Field of the Invention

The present invention pertains generally to the field of gas selective separation filters, in particular useful for gas mixture separation, notably in the context of carbon capture resulting from the separation of CO2 from N2, for instance from gas waste or effluents. The invention more specifically relates to filters using atom-thick graphene porous membranes.

Background of the Invention

The conventional gas separation processes such as such as cryogenic distillation, absorption, and adsorption have been quite successful for a wide range of gas separation applications, e.g., air fractionation, carbon capture, natural gas sweetening, biogas processing, olefin/paraffin separation, etc. These processes primarily rely on thermal energy, exploiting the differences in volatilities, solubilities, and binding energies of the gas molecules, and are generally attractive only for mid- to-large- scale operations attributing to their capital-intensive setups (Feron et al., 2009, Energy Procedia, 1 (1), 1067 1074). However, energy- and cost-efficient routes have become increasingly important in the light of tightening environmental regulations and increasing carbon emission taxes. Membrane-based separation processes, by the virtue of non-reliance on thermal energy can be highly efficient for a number of separations (Khalilpour et al, 2015, J. Clean. Prod., 103, 286- 300). Additionally, they are, by design, environment-friendly (no chemical is used, no waste is produced) (Sholl et al., 2016, Nature, 532 (7600), 435-437). Because of their simple design and steady-state operation, membrane processes can be conveniently retrofitted to an existing infrastructure.

High-performance membranes have been shown to reduce the capture penalty compared to other separation processes because membranes do not rely on thermal energy. For membranes, developing selective films with high CO2 permeance is highly desired to reduce the needed area and the capture penalty. The state-of-the-art membranes based on polymer thin film composites (Zhang et al., 2021, J. Memb. Sci., 638, 119696), metal-organic frameworks (Qiao et al., 2019,. Nat. Mater., 18, 163-168), stacked nanosheet (Zhou et al., 2017, Nat. Commun., 8 (1)), and layers facilitating CO2 transport (Chen et al., 2022, J. Memb. Sci., 645, 120195; Han et al., 2022, J. Membr. Sci. Lett., 2 (1), 100014; Marius et al, 2022, Science, 376 (6588), 90-94) have shown promising CO2/N2 separation performance. The various techniques which have been used to engineer a selective membrane layer for CO2 capture can be briefly summarized based on chosen material and engineered structure. In dense polymeric membranes, a thin polymer film is used as a selective layer. CO2 has a higher solubility and diffusivity in most of the polymeric film compared to N2. Diffusivity of CChis higher in the adsorbed phase because it is smaller molecule (in size) than N2. Solubility depends on dispersion and electrostatic interaction. Polymers with polar groups (e.g., ether or amine-based polymer) have a high affinity for CO2, therefore, a high CO2/N2 selectivity. This type of material has been commercialized for CO2 capture membranes, however, the performance of such films is bound by the intrinsic property of polymer. State-of-the art membranes prepared by this method yield limited permeance (1’000-3’000 GPU; 1 GPU = 3.35 x 10' ¹⁰ mol m' ² s' ¹ Pa' ¹ ) combined with CO2/N2 selectivity in the range of 20-100. Further, nanoporous polymers tend to undergo aging of the chain packing which reduces the free volume as a function of time (Tiwari et al., 2014, Polymer, 55, 5788-5800). Films made of nanoporous crystalline materials (zeolite and metal-organic frameworks or MOFs) have ordered structure with well-defined pores. Zeolites and MOFs can be synthesized with pores which favor fast diffusion of CO2 over N2. For CO2/N2 separation, these films can be ideal, however, currently, there is no easy way to synthesize a thin film yielding high CO2 permeance (Babu et al., 2019, Adv. Mater., 31, 1900855; Kim et al., 2017, J. Mater. Chem. A, 5, 11246-11254).

