Performance Stability and Regeneration Property of Catalytic Membranes Coupled with Advanced Oxidation Process: A Comprehensive Review

Abstract

The performance stability and regeneration property of catalytic membranes should be considered from a practical viewpoint. In this review, recent works concerning this issue have been comprehensively summarized. The performance stability of catalytic membranes was first summarized in terms of the two commonly employed evaluation ways, i.e., multi-cycle experiments and long-time filtration. The main deactivation reasons included the adsorption of pollutants and its oxidation intermediates, the change in chemical properties of the membranes, as well as leaching of metal species. After that, the regeneration property of the catalytic membranes was discussed according to the different regeneration methods including solvent washing, heat treatment, advanced oxidation, and others. This review further highlights the current achievements and hurdles in the performance stability and regeneration property of catalytic membranes and proposes recommendations for future works, including more attention on changes in the kinetic constant, trans-membrane pressure/flux and removals of total organic carbon/chemical oxygen demand, the selection of suitable cycle numbers and operation time, the development of new regeneration methods, as well as more studies on the pilot and larger scales.

1. Introduction

Clean water is essential to the survival and development of human society, but the various organic pollutants in water, including dyes, antibiotics, pesticides, and endocrine disruptors, pose potential threats to the environment and human health. Among the various physical, chemical, and biological remediation technologies, the advanced oxidation process (AOP) is known as one of the most promising approaches considering its wide applicability to treat different pollutants and the ability to degrade or even mineralize them. Based on the oxidant or energy input used to produce active species, AOPs can be classified as persulfate oxidation, H₂O₂ oxidation (Fenton oxidation), ozone oxidation, photocatalytic oxidation, electrochemical oxidation, and others [1,2]. In the latter two processes, light (UV or visible light) and electric energy as external energy sources are introduced, respectively. In these AOPs, catalysts, especially the heterogenous ones, are generally utilized to boost the removal efficiency of those organic pollutants. However, after the catalytic oxidation processes, the suspended catalysts need to be separated from the treated water by filtration, centrifugation, or other methods, which is rather tedious and time-consuming.

The membrane technology is another widely investigated remediation method for water treatment. However, membrane separation is a physical process with which the organic pollutants can only be concentrated rather than degraded. Recently, by coupling AOP and membrane technology, catalytic membranes fabricated by loading catalysts on membranes have been developed and employed for the treatment of organic pollutants in water. The pristine membranes possess limited ability for pollutant degradation. When flowing through the catalytic membranes, the organic pollutants can be degraded under the catalytic effects of the loaded catalysts. After being treated with the membranes, water and catalysts are separated, and clean water as the filtrate is collected, avoiding the routine catalytic separation step. Meanwhile, during the filtration process, accelerated mass transfer of organic pollutants and oxidants from the bulk solution to the catalyst surface is achieved. Compared to conventional batch systems, the employment of catalytic membranes can lead to a higher production rate of oxidative species. Enhanced mass transfer can also promote the interaction between pollutants and the produced radicals, leading to increased reaction rate and enhanced removal efficiency. For instance, CoAl-layered double hydroxide (LDH) was loaded on polyvinylidene fluoride (PVDF) to fabricate the catalytic membrane of LDH_m, which was then employed for the degradation of ranitidine (RA) [3]. The LDH_m/peroxymonosulfate (PMS) system resulted in around 85% removal of RA in 10 min, while the removals with the pristine PVDF membrane and the bulk LDH/PMS system were only <10% and 37–44% [3]. The LDH_m can be facilely separated after use, while the powdered LDH catalysts need to be separated by filtration or centrifugation. One disadvantage of LDH_m compared to bulk LDH may be its increased fabrication cost, but the loading of LDH on the membrane can minimize the catalyst loss during repeated use. As another example, Co²⁺/Mxene loaded on PVDF showed 97% removal of tetracycline (TC) in the presence of PMS, while the removal with the PVDF/PMS system was only around 40% [4]. For comparison, the catalytic membrane was immersed in a solution containing both TC and PMS to conduct AOP in the batch mode, showing a poor TC removal of around 60%, which demonstrated the indispensable role of membrane filtration in the catalytic membrane-based degradation process [4].

Catalytic membranes can be fabricated in many approaches. The most widely used one is the filtration method, where catalytic particles were dispersed in solvents and then loaded on a substrate membrane by filtration. Other methods such as phase inversion, hydrothermal treatment, impregnation, electro-deposition, wet-spinning, atomic layer deposition, etc. were also widely used. To conduct AOP with catalytic membranes, four different experiment modes have been used based on the employed filtration mode (dead-end or cross-flow) and the operation mode (single pass or recirculation), which are shown in Figure 1a–d. The membranes can also be used in the batch mode where no filtration is performed (Figure 1e). This method is widely employed in the literature, but in this case, the membranes acted as structured catalysts rather than membranes, which is outside the scope of this current review and not discussed below.

Ideally, catalytic membranes should be repeatedly used with stable activity. However, although organic pollutants can be degraded with them, complete mineralization of the pollutants is hard to achieve. As a result, oxidation intermediates were produced and accumulated on the membrane, leading to membrane fouling and decreased catalytic activity. Considering this, in order to evaluate the lifespan of a certain catalytic membrane, its performance stability and regeneration property should be investigated. When evaluating the performance stability, the catalytic membrane was used for multi-cycles or for a long operation time without any treatment. In this process, the substrate membrane and the catalysts loaded on them were used as a whole, and there was no need to separate them. The regeneration property is investigated in a different way, where the used membrane was regenerated by a certain method (solvent washing, heat treatment, or other methods) and then used again. Note that in some cases, the catalysts on the substrate membrane were separated, regenerated, and loaded again, and performance of the re-fabricated catalytic membrane was compared with the fresh one to illustrate the regeneration efficiency. This method is only applicable to the catalytic membranes formed by filtration, where the catalysts generally weakly interact with the substrate membrane and can be easily separated by back washing or other physical methods.

In the past decades, catalytic membranes have been extensively investigated for water treatment, and several excellent reviews have been published faced with this fast-developing field. For example, in our previous work [5], the coupled process integrating membrane filtration and electrochemical oxidation was reviewed, where the strategies for process design, the membrane materials, and the specific features of the coupled process were discussed. Electroactive membranes were reviewed in another work [6], which included an energy analysis of the overall process. In a more recent review [7], the application of catalytic membranes coupled with electrochemical oxidation for the treatment of antibiotic residues was summarized, focusing on the selection of various membrane electrode materials and their modification methods. The coupling of membrane filtration with peroxide and persulfate was reviewed in several other works [8,9,10], where the different process configurations, membrane materials, fabrication methods, and reaction mechanisms were summarized and discussed. The catalytic membranes coupled with various AOPs were discussed in a recent work in terms of fabrication methods and reaction mechanisms [11]. In another work [12], the removal of organic pollutants from wastewater by AOP was reviewed, where the coupling of AOP with membrane filtration was discussed based on the process figurations.

In these previous review works, although the performance stability and regeneration property of several catalytic membranes were briefly mentioned, the contents were mainly concerned with material design and mechanism analysis. To the best of our knowledge, no recent review paper focusing on the performance stability and regeneration property of catalytic membranes for the removal of organic pollutants from water is available. In this work, we intend to present a comprehensive review focusing on this issue based on the literature in the recent five years (2017–2022). The performance stability of catalytic membranes was first summarized in terms of the two commonly employed evaluation ways, i.e., multi-cycle experiments and long-time filtration, where the different deactivation reasons were discussed. After that, the regeneration property of the catalytic membranes was summarized and discussed according to the different regeneration methods including solvent washing, heat treatment, and others. This review further highlights the current achievements and hurdles in the performance stability and regeneration property of catalytic membranes.

2. Reaction Mechanisms and Impacting Factors of Catalytic Membranes Coupled with Advanced Oxidation Process

When flowing through the catalytic membranes, the organic pollutants and the oxidants would react with each other thanks to the catalytic effects of the composite membranes, leading to the degradation of these organic pollutants. Note that adsorption, which is generally involved in heterogeneous reactions, may also play a role in this process. However, if adsorption is the main mechanism responsible for the removal of pollutants when treated with a membrane, this membrane should be considered as an adsorption membrane rather than a catalytic membrane. Saturated adsorption membranes cannot be reused again without desorption treatment, which is out of the scope of the current review.

In the catalytic filtration process, the direct oxidation of the pollutants by the oxidants may make some contributions, but in most cases, the reaction is not this straightforward. In this section, the reaction mechanisms for the catalytic membrane-based AOP will be briefly summarized based on previous reports [5,11,12].

As mentioned above, AOPs can be classified as persulfate oxidation, H₂O₂ oxidation (Fenton oxidation), ozone oxidation, photocatalytic oxidation, electrochemical oxidation, and others such as catalytic wet oxidation and sonolysis. In all of these processes, the input of oxidants and/or energy is required, which pave the oxidation of pollutants. The input oxidants include persulfate, H₂O₂, O₃, and O₂. The input energy is UV/visible light and electric energy for photocatalytic and electrochemical oxidation, while in other cases, heat should be considered as the input energy, although the reactions are usually conducted at room temperature. The oxidation pathways can be generally classified into the following two series (Figure 2):

(a)

the radical pathway, where various radicals are produced and then react with the pollutants;

(b)

the non-radical pathway, where the oxidation is achieved by direct electron transfer or with the assistance of other reactive species other than radicals.

Generally, various radical species including the hydroxyl radical (·OH), sulfate radical (SO₄^·−), superoxide radical (O₂^·−), and others may be involved in the radical pathways. These radical species are produced in different procedures among different AOPs. For photocatalysis, the absorption of the light source led to the excitation of the semiconductive catalyst, producing valence band holes (h⁺) and conduction band electrons (e⁻), and h⁺ further reacted with water to generate ·OH. For electrochemical oxidation, ·OH is produced through water oxidation on the anodes. For persulfate oxidation, Fenton oxidation, and ozone oxidation, ·OH is produced by the reactions between these oxidants (persulfate, H₂O₂, and O₃) and the catalytic species. SO₄^·− is a unique radical produced from decomposition of persulfate, but it may also become involved in electrochemical oxidation when sulfates are used as the supporting electrolyte. Several methods have been employed for the determination of oxidative species in AOP systems. Conducting quenching experiments is an indirect method, where they can be identified based on the varied reaction rates between scavengers and them. Electron spin resonance (ESR) is another widely used approach, where the radicals are trapped by 5,5-dimethyl-1-pyrroline N-oxide (DMPO) and other reagents to produce spin adducts, providing qualitative and quantitative information with their hyperfine coupling constants and signal intensities. The concentrations of the oxidative species can also be determined by the high-performance liquid chromatography method through the reaction of benzoic acid and other probes with them. The non-radical pathways are also complex, where various mechanisms including singlet oxygen (¹O₂), high-valent metal-oxo species, as well as direct electron transfer may all make some contributions.

Even in a specific AOP, the involved reaction pathways may vary from case to case, which depends on the type of catalyst, the structure of the target pollutant, the reaction conditions, and the co-existent species in the water matrix. More in-depth discussions on the reaction mechanisms in the AOPs are out of the scope of the current review. No matter what mechanisms are involved, all of these pathways will lead to the consumptions of oxidants and energy, resulting in the oxidation of the organic pollutants in water. To evaluate the performance of this process, the removal efficiency, which is calculated by measuring the change in concentration of the target pollutant before and after treatment, is the generally utilized index. For catalytic membrane-based AOP, other indexes, such as the reaction time and apparent kinetic constant in the recirculation mode as well as the trans-membrane pressure (TMP) and flux in the single pass mode, should also be paid attention to. In addition, remember that complete mineralization of the pollutants is hard to achieve, and the produced oxidation intermediates may pose a potential threat. To evaluate the mineralization degree, the removals of total organic carbon (TOC) and chemical oxygen demand (COD) can be also measured.

The performance of a certain catalytic membrane to removal a certain target pollutant may be impacted by many factors. Operation conditions, including initial concentration of the pollutant, the dosage of oxidant, the amount of energy input (temperature, intensity of light, current density, etc.), and flux, will certainly play some roles. Environmental factors, including pH, inorganic ions, and natural organic matter, may also influence the removal efficiencies. The pH will influence the charge properties of both the catalytic membrane and the pollutant molecules as well as the stability of the oxidants, altering their interactions and changing the reaction results. The inorganic ions and natural organic matter can take part in the radical reactions, playing either positive or negative roles. These factors may also influence the performance stability and regeneration property of catalytic membranes by changing the level of metal leaching or the concentration and distribution of oxidation intermediates.

In the following sections, the performance stability and regeneration property of catalytic membranes are reviewed with the results in previous works summarized in tables. The removal efficiencies as the main index evaluating the performances of catalytic membranes are listed in the tables for comparison, together with some important factors including the target pollutants, the initial concentrations, and the fluxes. The performance stability of catalytic membranes is first discussed in Section 3, after which the regeneration property of the catalytic membranes is discussed in Section 4.

3. Performance Stability of Catalytic Membranes Coupled with Advanced Oxidation Process

Generally, the performance stability of catalytic membranes can be evaluated in two ways. In the first method, after the fresh membrane was used for the first time, it was directly used again in the catalytic process without any regeneration treatment. The catalytic run can be performed in several cycles with the same membrane. The same operational factors were used, and the removal efficiencies of the target pollutant in all these catalytic runs were recorded and compared. In the second method, the catalytic run was conducted continually for a long period of time, and samples of the filtrate were withdrawn periodically to record the change in removal efficiency of the pollutant. In this section, recent works regarding these two evaluation methods will be discussed in Section 3.1 and Section 3.2, with the previously reported experimental results summarized in

