Nanocomposite Polymeric Membranes for Organic Micropollutant Removal: A Critical Review

Abstract

The prevalence of organic micropollutants (OMPs) and their persistence in water supplies have raised serious concerns for drinking water safety and public health. Conventional water treatment technologies, including adsorption and biological treatment, are known to be insufficient in treating OMPs and have demonstrated poor selectivity toward a wide range of OMPs. Pressure-driven membrane filtration has the potential to remove many OMPs detected in water with high selectivity as a membrane’s molecular weight cutoff (MWCO), surface charge, and hydrophilicity can be easily tailored to a targeted OMP’s size, charge and octanol–water partition coefficient (K_ow). Over the past 10 years, polymeric (nano)composite microfiltration (MF), ultrafiltration (UF), and nanofiltration (NF) membranes have been extensively synthesized and studied for their ability to remove OMPs. This review discusses the fate and transport of emerging OMPs in water, an assessment of conventional membrane-based technologies (NF, reverse osmosis (RO), forward osmosis (FO), membrane distillation (MD) and UF membrane-based hybrid processes) for their removal, and a comparison to the state-of-the-art nanoenabled membranes with enhanced selectivity toward specific OMPs in water. Nanoenabled membranes for OMP treatment are further discussed with respect to their permeabilities, enhanced properties, limitations, and future improvements.

1. Introduction

Organic micropollutants (OMPs), such as endocrine disrupting chemicals (EDCs), pharmaceutical and personal care products (PPCPs), and other trace organic compounds (TrOCs), are an emerging concerns in the water supply of developed and developing countries alike. OMPs are defined as natural or synthetic compounds that can be present in water at trace levels (ng/L to μg/L).¹ Despite many emerging OMPs having still unknown impacts to ecology and human health, the increased OMP quantities have raised global concern about their toxicity, bioaccumulation, and persistence,^2,3 leading to updated environmental guidelines and laws in developed societies. The United States Environmental Protection Agency (US EPA) publishes a contaminant candidate list (CCL) of unregulated contaminants that potentially require future regulation for public water systems every 5 years.⁴ Canada identified certain substances including nonylphenol ethoxylates (NPEs) to be assessed on a priority basis.⁵ The European Union (EU) has updated a watch list of substances, such as several azole pharmaceuticals and pesticides, for union-wide monitoring in surface waters in 2020.⁶

Current water treatment processes, i.e., coagulation–flocculation,⁷ biological treatment,^1,8 advanced oxidation,^9,10 adsorption,^10,11 ultrafiltration (UF), nanofiltration (NF), and low pressure reverse osmosis (RO) membrane filtration,^12,13 can at best partially remove OMPs. Coagulation–flocculation employing inorganic salts or polymer agents can only eliminate some dissolved organic matter (DOM) and their OMP removal (15%–75%) is limited.³ Biological techniques such as conventional activated sludge (CAS) processes and membrane bioreactors (MBRs) are effective to remove nutrients, some simple organic molecules, and pathogens, but these biological processes are less effective toward recalcitrant OMPs.⁸ MBR gives higher removal (20%–60%) to OMPs than CAS due to MBR’s higher sludge retention time, higher biodiversity, and greater microorganism adaption^3,8 but is still inadequate for the growing concern for OMPs. Advanced oxidation processes (AOPs) such as the Fenton reaction, UV/H₂O₂ oxidation, and ozonation can degrade OMPs quickly and efficiently. However, these processes are not selective toward OMPs. Formation of a strong oxidant, hydroxyl radicals (·OH), demands high energy and chemical input. Furthermore, AOPs can form oxidation byproducts in water and can potentially transform OMPs into other toxic species,¹⁴ which is the subject of ongoing research. Adsorption has been widely used for aqueous OMP removal due to ease of operation and high efficiency (up to 99%),¹⁵ but activated carbon (AC) adsorption of OMPs faces early breakthrough, slow kinetics, and interference from other matrix substances. Also, AC production and regeneration can be costly.¹¹

Membrane-assisted technology has played an increasingly significant role for water purification and resource recovery. Nanofiltration (NF) and reverse osmosis (RO) processes typically provide adequate water flux and simultaneous rejection of organic molecules and multivalent salts. While rejection to organic solutes can be high (>90%), NF membranes (pore size 0.5–2 nm, molecular weight cutoff (MWCO) 200–2000 Da) cannot fully remove small-sized OMPs (e.g., <500 Da for some antibiotics and EDCs, 200–400 Da for most pesticides).¹⁶⁻¹⁸ While RO membranes have denser membrane structures and therefore generally higher OMP removal rates, uncharged OMPs can readily dissolve and diffuse through RO membranes.¹⁹ NF (6–10 bar) and RO (14–70 bar) processes are energy intensive due to high feed pressure requirements²⁰ and suffer from membrane fouling, resulting in decreased flux and decreased selectivity for OMP removal.

Microfiltration (MF) and ultrafiltration (UF) are more economical membrane filtration processes due to high permeability, low energy consumption, and easy automation.²¹ These membranes however poorly remove OMPs due to their large pore sizes. The MWCO of UF membranes (1–100 kDa), for example, greatly exceeds that of most OMPs (<1 kDa).²² Alternatively, combining the MF or UF process with other treatment measures (e.g., powdered activated carbon/ultrafiltration (PAC/UF) process and micellar-enhanced ultrafiltration (MEUF)) can enhance OMP removal. However, the additional units/agents applied largely increase water treatment plants’ operating costs and plant footprints.

The recent development of nanocomposite polymeric membranes is an attractive option to increase process intensification while achieving high OMP removal. These advanced functional membranes contain nanomaterial modifiers that enhance the polymeric membranes’ physicochemical properties (e.g., hydrophobicity, surface charge, pore size) to help adsorb/reject/degrade specific OMPs during filtration or within a membrane reactor. Nanocomposite polymeric membranes can be prepared by (1) coating nanomaterials onto the surface of conventional membranes, (2) integrating nanomaterials into the membrane matrix, (3) sandwiching nanomaterials between the membrane and a polymer thin film, or (4) embedding nanomaterials into a polymer thin film on the membrane.

Current review papers have described technological advances of OMP removal in wastewater,^14,23,24 with some focused on applying membrane-based technologies.^22,25−27 However, there is a dearth of reviews on the synthesis and application of state-of-the-art nanocomposite polymeric membranes for OMP treatment. Given a large increase in the number of publications in this field over the past 10 years (Figure 1), we provide a systematic overview and novel perspective of the research advancements on nanocomposite polymeric membranes and their applications for OMP removal from water to inspire researchers to better design next-generation membranes for environmental applications.

Figure 1.

Total number of publications (2011–2021) for organic micropollutants removal using nanocomposite polymeric membranes.

2. Occurrence and Transport of OMPs in Water

With increased anthropogenic activities, more and more substances have been recognized and monitored as OMPs. Typical OMPs include steroidal hormones, pesticides and biocides and their degradation products, plasticizers (e.g., phthalates), polyhalogenated compounds (PHCs) (e.g., polychlorinated biphenyls (PCBs) and perfluorinated chemicals (PFCs)), disinfection byproducts (DBPs), pharmaceutically active compounds (PhACs) (e.g., antibiotics and anti-inflammatory drugs), personal care products (e.g., cosmetics and fragrances), industrial chemicals (e.g., phenols and synthetic dyes), and combustion byproducts (e.g., polyaromatic hydrocarbons (PAHs)). A summary of representative OMPs, their physical properties, and measured concentrations in the aquatic environments is presented in Table 1.

Table 1. Commonly Studied OMPs in Water and Their Physical Properties.

			Physical Properties					Reported Concentrations (ng/L)
Category	Subclass	Example	MW (Da)	Sa (mg/L)	pKa	Log Kow	ref	Surface waterc	Effluentd	ref
Endocrine disrupting chemicals (EDCs)	Steroid hormones	estrone (E1)	270	30	10.34	3.13–3.43	(38)	1.8	0.1–15.3	(39, 40)
		17-β-estradiol (E2)	272	13	10.46	2.69–4	(38)	0.56	0.6–5.8	(39, 40)
		estriol (E3)	288	13	10.38	2.81	(38)	3.7	0.4–6.8	(39, 40)
		testosterone	288	23.4		3.32	(41)	0.01	ND	(39, 42)
		17-α-ethinylestradiol (EE2)	296	4.7–19	10.4	3.67–4.2	(38)		<106	(43)
		progesterone	314	8.81		3.87	(41)	<0.01	ND	(39, 42)
	Pesticides	carbendazim	191	8	4.2	1.52	(44)
		simazine	202	6.2	1.6	2.2	(45)	0.2–488.8	26.3	(46, 47)
		atrazine	216	34.7	1.7	2.7	(45)	0.9–5170	4.2	(46, 47)
		diuron	233	42	2.68	2.7	(22, 45)	1.4–1362	61.7	(46, 47)
		diazinon	304	40	2.6	3.81	(3)	0.6–276	21.4	(46, 47)
	Phenolic estrogens	octylphenol (OP)	206	12.6	10.39	4.12	(48)	<226		(49)
		nonylphenol (NP)	220	6.35	10.25	5.99	(3)	<2760	13.4–471.6	(49)
		bisphenol A (BPA)	228	120	9.6	3.32	(3)	<2470	<1840	(43, 49)
	Phthalates	dimethyl phthalate (DMP)	194	4000b		1.6	(50)		0.062	(50)
		dibutyl phthalate (DBP)	278	15b		4.83	(51)	<4430	0.54	(50, 51)
		diethylhexyl phthalate (DEHP)	391	<1b		8.71	(51)	<5000	1.6	(50, 51)
	Polychlorinated biphenyls (PCBs)	2-chlorobiphenyl	189			4.56	(52)
		4,4′-dichlorobiphenyl	223			5.28	(52)
	Perfluorinated chemicals (PFCs)	perfluorooctanoic acid (PFOA)	414	3400	2.5	6.28	(53)	0.21–4.2	255	(47, 54)
		perfluorooctanesulfonate (PFOS)	500	519	–3.27		(55, 56)	0.04–6.9	62.5	(47, 54)
Pharmaceutically active compounds (PhACs)	Antibiotics	sulfamethoxazole	253	610	5.6	0.89	(41)	8.3	280	(47, 57)
		sulfamethazine	278	430	7.4, 2.65	0.28	(58)		12	(42)
		norfloxacin	319	161,000	5.77	–1.03	(59, 60)		1.6–260.4	(40)
		ciprofloxacin	331		10.58, 8.7, 6.14, 3.01	0.28	(61)	5.4–448.1	67	(42, 46)
		enrofloxacin	360		9.86, 7.59, 6.19, 3.85	0.7	(62)		0.4–2.6	(40)
		cefadroxil	363		9.64, 7.37, 2.48	–0.08	(62)
		amoxicillin	365		9.6, 7.4, 2.4		(63)	<2.08–223	20–1340	(64)
		cephalexin	365		6.88, 2.56	0.076	(61)
		tetracycline	444	231	3.3	–1.3	(41)	1.1–4278.7	0.8–6000	(64)
		erythromycin	734	459	8.9	3.06	(3, 22)	0.84	6.5–2350	(57, 64)
	Analgesic and anti-inflammatory drugs	acetaminophen	151	14,000	9.4	0.46	(44)	1.7	79	(42, 57)
		ibuprofen	206	21	4.9	3.97	(65)	22	460	(42, 57)
		naproxen	230	51.1	4.2	3.18	(3, 22)	1.4	26.7	(47, 57)
		diclofenac	296	2.37	4.15	4.51	(12)	8.7	49.5	(47, 57)
	Blood lipid regulators	carbamazepine	236	17.7	<2–2.45, 7	2.18–2.93	(58)	4.9	832	(47, 57)
		bezafibrate	362	1.55	3.83	3.97	(3)		<4800	(43)
	Stimulants	caffeine	194	21,600	10.4	–0.07	(66)	160	191	(47, 57)
		cocaine	303		8.6	2.3	(67)
	β-Blocker	atenolol	266	300	9.6	0.16	(3)		0.8–515.7	(40)
		metoprolol	342	402	14.09	1.88	(3)	5.4	9.5–335.9	(57)
Personal care products (PCPs)	Antiseptics/Disinfectants	2-nitrophenol	139	2100b	7.2	1.76	(68)
		4-nitrophenol	139	16000b	7.15	1.91	(68)
		triclosan	290	10	7.9	4.76	(66)	1.5–534	74.8	(46, 47)
		triclocarban	316		11.4	4.9	(22)	13	<5860	(43, 57)
	Preservatives	methylparaben	152		8.4	2	(69)	<1062	<262	(70, 71)
		propylparaben	180		8.2	3	(69)	<3142	<231	(70, 71)
		benzylparaben	228		8.2	3.6	(69)	<3.93	<2.9	(70, 71)
	Fragrances	galaxolide	258			5.9	(72)
		tonalide	258			5.7	(72)
Other organic pollutants	Disinfection byproducts (DBPs)	dichloroacetic acid (DCAA)	129	1,000,000	1.26	0.92	(12)	<2.3	3.3–9.3	(73)
		trichloroacetic acid (TCAA)	163	44,000	0.51	1.33	(12)	<4.2	3.4–28	(73)
	Organic dyes	methylene blue (MB)	320	43,600b	5.6		(74, 75)
		methyl orange (MO)	327	5000	3.47		(76)

