Effect of granular activated carbon adsorption on mitigating microfiltration membrane fouling by algal organic matter

Climate change, population growth, and increased urbanization have contributed to the increasing frequency of eutrophication worldwide (Hoslett et al. 2018). The occurrence of harmful algal blooms in surface water has increased markedly over the last decade (Kudela et al. 2015). The metabolites of algae and other planktonic species are the major constituents of natural organic matter (NOM) in many surface water bodies, which are the sources of potable water in many areas. These substances cannot be easily removed by the traditional drinking water treatment processes such as coagulation–flocculation and sedimentation, creating problems for downstream units such as clogging of filters, increasing biofouling, reducing the efficiency of adsorption beds for the removal of trace contaminants, and increased disinfection byproducts formation (Lopes et al. 2017).

The extensive application of membrane processes, including microfiltration (MF) and ultrafiltration (UF), for drinking water treatment has significantly increased in the last two decades for their effective removal of pathogens such as Cryptosporidium oocysts and Giardia cysts and reduction of water turbidity with a comparable cost to conventional sand–charcoal filtration systems (U.S. Environmental Protection Agency 2005; Wong 2012). Microfiltration (MF, 0.1–10 μm) is widely applied in water treatment plants to remove particulate materials (Lehman & Liu 2009). However, membrane fouling caused by DOM significantly affects the filtration efficiency in water treatment. Membrane fouling due to the surrogate NOM such as commercial humic acid (HA) and Suwannee River NOM (SRNOM) is well-researched (Fang et al. 2010; Chiu et al. 2013; Kim & Dempsey 2013; Zhang et al. 2019a); however, fouling due to real cellular materials of algae or Cyanobacteria (most common eutrophic species) needs further attention due to the ubiquitous nature of the issue and absence of comprehensive research.

A previous investigation of membrane fouling by algal organic matters (AOMs) indicated that algal species and the derived AOM compositions significantly affected membrane fouling behavior (Huang et al. 2014b). An extended Derjaguin–Landau–Verwey–Overbeek (XDLVO) theory was applied to investigate the fouling behavior of AOM fractions from Aphanizomenon flos-aquae and Anabaena flos-aquae. The results indicated that the interface between membrane and neutral hydrophilic fraction presented the highest attractive energy, and controlled the membrane fouling in AOM microfiltration process (Huang et al. 2014a). Activated carbon (AC) adsorption is a process commonly used in water treatment plants to remove the residual organic pollutants, such as humic and fulvic DOM and ammonia. Growing problems with algal blooms have also indicated that AC adsorption is a promising solution for enhancing the removal of recalcitrant AOM (Pivokonsky et al. 2021). Granular activated carbon (GAC) adsorption, as one of the cost-effective and environmentally friendly processes for water treatment plants to remove organic matter, has been extensively applied as a pretreatment process for membrane filtration to mitigate membrane fouling (Rasouli et al. 2017; Mayer et al. 2019). It is generally acknowledged that the performance of MF is influenced by the membrane type, feedwater characteristics, and operational conditions (Lehman & Liu 2009). A hybrid membrane-activated carbon process was applied for the treatment of oil field produced water and the results presented that GAC pretreatment enhanced the removal efficiency of COD and conductivity, and also reduced cake layer formation on membrane surfaces (Kose-Mutlu et al. 2017). A previous study presented that biopolymers, such as proteins and polysaccharides, could be effectively removed by GAC pretreatment prior to UF filtration to mitigate membrane fouling (Huck et al. 2009). Another study (Hatt et al. 2013) further found that coupling GAC to the downstream MF process provided a significant reduction in membrane fouling with improved product water quality and lower carbon usage rate than powder-activated carbon (PAC). Zhang et al. (2019b) investigated the effect of PAC on fouling by algal solution during ultrafiltration using two modes, i.e., the addition of PAC to the bulk feed and pre-depositing PAC onto the membrane surfaces. Both modes improved the removal of EOM from the algal solution; however, the influence of PAC addition on the EOM fouling was weak. The pH value, ionic strength, temperature, and chemistry are the main solution properties that play an important role in the AOM adsorption process (Kopecka et al. 2014). AOM molecules can be predominately positively or negatively charged depending on the solution pH (Safarikova et al. 2013). pH will also alter the characteristics (deprotonation/protonation) of functional groups on the GAC surface. Protonated and deprotonated AOM and GAC functional groups participate in various types of interactions (electrostatic, hydrophobic, and hydrogen bonding) during adsorption, thus substantially affecting the efficiency of the whole process (Abouleish & Wells 2015).