Table 1. Performance stability of catalytic membranes tested by multi-cycle experiments.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Operation Mode	Target Pollutant	C0	Flux (L/(m2·h))	R1	Number of Cycles	η	Ref.
PMS	PVDF	CoAl-LDH	filtration	flat-sheet	dead-end	recirculation	RA	2.5 mg/L	80.3	94%	10	~100%	[3]
PMS	PVDF	Co2+/Mxene	filtration	flat-sheet	dead-end	single pass	TC	10 mg/L	65.7	98.2%	3	84%	[4]
PMS	PVDF	FeOCl/MoS2	filtration	flat-sheet	dead-end	single pass	RhB	10 mg/L	400	70%	5	87%	[13]
PMS	PVDF	MnO2/CNT	filtration	flat-sheet	dead-end	single pass	RhB	50 mg/L	248.51	100%	8	100%	[14]
PMS	PVDF	FeCoS@N-rGO	filtration	flat-sheet	dead-end	single pass	SMX	30 mg/L	632.12	99.2%	5	91%	[15]
PMS	PVDF	Co@N-C	phase inversion	flat-sheet	cross-flow	single pass	TC	20 mg/L	636.0	99.3%	5	91%	[16]
PMS	PTFE	Co3O4@NCNTs/g-CN	filtration	flat-sheet	dead-end	recirculation	SMX	10 mg/L	276.86	98.9%	5	89%	[17]
PMS	PTFE	FeCo@GCTs/GO	filtration	flat-sheet	dead-end	recirculation	SMX	10 mg/L	487.3	98.4%	5	91%	[18]
PDS	PTFE	N-doped GO	filtration	flat-sheet	dead-end	recirculation	phenol oxalic acid	50 mg/L 90 mg/L	1125	94% 100%	3	29% 80%	[19]
PMS	PAN	Fe-doped LaCoO3	phase inversion	flat-sheet	cross flow	single pass	TC	10 mg/L	220.7	99%	5	99%	[20]
PMS	PSF	SCCM	phase inversion	flat-sheet	cross-flow	single pass	RhB	20 mg/L	126.90	100%	5	~100%	[21]
PMS	nylon	Fe2O3@CNT	filtration	flat-sheet	dead-end	single pass	TC	0.04 mmol/L	16.3	96.1%	5	91%	[22]
PDS	nylon	NG/rGO/CNT	filtration	flat-sheet	dead-end	recirculation	SMX	0.5 mg/L	46.15	94.8%	5	85%	[23]
PMS	ceramic membrane	CuO	phase-inversion and liquid-phase sintering	hollow fiber	cross-flow	single pass	RhB	20 mg/L	500	81.5%	5	93%	[24]
PMS	ceramic membrane	CuO	impregnation and calcination	hollow fiber	cross-flow	single pass	SDZ	1 mg/L	70	91.7%	5	93%	[25]
PMS	ceramic membrane	CuO	impregnation and calcination	hollow fiber	cross-flow	single pass	BPA	10 mg/L	70	91.4%	5	~100%	[26]
PMS	ceramic membrane	Co3O4@CNT	filtration	flat-sheet	dead-end	single pass	MB	15 mg/L	224	84.4%	6	83%	[27]
PMS	ceramic membrane	Mn2O3	spraying and calcinating	flat-sheet	dead-end	single pass	acetaminophen	1 mg/L	60	99%	10	~100%	[28]
PMS	ceramic membrane	Co@GAC	filling the channels	flat-sheet	dead-end	single pass	BPA	10 mg/L	35	92.8%	6	97%	[29]
H2O2	FEP	Fe2O3	thermoforming method	flat-sheet	cross-flow	single pass	MB	40 mg/L	15	99.69%	3	73%	[30]
H2O2	PAN	goethite/maleate ferroxane	phase inversion	flat-sheet	dead-end	single pass	amoxicillin	105 mg/L	23.2	86.3%/92.3%	4	97%/98%	[31]
O3	ceramic membrane	CuO	filtration and calcination	tubular	dead-end	recirculation	1,4-dioxane	200 mg/L	363.81	46%	4	98%	[32]
O3	ceramic membrane	MnCe oxide	filtration, in situ reaction and calcination	flat-sheet	dead-end	recirculation	atrazine	1 mg/L	60	90%	5	94%	[33]
O3	ceramic membrane	CuMn2O4	impregnation and calcination	tubular	cross-flow	single pass	BP-3	2 mg/L	63.157	75.3%	30	91%	[34]
O3	ceramic membrane	N-rGO	pneumatic method	tubular	cross-flow	single pass	benzotriazole	0.084 mol/L	660.86	100%	18	92%	[35]
Photo	PVDF PTFE	TiO2	impregnation and hydrothermal treatment	flat-sheet	dead-end	recirculation	diclofenac/ ethinylestradiol	1.01 µmol/L	1144/1331	92%, 94%; 89%, 92%	3	~100%, 96%; 67%, 83%	[36]
Photo	ceramic membrane	TiO2	impregnation and calcination	flat-sheet	cross-flow	recirculation	MB	0.083 mmol/L	125	80%	4	~100%	[37]
Photo	ceramic membrane	GO–TiO2	filtration and calcination	flat-sheet	dead-end cross-flow	recirculation recirculation	naproxen diclofenac CBZ	50 ppb	92.6	100%, 100%, 90%; 100%, 100%, 30%	7 11	–	[38]
Photo	cellulose acetate	GO–TiO2	filtration	flat-sheet	dead-end	single pass	CR	50 mg/L	483.5	97%	3	98%	[39]
Photo	PSF	GO–N-TiO2	filtration	flat-sheet	dead-end	single pass	MB	50 mg/L	70	50%	8	66%	[40]
Photo	None: free standing	rGO@TiO2	electro-spinning and calcination	flat-sheet	cross-flow	recirculation	propranolol	2 mg/L	342.86	76.1%	10	~100%	[41]
Photo	PVDF	ZnIn2S4	filtration	flat-sheet	dead-end	recirculation	fluvastatin	10 mg/L	270	97.19%	6	94%	[42]
Photo + H2O2	ceramic membrane	α-Fe2O3	spraying and calcinating	flat-sheet	dead-end	recirculation	TC	20 mg/L	158.73	82%	5	88%	[43]
Photo + H2O2	PAN	β-FeOOH	impregnation and in situ mineralization	flat-sheet	cross-flow	single pas	methyl blue	20 mg/L	11–13	99.8%	5	97%	[44]
Photo + H2O2	air-laid paper	β-FeOOH	impregnation and in situ mineralization	flat-sheet	dead-end	single pass	TC	40 mg/L	600	99.5%	5	98%	[45]
Photo + H2O2	PSF	ferrocene	layer-by-layer interfacial polymerization and ion-exchange	flat-sheet	dead-end	single pass	BPS	20 mg/L	4.2	99.5%	3	–	[46]
Photo + PMS	PP	ZIF-67	in situ synthesis	flat-sheet	cross-flow	single pass	MB MO	20 mg/L 20 mg/L	216.8	TOC removal: 74.2% 83.5%	5	99% 98%	[47]
Photo + PMS	PTFE	α-Fe2O3/bacterial cellulose	filtration	flat-sheet	dead-end	single pass	RhB	20 mg/L	4.93	100%	5	97%	[48]
Photo + PMS	PVDF	CoFe2O4/carbon nanofiber	phase inversion	flat-sheet	dead-end	recirculation	berberine	10 mg/L	368	93.2%	4	97%	[49]
Photo + PDS	PVDF	–	–	flat-sheet	dead-end	single pass	ofloxacin	5 mg/L	–	54%	10	26%	[50]
Electro	porous titanium plate (anode) titanium mesh (cathode)	Pd (anode) –(cathode)	spray coating and calcination	flat-sheet	dead-end	single pass	SMX	100 μg/L	5000	96.3%	10	~100%	[51]
Electro	RuO2 plated titanic mesh (anode) RuO2 plated titanic mesh (cathode)	graphene/SnO2/carbon nanofibers (anode) –(cathode)	electrospinning and carbonization	flat-sheet	dead-end	recirculation	SMX	15 mg/L	216.4	85%	6	94%	[52]
Electro	porous Ti plate (anode) steel mesh (cathode)	blue TiO2 (anode) –(cathode)	electrooxidation, calcination and electroreduction	flat-sheet	dead-end	recirculation	triclosan	10 mg/L	3582	98.5%	4	~100%	[53]
Electro	stainless steel network/PTFE (anode) stainless steel network (cathode)	CNT (anode) –(cathode)	filtration	flat-sheet	dead-end	single pass	SMX ciprofloxacin amoxicillin	50 mg/L	127.48	90% 76% 98%	4	72% 79% 92%	[54]
Electro	titanium ring/PTFE (anode) stainless steel (cathode)	single cobalt atom and nitrogen atom co-doped graphene (anode) –(cathode)	filtration	flat-sheet	dead-end	single pass	MB	5 mg/L	15.29	99.5%	5	90%	[55]
Electro	stainless-steel mesh/PVDF (anode) titanium foil (cathode)	nano porous carbon/CuFeO2 (anode) –(cathode)	phase inversion	flat-sheet	dead-end	single pass	MO	10 mg/L	125	98.72%	4	69%	[56]
Electro	outer CNT layer (anode) inner CNT layer (cathode)	–(anode) –(cathode)	layer-by-layer coating	hollow fiber	dead-end	single pass	MC-LR	0.5 mg/L	500	99.8%	3	~100%	[57]
Electro	ceramic membrane (anode)/Ti mesh titanium mesh (cathode)	TiO2/SnO2-Sb (anode) –(cathode)	adhesive assemble	flat-sheet	dead-end	single pass	2,4-D	1 mg/L	278	62.4%	8	~100%	[58]
Electro + PMS	titanium mesh (anode) PTFE (cathode)	–(anode) MnFe2O4-rGO (cathode)	filtration	flat-sheet	dead-end	recirculation	OTC	10 mg/L	553.6	88.7%	5	86.1%	[59]
Electro + PMS	ceramic membrane (anode and cathode)	Pd (anode and cathode)	confocal magnetron co−sputtering	flat-sheet	cross flow	single pass	MB	10 μmol/L	461.34	94.5%	8	95%	[60]
Electro + Photo + PMS	stainless steel (anode) titanium plate (cathode)	TiO2-ZnO (anode) –(cathode)	atomic layer deposition	flat-sheet	cross flow	single pass	atrazine MB	5 mg/L 5 mg/L	59.9	100%	5	90% 85%	[61]

Abbreviations: BPA: Bisphenol A; BPS: Bisphenol S; BP-3: Benzophenone-3; CBZ: Carbamazepine; CR: Congo red; CNT: Carbon nanotube; OTC: Oxytetracycline; TOC: Total organic carbon; GAC: granular activated carbon; LDH: layered double hydroxide; MB: Methylene blue; MC-LR: Microcystin-LR; MO: Methyl orange; N-rGO: Nitrogen-doped reduced graphene oxide; PAN: Polyacrylonitrile; PDS: Peroxydisulfate; PMS: Peroxymonosulfate; PP: Polypropylene; PSF: Polysulfone; PTFE: Poly(tetrafuoroethylene-co-hexafuoropropylene); PTFE: Polytetrafluoroethylene; PVDF: Polyvinylidene fluoride; RA: Ranitidine; RhB: Rhodamine B; SDZ: Sulfadiazine; SMX: Sulfamethoxazole; TC: Tetracycline; ZIF-67: Zeolitic imidazole framework-67; 2,4-D: 2,4-Dichlorophenoxyacetic acid.

and Table 2, respectively.

3.1. Performance Stability of Catalytic Membranes Tested by Multi-Cycle Experiments

In this section, the performance stability of catalytic membranes evaluated by multi-cycle experiments was summarized and analyzed (Table 1), where the related reports were classified based on the type of advanced oxidation process. The employed substrate membrane, catalyst, fabrication method, membrane type (flat-sheet, tubular, or hollow fiber), filtration type (dead-end or cross-flow), operation mode (single pass or recirculation), target pollutant, initial concentration of pollutant, flux, removal efficiency of pollutant, and number of cycles are provided in the table.

For comparison of the performance stability results in these works, a dimensionless factor (η) was determined as follows:

η = R_n/R₁ × 100%,

(1)

where n stands for the number of catalytic runs conducted for the same membrane. R_n represents the removal efficiency of the target pollutant obtained in the last catalytic run, while R₁ stands for the value obtained with the fresh membrane. Thus, η should be in the range of 0–100%, and a larger η indicates better performance stability within the same number of catalytic cycles.

As summarized in Table 1, for the performance stability tests of these catalytic membranes tested by multi-cycle experiments, the employed number of cycles were in the range of 3–30, obtaining η values in the range of 26–100%. Obviously, a larger number of cycles and a higher η value should indicate better performance stability. From this point of view, three catalytic membranes, including the CoAl-LDH/PVDF membrane, the Mn₂O₃/ceremic membrane, and the free standing reduced graphene oxide (rGO)@TiO₂ membrane showed outstanding performances, all of which could be repeatedly used for 10 cycles with almost no deterioration in the removal efficiencies towards their target pollutants. Another two membranes, i.e., the CuMn₂O₄/ceramic membrane and the N-rGO/ceramic membrane, also performed very well, showing η values over 90% after repeatedly used for more than 10 cycles (30 and 18 cycles), respectively.

In the following sub-sections, the summarized results in Table 1 are further discussed based on the type of advanced oxidation process.

3.1.1. Multi-Cycle Performance of Catalytic Membranes Coupled with Persulfate Oxidation

In persulfate-based oxidation processes, persulfates including PMS and peroxydisulfate (PDS) are utilized as oxidants. In catalytic persulfate oxidation systems, various reactive oxidative species (ROSs) can be generated, such as ·OH, SO₄^·−, O₂^·−, ¹O₂, and high-valent iron-oxo, which then react with the organic pollutants. In addition, the pollutants can also be oxidized through non-radical electron transfer, where the catalyst acts as a medium and bridges the electron transfer between persulfate and the pollutants. Thus, both ROSs and the non-radical electron transfer pathway can lead to oxidation, decomposition, and mineralization of the various organic pollutants in water. In powder-based heterogenous systems, metal-based and carbon-based catalysts have been widely investigated and used for the activation of persulfates. When coupled with the membrane process, metal or carbon catalysts need be introduced to the substrate membranes to fabricate catalytic membranes.

The substrate membranes can be either organic or inorganic. The PVDF membrane is the most frequently used organic substrate membrane. For example, the powder catalyst of CoAl-layered double hydroxide (LDH) was loaded on a PVDF substrate membrane by vacuum filtration. The catalytic membrane was then used for the activation of PMS to degrade RA, which was a histamine H-receptor antagonist for the treatment of gastrointestinal disease, achieving a removal efficiency of 94% under optimum conditions [3]. The removal efficiency remained stable in 10 cycles without obvious changes, indicating the high stability of the LDH catalytic membrane [3]. In another report [4], Mxene modified by Co²⁺ was also loaded on a flat PVDF membrane by vacuum filtration. The resulting Co²⁺/Mxene catalytic membrane could efficiently degrade 98.2% TC, which was an antibiotic, in the presence of PMS. When reusing the membrane, the removal efficiency dropped to 90.2% and 82.5% in the second and third run, respectively [4]. When FeOCl/MoS₂ was used as the catalyst in another work [13], a similar vacuum filtration method was utilized to immobilize the catalyst to a PVDF membrane, except that the substrate membrane was modified first by coating with polydopamine (PDA). The FeOCl/MoS₂ membrane was employed to remove Rhodamine B (RhB) in five cycles, showing acceptable performance stability with a η value of 87% [13]. In some reports, composite catalysts constructed from both metal and carbon materials are investigated. For instance, MnO₂/carbon nanotube (CNT) [14] and FeCoS@N-rGO [15] were combined with PVDF, and the obtained catalytic membranes were used to degrade RhB and sulfamethoxazole (SMX), respectively. Both membranes could maintain high removal efficiencies in multicycle tests, showing η values of 100% [14] and 91% [15] after 8 and 5 cycles, respectively.

Besides the vacuum filtration method, the phase inversion method is another commonly employed approach to fabricate PVDF-based catalytic membranes. For instance, cobalt nanoparticles were encapsulated into nitrogen-doped carbon shells, and the resulting composite material (Co@N-C) was introduced within the pores of PVDF membrane by the phase inversion method [16]. Using this membrane, the degradation of TC was operated in the cross-flow mode, achieving 99.3% removal in 60 min. After being used for 5 cycles, the removal efficiency was still above 90%. The high stability of the PVDF/Co@N-C membrane was not only attributed to the stability of the Co@N-C catalyst but also to the strong adhesion of the catalysts to the membrane channels [16].

In addition to PVDF, other types of organic substrate membranes are also investigated, including polytetrafluoroethylene (PTFE), polyacrylonitrile (PAN), polysulfone (PSF), nylon, and others. Similar to the case of PVDF, these organic membranes could not activate persulfate, acting as supports only. To introduce catalytic species, various metal-based and carbon-based catalysts as well as their composite were loaded, including Co₃O₄@NCNTs/g-CN [17], Fe₂O₃@Co_0.08Fe_1.92@nitrogen-doped rGO@CNT (FeCo@GCTs/GO) [18], N-doped GO [19], Fe-doped LaCoO₃ [20], SrCu_xCo_1−xO_3−λ-deposited MCM-41 (SCCM) [21], Fe₂O₃@CNT [22], and nitrogen-doped graphene (NG)/rGO/CNT [23]. These as-synthesized catalytic membranes possessed high removal efficiencies of over 90% for the target pollutants, but the efficiencies declined after being used cyclically for 3–5 times. For the membrane of SCCM/PSF, although a η value of ~100% was achieved in the fifth cycle for the degradation of Rhodamine B, the authors noted that a minor decline in the removal efficiency could be observed and attributed this to the adsorbed organic molecules and their degradation products [21]. This is a generally mentioned reason when discussing the declined performance in multi-cycle degradation tests. The adsorbed organic species could block the active sites, hindering the adsorption and activation of continuously entering persulfates and consuming them as well. Consequently, the performance of the catalytic membrane gradually deteriorates in multi-cycle tests due to the accumulation of these organic species.