^aValues for solubility (S) in water were determined at 20 °C unless otherwise stated.

^bTemperature at which S was measured is 25 °C.

^cRefs (39 and 57): mean concentrations based on 217 surface water samples across 31 provinces in China. Ref (46): concentration for surface waters from seven provinces in China, 25 states in the US, and seven countries in EU. Refs (49 and 51): concentration for surface waters from some East Asian and European countries. Ref (54): concentrations based on 40 river water samples in Northern Europe. Ref (64): mean concentrations of surface waters in Asia, North America, and Europe. Ref (70): concentration based on multiple river waters in China. Ref (73): concentrations based on 51 surface waters from Beijing and Wuhan, China.

^dRef (40): concentrations of effluents from 14 WWTPs in China. Ref (42): mean concentrations of effluents from 50 WWTPs in the US. Refs (43, 64 and 71): concentrations of effluents from some Asian, North American, and European countries. Ref (47): mean concentrations of effluents from 90 WWTPs in EU.. Ref (50): mean concentration of effluents from 15 WWTPs in Europe. Ref (73): concentrations of effluents based on five WWTPs from Beijing and Wuhan, China.

OMPs may originate from both point sources and diffuse sources, and they can be emitted into domestic, industrial, agricultural, and hospital effluents.^22,28−30 For example, antibiotics overuse for medical treatments and livestock agriculture, especially recently in China, poses significant threats to ecosystems and human health.³¹ OMPs have been detected in surface water,^24,32 groundwater,³³ seawater,³⁴ and drinking water.^35,36 Wastewater treatment plant (WWTP) effluent discharge is the dominant pathway for OMP emission to surface waters and potable water supplies due to insufficient removal.^24,35,36

OMP pollution severity in surface waters can be influenced by effluent volume and effluent concentration (as impacted by regional usage and treatment efficiency).³² The fates of some OMPs, including many pharmaceuticals, are subject to changes through dilution, partition/adsorption, degradation, biotransformation, and photolysis.³² Other OMPs such as PFCs and PCBs are well known to be extremely persistent in surface water. Groundwater is usually less impacted by OMP pollution and accordingly has been less researched.

OMPs in groundwater primarily result from landfill and sewer leachates, and the infiltration of OMP-polluted water (e.g., from septic systems) is primarily controlled by both the adsorption/partition to soils as well as soil microbial oxidative processes.^24,33 OMPs’ fates and transfers into subsurface soil and groundwater are also impacted by aquifers’ hydraulic conditions, and soil properties.³³ OMPs with low octanol–water partition coefficient (K_ow) values (log K_ow < 3) such as some pharmaceuticals, pesticides, and DBPs likely have high mobility in soils, low affinity to organic sediments, and thus higher probability of occurring in groundwater. Adsorption/partition can only retard OMP transport, whereas microbial degradation reduces the concentrations of OMPs in groundwater.³⁷ The different fates and transports of OMPs through the subsurface have implications for policy makers, environmental regulatory bodies, and water operators to manage recalcitrant OMPs such as phenols in some water sources including drinking water wells.

3. Conventional Membrane-Based Processes for OMP Removal

While an exhaustive review on conventional membrane-based routes to remove OMPs is beyond the scope of this review, this section briefly presents the current performance and challenges of conventional membranes in treating OMPs. This provides mechanistic insights to motivate the use of nanomaterials within membranes.

3.1. Advanced Membrane Processes

OMPs in real-world water and wastewater treatment plants have been separated by nanofiltration (NF)), reverse osmosis (RO), forward osmosis (FO),⁷⁷ and membrane distillation (MD),⁷⁸ but the solute transport/removal mechanisms under different operating conditions and OMP properties are still not entirely understood. Also, none of these methods demonstrate full removal capability.

3.1.1. Nanofiltration (NF) and Reverse Osmosis (RO)

Tight NF and RO membranes demonstrate OMP retentions commonly above 80%,²⁷ whereas loose NF membranes do not retain OMPs as effectively. Uncharged, nonadsorptive OMPs such as some antibiotics⁷⁹ have been predominantly removed by size exclusion (steric hindrance), and as such, these OMPs’ removals are governed by their molecular weight (MW). The polarity (indicated by dipole moment) and hydrophobicity of uncharged OMPs, including some EDCs and PPCPs, impact their adsorption to NF and RO membranes, influencing their separation.^13,80 OMP speciation caused by pH change has been shown to enhance electrostatic repulsion with negatively charged membrane surfaces, which can increase the initial rejection but may decrease adsorption-mediated removal.^80,81 Generally, small-charged OMPs, such as dyes and some PhACs, are more affected by charge effects than size exclusion.⁸² Although RO membranes provide greater OMP removal due to their dense membrane structure, they are not an absolute barrier.⁸¹ Even conventional RO membranes struggle to remove low MW uncharged and nonadsorptive OMPs such as some drugs (2-naphthol, phenacetine, and primidone).¹²

Coexisting compounds in feed waters can affect OMP removal through solute–solute interactions or solute–membrane charge interactions.⁸³ For instance, natural organic matter (NOM) can complex with neutral OMPs to enhance size exclusion across loose NF membranes and produce membrane surface fouling that enhances separation by charge repulsion to negatively charged OMPs.²² Alternatively, Ca²⁺ and Mg²⁺ can neutralize membrane surface charges which reduces the membranes’ interactions with charged OMPs.²² This effect primarily impacts loose NF membranes rather than tight NF and RO membranes as the former are less able to separate OMPs by size exclusion or diffusive processes.

3.1.2. Forward Osmosis (FO)

FO processes use spontaneous water transport by an osmotic pressure difference from the less concentrated feed solution to a highly concentrated draw solution (DS) across a semipermeable membrane, achieving feedwater separation from the solutes. FO has advantages of low energy consumption, low fouling potential, high recovery, and simplicity⁸⁴ and has wide applications in desalination, wastewater reclamation, and concentration of polluted water (e.g., landfill leachate).⁸⁴ FO applications for OMP removal are rare,^77,84−86 and OMP removal efficiency is affected by characteristics of FO membranes and pollutants.

Typically, FO rejects charged OMPs well, where Hancock et al.⁷⁷ obtained different rejections for TrOCs that were charged (80%–98%) and uncharged (40%–90%) at bench scale. The rejection increased with MW due to increased diffusive hindrance.^77,84 Hydrophobic adsorption is also among the governing mechanisms for short-term OMP removal.⁸⁴ Greater OMP K_ow would increase adsorption to membranes, reduce the diffusion of trace solutes across membranes, and thus achieve higher removal.⁸⁴ Foulants on FO membranes can potentially affect interactions with OMPs in that the fouled FO membranes have reduced mass transport capacities, transformed hydrophilicities, and altered surface charges, promoting the removal of like-charged OMPs by charge repulsion and the removal of neutral hydrophobic OMPs by adsorption.⁸⁵

For most FO applications, DS recovery and solute leakage prevention are essential at industrial scale. A high total dissolved solids (TDS) DS will achieve high flux and high water recovery but incurs high cost for the DS concentrating process. When dewatering OMP-impaired water, future studies should explore new operating conditions (e.g., hybrid FO-RO process⁷⁷) and develop more cost-effective DS (e.g., polyelectrolytes⁸⁷) to improve removal and save energy.

3.1.3. Membrane Distillation (MD)

Membrane distillation (MD) is a thermally driven membrane process in which separation occurs by phase change at the membrane active surface. MD allows water vapor and volatile molecules to pass through a microporous hydrophobic membrane from a hot aqueous solution, in which the driving force is the vapor pressure caused by the temperature differences across the membrane surface from the feed side to the distillate side.⁷⁸ MD process could achieve high rejection performance (for nonvolatile solutes) and only requires mild operating temperatures (40–80 °C).⁸⁸ MD has some potential for removing OMPs from wastewater.⁸⁹ MD separation efficiency is related to the solutes’ volatilities and their hydrophobicities and the membranes’ properties. OMP rejections by direct contact membrane distillation (DCMD) have been shown to be governed by solutes’ volatilities rather than hydrophobicities.^78,90 Nonvolatile hydrophilic TrOCs (pK_H > 9) can be effectively rejected in the feed, whereas moderately volatile (pK_H < 9) and hydrophobic compounds can enter the permeate by evaporation or adsorption to the membrane, reducing selectivity.⁷⁸ One problem however is that some molecules may be susceptible to thermal degradation, leading to the permeation of degradation byproducts.⁸⁹ Overall, MD for OMP removal is still an emerging technology demanding further large-scale optimization and commercialization due to membrane wetting and fouling issues.⁸⁸

3.2. Membrane-Based Hybrid Processes

Advanced membrane processes are often expensive and prone to fouling; hence, some hybrid processes in which UF membranes are coupled with adsorption or coagulation have also been investigated to remove OMPs.

