Breaking Boundaries: Innovations in Two- and Three-Dimensional Metal-Organic Framework (MOF)-Based Mixed-Matrix Membranes for Effective Post-Combustion CO₂ Capture

Abstract

The escalating levels of carbon dioxide (CO₂) emissions have warranted countries worldwide to address the undesirable production of anthropogenic CO₂. It is anticipated that membrane processes can significantly improve overall CO₂ capture performance compared to alternative unit operations (e.g., adsorption, scrubbing) due to the small plant footprint requirement. This comprehensive review focuses on utilising metal-organic frameworks (MOFs) with large porosities and tunable functionalities as filler materials in the fabrication of mixed-matrix membranes (MMMs). The review focuses on different filler configurations, including two-dimensional, three-dimensional, and composites. The separation performance is systematically summarized and compared against the upper bound limit, along with the filler enhancement index (F_index) for performance validation. The review also provides insights into future perspective of MOF-based MMM in gas separation to enhance the practical application and relevance of MMM in addressing global CO₂ emissions.

Keywords:

• Metal-organic framework
• Mixed-matrix membrane
• CO₂ capture
• Gas separation

1. Introduction

Greenhouse gas emissions, particularly carbon dioxide (CO₂), have been a long-standing concern due to the substantial increase in environmental-related events, such as climate change, rising sea levels, and global warming (Chuah et al. 2018). The Paris Climate Agreement and reports from the Intergovernmental Panel on Climate Change (IPCC) underscore the urgency of reducing the global average temperature to 1.5 °C. To achieve this, there is a target to reduce global anthropogenic CO₂ emissions by 45% by 2030, based on 2010 levels (Masson-Delmotte 2022). Nations worldwide are striving to achieve net zero carbon emissions by 2050, aiming to achieve a balance between emitting and absorbing carbon from the atmosphere. While incorporating renewable energies (e.g., tidal, wind, and solar) is crucial, practical challenges, including technical difficulties and limited availability, hinder their widespread adoption to combat the rise in CO₂ emission (Chuah, Li, et al. 2019; Pires et al. 2011). As of now, carbon-based fuels (i.e., fossil fuels, natural gas, and coal) remain the primary sources of energy generation. However, mitigating CO₂ emissions from flue gas requires the incorporation of carbon capture, storage, and sequestration (CCUS) to remove CO₂ before its release into the environment (Khalilpour et al. 2015; Luis et al. 2012).

The success of effective CO₂ removal in the CCUS unit heavily relies on the CO₂ capture step, with approximately 70% of the cost attributed to this phase (Chuah, Kim, et al. 2019). Therefore, appropriate selection of unit operation is critical to ensure its effectiveness in CO₂ removal. Membrane separation stands out as a promising technology for this purpose due to its cost-effectiveness, ease of installation with a modular design, and energy efficiency (Lee et al. 2022; Miltner et al. 2017). While polymeric membranes have been extensively studied, the undesirable trade-off relationship between permeability and selectivity (Robeson upper bound) has limited overall performance (Comesaña-Gándara et al. 2019; Robeson 2008). Hence, in this review, the emphasis is on the use of mixed-matrix membranes (MMMs), proven to be feasible in overcoming the trade-off relationship (Barooah et al. 2024; Choi et al. 2024). The discussion begins by highlighting the type of CO₂ capture process (pre-combustion, post-combustion, and oxy-fuel combustion). Subsequently, a comparison of different separation technologies (e.g., swing adsorption, absorption, and cryogenic distillation) is provided, highlighting their merits and limitations. Thereafter, recent progress on metal-organic framework (MOF)-based MMMs is summarised, considering their attractive properties compared to other porous materials, which will be elaborated on in this section. This review will focus on the use of MOFs with different configurations (two-dimensional (2D), three-dimensional (3D), and composites), as such analysis is reviewed with much little emphasis as compared to other published review articles in MOF-based MMMs (Goh et al. 2022; Tanvidkar et al. 2022; Yang et al. 2023) or MMMs (Guo et al. 2019; Hassan et al. 2023; Kamble et al. 2021; Katare et al. 2023) in general. Subsequently, the performance analysis of these materials is conducted, followed by a discussion on the future perspective of this membrane technology to pave future research efforts.

2. Integration of CO₂ Capture with Combustion Processes

There are various approaches for performing CO₂ capture, depending on the feed condition. CCUS technologies, in general, can be categorised into three major processes, namely (i) post-combustion; (ii) pre-combustion; and (iii) oxy-fuel combustion (Figure 1). In this section, a brief outline for each technology will be discussed, and a summary comparing each technology is presented in Table 1.

Figure 1. Illustration of post-combustion, pre-combustion, and oxy-fuel combustion process. The most appropriate location to install the CO₂ capture system is shaded for clarity.

Table 1. Comparison between post-, pre-, and oxy-fuel combustion carbon capture (Akpasi and Yusuf 2022; Chuah 2019; Leung et al. 2014; Sumida et al. 2012).

Parameters	Post-combustion	Pre-combustion	Oxy-fuel combustion
Definition	Fuels are burned with excess air	Fuels are decarbonized into shifted syngas before H2 conversion	Fuels are burned under oxygen at high purity
Source of CO2 emission	Power plants (coal or gas-fired)	Integrated gasification combined cycle and turbines (IGCC)	Power plants (coal or gas-fired)
Common gas pairs	CO2/N2	CO2/H2	CO2/H2O
Flue gas condition
Temperature	21–40 °C	40–80 °C	High temperature
Pressure	5–40 bar	1 bar	Low pressure
CO2 concentration (%)	5–20	15–40	75–80
CO2 capture efficiency (%)	85–90	85–90	90–100
Impurities	N2, O2, H2S, H2O, CO, NOx, SOx	H2, CH4, CO, H2S	H2O, sulphur-based compounds
Treated CO2 purity (%)	99.6–99.8	95.0–99.0	87.0–94.8
Merits	Flue gas stream is present at ambient conditions.	Lower energy penalty than post-combustion. Higher concentration of CO2 in the feed. H2 (blue hydrogen) can be utilized as an energy source.	High CO2 capture efficiency. Absence of hazardous impurities (NOx).
Limitations	Low CO2 concentration in flue gas. Presence of other impurities. Higher energy penalty for regeneration.	Syngas treatment and drying are required before CO2 capture. High initial capital and investment costs.	High operational and capital costs. Expensive O2 purification. Dilution of O2 with CO2 is required to minimize the heat of reaction during the combustion process.

Post-combustion CO₂ capture typically takes place after the combustion process, where the flue gas contains CO₂ at a relatively low concentration (5–20 vol%) (Mason et al. 2015; Sumida et al. 2012), depending on the type of raw materials (e.g., coal or natural gas) being combusted. Other components present in the stream are primarily nitrogen (N₂), along with other minor components such as water vapor (H₂O), oxygen (O₂), carbon monoxide (CO), nitrogen oxides (NO_x), and sulphur oxides (SO_x) (Chuah, Li, et al. 2020). This gas stream is expected to be predominantly a CO₂/N₂ binary mixture due to the high N₂ content in the air during combustion. The gas stream is released at the pressure of 1 bar, with an operating temperature of 40–80 °C (Mason et al. 2015; Sumida et al. 2012). This capture process typically occurs after SO_x removal (Sumida et al. 2012).

Pre-combustion CO₂ capture generally occurs before the combustion process and can be achieved through an integrated coal gasification combined cycle (IGCC). In this process, fuels (e.g., coal, municipal solid waste (MSW), and rubber crumb) are gasified at high temperature and pressure to produce synthesis gas (syngas). The syngas mixture mainly consists of hydrogen (H₂), CO, CO₂, and H₂O, together with contaminants, such as methane (CH₄) and hydrogen sulphide (H₂S). A water-gas shift reaction is then conducted on this gas mixture to produce shifted syngas (predominantly composed of H₂ and CO₂) at high pressure (5–40 bar) and slightly elevated temperature (40 °C). Pre-combustion CO₂ capture is employed to separate the H₂ and CO₂ gas mixture, yielding pure H₂ for electricity generation. This approach is commonly referred to as the production of blue (decarbonized) hydrogen. It is a cost effective solution at present time compared to the relatively costly green hydrogen production (US$ 3.00–6.55 per kg green H₂ vs. US$ 1.40–2.40 per kg blue H₂) (Chuah, Jiang, et al. 2021).

In contrast, oxy-fuel combustion involves burning carbon-based fuel (e.g., pulverized coal) under a pure oxygen (O₂) environment. This process generates the gas stream that mainly consists of CO₂ and H₂O, which can be easily separated through condensation due to a drastic difference in boiling point (Sumida et al. 2012). While oxy-fuel combustion simplifies the CO₂ capture process by creating flue gas that is almost entirely CO₂, the energy intensity of separating air into pure O₂ and N₂ can be considerably high due to their similar properties (Li et al. 2009; Madejski et al. 2022).