Moreover, other reasons may also lead to decreased activities in multi-cycle tests. For metal-based catalysts, leaching of metal species is always observed. Especially, for persulfate-based oxidation processes, the addition of persulfates would shift the solution pH to more acidic values and enhance metal leaching. These leached ions may promote the degradation of organic pollutants through homogeneous reactions. Thus, the loss of metal ions through leaching may be a possible reason for the weakened catalytic performance of the membrane. Considering this, the concentration of leached metal ions is generally measured. For instance, with the FeCo@GCTs/GO membrane, the leached concentrations of total Fe and Co ions were measured to be 0.24 and 0.31 mg/L [18]. Furthermore, when conducting SMX degradation with the dissolved ions, low removal efficiencies less than 7% were observed for both Fe²⁺ and Co²⁺ [18]. This result suggested that homogeneous activation under this low metal leaching level only made a minor contribution, and the degradation of the pollutants was mainly achieved through heterogenous reactions. However, in the membrane filtration process, metal species continuously leached out, and the content of metal catalysts in the membrane became lower and lower. Thus, even a low leaching level may lead to distinct decrease in the heterogenous catalytic activity due to the accumulation of metal loss in multi-cycle catalytic filtration runs. Another reason for the decreased activity of metal-based catalysts is the change in valence states. Although metal species of both low and high-valence states could activate persulfates through cyclic conversion, the transformation from high-valence species (Co(III), Fe(III), etc.) to low-valence species (Co(II), Fe(II)) is much slower. As a result, when the fraction of high-valence species in the catalysts is increased, lower catalytic performance will be expected. As an example, after being used for the degradation of TC, both the ratios of Co(II)/Co(III) and Fe(II)/Fe(III) in the Fe-doped LaCoO₃/PAN membrane decreased, which may result in the slightly declined removal efficiency after five repeated runs [20].

When it comes to metal-free carbon-based catalytic membranes, other reasons may also make a contribution. As an example, the nitrogen-doped graphene (NG)/rGO/CNTs/nylon membrane was utilized to active PDS for the degradation of SMX [23]. The fresh membrane possessed an I_D/I_G ratio of 1.82 as characterized by Raman spectroscopy, indicating the defects-rich structure of the carbon catalysts and exhibiting strong adsorption and catalytic activity. After 5 filtration cycles, the I_D/I_G ratio was decreased to 1.50, indicating the consumption of these surface defects and the production of oxygen-containing functional groups, which was responsible for the reduction in removal efficiency [23]. In another report, the decreased efficiency of the N-doped GO/PTFE membrane after PDS activation was partially attributed to the attack of radical species on GO, which led to the formation of GO fragments and broke the structure of the membrane [19].

Compared to organic polymeric membranes, inorganic membranes are more costly due to their energy and time-consuming production procedures, but they also possess several merits including superior thermal, mechanical, and chemical stabilities as well as long service life. Thus, inorganic membranes are also widely investigated as substrate membranes to fabricate catalytic membranes. Inorganic membranes can be roughly classified as ceramic membranes, metal membranes, and carbon membranes, among which ceramic membranes are employed in a higher frequency. Similar to the case with polymeric membranes, the pristine ceramic membranes, which are often synthesized using Al₂O₃ as the main matrix material, show limited activity in persulfate-based oxidation processes. Thus, active metal species should be added to fabricate catalytic membranes. These additional species can be added in two ways, i.e., during the preparation of the ceramic membranes, or through post-treatment modification of synthesized ones. For instance, a CuO-doped ceramic membrane was constructed by the combined phase-inversion and liquid-phase sintering method starting from Al₂O₃, TiO₂, and CuO powders [24]. In several other reports [25,26], CuO was also employed as the active metal species, but it was introduced by post-treatment modification through an impregnation and calcination method using aqueous solutions containing Cu salts. Filtration can also be used to modify the pristine ceramic membrane, such as in the case of loading Co₃O₄@CNT [27]. When Mn₂O₃ was used to modify the ceramic membrane in another work [28], the pristine membrane was sprayed with Mn₂O₃-Al₂O₃ slurry and then calcinated, obtaining the catalytic membrane.

In a recent report [29], the authors intended to employ Co-coated granular-activated carbon (Co@GAC) as the catalytic species. The above methods are not applicable to introduce Co@GAC to the ceramic membrane because the GAC could not withstand the high-temperature calcination process. Thus, the authors designed a novel method by filling the channels of the ceramic membrane with Co@GAC (Figure 3). The Co@GAC-membrane module also contained a PP punching mesh, which could prevent Co@GAC from flowing out of the channels during filtration process [29].

The above-mentioned catalytic ceramic membranes were employed to degrade various organic pollutants including Rhodamine B, sulfadiazine (SDZ), BPA, and acetaminophen with the assistance of PMS, all possessing high performance stabilities with η values larger than 90% after being used for 5–10 cycles. Similar to the organic substrate membrane-based ones, the slightly decreased performance could be attributed to the adsorption of pollutants/oxidation products, the slight metal leaching, and/or the change in their surface properties such as the valence states of the active metal species.

3.1.2. Multi-Cycle Performance of Catalytic Membranes Coupled with H₂O₂ Oxidation

In the Fenton oxidation process, H₂O₂ was used as the oxidant. The reaction between H₂O₂ and catalysts, such as Fe²⁺ and Fe³⁺ ions, led to the production of ·OH, which then paved the oxidation of organic pollutants. Compared to homogenous ions, heterogeneous iron-based catalysts are more environmentally friendly and can be loaded on substrate membranes for Fenton oxidation applications. For example, a composite membrane was synthesized by combining poly(tetrafuoroethylene-co-hexafuoropropylene) (FEP) and Fe₂O₃ in a thermoforming method, where FEP resin, Fe₂O₃, and dioctyl phthalate (DOP) as a plasticizer were mixed and then treated with a thermo-compressor at 270 °C [30]. The membrane with 10 wt.% Fe₂O₃ could remove 99.69% methylene blue (MB) when operated in the cross-flow mode, and the removal efficiency declined to 95.15% and 72.85% in the second and third degradation runs. The main reason for the decreased performance was attributed to the blocking of pores in the membranes. By decreasing the content of Fe₂O₃ to 5 wt.%, the performance stability was increased, achieving ~91%, 90.85%, and 82.35% in three consecutive runs [30]. In another report, goethite and maleate ferroxane as green and inexpensive iron oxy-hydroxides (FeOOH) were combined with PAN by a phase inversion method, and the resulting catalytic membranes could efficiently remove amoxicillin in the dead-end filtration mode [31]. Close removal efficiencies of 86.3% and 92.3% were obtained with goethite and maleate ferroxane, respectively, and the slight reductions in multi-cycle runs were ascribed to the accumulation of organic molecules on the membrane surface and the concentration polarization phenomenon [31].

In the traditional Fenton oxidation process, the reduction of Fe(III) to Fe(II) is generally sluggish due to the inherently slow kinetics, and H₂O₂ should be added externally. Recently, electrical or light energy has been proved to speed up the regeneration of Fe(II) species. Meanwhile, H₂O₂ can sometimes be produced in situ in these processes, avoiding the safety issues related to its storage and transportation. These emerging Fenton oxidation processes, known as electro-Fenton and photo-Fenton oxidation processes, are discussed later in Section 3.1.4 and Section 3.1.5 below.

3.1.3. Multi-Cycle Performance of Catalytic Membranes Coupled with Ozone Oxidation

Ozone is another widely employed oxidant for pollutant degradation. Although ozone can directly oxidize some pollutants, low removal efficiency and limited mineralization are achieved with sole ozone. In contrast, catalysts can trigger the decomposition of ozone to produce ·OH with a higher redox potential, favoring the oxidation and mineralization of organic pollutants.

The multi-cycle performance of catalytic membranes in the ozonation process has been reported in several reports. CuO was loaded on a tubular ceramic membrane, and the fabricated catalytic membrane could remove 1,4-dioxane stably in 4 cycles in the presence of ozone [32]. The performance stability of the composite membrane was attributed to the low leaching of Cu, which was measured to be 0.58 ppm in each cycle [32]. In another work, MnMe oxides (Me = Fe, Co, Ce) were embedded on ceramic membranes through filtration, in situ reaction with KMnO_4, and calcination. The one loaded with MnCe oxide showed the best catalytic ozonation performance among them three, and the η value was found to be 94% after 5 degradation runs, which was attributed to the slight membrane fouling as well as leaching of Mn and Ce [33]. For the degradation of benzophenone-3 (BP-3), another bi-metal oxide CuMn₂O₄ was loaded on a ceramic membrane through impregnation and calcination [23]. The prepared catalytic membrane showed excellent performance stability, showing a high η value of 91% after being used for 30 cycles. The main reason for the slightly decreased removal efficiency was inferred to be leaching of Cu and Mn, which were around 0.032 and 0.230 mg/L, while the contribution from the accumulated oxidation intermediates was considered to be negligible [34]. In contrast, the declined removal efficiency for the nitrogen-doped reduced graphene oxide (N-rGO) on the ceramic membrane, which was a metal-free catalyst, was mainly attributed to the adsorption of oxidation products on its active sites [35].

3.1.4. Multi-Cycle Performance of Catalytic Membranes Coupled with Photocatalysis

In the advanced oxidation process, photo irradiation can be introduced as an external energy source, promoting the production of holes, ·OH, and other oxidative species. In some cases, additional oxidants such as H₂O₂ and persulfates are added, which are known as photo-Fenton and photo-driven persulfate activation processes. Traditionally, ultraviolet (UV) radiation is usually used as the light source in the photocatalysis process. However, UV only takes a small part in the natural sunlight (<5%), and the utilization of visible light is receiving increasing attention recently.

By loading photocatalysts on various substrate membranes, a series of composite membranes have been fabricated and successfully utilized for the degradation of different organic pollutants. In a recent report, TiO₂, the most widely investigated photocatalyst, was loaded on three different polymer membranes including PSF, PVDF, and PFTE [36]. The three catalytic membranes were then used for the photocatalytic degradation of diclofenac and ethinylestradiol. Interestingly, although the TiO₂/PSF membrane possessed the highest activity, it was damaged after the first run due to the low resistance of PSF to UV exposure. The other two membranes were tested for three consecutive runs, and the declined removal efficiencies were attributed to the adsorption of pollutants on the membrane as well as leaching of TiO₂ [36]. In another work, TiO₂ was loaded on a ceramic membrane, which could stably degrade MB in four cycles [37]. The composite of graphene oxide (GO) and TiO₂ was loaded on another ceramic membrane, and the degradation of three organic pollutants was conducted [38]. Under both dead-end and cross-flow filtrations modes, the multi-cycle performances were investigated by calculating the rate constants, and different results were revealed. No activity loss but a slight enhancement was observed for the dead-end filtration mode owing to the compaction effect during pressure-driven filtration. For the cross-flow filtration mode, some loss in activity was found, which was ascribed to the shear force under high flow rate [38]. The GO-TiO₂ composite was loaded on a cellulose acetate substrate membrane in another work [39], and the catalytic membrane showed consistent dye removal ability for Congo red (CR) in three cycles. To improve the photocatalytic efficiency, TiO₂ was modified by nitrogen doping to obtain N-TiO₂ and then loaded on a PSF membrane together with GO [40]. The prepared catalytic membrane could remove MB under the irradiation of sunlight, but the photocatalytic ability decreased along with repeated runs, which was caused by the loss of catalysts. Different from the filtration fabrication method used in these works, the catalyst of reduced GO (rGO)@TiO₂ was loaded on a stainless-steel mesh by electro-spinning and calcination to fabricate the rGO@TiO₂ catalytic membrane [41]. The membrane possessed a highly stable degradation performance for propranolol with no obvious deterioration over the ten-cycle tests, and this was probably owing to the unique preparation method, which allows the catalysts to form a free-standing membrane [41]. Besides TiO₂-based catalysts, other types of catalysts are also investigated. For instance, ZnIn₂S₄, which has a narrow band gap and shows high activity under the irradiation of visible light, was deposited on PVDF and then utilized for the degradation of fluvastatin. After six catalytic filtration cycles, a high η value of 94% was revealed, and the decreased removal efficiency was caused by the occupation of the active sites and catalyst loss due to the weak interaction between the PVDF substrate membrane and the ZnIn₂S₄ catalyst [42].

When H₂O₂ is added as an external oxidant in the photocatalytic process, the degradation of organic pollutants may be enhanced. For instance, the ceramic membrane with TiO₂ loaded on it could more efficiently remove MB under visible light in the presence of additional H₂O₂, compared to the case without it [37]. The authors noted that the reactive species resulted from the transformation between Ti³⁺ and Ti⁴⁺ in the TiO₂ catalyst, operating in a Fenton-like mechanism similar to the couple of Fe²⁺ and Fe³⁺ [37]. More commonly, the photo-Fenton process is operated with iron-based catalysts. For example, α-Fe₂O₃ as a widely used heterogenous Fenton catalyst was loaded on ceramic membrane, which was then employed for the photo-Fenton degradation of TC, showing a η value of 88% after 5 catalytic runs [43]. Another widely investigated Fenton catalyst, β-FeOOH, was loaded on PAN membrane [44] and air-laid paper [45], respectively. The photo-Fenton activities of the two fabricated membranes were then investigated, and both declined slightly after five catalytic cycles. In another work, a novel ferrocene-based membrane was fabricated by layer-by-layer interfacial polymerization and ion-exchange using PSF membrane as the substrate, which could remarkably remove bisphenol S (BPS) through the photo-Fenton process [46]. The reusability of the membrane was tested by three consecutive runs. The removal efficiencies of BPS in the second and third cycles were not provided, but the degradation rate in the three cycles calculated by the pseudo-first-order kinetic model was reported, which declined from 0.092 to 0.068 min⁻¹ [46]. In these works, the reasons for the weakened performances of the used membranes were not discussed. Based on the low metal leaching and stable properties of the membranes before and after use in these works, the declined activities were probably attributed to the adsorbed degradation intermediates, which were generally detected and analyzed in these reports [43,44,45,46].

When persulfate instead of H₂O₂ is added, the photo-driven persulfate activation process is developed. In this process, classical metal catalysts based on Co or Fe are generally employed. For example, ZIF-67, a Co-containing metal organic framework (MOF), was loaded on a PP membrane by in situ synthesis to remove MB and methyl orange (MO) in the co-presence of visible light and PMS [47]. In another work, α-Fe₂O₃ and bacterial cellulose were together loaded on PTFE membrane, where bacterial cellulose helped to disperse α-Fe₂O₃ particles and thus enhanced the visible light absorption ability of the catalytic membrane [48]. For the treatment of berberine, which was an antibiotic, CoFe₂O₄ and carbon nanofiber were loaded on PVDF membrane. Interestingly, with the assistance of Ag/ZnO nanorods array/Ni foam as the photo-anode, the CoFe₂O₄/carbon nanofiber/PVDF membrane cathode also possessed electricity production, providing a promising strategy towards water treatment with energy recovery [49]. High η values after several cycles were observed for these three membranes, and the slight reduction was attributed to the incomplete degradation of contaminants during the former cycles [47].

In the above three reports, persulfate as the additional oxidant was added along with the feed solution. Interestingly, in another work, the persulfate of PDS was not added in the feed but loaded on the membrane. In that work, PVDF was used as the substrate membrane, and it was modified by N,N-dimethylaminoethyl methacrylate to introduce positively charged sites for the electrostatic attraction of negatively charged PDS anions [50]. The membrane with prior loaded PDS could degrade ofloxacin under UV radiation thanks to the photo-driven activation of PDS. However, due to the gradual consumption of PDS, the membrane could only be used for several times. The author noted that by combining with cathode electrolysis or other technologies, the UV activation process of PDS may be revisable, extending the service life of the membrane [50].

3.1.5. Multi-Cycle Performance of Catalytic Membranes Coupled with Electrocatalysis

Similar to the case in photocatalysis, electron energy can also be introduced into AOP as an external energy source. When conducting electrochemical advanced oxidation, two electrodes are required, i.e., an anode and a cathode. In some cases, the oxidation of organic pollutants is achieved on the anode through either direct oxidation or indirect oxidation, and the cathode only acts as a counter electrode. In some other cases, the cathode also plays an important role. For example, in the electro-Fenton process, the cathode promoted the in situ generation of H₂O₂ with the supply of O₂. In the electron-activated persulfate oxidation process, the cathode contributed to the activation of persulfate. In these two processes, the regeneration of low-valent metal species was also achieved on the cathode, which provided electrons for the reduction of their high-valent counterparts. When coupled with the membrane process, the catalytic membranes can act as either or both, and the compositions of both anodes and electrodes are listed in Table 1 for clarity.