As mentioned above, most of the previous studies on the GAC/PAC adsorption-microfiltration process focused on NOM, and artificial organic micropollutants removal from drinking water treatment, and only a few of the investigations reported an integrated system for the treatment of algae-laden water. A comprehensive investigation of the combination of GAC and microfiltration membranes is not available. Further experimental data are necessary for process optimization and for designing such units. The pH point of zero charge (pHpzc) of GAC used in our work was determined to be 9.5 (Niasar et al. 2016), which shows the surface of GAC was positively charged within the working pH between 5.0 and 8.0. AOM has negative zeta potential values in the pH range of 2–10. If electrolyte/salt was added to increase the ionic strength, the AOM adsorption onto GAC would be reduced, thus the influence of ionic strength was not investigated from the practical perspective (Zhao et al. 2022). The objective of the current study is to investigate the effect of GAC dosage and solution pH on fouling potential and the flux of microfiltration due to several species of algae and Cyanobacteria.

The three species, Chlorella vulgaris (CV), Microcystis aeruginosa (MA), and Phaeodactylum tricornutum (PT), were obtained from the Canadian Phycological Culture Centre (CPCC) at Waterloo University (Waterloo, ON, Canada). The algal cells were cultivated in 2 L flasks in High Salt, 3N-BBM, and F/2, respectively, at 23 ± 2 °C under a fluorescent lamp (3,000 lx) with 16/8 h of light/dark cycle (Villacorte et al. 2015a). Algae and Cyanobacteria were harvested at the stationary growth phase and monitored by cell counting following the previous study (Zhao et al. 2020).

AOM solution was extracted by the following steps: (1) centrifugation of the harvested algal cultures at 3,700 rpm and 23 °C for 30 min (Thermo Scientific Sorvall, Legend T Plus); (2) subsequent filtration of the supernatant by a 1.2 μm filter (hydrophilic acrylic copolymer, Pall Corporation) to obtain extracellular organic matter (EOM); (3) the deposited algae on the filter were washed three times using Milli-Q water, then subjected to three freeze/thaw cycles (−18 °C for 12 h/40 °C for 2.0 h) to destroy the cells (Li et al. 2012), subsequently followed by centrifugation and filtration process as described above to obtain intracellular organic matter (IOM). The obtained AOM stock solutions were stored at 4 °C in a fridge for no more than 48 h before characterization or preparing the feed solution with dissolved organic matter (DOC) of 8.0 ± 0.5 mg/L for GAC adsorption and microfiltration after pH adjustment by 1.0 mol/L NaOH and 1.0 mol/L HCl solution. For comparison with the fouling behavior of AOM, humic acid (98% grade, Thermo Fisher Scientific Chemicals, Inc., USA) solution was used to prepare a working solution as the surrogate of NOM for the GAC-MF experiment.

The commercial GAC (Norit ROW 0.8 SUPRA, CAS Number: 7440-44-0) used in this study was purchased from Sigma-Aldrich Canada Co. The properties of GAC were investigated in a previous study in our group (Wan et al. 2019) shown as follows: surface area ≈ 1,400 m²/g, pore size ≈ 2 nm, total pore volume ≈ 0.7 cm³/g, mesoporous area ≈ 634 m²/g, and microporous area ≈ 766 m²/g. The GAC was screened by mesh sieves to collect the GAC with a size range of 0.42–0.60 mm and followed by washing to remove the fines, then dried in an oven at 105 °C, and stored in a desiccator before adsorption experiments.

The adsorption experiments were carried out in 500 mL Erlenmeyer flasks containing 400 mL of AOM solution using a Bench-top Orbital Shaker (Max Q 400, Thermo Scientific, ON, Canada) operated at temperatures of 23 ± 1.5 °C under 200 rpm of shaking speed. Since surface and groundwater contain DOC in the range of 2–10 mg/L (Gumus & Akbal 2017), 1.0 g/L GAC was added into the DOM solution with an initial DOC of 8.0 ± 0.5 mg/L. The pH of the solution was adjusted using 1.0 mol/L HCl or 1.0 mol/L NaOH to reach the pH values of 5–8 before adsorption. After GAC adsorption with a retention time of 1.0 h, the solution was filtered using a 1.2 μm filter (hydrophilic acrylic copolymer, Pall Corporation) to remove GAC particles. The duplicated experiments have been conducted and reported with average values in this study.