3. CO₂ Capture: Membranes vs. Other Technologies

Considering the flue gas is not present as pure CO₂, additional purification systems are necessary. Based on the impurities listed in Table 1, various approaches have been investigated for their potential utility in CO₂ capture. In this section, we compare membranes with other separation processes, including cryogenic distillation, amine scrubbing (absorption), and adsorption, to emphasize the merits of the membrane process (Figure 2).

Figure 2. Illustration of (a) cryogenic distillation, reprinted with permission from Shen et al. (2022), Copyright 2022 Elsevier; (b) amine scrubbing (absorption), reprinted with permission from Dutcher et al. (2015), Copyright 2015 American Chemical Society; (c) adsorption (temperature-vacuum swing adsorption (TVSA)), reprinted with permission from Nastaj et al. (2006), Copyright 2006 Elsevier; (d) membrane. Reprinted with permission from Freeman et al. (2014), Creative Common license CC BY 3.0.

Cryogenic distillation, a conventional technology, involves multiple stages of cooling and compression to allow phase change, resulting in pure CO₂. Originally adopted by Ryan and Holmes (Holmes and Ryan 1982; Lin, Zhang, et al. 2023; Ryan and Schaffert 1984; Shen et al. 2022), a more recent cryogenic distillation model optimized various parameters such as feed composition, CO₂ purity, pressure, temperature reflux ratio, and number of trays (Yousef et al. 2018). Through this approach can achieve high purity liquid CO₂ (99.9 vol%) with two distillation columns (c.a. 11–13 trays) at operating pressure and temperature of c.a. 46–50 bar and c.a. −80 to −74 °C, respectively, its limitations include high energy penalty, poor separation performance, and expensive installation costs. Hence, this process is recommended when CO₂ with high feed concentration (≥60 vol%) is present to ensure effective capture and recovery of CO₂ (Chuah, Lee, Bae 2021; Shen et al. 2022), without jeopardizing overall CO₂ removal efficiency. In this regard, the applicability of cryogenic distillation in post-combustion CO₂ capture is not recommended to be incorporated.

Next, amine scrubbing, achieved through CO₂ absorption, has been widely used since 1930 for CO₂ removal from natural gas (predominantly CH₄) and H₂ (Rochelle 2009). The system contains both absorption and stripping units, where CO₂ is first absorbed from the flue gas into the solvent at near ambient conditions. Solvent regeneration occurs in the stripping unit at an elevated temperature (100–120 °C) to recover the solvent for the subsequent absorption. The choice of solvents (a.k.a. absorbents) is critical for achieving high CO₂ absorption capacity, and such parameters are quantified using CO₂ solubility at equilibrium (Goto et al. 2009; Kim et al. 2023). Depending on the type of amines (primary, secondary, and tertiary), carbamate and bicarbonate species can be formed through reversible chemical reactions between CO₂ and amine. Monoethanolamide (MEA) is the most commonly studied amine in CO₂ absorption, typically in the range of 20–30 wt% in an aqueous solution (Chabanon et al. 2015). Undeniably, the required volume for CO₂ absorption can be decreased at a much smaller value with the increase in MEA concentration; however, undesirable corrosion of metallic parts may hinder its feasibility (Chuah, Kim, et al. 2019). Furthermore, the high energy penalty (−50 to −100 kJ/mol CO₂) for the regeneration process is attributed to the large heat capacity of water that is present in the amine solution (Sumida et al. 2012). Recent research has explored the use of lean solvents (e.g., thermomorphic biphasic solvent and lipophilic amines) or new designs (e.g., membrane contactors) to improve separation performance. However, challenges remain, including the loss of amines during desorption, contributing to increased CO₂ capture costs (Arif 2016; Zhang et al. 2011).

Adsorption relies on the difference in polarizabilities between CO₂ and N₂ gas (Li et al. 2009) for post-combustion CO₂ capture. Adsorbents in the adsorption column undergo repetitive adsorption–desorption cycling, analogous to the absorption process, requiring two adsorbent beds. Adsorbents saturated after adsorption are regenerated through temperature (TSA), pressure (PSA) or vacuum (VSA) swing adsorption. While PSA is not applicable due to the feed pressure being at ambient conditions, TSA presents the most viable option in this case despite challenges in real practice. This is due to the limited driving force to push out the adsorbed gas from the adsorption column during the desorption process. On the laboratory scale, such action is achieved through the addition of inert purge gas (Chuah, Li, et al. 2020; Yang et al. 2021), However, such an approach has been criticized heavily by researchers, considering the need to separate the purged gas from the absorbed gas. A hybrid approach, temperature and vacuum swing adsorption (TVSA), has been considered for effective regeneration of adsorbents (Gholami et al. 2023), although energy penalty must be carefully considered despite being comparatively lower than absorption in terms of the heat capacity values (Sumida et al. 2012).

Membrane-based separation has shown its competitiveness in generating high-purity CO₂ compared to cryogenic distillation, amine absorption, and swing adsorption process. Membranes function by separating gas components based on the difference in permeation rate between CO₂ and other impurities, utilising a solution-diffusion mechanism due to their non-porous structure. In the context of post-combustion CO₂ capture, CO₂ is expected to permeate faster than N₂ gas considering its comparatively high polarizability and smaller kinetic diameter (Lee et al. 2022; Li et al. 2024). Hence, this competitive advantage is evident when compared to cryogenic distillation and TVSA, offering cost-effectiveness and high energy efficiency. Additionally, membrane cascades can be set up with a small plant footprint. However, the membrane system at the present stage is still limited to a smaller feed flow rate with moderate CO₂ purity at the outlet streams (Chuah, Jiang, et al. 2021).

Undeniably, polymeric membranes have demonstrated dominance in gas separation due to their well-established synthesis procedure and scalability. However, as indicated in the introduction, the permeability-selectivity trade-off remains a major hurdle in achieving optimal performance. Hence, MMMs, created by integrating porous materials into the polymer matrix, present an attractive option for improving separation performance. Such integration allows the solubility and/or diffusivity of gases into the membrane to be tailored to ensure that the overall performance approach or exceed the upper bound limit. In this review, our focus will delve into MOFs as filler materials, which have shown attractive performance in post-combustion CO₂ capture.

4. MOF-Based Mixed-Matrix Membrane for Post-Combustion CO₂ Capture

MOF, also termed porous coordination polymers (PCPs) or porous coordination networks (PCNs), have proven high applicability in various fields, including molecular separation, gas storage, and heterogeneous catalysts. This is attributed to their unique structural properties, such as low densities (≤0.19 g cm⁻³), high internal surface areas (≤7000 m² g⁻¹), high void volumes (≤90%), and narrow pore size distributions (Farha et al. 2012; Stylianou and Queen 2015). MOFs are developed based on coordinative bonding between metal-based nodes and organic ligands, which can be synthesized through various approaches, including hydrothermal, solvothermal, mechanochemical, and sonochemical methods (Chuah et al. 2018). In this section, we will focus on the application of 2D, 3D and 2D–3D composite-based MOF as filler materials for post-combustion CO₂ capture. The performance of the reported membranes is summarized in Tables 2–4.

Table 2. Enhancement in CO₂ permeability and CO₂/N₂ selectivity of 2D MOF-based mixed-matrix membranes.

