the limited statistical significance of PS dosage and Fe(II) concentration in relation to temperature. Low temperatures limit the rate of PS decomposition, which is insufficient to maintain an effective radical flux irrespective of oxidant or catalyst concentration. Under these effects, radical flux is secondary and highly temperature-dependent. The presence of Fe(II) promotes greater Fe(III) concentration by facilitating electron-transfer pathways. Incremental thermal activation of Fe(II) radical generation decreases in these circumstances, suggesting that thermal activation controls radical generation.

These patterns align with the idea that activation energy limits PS activation. The statistical relevance of each factor within the response surface model was further examined through linear, quadratic, and interaction terms in Figure 6(a). This T-value ranks the standardized visual hierarchy of temperature (A) overwhelmingly dominated the system response, accounting for more than half of the total effect, which is consistent with the high activation energy required for PS decomposition. PS dosage (B) and the quadratic term B² also contributed significantly, although to a lesser extent, followed by the Fe(II) concentration (C), quadratic terms A² and C², indicating nonlinear behavior at higher temperatures and oxidant levels. In contrast, interaction terms (AB, AC, BC) exhibited negligible influence on TOC removal, confirming that the process is primarily governed by the individual effects of the variables rather than their combined interactions.

Figure 6.

(a) Pareto diagram of the model terms and (b) plot of predicted TOC removal versus experimental TOC removal.

Figure 6(b) presents the correlation between predicted and experimental TOC removal values. The close clustering of data points around the 1:1 line demonstrates the excellent predictability of the fitted polynomial model. Only minor deviations were observed, indicating limited dispersion and confirming that the model is robust and suitable for process optimization. This strong agreement between predicted and experimental values supports the adequacy of the quadratic model for capturing the main response trends, despite the statistically significant lack-of-fit observed in the ANOVA analysis.

The model determined that the ideal operating parameters were 90°C, 100% PSD, and 10 mg L⁻¹ Fe(II) based on the desire function and response surface optimization. This would result in a predicted TOC removal of 265.35 mg L⁻¹ from an initial concentration of 412 mg L⁻¹, corresponding to approximately 64% removal. Balcik-Canbolat et al. reported a maximum TOC removal of 80.8% for Fenton treatment of a dye- and sulfate-rich textile NF membrane concentrate. ³⁸ Although such highly reactive systems can achieve higher degrees of mineralization, the removal observed in the present study remains noteworthy, particularly given the high salinity and refractory nature of the NF retentate. In highly saline concentrates, the efficiency of AOPs is often limited by matrix effects such as radical scavenging, high ionic strength, and mass-transfer constraints, which restrict the extent of mineralization.^13,39 For this reason, recent studies frame oxidation not as a standalone solution, but as a conditioning step within integrated treatment schemes. ⁴⁰ This broader process-integration perspective also applies to other textile wastewater polishing approaches, such as adsorption-based post-treatment, which may significantly reduce dye content and part of the residual pollutant load, yet still require further treatment to fully meet discharge or reuse targets. ⁴¹ Within this context, TAP can be understood as a targeted conditioning step that improves the manageability of the concentrate.

Optimizing surface conditions achieves a balance between oxidative efficiency and reagent cost, particularly for Fe(II), where increasing dosages did not notably enhance mineralization. Since the number of pollutants in the stream ultimately dictates the MLD system’s operational and environmental footprint, the relatively small TOC reduction, which is typical of very saline retentates, represents a significant improvement in wastewater quality. Residual PS was detected following the reaction, but the concentrations were low, indicating that most of the oxidant was consumed during the process. PS decomposition produces sulfate ions, which remain in solution and contribute to the overall salinity of the treated stream. Given the relatively high PS dosages applied, this contribution may be significant. However, as the TAP process was applied to the NF retentate, already characterized by high EC and elevated salt concentrations, this increase occurs within a highly saline stream intended for further management within an MLD framework. In this context, sulfate formation should be regarded as part of the concentrate conditioning rather than as a secondary pollution pathway.

