Dynamic evolution of ultrafiltration membrane fouling affected by S(IV)–Fe(II)-activated permanganate pretreatment during a long-term operation

Due to stringent drinking water quality standards and the limited availability of high-quality water sources, ultrafiltration (UF) technology has gained recognition as an exceptionally effective solution for safe drinking water production. This recognition is attributed to its ability to efficiently retain particles and microorganisms, along with its relatively lower investment and operational costs. Nevertheless, membrane fouling has emerged as a significant hurdle, impeding its widespread adoption in drinking water production (Gao et al. 2011; Zhang et al. 2023). Membrane fouling not only diminishes water permeability but also leads to increased operating expenses and reduced membrane lifespan (Peters et al. 2021). Consequently, the effective management of membrane foulant in UF systems has become a prominent concern in the field of UF drinking water treatment in recent years.

Presently, methods for mitigating UF membrane fouling encompass membrane surface modification (Song et al. 2012), membrane pretreatment (Liu et al. 2023b), optimization of operational parameters (Taheri et al. 2019), and the application of electromagnetic fields (Rouina et al. 2016). Among these methods, membrane pretreatment (e.g., coagulation, adsorption, and oxidation) is the most prevalent approach for controlling membrane fouling in practical production (Ma et al. 2018; Yu et al. 2018; Li et al. 2020; Xing et al. 2021; Yan et al. 2021; Liu et al. 2022). Gao et al. (2011) analyzed the effectiveness of coagulation, adsorption, and oxidation pretreatment, in addition to optimizing operational aspects such as mode of operation, rinsing procedures, and chemical cleaning, to manage membrane fouling. They highlighted certain limitations when using coagulation, oxidation, and adsorption pretreatment methods individually. For example, coagulation pretreatment demonstrates notably low efficiency in removing algal organic matter (AOM), especially extracellular organic matter, which results in residual AOM that can lead to substantial membrane fouling (Yan et al. 2017). Adsorption pretreatment, such as powdered activated carbon adsorption, met with controversy due to the potentially complex interactions between powdered carbon and organic contaminants which aggravated membrane fouling (Shao et al. 2016, 2017). Additionally, pre-oxidation may exacerbate membrane fouling when dealing with algae-containing water by causing the release of organic substances and toxins from algal cells (Wert et al. 2014; Qu et al. 2015).

Recognizing the limitations of traditional pretreatment technologies and the complexity of raw water quality, researchers have increasingly explored combined pretreatment processes, such as oxidation–coagulation, adsorption–coagulation, and oxidation–adsorption (Bu et al. 2019; Xing et al. 2019a; Cheng et al. 2021a, 2021b). Xing et al. (2019b) introduced Fe(II) into the UV/chlorine pretreatment technology system, upgrading the system to a purification system integrating flocculation, UV/Fe(II), and chlorine/Fe(II), which resulted in a decrease in the resistance to irreversible membrane fouling of subsequent UF membrane by >30%. Sarasidis et al. (2017) coated powdered activated carbon with Fe(II) and harnessed the oxidation of Fe(II)/H₂O₂ along with the adsorption capabilities of powdered activated carbon to enhance the removal of irreversible membrane contaminants. In a continuous-flow small-scale trial spanning 3 h, they successfully achieved consistent removal of irreversible membrane contaminants, maintaining a stable water production flux. Chang et al. (2020) achieved the mitigation of irreversible membrane fouling by reinforcing free radical generation through the addition of Fe(II) and introducing Fe(III) flocculation in the UV/persulfate system. These findings suggest that the combination of oxidation or adsorption with flocculation can enhance the removal of irreversible membrane contaminants before the membrane.