3.2.1. Ultrafiltration with Adsorption

The single UF process is inadequate to retain most OMPs due to their large pore sizes, and hence, the UF/adsorption hybrid process can be utilized (Figure 2). Permeate post-treatment with adsorption units is an alternative to enhance OMP removal. Gerrity et al.⁹¹ effectively reduced TrOC concentration in a potable water stream when setting biological activated carbon (BAC) filters post UF and ozone/H₂O₂ treatment. An adsorption column can also precede the UF unit to adsorb low MW OMPs and reduce organic loads to mitigate UF membrane fouling. Acero et al.⁹² used powdered activated carbon (PAC) adsorption as a pretreatment step before a cross-flow UF process to remove WWTP OMP-rich effluents. Although less adopted, Löwenberg et al.⁹³ used a mix tank allowing contact between PAC and PhACs, followed by adding FeCl₃ coagulant. The treated water underwent a dead-end UF process and obtained 60%–95% removal for PhACs. Generally, only limited types of OMPs, such as PhACs,^92,93 phenols,⁹⁴ dyes,⁹⁵ and pesticides⁹² have been investigated using UF/adsorption processes. There is demand to simplify the configuration, better recirculate the sorbents, and lessen membrane fouling and pore blockage to minimize cost and operational complexity.

Figure 2.

Activated carbon adsorption-ultrafiltration (AC/UF) hybrid system for organic micropollutant removal.

3.2.2. Micellar-Enhanced Ultrafiltration (MEUF)

Micellar-enhanced ultrafiltration (MEUF), introduced in the 1980s,⁹⁶ combines surfactants in the feed with membrane separation. Above a surfactants’ critical micellar concentration (CMC), UF membranes retain the surfactant micelles with trapped pollutants in the feed by size exclusion, and only the small unbound solutes and some surfactant monomers can pass to the permeate. MEUF performance is impacted by membrane type, operating conditions, solution conditions, surfactant type, and target compound characteristics.⁹⁷ MEUF for removing aqueous pollutants is pictorially shown in Figure 3.

Figure 3.

Basic principles for micellar enhanced ultrafiltration (MEUF) for organic micropollutant removal.

MEUF has been traditionally investigated for heavy metal ions and anions removals,⁹⁸ but OMP removal was also reported, including PhACs,^99,100 phenols,¹⁰¹ dyes,¹⁰² and pesticides.¹⁰³ Ionic surfactants can remove charged OMPs.¹⁰²

Generally, high CMC surfactants are not beneficial for OMP retention because of greater surfactant monomer losses into the permeate. CMCs can be lowered by adding salts or ionic/nonionic surfactants. Cationic surfactants (e.g., cetryltremethyl ammonium bromide (CTAB) and cetylpyridinium bromide (CPB)) have good commercial availabilities, large micellar sizes, and lower CMC values than anionic surfactants and are thus promising for OMP removal in MEUF.^99,100 Nonionic surfactants have low CMCs but cannot form ion-pair complexes with pollutants,⁹⁷ which remove noncharged OMPs based on hydrophobic solubilization. Doulia and Xiarchos¹⁰³ reported alachlor removal using nonionic surfactant-assisted MEUF. Recently, gemini surfactants have aroused increasing interests due to low CMC and promising organic solubilization effects.¹⁰¹ MEUF has the potential to remove OMPs but has not been widely commercialized. Future studies should improve the recovery of surfactants and organic compounds in the permeation.⁹⁸

4. Nanocomposite Polymeric Membranes for OMP Removal

Nanocomposite polymeric membranes feature significant functional changes through surface or bulk modifications of polymer substrates by nanofillers, often providing higher OMP removal than traditional composite and thin film composite (TFC) membranes. Composite membranes can either be surface modified or bulk modified (nanocomposites), while TFCs can either be modified within the thin film or below the thin film as shown in Figure 4. OMP separation by nanocomposite polymeric membranes is dictated by fabrication conditions (e.g., modifier type, position, and loading), filtration conditions (e.g., applied pressure, flow rate, feed concentration), solution (e.g., pH and coexisting compounds), and pollutant characteristics (e.g., MW, pK_a, and K_ow). These membranes can be categorized by the separation mechanisms involved, (e.g., size-exclusion, adsorption, photo- and electro-catalysis, charge exclusion, and chemical-assisted membrane separations), which are discussed below.

Figure 4.

Typical nanocomposite polymeric membranes. Membranes formed either with (a) nanoparticles added to membrane surfaces or (b) nanoparticles blended into the membrane bulk, and thin film nanocomposites (TFNs) formed either with (c) nanoparticles sandwiched between the support membrane and the thin film (TF) or (d) nanoparticles embedded within the bulk TF phase.

4.1. Surface-Modified Nanocomposite Membranes

4.1.1. Fabrication

Surface-modified nanocomposite membranes, known as surface-located membranes,¹⁰⁴ have been rapidly developed for OMP separation over the past decade. Common procedures to modify membrane surface by nanoparticles (NPs) include coating and deposition,^65,105,106 chemical bonding and grafting,^107−109 and layer by layer (LBL) assembly.¹¹⁰ Pressure deposition or dip coating the membrane in a precursor solution is the easiest approach to introduce nanomodifiers onto the membrane surface. Pressurized filtration has been used to coat carbon-based NPs, including carbon nanotubes (CNTs),^44,65 carbon nanofibers (CNFs),¹¹¹ cellulose nanocrystals (CNCs),¹¹² graphene oxide (GO),¹¹³ and reduced graphene oxide (rGO).^44,114 Of them, two-dimensional (2D) nanomaterials such as GO and rGO have emerged as popular coating materials to make adsorptive membranes due to ultrathin layers, high surface areas, and good chemical stabilities.^44,115 rGO can also make membrane conductive layers for electrochemical degradation of various OMPs.^110,116 Nonetheless, challenges exist as GO-coated membranes show poor stability under cross-flow conditions due to GO swelling.¹¹⁷ Pristine GO- or rGO-coated membranes have low flux due to short d-spacings of GO and rGO laminates (0.8 and 0.35 nm, respectively).¹¹⁸ This dense membrane structure would greatly hinder water transport leading to extremely low water permeation due to strong capillary forces in the compact nanochannels.¹¹⁹ While dense graphene membranes are effective at rejecting OMPs, the extremely low water fluxes through these membranes limit their use for most applications requiring OMP removal. Additives have been intercalated to expand the spacing between nanosheets, including organic spacers (e.g., CNTs,^44,115,116 graphitic carbon nitride (g-C₃N₄),¹²⁰ covalent organic frameworks (COFs),¹²¹ and chitin nanoycrystals (ChNC)¹²²) and inorganic spacers (e.g., metal organic frameworks (MOFs),^123−125 transition metal dichalcogenides (TMDs),^106,117,126 attapulgite nanorods (APTs),¹²⁷ halloysite nanotubes (HNTs),¹²⁸ MgSi,¹²⁹ SiO₂,¹³⁰ and Fe₃O₄¹³¹). For instance, intercalating CNT¹¹⁵ and COF material (COF-1)¹²¹ increased GO interlayer spacings to 1.09 and 1.03 nm, respectively. These modifications greatly enhanced the water permeation and provided high rejection to water-soluble dyes (>95%).

Recently, with a deeper understanding of inorganic 2D nanosheets, including MXene,¹³² hexagonal boron nitride (h-BN),¹³³ and MoS₂,^106,126 they have been used to modify membrane surfaces. Tong et al.¹³² developed MXene and tannic acid (TA)–metal complex surface-coated polyvinylidene fluoride (PVDF) membranes, giving high selectivity toward hydrophobic OMPs. Hafeez et al.¹³³ surface modified PVDF membranes with poly(ethylene glycol) (PEG)-grafted amine-functionalized h-BN (BN(NH₂)), leading to superhydrophilicity and subsequently a high water flux (>840 L·m^–2·h^–1·bar^–1). Ma et al.¹⁰⁶ intercalated MoS₂ into GO nanosheets through van der Waals interactions, which supported the membrane robustness at a wide range of pH conditions (3–11) while allowing high water flux and strong sieving of dyes.

Stabilizing NPs onto membrane surfaces during coating is a major challenge, where bioinspired chemistry has gained much attention to prepare OMP removal membranes. Mussel-inspired protein polydopamine (PDA), prepared by self-polymerization of dopamine (DA), has been widely adopted to stabilize membrane NPs due to strong adhesive forces. The reaction is favored in DA-containing Tris-HCl buffer solution at pH 8.5. PDA-assisted coating has been used to stabilize Au-TiO₂,¹³⁴ rGO-Cu,¹⁰⁵ and MOFs¹³⁵ on membranes via dip coating or pressurized filtration onto membrane surfaces. PDA codeposition with polyethylenimine (PEI) through interfacial cross-linking is an effective way to improve the surface hydrophilicity and introduce positive charges.^109,136 Besides, plant polyphenols, such as tannic acid (TA) and gallic acid (GA), are strong adhesives and cheap cross-linkers to bind thiol and amino groups. Chen et al.¹³⁷ investigated dye removal using hydrophilic membranes prepared by cross-linking TA with PEI via a Michael addition or Schiff base reaction and codepositing it with HNTs onto PVDF membrane. Further, TA could form coordination complexes with metal ions. Li et al.¹³⁸ modified PAN membrane surfaces via doping TA-aminopropyltriethoxysilane (APTES) with Fe³⁺-mediated GO sheets to remove aqueous OMPs. GO nanosheets can also be cross-linked by other cross-linkers such as ethylenediamine, glutaraldehyde, PEI, dicarboxylic acids, and borate,^113,139 which improves the membranes’ stabilities and permeabilities. These strategies provide new perspectives for obtaining stable, multifunctional, industrially practical membranes for OMP removal.

4.1.2. Application for OMP Removal

Surface modification by NPs enhance OMP removal, with additional benefits including formation of surfaces that are antifouling, chlorine resistant, and more selective for solutes. Surface-modified membranes for OMP removal are summarized in Table 2. Impacts of surface-anchored NPs to OMP removal include increased surface area for adsorption and tuned surface charge and pore size, and they provide catalytic properties. Oftentimes, multiple OMP removal mechanisms occur simultaneously.

Table 2. Surface-Modified Nanocomposite Membranes for Removing OMPs in Water.