The findings demonstrate that TAP oxidation can effectively complement membrane processes by selectively targeting the most recalcitrant fraction of the wastewater (the equilibrium between chemical input and thermal activation needed to accomplish efficient oxidation during membrane treatment), thereby providing a technically viable and energy-conscious conditioning step that reduces the recalcitrance of the concentrate and improves its suitability for subsequent management within MLD frameworks. It should be noted that TAP oxidation is not intended to eliminate the need for downstream treatment, but rather to transform the retentate into a more manageable stream by reducing its recalcitrance and improving its overall treatability.

From an implementation perspective, the proposed hybrid NF–RO–TAP configuration can be integrated into existing textile wastewater treatment schemes as a downstream polishing and retentate conditioning stage following conventional biological treatment. The membrane units can be incorporated into existing treatment lines, while the TAP process is applied selectively to the generated retentate stream, thereby minimizing the required treatment volume. This configuration is inherently modular and can therefore be adapted to existing facilities with limited modifications, enabling gradual implementation without major disruption to plant operation. The modularity of such treatment trains is consistent with recent membrane-based textile wastewater studies that combine separation, biological conversion, or draw-solution-assisted reuse in staged layouts rather than in a single standalone unit.^33,42 In terms of footprint, the use of membrane processes combined with targeted retentate treatment supports a relatively compact system design compared with conventional treatment trains. Nevertheless, a detailed assessment of space requirements and site-specific retrofit constraints would be necessary for full-scale implementation.

Although thermal activation was achieved under controlled laboratory conditions using a thermostatic water bath, the associated energy demand was not quantified in this study. In practical applications, the required temperature range (up to 90°C) could be supplied by low-grade waste heat or steam streams commonly available in textile processing facilities, thereby reducing the net energy penalty of the oxidation step. Still, a comprehensive evaluation of the overall process cost would require consideration of chemical consumption, thermal energy requirements, and specific operational costs, including membrane replacement and cleaning, and pumping/maintenance. These factors may influence the economic feasibility of the treatment and should be addressed in future studies aimed at process scale-up and techno-economic evaluation.

Conclusions

This study has demonstrated that the integration of pressure-driven membrane processes with persulfate-based advanced oxidation provides an effective and energy-efficient strategy for treating biologically pretreated textile wastewater within an MLD-oriented framework. Both NF and RO produced permeates with very low conductivity and near-complete removal of dissolved ions, while also concentrating the refractory organic matter and salts in the retentate streams. Among the NF operating conditions evaluated, 6 bar was identified as a balanced compromise, delivering high salt rejection, stable permeate flux, and comparatively lower energy demand.

Both membrane systems generated retentates with high salinity and refractory organic matter accumulation, highlighting the need for a complementary treatment step. TAP oxidation, applied to the NF retentate, enabled substantial removal of the concentrated organic load. Response surface methodology enabled the identification of the temperature factor as the dominant operational parameter, while both PS dosage and Fe(II) concentration wielded secondary effects. The optimal conditions (100% PSD, 10 mg L⁻¹ Fe(II), 90°C) achieved a predicted TOC removal of ~64%, significantly reducing the organic load of the concentrate.

The performance achieved in this study is consistent with key MLD targets, particularly in terms of producing low-salinity permeate at controlled specific energy consumption and reducing the residual stream to a concentrated fraction amenable to targeted post-treatment. The NF–RO configuration concentrates salts and refractory organics into a reduced-volume retentate, while subsequent TAP oxidation enhances its manageability, demonstrating how integrated process design can simultaneously support water reuse and concentrate minimization in textile systems. The possibility of coupling TAP with available low-grade waste heat or steam streams further strengthens the practical feasibility of the proposed approach by mitigating the additional energy demand associated with thermal activation. Nevertheless, a comprehensive assessment of operational costs, including chemical consumption and thermal energy requirements, remains necessary to fully evaluate the economic viability of the proposed system on a larger scale. It should be noted that the membrane experiments were conducted in batch mode and therefore do not capture the progressive fouling behavior and cleaning requirements associated with long-term continuous operation. Future work should address membrane stability, long-term flux evolution, fouling control strategies, and the development of appropriate cleaning protocols under realistic operating conditions to support scale-up and industrial implementation.