Combined pretreatment processes, such as permanganate (PM) oxidation or activated PM oxidation along with coagulation (PM-coagulation), have recently garnered attention for their impressive performance in algae removal, pollutant degradation, and mitigating membrane fouling (Qi et al. 2016; Yan et al. 2017; Liu et al. 2023a, 2023b). This is primarily due to their ability to simultaneously employ oxidation, adsorption, and coagulation properties. However, these processes typically involve two stages. Inspired by this, we introduce the S(IV)–Fe(II) binary-activated PM pretreatment technology (Liu et al. 2023c). The S(IV)–Fe(II)/PM system effectively combines oxidation, adsorption, and coagulation by generating Fe(III) (serving as a flocculating agent), newly formed MnO₂ (serving as an adsorbent and coagulation aid), and oxidizing agents (activated manganese species and free radicals) in situ. Moreover, it simplifies the process into a one-stage operation. The previous experimental results demonstrated that the S(IV)–Fe(II)/PM pretreatment reduced the amount of membrane contaminants in the source water and mitigated the subsequent development of the membrane surface fouling layer by regulating the deposition environment at the water-membrane interface, which initially proved the feasibility of S(IV)–Fe(II)/PM as a membrane pretreatment technology. However, the long-term control effect of S(IV)–Fe(II)/PM pretreatment on UF membrane fouling needs to be further verified.

In this study, a continuous-flow experimental system of S(IV)–Fe(II)/PM pretreatment-UF process was established using real lake-type raw water as the test water source. S(IV)–Fe(II)/PM pretreatment was compared with a conventional Al(III) coagulation pretreatment, commonly employed in waterworks. This work aims to (i) assess the fouling alleviation potential of S(IV)–Fe(II)/PM over a long operation period (32 days), (ii) analyze the dynamic evolution in UF membrane fouling behavior, and (iii) quantify the individual and combined contributions of organic, inorganic, and biological contaminants via physicochemical and biological characterization and assisted with statistical techniques. The findings of this study will offer both technical and theoretical support for the practical implementation of S(IV)–Fe(II)/PM as UF pretreatment in drinking water treatment.

Potassium permanganate, ferrous chloride, sodium chloride, sodium hypochlorite, sodium hydroxide and hydrochloric acid were purchased from Sinopharm Chemical Reagents Co. Polymeric aluminum chloride (PAC), anhydrous sodium sulfite and anhydrous ethanol were purchased from Aladdin Reagent (Shanghai) Co. Unless otherwise stated, deionized water (Smart-N, Heal Force) was used to prepare solutions for this study. Sodium hydroxide, hydrochloric acid, and sodium hypochlorite, used for chemical cleaning, were ordered from J&K Scientific Ltd (Shanghai).

Raw water samples were taken from a lake-type reservoir (March to April) in Zhejiang Province, China. The collected raw water was stored in a refrigerator at a temperature of 4 °C. The specific water quality parameters of the water samples are shown in Supplementary material, Table S1.

The UF membrane utilized in this study was a polyethersulfone (PES) hollow fiber membrane with a molecular weight cutoff of 10 kDa, procured from Hangzhou Yuanxiang Membrane Technology Co., featuring an effective filtration area of 200 cm². Prior to experimentation, the UF membrane underwent immersion in ultrapure water for a minimum of 24 h, with water replacement every 12 h. Following this, the membrane was pre-filtered using ultrapure water for 12 h to attain a stable flux. To mimic practical operational conditions, the UF membrane was operated at a flux of 10 L/(m²·h), with periodic adjustments to the water production rate facilitated by a peristaltic pump to maintain a consistent flux. Regularly cut quantities of membrane fibers from the UF modules were subjected to physicochemical and biological characterization, with the water temperature maintained at 25 °C through the use of a heating rod in the pool. To accelerate the evolution of distinct membrane fouling stages and minimize interference from backwashing on membrane surface characterization and measurements, no backwashing was employed in the UF unit. The experimental period spanned 32 days, and actual sampling occurred at intervals of 1, 2, 4, 7, 10, 14, 18, 25, and 32 days. The apparatus was equipped with a pressure gauge to transmit transmembrane differential pressure data to a computer in real time.

A Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR-MS, Bruker SolariX) was used to analyze the dissolved organic matter composition. The extraction and the fluorescence excitation–emission matrix (EEM) detection of organic matter in the contaminated layer on the membrane surface was carried out as follows: the cake layer was carefully scraped from the UF membrane and then mixed with 0.05% NaCl solution, the resulting mixture was ultrasonicated for 3 min and then filtered through a 0.45-μM membrane, and the post-filtered water was used to measure EEM of dissolved organic matter in it using an RF-5301 Shimadzu spectrofluorometer (excitation range: 220–500 nm, data interval: 5 nm; emission range: 200–500 nm, data interval: 2 nm), and parallel factor analysis and regional integration were carried out using the drEEM toolbox of MATLAB 9.11.