Modifier	Nanomaterial	Substrate	Additive	Pore size/MWCO	Pollutant	Feed concentration (ppm)	Removal (%)	Removal mechanism	ref
Organic	GO-Fea	PAN	TA-APTES	0.56 nm	(1a) naphthalene, 6-methoxytetralin, (1b) pyrene, (1c) bisphenol A, tetracycline hydrochloride, (2) methylene blue, (3) methyl orange	(1a, 2, 3) 10, (1b) 1, (1c) 5	(1a, 1b, 1c) > 60, (2) 99.5, (3) 90	(1a, 1b, 1c) size exclusion, (2) adsorption, (3) electrostatic repulsion	(138)
	GO/COF-1	PAN		1.5 nm	Congo red, methylene blue, reactive black 5, direct red, chrome black T	200	>98	size exclusion, electrostatic repulsion	(121)
	GO/chitin nanocrystal (ChNC)	CA	PDA		methylene blue, Congo red	20	99.6, 98.3	adsorption	(122)
	GO-Feb/rGO	PTFE		1000 Da	florfenicol	1	90	electro-Fenton oxidation	(110)
	GO/CNT	CA			Congo red, methyl blue	100	98.7, 94.1	adsorption	(115)
	GO/N-CNT/O-g-C3N4	PC			rhodamine 6G	10	∼80	photocatalytic degradation	(147)
	GO/sodium alginate (SA)	PVDF			direct red 80, Congo red, methylene blue	20	99.8	size exclusion, electrostatic repulsion	(139)
	rGO/CNT-Feb	PTFE		1.29 nm	(1) amoxicillin, ampicillin, cefalexin, ofloxacin, sulfadiazine, sulfamethoxazole, florfenicol, (2) chloramphenicol, thiamphenicol, atenolol, carbamazepine	1	(1) 79–100, (2) 71–82	(1) electrostatic repulsion, electro-Fenton oxidation, (2) adsorption, size exclusion, electro-Fenton oxidation	(116)
	rGO/MWCNT	PVDF			acetaminophen, caffeine, carbendazim, triclosan	1	>76	adsorption	(44)
	rGO/g-C3N4	CA	PDA		rhodamine B	10	98.5	size exclusion, photocatalytic degradation	(120)
	SWCNT	PVDF		>10 nm	triclosan, ibuprofen, acetaminophen;	1	90, 59, 62	adsorption	(65)
Organic/Metallic	GO/TiO2	PSF	PDA/PEI	0.87 nm, 978.8 Da	crystal violet, safranine T, eriochome black T, alizarin yellow GG	100	99.99, 96, 99.99, 89.4	size exclusion, electrostatic repulsion	(109)
	GO/NH2-MIL-88B	Nylon	PAA		(1, 2) methylene blue, (1) Congo red, crystal violet	20	(1) > 97, (2) 98.79	(1) adsorption, size exclusion, (2) adsorption, photo-Fenton degradation	(148)
	GO/MIL-88A	PVDF			methylene blue, rhodamine B, methyl orange	10	>98	size exclusion, photo-Fenton degradation	(124)
	GO/attapulgite nanorods (APT)	PC			rhodamine B	7.5	>99.9	adsorption, size exclusion	(127)
	GO/SiO2	CA			methyl blue	10–40	>95	adsorption, size exclusion	(130)
	GO/MoS2	CA			Congo red, direct red 80, methylene blue, rhodamine B	20	>99	size exclusion	(106)
	GO/MoS2	PVDF	PVA		Congo red, methylene blue, methyl orange, rhodamine WT	10	99.6, 97.4, 96.3, 94.6	size exclusion	(126)
	GO/WS2	Nylon			rhodamine B, methylene blue	10	97.7, 96.3	size exclusion	(117)
	GO/NH2-Fe3O4	PVDF	PDA		(1) Congo red, methyl orange, (2) methylene blue	100	98, 75, 70	(1) size exclusion, electrostatic repulsion, (2) adsorption	(131)
	rGO/MgSi	PAN		1.27 nm, 536 Da	acid brilliant blue, chrome blue-black R, methyl orange	100	98.2, 95.2, 73.4	size exclusion, electrostatic repulsion	(129)
	rGO/Cu	PAN	PDA	1510 Da	Congo red, direct red 23, reactive blue 2	500	>98	size exclusion, electrostatic repulsion	(105)
	rGO/UiO-66	PC			rhodamine B	10	95	size exclusion	(123)
	rGO/Ag3PO4	PVDF	PDA		methylene blue	30	99.1	adsorption, photocatalytic degradation	(146)
	rGO/SiO2	PVDF	PDA		methylene blue	10	99.8	adsorption	(151)
	rGO/HKUST-1	CA	PDA		methylene blue, Congo red	40	99.8, 89.2	adsorption, size exclusion	(125)
	rGO/HNT	CA	PDA		methylene blue, Congo red	100	>99	adsorption, size exclusion	(128)
	prGO/UiO-66-(COOH)2	Nylon			Congo red	10	>97.2	electrostatic repulsion	(152)
	CNT/ZnO/TiO2	CA			acid orange 7	50	>95	adsorption, photocatalytic degradation	(145)
	SWCNT/DWCNT/nZVI	PES			methyl orange	81.8	87.3	adsorption, electrocatalytic reduction	(144)
	MWCNT/TiO2	CA			carbamazepine, ibuprofen, acetaminophen	10	80, 45, 24	adsorption	(153)
	MWCNT/nZVI	PTFE		41 nm	metoprolol	2	97	adsorption, electrochemical oxidation	(154)
Metallic	Mxene	PVDF	TA-Fe	1.3 nm	4-hydroxybenzoic acid, cinnamic acid	5	66, 53	size exclusion	(132)
	h-BN	PVDF		1.4–3 nm	methylene blue	5	98	size exclusion	(133)
	UiO-66	GO/PAN	PDA	2.88 nm	methyl orange, tetracycline hydrochloride	10	94.84 95.5	size exclusion, electrostatic repulsion	(135)
	HNT	PVDF	TA/PEI	45.2 nm	direct blue 14, direct red 28, direct yellow 4	100	96, 92.8, 90.7	adsorption, size exclusion	(137)
	PCu2W11	PVDF		200–750 nm	tetracycline	300	88.6	adsorption	(107)
	nZVI	PVDF			2-chlorophenol	10	91.63	adsorption, reductive degradation	(150)
	Au/TiO2	PVDF	PDA		tetracycline	10	92	photocatalytic degradation	(134)
	Ag/TiO2	PAN	PDA/PEI		Congo red, reactive orange 16, reactive black 5	200	99.6, 96.2, 99.5	size exclusion, photocatalytic degradation	(136)
	ZIF-8	PAN	PEI		acid fuschin, methyl orange, methyl blue, Congo red	100	94.4, 81.2, 99.6, 99.2	adsorption, size exclusion	(108)

^aFe³⁺ was introduced to the edge of GO sheets to complex with TA-APTES.

^bFe²⁺ served as the Fenton catalyst and was grafted into carboxylic CNT or rGO through electrochemical reduction of −COOFe³⁺.

4.1.2.1. Impact of Adsorption, Charge Interaction, and Size Exclusion

Many nanosorbents are characterized by high surface areas that can increase the adsorption sites in membrane matrices, which enhances OMP removal in loose structured MF or UF membranes, by reducing the free volume present in these high porosity membranes. For example, monolayer GO and single-walled carbon nanotubes (SWCNTs) have calculated surface areas up to 2400¹⁴⁰ and 1315 m²/g,¹⁴¹ respectively. Wang et al.⁴⁴ used multiwalled carbon nanotube (MWCNT)/rGO surface-modified PVDF microfiltration membranes to treat acetaminophen, caffeine, carbendazim, and triclosan and obtained >76% rejection through adsorption.

Huang et al.¹¹⁵ found that CNT π–π bonds and hydrophobic interaction enhanced aqueous dye adsorption when assembled with GO on cellulose acetate (CA) UF membrane surfaces. At only 1 bar transmembrane pressure (TMP), Congo red (CR) and methyl blue reached 98.7% and 94.1% removal, respectively. Wang et al.⁶⁵ found that SWCNTs coated at PVDF membrane surfaces favored adsorption of triclosan through π–π donor–acceptor interaction and acetaminophen through hydrogen bonding at neutral pH, respectively. Nanomaterials could alter membrane surface charges to enhance adsorption by electrostatic attraction. Ou et al.¹²² established a dopamine-modified chitin nanocrystal (D-ChNC)/GO coating on CA membranes, and charged dyes achieved >98% removal while filtered at 2 bar TMP. Adsorption of cationic dye methylene blue (MB) and anionic dye CR were, respectively, favored by negatively charged GO and positively charged ChNC. Electrostatic repulsion may play a role when OMPs and membrane surfaces carry opposite charges. Fang et al.¹³⁵ produced a MOF (UiO-66)/PDA thin selective layer atop the GO-modified polyacrylonitrile (PAN) NF membrane. The high electronegativity of UiO-66 and GO and PDA layer’s sieving (pore size 2.88 nm) contributed to repulsive exclusion of anionic dye methyl orange (MO) (removal >94%). Zhang et al.¹²¹ investigated GO/COF-1 nanocomposite-coated PAN membrane for dye separation, which led to >95% removal of anionic dyes. Here, negatively charged GO enhanced membranes’ selectivities, whereas the intercalated COF between GO nanosheets had a 0.33 nm spacing, much smaller than that of dyes (>1.5 nm), while allowing water molecules (∼0.28 nm) to pass through. In both works, low TMP (3–4 bar) was needed to reach high removal, much lower than normal NF or RO operating pressures. Further, weakening hydrophobic interactions between solutes and membrane surfaces can enhance size exclusion. For example, MXene-TA decreased the PVDF membrane surface hydrophilicity and pore size (1.3 nm), thereby increasing the separation of hydrophobic 4-hydroxybenzoic acid and cinnamic acid at 2 bar TMP.¹³²

4.1.2.2. Impact of Electro-Catalytic Degradation

While surface-coated nanomaterials give a more purified permeate through physical interactions, OMPs are concentrated in the retentate or at the membrane surface. Membranes coupling OMP separation and degradation are novel research areas that promise to permanently remove contaminants from solution offering the possibility of both high conversion of OMPs and high selectivity.

Electrically conductive membranes (ECMs) have gained great interest to degrade OMPs at membrane surfaces under an applied voltage. CNT and rGO are the most common conductive coating nanomaterials. Liu et al.¹⁴² and dos Santos Cunha et al.¹⁴³ developed a CNT coating on polytetrafluoroethylene (PTFE) membranes for electrochemical oxidation of trace levels of the antibiotic tetracycline and estrogenic compounds (E2 and EE2) using Na₂SO₄ as an electrolytic solution under a positive cell voltage applied by an external DC supply. More than 90% of these three compounds was shown to be removed by the combination of membrane adsorption and electro-oxidation. The high oxidative capacities could be explained by the formation of oxidant species including hydroxyl radicals (·OH) and/or sulfate radicals (SO₄·^–) generated from peroxydisulfate ions (S₂O₈^2–) due to oxidation of SO₄^2–. Jiang et al.¹¹⁶ developed an electro-Fenton membrane for PhACs removal. Carboxylic CNT-intercalated rGO was used as a membrane cathode, where CNT-loaded Fe³⁺ became reduced to Fe²⁺ to serve as the catalyst, and ·OH was continuously electro-generated with dissolved oxygen (DO). Florfenicol under dead-end filtration exhibited 61.6% removal through size exclusion and electrostatic repulsion but increased to 95.3% when applied with electro-Fenton oxidation reaction. Sutherland et al.¹⁴⁴ studied the concurrent dead-end filtration and electrochemical reduction of MO using nano zerovalent iron (nZVI)/CNT surface-coated polyethersulfone (PES) membranes at 2.75 bar. Under −2 V applied voltage on the membrane cathode, MO removal reached 87.3%, significantly higher than the control test (no voltage, rejection 2.9%). This result was attributed to nZVI’s high surface area and strong reductive potential leading to adsorption and electrocatalytic reduction at the membrane surface. As of writing this review, only limited ECMs have been reported to electrocatalytically degrade OMPs. Future studies need to overcome high catalyst cost and operating complexity. Current modules are restricted to bench scale and require optimization for treating realistic wastewaters in industry. For instance, critical research and development work includes improving the bonding stability of the catalytic layer, matching the separation rate with catalytic degradation, and exploring OMP removal with coexisting substances.