Conductive porous supports made of metal such as stainless steel and titanium can be directly modified to fabricate electrochemical catalytic membranes. For example, Pd was supported on porous titanium plate by spray coating and calcination to fabricate a membrane anode. Using titanium mesh as a cathode, the electrochemical membrane system could remove SMX stably in 10 cycles [51]. In another work, the same RuO₂-plated titanic mesh was used as both anode and cathode, except that the anode was further modified through electrospinning and carbonization to load a layer of graphene/SnO₂/carbon nanofibers. The system also possessed high stability for SMX degradation with a η value of 94% after 6 cycles [52]. A porous Ti plate was modified to introduce blue TiO₂ on its surface, which could efficiently and stably remove triclosan by sequential adsorption and electrochemical degradation [53].

Non-conductive supports such as PTFE and PVDF can also be modified to construct electrochemical catalytic membranes, but an additional conductive support is generally required. For instance, a CNT/PTFE membrane was combined with a stainless-steel network to act as the anode, and a pristine stainless-steel network served as the cathode. The system could degrade several organic pollutants including SMX, ciprofloxacin, and amoxicillin, showing different η values of 72%, 79%, and 92% after 4 cycles. The weaker performance stability of the former two pollutants was probably attributed to the high molar weight oxidation products formed during their degradation, which accumulated on the membrane and led to higher mass transfer resistance [54]. Single cobalt atom/nitrogen atom co-doped graphene [55] and nano porous carbon/CuFeO₂ [56] were employed for the modification of PTFE and PVDF membranes in another two reports, which were then used as anodes for the electrochemical oxidation of two dye pollutants (MB and MO), respectively. In the latter case, the deceased removal efficiency upon repeated use of the membrane was attributed to the reduction of active sites caused by membrane fouling [56].

In the above reports, the modified catalytic membrane is generally employed as the anode, and an additional cathode is required. Interestingly, a novel CNT-based membrane with a sandwich-like structure was designed and fabricated, which consisted of an outer CNT layer, a middle PVDF layer, and an inner CNT layer [57]. The composite membrane served as a complete electrochemical system where the two CNT layers acted as two electrodes and the PVDF layer as an insulating separator. With this system, microcystin-LR (MC-LR) can be stably removed through adsorption and electrochemical oxidation cycles [57].

In a recent work, an electrocatalytic ceramic membrane filtration system was developed for the degradation of 2,4-dichlorophenoxyacetic acid (2,4-D). H₂O₂ played an indispensable role in the reaction system, which was generated from continuously supplied oxygen [58]. This was different from the photo-Fenton process discussed above, where H₂O₂ was added rather than generated in situ. Reusability tests showed that the reaction system could remove 2,4-dichlorophenoxyacetic acid stably in 8 cycles, demonstrating the excellent stability of the membrane [58].

A MnFe₂O₄-rGO-PVDF composite membrane was fabricated by vacuum filtration and then employed as the cathode in the electron-activated persulfate activation system. The removal efficiency of oxytetracycline (OTC) gradually decreased after multiple use, which was ascribed to inevitable material loss caused by mechanical wear during assembly and disassembly of the filtration apparatus [59].

As mentioned above, an anode and a cathode are generally required to construct an electrochemical reaction system. Interestingly, in a recent work, a Janus electrified ceramic membrane was fabricated by sputtering palladium particles on both sides of a ceramic substrate membrane, which could then act as both anode and cathode (Figure 4) [60]. The Janus membrane showed well retained efficiency for electrochemical activation of PMS towards the oxidative degradation of MB with a η value of 95% after 8 repeated runs [60].

The introduction of both photo and electron energy may further boost the efficiency of the persulfate oxidation process. For instance, a g-C₃N₄/CNT/ceramic membrane was used in the photo-electrocatalysis process, showing a phenol removal efficiency 2.7 and 2.0 times higher than that with photo irradiation and electron energy supply only, respectively [62]. In another work, TiO₂-ZnO deposited on stainless steel support was also utilized in the photo-electrocatalysis process. In the presence of PMS, the reaction system could remove atrazine and MB almost completely [61]. After 5 runs, a loss of 15% and 10% in removal efficiency was observed for MB and atrazine, which was mainly due to the adsorption of oxidation intermediates accumulated over time considering the stable morphology and crystal structure of the catalytic membrane [61].

3.2. Performance Stability of Catalytic Membranes Tested by Long-Time Operation

In this section, the performance stability of catalytic membranes evaluated by long-time operation was summarized and analyzed (Table 2), where the related reports were also classified based on the type of AOP and discussed in sub-sections. The factors listed in Table 2 are similar to those in Table 1. However, it should be noted that when the long-time operation method is used, the operation mode must be single pass while the recirculation mode is not applicable. Thus, the column of operation mode is omitted in Table 2.

For comparison of the performance stability results, another dimensionless factor (η′) was determined as follows:

η′ = R_t/R₀ × 100%,

(2)

where t stands for operation time of the long-time stability test. R_t represents the removal efficiency of the target pollutant obtained at the end of the test, while R₀ stands for the value obtained during the early time of the test. Note that R₀ was defined in this way rather than as the removal efficiency of the first permeate sample. This was because in some reports, the membrane was saturated in advance by the pollutant solution. In this case, the removal efficiency was limited due to the adsorbed pollutants when the first sample was withdrawn, and it took some time for the removal efficiency to increase and reach a stable value. Similar to η defined in Section 3.1, η′ should also be in the range of 0–100%, and a larger η′ indicates better performance stability within the same operation time.

As summarized in Table 2, for the performance stability tests of these catalytic membranes tested by long-time operation, the employed operation time is in the range of 2.5–240 h, achieving η′ values in the range of 70–100%. A longer operation time and a higher η′ value should indicate better performance stability. Thus, among all the reported cases summarized in Table 2, the Co-TiO_x/mixed cellulose ester membrane and the Co-CuO_x/PVDF membrane stood out, both of which could remove their target pollutants completely during a long operation time of 100 h and 120 h, respectively. Another two membranes were tested for prolonged periods of time. The Fe-doped graphitic carbon nitride/cellulose acetate membrane was tested for PMS activation towards bisphenol A (BPA) removal for 170 h, and the Co₃O₄/Ti membrane was tested for electrochemical removal of COD for an even longer time of 240 h. In both cases, the η′ values at the end of the long-time operations surpassed 90%, demonstrating the outstanding performance stabilities of these two catalytic membranes.

In the following sub-sections, the summarized results in Table 2 are further discussed based on the type of advanced oxidation process.

Table 2. Performance stability of catalytic membranes tested by long-time operation.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Target Pollutant	C0	Flux (L/(m2·h))	R0	Operation Time (h)	η′	Ref.
PMS	PVDF	β-FeOOH	impregnation and in situ mineralization	flat-sheet	cross flow	RhB	10 mg/L	4097.9	99%	12	99%	[63]
PMS	PVDF	MoS2	filtration	flat-sheet	dead-end	BPA	2 mg/L	30.8	100%	6	90%	[64]
PMS	PVDF	Co-CuOx	filtration	flat-sheet	dead-end	RA	5 mg/L	357	100%	120	~100%	[65]
PMS	PVDF	ZIF-67-derived Co-carbon	flash freezing	flat-sheet	cross-flow	TC	10 mg/L	400	97%	84	95%	[66]
PMS	PVDF	CoAl-LDH	filtration	flat	dead-end	RA	2.5 mg/L	80.3	91%	29	~100%	[11]
PMS	PES	Co3O4/C@SiO2	phase inversion	flat-sheet	dead-end	BPA	10 mg/L	229	95%	42	90%	[67]
PMS	cellulose acetate	Fe-doped graphitic carbon nitride	filtration	flat-sheet	dead-end	BPA	20 mg/L	28.66	100%	170	90%	[68]
PMS	cellulose acetate	Co3O4	filtration	flat-sheet	dead-end	RA	5 mg/L	158.4	100%	13	90%	[69]
PMS	mixed cellulose ester	Co-TiOx	filtration	flat-sheet	dead-end	RA	5 mg/L	130	100%	100	~100%	[70]
PMS	nylon	Co@N CNT	filtration	flat-sheet	dead-end	TC	10 mg/L	40	100%	24	90%	[71]
PMS	nylon	Fe2O3@CNT	filtration	flat-sheet	dead-end	TC	0.04 mmol/L	16.3	96.1%	48	89%	[22]
PDS	nylon	rGO/NCNT	filtration	flat-sheet	dead-end	NSMX	0.5 mg/L	46.15	90%	50	79%	[72]
PDS	nylon	NG/rGO/CNT	filtration	flat-sheet	dead-end	SMX	0.5 mg/L	46.15	99.8%	24	99.8%	[23]
PMS	ceramic membrane	nitrogen doped carbon	dip-coating and pyrolysis	flat-sheet	dead-end	phenol	0.1 mmol/L	689.76	100%	48	100%	[73]
PMS	ceramic membrane	Mn2O3	solid state sintering and calcination	flat-sheet	dead-end	BPA EE2 NP	0.1 mg/L 0.1 mg/L 0.1 mg/L	60	92.2% 97.6% 97.3%	8	96% 94% 95%	[74]
PMS	ceramic membrane	Mn2O3	spraying and calcinating	flat-sheet	dead-end	acetaminophen	1 mg/L	60	95%	24	92%	[28]
PMS	carbon nanofibrous membrane	CoOx	impregnation and pyrolysis	flat-sheet	dead-end	TC	30 mg/L	30	98%	10	90%	[75]
H2O2	PVDF	Prussian blue	in situ growth	flat-sheet	cross-flow	MB	30 mg/L	300	99%	24	100%	[76]
H2O2	PVDF	Fe3O4	phase inversion	hollow fiber	cross-flow	MB	100 mg/L	175.8	97%	12	100%	[77]
H2O2	PVDF	BiOI	filtration	flat-sheet	dead-end	paracetamol	10 mg/L	120	96%	100	88%	[78]
H2O2	PP	Prussian blue/GO	filtration	flat-sheet	dead-end	MB	10 mg/L	27	98%	24	100%	[79]
H2O2	ceramic membrane	MoSx	hydrothermal	tubular	dead-end	clothianidin	10 mg/L	25	99.9%	60	70%	[80]
H2O2	ceramic membrane	FeOCl	impregnation and calcination	flat-sheet	dead-end	pCBA	50 μmol/L	100	100%	120	80%	[81]
H2O2	ceramic membrane	Co3O4/MCM-41	packing	tubular	dead-end	AO7	50 mg/L	11.25	96%	40	103%	[82]
Photo	PVDF	ZnIn2S4	filtration	flat-sheet	dead-end	TC	100 μg/L	84.06	90%	36	100%	[83]
Photo	PVDF	Ag2CO3@UiO-66-NH2	filtration	flat-sheet	dead-end	MB	20 mg/L	50	100%	3.3	100%	[84]
Photo	PES	B doped- TiO2-SiO2/CoFe2O4	phase inversion	flat-sheet	cross-flow	COD of biologically treated palm oil mill effluent	1000 mg/L	38.1	100%	12	100%	[85]
Photo	hydrogel membrane	ZnCeOx/graphitic carbon nitride	crosslinking	flat-sheet	cross-flow	MB	20 mg/L	640	95%	4	97%	[86]
Photo	ceramic membrane	TiO2	atomic layer deposition and calcination	flat-sheet	dead-end	MB	1 mg/L	150	40%	33	75%	[87]
Electro	Ti membrane (anode) stainless steel mesh (cathode)	Co3O4 (anode) –(cathode)	hydrothermal and calcination	tubular	dead-end	phenol	5 mmol/L	3.22	99% (COD removal)	240	91%	[88]
Electro	coal-based carbon membrane (anode) titanium plate (cathode)	–(anode) –(cathode)	–	tubular	dead-end	BPA	50 mg/L	43.73	97%	24	90%	[89]
Electro	carbon nanofiber membrane (anode) titanium mesh (cathode)	RuO2/TiO2 (anode) –(cathode)	hydrothermal and calcination	flat-sheet	dead-end	BPA SMX	500 μg/L 500 μg/L	360	98% 98%	72	97% 97%	[90]
Electro	ceramic membrane/Ti mesh (anode) Ti mesh (cathode)	TiO2-SnO2-Sb (anode) –(cathode)	adhesive assemble	flat-sheet	dead-end	PCA	1.28 mg/L	17.4	85.5%	72	97%	[91]
Electro	ceramic membrane/Ti mesh (anode) titanium mesh (cathode)	TiO2/SnO2-Sb (anode) –(cathode)	adhesive assemble	flat-sheet	dead-end	2,4-D	1 mg/L	278	62.4%	84	100%	[58]
Electro	Pt foil (anode) PTFE (cathode)	–(anode) GO-COOFe2+ (cathode)	filtration	flat-sheet	dead-end	florfenicol	1 mg/L	21.5	95%	72	95%	[92]
Electro + O3	titanium mesh (anode) PTFE (cathode)	–(anode) CNT (cathode)	filtration	flat-sheet	dead-end	ibuprofen	2 mg/L	140	95.19%	10	~100%	[93]
Electro + PMS	perforated titanium plate (anode) PTFE (cathode)	–(anode) zero valence copper-CNT (cathode)	filtration	flat-sheet	dead-end	CR	10.4 mg/L	34.6	100%	2.5	100%	[94]

Abbreviations: AO7: Acid orange 7; BPA: Bisphenol A; BPS: Bisphenol S; CNT: Carbon nanotube; COD: Chemical oxygen demand; CR: Congo red; EE2: 17α-ethynylestradiol; LDH: layered double hydroxide; MB: Methylene blue; NP: Nonylphenol; NSMX: N4-acetyl-SMX; PCA: p-chloroaniline; pCBA: Para-chlorobenzoic acid; PDS: Peroxydisulfate; PES: Polyethersulfone; PMS: Peroxymonosulfate; PP: Polypropylene; PTFE: Polytetrafluoroethylene; PVDF: Polyvinylidene fluoride; RA: Ranitidine; rGO: reduced graphene oxide; RhB: Rhodamine B; SMX: Sulfamethoxazole; TC: Tetracycline; ZIF-67: Zeolitic imidazole framework-67; 2,4-D: 2,4-Dichlorophenoxyacetic acid.

