2-Aminoimidazole Reduces Fouling and Improves Membrane Performance

Abstract

Biofouling is difficult to control and hinders the performance of membranes in all applications but is of particular concern when natural waters are purified. Fouling, via multiple mechanisms (organic-only, biofouling-only, cell-deposition-only, and organic+biofouling), of a commercially available membrane (control) and a corresponding membrane coated with an anti-biofouling 2-aminoimidazole (2-AI membrane) was monitored and characterized during the purification of a natural water. Results show that the amount of bacterial cell deposition and organic fouling was not significantly different between control and 2-AI membranes; however, biofilm formation, concurrent or not with other fouling mechanisms, was significantly inhibited (95-98%, p<0.001) by the 2-AI membrane. The limited biofilm that formed on the 2-AI membrane was weaker (as indicated by the polysaccharide to protein ratio) and thus presumably easier to remove. The conductivity rejection by the 2-AI and control membranes was not significantly different throughout the 75-hour experiments, but the rejection of dissolved organic carbon by biofouled (biofouling-only, cell-deposition-only, and organic+biofouling) 2-AI membranes was statistically higher (10-12%, p=0.003-0.07). When biofouled, the water permeance of the 2-AI membranes decreased significantly less (p<0.05) over 75 hours than that of the control membranes, whether or not other additional types of fouling occurred concurrently. Despite the initially lower water permeances of 2-AI membranes (11% lower on average than controls), the 2-AI membranes outperformed the controls (10-11% higher average water permeance) after biofilm formation occurred. Overall, 2-AI membranes fouled less than controls without detriment to water productivity and solute rejection.

Graphical Abstract

1. Introduction

One of the most widespread challenges in the application of high-pressure water purification membranes is the accumulation or growth of substances on their surfaces, otherwise known as fouling. Fouling increases operational costs and negatively impacts membrane performance. For example, fouling increases the frequency of membrane cleaning and membrane replacement, including the need for chemicals and downtime associated with these procedures, and decreases membrane water productivity and the quality of purified water [1-7].

A variety of substances from the feed water can foul membranes, including precipitated inorganics (scaling), organic matter (organic fouling), colloids (colloidal fouling), and bacterial biofilms (biofouling). Among these four types of fouling, biofouling is the most difficult to prevent and control [2,6,8-11]. Biofouling occurs initially when planktonic microbes in the feed, sensing a suitable surface in their environment, attach to the membrane [1,9,12,13]. These sessile bacteria grow, reproduce, and excrete extracellular polymeric substances (EPS) to form a biofilm. Most biofouling control strategies currently used at treatment facilities aim to kill, inactivate, or remove bacteria before attachment using technologies such as micro-/ultrafiltration and feed water disinfection (e.g., chlorination or chloramination), not directly addressing the attachment of bacteria to the membrane or the excretion of EPS, which constitutes up to 90% of biofilms by volume [1,9,14-17]. An ideal biofouling control strategy would actively prevent both attachment and excretion of EPS.

In our recent work [18,19], we describe such a control technology, where an anti-biofilm 2-aminoimidazole (2-AI) compound, shown in Fig 1, was incorporated into the polyamide active layer of RO/NF membranes. These “2-AI membranes” inhibited biofilms significantly (p=0.001-0.12) and substantially—up to 96% compared to corresponding control 2-AI lacking membranes. The 2-AI is a bioactive, but non-biocidal compound, that blocks a wide range of bacteria from sensing and responding to their environment by disrupting the bacteria’s two-component regulatory system [20-23]. By disrupting this system, the bacteria stay in a planktonic state, do not attach to surfaces, do not excrete EPS, and thus do not form biofilms [21,24-28]. 2-AIs are the only known non-biocidal class of compounds that are effective at preventing and dispersing biofilms of bacteria across different phyla, classes, and orders [21,23-28].

Fig 1.

The 2-AI, 5-(4-aminophenyl)-1H-imidazol-2-amine (2-AI-para), used in this study, and incorporated into the best performing 2-AI membranes from our previous study [18].

The effects of 2-AI chemistry and physico-chemical properties of 2-AI membranes (i.e., hydrophobicity, surface charge, and roughness) on biofilm inhibition were evaluated in our previous work [18,19]. The 2-AI membranes performed the best when 2-AI was grafted onto commercial membranes instead of incorporated during polyamide casting [18,19]. The membranes with 2-AI incorporated into the polyamide active layer of commercial membranes significantly (p=0.01-0.12) inhibited Pseudomonas aeruginosa biofilms by 61-96%, when tested under static conditions (i.e., uninhibited by hydrodynamic forces) with pure bacterial solutions (i.e., no other types of foulants present) and ample nutrient supply [18,19]. No substantial change in physico-chemical properties (i.e., hydrophobicity, surface charge, and roughness) was found and thus inhibition was attributed to 2-AI chemistry [18,19]. Grafting 2-AI onto commercial membranes did not statistically significantly affect salt rejection and resulted in lower decreases in pure water permeance (ranging from 0 to 25%) than the other 2-AI incorporation approach [18,19]. Although these levels of static biofilm inhibition are promising, it is important to test these membranes under dynamic and realistic conditions for proof-of-concept towards full-scale operation. A proof-of-concept for anti-biofouling membranes necessitates tests under cross-flow filtration with natural water matrices to demonstrate how performance is affected by 2-AI incorporation and to better understand fouling mechanisms—including whether 2-AI incorporation alters organic matter accumulation, bacterial deposition, and biofilm formation. A more complete, systematic understanding of fouling will enable us to optimize the anti-biofouling technology, as well as appropriately consider it for implementation.