Filler	Polymer	Filler loading (wt%)	Thickness (µm)	Testing condition			P(CO2) (barrer)	% En.	α (CO2/N2)	% En.	Findex	Ref.
				Pressure (bar)	Comp. (CO2/N2)	Temp. (°C)
BCoC-ZIF	PIM-1	10	40–50	2	50/50	25	7325	74.4	32.5	22.6	1.15	22′ (Sun et al. 2022)
CuBDC	PES-g-PEG	7	80	1	–	35	66.6	367.0	32.6	48.5	2.68	22′ (Ma, He, Xu, et al. 2022)
CuBDC	PES-g-PEG	7	80	1	50/50	35	55.7	290.8	34.8	58.6	2.70	22′ (Ma, He, Xu, et al. 2022)
Cu(dhbc)2(bpy)	PEBAX MH 1657	15	30	12	15/85	25	102.1	70.2	73.9	105.2	2.61	19′ (Ying et al. 2019)
Cu(dhbc)2(bpy)	PEBAX MH 1657	15	30	12	15/85 (RH = 85%)	25	466.2	66.5	59.2	85.0	2.29	19′ (Ying et al. 2019)
Cu-TCPP NS	PEBAX MH 657/PEGMEA	0.1	160	3.5	–	25	1183	92.7	57.6	38.1	1.51	23′ (Wang et al. 2023)
MUF-15-Ns	PIM-1	5	80	1	–	20	16120	34.3	19.8	10.0	0.57	20′ (Yin et al. 2020)
MUF-15-Ns	PIM-1	5	80	1	50/50	20	13180	9.8	19.6	8.9	0.34	20′ (Yin et al. 2020)
NUS-8	PIM-1	2	54–78	1	50/50	25	6725	72.7	26.8	71.8	2.11	18′ (Cheng et al. 2018)
NUS-8-COOH	PIM-1	2	–	2	50/50	298	11512	77.1	30.9	54.5	1.82	22′ (Wang et al. 2022)
NUS-8-NH2	PIM-1	10	54	2	20/80	25	14638	87.0	29.2	77.0	2.27	22′ (Pu et al. 2022)
PEI-F-Ce-2	XLPEO	2	200	2	–	35	641	–	70.1	–	–	23′ (Zhao et al. 2023)
SUM-1	PEBAX MH 2533	5	70	2	10/90	35	380.9	59.2	20.5	32.3	1.27	23′ (Qin et al. 2023)
SUM-1	PEBAX MH 2533	5	70	2	10/90 (RH = 100%)	35	401.7	18.4	18.1	−1.1	0.14	23′ (Qin et al. 2023)
SUM-1	PEBAX MH 2533	7	30–90	6	–	35	443.3	230	27.2	22	1.27	23′ (Khalilpour et al. 2015)
SUM-9	PEBAX MH 2533	1	70	2	10/90	35	390.9	63.4	18.5	19.4	1.00	23′ (Qin et al. 2023)
SUM-9	PEBAX MH 2533	1	70	2	10/90 (RH = 100%)	35	452.4	33.3	19.7	7.7	0.50	23′ (Qin et al. 2023)
SUM-9	PEBAX MH 2533	1	30–90	6	–	35	539	124.5	24.7	12.3	1.14	23′ (Feng et al. 2023)
V-ZIF-8	PSF	–	2.5	3	–	30	97.5	1089.0	31.3	5.7	2.63	22′ (Wang, Xu, et al. 2021)
V-ZIF-8	PSF	–	2.5	3	50/50	30	89.7	1181.4	30.0	7.1	2.74	22′ (Wang, Xu, et al. 2021)
ZIF-67 NS	PEBAX MH 1657	5	60–100	1	–	25	139.4	50.7	73.2	76.0	2.04	20′ (Feng et al. 2020)
ZIF-C	PEBAX MH 1657	20	–	2	10/90	24	140	85	40	30	1.33	20′ (Deng et al. 2020)
ZIF-C	PEBAX MH 1657	20	–	1	10/90 (RH = 100%)	24	387.2	100	47.1	34	2.30	20′ (Deng et al. 2020)
ZIF-L(Co)	PEBAX MH 1657	2	–	4	–	25	56	9.8	67	52.3	1.31	23′ (Maleh and Raisi 2023)
ZIF-L(Zn)	PEBAX MH 1657	2	–	4	–	25	54	5.9	76	72.7	1.64	23′ (Maleh and Raisi 2023)
ZIF-L(Zn)@ZIF-67	PEBAX MH 1657	2	–	4	–	25	64	25.5	94	113.6	2.42	23′ (Maleh and Raisi 2023)
Zn2(bim)4	PIM-PMDA-OH	5	25	1	–	35	341.8	72.3	54.8	91.5	2.52	22′ (Ma, He, Tang, et al. 2022)
Zn2(bim)4	PIM-PMDA-OH	4	25	1	50/50	35	173.3	−12.6	57.7	101.6	1.89	22′ (Ma, He, Tang, et al. 2022)

Table 3. Enhancement in CO₂ permeability and CO₂/N₂ selectivity of 3D MOF-based mixed-matrix membranes.