Heterotrophic plate count (HPC) measurement was used to quantify the bacteria in water. Laser scanning confocal microscopy (CLSM) was applied to analyze the bacteria on the membrane surface: after staining with SYTO9 and propidium iodide (LIVE/DEAD Biofilm Viability Kit, FilmTracer, Inc.), microbial fluorescence images were observed using CLSM (LSM780, Zeiss, Germany), observations were analyzed with the ZEN2010 software, and biomass content was calculated using the ImageJ add-on plug-in Comstat2 (Singhal et al. 2012). Scanning electron microscopy (SEM, Quanta FEG 650, FEI, USA) was used to observe the frontal and cross-sectional morphology of the UF membranes operated for different durations. X-ray spectroscopy (EDS) was used in conjunction with SEM to analyze the elemental species and content of contaminated layers on the surface of the UF membranes.

The tandem resistance model was used to calculate the total fouling resistance (R_t), the irreversible fouling resistance (R_ir), and the reversible fouling resistance (R_r) (Lin et al. 2009). The membrane fouling index (FI) was then calculated based on the tandem resistance model and assuming that the membrane fouling resistance is proportional to the amount of filtered water (Nguyen et al. 2011). The detailed calculation methods are described in Supplementary material, Text S1.

Variance Partitioning Analysis (VPA) is the process used to assess the impact of independent variables on the outcomes of a study by examining how much they contribute to changes in the dependent variable (Lin et al. 2019). The extent of this contribution can be measured using the sum of squared deviations, denoted as R², which quantifies the influence of the independent variables on the dependent variable. The contributions of organic, inorganic, and biological foulants to overall membrane fouling in the fouling process were analyzed. The detailed analysis methods are described in Supplementary material, Text S2.

Cluster analysis (CA) is a statistical method based on a large amount of data or samples to cluster the closest samples into one category based on the similarity or distance between the samples. In this study, UF membrane samples with different levels of fouling were categorized using CA (Liu et al. 2018), to achieve a scientific division of membrane fouling development stages. The detailed analysis methods are described in Supplementary material, Text S3.

Considering the absence of cleaning measures during operation, the thinner cake layer exhibited by the M1 membrane at 32 days suggests a weaker adhesion with the M1, allowing for detachment once the cake layer reaches a certain thickness. It can be verified by the almost identical appearance of the M1 membrane surface to that of an unused membrane after cleaning (Figure 4). The weak adhesion of the filter cake layer on the M1 membrane surface can be attributed to the S(IV)–Fe(II)/PM pretreatment resulting in cake layer composed of medium-uniform particle size particles and organic matter with a lower proportion of polysaccharides/proteins (Liu et al. 2023c). Such a cake layer is considered to have weak binding to the membrane surface (Meng et al. 2007).

The chemical composition of the cake layer plays a crucial role in determining the physicochemical properties of the cake layer, subsequently influencing the filtration performance of the cake layer-UF membrane composite and the accumulation of foulants within membrane pores. Therefore, it is imperative to analyze the dynamic changes in the organic, inorganic, and biological components of the cake layer. Consequently, the composition of the cake layer on the surfaces of M1 and M2 membranes was continuously monitored.

Notably, the M2 membrane showed a higher percentage of dead microorganisms (>40%) on its surface after 2 days of operation compared to the case of the first day (97%). Considering that Al(III) pretreatment cannot effectively remove microorganisms (Supplementary material, Figure S5(a)), it may mean that PAC coagulation pretreatment can reduce the concentration of C/N nutrients in the UF membrane influent (indexed by CHO and CHON fractions (Supplementary material, Figure S5(b)), thereby affecting the survival and reproduction of microorganisms on the membrane surface. Additionally, considering the distribution of microorganisms in the filter cake layer (Supplementary material, Figure S4), the masking effect of the filter cake layer on microorganisms possibly influences the availability of nutrients, which may affect the proportion of dead microorganisms.