The MF membrane used was a 0.45 μm nominal pore size hydrophilic PVDF membrane (Millipore Corporation, USA) with an effective filtration area of 1.59 × 10⁻³ m² in a dead-end stainless steel filter holder at a constant transmembrane pressure (TMP) of 50 ± 0.5 kPa by compressed air and operating temperature of 25 ± 0.5 °C. Before filtration, all fresh membranes were soaked in Milli-Q water for at least 24 h to remove possible organic contaminants. The filtrate weight was measured constantly by a digital balance (Denver SI-4002, Denver Instrument Co., USA) and data were automatically logged to a connected computer equipped with a data acquisition system shown in Supplementary Figure S1.

To elucidate fouling mechanisms, the classic filtration models, including complete blocking, standard blocking, intermediate blocking, and cake filtration, were applied to understand the flux decline during the MF of the DOM solution under constant pressure. The instantaneous flux was calculated by numerically differentiating the cumulative volume filtered (V) per unit membrane area and analyzing it using blocking laws listed in Supplementary Table S1.

DOC of AOM solution was measured using a TOC–V_CPN analyzer (TOC–V_CPN, Shimadzu, Japan) with a detection limit of 0.1 mg/L calibrated by a standard glucose solution. Temperature and pH were measured using a pH meter (Orion Model STAR A111, USA). The UV absorbance at 254 nm (UV₂₅₄) was measured by a UV/Vis spectrophotometer (UV-3600, Shimadzu, Japan) and the specific UV absorbance (SUVA, L mg⁻¹ cm) was calculated from UV₂₅₄ value divided by DOC concentration.

It was indicated that GAC adsorption was effective to promote DOM removal; similar results also were proposed that DOCs of lower molecular weight substance were adsorbed onto GAC and DOCs of higher molecular weight can be removed by the membrane formed on the surface of the MF membrane (Kim et al. 2009). In comparison with DOM from AOM, humic acid presented relatively lower removal efficiency. It was also noted that the addition of GAC (1.5 g/L) presented (Figure 1) approximately 16.49 and 16.67% greater DOC and UV₂₅₄ removal than without GAC adsorption.

Specifically, the feed solution from AOM presented a significantly greater flux decline compared to humic acid, with more than 80% flux decline obtained at the end of the single cycle filtration (Figure 2(a)). A similar trend was also observed for the filtration of humic acid and AOM by Zhang et al. (2018). With 1.5 g/L GAC adsorption (Figure 2(b)), the filtration flux for each DOM was improved, with a maximum 15% improvement in flux occurring for HA, followed by PT-EOM, CV-IOM, CV-EOM, MA-EOM, and PT-IOM. It was noted that even though filtration flux was increased for all AOM after GAC adsorption, about an 80% decline in flux occurred for AOM. In comparison, flux for the humic acid solution declined by 60%.

The pH of the DOM solution not only alters the surface charge of the adsorbent, the dissociation of functional groups on the active sites of the adsorbent but also affects the ionization degree of the DOM in the solution (Kuśmierek & Świątkowski 2015; Zhang et al. 2019c). In this study, the pH dependence of DOM removal by microfiltration with GAC adsorption was performed in the pH range of 5.0–8.0. The initial pH of the solution can affect the characteristics (deprotonation/protonation) of functional groups on the GAC surface. Consequently, protonated and deprotonated AOM and GAC functional groups participate in multiple types of interactions during adsorption, including electrostatic, hydrophobic, and hydrogen bonding, thus substantially affecting the efficiency of the whole process (Abouleish & Wells 2015).