Filler	Polymer	Filler loading (wt%)	Thickness (µm)	Testing condition			P(CO2) (barrer)	% En.	α (CO2/N2)	% En.	Findex	Ref.
				Pressure (bar)	Comp. (CO2/N2)	Temp. (°C)
Ag^+@UiO-66-NH2	PIM-1	5	120–140	2	–	25	9152.1	55.3	22.3	47.3	1.56	23′ (Lin, Yuan, et al. 2023)
Azo-UiO-66	Matrimid	24	80	–	–	–	10.0	44.9	37.0	24.0	0.99	18′ (Prasetya et al. 2018)
C-ZIF-700	PEBAX MH 1657	10	55	4	15/85	35	161.3	83.3	51.4	51.2	1.80	23′ (Cheng et al. 2023)
CeZr-UiO-66-NH2	PEBAX MH 1657	3	–	3	–	35	100.7	86.4	76.4	9.1	0.88	23′ (Du et al. 2023)
Cu-BDC	Polyactive	4	0.7	2	15/85	25	40 GPU	–	77	–	–	19′ (Sabetghadam et al. 2019)
Cu-TCPP NF	PEBAX MH 657/PEGMEA	0.1	160	3.5	–	25	550	−15.3	54	28.6	0.56	23′ (Wang et al. 2023)
GMA-UiO-66	APUA	15	–	1	–	25	14.2	125.4	53	76.7	2.46	19′ (Molavi and Shojaei 2019)
HKUST-1 (bulk)	6FDA-DAM	30	–	1	–	35	2360	970	14.6	17.7	0.33	19′ (Chi et al. 2019)
HKUST-1	Matrimid	30	10	10	35/65	35	18 GPU	80	23	35.3	1.46	11′ (Basu et al. 2011)
HKUST-1	Chitosan	10	120	–	–	25	2000	–	2	–	–	15′ (Casado-Coterillo et al. 2015)
HKUST-1/IL	Chitosan	10	146	–	–	25	1400	–	0.5	–	–	15′ (Casado-Coterillo et al. 2015)
HKUST-1-0NH2	Matrimid	20	50–70	1	20/80	35	22.2	113.5	33.8	9.0	1.01	19′ (Chuah, Li, et al. 2019)
HKUST-1-25NH2	Matrimid	20	50–70	1	20/80	35	13.0	25.0	42.7	37.7	1.15	19′ (Chuah, Li, et al. 2019)
HKUST-1 (branched)	6FDA-DAM	30	–	1	–	35	2480	970	16.3	17.7	0.77	19′ (Chi et al. 2019)
Ln-MOF	Matrimid	30	–	10	–	35	20.5	183.1	48.5	42.2	2.06	22′ (Bano et al. 2022)
Ln-MOF	Matrimid	30	–	10	50/50	35	15.2	145.2	45.0	45.2	1.97	22′ (Bano et al. 2022)
lys-c-Zr BDC	Chitosan	7	–	–	–	–	101	–	43	–	–	23′ (Katare and Mandal 2023)
lys-c-Zr BDC	Chitosan	7	4	2.21	–	85	135 GPU	980	71	31.5	3.17	23′ (Katare and Mandal 2023)
Mg2(dobdc)	PDMS	20	–	2	–	25	2100	−32.3	12	26.3	0.28	13′ (Bae and Long 2013)
Mg2(dobdc)	PI	10	–	2	–	25	850	30.8	23	64.3	−0.05	13′ (Bae and Long 2013)
Mg2(dobdc)	XLPEO	10	–	2	–	25	250	−34.2	25	13.6	1.70	13′ (Bae and Long 2013)
MIL-53	Matrimid	30	10	10	35/65	35	17 GPU	41.6	23	35.3	1.22	11′ (Basu et al. 2011)
MIL-96(Al)-NP3	6FDA-DAM	25	–	2	15/85	25	1122	47.6	26.4	10.0	0.66	17′ (Benzaqui et al. 2017)
MIL-101 (Cr)	PU	24	–	–	–	–	83.1	116.5	42.4	8.2	0.99	18′ (Rodrigues et al. 2018)
MIL-140C	PGO	20	0.60	1	–	25	1768 GPU	98.9	38	22.6	1.27	22′ (Kang et al. 2022)
MUF-15-Bulk	PIM-1	5	80	1	–	20	24000	100.0	15.0	−16.7	0.17	20′ (Yin et al. 2020)
NbOFFIVE-byp-Ni	PEBAX MH 1657 + PEGDME	2	–	–	–	25	474	14.2	58.5	7.5	0.37	24′ (Hou et al. 2023)
NbOFFIVE-2-Cu-i	PEBAX MH 1657 + PEGDME	2	–	–	–	25	578	39.3	70.4	29.3	1.07	24′ (Hou et al. 2023)
NbOFFIVE-2-Ni-i	PEBAX MH 1657 + PEGDME	2	–	–	–	25	482	16.1	58.9	8.2	0.37	24′ (Hou et al. 2023)
NH2-MIL-53 (Al)	6FDA-ODA	20	45–55	2	15/85	35	36.9	24.7	13.0	5.7	0.38	21′ (Loloei et al. 2021)
Ni2(L-asp)2bipy	PEBAX MH 1657	20	90–110	1	–	25	82.3	47.4	52.4	30.4	1.15	18′ (Fan et al. 2018)
Ni2(L-asp)2pz	PEBAX MH 1657	30	90–110	1	–	25	89.6	60.4	28.9	−28.0	−0.48	18′ (Fan et al. 2018)
Ni2(L-asp)2bipy	PEBAX MH 1657	20	90–110	1	50/50	25	91.7	64.1	35.2	−12.4	0.11	18′ (Fan et al. 2018)
PHZ-2	PEBAX MH 1657	10	–	1	–	25	172.9	140.1	87.9	75.8	2.51	21′ (Ding et al. 2021)
UiO-66	APUA	12	–	1	–	25	12	90.5	39	30.0	1.40	19′ (Molavi and Shojaei 2019)
UiO-66	Matrimid	24	80	–	–	–	7.8	13.0	29.4	−1.4	0.08	18′ (Prasetya et al. 2018)
UiO-66	ODPA-TMPDA	20	169	1	20/80	35	169	92.0	31.9	−3.6	0.54	20′ (Chuah, Lee, et al. 2020)
UiO-66	PEBAX MH 1657	10	–	3	50/50	25	963.0	87.0	56.6	34.4	1.48	16′ (Shen et al. 2016)
UiO-66	PEBAX MH 1657	10	–	3	50/50	25	139.7	171.0	61.1	45.1	2.07	16′ (Shen et al. 2016)
UiO-66	PEGDA	20	170–230	1	–	35	187.9	61.7	41.9	−6.1	0.30	21′ (Lee, Ozcan, et al. 2021)
UiO-66	PIM-1	23.1	–	1	–	25	7610	59.5	20.7	−5.1	0.31	17′ (Khdhayyer et al. 2017)
UiO-66	PSF	20	65	3	–	35	16.0	186.0	26.2	−11.0	0.71	16′ (Su et al. 2016)
UiO-66	PU	24	–	–	–	–	75.2	95.8	34.2	−12.8	0.27	18′ (Rodrigues et al. 2018)
UiO-66-Br	ODPA-TMPDA	20	200	1	20/80	35	200	127.0	34.5	4.2	0.94	20′ (Chuah, Lee, et al. 2020)
UiO-66-Br	PIM-1	10	80–100	4	–	25	2846	−6.8	21.6	34.2	0.78	17′ (Ghalei et al. 2017)
UiO-66-COOH	PIM-1	23.1	–	1	–	25	9980	109.2	21.6	−0.9	0.71	17′ (Khdhayyer et al. 2017)
UiO-66-H	PIM-1	20	80–100	4	–	25	2606	−14.7	24.6	52.8	1.08	17′ (Ghalei et al. 2017)
UiO-66-NB	PEG/PPG-PDMS	3	40–50	1	–	30	585	33.8	52.7	−6.2	0.11	20′ (Hossain et al. 2020)
UiO-66-NH2	APUA	15	–	1	–	25	13.6	115.6	44	46.7	1.88	19′ (Molavi and Shojaei 2019)
UiO-66-NH2	COOH-PI	20	100–120	4	–	25	461	25.3	18.1	30.2	0.99	21′ (Wang, Tian, et al. 2021)
UiO-66-NH2	(ODPA0.25/6FDA0.75)-DAM (P4)	29	–	3	–	35	516.5	145.3	11.2	−30.1	−0.13	23′ (Li, Qi, et al. 2023)
UiO-66-NH2	ODPA-TMPDA	20	142	1	20/80	35	142	61.3	37.1	12.0	0.81	20′ (Chuah, Lee, et al. 2020)
UiO-66-NH2	PIM-PI	15	–	1	–	30	1470	−31.8	21	43.8	0.67	21′ (Husna et al. 2021)
UiO-66-NH2	PEG/PPG-PDMS	3	40–50	1	–	30	180.3	−58.7	41.5	−26.2	−1.76	20′ (Hossain et al. 2020)
UiO-66-NH2	PIM-1	10	80–100	4	–	25	2869	−6.1	27.5	70.8	1.48	17′ (Ghalei et al. 2017)
UiO-66-NH2	PIM-1	23.1	–	1	–	25	5070	6.3	20.0	−8.3	−0.19	17′ (Khdhayyer et al. 2017)
UiO-66-NH2	PIM-1	5	120–140	2	–	25	7777.4	32.0	14.6	−0.40	0.17	23′ (Lin, Yuan, et al. 2023)
UiO-66-NH2	PEBAX MH 1657	10	–	3	50/50	25	87.0	68.9	66.1	57.0	1.82	16′ (Shen et al. 2016)
UiO-66-NH2	PEBAX MH 1657	10	–	3	50/50	25	130.2	153.0	72.2	71.5	2.48	16′ (Shen et al. 2016)
UiO-66-NM@PEG	PVDF	23.5	100	1	–	25	611.1	–	44.5	–	–	23′ (Li, Liu, et al. 2023)
UiO-66-(OH)2	ODPA-TMPDA	20	125	1	20/80	35	125	42.0	38.9	17.5	0.91	20′ (Chuah, Lee, et al. 2020)
UiO-66-(OH)2	PEBAX MH 1657	5	–	2	–	25	130	62.5	62	29.2	1.22	22′ (Yang et al. 2022)
UiO-66-(OH)2@PIL	PEBAX MH 1657	20	–	2	–	25	131	63.8	71	47.9	1.62	22′ (Yang et al. 2022)
UiO-66-Ref	PIM-1	20	80–100	4	–	25	6981	128.6	13.0	−19.3	0.21	17′ (Ghalei et al. 2017)
UiO-66-TF6	PEGDA	20	170–230	1	–	35	258.3	122.3	41.1	−7.8	0.56	21′ (Lee, Ozcan, et al. 2021)
UiO-67	PGO	20	0.60	1	–	25	1745 GPU	96.3	35	12.9	1.02	22′ (Kang et al. 2022)
ZIF-7	PEBAX MH 1657	22.1	812		–	20	111	54.2	97	185.3	3.46	13′ (Li et al. 2013)
ZIF-L(Co)	PEBAX MH 1657	2	–	4	–	25	60	17.6	63	43.2	1.20	23′ (Maleh and Raisi 2023)
ZIF-8	Chitosan	20	125	–	–	25	2800	–	6	–	–	15′ (Casado-Coterillo et al. 2015)
ZIF-8	PEBAX MH 2533	15	60–69	2	–	24	230.0	84.0	30	66.7	2.09	22′ (Li, Kujawski, Tonkonogovas, et al. 2022)
ZIF-8	PEBAX MH 1657	10	–	1	–	25	150.0	108.3	40	−20	1.79	21′ (Ding et al. 2021)
ZIF-8	PVAm/PSF	–	–	10	15/85	25	297 GPU	–	83	–	–	15′ (Zhao et al. 2015)
ZIF-8	Matrimid	30	10	10	35/65	35	19 GPU	58.3	18	5.9	0.62	11′ (Basu et al. 2011)
ZIF-8/IL	Chitosan	20	125	–	–	25	2500	–	0.8	–	–	15′ (Casado-Coterillo et al. 2015)
ZIF-8-PEI-IL	PEBAX MH 2533	15	60–69	2	–	24	285.0	128.0	76	322.2	4.98	22′ (Li, Kujawski, Tonkonogovas, et al. 2022)
ZIF-67-L	PEBAX MH 1657	10	90–120	2	–	30	91.6	76.1	51.8	23.3	1.17	23′ (Zhao et al. 2023)
ZIF-67 MP	PEBAX MH 1657	5	60–100	1	–	25	101.0	9.2	45.1	8.4d	0.32	20′ (Feng et al. 2020)
ZIF-67 NP	PEBAX MH 1657	5	60–100	1	–	25	95.7	3.5	46.0	10.6	0.32	20′ (Feng et al. 2020)
ZIF-94	PEBAX MH 1657	20	45	3	15/85	35	128.0	197.6	26.0	−21.2	0.40	21′ (Hasan et al. 2022)
ZIF-300	PEBAX MH 1657	30	–	4	–	20	83.0	59.6	84.0	52.7	1.69	17′ (Yuan et al. 2017)
ZIF-UC-6	PEBAX MH 1657	25	40–53	1	50/50	30	576	476.0	86.2	146.3	4.35	23′ (Li, Ma, et al. 2023)
ZIF-L-Co	PEBAX MH 1657	5	90–120	2	–	30	68	28.3	47	6.8	0.44	23′ (Zhao et al. 2023)
Zr-BDC	Chitosan	7	4	2.21	–	85	79	532.0	60	11.1	2.15	23′ (Katare and Mandal 2023)

1 GPU = 10⁻⁰ cm³ (STP) cm⁻⁽ s^−s cmHg^−c.

Table 4. Enhancement in CO₂ permeability and CO₂/N₂ selectivity of MOF composite-based mixed-matrix membranes.