Table 2. Performance stability of catalytic membranes tested by long-time operation.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Target Pollutant	C0	Flux (L/(m2·h))	R0	Operation Time (h)	η′	Ref.
PMS	PVDF	β-FeOOH	impregnation and in situ mineralization	flat-sheet	cross flow	RhB	10 mg/L	4097.9	99%	12	99%	[63]
PMS	PVDF	MoS2	filtration	flat-sheet	dead-end	BPA	2 mg/L	30.8	100%	6	90%	[64]
PMS	PVDF	Co-CuOx	filtration	flat-sheet	dead-end	RA	5 mg/L	357	100%	120	~100%	[65]
PMS	PVDF	ZIF-67-derived Co-carbon	flash freezing	flat-sheet	cross-flow	TC	10 mg/L	400	97%	84	95%	[66]
PMS	PVDF	CoAl-LDH	filtration	flat	dead-end	RA	2.5 mg/L	80.3	91%	29	~100%	[11]
PMS	PES	Co3O4/C@SiO2	phase inversion	flat-sheet	dead-end	BPA	10 mg/L	229	95%	42	90%	[67]
PMS	cellulose acetate	Fe-doped graphitic carbon nitride	filtration	flat-sheet	dead-end	BPA	20 mg/L	28.66	100%	170	90%	[68]
PMS	cellulose acetate	Co3O4	filtration	flat-sheet	dead-end	RA	5 mg/L	158.4	100%	13	90%	[69]
PMS	mixed cellulose ester	Co-TiOx	filtration	flat-sheet	dead-end	RA	5 mg/L	130	100%	100	~100%	[70]
PMS	nylon	Co@N CNT	filtration	flat-sheet	dead-end	TC	10 mg/L	40	100%	24	90%	[71]
PMS	nylon	Fe2O3@CNT	filtration	flat-sheet	dead-end	TC	0.04 mmol/L	16.3	96.1%	48	89%	[22]
PDS	nylon	rGO/NCNT	filtration	flat-sheet	dead-end	NSMX	0.5 mg/L	46.15	90%	50	79%	[72]
PDS	nylon	NG/rGO/CNT	filtration	flat-sheet	dead-end	SMX	0.5 mg/L	46.15	99.8%	24	99.8%	[23]
PMS	ceramic membrane	nitrogen doped carbon	dip-coating and pyrolysis	flat-sheet	dead-end	phenol	0.1 mmol/L	689.76	100%	48	100%	[73]
PMS	ceramic membrane	Mn2O3	solid state sintering and calcination	flat-sheet	dead-end	BPA EE2 NP	0.1 mg/L 0.1 mg/L 0.1 mg/L	60	92.2% 97.6% 97.3%	8	96% 94% 95%	[74]
PMS	ceramic membrane	Mn2O3	spraying and calcinating	flat-sheet	dead-end	acetaminophen	1 mg/L	60	95%	24	92%	[28]
PMS	carbon nanofibrous membrane	CoOx	impregnation and pyrolysis	flat-sheet	dead-end	TC	30 mg/L	30	98%	10	90%	[75]
H2O2	PVDF	Prussian blue	in situ growth	flat-sheet	cross-flow	MB	30 mg/L	300	99%	24	100%	[76]
H2O2	PVDF	Fe3O4	phase inversion	hollow fiber	cross-flow	MB	100 mg/L	175.8	97%	12	100%	[77]
H2O2	PVDF	BiOI	filtration	flat-sheet	dead-end	paracetamol	10 mg/L	120	96%	100	88%	[78]
H2O2	PP	Prussian blue/GO	filtration	flat-sheet	dead-end	MB	10 mg/L	27	98%	24	100%	[79]
H2O2	ceramic membrane	MoSx	hydrothermal	tubular	dead-end	clothianidin	10 mg/L	25	99.9%	60	70%	[80]
H2O2	ceramic membrane	FeOCl	impregnation and calcination	flat-sheet	dead-end	pCBA	50 μmol/L	100	100%	120	80%	[81]
H2O2	ceramic membrane	Co3O4/MCM-41	packing	tubular	dead-end	AO7	50 mg/L	11.25	96%	40	103%	[82]
Photo	PVDF	ZnIn2S4	filtration	flat-sheet	dead-end	TC	100 μg/L	84.06	90%	36	100%	[83]
Photo	PVDF	Ag2CO3@UiO-66-NH2	filtration	flat-sheet	dead-end	MB	20 mg/L	50	100%	3.3	100%	[84]
Photo	PES	B doped- TiO2-SiO2/CoFe2O4	phase inversion	flat-sheet	cross-flow	COD of biologically treated palm oil mill effluent	1000 mg/L	38.1	100%	12	100%	[85]
Photo	hydrogel membrane	ZnCeOx/graphitic carbon nitride	crosslinking	flat-sheet	cross-flow	MB	20 mg/L	640	95%	4	97%	[86]
Photo	ceramic membrane	TiO2	atomic layer deposition and calcination	flat-sheet	dead-end	MB	1 mg/L	150	40%	33	75%	[87]
Electro	Ti membrane (anode) stainless steel mesh (cathode)	Co3O4 (anode) –(cathode)	hydrothermal and calcination	tubular	dead-end	phenol	5 mmol/L	3.22	99% (COD removal)	240	91%	[88]
Electro	coal-based carbon membrane (anode) titanium plate (cathode)	–(anode) –(cathode)	–	tubular	dead-end	BPA	50 mg/L	43.73	97%	24	90%	[89]
Electro	carbon nanofiber membrane (anode) titanium mesh (cathode)	RuO2/TiO2 (anode) –(cathode)	hydrothermal and calcination	flat-sheet	dead-end	BPA SMX	500 μg/L 500 μg/L	360	98% 98%	72	97% 97%	[90]
Electro	ceramic membrane/Ti mesh (anode) Ti mesh (cathode)	TiO2-SnO2-Sb (anode) –(cathode)	adhesive assemble	flat-sheet	dead-end	PCA	1.28 mg/L	17.4	85.5%	72	97%	[91]
Electro	ceramic membrane/Ti mesh (anode) titanium mesh (cathode)	TiO2/SnO2-Sb (anode) –(cathode)	adhesive assemble	flat-sheet	dead-end	2,4-D	1 mg/L	278	62.4%	84	100%	[58]
Electro	Pt foil (anode) PTFE (cathode)	–(anode) GO-COOFe2+ (cathode)	filtration	flat-sheet	dead-end	florfenicol	1 mg/L	21.5	95%	72	95%	[92]
Electro + O3	titanium mesh (anode) PTFE (cathode)	–(anode) CNT (cathode)	filtration	flat-sheet	dead-end	ibuprofen	2 mg/L	140	95.19%	10	~100%	[93]
Electro + PMS	perforated titanium plate (anode) PTFE (cathode)	–(anode) zero valence copper-CNT (cathode)	filtration	flat-sheet	dead-end	CR	10.4 mg/L	34.6	100%	2.5	100%	[94]

Abbreviations: AO7: Acid orange 7; BPA: Bisphenol A; BPS: Bisphenol S; CNT: Carbon nanotube; COD: Chemical oxygen demand; CR: Congo red; EE2: 17α-ethynylestradiol; LDH: layered double hydroxide; MB: Methylene blue; NP: Nonylphenol; NSMX: N4-acetyl-SMX; PCA: p-chloroaniline; pCBA: Para-chlorobenzoic acid; PDS: Peroxydisulfate; PES: Polyethersulfone; PMS: Peroxymonosulfate; PP: Polypropylene; PTFE: Polytetrafluoroethylene; PVDF: Polyvinylidene fluoride; RA: Ranitidine; rGO: reduced graphene oxide; RhB: Rhodamine B; SMX: Sulfamethoxazole; TC: Tetracycline; ZIF-67: Zeolitic imidazole framework-67; 2,4-D: 2,4-Dichlorophenoxyacetic acid.

3.2.1. Long-Time Operation Performance of Catalytic Membranes Coupled with Persulfate Oxidation

The long-time operation method is commonly used to test the performance stability of catalytic membranes in the persulfate oxidation process. PVDF, the widely employed organic substrate membrane, was loaded with β-FeOOH, and the composite membrane was tested to remove RhB in the presence of PMS, showing high removal efficiency for 12 h. The flux of the membrane under the same transmembrane pressure also remained stable during that time, indicating the high stability of the membrane [63]. Several other catalysts including MoS₂ [64], Co-CuO_x [65], and ZIF-67-derived Co-carbon [66] were also loaded on PVDF, and the resulting catalytic membranes all possessed high performance stability with η′ values of 90–100%. The CoAl-LDH/PVDF membrane mentioned in Section 3.1.1 was also tested in the long-time operation by changing the operation mode from recirculation to single pass, and the removal efficiency of RA remained stable during the 29 h operation. Interestingly, the flux of the membrane was observed to increase slightly in the first 2 h, which was possibly due to swelling of the pores caused by hydration of the membrane [11]. Organic substrate membranes made from other materials including PES [67], cellulose acetate [68,69], mixed cellulose ester [70] and nylon [22,71] were also investigated. After loaded with various Co- or Fe-containing catalysts, the fabricated catalytic membranes generally showed high performance stability with η′ values above 90%, and limited metal leaching is generally detected in these reports.

In a recent work, a metal-free catalyst, i.e., rGO/N-doped CNT (NCNT), was loaded on a nylon substrate membrane, which was then employed for the degradation of N₄-acetyl-SMX (NSMX), the main refractory metabolite of SMX. The membrane exhibited over 90% NSMX removal efficiency within the first 10 h. However, at the end of the 50 h continuous filtration, the removal efficiency decreased to 70.7%, corresponding to a η′ value of 79% [72]. XPS measurements revealed the accumulation of acid oxygen-containing groups, especially phenolic and carboxylic groups, in the used membrane, which hindered PDS activation with electron-withdrawing effects at the reactive sites. In addition, the partial transformation of graphitic and pyridinic N to pyrrolic N as well as the oxidation of structural defects also contributed to the decreased removal efficiency [72]. Another metal-free catalytic membrane (NG/rGO/CNT/nylon), which was mentioned in Section 3.1.1, was fabricated by the same research group [23]. The membrane showed stable SMX removal in the 24 h long-time operation when deionized water or tap water was used as the background. However, when surface water or effluents of a municipal wastewater treatment plant were utilized, declined removal efficiency was observed. This was attributed to membrane fouling as evidenced from the formation of gel-like cake layers after filtration. In addition, the decrease in the degree of carbon defects also contributed to the decreased removal efficiency [23].

Catalytic membranes constructed of inorganic membranes including ceramic membrane and carbon membrane were also investigated. Nitrogen-doped carbon supported on ceramic membrane was investigated for phenol degradation in a long-time period of 48 h. Specifically, real surface water was used as the background, and phenol removal remained 100% during the whole operation time [73]. In another work, Mn₂O₃-containing ceramic membrane was prepared by solid state sintering and calcination [74]. The membrane could efficiently remove three organic pollutants including BPA, 17α-ethynylestradiol (EE2), and nonylphenol (NP), showing similar removal efficiencies (92.2–97.6%) and η′ values (94–96%). Another Mn₂O₃-containing ceramic membrane mentioned in Section 3.1.1 also showed slightly decreased removal efficiency from 95% to 87% (η′ = 92%) in the 24 h long-time operation [28]. Meanwhile, TOC removal during that time was also recorded, which declined from 64% to 58% [28]. CoO_x/carbon nanofibrous membrane as a gravity-driven membrane was constructed in another report. The long-term stability of the membrane for TC removal was investigated using consecutive gravity filtration experiments, showing a constantly high removal efficiency over 90% and stable permeability in 10 h [75].

3.2.2. Long-Time Operation Performance of Catalytic Membranes Coupled with H₂O₂ Oxidation

Two iron-based catalysts, Prussian blue [76] and Fe₃O₄ [77], were loaded on PVDF substrate membrane and then used for the oxidative degradation of MB in the Fenton process. Both membranes showed stable removal efficiencies in the 24 h and 12 h operation time, respectively, with constant [76] or slightly decreased water flux [77]. In another report, the loaded catalyst was changed to BiOI, and the synthesized membrane also showed stable removal efficiency in the 24 h operation period [78]. Nevertheless, when the operation time was extended to 100 h, the removal efficiency declined with a η′ value of 88%. Leaching of Bi and I was measured throughout the whole operation process. Although the leached concentration of I was much higher than that of Bi, the role of I in the catalytic process was found to be negligible [78]. Instead of PVDF, PP membrane as another organic polymer membrane was utilized as the substrate to load Prussian blue/GO [79], which could stably remove MB for 24 h with almost constant flux.

To modify ceramic membranes for the Fenton oxidation process, MoS_x [80] and FeOCl [81] were loaded, and the degradation of clothianidin and para-chlorobenzoic acid (pCBA) was investigated, respectively. During the 24 h operation time in the former case, the removal efficiency could maintain above 90%. When extending the operation time to 60 h, the removal efficiency further declined and approached 70% [80]. In the latter case, the removal efficiency could maintain almost complete for 24 h, but finally reached 80% after another 96 h [81]. Different from the BiOI catalyst mentioned above, for this FeOCl catalyst, the reason for the decreased catalytic activity was attributed to the leaching of Cl rather than Fe. It was noted that the Fenton reaction was initiated by the binding of H₂O₂ to the Fe site which was coordinated with a labile Cl ligand. Since H₂O₂ possessed a higher affinity to the original Fe site bound to Cl compared to the Fe site formed after each reaction cycle, the gradual loss of Cl resulted in gradually deteriorated removal efficiency [81].

In a recent report, to test the applicability of the Co₃O₄/MCM-41 catalyst packed in the lumen of tubular ceramic membrane, humic acid (HA) as a typical natural organic matter was fed together with acid orange 7 (AO7), which was the target organic pollutant in the Fenton oxidation process [82]. During the 40 h operation time, water flux of the integrated membrane gradually declined by 10%, indicating the fouling effect of HA. Interestingly, the removal efficiency of AO7, which was 96% first, was gradually increased to more than 99% at the end of 40 h [82], indicating an unusual η′ higher than 100%. The reason behind this enhanced removal efficiency was not discussed in the original work. It was inferred that HA gradually accumulated on the membrane and led to decreased pore size, which probably benefitted the rejection of AO7 and contributed to the enhanced removal efficiency.

3.2.3. Long-Time Operation Performance of Catalytic Membranes Coupled with Photocatalysis

Using PVDF membrane as the substrate, ZnIn₂S₄ [83] and Ag₂CO₃@UiO-66-NH₂ [84] were respectively loaded by filtration and tested for the photocatalytic degradation of TC and MB. Stable removal efficiencies were achieved in both cases. In another work, B-doped TiO₂-SiO₂/CoFe₂O₄ were loaded in PES substrate membrane by phase inversion [85]. Specifically, different from the single pollutants generally investigated in other reports, the COD of biologically treated palm oil mill effluent was treated by this photocatalysis membrane filtration system. During the operation time of 12 h, complete COD removal could be stably achieved, although a reduction in flux was observed simultaneously [85].

Different from the commonly employed organic polymer substrate membrane, a self-standing hydrogel membrane made from sodium alginate and porous carbon nitride with uniformly distributed ZnCeO_x nanoparticles was prepared in a recent report [86]. During the operation time of 4 h, the integrated hydrogel could stably reject MB both with and without visible light irradiation. However, under visible light conditions, the water flux remained stable after a drop by 21%, while it dropped dramatically by 53% in the dark. This clearly demonstrated the indispensable contribution of photocatalytic degradation during the filtration process [86].

A ceramic substrate-based catalytic membrane was constructed after the atomic layer deposition of TiO₂ in another report [87]. The as-synthesized membrane was tested for the continuous degradation of MB under UV radiation during a time period of 33 h. The removal efficiency of MB maintained stable at around 40% during the first 10 h. After that, it declined slowly and stabilized again at around 30%. The authors noted that the decrease in MB removal efficiency was unexplained and possibly within system error considering that other recorded variables including temperature, pressure, conductivity, and flow rate were all constant throughout the whole operation time [87].

3.2.4. Long-Time Operation Performance of Catalytic Membranes Coupled with Electrocatalysis

Ti membrane as a conductive substrate was modified with Co₃O₄ by hydrothermal treatment and calcination, which was then employed for the removal of phenol containing wastewater [88]. The removal efficiency of COD maintained stable during a long operation time of 240 h, and stable hydrogen production could be simultaneously achieved with the reaction system. When further extending the operation time, a clear decline trend was observed, showing a η′ value of 60% at 2400 h. X-ray diffraction (XRD) and scanning electron microscopy (SEM) tests revealed the detachment and collapse of Co₃O₄ particles on the membrane surface, which was responsible for the deteriorated removal efficiency [88].

Carbon-based membranes as another type of conductive membrane have been investigated as well. Coal as a cheap and abundant precursor was fabricated into a carbon membrane, which was then directly used as the anode for the electrochemical oxidation of BPA, showing a η′ value of 90% after 24 h [89]. In another report, a carbon nanofiber membrane was modified with RuO₂/TiO₂, and the integrated membrane could stably remove BPA and SMX during an operation time of 72 h [90].

With continuous supply of oxygen, the integrated membrane loaded with TiO₂-SnO₂-Sb was reported to stably remove p-chloroaniline (PCA) [91] and 2,4-dichlorophenoxyacetic acid [58]. In another report, PTFE membrane modified with GO-COOFe²⁺ was employed for the electrochemical oxidation of florfenicol, which could stably remove florfenicol during 72 h. The modified membrane served as the cathode, promoting the electro-Fenton process through the production of H₂O₂ from oxygen, while the contribution of ·OH generated on the anode was negligible [92].

Additional oxidants, such as O₃ and persulfates, can be added in the electrocatalysis process to boost the removal efficiency. In a recent work, a CNT modified PTFE membrane served as the cathode, which was responsible for the decomposition of both in situ-generated H₂O₂ and additionally added O₃ [93]. The reaction system could remove ibuprofen and another four organic pollutants including carbamazepine (CBZ), caffeine, triclosan, and acetaminophen with removal efficiencies over 95% and η′ values approaching 100% [93]. Another PTFE membrane modified with zero valence copper and CNT was also employed as the cathode for the electrochemical activation of PMS, maintaining complete CR removal during 2.5 h continuous filtration [94].