To probe the effect of feed solution pH on membrane permeate and degree of fouling after GAC adsorption for each DOM, experiments were performed at pH 5.0 and 8.0 for GAC adsorption with conditions mentioned above followed by three cycles of MF filtration without any additional pH control. As pH can alter the degree of ionization of the DOM presented in the solution, the electrostatic interaction between DOM and membrane would play a significant role. At higher pH, the repulsive force between negatively charged membrane and anionic species of DOM caused lower flux decline. On the contrary, at lower pH, the negatively charged membrane would attract protonated and positively charged DOM species, thus leading to increased fouling and a decrease in permeate flux (de la Casa et al. 2007).

As shown in Table 1, among the four fouling models for each DOM solution, for CV-EOM and MA-IOM, membrane fouling was controlled by standard blocking with R² values of 0.997 for CV-EOM and 0.982 for MA-IOM. For the other DOM solutions, including humic acid, CV-IOM, MA-EOM, PT-EOM, and PT-IOM with the maximum R² values 0.977, 0.973, 0.958, 0.995, and 0.977, respectively, the fouling mechanisms were predominated by intermediate blocking and standard blocking. This result can mostly be attributed to the MW distribution of the DOM, in which the high-MW biopolymers (>13.5 kDa) might be removed by initial filtration through a 1.2 μm membrane (Fang 2010; Zhang et al. 2014), in addition to the fact that the peptides from AOM (MW < 10 kDa) and medium-MW compounds (i.e., humic-like substances, ∼ 1 kDa) with high affinity with GAC surface might more susceptible to adsorption onto GAC (Pivokonsky et al. 2021).

It could also be inferred that the comparable R² values of the fouling models implied that the fouling process was controlled by multiple mechanisms. This is attributed to the broad MW distribution of AOM (Fang et al. 2010; Li et al. 2018). The low-MW substance may be trapped inside membrane pores resulting in standard blocking, and the high-MW components may be deposited on the membrane surface to form a cake layer. A previous study on the UF membrane fouling potential of EOM also demonstrated that multiple mechanisms, including cake filtration and standard blocking, dominated the fouling formation (Li et al. 2014; Yan et al. 2017).

In a comparative study on membrane fouling potentials of algal extracellular and IOM, cake filtration has been identified as an important mechanism for flux decline.

It could be noted that multiple mechanisms might also take effect during filtration considering the relatively higher R² value (0.958 ∼ 0.997); for instance, intermediate blocking and cake filtration dominated the fouling formation for humic acid, and standard blocking and intermediate blocking controlled the membrane fouling for most AOM, except for CV-IOM and humic acid for which intermediate blocking and cake filtration mechanisms governed the fouling formation. The bold figures of in Table 1 show the largest R2 value among four fouling models for each DOM, which indicated the principal mechanism dominated the fouling formation for the DOM during the first cycle of MF after GAC adsorption. An earlier study demonstrated that cake filtration dominated the fouling during ultrafiltration of AOM from MA; however, the difference is mainly due to the type of filtration (Liu et al. 2017).

The influence of GAC dosage and initial pH on microfiltration to remove DOM derived from three different algae, as well as humic acid, was investigated using a dead-end down-flow MF unit in batch scale. The combination of GAC adsorption and MF can significantly enhance DOM removal up to 72.23 and 85.95% for DOC and UV₂₅₄ for CV-IOM. The addition of GAC can not only promote DOM removal but also mitigate the flux decline and reduce irreversible fouling. A lower initial pH value within the experimental range (5–8) showed positive effects for DOM removal and membrane reversibility. The total removal efficiency of AOM was higher than humic acid; however, a greater flux decline and higher irreversible fouling were observed from AOM than that of humic acid. The AOM derived from CV presented a better removal efficiency with less flux decline and irreversible fouling, followed by the Cyanobacteria, MA, and the diatom, PT. The fouling models implied that intermediate blocking and standard blocking were the dominant membrane fouling mechanisms for most DOM except the CV-IOM and humic acid where intermediate blocking and cake filtration controlled the fouling process. Although IOM from each algal species demonstrated relatively higher removal performance than EOM, considering the greater flux decline and irreversible fouling compared to EOM, to maintain the algal cell integrity and avoid lysis to release IOM are important considerations for membrane treatment of algae-laden water.

This work was financially supported by the Ontario-China Research and Innovation Fund (OCRIF) ‘Ensuring Water Supply Safety in Beijing: Water Diversion from South to North China’ (No. 2015DFG71210).

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

The authors declare there is no conflict.