Conf.	Filler	Polymer	Filler loading (wt%)	Thickness (µm)	Testing condition			P(CO2) (barrer)	% En.	α (CO2/N2)	% En.	Findex	Ref.
					Pressure (bar)	Comp. (CO2/N2)	Temp. (°C)
2D/2D	CuBDC-ns@MOS2	PEBAX MH 1657	2.5	–	4	15/85	35	123.0	61.8	69.0	109.1	2.61	22′ (Liu et al. 2022)
2D/3D	GO-ZIF	PVI-POEM (PSF support)	5	0.1	1	–	25	995.1 GPU	58.9	45.9	20.7	1.01	21′ (Lee, Song, et al. 2021)
2D/3D	MIL-GO-2	PEBAX MH 2533	9.1	–	2	–	20	303	51.5	24	41.2	1.41	22′ (Li, Kujawski, Knozowska, et al. 2022)
2D/3D	PIM-g-MOF	PEBAX MH 1657	1	–	1	–	30	247	74.7	56.1	58.9	1.90	22′ (Husna et al. 2022)
2D/3D	PSA@ZIF-8	PEBAX MH 1657	10	–	1	–	25	120.0	66.7	50.0	0	0.51	21′ (Ding et al. 2021)
2D/3D	UiO-66-NH2/GO (UN@GO)	Matrimid	5	–	3	–	25	7.3	–	52.0	–	–	19′ (Jia et al. 2019)
2D/3D	ZIF-8@BNNS	PEBAX MH 1657	20	75–90	5	–	25	106.5	35.5	83.2	70.8	1.85	23′ (Guo et al. 2023)
2D/3D	ZIF-90@C3N4	PEBAX MH 1657	8	50–70	2	–	25	110.5	41.7	84.4	79.6	2.04	22′ (Guo et al. 2022)

1 GPU = 10⁻⁰ cm³ (STP) cm⁻⁽ s^−s cmHg^−c.

4.1. Two-dimensional MOF

2D MOFs, also described as nanosheet (NS) MOFs due to their high aspect ratios, have demonstrated their applicability in MMMs by enhancing gas selectivity at low filler loading (<5 wt%) in most circumstances (Table 2). This contrasts with 3D MOF, which requires higher filler loading (>10 wt%) to observe significant enhancement in separation performance (Gong et al. 2018; Samarasinghe et al. 2018). In this section, the discussion will be categorized based on the use of different metal clusters, specifically copper (Cu²⁺), zirconium (Zr⁴⁺), and indium (In³⁺).

First, 2D MOF NSs in MMM fabrication begin with the development of NS 2D copper 1,4-benzenedicarboxylate (ns-CuBDC), as reported by Rodenas et al. (2015). The approach utilizes a bottom-up strategy via a diffusion-mediated synthesis method between Cu²⁺ ions and 1,4-benzenedicarobyxlic acid. The potential utility of this MOF in MMM fabrication has been validated using different polymeric membranes and gas pairs (CO₂/CH₄) (Cheng et al. 2017; Rodenas et al. 2015; Samarasinghe et al. 2018; Yang, Goh, et al. 2017), demonstrating the capability to surpass the upper bound limit. A recent study on utilizing ns-CuBDC in CO₂/N₂ separation involves the use of polyethylene glycol (PEG) as reported by Ma, He, Xu, et al. (2022). The gas permeation results indicate an increase in both CO₂ permeability and CO₂/N₂ selectivity up to 5 wt% loading (Figure 3(a)). This co-current enhancement was not observed in previous investigations (Cheng et al. 2017; Rodenas et al. 2015; Samarasinghe et al. 2018; Yang, Goh, et al. 2017), presumably due to PEG, which exhibits comparatively lower gas permeability than the highly permeable glassy polymers used in previous works. Thus, ns-CuBDC is expected to enhance both CO₂ solubility and diffusivity in MMM, as verified by solubility-diffusivity analysis conducted by the authors.

Figure 3. (a) Investigation of CO₂ permeability and CO₂/N₂ selectivity of ns-CuBDC based MMM with PEG as the polymer matrix, reprinted with permission from Ma, He, Xu, et al. (2022), Copyright 2022 Elsevier; (b) reaction scheme for the preparation of 2D and 3D Cu-TCPP; (c) CO₂ permeability and CO₂/N₂ selectivity of Cu-TCPP-based mixed-matrix membrane with the increase in Cu-TCPP (2D: nanosheet (NS); 3D: nanoflake (NF) content at 3.5 bar and 25 °C. Reprinted with permission from Wang et al. (2023), Copyright 2023 Elsevier; (d) effect of feed pressure on CO₂ permeability, N₂ permeability as well as CO₂/N₂ separation factor for MMM. Reprinted with permission from Ying et al. (2019), Copyright 2019 Elsevier; effect of increased (e) NUS-8, (f) NUS-8-NH₂, and (g) NUS-8-COOH loading towards the change in CO₂/N₂ separation performance (25 °C and 1 bar operating pressure for NUS-8; 25 °C and 2 bar for NUS-8-NH₂ and NUS-8-COOH). Reprinted with permission from Cheng et al. (2018), Copyright 2018 Elsevier; Pu et al. (2022), Copyright 2022 Elsevier and Wang et al. (2022), Copyright 2022 Elsevier.

Subsequent research endeavours have explored the utilisation of Cu²⁺ as the metal cluster in producing NSs with different ligands. First, a surfactant-assisted bottom-up synthesis approach, developed by Zhao et al. (2015), was employed for the creation of sub-10 nm thick 2D-based copper (tetrakis(4-carboxyphenyl)porphyrin), known as Cu-TCPP. To validate the effectiveness of this 2D material in CO₂ capture, a systematic comparison was conducted between 2D Cu-TCPP NS and 3D Cu-TCPP nanoflower. The latter was developed by adjusting surfactant dosages (Figure 3(b)) and evaluated using PEBAX MH-1657 as the polymer matrix. Based on the same measurement condition, 0.1 wt% of 2D Cu-TCPP NS have demonstrated 82% and 37% increase in CO₂ permeability and CO₂/N₂ selectivity with reference to the polymeric membrane. In contrast, 3D Cu-TCPP is unable to showcase its competitiveness, as evidenced by an 18% decrease in CO₂ permeability at the same loading (Figure 3(c)). This decline is attributed to reduced CO₂ diffusivity, stemming from particle agglomeration due to poor dispersibility – a common observation in MMM fabrication when 3D MOFs are utilized (Li et al. 2019).

Another notable investigation of Cu-based 2D MOF involves the utilization of flexible Cu(dhbc)₂(bpy) MOF (Hdhbc = 2,5-dihyroxybenzoic acid; 4,4′-bpy = 4,4′-bipyridine) that was developed by Kitaura et al. (2003) for potential use in CO₂/N₂ separation. However, the effectiveness of Cu(dhbc)₂(bpy) in mitigating CO₂ plasticization, a phenomenon leading to decreased mixed-gas selectivity at elevated pressure, was not observed (Figure 3(d)). It is important to note that while the significance of CO₂ plasticization is lower in post-combustion CO₂ capture due to its low feed pressure (Yang et al. 2021), this behaviour could serve as a crucial starting point for investigating other gas pairs, such as natural gas purification and biogas upgrading (CO₂/CH₄) where high CO₂ partial pressure is present (Chuah et al. 2018).