Although the performance stability of catalytic membranes is an important issue, information on this is missing for some reported membranes with inspiring catalytic activities. Thus, the practical applications of these membranes may be doubtful. Sometimes the stability was tested by measuring the change in pure water flux. Although pure water flux is an important parameter for membranes, it could not directly reflex the stability of membranes in the catalytic process.

Summarizing the discussions above, it can be found that the performance of catalytic membranes generally deteriorates upon multi-cycle use or long-time operation. In some cases, the removal efficiency of the target pollutant maintains high. However, stable removal efficiencies may not always indicate high performance stability of the membranes.

In the recirculation operation mode, stable removal efficiency may be obtained at the end of the multi-cycle degradation runs, but a decrease in the kinetic constant may take place. In the single pass operation mode, the removal efficiency may also remain stable during the long-time operation, but the increased TMP at constant flow rate, the decreased flow rate at constant TMP, or the increased cell voltage at constant current density may indicate the performance reduction of the membranes, leading to increased operation cost of the filtration process. Without measuring changes in these factors, the stability of catalytic membranes may not be clearly revealed.

In addition, most of the stability tests were conducted in pure water without co-existing ions and natural organic matter. For practical applications, the presence of these species is inevitable, which may lead to membrane fouling and strongly weaken the performance stability of these catalytic membranes.

The removal efficiency of TOC or COD is generally measured for fresh catalytic membranes only. The ability of catalytic membranes for the removal of TOC or COD may also decline upon repeated use, which has been paid little attention.

In summary, from a practical viewpoint, regeneration of the used catalytic membranes is required sooner or later to restore their ability towards efficient and sustained degradation of organic pollutants, which is discussed in the following section.

4. Regeneration Performance of Catalytic Membranes Coupled with Advanced Oxidation Process

As discussed above in Section 3, a series of reasons may lead to decreased activity of the membranes, including adsorption of organic pollutants and its oxidation intermediates, changes in chemical properties of the membrane such as valence states or surface functionalities, as well as leaching of metal species. Faced with these inactivation reasons, various regeneration methods have been developed. To remove the adsorbed organic species, washing the used membrane with water and other solvents is a widely used approach. A heat treatment procedure or an addition oxidation step can also help to remove the adsorbed organic species. To restore the chemical properties of the carbon-based membranes, heat treatment may help to remove the oxygen functionalities on their surface, which were produced during the degradation process. To restore the valence states of a metal-based membrane, electrochemical or chemical reduction of the used membrane may make a contribution. Note that none of these methods can solve the leaching problem, and it is important to carefully regulate the level of metal leaching through optimization of the membrane structure in order to extent the lifespan of metal-based catalytic membranes.

For comparison of the regeneration performance, the dimensionless factor (η″) was determined as follows:

η″ = R_r/R₀ × 100%,

(3)

where R_r represents the removal efficiency of the target pollutant obtained with the regenerated membrane, while R₀ stands for the value obtained with the fresh one.

The recent works concerning the regeneration of catalytic membranes using solvent washing, heat treatment, and other methods are summarized in Table 3, Table 4, and Table 5, respectively. For the solvent washing method, the CoCu-LDH/polyethylene glycol (PEG)/PVDF membrane stood out, which provided stable SMX removal with a η″ value of 100% in 10 regeneration cycles using ethanol as the washing solvent. The two membrane electrodes, carbon black/polyaniline/carbon felt and CNT-doping polypyrrole/PVDF, which were regenerated by washing with water, showed stable performance with η″ values of 95% after 15 and 20 cycles, respectively, indicating their promising regeneration properties. The employed number of cycles for the heat treatment regeneration method (1–5 cycles) was much smaller compared to that of solvent washing (3–20 cycles). Although η″ values approaching 100% were reported for several catalytic membranes, the heat treatment was conducted only once for them, and the trend in more heat treatment cycles remains unknown. For the Mo/ceramic membrane and the CoFe₂O₄/diatomite/stainless steel mesh, 3 and 5 heat treatment regeneration cycles in air were conducted, respectively, but the achieved η″ values deviated from 100%, reaching 94% and 87%. When other regeneration methods were employed, the number of cycles was in the range of 1–8 cycles. The three membranes of Mn₂O₃/ceramic membrane, TiO₂/Poly(sodium-p-styrenesulfonate) (PSS)/PVDF, and ZIF-67/Ni-foam, which were regenerated by advanced oxidation, all showed outstanding η″ values approaching 100% after 5 cycles. The CoFe₂O₄/ceramic membrane was also promising, showing a η″ value of 95% after 8 regeneration cycles by advanced oxidation.

In the following sub-sections, the summarized results in Table 3, Table 4 and Table 5 will be separately discussed based on the type of regeneration method.

4.1. Regeneration Performance of Catalytic Membranes by Washing

Washing is probably the most widely employed method for membrane regeneration, during which the adsorbed organic species are washed away with the solvents, and thus, the active sites are restored. The washing process can be conducted in many approaches listed as follows (Figure 5):

(a)

The used membrane remains in the filtration apparatus, and filtration with the washing solvent is conducted. This method is applicable to most of the catalytic membranes.

(b)

The used membrane remains in the filtration apparatus, and back washing with the washing solvent is conducted. This method is not applicable to the catalytic membranes formed by filtration.

(c)

The used membrane is removed from the filtration apparatus and then immersed in the solvent, and sonication, stirring or another mechanical force is introduced to facilitate the desorption/detachment of organic pollutants on the membrane. This method is not applicable to the catalytic membranes formed by filtration and those with weak mechanical strength.

(d)

The used membrane is removed from the filtration apparatus, and the catalytic particles on the membrane were removed from the substrate membrane. After that, the separated catalytic particles were washed with the solvent with the assistance of sonication or stirring. After washing, the regenerated catalytic particles were reloaded on the substrate membrane. This method is only applicable to the catalytic membranes formed by filtration.

Various washing solvents have been utilized, and water is the most extensively investigated one due to its green and economical nature. The recent works concerning regeneration of catalytic membranes using solvent washing are summarized in Table 3. An additional column showing the employed washing solvent is provided in the table. As shown in Table 3, water has been widely utilized for membrane regeneration in PMS oxidation [95,96,97,98,99,100], H₂O₂ oxidation [101,102,103,104], O₃ oxidation [35,105], photocatalytic oxidation [106], and electrochemical oxidation [107,108,109,110,111] processes. After washing with water, the adsorbed/attached pollutants can be removed, resulting in restored catalytic performance. In these reports, high η″ values in the range of 88–100% were obtained, indicating the efficiency of this regeneration method.

However, in most of these reports, the performance of the membrane cannot be completely restored. On the one hand, although the adsorbed/attached pollutants can be removed by washing with water, other reasons may also lead to decreased catalytic activity of the membrane. For example, when the NH₂-MIL-88B(Fe)/ceramic membrane was investigated, the catalyst loss was found to be 12.5 mg after the first run, and the value was in the range of 2–5 mg during the subsequent runs [107]. This catalyst loss partly resulted in the decreased activity of the membrane, which could not be recovered by washing. As another example, the ceramic tube membrane loaded with Cu-UiO-66 showed high regeneration ability with a η″ value of 90% after 5 regeneration cycles by washing with water, and the slightly decreased activity was attributed to the leaching of Cu [104]. However, when Cu was replaced by Mn, the Mn-UiO-66/ceramic tube membrane showed dramatically deteriorated phenol removal after 5 cycles with a η″ value of only 11%. This was ascribed to the low stability of the Mn-UiO-66 catalyst, whose decomposition resulted in substantial leaching of Mn and sharply declined removal efficiency [104].

On the other hand, some organic pollutants may have possessed low solubility in water and could not be completely removed after washing, which consumed reactive species in subsequent cycles and led to decreased removal efficiency. Consequently, other washing agents were also investigated for membrane regeneration, including organic solvents as well as acidic and alkaline solutions.

Table 3. Regeneration performance of catalytic membranes by washing.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Operation Mode	Target Pollutant	C0	Flux (L/(m2·h))	R0	Washing Solvent	Number of Washing Cycles	η″	Ref.
PMS	ceramic membrane	MnO2	ball-milling and calcination	flat-sheet	cross-flow	single pass	4-hydroxylbenzoic acid	80 mg/L	19.50	100%	water	6	91%	[95]
PMS	ceramic membrane	CoFe2O4	impregnation and calcination	tubular	dead-end	single pass	MB	25.0 mg·L−1	150	98%	water	4	92%	[96]
PMS	ceramic membrane	Fe-doped CoTiO3	impregnation and calcination	flat-sheet	dead-end	single pass	nimesulide	10 mg/L	300	96.3%	water	5	88%	[97]
PMS	ceramic membrane	CoOx granular activated carbon	dipcoating and calcination; filling the channels	flat-sheet	dead-end	single pass	BPA	10 mg/L	35	95%	water	5	95%	[98]
PMS	Ti membrane	Co3O4-Bi2O3	impregnation and calcination	flat-sheet	dead-end	single pass	MB	20 μmol/L	76.8	98.7%	water	4	~100%	[99]
PMS	Ti membrane	ZIF-67	electro-deposition and calcination	flat-sheet	dead-end	single pass	SMX	0.04 mmol/L	56.9	96.3%	water	5	93%	[100]
H2O2	bacterial cellulose membrane	tetra-amino cobalt (II) phthalocyanine	impregnation and in situ reaction	flat-sheet	dead-end	recirculation	red X-3B	100 µmol/L	1146.50	95%	water	5	97%	[101]
H2O2	ceramic membrane	PDA-β-FeOOH	impregnation and in situ reaction	tubular	cross-flow	single pass	MO	2 mg/L	38	98%	water	6	~100%	[102]
H2O2	CNT membrane	Fe0	wet-spinning and calcination	hollow fiber	dead-end	single pass	BPA	10 mg/L	545.9	97.8%	water	5	97%	[103]
H2O2	ceramic membrane	Cu-UiO-66 Mn-UiO-66	solvothermal and impregnation	tubular	dead-end	single pass	phenol	100 mg/L	3.18	100% 90%	water	5	90% 11%	[104]
O3	ceramic membrane	Mn/FeOx	co-precipitation and calcination	flat-sheet	dead-end	single pass	trimethoprim	5 mg/L	10	98.6%	water	5	91%	[105]
O3	ceramic membrane	N-rGO	pneumatic method	tubular	cross-flow	single pass	benzotriazole	0.084 mol/L	660.86	100%	water	1	95%	[35]
Photo	ceramic membrane	TiO2	dip-coating and calcination	tubular	cross-flow	recirculation	OTC	5 mg/L	960	99%	water	3	~100%	[106]
Electro	ceramic membrane	NH2-MIL-88B(Fe)	hydrothermal	flat-sheet	dead-end	recirculation	naproxen	0.060 mmol/L	15.9	97%	water	5	89%	[107]
Electro	coal-based carbon membrane (anode) the titanium plate (cathode)	polyaniline (anode) –(cathode)	electrochemical polymerization deposition	tubular	dead-end	single pass	phenol	50 mg/L	77.42	99%	water	10	97%	[108]
Electro	coal-based carbon membrane (anode) the titanium plate (cathode)	polyaniline (anode) –(cathode)	electrochemical polymerization deposition	tubular	dead-end	single pass	phenol	50 mg/L	–	99.3%	water	6	96%	[109]
Electro	titanium (anode) carbon felt (cathode)	RuO2-IrO2 (anode) carbon black and polyaniline (cathode)	phase inversion	flat-sheet	dead-end	recirculation	TC	50 mg/L	3857.14	92.9%	water	15	95%	[110]
Electro + PMS	titanium plate (anode) PVDF (cathode)	–(anode) CNT doping polypyrrole (cathode)	phase inversion	flat-sheet	cross-flow	single pass	CBZ	2 mg/L	142.5	95%	water	20	95%	[111]
PMS	PTFE	Mn3O4/CNNS	filtration	flat-sheet	dead-end	recirculation	4-chlorophenol	0.1 mM	15.92	90%	ethanol	6	98%	[112]
PMS	PTFE	NiCo@NCNT	filtration	flat-sheet	dead-end	single pass	ibuprofen	20 mg/L	305.73	100%	ethanol	3	95%	[113]
PMS	PVDF	Fe-Co@NCNT	phase inversion	flat-sheet	cross-flow	single pass	BPA	30 mg/L	45.5	100%	ethanol	10	60%	[114]
PMS	PES	CoCu-LDH/PEG/calotropis gigantean fiber	filtration	flat-sheet	dead-end	single pass	SMX	10 mg/L	124.78	93%	ethanol	10	100%	[115]
PMS	nylon	MnOOH	filtration	flat-sheet	dead-end	recirculation	2,4-Dichlorophenol	25 mg/L	238.8	97.9%	ethanol	4	99%	[116]
Electro	coal-based carbon membrane (anode) Ti plate (cathode)	–(anode) –(cathode)	–	tubular	dead-end	single pass	RhB	250 mg/L	88.7	100%	ethanol	6	~100%	[117]
Electro	coal-based carbon membrane (anode) Ti plate (cathode)	CuO (anode) –(cathode)	electrodeposition	tubular	dead-end	single pass	RhB	300 mg/L	66.51	99.9%	ethanol	4	~100%	[118]
PMS	Ag-La0.8Ca0.2Fe0.94O3−δ	–	phase inversion and calcination	hollow fiber	dead-end	recirculation	MB	10 mg/L	2228.57	90%	H2SO4	5	78%	[119]
H2O2	ceramic membrane	FeOCl	impregnation and calcination	flat-sheet	dead-end	single pass	pCBA	50 μmol/L	100	100%	HCl	4	80%	[81]
Photo	PVDF	PDA/BiOCl0.875Br0.125	pneumatic method	flat-sheet	dead-end	recirculation	roxarsone	17.5 mg/L	380.4	100%	NaOH	4	79%	[120]
Photo +H2O2	ceramic membrane	FeOCl	cross-linking	flat-sheet	dead-end	single pass	nitrobenzene	10 μmol/L	69.2	100%	NaOH	5	100%	[121]
Electro	Ti sheet (anode) PTFE (cathode)	– (anode) CNT@MIL-101(Fe) (cathode)	filtration	flat-sheet	dead-end	recirculation	TC	17.8 mg/L	25.35	94.3%	NaOH	4	88%	[122]
Electro + PMS	PTFE	CNT-Fe3O4	filtration and hydrothermal	flat-sheet	dead-emd	recirculation	roxarsone	1.5 mg/L	–	91.8%	NaOH	4	90%	[123]

Abbreviations: BP-3: Benzophenone-3; BPA: Bisphenol A; CBZ: Carbamazepine; CNT: Carbon nanotube; LDH: layered double hydroxide; MB: Methylene blue; MO: Methyl orange; NCNT: N-doped CNT; OTC: Oxytetracycline; pCBA: para-chlorobenzoic acid; PDA: Polydopamine; PES: Polyethersulfone; PMS: Peroxymonosulfate; PTFE: Polytetrafluoroethylene; PVDF: Polyvinylidene fluoride; rGO: reduced graphene oxide; RhB: Rhodamine B; SDZ: Sulfadiazine; SMX: Sulfamethoxazole; TC: Tetracycline; ZIF-67: Zeolitic imidazole framework-67.