4.1.2.3. Impact of Photocatalytic Degradation

Photocatalytic membranes, based on semiconductor NPs, such as TiO₂,^134,145 ZnO,¹⁴⁵ and Ag₃PO₄,¹⁴⁶ and other nanomaterials such as C₃N₄¹⁴⁷ and MOFs¹⁴⁸ have also been used to enhance OMP removal via photocatalytic degradation. TiO₂ has garnered the most attention due to its low cost, excellent photocatalytic activity, and chemical stability.¹³⁴ Bai et al.¹⁴⁵ used CNT/ZnO/TiO₂-coated CA membranes to photocatalytically degrade acid orange 7 and achieved a >95% removal through absorption followed by phot-oxidation upon contact under 30 min UV light exposure, which was greater than using CNT alone (∼80%). Nevertheless, TiO₂’s broad energy gap (E_g = 3–3.2 eV) can only be excited under UV light.¹³⁴ Recent studies investigated NP-decorated photocatalytic membranes having visible light response. Wang et al.¹³⁴ developed an Au-TiO₂/PDA surface-coated PVDF membrane to remove tetracycline, where Au NPs increased the optical absorption, and PDA broadened TiO₂’s wavelength response. Here, 120 min visible light irradiation rendered 92% removal, much higher than performed in darkness (28%) and higher than adsorption to a pristine PVDF membrane (<0.5%). Ag₃PO₄ and g-C₃N₄ have narrow band gaps (Eg = 2.36 and 2.7 eV, respectively),^146,147 enabling visible light response. Zhang et al.¹⁴⁶ examined MB removal at 1 bar TMP by dead-end filtration across a PDA/rGO/Ag₃PO₄ surface-modified PVDF membrane under visible light irradiation. MB removal (up to 99. 1%) increased with surface nanocomposite loading resulting from increased adsorption and photodegradation, much higher than using a pristine PVDF membrane (15.2%). Qu et al.¹⁴⁷ deposited O-g-C₃N₄/GO/N-CNT onto a polycarbonate (PC) membrane to remove rhodamine 6G in a batch stirring reactor. GO and N-CNT greatly enhanced O-g-C₃N₄’s photocatalytic activity, as 360 min of visible light illumination gave 80% removal, and that was only 39% using pure O-g-C₃N₄ in the solution. Recently, heterogeneous photo-Fenton reactions on membranes have become a hot topic. Gao et al.¹⁴⁸ intercalated a polyacrylic acid (PAA)-modified iron-based MOF (NH₂-MIL-88B) in GO nanosheets and coated it onto a Nylon MF membrane. With the addition of H₂O₂, the as-prepared membrane degraded 98.8% MB during exposure to visible light for 40 min, whereas the pristine Nylon membrane only removed 43.1% MB through adsorption. Such improvement can be attributed to GO’s high surface area, which facilitated MB adsorption, as well as its good carrier mobility, which enhanced NH₂-MIL-88B’s charge separation capabilities for photo-Fenton reactions.

Results in this section suggest that post membrane surface modification by nanomaterials, adsorption (electrostatic attraction,¹²² hydrophobic effect,¹¹⁵ π–π interaction, and hydrogen bonding^44,65), size exclusion,^106,117,126 electrostatic repulsion,^109,131 and nanomaterial-assisted degradation (electro-oxidation,^{110,142,143,149} electro-¹⁴⁴ or chemical-reduction,¹⁵⁰ and photocatalytic degradation¹²⁰) favored the removal of charged/uncharged dyes, phenols, and aromatic PPCPs. The reported effectiveness of each of these processes varies, where we for instance found that most dye compounds were shown to have high removals (>90%) merely based on nanoenhanced physical interactions, whereas removals of PPCPs with various sizes, charges, hydrophilicities, and chemical structures were generally lower (>60%), even combined with nanoassisted catalytic degradations. These results highlight the necessity to better understand removal mechanisms and explore new nanomodifiers in accordance with the characteristics of the target OMPs in waste streams so that separation performance can be elevated. Interlab comparisons are also challenging since each lab uses different OMPs, nanomaterials, membranes, modifications, and operating parameters, but the general trend is that nanomaterial modifications improve OMP removal between 60%–100% at low operating pressures (1–5 bar).

4.2. Bulk-Modified Nanocomposite Membranes

4.2.1. Fabrication

Bulk-modified nanocomposite membranes introduce nanofillers into the polymeric matrix and are also classified as mixed matrix membranes. Bulk-modified nanocomposite membranes for OMP removal have been synthesized with nanospheres, nanosheets, and nanotubes, including organic (e.g., CNT,¹⁵⁵ CNC,¹⁵⁶ GO,^157,158 rGO,¹⁵⁹ and carbon dots (CDs)¹⁶⁰), inorganic (e.g., metals,¹⁶¹ metal oxides,^162−164 MOFs,^165,166 SiO₂,^167−170 and clays^171−173), or organic–inorganic hybrids thereof.^174−177 These nanoenabled membranes possess enhanced functionalities over commercial UF/NF membranes, such as greater hydrophilicities (thus higher permeability),^157,160,178 tuned pore channels,^161,179 increased antifouling properties,^172,174,180 and better mechanical stabilities.¹⁵⁵

Phase inversion via in situ nanofiller blending has been extensively adopted to fabricate bulk-modified nanocomposite membranes due to low cost and efficiency. This can be achieved by thermally induced phase inversion (TIPS) and nonsolvent-induced phase inversion (NIPS). The former approach increases the temperature to evaporate the solvents and thereby trigger polymer precipitation, whereas the latter approach immerses the polymer–solvent solution into a nonsolvent (typically water) to initiate the demixing process, in which asymmetric membrane structures form during solvent and nonsolvent exchanges at the polymer interface. Bulk-modified nanocomposite membranes for OMP removal often employ NIPS during fabrication. NPs are first dispersed into polymer casting solutions before casting. Solvent and polymer selection impacts NP dispersion and the resultant membrane characteristics. The solvent and polymer collectively impact the NP dispersion and the resultant membrane characteristics. DMF,^155,162,173 DMAc,^156,160,164 NMP,^161,174,176 and DMSO¹⁸¹ have been used as casting solutions. PES and polysulfone (PSF) are some of the most frequently investigated base polymers due to good mechanical, structural, and chemical stabilities.¹⁶⁰

Unlike standard polymeric membranes, bulk-modified nanocomposite membranes’ performances sometimes decrease because of NP agglomeration due to their high surface areas and strong interactions in the hydrophobic bulk phase, as well as the reduced NP reactivity due to oxidation or hydration reactions (e.g., metallic NPs and MOFs).^164,174 Improving NP dispersion and chemical stability becomes vital to maintain efficient OMP removal. Stirring/sonication in the casting step, dosing surfactants to the solvents, and modifying NPs’ surfaces with functional groups, grafted oligomers, or noncovalently bound modifiers can provide better dispersion. Increased inorganic nanofiller–polymer compatibilities of several dye separation membranes were achieved through modification with silane coupling agents,^170,172 polymers,¹⁶⁷ and carbon-based NPs.^174,175,178 These efforts overall improved membrane durability, hydrophilicity, and selectivity toward OMPs. Nonetheless, most bulk-modified nanocomposite membranes were studied at bench scale. NP-modified phase inversion requires optimization so that bulk-modified nanocomposite membranes can be industrially scaled.

4.2.2. Application for OMP Removal

OMP removal by bulk-modified nanocomposite membranes rely on adsorption by weak forces (van der Waals, dipole, and π – π interactions),^{161,164,165,182} electrostatic repulsion,^158,160,173 size exclusion,^156,167,183 catalytic oxidation/reduction,¹⁷⁷ and photocatalytic degradation.^176,178,181 Nanocomposite membranes for OMP removal are summarized in Table 3. We discuss the effects of three key factors on bulk-modified nanocomposite membranes for removing OMPs: (1) NP type and loaded mass, (2) operating conditions of pressure and feed concentration, and (3) solution pH.

Table 3. Bulk-Modified Nanocomposite Membranes for Removing OMPs in Water.

Modifier	Nanomaterial	Substrate	Pore size/MWCOa	Pollutant	Feed concentration (ppm)	Removala (%)	Removal mechanism	ref
Organic	PEI	PES	1.384 nm, 781 Da	reactive red 49, reactive black 5	500	92, 96	adsorption, size exclusion	(186)
	hollow mesoporous carbon nanospheres (HMCNs)	PES	3.5 nm, 336 kDa	tetracycline, 17β-estradiol	0.1, 0.0001	97, 94	adsorption	(187)
	GO	PES	3.8 nm	direct red 16	30	99	electrostatic repulsion	(157)
	carbon	PSF	5.93 nm	benzene, toluene, phenol	50	97.7, 82.8, 79.2	adsorption, size exclusion	(183)
	N-doped porous graphene oxide (N-PGO)	PES	9.14 nm, 1096 Da	reactive red 195	100	95.6	electrostatic repulsion	(188)
	carbon dots (CDs)	PES	12 nm	RR 198	100	98.9	electrostatic repulsion	(160)
	rGO-PDA	PSF	50.6 nm	direct red 80, methyl blue	25	98.8, 87	size exclusion, electrostatic repulsion	(159)
	GO	PSF	71.28 nm	bisphenol A	7.5	93	electrostatic repulsion	(158)
	P(4-VP-co-TRIM)	PES	565 nm	bisphenol A	0.25	87.4	adsorption	(182)
	SWCNT	PES	∼80 kDa	bisphenol A, 4-nonylphenol	0.1	78, 85	adsorption	(155)
	CNC	PES		direct red 80		95.8	size exclusion	(156)
Organic/Metallic	GO/Cu(tpa)	PES	3.6 nm for the filler	methylene blue, Congo red, methyl orange	100	65, 92	size exclusion, electrostatic repulsion	(174)
	GO/ZIF-L	PES	∼3.1 nm	amoxicillin	25	98.9	electrostatic repulsion	(180)
	GO/CuS	PES	∼ 3.7 nm	oxybenzone, bisphenol A	25	98, 95	electrostatic repulsion	(175)
	g-C3N4/ZnO	CA	6.1 nm	lanasol blue 3R	200	93.7	adsorption	(189)
	mpg-C3N4/TiO2	PSF	80 kDa	sulfamethoxazole	10	69	photocatalytic degradation	(176)
	MWCNT/ZnO	PES		direct red 16	30	96	photocatalytic degradation	(178)
	HPEI/MWCNT/Fe–Cu	PES		2,4,6-trichlorophenol	0.025	99.4	adsorption, catalytic degradation	(177)
Metallic	SiO2-PSS	PES	0.64 nm, 655 Da	reactive red 49, reactive black 5	500	∼90	size exclusion, electrostatic repulsion	(167)
	HNT (sulfonated)	PES	1.3 nm, 682 Da	reactive red 49, reactive black 5	1000	90, 94	size exclusion	(171)
	HNT (modified by APTES)	PVDF	1.922 nm	direct red 28	100	94.9	adsorption, electrostatic repulsion	(172)
	Fe3O4@SiO2-NH2	PES	9.42 nm	methyl red	30	97	adsorption	(164)
	montmorillonite	PES	9.59 nm	3,5-dinitrosalicylic acid, 2,4-dinitrophenol	22.8, 18.4	94, 90	adsorption, electrostatic repulsion	(173)
	alumina	CAP	22.8 nm, 122 kDa	catechol, p-nitrophenol	100	87, 89	adsorption	(162)
	CeO2	PES	∼23.4 nm	direct red 23, direct red 243, Congo red	100	>99	size exclusion, electrostatic repulsion	(163)
	Pd	PSF	∼40 nm	crystal violet	40.8	99	adsorption	(161)
	SiO2-DES	PI	>46.4 nm	phenol	30	96	adsorption, electrostatic repulsion	(168)
	Au0.1Ag0.9/TiO2	CA	430.12 nm	tetracycline	5	88.7	photocatalytic degradation	(181)
	MIL-68(Al)	PVDF	700 nm	p-nitrophenol, methylene blue	5	82.7, 95	adsorption	(165)
	MIL-125(Ti)	PVDF		rhodamine B	10	99.7	adsorption, photocatalytic degradation	(166)
	SiO2	PES		bisphenol A	0.1	88	adsorption	(169)
	SiO2 (modified by SBMA)	PES		reactive black 5, reactive green 19		97.9, 99	electrostatic repulsion	(170)

^aData were obtained either from the article contents or extracted from figures using GetData Graph Digitizer software.