The investigation of 2D MOFs in CO₂ capture also involves the utilization of zirconium (Zr)-based metal sites. The creation of this 2D MOF (NUS-8), which was developed by the Zhao research group from the National University of Singapore (NUS) is referred to as NUS-based MOF. This synthesis employed hydrothermal method with zirconium tetrachloride (ZrCl₄) and 1,3,5-benzenetribenzoic acid (H₃BTB). NUS-8 MOF was then investigated for CO₂/N₂ separation with PIM-1 (PIM = polymer of intrinsic microporosity) as the polymer matrix. Cheng et al. (2018) showcased the optimal CO₂ permeability (6500 barrer, with 1 barrer = 10⁻¹⁰ cm³ (STP) cm cm⁻² s⁻¹ cmHg⁻¹) and CO₂/N₂ selectivity (26.8) at 2 wt% loading (Figure 3(e)). Hence, to further enhance the CO₂/N₂ separation performance (i.e., transcend the upper bound limit), pre-synthetic functionalization was performed using amines (H₃BTB-NH₂) (Pu et al. 2022) or carboxylate (H₃BTB-COOH) (Wang et al. 2022) ligands to create NUS-8-NH₂ and NUS-8-COOH MOF, respectively. The creation of functionalized MOF has resulted in a much higher CO₂ permeability (14368 barrer and 18000 barrer for NUS-8-NH₂ and NUS-8-COOH, respectively), with reference to the PIM-1 polymer matrix (Figure 3(f,g)).

Besides, the application of Zr-based MOF as filler materials was achieved through hydroxamate-based MOFs that were synthesized by the Dai research group from Sichuan University Materials (SUM) (Lai et al. 2022). Subsequently, Feng et al. (2023) utilize SUM-1 as the filler material for CO₂/N₂ separation, in conjunction with the polymer matrix PEBAX MH-2533. Gas permeation data indicated that high gas permeability could be achieved with SUM-1 at 7 wt% loading. However, the reported separation performance (CO₂ permeability and CO₂/N₂ selectivity 443 barrer and 27.2, respectively) proved challenging to surpass the upper bound limit. Hence, the same author has created indium(III)-based MOF (SUM-9), which is capable of achieving comparatively higher performance (539 barrer and 24.7 for CO₂ permeability and CO₂/N₂ selectivity, respectively) at a much lower loading. Such performance was hypothesized by the uniform distribution of SUM-9 in the polymer matrix at comparatively low filler loading (1 wt%) as compared to SUM-1, which is only capable of generating such an effect at higher loading (5 wt%).

Last but not least, a comparative study involving structural and morphological changes was performed on MOFs. This involves the use of ZIF-67, developed by bridging cobalt (Co) cations and 2-methyl imidazolate anions, was employed for CO₂/N₂ separation. The aperture size (3.4 Å) of ZIF-67 falls between the kinetic diameter of CO₂ and N₂. The study focused on tuning the overall morphology to improve CO₂ permeability in MMM. This involves the creation of 3D ZIF-67 (leaflets (ZIF-67-L), microparticles (MPs), and nanoparticles (NPs)) and 2D ZIF-67 NSs by Zhao et al. (2023) and Feng et al. (2020), respectively. The data reported that the creation of ZIF-67 NP and MP resulted in structural instability due to the dissolution of metal cluster and ligands during the fabrication of polymer PEBAX MH-1657. Conversely, ZIF-67 NS exhibited increased structural stability, attributed to the higher proportion of [2 1 1] crystal faces, which contains the least Co–N bonds per unit area. Hence, the comparative study between ZIF-67 NP, ZIF-67 MP, and ZIF-67 NS in PEBAX MH-1657 membrane demonstrated that incorporation of ZIF-67 NS in the membrane improves both CO₂ permeability (50.7%) and CO₂/N₂ selectivity (76.0%) substantially compared to ZIF-67 NP and ZIF-67 MP, which demonstrates an improvement of less than 10% in both parameters.

In conclusion, the potential of 2D MOFs is evident in their ability to enhance performance at small filler loading. However, the structural vulnerability of most 2D MOFs poses a challenge due to their inherent nature. Hence, before MMMs fabrication, the samples are required to be suspended in organic solvent to minimize agglomeration of 2D NS before membrane casting. The subsequent sections elaborated below (3D MOFs) which are commonly utilized in membrane fabrication, will be discussed.

4.2. Three-dimensional MOF

3D MOFs are commonly adopted as filler materials in membrane fabrication due to their ease of synthesis protocol compared to 2D MOFs. Based on the summary data from Tables 2–4, up to 71% of the reported literature rely on 3D MOF for membrane fabrication. This section focuses on various commonly reported MOFs, including HKUST-1 (MOF-199), UiO-66, and UiO-67, as well as ZIF-based series considering the attractive features in the context of MMM fabrication.

HKUST-1 stands out as a frequently explored MOFs in MMM fabrication. It is constructed using multiple networks of paddlewheel Cu₂(COO)₄ nodes and 1,3,5-benzenetricarboxylate (btc³⁻) as the linker (Basu et al. 2011; Chuah et al. 2017; Chuah, Samarasinghe, et al. 2020; Chui et al. 1999). The removal of solvent molecules in HKUST-1 creates coordinatively unsaturated open metal sites, facilitating preferential adsorption of CO₂. The well-defined pore channels of 9 × 9 Å allow strong facilitation for gas molecules in MMM, leading to a 15% increase in CO₂ diffusivity (for 20 wt% HKUST-1). However, the presence of pore channels significantly larger than the studied gas species (i.e., CO₂ and N₂) hinders HKUST-1 from demonstrating improved CO₂/N₂ selectivity.

To address this limitation, Chuah, Li, et al. (2019) conducts a study where the metal sites in HKUST-1 are post-synthetically functionalized with amines to increase CO₂ adsorption at low partial pressure (Yang et al. 2019). Gas permeation result indicated that amine-functionalized HKUST-1 (i.e., HKUST-1-25NH₂ at 20 wt% loading) could increase CO₂/N₂ selectivity by 38% with reference to polymeric membrane. However, excessive amines incorporation will lead to drastic decrease in accessible surface area, thus reducing the number of available active sites for CO₂ adsorption (Figure 4(a,b)). An alternative approach by Chi et al. (2019) adopted branched HKUST-1 NPs through carboxylate as the modulator. Unlike bulk HKUST-1, branched HKUST-1 demonstrated the advantage of maintaining good dispersion with minimal agglomeration (Park et al. 2023). This has ensured an increased CO₂ permeability with up to 30 wt% loading of branched HKUST-1, while minimising the compromise in CO₂/N₂ selectivity (Figure 4(c)). In contrast, bulk HKUST-1 exhibited a 17% decrease in CO₂/N₂ selectivity at 30 wt% loading. The findings suggest that branched HKUST-1 offers the potential for uniform dispersion of NPs in MMM compared to traditional bulk HKUST-1 particles. However, the nature of HKUST-1 relying on coordinatively unsaturated open metal sites, poses challenges for its use under humid conditions, as verified by various literature studies (Chuah, Li, et al. 2020; Yang, Chuah, et al. 2017). Such observation was also valid for the case of M-MOF-74, investigated in MMM for CO₂/N₂ separation due to the presence of M²⁺ sites (Chuah et al. 2018; Li, Chuah, et al. 2018), allowing attractive CO₂ adsorption, as reported by Bae and Long (2013).

Figure 4. (a, b) Comparison of N₂ physisorption isotherm (77 K) and CO₂ adsorption (25 °C) of HKUST-1 with varying percentages of amine loading. Reprinted with permission from Chuah, Li, et al. (2019), Copyright 2019 Elsevier; (c) comparison between bulk and branched HKUST-1 with the increase in loading (10, 20, and 30 wt%), through 6FDA-DAM as the polymer matrix. Reprinted with permission from Chi et al. (2019), Copyright 2019 Wiley-VCH; (d, e) Effect of humidity on CO₂ permeability and CO₂/N₂ selectivity of MMM, with the increase in UiO-66-NH₂ loading on PEBAX MH-1657 membrane. Reprinted with permission from Shen et al. (2016), Copyright 2016 Elsevier; (f) fabrication process of PEG-grafted UiO-66 on PVDF support. Reprinted with permission from Li, Liu, et al. (2023), Copyright 2023 Elsevier; (g, h) CO₂/N₂ separation performance of MMM based on different loadings of UiO-66-(OH)₂ and UiO-66-(OH)₂@PIL in PEBAX MH-1657 membrane. Reprinted with permission from Yang et al. (2022), Copyright 2022 American Chemical Society.