Table 3. Regeneration performance of catalytic membranes by washing.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Operation Mode	Target Pollutant	C0	Flux (L/(m2·h))	R0	Washing Solvent	Number of Washing Cycles	η″	Ref.
PMS	ceramic membrane	MnO2	ball-milling and calcination	flat-sheet	cross-flow	single pass	4-hydroxylbenzoic acid	80 mg/L	19.50	100%	water	6	91%	[95]
PMS	ceramic membrane	CoFe2O4	impregnation and calcination	tubular	dead-end	single pass	MB	25.0 mg·L−1	150	98%	water	4	92%	[96]
PMS	ceramic membrane	Fe-doped CoTiO3	impregnation and calcination	flat-sheet	dead-end	single pass	nimesulide	10 mg/L	300	96.3%	water	5	88%	[97]
PMS	ceramic membrane	CoOx granular activated carbon	dipcoating and calcination; filling the channels	flat-sheet	dead-end	single pass	BPA	10 mg/L	35	95%	water	5	95%	[98]
PMS	Ti membrane	Co3O4-Bi2O3	impregnation and calcination	flat-sheet	dead-end	single pass	MB	20 μmol/L	76.8	98.7%	water	4	~100%	[99]
PMS	Ti membrane	ZIF-67	electro-deposition and calcination	flat-sheet	dead-end	single pass	SMX	0.04 mmol/L	56.9	96.3%	water	5	93%	[100]
H2O2	bacterial cellulose membrane	tetra-amino cobalt (II) phthalocyanine	impregnation and in situ reaction	flat-sheet	dead-end	recirculation	red X-3B	100 µmol/L	1146.50	95%	water	5	97%	[101]
H2O2	ceramic membrane	PDA-β-FeOOH	impregnation and in situ reaction	tubular	cross-flow	single pass	MO	2 mg/L	38	98%	water	6	~100%	[102]
H2O2	CNT membrane	Fe0	wet-spinning and calcination	hollow fiber	dead-end	single pass	BPA	10 mg/L	545.9	97.8%	water	5	97%	[103]
H2O2	ceramic membrane	Cu-UiO-66 Mn-UiO-66	solvothermal and impregnation	tubular	dead-end	single pass	phenol	100 mg/L	3.18	100% 90%	water	5	90% 11%	[104]
O3	ceramic membrane	Mn/FeOx	co-precipitation and calcination	flat-sheet	dead-end	single pass	trimethoprim	5 mg/L	10	98.6%	water	5	91%	[105]
O3	ceramic membrane	N-rGO	pneumatic method	tubular	cross-flow	single pass	benzotriazole	0.084 mol/L	660.86	100%	water	1	95%	[35]
Photo	ceramic membrane	TiO2	dip-coating and calcination	tubular	cross-flow	recirculation	OTC	5 mg/L	960	99%	water	3	~100%	[106]
Electro	ceramic membrane	NH2-MIL-88B(Fe)	hydrothermal	flat-sheet	dead-end	recirculation	naproxen	0.060 mmol/L	15.9	97%	water	5	89%	[107]
Electro	coal-based carbon membrane (anode) the titanium plate (cathode)	polyaniline (anode) –(cathode)	electrochemical polymerization deposition	tubular	dead-end	single pass	phenol	50 mg/L	77.42	99%	water	10	97%	[108]
Electro	coal-based carbon membrane (anode) the titanium plate (cathode)	polyaniline (anode) –(cathode)	electrochemical polymerization deposition	tubular	dead-end	single pass	phenol	50 mg/L	–	99.3%	water	6	96%	[109]
Electro	titanium (anode) carbon felt (cathode)	RuO2-IrO2 (anode) carbon black and polyaniline (cathode)	phase inversion	flat-sheet	dead-end	recirculation	TC	50 mg/L	3857.14	92.9%	water	15	95%	[110]
Electro + PMS	titanium plate (anode) PVDF (cathode)	–(anode) CNT doping polypyrrole (cathode)	phase inversion	flat-sheet	cross-flow	single pass	CBZ	2 mg/L	142.5	95%	water	20	95%	[111]
PMS	PTFE	Mn3O4/CNNS	filtration	flat-sheet	dead-end	recirculation	4-chlorophenol	0.1 mM	15.92	90%	ethanol	6	98%	[112]
PMS	PTFE	NiCo@NCNT	filtration	flat-sheet	dead-end	single pass	ibuprofen	20 mg/L	305.73	100%	ethanol	3	95%	[113]
PMS	PVDF	Fe-Co@NCNT	phase inversion	flat-sheet	cross-flow	single pass	BPA	30 mg/L	45.5	100%	ethanol	10	60%	[114]
PMS	PES	CoCu-LDH/PEG/calotropis gigantean fiber	filtration	flat-sheet	dead-end	single pass	SMX	10 mg/L	124.78	93%	ethanol	10	100%	[115]
PMS	nylon	MnOOH	filtration	flat-sheet	dead-end	recirculation	2,4-Dichlorophenol	25 mg/L	238.8	97.9%	ethanol	4	99%	[116]
Electro	coal-based carbon membrane (anode) Ti plate (cathode)	–(anode) –(cathode)	–	tubular	dead-end	single pass	RhB	250 mg/L	88.7	100%	ethanol	6	~100%	[117]
Electro	coal-based carbon membrane (anode) Ti plate (cathode)	CuO (anode) –(cathode)	electrodeposition	tubular	dead-end	single pass	RhB	300 mg/L	66.51	99.9%	ethanol	4	~100%	[118]
PMS	Ag-La0.8Ca0.2Fe0.94O3−δ	–	phase inversion and calcination	hollow fiber	dead-end	recirculation	MB	10 mg/L	2228.57	90%	H2SO4	5	78%	[119]
H2O2	ceramic membrane	FeOCl	impregnation and calcination	flat-sheet	dead-end	single pass	pCBA	50 μmol/L	100	100%	HCl	4	80%	[81]
Photo	PVDF	PDA/BiOCl0.875Br0.125	pneumatic method	flat-sheet	dead-end	recirculation	roxarsone	17.5 mg/L	380.4	100%	NaOH	4	79%	[120]
Photo +H2O2	ceramic membrane	FeOCl	cross-linking	flat-sheet	dead-end	single pass	nitrobenzene	10 μmol/L	69.2	100%	NaOH	5	100%	[121]
Electro	Ti sheet (anode) PTFE (cathode)	– (anode) CNT@MIL-101(Fe) (cathode)	filtration	flat-sheet	dead-end	recirculation	TC	17.8 mg/L	25.35	94.3%	NaOH	4	88%	[122]
Electro + PMS	PTFE	CNT-Fe3O4	filtration and hydrothermal	flat-sheet	dead-emd	recirculation	roxarsone	1.5 mg/L	–	91.8%	NaOH	4	90%	[123]

Abbreviations: BP-3: Benzophenone-3; BPA: Bisphenol A; CBZ: Carbamazepine; CNT: Carbon nanotube; LDH: layered double hydroxide; MB: Methylene blue; MO: Methyl orange; NCNT: N-doped CNT; OTC: Oxytetracycline; pCBA: para-chlorobenzoic acid; PDA: Polydopamine; PES: Polyethersulfone; PMS: Peroxymonosulfate; PTFE: Polytetrafluoroethylene; PVDF: Polyvinylidene fluoride; rGO: reduced graphene oxide; RhB: Rhodamine B; SDZ: Sulfadiazine; SMX: Sulfamethoxazole; TC: Tetracycline; ZIF-67: Zeolitic imidazole framework-67.

Ethanol as a cheap and green organic solvent has been widely investigated [112,113,114,115,116,117,118]. In some cases, the washing procedure was conducted under solvothermal conditions. For example, for the regeneration of Mn₃O₄ nanodots-g-C₃N₄ nanosheets (Mn₃O₄/CNNS), the used membrane was washed with ethanol and water and then used for the next run, and the η″ approached 89% after 5 cycles. After that, the used membrane was further treated with ethanol at 150 °C for 3 h, obtaining an increased η″ value of 98% [112]. This may be ascribed to the enhanced solubility of organic pollutants in ethanol under the increased treatment temperature.

Acid and alkaline solutions have long been used for the chemical cleaning of fouled membranes. The change in pH may alter the charge status of both the membrane surface and the adsorbed pollutants, enhancing desorption of the adsorbed pollutants and, thus, the regeneration efficiency. Various acids including H₂SO₄, HCl, and H₃PO₄ have been employed. H₂SO₄ was used for the regeneration of perovskite oxide membrane, achieving a η″ value of 78% after 5 cycles. The declined catalytic activity was mainly ascribed to leaching of iron, which changed the surface composition of perovskite and caused a loss in activity over time [119]. For the regeneration of FeOCl/ceramic membrane, HCl was utilized [81]. As mentioned above in Section 3.1.1, the decreased activity of the membrane was partly ascribed to the leaching of Cl. It was reported that the activity of the used membrane could be effectively restored by HCl treatment and subsequent heat treatment at 220 °C. HCl treatment helped to remove the pollutants on the used membrane. More importantly, the regeneration procedure led to Cl intercalation and recovery of the Fe-Cl bonds. The heat treatment step was crucial to restore the Fe-Cl bonds, as evidenced by the low regeneration efficiency with HCl treatment only [81]. For alkaline treatment, NaOH is commonly employed. For example, NaOH was used to regenerate the PDA/BiOCl_0.875Br_0.125/PVDF membrane after the treatment of roxarsone, which was an arsenic-containing organic pollutant [120]. During the photocatalytic process, the released inorganic arsenic could be converted to As (V) and then immobilized on the membrane surface. Upon NaOH treatment, OH⁻ could replace the adsorbed As (V) anions, while Na⁺ helped to weaken the electrostatic interaction between PDA and As (V) anions [120]. NaOH was also used to regenerate other catalytic membranes in several photocatalytic [121] and electrochemical catalytic processes [122,123]. To further improve the regeneration performance, acid and alkaline treatment can be used in combination, such as in the case of α-FeOOH/ceramic membrane where NaOH and H₃PO₄ washing was performed successively [124].

It should be noted that when organic solvents or acid/alkaline solutions were used for membrane cleaning, a final washing step with water is generally required to remove the residue detergent, avoiding the possible interference in subsequent operations. In addition, although the utilization of these agents could lead to improved regeneration performance, the increased cost and potential environmental pollution should also be taken into consideration.

Solvent washing can help to remove the adsorbed pollutants to some extent. However, even after solvent washing with different solvents, some of the pollutants may still remain on the membranes. Another method to remove the adsorbed pollutants in the heat treatment approach, which is discussed in the following section.

4.2. Regeneration Performance of Catalytic Membranes by Heat Treatment

During heat treatment, the adsorbed species volatilize or decompose into gaseous products under the applied high temperature and leave the occupied active sites on the membrane. The recent works concerning regeneration of catalytic membranes using heat treatment are summarized in Table 4. The employed treatment temperature and atmosphere were provided in two additional columns in the table. As shown in Table 4, different types of atmospheres, including nitrogen [125], argon [113], and air [126,127,128,129,130,131], have been applied for the thermal regeneration of catalytic membranes.

Table 4. Regeneration performance of catalytic membranes by heat treatment.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Operation Mode	Target Pollutant	C0	Flux (L/(m2·h))	R0	Treatment Temperature/°C	Atmosphere	Number of Heat Treatment Cycles	η″	Ref.
PMS	PTFE	NiCo@NCNT	filtration	flat-sheet	dead-end	single pass	ibuprofen	20 mg/L	305.73	100%	350	Ar	1	100%	[113]
PDS	CNT	–	wet-spinning and pyrolyzation	hollow fiber	dead-end	single pass	phenol	10 mg/L	–	97%	800	N2	1	100%	[125]
PMS	ceramic membrane	CNT@nitrogen-doped carbon	filtration	flat-sheet	dead-end	single pass	SMX	20 mg/L	6.8	65%	250	air	1	100%	[126]
PMS	nylon	nitrogen-doped rGO/CNT	filtration	flat-sheet	dead-end	single pass	4-chlorophenol	0.16 mM	72.55	100%	90	air	1	94%	[127]
PMS	PVDF	CoxFe3−xO4	filtration	flat-sheet	dead-end	single pass	BPA	10 mg/L	100	99.9%	100	air	1	~100%	[128]
PMS	steel mesh/non-woven fabric	LaFexCo1−xO3−λ/SiO2	filtartion	flat-sheet	dead-end	recirculation	TC	20 mg/L	1064.5	99%	450	air	1	~100%	[129]
PMS	stainless steel mesh	CoFe2O4/diatomite	filtration	flat-sheet	dead-end	recirculation	RhB	50 mg/L	98.68	99.9%	200	air	5	87%	[130]
Air	ceramic membrane	Mo	hydrothermal and calcination	flat-sheet	dead-end	single pass	safranine O	10 mg/L	324	96%	300	air	3	94%	[131]

Abbreviations: BPA: Bisphenol A; CNT: Carbon nanotube; NCNT: N-doped CNT; PDS: Peroxydisulfate; PMS: Peroxymonosulfate; PTFE: Polytetrafluoroethylene; PVDF: Polyvinylidene fluoride; rGO: reduced graphene oxide; RhB: Rhodamine B; SMX: Sulfamethoxazole; TC: Tetracycline.

Table 4. Regeneration performance of catalytic membranes by heat treatment.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Operation Mode	Target Pollutant	C0	Flux (L/(m2·h))	R0	Treatment Temperature/°C	Atmosphere	Number of Heat Treatment Cycles	η″	Ref.
PMS	PTFE	NiCo@NCNT	filtration	flat-sheet	dead-end	single pass	ibuprofen	20 mg/L	305.73	100%	350	Ar	1	100%	[113]
PDS	CNT	–	wet-spinning and pyrolyzation	hollow fiber	dead-end	single pass	phenol	10 mg/L	–	97%	800	N2	1	100%	[125]
PMS	ceramic membrane	CNT@nitrogen-doped carbon	filtration	flat-sheet	dead-end	single pass	SMX	20 mg/L	6.8	65%	250	air	1	100%	[126]
PMS	nylon	nitrogen-doped rGO/CNT	filtration	flat-sheet	dead-end	single pass	4-chlorophenol	0.16 mM	72.55	100%	90	air	1	94%	[127]
PMS	PVDF	CoxFe3−xO4	filtration	flat-sheet	dead-end	single pass	BPA	10 mg/L	100	99.9%	100	air	1	~100%	[128]
PMS	steel mesh/non-woven fabric	LaFexCo1−xO3−λ/SiO2	filtartion	flat-sheet	dead-end	recirculation	TC	20 mg/L	1064.5	99%	450	air	1	~100%	[129]
PMS	stainless steel mesh	CoFe2O4/diatomite	filtration	flat-sheet	dead-end	recirculation	RhB	50 mg/L	98.68	99.9%	200	air	5	87%	[130]
Air	ceramic membrane	Mo	hydrothermal and calcination	flat-sheet	dead-end	single pass	safranine O	10 mg/L	324	96%	300	air	3	94%	[131]

Abbreviations: BPA: Bisphenol A; CNT: Carbon nanotube; NCNT: N-doped CNT; PDS: Peroxydisulfate; PMS: Peroxymonosulfate; PTFE: Polytetrafluoroethylene; PVDF: Polyvinylidene fluoride; rGO: reduced graphene oxide; RhB: Rhodamine B; SMX: Sulfamethoxazole; TC: Tetracycline.

When conducting thermal regeneration, the employed atmosphere and heating temperature should be carefully designed considering the properties of the catalytic membrane. For carbon-based membranes, the regeneration process should be generally conducted under inert atmosphere to avoid the oxidation of the membrane. For example, the regeneration of NiCo@NCNT and CNT membranes was conducted under argon [113] and nitrogen [125] atmospheres, respectively. The air atmosphere can be applied as well, but the employed temperature should be much lower. For example, the regeneration of CNT@nitrogen-doped carbon and nitrogen-doped rGO/CNT membranes was performed at 250 °C [126] and 90 °C [127], respectively, lower than the temperatures of 350 °C and 800 °C applied in the former two cases. To avoid the oxidation of the substrate membrane, the loaded catalyst could be separated first and then treated by heat treatment. After heat treatment, the regenerated catalytic particles could be reloaded on the substrate membrane. For example, the LaFe_xCo_1−xO_3−λ/SiO₂ catalyst could withstand the high temperature treatment, but the substrate membrane of steel mesh/non-woven fabric could not. Therefore, the functional layer formed by the catalyst was separated and calcined in air at 450 °C, and then, the catalytic membrane was reformed by filtration, achieving a high η″ value close to 100% [129]. In another case where CoFe₂O₄/diatomite was supported on stainless steel mesh, a similar separation, calcination, and reforming procedure was applied [130].