The films made by stacking two-dimensional nanosheets (e.g., graphene oxide nanosheets) where micrometer-sized nanosheets with individual thickness of just 1-2 nm are deposited on top of each other create a transport pathway defined by the intersheet gap. The intersheet gap can be controlled by a number of chemistries. This approach has been used to obtain high selectivity. However, CO2 permeance has remained low (100-500 GPU) because the transport path through intersheet gap is highly tortuous and long (Wang et al, 2016, Energy Environ. Sci., 9, 3107-3112).

Finally, facilitated transport membranes are being commercialized for carbon capture. The key advantage is high selectivity between CO2 and N2. This high selectivity comes from the fact that CO2 reacts strongly with mobile (or fixed) carrier in the membrane which then allows selective transport of CO2. Selectivity in the range of 100-1000 has been reported. However, the CO2 permeance is typically low (100-1000 GPU).

Therefore, while high CO2/N2 selectivity has been reported (e.g., facilitated transport membranes), high CO2 permeance (beyond 1’000 GPU) combined with high selectivity is challenging. CO2 permeance is bound by the thickness of the selective layer which is challenging to reduce below 50-100 nm while avoiding selectivity-deteriorating pinhole defects.

Two-dimensional (2D) films hosting CCh-selective nanopores can yield rapid and selective transport. Porous single-layer graphene is an ideal selective layer because the atom-thick pores in graphene maximize gas transport for given pore size. The gas transport mechanism across an atom- thick pore depends on the relative size of molecules with respect to the pore. When the pore size is large (>2 nm), molecular effusion is the primary transport pathway. Effusion yields extremely large permeance, however, at the expense of gas pair selectivity which is determined by the mass ratio. In contrast, for smaller A-scale pores, the transport rate is dependent on the energy barrier that the molecule experiences during pore translocation. The size-dependent energy barrier has been demonstrated to differentiate gas molecules (Zhao et al., 2019, Sci. Adv., 5 (1), eaavl851; Sun et al., 2021, Nat. Commun., 12 (1), 7170).

However, given the small size difference between CO2 and N2 (0.3 A), one requires a precise pore incorporation technique to achieve a narrow pore-size-distribution. This becomes challenging for high permeance membranes where a high pore density is needed. As a result, recent progress in lattice engineering of graphene has yielded limited CO2/N2 selectivity (20) with high CO2 permeance (10000 GPU).

Functionalization of graphene oxide with amine-functional groups has been investigated as a strategy for CO2 adsorption (Zhou et al., 2019, J. Memb. Sci., 573, 184-191. Zhang et al., 2019, J. Memb. Sci., 570-571, 343-354). However, neither atom-thick membranes, nor high permeance or high CO2/N2 selectivity has been reported for those materials.

Molecular simulation studies probing the interaction of CO2 with porous graphene indicate that substituting pore edge C atom with oxygen atoms through the theoretical formation of O-crown nanopores (18-crown-6 nanopores) into the graphene layer leads to CCh-philic while rejecting N2 and O2 molecules owing to the charge distribution in the vicinity of the pore which could lead to promising properties of those 18-crown-6 nanopore- containing graphene membranes for CO2 separation/capture but also fast transport rates as compared to the same virtual graphene membrane with all carbon atoms (no oxygen termination) or graphene nanopores with nitrogen terminations or graphene nanopores with larger oxygen terminations (Luan et al, 2022, ACSNano, 16 (4), 6274 6281).

Further, beyond the theoretical obtaining of crown nanopores in graphene, the experimental obtaining of those crown nanopores suitable for CO2 capture and filtration is rather challenging for the following reasons: there is no known method to incorporate crown nanopores in graphene.

Therefore, in spite of the fact that the development of membranes with high CO2 permeance is extremely attractive and desired for carbon capture applications because the required membrane area, and therefore, a large fraction of capital cost can be cut down, the development of those membranes that has been hampered so far by the above-described technical limitations. Summary of the Invention

A general object of this invention is to provide an efficient gas selective filter using a graphene membrane for gas separation (e.g., CO2/N2 separation).

One of the specific objects of this invention is to provide an efficient gas selective filter for CO2 capture.