4.2.2.1. Impact of NP Type and Loading

First, OMP removal by bulk-modified nanocomposite membranes is dependent on NP type, loading, and how the membrane properties were impacted by NP type and loading. NP inclusion into membranes primarily impacts membrane hydrophilicity, pore size, and porosity, which impacts the membrane’s permeability and OMP rejection. Membrane pore formation is affected by NPs during the blending phase inversion step. Thus, evaluating and optimizing NP loading is crucial. Ghemei et al.¹⁷³ studied the effect of organically modified montmorillonite (OMMT) in NF PES membranes on pesticide removal. Membrane hydrophilicity increased while pore size decreased when increasing OMMT loading from 0 to 4 wt %. This benefited 3,5-dinitrosalicylic acid and 2,4-dinitrophenol rejection (>90%). However, adding more OMMT (>6 wt %) caused dense skin layer formation and layered silicate aggregation, which decreased the water flux and impeded NF separation. Balcik-Canbolat and Van der Bruggen¹⁵⁶ observed that high content (up to 1 wt %) of CNC in the PES membrane matrix contributed to higher membrane hydrophilicity, pore size, and looser skin layers. This gave higher water flux but sacrificed some direct red 80 removal due to a weakened sieving effect. Despite losing selectivity, all the membranes obtained >90% rejection. Modi and Bellare¹⁸⁰ compared the amoxicillin removal efficiencies using PES hollow fiber membranes embedded by carboxylated GO, ZIF-L nanoflakes, and their combined nanocomposites, respectively. The latter membrane exhibited the highest hydrophilicity, negative surface charge, and largest pore size, while rendering the best water flux and removal to negatively charged amoxicillin. At 1.5 bar TMP in the cross-flow condition, the removal reached 98.9%, more than 2 times higher than using bare PES membranes. These studies indicate that the impacts of NP incorporation are not easily determined because the resulting physicochemical properties of nanocomposite membranes are affected by multiple parameters such as nanomaterial concentration, dispersion, surface chemistry, and crystal structures. Unfortunately, NP loading must be experimentally optimized to maximize OMP removal efficiency, and we argue that researchers should design their experiments to understand what nanomaterial parameters, for instance, the NP loading, size, porosity, and chemical characteristics, have the greatest impact on membranes’ OMP removal, rather than simply trying to demonstrate high OMP removal with the addition of a nanomaterial.

4.2.2.2. Impact of Filtration and Feedwater Condition

Second, operating conditions, OMP concentration, and coexisting species impact bulk-modified nanocomposite membrane performance, as these factors affect the residence time and interaction between OMPs and the membrane. Nasseri et al.¹⁵⁸ investigated the dead-end filtration of bisphenol A (BPA) using PSF/GO nanocomposite membranes. Low TMP and high feed concentration gave better separation characteristics because the electrostatic repulsion effect dominated the filtration. The negatively charged BPA tended to be rejected by PSF/GO membranes due to electrostatic repulsion. A high BPA concentration and/or an increase of aqueous pH (thus an increase of net negative charge) would cause stronger electrostatic repulsion to the membrane surface which could increase OMP removal by increasing the residence time between solute and membrane. However, at high TMP, the hydraulic pressure caused high convective mass transfer of BPA, which exceeded the electrostatic repulsive force on BPA molecules and thereby increased the amount of BPA that crossed the membrane. Further, high TMP increased the solution passage rate through the GO layers, thereby reducing the contact time between the solute and the membrane. This finding aligns well with established adsorption theory that a higher retention time, achieved at lower TMPs, leads to greater adsorption. We also hypothesize that a higher feed concentration is likely to induce greater membrane fouling which hinders permeation and increases rejection. Mukherjee and De¹⁶² removed catechol using CA/alumina nanocomposite membranes, and rejection rates significantly dropped in the presence of NaCl electrolytes, which is likely the result of weakening the electrostatic force between OMPs and the membrane due to electrostatic shielding. Additionally, gradual saturation of deposited OMPs and other coexisting compounds (e.g., NOM, calcium and magnesium salts, and soluble microbial products (SMPs)¹⁸⁴) on the NPs in the membrane surfaces and internal pores may cause fouling, consume existing reactive/adsorptive sites, and block nanochannels which impact permeate flux and selectivity and thus membrane service life. These findings suggest that nanocomposite membranes do not necessarily primarily rely on sieving and/or solute diffusion to remove OMPs, unlike traditional NF/RO processes. Rather, their effectiveness for OMP removal is dependent on the adsorptive properties of the OMP with NPs and polymer materials of the membrane and requires a deep understanding of membrane–nanofiller–solute interaction mechanisms.¹⁵⁸

4.2.2.3. Impact of pH and OMP Characteristics

Third, solution pH critically impacts the bulk-modified nanocomposite membrane performance by affecting membrane surface charge (reflected by zeta potential), nanocomposite stability, dissociation of the charged functional groups from OMPs, and solute solubility.^165,182,185 Tan et al.¹⁶⁵ developed MIL-68(Al)-blended PVDF membrane adsorbers to remove aqueous p-nitrophenol (pK_a = 7.15). Maximum adsorption (up to 82.7%) of p-nitrophenol was obtained at pH values (4–7) below pK_a, whereas the adsorption capacities decreased remarkably at pH above pK_a, because at higher pH p-nitrophenol became anionic and was electrostatically repelled from the negatively charged membrane surface. pH also affects NP stability, where the inherent MOF structure was destroyed at pH < 4, leading to poor OMP removal. Mukherjee and De¹⁶² found maximum catechol adsorption on cellulose acetate phthalate (CAP)/alumina nanocomposite membrane surfaces at aqueous pH above the membrane’s isoelectric point (pH_IEP 5.4) and below the pK_a (9.5) of catechol. Under these conditions, the CAP acetate groups were deprotonated such that the membrane surface was negatively charged, which facilitated the protonated catechol transport to the membrane due to their opposite polarity. Niedergall et al.¹⁸² investigated adsorption to remove OMP from water. A sharp decline of BPA adsorption on the polymer NP (P(4-VP-co-TRIM))-mixed PES membrane was observed at high pH above BPA’s pK_a (9.6–10.2). This can be explained by BPA deprotonation resulting in electrostatic repulsion against the negatively charged membrane, as well as increased solubility leading to less hydrophobic adsorption. Manipulating solution pH can maximize electrostatic attraction of OMPs to adsorption sites; however, OMP rejection will eventually decrease as adsorption sites are saturated.

4.3. Thin Film Nanocomposite (TFN) Membranes

4.3.1. Fabrication

Thin film composite (TFC) membranes consist of an ultrathin barrier layer (commonly polyamide (PA), thickness 50–300 nm¹⁹⁰) atop a polymeric membrane support. Compared to integrally skinned asymmetric membranes, TFC membranes have an independently controlled and optimized top selective layer for enhancing selectivity and pollutant rejection.^191,192 Interfacial polymerization (IP), first reported by Morgan et al. in the 1960s,²²² is the most common route to prepare commercial PA TFC membranes.¹⁹³ Typically, the support membrane is impregnated with an aqueous solution of diamine or polyamine, such as PIP^16,69,191 and MPD,^18,192 followed by removing excessive solution on the membrane surface. The soaked membrane is then put into contact with a water immiscible organic solution (e.g., hexane) containing acryl chloride monomers, such as TMC,^16,191,192 to form a PA thin film.

While traditional PA TFC membranes were primarily optimized to make high pressure NF or RO membrane systems for NOM removal, desalination, and heavy metal removal (monovalent ions can permeate through NF membranes), the inherent permeability–selectivity trade-off, fouling-prone nature, and high energy costs limit their application for OMP separation under low pressures. NPs, such as CNT,¹⁹⁴ CNC,¹⁹¹ GO,^195,196 MOFs,^69,192,197 SiO₂,¹⁸ clays,^16,198,199 and MoS₂,^200,201 have high surface areas and high adsorption potentials, and hence, the exploitation of these NPs into TFC membranes provides thin film nanocomposite (TFN) membranes with great potential for OMP separation. This is due to TFN membranes’ improved mechanical and chemical stabilities and reactive and adsorptive capacities. The porous fillers can be (i) assembled flexibly within the thin film,^{18,69,191,195} by premixing the NPs in the aqueous solution or the organic solvent before the IP reaction and (ii) formed as an interlayer between the thin film,^{16,192,194,197} through in situ growth (e.g., by phase inversion) or pressurized filtration of the NPs, followed by conventional IP reaction. Although less common, nanofillers have been combined with non-PA-based thin films to design TFN membranes for OMP separations. For instance, Ghaemi and Safari²⁰² polymerized pyrrole to form hydrophobic polypyrrole (PPy) on the PES membrane, in which nanozeolite SAPO-34 was blended. Zhang et al.¹⁹³ incorporated graphene oxide quantum dots (GOQDs) into a tannic acid (TA) layer during an IP process, which yielded a loose structured NF membrane. Compared with bulk-modified nanocomposite membranes, TFN membranes suffer less NP dissolution, which reduces the potential for secondary contamination caused by toxic and harmful NPs.¹⁹¹

4.3.2. Application for OMP Removal

TFN membranes have been increasingly explored for NF²⁰³ and RO²⁰⁴ processes for OMP separation, and the tailored top layer plays a crucial role for membrane performance, particularly with respect to membrane flux and selectivity. Here, we focus on NF applications, which are much less energy intensive. The selective layer and the added NPs collectively determine TFN membranes’ surface properties, and OMP removal can be attributed to one or more mechanism, which includes size exclusion,^191,201,205 electrostatic interactions,^16,194,206 and adsorption/diffusion.^{18,69,197,207} TFN membranes for OMP removal are summarized in Table 4. Integrated NPs into conventional TFC membranes often intentionally produce nanovoids either as nanochannels within porous NPs and/or between the NPs and PA selective layers.²⁰⁸ While these nanochannels increase water permeance, they often provide no improvement in OMP removal and sometimes cause a decrease in OMP rejection due to impaired membrane integrity and defect formation.²⁰⁸ In this section, we discuss how assembly structure and nanofiller loading impact separation performance, as well as the benefits and challenges encountered for OMP removal.