Although the thermal regeneration method may be highly efficient, the energy cost may be a potential concern. In addition, a higher treatment temperature may lead to better regeneration performance, but the cost is also considerably increased. Consequently, rather than conducting heat treatment after each catalytic run, it can be performed periodically in the whole lifespan of the membrane to make a compromise between regeneration cost and removal efficiency. For example, when the CNT membrane was employed in combination with PDS for phenol degradation, it was used for five times without regeneration, after which thermal regeneration was conducted [125]. Similarly, the CNT@nitrogen-doped carbon/ceramic membrane was used for three times without regeneration followed by a calcination step in air [126]. The thermal regeneration method can also be used in combination with the solvent washing approach. For instance, the nitrogen-doped rGO/CNT membrane was regenerated by heating at 90 °C and subsequent washing with NaOH [127]. Among the summarized reports, the heating temperature in this work was the lowest. This may be attributed to the assistance of NaOH washing, which shared some responsibility. In another work [113], for the regeneration of the NiCo@NCNT membrane, the thermal method and solvent washing method were also used in combination but operated in a different way. After each of the first three runs, solvent washing with ethanol was conducted, while heat treatment was performed after the fourth run [113]. Twelve runs were performed in total, with nine times of solvent washing and three times of thermal regeneration, achieving an acceptable η″ value of 78% in the end.

Besides solvent washing and heat treatment, other methods have also been employed for membrane regeneration, which are discussed in the following section.

4.3. Regeneration Performance of Catalytic Membranes by Other Methods

The recent works concerning regeneration of catalytic membranes by other methods are summarized in Table 5, where the employed regeneration method is shown in a separated column.

Table 5. Regeneration performance of catalytic membranes by other methods.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Operation Mode	Target Pollutant	C0	Flux (L/(m2·h))	R0	Regeneration Method	Number of Regeneration Cycles	η″	Ref.
PMS	ceramic membrane	CoFe2O4	hydrothermal	flat-sheet	dead-end	single pass	ofloxacin	40 μmol/L	98	100%	PMS	8	95%	[132]
PMS	ceramic membrane	CoFe2O4	impregnation and calcination	flat-sheet	dead-end	single pass	SMX	10 mg/L	236	50%	PMS	3	95%	[133]
PMS	ceramic membrane	Co3O4	impregnation and calcination	flat-sheet	dead-end	single pass	SMX	0.08 mmol/L	236	46%	PMS	3	54%	[134]
PMS	ceramic membrane	Co3O4	in situ reaction and calcination	flat-sheet	dead-end	recirculation	SMX	10 mg/L	211.68	75%	PMS	1	95%	[135]
O3	ceramic membrane	Mn2O3	impregnation and calcination	flat-sheet	dead-end	single pass	SMX	20 mg/L	35	81.3%	O3	5	~100%	[136]
Photo	PVDF	TiO2/PSS	layer-by-layer deposition	flat-sheet	cross-flow	recirculation	Lanasol Blue 3R	30 mg/L	300	91.42%	photo	5	~100%	[137]
PMS	Ni-foam	ZIF-67	electrodeposion and calcination	flat-sheet	dead-end	recirculation	BPA	0.044 mmol/L	12.68	100%	photo	5	~100%	[138]
Electro	porous Ti plate (anode) porous Ti plate (cathode)	blue TiO2 (anode) blue TiO2 (cathode)	electrooxidation, calcination and electroreduction	flat-sheet	dead-end	recirculation	CBZ	10 mg/L	11464.97	95.3%	electrochemical reduction	4	99%	[139]
Electro	Ti membrane (anode) Ti mesh (cathode)	blue TiO2 (anode) –(cathode)	electrooxidation, calcination and electroreduction	flat-sheet	dead-end	single pass	MB	10 mg/L	–	99.5%	electrochemical reduction	2	~100%	[140]
Electron	ceramic membrane (anode) stainless steel mesh (cathode)	CNT (anode) –(cathode)	filtration and pyrolyzation	tubular	cross-flow	single pass	TOC of water from Lingshui Reservoir	3.281 mg/L	347.84	90%	electrochemical back washing	4	95%	[141]
H2O2	PVDF	Fe/PDA/PEI	deposition, filtration and reduction	flat-sheet	cross-flow	single pass	RhB	10 mg/L	400	95%	chemical reduciton with NaBH4	1	~100%	[142]

Abbreviations: BPA: Bisphenol A; CBZ: Carbamazepine; CNT: Carbon nanotube; MB: Methylene blue; PDA: Polydopamine PMS: Peroxymonosulfate; PSS: polystyrene; PVDF: Polyvinylidene fluoride; RhB: Rhodamine B; SMX: Sulfamethoxazole; TOC: Total organic carbon; ZIF-67: Zeolitic imidazole framework-67.

Table 5. Regeneration performance of catalytic membranes by other methods.

AOP Type	Substrate Membrane	Catalyst	Fabrication Method	Membrane Type	Filtration Mode	Operation Mode	Target Pollutant	C0	Flux (L/(m2·h))	R0	Regeneration Method	Number of Regeneration Cycles	η″	Ref.
PMS	ceramic membrane	CoFe2O4	hydrothermal	flat-sheet	dead-end	single pass	ofloxacin	40 μmol/L	98	100%	PMS	8	95%	[132]
PMS	ceramic membrane	CoFe2O4	impregnation and calcination	flat-sheet	dead-end	single pass	SMX	10 mg/L	236	50%	PMS	3	95%	[133]
PMS	ceramic membrane	Co3O4	impregnation and calcination	flat-sheet	dead-end	single pass	SMX	0.08 mmol/L	236	46%	PMS	3	54%	[134]
PMS	ceramic membrane	Co3O4	in situ reaction and calcination	flat-sheet	dead-end	recirculation	SMX	10 mg/L	211.68	75%	PMS	1	95%	[135]
O3	ceramic membrane	Mn2O3	impregnation and calcination	flat-sheet	dead-end	single pass	SMX	20 mg/L	35	81.3%	O3	5	~100%	[136]
Photo	PVDF	TiO2/PSS	layer-by-layer deposition	flat-sheet	cross-flow	recirculation	Lanasol Blue 3R	30 mg/L	300	91.42%	photo	5	~100%	[137]
PMS	Ni-foam	ZIF-67	electrodeposion and calcination	flat-sheet	dead-end	recirculation	BPA	0.044 mmol/L	12.68	100%	photo	5	~100%	[138]
Electro	porous Ti plate (anode) porous Ti plate (cathode)	blue TiO2 (anode) blue TiO2 (cathode)	electrooxidation, calcination and electroreduction	flat-sheet	dead-end	recirculation	CBZ	10 mg/L	11464.97	95.3%	electrochemical reduction	4	99%	[139]
Electro	Ti membrane (anode) Ti mesh (cathode)	blue TiO2 (anode) –(cathode)	electrooxidation, calcination and electroreduction	flat-sheet	dead-end	single pass	MB	10 mg/L	–	99.5%	electrochemical reduction	2	~100%	[140]
Electron	ceramic membrane (anode) stainless steel mesh (cathode)	CNT (anode) –(cathode)	filtration and pyrolyzation	tubular	cross-flow	single pass	TOC of water from Lingshui Reservoir	3.281 mg/L	347.84	90%	electrochemical back washing	4	95%	[141]
H2O2	PVDF	Fe/PDA/PEI	deposition, filtration and reduction	flat-sheet	cross-flow	single pass	RhB	10 mg/L	400	95%	chemical reduciton with NaBH4	1	~100%	[142]

Abbreviations: BPA: Bisphenol A; CBZ: Carbamazepine; CNT: Carbon nanotube; MB: Methylene blue; PDA: Polydopamine PMS: Peroxymonosulfate; PSS: polystyrene; PVDF: Polyvinylidene fluoride; RhB: Rhodamine B; SMX: Sulfamethoxazole; TOC: Total organic carbon; ZIF-67: Zeolitic imidazole framework-67.

Since AOP can help to remove the pollutants in the feed water, it can also be used to remove the pollutants on the used catalytic membranes. In the regeneration process, no additional catalyst is required because the membrane itself can act as the catalyst, and the oxidant or external power is provided in the absence of pollutants in feed water. Generally, the same oxidant or external power as that employed for the treatment of polluted water is used for membrane regeneration. This is because the active species in different AOPs may differ from each other, and the used membrane after a specific AOP may not act as an efficient catalyst during regeneration by another AOP. For instance, after being used in the PMS oxidation process, several membranes loaded with Co-based catalytic species were regenerated by PMS oxidation [132,133,134,135,143]. After PMS treatment, the accumulated pollutants on the membranes were further degraded and removed, leading to high η″ values over 90%. The relatively low η″ reported in Ref. [134] was ascribed to the co-existence of humic acid in feed water, which could not be efficiently degraded in the catalytic oxidation process. Similarly, the Mn₂O₃/ceramic membrane was employed in the catalytic ozonation process and regenerated by ozone oxidation [136]. The FeOCl/PVDF membrane was tested for Fenton oxidation and regenerated by Fenton oxidation [144]. TiO₂/PSS/PVDF was utilized in the photocatalytic process and regenerated by photo treatment [137]. Interestingly, after ZIF-67/Ni-foam was tested for PMS activation, its regeneration was performed not by PMS oxidation but by photo irradiation, achieving a high η″ close to 100% after 5 regeneration cycles [138]. This may be ascribed to the specific feature ofZIF-67, which could act as the catalyst not only for PMS activation but also in the photocatalytic process.

Several catalytic membranes, after being used for electrochemical oxidation, were regenerated by electrochemical treatment. However, different from the cases in other AOPs discussed above, the mechanism during the regeneration process is not the same as that in the degradation process. For example, the Ti-based membrane anodes loaded with blue TiO₂ were regenerated by electrochemical reduction, during which the active Ti³⁺ sites were restored, leading to recovered catalytic activities [139,140]. In another two cases, both the ZnO/coal-based carbon membrane [145] and the CNT/ceramic membrane [141] were used as anodes in the electrochemical oxidation process, but their regeneration was performed by electrochemical back washing where they acted as cathodes.

Chemical reduction with reduction agents such as NaBH₄ has also been employed for metal-based membranes. For example, after being used in the Fenton process, the membrane of Fe/PDA/PEI/PVDF was regenerated by NaBH₄ (Figure 6) [142]. Similarly, the Fe_xO_y/PVDF membrane used for PDS oxidation was also regenerated by NaBH₄ treatment [146]. In both cases, chemical reduction with NaBH₄ helped to reduce the high-valent iron species, resulting in restored catalytic activity.

Summarizing the above discussions, the regeneration properties of catalytic membranes by various methods have been investigated in many references, but there are still some problems. As mentioned above, the regeneration method by solvent washing can be conducted in several approaches, considering the fabrication method and the mechanical strength of the catalytic membranes. However, in many previous works, the regeneration method is simply described as “washing” without the detailed procedures. Similarly, detailed information for the thermal regeneration process, such as the temperature and calcination time, is also missing in some references.

In addition, similar to the case when evaluating the performance stability, the removal efficiency of the target pollutant is also the main (in most cases the only) index to evaluate the regeneration property. More attention should be paid to other factors including flux, TMP, and TOC/COD removals. The presence of inorganic ions and natural organic matter in the water matrix may lead to more serious membrane fouling, making the current regeneration methods less effective.

Up until now, when AOP was used for membrane regeneration, the same oxidant or external power as that employed for the treatment of polluted water was generally utilized. Meanwhile, none of the reported regeneration methods can solve the problem of metal leaching.

5. Recommendations for Future Works

Although the performance stability and regeneration of catalytic membranes have been extensively investigated, further research in this field is still required. Several recommendations for future works are listed as follows.

• In future works, it is suggested to always include the performance stability or regeneration property of catalytic membranes when evaluating their abilities for pollutant removal. Catalytic membranes with high activity but poor stability and regeneration property are far away from practical applications.
• When evaluating the stability and regeneration property, it is suggested to clearly describe the process with all the details provided. Besides the removal efficiency of pollutants, other factors including change in kinetic constant, TMP, flux, and TOC/COD removal should also be paid attention to present a more comprehensive profile.
• Although the effects of inorganic ions and natural organic matter for the activity of fresh catalytic membranes are generally reported, their effects on the performance stability and regeneration property have been rarely investigated. In future works, more attention should be paid to this issue because these species may lead to serious membrane fouling, weakening the performance stability and making the current regeneration methods less effective.
• The number of cycles in the multi-cycle run and the operation time in the long-time run summarized in Table 1 and Table 2 have been statistically analyzed. As shown in Figure 7a, for the number of cycles, most of the data were in the range of 5–10 cycles, and the median and average values were 5 and 6.3 cycles (n = 52). The operation time mostly fell in the range of 0–25 h with a median value of 33 h and an average value of 50.2 h (n = 37) (Figure 7b). For future works, it is suggested to select suitable cycle numbers and operation times based on these statistics. When possible, a large cycle number and a longer operation time are suggested, which can provide better references for potential practical applications of the investigated catalytic membranes.
• The types and distributions of organic pollutants on the fouled membrane are different from those in the polluted water, and the utilization of another AOP for its regeneration may be more efficient. In other words, a certain AOP may be suitable to regenerate the fouled membrane, although it is not suitable to treat the polluted water. Through the combination of different AOPs, better process optimization and lower whole process cost may be achieved.
• Faced with the decreased activity of metal-based membranes caused by metal-leaching, re-loading of metal species may be a possible solution, which is more economic than simply discarding the used membrane. However, the organic pollutants on the used membrane may impact the metal loading process, and more research on this issue is still in demand.
• Almost all of the reports summarized and discussed above were conducted at bench-scale. To the best of our knowledge, there is no industrial scale application of this new technology at the current stage. Only several pilot-scale studies are available, where catalytic membranes were coupled with ozone oxidation [147,148,149], electrochemical oxidation [150,151], and electro-Fenton [152] processes. In these studies, to increase the throughput of the catalytic filtration process, two strategies have been utilized. One is to build catalytic membranes with larger sizes, as in the cases of Ti-Mn/ceramic membrane (Figure 8a) [147] and PbO₂/Ti membrane (Figure 8b) [151]. Another is the numbering up strategy where multiple bench-scale membranes are integrated, such as in the cases of Ti-Mn/TiO₂/ceramic membrane (Figure 8c) [148]. In these pilot studies, natural water sources with complex compositions were treated, leading to membrane fouling in the long run and making membrane regeneration inevitable. From a practical viewpoint, the large-sized membranes are difficult to fabricate and difficult to regenerate by some of the above-mentioned methods such as the heat treatment method, during which large-sized furnaces are required to fit the large-sized membranes, increasing the equipment investment. The numbering up strategy is convenient for membrane fabrication, but complex pipe systems would be required, making it difficult to maintain, disassemble, and regenerate the integrated multiple membranes. In future works, more research is required regarding the stability and regeneration property of pilot-scale catalytic membranes.

In summary, the current review further highlights the current achievements and hurdles in the performance stability and regeneration property of catalytic membranes, proposing recommendations for future works, including more comprehensive evaluation of the performance stability and regeneration property based on changes in the kinetic constant, TMP/flux, and TOC/COD removals as well as the effects of inorganic ions and natural organic matter, the selection of suitable cycle numbers and operation time, the utilization of a different AOP for membrane regeneration, the development of new regeneration methods, and more studies on the pilot and larger scales.

6. Conclusions

Catalytic membranes coupling AOP and membrane filtration have been widely employed for the treatment of various organic pollutants in water. In this review, the recent works concerning performance stability and regeneration property of these membranes have been summarized in terms of the different reaction systems including persulfate oxidation, Fenton oxidation, ozone oxidation, photocatalytic oxidation, and electrochemical oxidation. The performance stability of catalytic membranes was commonly evaluated by either multi-cycle experiments or long-time filtration, and the main deactivation reasons included the adsorption of pollutants and its oxidation intermediates, the changes in chemical properties of the membranes (valence state or surface functionalities), as well as leaching of metal species. Faced with these inactivation reasons, various regeneration methods have been developed. Washing with solvents is the most extensively used regeneration method, and different solvents including water, ethanol, and acidic/alkaline solutions have been investigated. Heat treatment is another popular approach, during which the employed atmosphere and heating temperature should be carefully selected considering the properties of the catalytic membrane. Other regeneration methods were also discussed, including AOP, electrochemical reduction, electrochemical backwashing, and chemical reduction. This review further highlights the current achievements and hurdles in the performance stability and regeneration property of catalytic membranes, proposing recommendations for future works.