The use of MOFs with relatively higher stability in humid conditions has proven pivotal in expanding its applicability in post-combustion CO₂ capture. One notable approach involves the use of isoreticular UiO-series (UiO-66 and UiO-67, with UiO = Universitetet i Oslo) that forms relatively stable coordination bonding between Zr metal sites and ligands, particularly under humid conditions. In a representative study by Shen et al. (2016), both CO₂ permeability and CO₂/N₂ selectivity increase (Figure 4(d,e)) after the relative humidity of the feed increases from 20% to 85%, with a simultaneous rise in weight fraction of UiO-66-NH₂ in the polymer matrix to 20 wt%. The focus of UiO-66 investigation has predominantly been on pre-synthetic functionalization, in contrast to post-synthetic functionalization discussed earlier for HKUST-1. This is attributed to the inherent difficulty in enhancing CO₂/N₂ selectivity using UiO-66 alone, despite with the feasibility of improving CO₂ permeability (Li, Qi, et al. 2023; Rodrigues et al. 2018). Thus, the ligand (2,5-dihydroxyterephthalic acid) has been subject to pre-synthetically functionalized with various ligands, including amines (Chuah, Lee, et al. 2020; Ghalei et al. 2017; Hossain et al. 2020; Husna et al. 2021; Lin, Yuan, et al. 2023; Shen et al. 2016; Su et al. 2016), azobenzene (Prasetya et al. 2018), bromine (Chuah, Lee, et al. 2020), carboxylic acid (Khdhayyer et al. 2017; Wang, Tian, et al. 2021), and hydroxyl (Yang et al. 2022). As reported in the literature, the addition of these functionalities predictably enhances mixed-gas selectivity while partially sacrificing CO₂ permeability. This trade-off is a result of the decreased accessible surface area in the UiO-66 framework due to the introduction of such functionalities. Alternatively, the polymers can be modified with various functional groups (instead of MOFs) to enhance the separation performance. For instance, the study by Carja et al. (2021) incorporated UiO-66 into functionalized PIM-1 (with amidoxime (AO), tetrazole (TZ), and N-((2-ethanolamino)ethyl)carboxamide (EA)). With the aid of the combined computations and experimental investigations, the hydrogen bonding between MOF fillers and functional groups in the polymer chains has assisted in minimizing the nanoscale gap between MOF/polymer interface, leading to an improved separation performance. Hence, the computational finding also revealed that MOF/polymer adhesion can be enhanced, not entirely by the functionalization of both MOF and polymers.

Recent studies have explored the expansion of isoreticular UiO-series beyond these four functionalities to improve the interfacial interaction between the porous filler and polymer matrix. In a study by Li, Liu, et al. (2023), maleic anhydride (MAH) serves as the mediator between MOF (UiO-66-NH₂) and polymer (poly(vinylidene fluoride), PVDF) to improve the overall compatibility (Figure 4(f)). Furthermore, MAH could also be utilized as the active sites to be impregnated with poly(ethylene glycol) (PEG), enhancing CO₂/N₂ separation performance without blocking the active sites in UiO-66-NH₂. This led to a notable enhancement in both CO₂ permeability (7367%) and CO₂/N₂ selectivity (7142%) compared to pure PVDF membrane. In a similar vein, Yang et al. (2022) grafted poly(ionic liquid) (PIL) onto the surface of the hydroxyl-modified UiO-66 (UiO-66-(OH)₂). Under a similar loading (20 wt%), the addition of PIL (UiO-66-(OH)₂@PIL) demonstrated significantly higher CO₂ separation performance (Figure 4(g,h)) compared to UiO-66-(OH)₂. This highlights the importance of PIL in mitigating potential defects that may arise in MMM fabrication.

Apart from this, zeolitic imidazolate frameworks (ZIFs), a subclass of MOFs, have been incorporated as filler materials in gas separation. ZIFs demonstrate stronger chemical and thermal stability than MOFs, attributed to the generally less stable metal–ligand bond in MOFs in contrast to imidazolate–base linkers used in ZIFs. Notably, ZIF-8, built with zinc ion centres coordinated with 2-methyl imidazolate, has displayed robust stability against water and methanol, maintaining crystallinity even at temperature up to 200 °C (Chuah et al. 2018; Park et al. 2006). However, similar to observations in the literature studies, the utilization of ZIF-8 alone tends to improve CO₂ permeability without a concurrent improvement in mixed-gas selectivity (Li, Samarasinghe, et al. 2018; Nafisi and Hägg 2014; Samarasinghe et al. 2018).

To address this, ZIF-8 was incorporated with CO₂-philic groups to improve the CO₂/N₂ selectivity and improve compatibility between the filler and polymer matrix. In a study by Li, Kujawski, Tonkonogovas, et al. (2022), the facilitation of CO₂ capture performance was performed through incorporation with branched polyethyleneimine (PEI) and ionic liquid (IL). The gas permeation data (Figure 5(a,b)) based on the comparison across different filler materials (ZIF-8, ZIF-8-PEI, and ZIF-8-PEI@IL) in PEBAX MH-2533 polymer have indicated that the addition of IL will increase the CO₂/N₂ ideal selectivity at a much substantial extent as compared to ZIF-8 and ZIF-8-PEI. This result confirmed the role of IL in improving the CO₂ solubility in MMMs. However, it is also critical to optimize the quantity of IL in the ZIF-8-PEI sample, considering a potential trade-off between the accessible surface area and CO₂ adsorption capacity can be potentially present in the process (Ban et al. 2015; Hao et al. 2013).

Figure 5. (a, b) Effect of various ZIF-8-based filler materials in influencing CO₂ permeability and CO₂/N₂ selectivity in PEBAX MH-2533-based MMM. The measurement is performed at 24 °C and 2 bar. Reprinted with permission from Li, Kujawski, Tonkonogovas, et al. (2022), Copyright 2022 Elsevier; (c) comparison between the gas mass transfer distance between modified hollow ZIF-8 (PHZ)-based MMM and ZIF-8-based MMM. Reprinted with permission from Ding et al. (2021), Copyright 2021 Elsevier.

On the other hand, Ding et al. (2021) adopted a similar analogy to improve CO₂ separation performance by creating modified hollow ZIF-8 (PHZ) through a hard template-incomplete etching method using polystyrene-acrylate (PSA). This creation effectively reduced the overall mass transfer distance (D_E vs. D_N in Figure 5(c)), leading to an increase in both CO₂ permeability (140%) and CO₂/N₂ selectivity (76%) compared to the pure PEBAX MH-1657 membrane.

In general, the use of 3D MOF as porous materials in MMMs is still expected to demonstrate its promise due to its capability to enhance the separation performance with a reasonable increase in filler loading. However, despite the presence of ligands that may help to improve the interfacial compatibility between MOF and polymer (i.e., sieve-in-a-cage behaviour in zeolites; Chuah, Goh, et al. 2021; Chuah, Lee, Bao, et al. 2021; Li et al. 2020; Nguyen et al. 2016), undesirable agglomeration at high filler loadings remains the challenge. Thus, research efforts on creating mixed-dimensional composites (Park et al. 2023) have gained interest among researchers to improve the CO₂/N₂ separation performance.

4.3. MOF composites

As elaborated above, the properties of MOF towards improved gas separation performance have been performed through various adjustments in physicochemical properties, such as adjustment in structure length to tune the accessible surface area and post-synthetic grafting to improve the CO₂ adsorption capacities (Chuah et al. 2018; Sumida et al. 2012). However, adsorbents developed solely from a single type of porous materials often show limited capacities for gas separation, due to limited sites capable of generating strong adsorptive forces for small molecules. Thus, MOF composites with 3D architecture, integrating with 2D materials (e.g., graphene oxide (GO), graphitic carbon nitride (g-C₃N₄), and molybdenum disulfide (MoS₂)) are utilized as fillers to improve the CO₂/N₂ separation performance in MMMs.

Graphene oxide, a derivative of graphene, exhibits oxygen-containing functional groups (carboxyl, carbonyl, epoxy, and hydroxyl) at the basal planes and edges that allow strong affinity towards CO₂ (Nie et al. 2021). The preparation of MOF/GO composites as filler materials for CO₂/N₂ separation was successfully prepared by Jia et al. (2019) and Lee, Song, et al. (2021). The synthesis in general involves in situ growth through the co-precipitation method, where GO powders are dispersed in a homogenous solution containing precursors to make 3D adsorbents. This process ensures improved homogeneity and dispersibility of the fabricated composite. Gas permeation test revealed that the presence of interconnected MOF particles in MOF/GO at higher loadings substantially increased gas permeability in MMM (Figure 6(a,b)). In contrast, the incorporation of GO only (2D material) creates a tortuous path, making it difficult to enhance the gas permeabilities in MMM (Figure 6(c)).