Accordingly, the overall goal of this proof-of-concept study was to evaluate the differences between 2-AI membranes and corresponding 2-AI-lacking control membranes in terms of overall performance (i.e., fouling, water permeance, and solute rejection), and the mechanisms at play in those differences, when operated under cross-flow filtration with a natural water containing multiple foulants. A natural water was used to promote all fouling mechanisms that would occur at a treatment plant and for some experiments was seeded with bacteria to further promote biofouling. To achieve our overall goal, we sought to meet the following specific objectives: (1) to evaluate changes in water permeance and contaminant rejection of 2-AI and control membranes when biofilm formation, bacterial cell deposition, and organic matter accumulation occur independently and co-occur; (2) to quantify and characterize accumulated foulants on 2-AI and control membranes; and (3) to relate the differences in fouling to differences in water flux and solute rejection between 2-AI and control membranes.

2. Materials and Methods

2.1. Reagents, Control Membranes, and 2-AI Membranes

Unless otherwise noted, all reagents were purchased from commercial sources, were of ACS reagent grade or better, and were used without further purification. The chemical structure of the 2-AI molecule (i.e., 5-(4-aminophenyl)-1H-imidazol-2-amine, or 2-AI-para for short) that was incorporated into commercial membranes is shown in Fig 1. 2-AI-para was synthesized in-house, and the purity was confirmed as described previously [18]. ESPA3 commercial membranes (Hydranautics, Oceanside, CA, USA) were selected as the membranes into which 2-AI-para was incorporated because ESPA3 with 2-AI incorporated was previously shown to inhibit biofilm effectively (71-92%) while having modest differences in water permeance (25% decrease) and minimal differences in salt rejection (0.1% increase) compared to corresponding control (2-AI-lacking) membranes [18]. The 2-AI membranes were prepared as described previously [18]. Briefly, ESPA3 membranes were exposed to a basic (pH=9) aqueous solution of 2-AI-para (1.6 mmol), 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (1.0 mmol), N-methylmorpholine (1.5 mmol), and hydroxybenzotriazole (1.0 mmol) overnight (≈18 hours), and then rinsed with ultrapure water. The membranes were prepared immediately before use for each experiment.

2.2. Cross-flow Apparatus

A conventional laboratory-scale flat sheet membrane cross-flow system with four membrane cells in series was used for the fouling experiments. The system design enabled continuous control and monitoring of pH, dissolved oxygen (DO), and temperature of the feed water throughout the experiments. The effective filtration area in each membrane cell was 35.6 cm², and the pressure drop between the first and fourth membrane cells was insignificant (<2 psi). The permeate flow rate (Q_w, m³.s⁻¹) from each membrane tested was calculated as the mass of permeate water collected per unit time divided by water density (0.998 g.cm⁻³). A more detailed description of the cross-flow system is provided in the Appendix.

2.3. Feed Solutions

Table 1 describes the different feed solutions that were used in the cross-flow fouling experiments. The solution contents were modeled after Herzberg and Elimelech [29]. The natural water used was a raw drinking water source, University Lake. The natural water used for all experiments was collected during one sampling event at the OWASA Jones Ferry Drinking Water Treatment Plant in Carrboro, NC (USA), and stored at 4 °C for a maximum of one month before use. The non-purgeable organic carbon concentration (TOC) of the natural water was 7.2 mg.L⁻¹, the UV 254 absorbance was 0.23 cm⁻¹, and the conductivity was 150-300 μS.cm⁻¹. The water was filtered using a 1.2 μm G4 glass fiber filter (Fisher Scientific, Hampton, NH, USA), followed by a 0.45 μm hydrophilic mixed cellulose ester filter (EMD Millipore, Billerica, MA, USA). Sodium azide (2 mM) was added to some of the feed solutions to inhibit bacterial growth. In all feed solutions, Lennox LB broth (1.0 mL.L⁻¹) and potassium phosphate (0.45 mM) were added to provide nutrients and stabilize the pH, respectively. In the biofouling-only and cell-deposition-only feeds, there was relatively low background conductivity (<10 μS.cm⁻¹), so the following combination of monovalent and divalent salts [29] were added to achieve conductivity similar to the natural water: 2.6 mM NaCl, 0.55 mM NaHCO₃, 0.4 mM MgSO₄, and 0.6 mM NH₄Cl. Feed solutions were autoclaved at 121 °C for 60 minutes on a liquid sterilization cycle to prevent undesired bacterial growth. Feed waters were always autoclaved before adding the sodium azide or the bacteria.

Table 1.

Description of feed solutions used in cross-flow fouling experiments.