Table 4. Thin Film Nanocomposite (TFN) Membranes for Removing OMPs in Water.

Modifier	Nanomaterial	Location	Substrate	Thin film	Pore size/MWCOa	Pollutant	Feed concentration (ppm)	Removala (%)	Removal mechanism	ref
Organic	CNC	within TF	PES	PA	0.41 nm, 312.06 Da	rose bengal, Congo red, methyl orange, crystal violet, methylene blue		>99	size exclusion	(191)
	quaternized carbon quantum dots (QCQDs)	within TF	PVDF	PA	0.42 nm	sulfamethoxazole, carbamazepine, atenolol, trimethoprim	1	98.2, 98.6, 99.5, 99.7	size exclusion, electrostatic repulsion	(209)
	β-CD-PIM	within TF	PSF	PA	0.62 nm	erythromycin	100	97	size exclusion	(210)
	GOQDs	within TF	PES	PA	0.86 nm, 525 Da	orange GII, Congo red	100	95, 99.9	size exclusion	(205)
	GO	within TF	PSF	PA	1764 Da	new coccine, ponceau S, direct red 80	500	94,3, 96.2, 99.2	size exclusion, electrostatic repulsion	(196)
	GO	within TF	PSF	PA		norfloxacin, sulfamethoxazole	1	53.32, 41.85	size exclusion, electrostatic repulsion	(195)
	GOQDs	within TF	PAN	TA		(1) methyl orange, Congo red, methyl blue, (2) methylene blue	100	84, 100, 98, 74	(1, 2) size exclusion, (1) electrostatic repulsion	(193)
	CNT (modified by PDA)	interlayer	PES	PA	0.878 nm, 365 Da	(1) methylene blue, crystal violet, (2) methyl orange, methyl violet, acid fuchsin, Congo red	100	86.4, 98, 92.5, 99.5, 100, 100	(1, 2) size exclusion, (2) electrostatic repulsion	(194)
	TpPa-1 (modified by PDA)	interlayer	PAN	PA	16.36 nm	orange GII	100	93	adsorption, size exclusion	(207)
	GO	interlayer	PSF	PA		safranine T, indigo carmine, coomassie brilliant blue, rose bengal	100	>92	size exclusion	(211)
Metallic	SiO2 (modified by oleic acid)	within TF	PSF	PA	0.32 nm	atrazine, propazine, prometryn	10	>98	adsorption, size exclusion	(18)
	SAPO-34	within TF	PVDF/PES	PA	0.45 nm, 320 Da	raffinose, saccharose, glucose		98.6, 95.3, 82.1	size exclusion	(198)
	MIL-101(Cr)	within TF	PES	PA	0.6 nm	methylparaben, propylparaben, benzylparaben, bisphenol A	0.2	47, 46, 51, 80	adsorption, size exclusion	(69)
	Zn-TCPP	within TF	PSF	PA	1.54 nm	Congo red, methylene blue, direct red 23	200	>96	size exclusion	(212)
	SAPO-34	within TF	PES	PPy	220 Da	(1) methyl violet 6B, (2) reactive blue 4, acid blue 193	50	100	(1) adsorption, (2) size exclusion, electrostatic repulsion	(202)
	HNT	within TF	PSF	PA		setazol red, reactive orange	100	99.7, 99.7	electrostatic repulsion	(206)
	ED-MIL-101(Cr)	within TF	PES	PA		terbutaline, atenolol, fluoxetine, ketoprofen, diclofenac, bezafibrate	0.2	>82.7	size exclusion, electrostatic repulsion	(203)
	MoS2 (modified by TA-Fe3+)	within TF	PSF	PA		glucose, sucrose, raffinose	150	70, 89, 91	size exclusion	(201)
	h-BN	within TF	PES	PA		methylene blue	10	73	size exclusion	(213)
	TiO2	withn TF and interlayer	PES	PA		Congo red, alcian blue, orange GII	100	98, 96, 85	size exclusion	(214)
	MoS2	interlayer	PES	PA	0.53 nm	methylparaben, ethylparaben, propylparaben, benzylparaben	0.2	53.7, 69.1, 79.1, 91.3	adsorption, size exclusion	(200)
	ZIF-8	interlayer	PSF	PA		acetaminophen	100	55	size exclusion	(192)
	zeolite	interlayer	PSF	PA		21 types of PhACs	0.02	>80	size exclusion, and/or electrostatic repulsion	(16)
	MIL-53(Al)	interlayer	PSF	PA		phenacetine, nalidixic acid, carbamazepine, sulfamethoxazole, atenolol, sulpiride	0.05	67, 88, 78, 88, 80, 92	adsorption, size exclusion	(197)

^aData were obtained either from the article contents or extracted from figures using GetData Graph Digitizer software.

4.3.2.1. Impact of NP Characteristics and Their Assembly Location

The membrane permeability and OMP separation capacities are greatly impacted by the NP properties and thus their location in the TFN membrane. Basu and Balakrishnan¹⁹² and Dong et al.¹⁶ intercalated ZIF-8 and zeolites, respectively, between the PA thin film and the PSF support layer. This approach enhanced the stability of the TFN membrane because blending these NPs into PA during the IP process can cause microsized cavities, reducing separation efficiencies for OMPs. The interlayer setup preserved the PA’s size exclusion effect and enhanced the water permeability by creating nanochannels under the PA which acted as water channels. In this manner, Basu and Balakrishnan¹⁹² obtained 9% higher rejection for acetaminophen (55%) and a 2-fold permeation increase compared with the unfilled PSF/PA membrane, when tested under dead-end flow conditions at only 4 bar TMP. Dong et al.¹⁶ also observed a similar or higher rejection (>90% for most compounds tested) of 21 types of PhACs when using the zeolite–PA-blended TFN membrane in cross-flow conditions at only 5 bar TMP compared to the bare PA membrane. In many cases, incorporating NPs into the PA can be challenging; however, some NPs can exhibit high compatibilities with PA and high stabilities during thin film formation by IP. Zhao et al.¹⁹⁷ investigated MOF-PA interaction patterns and found that MIL-53(Al) can tightly bind with PA post IP. Under either setup, the obtained TFN membrane achieved 1.3 times higher permeability and significantly higher nalidixic acid and sulfamethoxazole rejections than the TFC control membrane.

4.3.2.2. Impact of NP Content

Furthermore, NP content in the TFN membranes critically impact the membrane–solute affinity (e.g., through changing membrane hydrophilicity, thin film thickness, and available adsorption sites), size exclusion (e.g., through changing pore size), and Donnan exclusion (charge repulsion) (through changing charge density), leading to a change in the dominant OMP removal mechanism. It has been observed that incorporating NPs into TFCs leads to various trade-offs in selectivity and permeability of different OMPs. As such, it has been challenging to remove a wide range of OMPs using a single NP modification to a TFC. For example, Dai et al.⁶⁹ incorporated MIL-101(Cr) into a PA active layer on the PES membrane and used the TFN membrane for the removal of EDCs and salts. Increased MIL-101(Cr) content in the PA layer led to a moderately higher rejection of hydrophobic methylparaben, propylparaben, benzylparaben, and BPA through stronger hydrophobic adsorption, but greater MIL-101(Cr) content also led to slightly lower rejection of some monovalent and divalent salts. Similarly, Wang et al.¹⁹⁵ found a trade-off in incorporating GO into PA thin films. They found that increased GO content in the PA top layer on PSF support provided the TFN membrane with a higher negative charge. At moderate GO loading (0.004 wt %) in the PA layer, the TFN membrane showed 41.85% removal toward negatively charged sulfamethoxazole due to charge repulsion, approximately 3 times higher than the control TFC membrane. However, despite the presence of GO, the TFN membrane showed equally poor removal (<5%) toward paracetamol as the control TFC membrane, which was a result of its low MW and high hydrophilicity. Gong et al.¹⁹⁴ preloaded PDA-modified CNT onto PES membranes, followed by adding a PA top layer. They found that increased PDA–CNT content increased the interlayer thickness and the zeta potential, thus leading to lower MWCO and a more negatively charged surface. Having similar MWs, anionic dye methyl violet obtained higher removal than cationic dye MB (99.5% vs 86.4%) due to the Donnan effect. However, despite high dye removal, the membrane exhibited low rejection toward Cl^–, a monovalent ion, due to the disruptions within the PA layer by the CNTs. This suggested promising recovery of aqueous dyes from some saline wastewaters but without compromising the monovalent ion rejection. To conclude, the engineered TFN membranes have great potential for purification and recovery of different OMPs, but their properties should be tailored to suit the characteristics of the target compounds.

5. Future Perspectives

Significant progress on state-of-the-art nanocomposite polymeric membranes and TFNs has been made over the past 10 years to efficiently remove OMPs in aqueous environmental feeds. Nanomaterial incorporation into traditional polymeric substrates was shown to be promising for increasing membrane selectivity toward OMPs by customizing the membrane characteristics and functionality for target contaminants.

5.1. Improvements for Physical Separation

Based on the synthesized results in Section 4, OMP removal can be improved with nanomaterials by matching specific OMP properties (e.g., MW, adsorptive properties, and charge) with specific nanomaterial properties. There is unlikely to be a one-size-fits-all approach to removing all OMPs; however, broad classes of OMPs can be targeted by rational selection and effective use of nanomaterials.

OMP adsorption to membrane walls and nanomaterials is a major pathway to OMP removal. We recommend that researchers develop porous nanomodifiers with high surface areas (e.g., MOFs, CNTs, and zeolite) to maximize OMP adsorption into nanofillers while maintaining high water flux through nanochannels in the membrane and through the nanomodifiers. Because OMPs (e.g., most PhACs, DBPs, and dyes in Table 1) with high water solubility and low log K_ow values tend to exhibit low affinity to organic materials, introducing hydrophilic NPs or NPs grafted by hydrophilic polar moieties (e.g., hydroxyl groups) into as-prepared membranes can be beneficial to enhance OMP–NP interaction, thus the OMP adsorption. Physical adsorption is rapid and efficient, and for low concentrations of OMPs, it is mostly only limited by OMP diffusion to adsorption sites. As shown in Table 1, the concentrations of most OMPs in surface waters are very low. With the exception of some EDCs, PCPs, and tetracycline, most OMP concentrations range from 0.01 mg/L (ppm) to only a few hundred mg/L (ppm). As such, adsorption is especially helpful when treating small-volume OMP-rich waters (e.g., low-severity point-source leakage) in which the OMP quantities usually will not surpass membrane adsorption capacity. Adsorption sites, however, may become saturated rapidly when treating EDCs and PCPs as their concentrations in surface waters have been reported to be as high as over 4000 mg/L. For these OMPs, advancing nanomaterial adsorption site regeneration and/or catalytic and electrocatalytic nanomaterials will be critical.