It is advantageous to provide a gas selective filter, having a high selectivity for similar sized molecular pairs such as CO2 and N2 not only for low-concentration CO2-feed but also at high concentration CO2-feeds compatible with combined heat and power (CHP) plants.

It is in particular advantageous to provide a gas selective filter, having attractive CO2/N2 selectivity of from about 20 to about 2’000, in particular higher than 20 even for low concentration CCh-feed (e.g. 0.5-1% CO ₂ ).

It is advantageous to provide a gas selective filter, having a large CO2 permeance, in particular exceeding 200, in particular exceeding 5’0000 GPU combined with attractive CO2/N2 selectivity.

It is advantageous to provide a gas selective filter made of inorganic material with a rigid structure which does not densify over time, increasing the lifespan of the membrane to over 5 years.

An object of this invention is to provide a gas selective filter comprising a graphene membrane, and a method for the preparation of a gas selective filter comprising a graphene membrane, which is cost effective, has good gas selectivity, which has high performance for CO2 capture.

It is advantageous to provide a facile and scalable method for the preparation of a gas selective filter comprising a graphene membrane which allows achieving the combination of the above properties.

Objects of this invention have been achieved by providing a gas selective separation filter according to claim 10 and uses thereof according to claim 16.

Disclosed herein is a method for the preparation of a gas selective separation filter comprising the steps of: a) Providing a supported nanoporous single-layer graphene membrane comprising nanopores with a pore density from about 10 ¹⁰ cm’ ² to about 10 ¹³ cm’ ² functionalized with O-functional groups with a molar O/C ratio from about 0.01 to about 0.5, said single-layer graphene membrane being supported on a mechanical support; b) Subjecting the said supported nanoporous single-layer graphene membrane to a N- functionalization of the graphene nanopores; c) Optionally transferring the obtained supported N-functionalized graphene membrane from the said mechanical support onto a porous support.

Also disclosed herein is a gas selective filter comprising a N-functionalized graphene membrane comprising wherein the nanopores are N-functionalized and 16 to 20 in average of pyridinic N are located on each pore edge.

Also, disclosed herein is a use of a gas selective filter comprising a graphene membrane according to the invention, for gas separation, in particular for separating N2 and/or from CO2.

In an advantageous embodiment, the gas selective filters according to the invention have CO2 permeance of the graphene membrane is from about 200 to about 50’000 GPU depending upon the porosity in graphene.

In an advantageous embodiment, the gas selective filters according to the invention have N2 permeance of the graphene membrane is from about 3 to about 2’500 GPU depending upon porosity in graphene (e.g., 100 GPU).

In an advantageous embodiment, the gas selective filters according to the invention have CO2/N2 from about 20 to about 2’000 depending upon CO2 concentration and pore size in graphene.

Also, disclosed herein is a use of a gas selective filter comprising a graphene membrane according to the invention, for gas separation, in particular for separating N2 and/or from CO2.

Other features and advantages of the invention will be apparent from the claims, detailed description, and figures.

Brief Description of the drawings

Figure 1 illustrates a method of preparation of gas selective separation filter according to the invention.

Figure 2 illustrates the chemical properties and structure of N-functionalized graphene membranes of the invention hosting pyridinic-N-substituted nanopores as characterized in Example 3. (a) Schematic illustration of the N-functional groups (pyridinic-N-substituted on the pore edges and - NH2 grafted near pore edges) and their CO2 adsorption routes on the single-layer graphene membrane, (b) XPS spectrum of N-functionalized graphene hosting -NH2, its derivatives, pyridinic N, and Py.CCE. (c) Peak shift of the adsorbed N-functionalized graphene curve and its desorbed curve (after 150°C heating), (d) the density of N-functional groups before and after 150°C heating. The heating leads to the increase of -NH2 and pyridinic N density, but the decrease of carbamate, ammonium, and Py.CCh. (e) The change of N-functional groups and HCCh' density during 150°C heating showing that the desorption follows the zwitterion reaction, (f) Schematic representation of the NH3 reaction mechanism with epoxy and semiquinone groups in the graphene lattice (adapted from Vacchi et al., 2016, Nanoscale, 8 (28), 13714 -13721; Kolle et al, 2021, Chem. Rev., 121 (13), 7280-7345).