Feed solution	Expected fouling mechanisms	Solution contents
Organic-only	Independent mechanism: Organic matter deposition	Natural water, nutrients (LB broth), buffer (potassium phosphate) —autoclaved Growth inhibitor (sodium azide)
Biofouling-only	Independent mechanism: Biofilm formation and cell deposition	Ultrapure water, nutrients (LB broth), PA14 (107 cells.mL−1), salts, buffer (potassium phosphate)
Cell-deposition-only	Independent mechanism: Cell deposition	Ultrapure water, nutrients (LB broth), PA14 (109 cells.mL−1), salts, buffer (potassium phosphate)—autoclaved Growth inhibitor (sodium azide)
Organic+biofouling	Co-occurrence of all mechanisms: Organic matter deposition, biofilm formation, and cell deposition	Natural water, nutrients (LB broth), buffer (potassium phosphate) —autoclaved PA14 (107 cells.mL−1)

PA14= Pseudomonas aeruginosa strain 14

Pseudomonas aeruginosa strain 14 (PA14) was used to biofoul the membranes because P. aeruginosa are ubiquitous in the environment, are known to form biofilms aggressively and have been found to contribute to membrane biofouling at treatment plants [13,30]. PA14 was cultured overnight in LB broth to the exponential growth phase. The overnight culture was then washed by centrifuging the cultures at 3670 rpm for 10 minutes, decanting the broth, and then vortexing the pellet in a small volume (e.g., 15 mL) of the appropriate feed solution. This sequence (centrifugation, decanting, and vortexing in the new feed solution) was performed twice more before the concentration of the PA14 solution was measured as optical density at 600 nm. The PA14 solution was diluted in the feed water to achieve the target concentration, as indicated in Table 1.

2.4. Cleaning of Cross-Flow Membrane System between Experiments

The cross-flow system was cleaned before and after every experiment to disinfect and remove bacteria and trace contaminants as described elsewhere [29]. The following cleaning solutions were circulated through the system in sequence: 0.5% sodium hypochlorite for 2 hours, house-prepared ultrapure water (>17.9 MΩ.cm⁻¹) for 15 minutes twice, 5 mM ethylenediaminetetraacetic acid and 2 mM sodium dodecylbenzenesulfonate in ultrapure water for 15 minutes three times, 70% ethanol for 1 hour, and ultrapure water for 15 minutes three times. Bacterial enumeration and conductivity of final ultrapure rinse water was measured to ensure the system was clean for the next experiment.

2.5. Fouling Experiments

After cleaning, four membranes were placed in the cells with fresh feed and permeate spacers. Control membranes were placed in cells 1 and 3, and 2-AI membranes in cells 2 and 4. The membranes were compacted with ultrapure water adjusted to pH=8 with sodium hydroxide, at 13.8 bar for 24 hours, ensuring the pure water permeance was stable (<2% change per hour). The ultrapure water was then replaced with the appropriate feed solution, as listed in Table 1. The pH of feed waters was initially between 7.0 and 7.9 (no adjustment to the ambient feed pH) with <0.4 pH unit change during the experiments. The DO concentration in the feed was initially between 6.2 and 7.0 mg.L⁻¹ during all runs. The feed water temperature was kept constant at 22.0 °C. To ensure the integrity of membranes, the initial conductivity rejection before fouling was measured and found to be <0.1% to 1% different across membranes for each fouling experiment. Fouling experiments were conducted at constant pressure (13.8±0.3 bar) and cross-flow velocity (14 cm.s⁻¹) for 75 hours total. Feed and permeate samples were collected at 15, 30, 40, 55, 65, and 75 hours for analyses of water permeance and solute rejection, as described in subsequent sections. At 75 hours of filtration, the pressure was slowly and steadily decreased over 5 minutes to prevent drastic changes in pressure that could alter the foulant layer. The membranes were carefully taken out of the cross-flow cells, very lightly rinsed with ultrapure water to remove loose bacteria and residual feed, and analyzed as described below.

2.6. Scanning Electron Microscopy (SEM)

SEM images of membrane surfaces were obtained to capture the foulant layers of 2-AI and control membranes visually. Both fouled and non-fouled membrane samples of each 2-AI and control membranes were imaged. We examined the membranes at varying levels of magnification first and then collected replicate images (n=7-16) at the higher magnification levels that were representative for each fouled membrane type. Membrane samples were air-dried for >48 hours before SEM analysis, and when needed, the samples were coated with 2–5 nm Au/Pd alloy to prevent charging. SEM imaging was performed using a Helios Nanolab 600 dual beam system (FEI, Hillsboro, OR), and accelerating voltage and current of 2.0 kV and 0.34 nA, respectively.

2.7. Attenuated Total Reflectance-Fourier Transform Infrared (ATR-FTIR) Spectroscopy

ATR-FTIR analysis was used to characterize the chemical signatures and relative thickness of foulant layers on control and 2-AI-membranes. The relative thickness of foulant layers was evaluated qualitatively by assessing the attenuation of the intensities of the FTIR peaks associated with the membrane materials, and intensities of the foulant-associated peaks, similarly as described by Hausman and Escobar [31]. All samples were gently rinsed with ultrapure water and then air-dried for >48 hours before analysis. Four replicate samples were analyzed for each sample type, and 24 scans were taken per replicate. The spectra reported are the average of the replicates. Analyses were performed on a sample area of 3.1 mm² over the 400-3997 cm⁻¹ wavenumber range with 2 cm⁻¹ resolution. Analyses were performed using a Bruker Alpha spectrometer (Bruker Optics, Billerica, MA), equipped with an IR source with a 45° incident angle and an Alpha-P ATR accessory equipped with a diamond ATR crystal.