In the later stage (18–32 days), the gap in the percentage of live microorganisms amount and dead microorganisms amount for between the M1 and M2 membranes narrowed, which may be attributed to the formation of a large pore structure membrane surface filter cake layer (Figure 4). This resulted in microorganisms entering the pores and remaining undetected.

As shown in Figure 8, concerning the individual contribution factors, the pollution contribution rates of F_Org to M1 and M2 membranes are 7.5 and 2.2%, respectively, while the contributions of F_Inorg and F_Bio are both <1.5%. This indicates that F_Org has a significant impact on membrane fouling, while the contributions of F_Inorg and F_Bio are relatively low. The severity of organic pollution may be related to the adsorption of organic substances on the membrane surface, while inorganic fouling is associated with the scaling of inorganic substances on the membrane surface. However, scaling typically requires a longer time and progresses slowly. Therefore, the contribution of scaling to membrane fouling is weak in this study. Biological fouling usually requires the auxiliary effect of other types of pollutants (such as organic substances) (Nguyen et al. 2012), so the contribution of individual biological fouling is also small.

It is noteworthy that, concerning the composite fouling contributing factors, the contribution of F_Org ∪ F_Inorg is low (<4.0%) for both M1 and M2 membranes. This finding contradicts some studies that reported more severe membrane fouling caused by mixed organic and inorganic contamination. Such discrepancies may be attributed to variations in the quality of raw water used in different studies (Tian et al. 2013; Ma et al. 2019). Additionally, the contribution of organic–biological interactions is higher (>12%) for both M1 and M2 membranes, indicating that the intricate interplay between organic and biological foulants can exacerbate membrane fouling. This phenomenon can be explained by the adsorption of organics on the membrane surface providing sites for microbial attachment (bridging), thus promoting the growth and aggregation of microorganisms. Moreover, the growth of microorganisms on the membrane surface may lead to the production of extracellular polymers (mainly organics), further intensifying the degree of organic fouling (Nguyen et al. 2012).

Concerningly, the combined contribution of F_Org ∪ F_Inorg ∪ F_Bio is 50.4% for the M1 membrane and 70.2% for the M2 membrane, signifying that membrane fouling during actual operation typically arises from the intricate interactions of organic, inorganic, and biological foulants. Notably, the contribution of F_Org ∪ F_Inorg ∪ F_Bio for the M1 membrane is 19.8% lower than that for the M2 membrane. This variance in contribution rates might be attributed to the superior removal efficiency of S(IV)–Fe(II)/PM pretreatment on organic, inorganic, and biological foulants (Figures 5–7). The superior effectiveness of S(IV)–Fe(II)/PM pretreatment in controlling UF membrane fouling is fundamentally rooted in its capacity to regulate membrane foulants at the source.

The continuous-flow UF experiments revealed that, over the 32-day operational period, S(IV)–Fe(II)/PM pretreatment outperformed conventional Al(III) coagulation pretreatment in controlling fouling of the UF membrane. The membrane surface cake layer exhibited greater looseness and smoothness under the influence of S(IV)–Fe(II)/PM pretreatment, diminishing subsequent contaminant deposition and favoring sustained membrane flux and surface cleaning. Notably, the S(IV)–Fe(II)/PM pretreatment demonstrated remarkable efficacy in removing organic matters, especially proteinaceous organic matters and soluble microbial by-products, resulting in a significant reduction in the accumulation of organic foulants on the membrane surface. Additionally, the S(IV)–Fe(II)/PM pretreatment effectively controlled Si in the UF feed water. Given its concurrent reduction of total microorganisms and living microorganisms in the UF feed water, biological fouling on the UF membrane surfaces was satisfactorily managed. VPA results affirmed the predominant contribution of the combined effects of organic, inorganic, and biofilm foulants to UF membrane fouling. The efficient removal of organic, inorganic, and biological membrane foulants by S(IV)–Fe(II)/PM pretreatment positioned it as superior to conventional Al(III) pretreatment for composite fouling control. Subsequent research should conduct pilot experiments under rigorously simulated actual operational conditions to further validate the long-term membrane fouling control capabilities of S(IV)–Fe(II)/PM pretreatment.

This research was funded by the National Science and Technology Major Projects for Water Pollution Control and Treatment (No. 2017ZX07201003) and the National Key Research and Development Program of China (No. 2023YFF0614500).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.