Figure 3 shows CO2 adsorption-desorption on N-functionalized graphene phenomenon as described in Example 3. (a) Schematic illustration, (b) corresponding STM images, and (c) 3D topography STM images of the filled, empty, and re-filled pyridinic-N-substituted nanopore during desorption and adsorption. The percentage revolution of -NH2 and its derivatives (d) and pyridinic N and Py.CCh (e) during desorption and adsorption at 20 mbar of CO2 in NAP-XPS. (f) the quantitative analysis for percentage change of N-functional groups during CO2 adsorption and desorption in NAP-XPS, indicating that the quantitative change of N-functional groups is similar to the CO2 adsorption reaction. The percentage of each N-functional group was extracted from the XPS spectrum.

Figure 4 provides the characterization of N-functionalized graphene of AC-HRTEM and Raman spectrum as described in Example 3. (a) EDS mapping (HAADF, carbon, nitrogen, and oxygen) in AC-HRTEM image and (b) the EDS spectrum of N-functional graphene showing the presence of N-functional groups on the graphene surface, (c) Raman spectrum of pristine graphene and N- functionalized graphene functionalized at 20 and 80°C. (d) ED/IG versus G peak position and (e) 2D peak position versus G peak position. The results are extracted from the Raman spectrum, (f) Evolution of clusters on Os-treated graphene and N-functionalized graphene under electron beam in TEM (electron dose of 4.7 x 10 ⁵ e“ A ^-2 ), showing that N-clusters on N-functionalized graphene are more stable under the electron beam.

Figure 5 presents the gas transport and carbon capture performance properties of N-functionalized graphene membranes as described in Example 3. (a) The 20% CO2/N2 mixed gas permeation results of graphene membranes before and after 1.5 or 24 hrs N-functionalization at 20°C. The single gas permeance (b) and gas pair ideal selectivity (c) of graphene membrane before and after N- functionalization, showing that a decrease of permeance and enhanced selectivity contributed by the narrower size-sieving pores shrunk by the N-functional groups and CO2. (d) the single-gas permeation results of N-functionalized graphene at 30, 60, and 100°C, showing preferential adsorption to CO2. (e) Comparison of the CO2/N2 mixture separation performance with the state- of-the-art membranes for post-combustion capture. A performance upper bound from polymeric membranes is shown for reference (assuming a selective layer thickness of 1 pm). The target area in black refers to membranes with a separation factor higher than 20 and CO2 permeance higher than 1000 GPU. (f) The stability test of N-functionalized graphene for carbon capture, showing that the performance remains similar after 8 cycles of regeneration. All the gas was measured at 2 bar feed pressure.

Figure 6 illustrates the CO2 adsorption and gas transport properties of N-functionalized graphene of the invention in dilute CO2 atmosphere as described in Example 3. (a) The experimental Py.CCh coverage extracted from XPS results fitted with the theoretical Langmuir single-site isotherm, (b) The normalized gas permeation results of N-functionalized graphene were measured under different CO2 partial pressure. The results are fitted with the gas transport model, (c) the carbon capture performance of N-functionalized graphene was measured at 12 and 8.4 mbar of CO2 partial pressure, (d) Photograph of the centimeter-scale N-functionalized membrane on the smoothened polymer support. Red lines highlight the edges of the membrane.

Detailed description of embodiments of the invention

The expression “graphene membrane” is a graphene layer, in particular a graphene monolayer such as obtained for example by CVD. For example, a single-layer graphene membrane has a thickness in a range of about 0.34 to 1 nm. The graphene membrane according to embodiments of the invention may however also include bilayer graphene, or portions with bilayer graphene, it being understood that achieving a highly homogeneous monolayer over the surface area of the membrane may not be efficient for an industrial scale manufacturing of the membrane.