2.8. Total (Non-Purgeable) Organic Carbon (TOC)

Permeate and feed samples for TOC analysis were collected at 75 hours of filtration time during cross-flow experiments. Fouled membrane samples (2 cm²) were submerged in ultrapure water (5 mL) immediately after the cross-flow experiment and alternatingly vortexed for 30 seconds and sonicated for 1 minute for three cycles. Calibration standards were prepared with potassium hydrogen phthalate diluted in ultrapure water. All TOC samples and standards were prepared by diluting to the appropriate concentration range (1-10 mg.L⁻¹), and then adjusting the pH to <2.5 with hydrochloric acid. The samples were analyzed on a Shimadzu TOC-V CPH analyzer equipped with an ASI-V autosampler.

2.9. EPS Extraction and Characterization of Polysaccharide and Protein Content

Immediately following cross-flow experiments, fouled membrane samples (4 cm²) were added to 20 mL of 0.1 M NaCl and shaken for 45 minutes at 4 °C. The EPS was then extracted as described by Liu and Fang [32], by adding formaldehyde and sodium hydroxide to the solution with the membrane still immersed in it, followed by centrifugation. The solution was filtered through a 0.22 μm nylon filter and then dialyzed against ultrapure water using a 3.5 kDa membrane (Spectra/Por). EPS extracts for non-fouled membranes were used as controls and were prepared in the same manner as for fouled membranes.

The polysaccharide content in EPS extract samples was quantified as described by Dubois et al. [33], using phenol and sulfuric acid as reagents, followed by measuring the absorbance at 492 nm against alginic acid standards prepared in ultrapure water. The protein content was quantified as described by Bradford et al. [34] using Coomassie protein assay reagent, standards prepared with bovine serum albumin in ultrapure water, and absorbance measurements at 595 nm.

2.10. Enumeration of Biofilm-forming Bacteria in Feed Solutions and Membrane Samples

Feed water samples were collected at least twice (at 0 and 75 hours) for bacterial enumeration during each cross-flow experiment. After the cross-flow run, fouled membrane samples (2 cm²) were rinsed with ultrapure water to remove loose bacteria from the surface. The rinsed membranes were then added to 2.0 mL LB broth. The membranes in solution were vortexed for 30 seconds, then sonicated for 30 seconds, for three vortex-sonication cycles to produce membrane extract solutions.

The bacteria were enumerated in feed water samples and membrane extract solutions by a colony count method where vortexed solutions were serially diluted (nine 10x dilutions) and spotted (5 μL) onto LB-agar plates, which were incubated at 37 °C overnight. The formed colonies were counted and the areal density of colony-forming units (CFU.cm⁻²) was calculated. The areal density of colony-forming units was used as the quantitative descriptor of the areal mass of biofilm on fouled membranes. No colonies were detected in the feed, nor on the membranes during the sterile runs (with organic-only and cell-deposition-only feed solutions, see Table 1).

2.11. Water Permeance and Solute Rejection

Water permeance is the volume of water that is produced by a membrane normalized by unit time, pressure, and membrane area. The permeate flow rate (Q_w, m³.s⁻¹) of each membrane sample tested in fouling experiments was measured at least eight times during each experiment (0, 5, 15, 30, 40, 55, 65, and 75 hours. The measured Q_w was used to obtain the membrane water permeance (A, m.s⁻¹.bar⁻¹) as

$$A=\frac{Q_w}{\Delta P\times a}.$$ (1)

where ΔP (13.8 bar) is the transmembrane pressure and a (0.00356 m²) is the effective membrane filtration area. When evaluating the decline in performance due to fouling, normalized water permeance (A_t/A₀) is reported and was calculated as the water permeance at time t (A_t) divided by the initial water permeance (A₀).

The conductivity rejection (%R_C) of membrane samples tested in fouling tests was also obtained alongside the water flux measurements, at least eight times during each experiment. %R_C was calculated as

$$\%R_C=\left[1-\left(\frac{C_p}{C_f}\right)\right]\times 100\%,$$ (2)

where C_p and C_f correspond to the conductivities of the permeate water and feed water, respectively. Additionally, TOC was measured in permeate and feed water samples collected after 75 hours of fouling. The corresponding TOC concentrations were used in Equation 2 (instead of conductivity values) to calculate the rejection of organics (%R_Org). Reported water permeance and rejection values correspond to the average of duplicate membrane samples.