Figure 6. (a) Influence of GO and GO-ZIF loading in PVI-POEM (poly(vinyl imidazole)-co-poly(oxyethylene methacrylate)) membrane in influencing CO₂/N₂ separation performance. The data are obtained from Lee, Song, et al. (2021). (b) Performance summary of MMM-containing GO, UiO-66-NH₂ (U-NH₂), UiO-66/GO (U@GO), UiO-66-NH₂/GO (UN@GO) in Matrimid polyimide (PI). The data are obtained from Jia et al. (2019); (c) effect of different GO and GO-ZIF loading in influencing the gas transport pathway in MMM. Reprinted with permission from Lee, Song, et al. (2021), Copyright 2021 American Chemical Society; (d) effect of different loading of MoS₂ in the composite in influencing the CO₂/N₂ and CO₂/CH₄ separation performance in PEBAX MH-1657-based MMM; (e) effect of different loading of ns-CuBDC@10MoS₂ composite in PEBAX MH-1657-based MMM in CO₂-based separation performance. Reprinted with permission from Liu et al. (2022), Copyright 2022 Elsevier.

The investigation of MOF composite beyond 2D/3D configuration was performed by Liu et al. (2022) through the development of ns-CuBDC that was in situ growth onto multilayered MoS₂ NS (i.e., 2D/2D composite). As elaborated in the previous section, the addition of ns-CuBDC alone is unable to achieve improved CO₂ permeability as ns-CuBDC is arranged perpendicular to the flow, leading to difficulty in enhancing gas transport behaviour. The incorporation of MoS₂ NS into the synthesis, on the other hand, can improve CO₂ adsorption capacity as compared to ns-CuBDC, due to the creation of a coordination complex between Cu ions on ns-CUBDC and S atoms in MoS₂ NS. However, optimization of the loading of MoS₂ in MMM is critical. This is because the overall porosity in the composite decreases with the increase in MoS₂ content in the composite (Figure 6(d)). With the optimized content in ns-CuBDC and MoS₂ loading in the composite (ns-CuBDC@10MoS₂), the investigation was performed with the increase in filler loading in the PEBAX MH-1657 membrane. An increase in filler loading beyond 2.5 wt% does not illustrate an improvement in the separation performance due to the formation of filler agglomerates that result in poor dispersion of the NSs.

Based on the performance summary in Table 4, the investigation of MOF composite in MMM is still limited as compared to 2D and 3D MOF. The study of MOF composites, in general, is challenged by the need to optimize the loading between each material to ensure that MOF composite is developed with the most optimal porosity during the membrane fabrication.

5. Performance Evaluation of MOF-Based MMM

5.1. Comparison with upper bound

The analysis of MOF as filler materials in MMM was accessed through various metrics. First, the CO₂/N₂ separation performance was compared with the upper bound limit, an important goal for researchers worldwide in the field of gas separation, as evident from shifting upper bound line from 2008 to 2019 (Comesaña-Gándara et al. 2019; Robeson 2008). The performance data reported in Tables 2–4 are included in the upper bound plot in Figure 7(a). From the profile, it is indisputable that only a limited number of filler materials (10%) can achieve performance beyond the recently constructed upper bound limit. Undeniably, this method has proven the simplest yet effective approach to validating the performance of MMM. However, this construction, initially relies on the correlations observed for pure polymeric membranes, is generally influenced by the selection of the polymers used in the membrane fabrication. Hence, if the adopted polymeric membranes possess reasonably high intrinsic gas permeability (e.g., PIM-1, PEBAX MH-1657, and PEBAX MH-2533), it is not surprising that the performance can more easily surpass the upper bound compared to membranes with low gas permeability (e.g., Matrimid^® and polysulfone).

Figure 7. (a) Comparison of CO₂/N₂ separation performance of the reported MMMs reported in Tables 2–4 with upper bound constructed in 2008 and 2019, respectively; (b) performance of MMM with reference to filler enhancement index (F_index) calculated for each MMM. The performances of several notable MMMs were indicated in the figure for ease of comparison.

5.2. Filler enhancement index (F_index)

Recent efforts in decoupling and downplaying the effect of the polymer matrix in MMM were computed with the calculation of filler enhancement index (F_index), as indicated in the following equation (Chuah et al. 2018): (1) $$F_{\mathrm{\text{index}}}=\mathrm{\text{ln}}\hspace{0.17em}\left(\frac{P_{\mathrm{\text{MMM}}}}{P_{\mathrm{p}}}\right)+\eta \mathrm{\text{ln}}\hspace{0.17em}\left(\frac{{\alpha }_{\mathrm{\text{MMM}}}}{{\alpha }_{\mathrm{p}}}\right)$$(1)

In this expression, P_MMM, P_p, α_MMM, and α_p are described as CO₂ permeability of MMM, CO₂ permeability of pure polymer, CO₂/N₂ selectivity of MMM, and CO₂/N₂ selectivity of pure polymer. η is described as the enhancement coefficient with the value of 3.409 to reflect the change in the slope computed from the recent upper bound plot. The reference guide that shows the range of F_index value is utilized based on the reference provided in the description (Chuah et al. 2018), based on the range of F_index value obtained across different filler materials. Based on the obtained results, it can be observed that the filler materials that showcase “exemplary” performance (4.0 ≤ F_index ≤ 8.0) are UiO-66-Br, ZIF-UC-6, and ZIF-7, with majority of the fillers falling in the range of “competent” (1.5 ≤ F_index ≤ 4.0) and “moderate” (0 ≤ F_index ≤ 1.5), as indicated in Figure 7(b). At the present stage, due to the limited data available in the literature, it is yet difficult to justify the effectiveness in enhancing the performance of filler materials, considering the difficulty of achieving a performance that is close to the ideal performance (F_index ≥ 8.0).

6. Conclusions and Future Direction

This review delves into the recent progress in MOF-based MMM for post-combustion CO₂ capture. Each configuration (2D, 3D, and composites (2D/3D and 2D/2D)) proves capable of enhancing CO₂/N₂ separation performance through distinct mechanisms. Generally, 2D MOFs improve mixed-gas selectivity by creating a tortuous path that restricts gas transport across all gases. In contrast, the incorporation of 3D MOFs enhances gas transport performance by leveraging the inherent porosity in 3D MOFs. MOF composites capitalise the synergistic effect of the two materials to optimise overall performance, emphasizing the importance of careful selection of the polymer loading.

Nevertheless, the practical application of MMMs requires further investigations. Despite the organic moieties in MOFs improves interfacial compatibility between filler and polymer (hence minimizing the potential non-idealities such as sieve-in-a-cage behaviour in zeolite-based MMMs), the agglomeration of fillers at high loading limits overall gas transport effectiveness. To mitigate this, it is often required to ensure effective dispersion of MOF solutions using sonication horn before membrane fabrication. Additionally, the addition of fillers can affect physicochemical properties, such as mechanical strength, highlighting the need for a comprehensive evaluation beyond performance indicators like the upper bound limit and F_index, which focus solely on separation capability.

Lab-scale study of MMMs in post-combustion CO₂ capture may not well-reflective the actual condition. Notably, 67% of the reported literature studies (Tables 2–4) mainly focus on pure gas permeation, with limited attention focussing on humid gas permeation. It should be noted that in Tables 2–4, most of the investigation on MMMs in humid conditions generally utilized PEBAX-based polymer, which are known to exhibit improved CO₂-based separation performance in humid conditions (Qin et al. 2023; Shen et al. 2016). This is attributed to the presence of water-facilitated CO₂ separation behaviour in the PEBAX membrane (Li et al. 2014; Wang et al. 2016). Comparatively, this is in stark contrast with the breakthrough studies that were adopted for porous materials investigating applicability under repetitive adsorption–desorption cycling processes (Chuah, Li, et al. 2020). Future research should this explore membrane performance under humid conditions more extensively to ensure overall practical feasibility.

Abbreviations
APUA	=	acrylate polyurethane/acrylate-diluent
Azo	=	azobenzene
GPU	=	gas permeation unit
NB	=	norbornene
ODA	=	4,4′-oxydianiline
ODPA	=	4,4′-oxydiphthalic anhydride
PDMS	=	polydimethylsiloxane
PEG	=	poly(ethylene glycol)
PEGDA	=	poly(ethylene glycol) diacrylate
PES	=	polyethersulfone
PGO	=	poly(glycidyl methacrylate-co-poly(oxyethylene methacrylate)
PI	=	polyimide
PIM	=	polymer of intrinsic microporosity
PMDA	=	pyromellitic dianhydride
POEM	=	poly(oxyethylene methacrylate)
PPG	=	polypropylene glycol
PSF	=	polysulfone
PU	=	polyurethane ether
TMPDA	=	2,4,6-trimethyl-m-phenylenediamine
XLPEO	=	crosslinked polyethylene oxide
ZIF	=	zeolitic imidazolate framework

Additional information

Funding

The author would like to thank the Yayasan Universiti Teknologi PETRONAS-Fundamental Research Grant (YUTP-FRG, Cost Centre: 015LC0-486) and International Collaborative Research Fund – UTP & UNILA (Cost Centre: 015ME0-360) for the funding support.