The expression “N-functionalization” refers to the incorporation of N atoms in the graphene lattice as pyridinic N. According to the present invention, N-functionalized nanopores is achieved through the reaction of the O functional groups (e.g., semiquinone (C=O), epoxy) of the nanoporous singlelayer graphene membrane with a source of nitrogen (e.g., ammonia or nitrogen plasma or small N containing molecules such as hydrazine, ethylene diamine, etc).

The expression “O functional groups”, refers to groups obtained by Os-treatment and include semiquinone, ether, and epoxy groups.

The expression “sacrificial support layer” is a suitable support (e.g. a Cu, Ni, Pt or any other metallic substrate on which single-layer graphene can be synthesized), in particular a non-porous support, for a graphene membrane that can be sacrified before or after the graphene membrane is applied to a structural (mechanical) support. The pore density according to the invention is commonly measured by high-resolution transmission electron microscopy (HRTEM), the pore size (van der Waals gap), O/C ratio and average N atoms in pore edge by a combination of HRTEM and spectroscopy (XPS). XPS is very common method in surface science to look at elemental (O, C) composition.

The expression “membrane performance” refers to the combination of the membrane gas permeance and its gas selectivity. Typically, in the field of gas separation, CO2 permeance of 1’000 GPU and CO2 /N2 selectivities of 20 or higher is considered as a good membrane performance. Further, O2/N2 selectivities of 3 or higher (Kiwon, et al., 2019, Angew. Chem. Int. Ed., 131, 16542- 16546) is considered as a good membrane performance.

According to a particular embodiment is provided a method for the preparation of a gas selective filter comprising: a) Providing a supported nanoporous single-layer graphene membrane comprising nanopores with a pore density from about IO ¹⁰ cm’ ² to about 10 ¹³ cm’ ² functionalized with O-functional groups with a molar O/C ratio from about 0.01 to about 0.5, said single-layer graphene membrane being supported on a mechanical support; b) Subjecting the said supported nanoporous single-layer graphene membrane to a N- functionalization of the graphene nanopores; c) Optionally transferring the obtained supported N-functionalized graphene membrane from the said mechanical support onto a porous mechanical support.

According to a particular embodiment, the mechanical support is a sacrificial support (e.g., annealed copper foil).

If the single-layer graphene membrane is provided on a sacrificial support, the sacrificial support is replaced by a porous mechanical-non sacrificial support either before the N-functionalization or after N-functionalization, wherein the replacement is achieved through i) the coating of the singlelayer graphene membrane on the sacrificial support with a porous support layer such as for example of polymeric layer having a high gas permeance (e.g. from about 5 x 10 ⁴ to about 6 x 10 ⁶ GPU) such as those referenced herein and then ii) the removal of the sacrificial support by wet-etching.

According to a particular embodiment, the porous mechanical support is a thin (typically from 100 to about 1000 nm) and porous polymeric film or a nanoporous carbon film, having a porosity from 10 to about 50 %. Typical porous mechanical support can be selected from a layer of poly [1 -(trimethyl silyl)propyne (PTMSP) or amorphous fluoropolymer such as TEFLON AF 2400, a nanoporous carbon (NPC) film, a carbon nanotube mesh, or combination of these.