2.12. Statistical Significance Testing

When appropriate, statistical significance testing was used to compare the performance (i.e., water permeance, solute rejection, and fouling) of control and 2-AI-membranes. Unless otherwise stated, a one-tail unpaired two-sample t-test assuming unequal variances was used. The null hypothesis (H₀) was that the mean of the control membrane was equal to the mean of the 2-AI membrane (${\overline{\mathrm{\mu }}}_{\text{ct1}}\text{=}{\overline{\mathrm{\mu }}}_{\text{2-AI}}$), and the alternative hypothesis (H₁) was that one of the means was greater than the other depending on the parameter compared (${\overline{\mathrm{\mu }}}_{\text{ct1}}>{\overline{\mathrm{\mu }}}_{\text{2-AI}}$ or ${\overline{\mathrm{\mu }}}_{\text{ct1}}<{\overline{\mathrm{\mu }}}_{\text{2-AI}}$). The calculated p-values are reported throughout the results and discussion section and p<0.05 is considered significant. When p-values were close to being significant (e.g., 0.05<p≤0.15), this is specifically highlighted and results are discussed in further detail.

3. Results and Discussion

3.1. Characterization of Membrane Foulant Layers in Control and 2-AI Membranes

3.1.1. Visual Appearance of Foulant Layers

Fig 2 shows representative SEM images of control and 2-AI membranes after being fouled with the various feed solutions, listed in Table 1, for 75 hours. Fouling the membranes with the organic-only feed solution produced a smooth fouling layer that appeared dense at the surface on both control and 2-AI membranes, with no distinctive visual differences in the foulant layer between the two membrane types. The fouling mechanism that appeared to occur was organic matter surface accumulation. The cracking in the organic fouling layers was due to the sample drying process and appeared to have similar spacing in control and 2-AI membranes and thus the cracks were not features further considered in this discussion.

Fig 2.

Representative surface SEM images of fouled control and 2-AI membranes at 1,000x (1^st and 3^rd rows) and 10,000x (2^nd and 4^th rows) magnifications. The scale bars on the 1,000x and 10,000x images are 50 μm and 5 μm, respectively. The headings at the top indicate feed water used during the fouling experiments (see Table 1). Where applicable, example bacteria have been outlined in red (10,000x images) and red arrows placed to indicate example dark patches associated with bacteria (1,000x images), to guide the eye.

When we used the biofouling-only feed solution to foul the membranes, both control and 2-AI membranes were covered in what appeared to be a biofilm (the expected fouling mechanism), with distinctive rod shapes of dimensions consistent with the bacteria (PA14). The only visual difference between control and 2-AI membranes was a slightly higher areal density of bacteria on the control membrane. We could not quantitatively evaluate the thickness and volumetric density of the biofouling layer by SEM.

When we fouled the membranes with the cell-deposition-only feed solution, there was not a clearly-defined, cohesive fouling layer as with the biofouling-only feed solution, but rather patchy accumulations of dark rod shapes from bacteria on top of the membrane. Although we targeted the cell concentration in solution to be consistent under biofouling-only and cell-deposition-only fouling conditions, the number of cells that deposited under cell-deposition-only conditions was less than the number of cells that either was deposited or that developed on the membranes during biofouling-only conditions. We speculate that bacterial cells stick better to biofilm than to the bare membrane surface; thus, due to the presence of biofilm in the biofouling-only condition, more cells adhered to the membrane surface than under cell-deposition-only conditions. The images clearly show (particularly at 10,000x magnification) that the membrane fouling mechanism was cell deposition, with more cell deposition on the control membrane than on the 2-AI membrane.

Finally, when we fouled the membranes using the organic+biofouling feed solution, the fouling layers appeared visually different than when fouled with the other feed solutions due to the potential co-occurrence of all fouling mechanisms. The control membrane was fully covered by the foulant layer, whereas the 2-AI membrane had some areas that were not covered with foulant (e.g., bottom left corner of 10,000x image). On the control membrane, the foulant layer covered or surrounded many bacteria (see 10,000X image), where the bacteria did not protrude out of the foulant layer as clearly in the biofouling-only or cell-deposition-only cases. The 2-AI membrane had fewer bacteria overall, with the foulant layer appearing to be underneath the bacteria rather than covering/surrounding them. The fouling mechanisms also appeared to be different between control and 2-AI membranes. Biofilms formed on the control membrane were similar to the foulant layer from the biofouling-only condition, while the biofilms formed on the 2-AI membrane were similar to the foulant layer from the cell-deposition-only condition.

To summarize, upon visual inspection, the 2-AI membrane appeared to foul less under all experimental conditions that involved biofilm formation and/or cell deposition (i.e., biofouling-only, cell-deposition-only, and organic+biofouling). When all foulant types were present, foulant layers of control membranes appeared to cover/surround the bacteria, and foulant layers of the 2-AI membranes appeared to be underneath the bacteria.

3.1.2. Chemical Signature

Fig 3 presents partial FTIR spectra of fouled membranes; the wavenumber ranges in the figure correspond to IR regions associated with specific molecule types (i.e., proteins, polysaccharides, aliphatics and aromatics). The full spectra are presented in Fig A.2 and Table A.1 summarizes the full list of wavenumber ranges/peaks associated with specific molecule types [31,35-37]. In the subsequent discussion we focus on the absorbance peaks with the most relevant differences between control and 2-AI membranes. This analysis focuses on the molecular differences in foulant layers of control and 2-AI membranes. Note that differences in specific peaks indicate only differences in the amounts of the corresponding specific molecule types present and do not indicate that all molecule types or all foulants were different, nor that total fouling was different.