Several OMPs are charged or polar compounds, as listed in Table 1, such as tetracycline, sulfamethoxazole, and acetaminophen. We encourage future research to further demonstrate using charged nanomodifiers to reject charged OMPs. For instance, negatively charged nanomodifiers (e.g., CNTs, GO, rGO, and Ag) that have been extensively studied can enhance electrostatic repulsion to like-charged OMPs such as some anionic dyes and PhACs. In contrast, positively charged OMP rejection is usually low because of the slight negative charge of most commercial membranes. The effect of electrostatic adsorption to the slightly negatively charged membranes is generally minimal. To remedy this limitation, research should focus on using positively charged NPs for membrane modification and test these nanoenabled membranes with cationic model OMPs, such as MB. One can obtain cationic NPs by cross-linking them with cationic polymers such as PEI¹⁰⁹ and/or stabilizing them with cationic surfactants such as quaternary ammonium surfactants.²¹⁵ Worth noting is that there are trade-offs between membrane retention (adsorption) and membrane rejection (electrostatic repulsion). An increase of electrostatic repulsions and/or a suppression of membrane–OMP hydrophobic interactions might weaken membrane adsorption mechanisms and vice versa. This trade-off suggests that when designing nanoenabled membranes the dominant OMP removal mechanism with a specific nanofiller must be investigated to maximize OMP removal.

Most OMPs have low MW, ranging from 100–400 g/mol (Da), as shown in Table 1. Size exclusion-based separation therefore requires tight porous membranes, which can be controlled via modifying nanofiller interior structure/size. By taking advantage of the highly organized nanochannels in some prevalent 2D nanomodifiers (e.g., GO, MOS₂, h-BN, and Mxene), membrane selectivity can be regulated via molecular sieving. Nevertheless, perfectly matching the pore size of a resultant nanoenabled membrane to a target OMP is challenging and likely to provide little benefit. Rather, the goal of using nanofillers should be to create membranes with a very narrow statistical distribution of small pore sizes (several nanometers or less) to achieve effective size exclusion of OMPs. In conventional size exclusion-based separations, there will always remain the trade-off between water permeability (i.e., energy requirements) and rejection: membranes with small pore sizes require energy-intensive separations due to low water permeability, as compared to membranes with larger pore sizes and thus high water permeability but which cannot attain sufficient OMP removal. Nevertheless, with OMP removal as the primary goal, nanofillers’ main advantages to size exclusion are their highly uniform sizes and structures, enabling uniform pore sizes.

Overall, OMP transport mechanisms across nanocomposite polymeric membranes are still not perfectly quantified and will likely never be simply ascribed to a single mechanism. Moving forward to next-generation nanocomposite polymeric membranes demands a knowledge of OMPs’ physicochemical characteristics (Table 1) and a rational alignment of their size/structure, charge, and hydrophilicity with those of a nanomodifier. Nanoenabled membrane development should be based on known nanomodifier and OMP properties, and the mechanisms of separation should be determined by systematic experiments and potentially supported by molecular simulations which can help optimize OMP removal.

5.2. Improvements for Catalytic Degradation

Electro-¹⁴⁹ and photo-assisted¹⁸¹ catalytic nanocomposite membranes have seen tremendous progress in their abilities to catalyze the conversion of OMPs into less harmful species. Conductive nanomaterials such as CNT and rGO have been widely used to make electrocatalytic membranes, but only a small number of other nanomodifiers have been studied, providing an opportunity for researchers. A further limitation corresponds to short lifespans of the electro-generated reactive radicals and their low reaction rates toward trace-level OMPs.¹¹⁰ We suggest that future studies should focus on synthesizing and/or using high surface area, highly conductive nanomaterials to prepare membrane electrodes which would enable rapid mass transport and effective OMP degradation. The advancements in highly porous, high surface area, conductive MOFs²¹⁶ may provide interesting foundations for fabricating small-scale ECMs for OMP degradation in the future. Another major challenge concerns the energy required for the electro-oxidation of OMPs, which can be scavenged by side reactions such as the oxygen evolution reaction (OER).²¹⁷ Precise control of applied potentials and knowledge of material redox potentials are required for optimal operation. In terms of nanoenabled photocatalytic membranes, while previous studies nearly exclusively researched nanosized TiO₂, future research should focus on incorporating low cost, easily synthesized functional nanophotocatalysts (e.g., g-C₃N₄) that can be activated by visible spectrum wavelengths (e.g., solar or LED light) to achieve contaminant photodegradation. Beyond materials, improvements to the photocatalytic membrane reactor designs are also critical. Reactor designs should be optimized using a systematic comparison of how different operating modes (e.g., continuous flow-through vs photoirradiation followed by filtration) impact OMP removal efficiencies for a given specific nanocatalysts’ surface area, pore size, and reactivity. Finally, for both electrocatalytic and photocatalytic membrane processes, OMP degradation pathways, degradation byproducts, and membrane removal efficiencies to the produced OMP intermediates should be identified to minimize their risks in permeates.

5.3. Scale-Ups from Bench to Pilot Scale

To transition these membranes from the lab to pilot scale, more research using realistic membranes and systems will be required. Developing durable membranes with longer operating lifetimes that can achieve high throughput and high removal efficiency is an urgent need for these membranes’ economic viabilities. There is a large variety of nanocomposite membranes in the literature with many different nanomaterials, polymers, formulations, and chemistries. To identify the most economically viable nanocomposite materials, we recommend that researchers perform techno-economic analyses (TEA) and life-cycle analyses (LCA) to identify which membranes would be most promising to scale up for industry.²¹⁸ In addition, there is no “one-size-fits-all” solution to remove all OMPs; their charges, polarities, MWs, and concentrations as well as the media in which OMPs are found all influence removal. As such, standard OMPs should be used to reveal different “nanocomposite polymeric membrane–OMP” interaction mechanisms which will help select the most appropriate membrane properties and operating conditions for specific OMPs. Relatedly, realistic complex pilot-scale operations have rarely been investigated and should be the continued focus of future research, as the effects of complicated environmental matrices on the removal of target OMPs are poorly understood and might negatively affect their selectivity. We recommend a deeper investigation into OMP transport behavior into the nanoscale confinement during filtration across the porous membrane under both laboratory and practical wastewater conditions for a better understanding of OMP removal by solute retention, adsorption, and/or catalytic degradation. Challenges, such as scalability, agglomeration, permeability, and selectivity, need to be addressed for the membrane types that have been developed so far. In addition, with respect to realistic systems, the robustness of the nanocomposite membranes for long-term OMP treatment under realistic working conditions (e.g., WWTPs) should be investigated.

Next, there are significant materials challenges that must be the focus of surface-modified and bulk-modified nanocomposite membranes. NPs’ adsorption capacities and reactivities inevitably decrease with use, which limits their industrial application, since frequent replacement of nanocomposite membranes/nanomaterials is economically inviable. Thus, it is important to develop means to regenerate these NPs in situ to allow stable separation performance. In our appraisal, chemical washing²¹⁹ and electro-regeneration¹¹⁰ of NPs are the preferable methods. NP agglomeration, especially at high loading content, on membrane surfaces or with polymeric matrices can result in defective membranes with poor OMP removal efficiencies. The translocation and depletion of NPs in the membrane during filtration over time would likely destroy the original structures, reduce the membrane life span, and eventually become a secondary nanomaterial-laden pollutant stream. Future work should develop better chemistries to stabilize or covalently bind the NPs to the membranes and study the retention and leaching of NPs in the nanocomposites. For instance, there is a whole field focusing on exciting research into bioinspired membranes which employs biological molecules including mussel proteins to adhere NPs to the skin layer of membranes or within the support. Another emerging tool is the use of atomic layer deposition (ALD), which allows deposition of various metallic or organic materials to be layered onto porous membranes to produce nanoscaled ultrathin films, which have good conformities, high layer thickness precisions and controls, and even regulated pore sizes.²²⁰ These membrane types have not been given much attention—likely because ALD has not been used at scale for membrane production—but these membranes have high potential for OMP separation.

Finally, TFNs are very promising technologies, but major challenges remain with respect to modifying PA thin films. PA TFC membranes have wide applicability in industry but are prone to swell in contact with organic solvents.²²¹ PA layers also have the drawbacks of low cross-linking degree and nonuniform functional groups, which make the thin films heterogeneous with respect to their hydrophilicities and polarities.¹⁷ A high priority is the development of TFN membranes aimed to address the limitations of low flux, swelling by solvents, heterogeneous surface properties (hydrophilicity and polarity), and sensitivity to foulants and chlorine. PA TFN membranes have potential to overcome current permeability—selectivity trade-offs, which would enable them to remove OMPs at much lower TMPs. Nonetheless, NP agglomeration and leaching may cause defective PA structures and thereby reduce the NP effective surface area. Future studies should seek proper surface modifications of NPs to increase the compatibility with the organic phase and employ new chemistries or adopt novel monomers for constructing the thin film. Finally, more effort should be paid to develop loose TFN membranes for low pressure operations (UF and NF processes), without sacrificing the rejection rates of OMPs and salts.

6. Conclusion

The presence of OMPs in water bodies used for the potable water supply has raised public concern about the long-term drinking water safety and quality. Natural ecological and physicochemical processes and conventional water treatment technologies cannot sufficiently degrade or remove these pollutants. In the absence of broad regulation or societal change in behaviors and consumption, new technologies are needed to address the growing environmental issue. Pressure-driven membrane-based technologies are proven and widely adopted and can be easily upgraded to treat various water types. However, most membrane technologies require enhancements to efficiently remove OMPs from feed waters. Nanocomposite polymeric membranes are one such technological improvement.

The reported OMP removal performance of nanocomposite polymeric membranes depends on membrane characteristics, operating conditions, and OMPs physicochemical properties. Among MF, UF, and NF membranes, NF membranes have the lowest MWCO and can achieve a high rejection toward small-sized OMPs based on size exclusion. However, NF membranes require high pressures and suffer from severe fouling issues in water treatment. Traditional UF or MF membranes demand lower operational pressures but can only partially remove large MW OMPs due to these membranes’ larger pore sizes and higher MWCOs. Recent studies have demonstrated that UF or MF membranes with modifications to their surface charges and hydrophilicities can enhance OMP rejection. The physicochemical properties of these loose membranes could be further improved through surface or matrix modification with nanomodifiers. In this regard, rejection rates for OMPs could be greatly improved, depending on their charges and polarities. Furthermore, there is an exciting possibility to reduce the membrane operating pressures if these membranes are combined with means of OMP degradation and removal, achieved either through photocatalysis, electrochemical redox reactions, or enhanced adsorption. This review of OMP removal using nanocomposite polymeric membranes provides a baseline organization for the first time so that general separation trends can be quantitatively determined through statistical approaches, such as meta-analyses of current data. Such quantitative analysis will justify optimization of specific nanocomposite polymeric membranes indicated as worthy of further research in this review. Should the researchers continue to optimize membrane material costs, long-term stabilities, scalabilities, and separation characteristics, we expect that nanocomposite polymeric membranes will be the dominant method to treat emerging pollutants, such as PhACs and EDCs, in complex aquatic environments.

The authors declare no competing financial interest.