According to another particular embodiment, the supported nanoporous single-layer graphene membrane comprising nanopores functionalized with O-functional groups can be obtained by subjecting a nanoporous single-layer graphene membrane to an O3 treatment to incorporate O functional groups on the graphene lattice. Typically, the introduction of nanopores functionalized with O-functional groups in the single-layer graphene film can be achieved by a well-controlled Os-millisecond etching process such as described in Hsu et al., 2021, ACS Nano, 15 (8), 13230- 13239). For example, first, the desired pore density and sizes (e.g. a pore density from about IO ¹⁰ cm’ ² to about 10 ¹³ cm’ ² with van der Waals gap in the pore less than 1 nm, typically from about 0.25 to about 1 nm, can be obtained by exposing a single-layer graphene in 13.5 torr of O3 at 250°C. Then, the O-functionalization can be carried out at 20°C for about 1 to 2 h to obtain O- functionalized porous single-layer graphene (typically functionalized with O-functional groups with a molar O/C ratio from about 0.01 to about 0.5, in particular from about 0.01 to about 0.2). Alternatively, a nanoporous single-layer graphene membrane comprising nanopores functionalized with O-functional groups with a molar O/C ratio from about 0.01 to about 0.5 can be obtained through the treatment of a single-layer graphene with O3 between about 25 and about 80°C for about 0.01 to about 10 h for the generation of pores and then heating above 100°C (typically from about 100°C to about 200°C) for about 1 h for the O-functionalization of the pores.

According to a particular embodiment, for achieving the above O-functionalization, the treatment of the single-layer graphene with O3 can be achieved with a O3 at 1-20% in O2 at a O3 pressure (1- 10 bar) at a functionalization temperature (0 to 80°C), and a reaction time (0.01 to 10 h).

The invention is based on the finding that the provision of at least 16-20 O-functionalized groups at the pore edge, they will react with nitrogen atoms of the nitrogen source such as NH3 to form pyridinic groups.

Referring to the figures, in particular first to Figure 1, is provided an illustration of a method for the preparation of a gas selective filter.

More specifically, the steps of the embodiment illustrated in Figure 1 comprise: al) Providing a single-layer graphene membrane on a sacrificial support

According to a particular embodiment, a single-layer graphene film can be prepared on a Cu foil by CVD to obtain a single-layer graphene membrane on a mechanical support. a2) Forming nanopores in the single-layer graphene film on a sacrificial support by O3 etching The single-layer graphene film on a sacrificial support is then subjected to a well-controlled Os- millisecond etching process to incorporate nanopores in the single-layer graphene film in a well- controlled manner.

Typically, the supported nanoporous single-layer graphene membrane is subjected to an O3/O2 mixture flow (e.g., 1 or 20 wt%, 1-1000 ml/min, from about 20 to about 80°C for about 0.01 to about 10 h).

According to a particular embodiment, the obtained nanoporous single-layer graphene film is subjected to a further treatment at high temperature in an H2 atmosphere to smoothen Cu in view of the subsequent removal of the sacrificial support.

According to a preferred embodiment, a single-layer graphene film with nanopores (pore density in the range of 10 ¹⁰ cm' ² to 10 ¹³ cm' ² with van der Waals gap in the pore less than 1 nm) is obtained. a 3) Subjecting the supported nanoporous single-layer graphene membrane to 0- functionalization of the graphene nanopores

According to a particular embodiment, the supported nanoporous single-layer graphene membrane is subjected to epoxy functionalization of the graphene lattices, for example by O3 treatment for about 0.01 to about 10 hours (e.g. from about 30 to about 60 min).

Typically, the supported nanoporous single-layer graphene membrane is subjected to an O3/O2 mixture flow (e.g., 1 or 20 wt%, 1-1000 ml/min, from about 100°C to about 200°C for about 1 h).

According to a particular aspect, a nanoporous single-layer graphene membrane comprising nanopores (pore density in the range of 10 ¹⁰ cm' ² to 10 ¹³ cm' ² with van der Waals gap in the pore less than 1 nm) functionalized with O-functional groups with a molar O/C ratio from about 0.01 to about 0.5 is obtained wherein said single-layer graphene membrane is supported on a sacrificial mechanical support. Pore density in the pore and extent of O-functionalization of the pores of nanoporous single-layer graphene membrane can be determined by STM (scanning tunneling microscope) XX and by X-ray photoelectron spectroscopy (XPS), respectively. a4) Preparing a single-layer graphene film on a porous mechanical support layer

According to a further particular embodiment, the single-layer graphene film on a sacrificial support is coated by a mechanical reinforcing layer before removing the sacrificial support. In particular, a porous polymeric layer such as those referenced herein is coated (e.g. by spin coating) on the single-layer graphene film on a sacrificial support before the removal of the sacrificial support by wet-etching. According to a particular embodiment, the sacrificial support is a Cu foil and the single-layer graphene film on the Cu foil is spin coated with a layer from about 0.2 to 1.0 pm of polyfl- (trimethylsilyl)propyne (PTMSP).