Fig 3.

ATR-FTIR spectra, at select wavenumbers associated with specific foulants, of fouled control membranes (black line) and fouled 2-AI membranes (red line). The membranes were fouled by using organic-only, biofouling-only, cell-deposition-only, and organic+biofouling feed solutions. We analyzed four membrane samples for each fouling condition—two controls and two 2-AIs—each from separate membrane cells. Each membrane was analyzed at four different areas, and we report the average of the eight replicate measurements.

The fouled control and fouled 2-AI membranes were statistically similar in the absorbance peaks associated with polysulfone (1488, 1145-1180 cm⁻¹) under each of the fouling conditions (Fig A.2). This indicates that the thickness of the foulant layers observed on the control and 2-AI membranes during each experiment were similar (i.e., no detectible difference).

Fig 3 shows when we used the organic-only feed solution to foul the membranes, the 2-AI membrane had statistically higher polysaccharide (1400-1500, 900-1200 cm⁻¹) and protein (1500-1700 cm⁻¹) fouling than the control membrane. Since there was no biofouling when we used the organic-only feed solution (i.e., sodium azide was present as bacterial growth inhibitor), we speculate the polysaccharide and protein fouling is microbially derived organic matter that has been observed with this water source in previous work [38]. These differences indicate that when the only mechanism of fouling is organic matter accumulation, the 2-AI membrane slightly alters the composition of the fouling layers, likely due to increased or decreased affinity of the membrane surface for specific types of organics (such as polysaccharides or proteins).

When fouled with the biofouling-only or cell-deposition-only feed solutions, the fouled control membranes had statistically higher protein (1500-1700 cm⁻¹), polysaccharide (1400-1500, 900-1200cm⁻¹), and aliphatic (2900-3000 cm⁻¹) fouling than the 2-AI membranes. Proteins and polysaccharides are associated with biofilm formation and bacterial cell deposition fouling mechanisms [37,39]. Therefore, biofilm formation and cell deposition were lower in the 2-AI membranes compared with control membranes.

When we used the organic+biofouling feed solution to foul the membranes, the absorbance peaks associated with polysaccharides (1400-1500, 900-1200 cm⁻¹) for control and 2-AI membranes were not statistically different, but two main absorbance peaks associated with proteins (1503 and 1584 cm⁻¹) were statistically higher on the 2-AI membranes. The other protein absorbance peaks (e.g. 1545 and 1650 cm⁻¹) were statistically similar across control and 2-AI membranes. Proteins (PN) and polysaccharides (PS) are the main components of EPS, and their mass ratio (PS/PN) is related to biofilm strength, where the higher the PS/PN ratio, the more adherent and cohesive the biofilm is [37,39]. If IR absorbance is used as a surrogate of PS and PN content, our FTIR results suggest that when we used the organic+biofouling feed solution to foul the membranes, the PS/PN ratio was lower for 2-AI membranes than for the control membranes. Thus, when all fouling mechanisms co-occurred, the biofilm on the 2-AI membrane was weaker than the biofilm on the control membranes.

To gain further insight on the strength of biofilms formed on 2-AI membranes, we extracted EPS from biofouled membranes and we obtained the areal mass of proteins and polysaccharides, with results shown in Fig 4. The average polysaccharide areal mass was higher on the fouled control membranes (50±5 μg.cm⁻²) than on the fouled 2-AI membranes (34±10 μg.cm⁻²), whereas the protein areal masses were similar (24±7 μg.cm⁻² and 21±4 μg.cm⁻², respectively). Therefore, the PS/PN ratio was lower for fouled 2-AI membranes (1.6±0.1) than for fouled control membranes (2.2±0.4), again indicating that the biofilm was weaker on the 2-AI membranes.

Fig 4.

Areal mass of polysaccharides and proteins on control (black dotted bars) and 2-AI membranes (red striped bars) fouled with the biofouling-only feed solution. We extracted the extrapolymeric substances (EPS) and analyzed it for both protein and polysaccharides from six samples of each membrane. The error bars correspond to the standard error of all replicate measurements.

Overall, the ATR-FTIR and EPS results suggest that there was less severe biofouling on 2-AI membranes than on control membranes and that the biofilms on 2-AI membranes were weaker (i.e., less adherent and cohesive) than on control membranes.

3.1.3. Areal Mass of Foulant Layers and Biofilms

Bulk measurements of foulant layers, as opposed to specific measurements of chemical species, include the total areal mass as carbon of the foulant layer (mg-C.cm⁻²) and areal biomass of biofilm (CFU.cm⁻²). The total areal mass of the foulant layer on the control and 2-AI membranes were not statistically different when organic-only feed was used to foul the membranes. There was also no statistical difference in the total areal mass of the foulant layer between the control and 2-AI membranes when the cell-deposition-only feed was used (see Appendix Fig A.3). By contrast, the total areal mass of the foulant layer on the 2-AI membranes was significantly lower than on the control membranes when we used the biofouling-only (p=0.02, 35% decrease) and organic+biofouling (p=0.01, 54% decrease) feed solutions to foul the membranes (see Appendix Fig A.3). These results indicate that when biofilm formation is a mechanism of fouling and when all fouling mechanisms co-occur, the 2-AI incorporation significantly decreased the total areal mass of the foulant layer. When all fouling mechanisms co-occur, biofouling cannot be definitively determined to be the only fouling mechanism reduced because (a) the total mass of foulant is reduced substantially more (54% vs. 35% reduction) when all fouling mechanisms co-occur (organic+biofouling) versus when biofouling-only occurs and (b) the different types of fouling may affect each other when they are co-occurring. For example, less accumulation of sticky biofilm in 2-AI membranes may reduce organic fouling or cell-deposition.

Fig 5 shows the areal biomass of biofilm (CFU.cm⁻²) on control and 2-AI membranes after fouling with the two feed solutions. Compared to the control membranes, the 2-AI membranes inhibited biofilms by 95% (p=0.10) when we used the biofouling-only feed solution to foul the membranes. Although there was an order-of-magnitude less biofilm mass on the 2-AI membrane, the statistical confidence (as indicated by p-value) in the difference between the 2-AI membrane and control membrane was slightly low (90% confidence). This lower confidence was because one of the control membranes had 10x more biomass than the other control replicates, producing a higher uncertainty (e.g., as shown by the error bar) in the biomass value for the control membrane. When all fouling mechanisms co-occurred, (i.e., use of organic+biofouling feed solution to foul the membrane), the 2-AI membranes inhibited biofilms by 98% (p<0.001).

Fig 5.

Areal biomass (CFU.cm⁻²) on fouled control (black dotted bars) and 2-AI membranes (red striped bars). The 2-AI membranes inhibited biofilm formation by 95% (p=0.10) when we used the biofouling-only feed to foul the membranes. The 2-AI membranes significantly (p<0.001) inhibited biofilm formation by 98% when we used the organic+biofouling feed solution to foul the membranes. The bars represent the average areal mass of the biofilms from 6-11 replicate measurements, with 3-6 measurements per membrane, and error bars correspond to the standard error of all replicate measurements.

Overall, the results for total areal mass of the foulant layer and areal mass of biofilm indicate that there was substantially less biofilm formation on 2-AI membranes than on control membranes and that 2-AI incorporation decreased total fouling by inhibiting biofilm formation.

3.2. Membrane Performance under Fouling Conditions and Relationship to Fouling Inhibition

3.2.1. Water permeance

Fig 6 shows the change in water permeance of the control and 2-AI membranes over 75 hours of filtration under each fouling condition. The water permeance of all membranes dropped over time as fouling took place. Under all fouling conditions, the water permeance of the control membranes decreased by 22-40% within 75 hours. The decrease in water permeance of the 2-AI membranes after 75 hours of operation differed by fouling type. The water permeance of the 2-AI membranes dropped the most (29% on average) when we used the organic-only feed solution to foul the membranes and this drop was statistically similar to that of the control membrane (34% on average). When we used the other feed solutions (i.e., biofouling-only, cell-deposition-only, and organic+biofouling) to foul the membranes, the water permeance of the 2-AI membranes decreased significantly less over 75 hours (p<0.05) than for the control membranes. Specifically, for tests with the biofouling-only, cell-deposition-only, and organic+biofouling feed solutions, the water permeance of the 2-AI membranes compared with the control membranes decreased by 17% vs. 30%, 22% vs. 31% and 12% vs. 22%, respectively, after 75 hours of filtration. Overall, the decreased biofilm formation and lower surface affinity for polysaccharides and/or cell deposition on 2-AI membranes translated into significantly lower decreases in water permeance over time, including when all fouling mechanisms co-occurred.

Fig 6.

Normalized water permeance (A_t/A₀) over 75 hours of operation under each fouling condition for control (black squares) and 2-AI (red diamonds) membranes. Each point represents the average normalized water permeance from two separate cells. The error bars show the range of normalized water permeance between the membranes from two cells.

Although the water permeance decreases were significantly lower for 2-AI membranes, it is also important to consider the absolute water permeance at each time point, not just the change over time. Fig 7 shows that in all cases, the initial (i.e., at time zero) water permeances of 2-AI membranes were moderately lower than those of the controls, 11% lower on average. After 75 hours of fouling with the organic-only feed solution, the water permeance of 2-AI membranes and control membranes were similar, with the 2-AI membranes having a 1% higher water permeance on average. However, when biofilm formation was a mechanism of fouling (i.e., when biofouling-only and organic+biofouling feed solutions were used), the water permeance of 2-AI membranes was higher than that of control membranes after 15-55 hours of fouling. After 75 hours of fouling with the biofouling-only and organic+biofouling feed solutions, the water permeances of 2-AI membranes were higher, by 11% and 10% on average, respectively. The only case in which the water permeances of the 2-AI membranes were lower than those of the control membranes after 75 hours, was when we fouled the membranes with the cell-deposition-only feed solution, where the 2-AI membranes had a 6.5% lower water permeance than the control membranes. This lower final water permeance is likely due to the lower initial water permeance of the 2-AI membrane (20% lower than control). We note that even though the initial water permeances of 2-AI membranes were moderately lower than those of control membranes, we made no effort to optimize water permeance in 2-AI membranes. Future work could include optimization of the 2-AI grafting method (e.g., experimenting with carboxyl activating agents, protection/deprotection of 2-AI), grafting 2-AI onto different types of commercial membranes, evaluating post-modifications (e.g., solvent activation), and investigating additives to improve water permeance of the 2-AI membrane further. Therefore the 2-AI membrane fabrication procedure may be optimized to minimize or eliminate initial water permeance differences between 2-AI and control membranes. Overall, decreased biofilm formation and total fouling when mechanisms co-occurred on 2-AI membranes translated into higher absolute water permeance after 15-55 hours of operation for 2-AI membranes than for control membranes.

Fig 7.

Absolute water permeance over 75 hours of operation under each fouling condition for control (black squares) and 2-AI (red diamonds) membranes. Each point represents the average water permeance from two separate cells. The error bars (in some cases hidden by the symbols) show the range of water permeance between the membranes from two cells.

3.2.2. Solute Rejection

The conductivity rejection by membranes changed only slightly (<1%) during all fouling experiments, except during cell deposition experiments where rejection increased approximately 6.7% throughout the 75 hours of operation. Notably, the conductivity rejection between the control and 2-AI membranes was statistically similar (p<0.05) in all of the fouling experiments; the difference in conductivity rejection between the control and 2-AI membranes was 0.1% on average after 75 hours and the difference in conductivity rejection between duplicate membranes for any experimental condition was at maximum 1% different (see Fig A.4).

Fig 8 shows the rejection of organic carbon (TOC) by the control and 2-AI membranes under each fouling condition after 75 hours of operation. When we used the organic-only feed solution to foul the membranes, the control and 2-AI membranes rejected TOC similarly (1% difference). By contrast, when we fouled the membranes using the biofouling-only, cell-deposition-only, and organic+biofouling feed solutions, the TOC rejection by the 2-AI membranes was substantially higher (by 11 percentage points (p=0.10), 13 percentage points (p=0.05), and 12 percentage points (p=0.07), respectively) than by the control membranes. Therefore, the fouled 2-AI membranes, with independent fouling mechanisms or co-occurring fouling mechanisms, rejected TOC better than the fouled control membranes.

Fig 8.

Rejection (%) of organic carbon (TOC) in the feed by control (black dotted bars) and 2-AI (red striped bars) membranes after 75 hours of operation under the various fouling conditions. The bars are the average rejections calculated from four permeate sample measurements (duplicate measurements from two separate membrane cells) and duplicate feed measurements. The error bars represent the differences in the averages for each cell.

Overall, results indicate that under fouling conditions, including the most realistic condition of co-occurring fouling mechanisms, 2-AI membranes exhibited a substantially lower water permeance reduction compared with control membranes, similar conductivity rejection compared with control membranes, and substantially higher TOC rejection compared with control membranes.

4. Conclusions

We evaluated performance (i.e., fouling, water permeance, and solute rejection) of 2-AI membranes and corresponding 2-AI-lacking control membranes under conditions similar to what would occur during full-scale water treatment application, including (1) co-occurrence of all fouling mechanisms, (2) a natural feed water, (3) cross-flow configuration, and (4) operational parameters mimicking those at a full-scale plant. Performance was also evaluated under conditions that isolated individual fouling mechanisms.

Our experimental results support the following main conclusions:

• 2-AI membranes and (2-AI-lacking) control membranes fouled similarly when the fouling mechanism was organic matter accumulation,

• 2-AI membranes significantly inhibited (98%, p<0.001) and weakened (as indicated by PS/PN) biofilm, whether or not fouling also occurred by other mechanisms; the ability of 2-AI membranes to inhibit biofilms under multiple fouling conditions is an important finding,

• When we used the biofouling-only and organic+biofouling feed solutions to foul the membranes, fouled 2-AI membranes had a significantly lower change in water permeance over time, higher absolute water permeance (10-11% on average after 75 hours), and higher rejection of organics (by 11-13 percentage points after 75 hours), outperforming fouled control membranes,

• Conductivity rejection was similar for control and 2-AI membranes and was unaffected by fouling.

The results presented constitute the proof-of-concept for 2-AI membranes as high-pressure membranes with comparable water permeance and solute rejection to commercial RO/NF membranes but with significantly lower susceptibility to biofouling and biofouling associated performance changes. These results demonstrate that 2-AI membranes could be useful for decreasing overall fouling and improving overall performance in water treatment applications and other applications where biofouling and multiple fouling mechanisms co-occur (e.g., industrial applications).

Highlights.

• 2-AI membranes fouled less than controls and maintained conductivity rejection

• 2-AI incorporation significantly inhibited (98%) and weakened biofilm

• Biofilm inhibition by 2-AI caused higher water permeance (~10%) after 15-55 hours

• Biofouled 2-AI membranes rejected TOC significantly more (~10%) than controls

• Mass of fouling from cell deposition and organics was similar on 2-AI membranes and controls