According to a particular embodiment, the spin coating of the single-layer graphene film on a sacrificial support is carried out with a solution of a gas permeable polymer at 1.25 to about 3 wt% (e.g., 1.25 wt%) in a solvent (e.g., toluene).

According to a further particular embodiment, the sacrificial support is removed from the polymer- coated single-layer graphene film on sacrificial support and the resulting the polymer-coated singlelayer graphene film is then transferred to a porous mechanical support such as porous tungsten (e.g. with a porosity in the range of 2% to 80%). b) Subjecting the supported nanoporous single-layer graphene membrane to N-functionalization of the graphene nanopores

According to a further particular embodiment, the mechanically supported nanoporous single-layer graphene membrane comprising nanopores functionalized with O-functional groups with a molar O/C ratio from about 0.01 to about 0.5 as described herein is then subjected to a gas-phase NH3 atmosphere for the incorporation of pyridinic groups at the graphene pore edges.

According to a further particular embodiment, the N-functionalization reaction in a saturated ammonia vapor.

According to a further embodiment, the N-functionalization reaction is carried out at a temperature from about 5 to about 100°C, typically from 20 to about 80°C.

Typically, the N-functionalization reaction is carried out for about 0.01 to 72 h, for example from 0.5 to about 24h.

For example, for about 24h at 20°C under about 0.8 bar and for about 1.5 h at 80°C under about 3.6 bar.

The N-functionalization of the graphene membrane proceeds with reaction between NH3 and O functional groups to achieve 16 to 20 pyridinic N on each pore edges, and amino groups in the vicinity of the pores.

According to another particular embodiment, is provided a N-functionalized nanoporous singlelayer graphene membrane obtainable by a method of the invention. According to another particular embodiment, is provided a N-functionalized nanoporous singlelayer graphene membrane wherein the nanopores are N-functionalized and 16 to 20 pyridinic N are introduced on each pore edge.

According to a particular embodiment, gas selective filters according to the invention have a poresize distribution such that pores in graphene have van der Waals gap less than 1 nm and greater than 0.25 nm.

According to another particular embodiment, is provided a N-functionalized nanoporous singlelayer graphene membrane wherein the density of N-functionalized pores is about IO ¹⁰ cm’ ² to 10 ¹³ -2 cm .

According to another particular embodiment, is provided a N-functionalized nanoporous singlelayer graphene membrane wherein pyridinic functionalization accounts from about 1 to about 80% of the total N-functionalizations of the nanopores (e.g. about 16% for a N-functionalization at 25°C and 50.3% for N-functionalization at 80°C). Typically, the extent of pyridinic functionalization can be determined by X-ray photoelectron spectroscopy (XPS).

According to another particular embodiment, is provided a N-functionalized nanoporous singlelayer graphene membrane wherein pyridinic functionalization accounts from about 10 to about 55% of the total N-functionalizations of the nanopores.

According to another particular embodiment, the N-functionalized graphene membrane can be assembled into a gas filter module after removal of the sacrificial support layer and provision of a reinforcement support by known techniques such as for example described in our previous report (Huang, et al., Nat. Commun., 2018, 9, 2632) and WO 2019/175162.

According to a particular aspect, the gas selective filters according to the invention can be advantageously used for carbon capture (e.g., CO2/N2 separation).

According to a particular aspect, the gas selective filters according to the invention have N2 permeance of about 3 to about 2500 GPU depending upon porosity in graphene (e.g., 100 GPU).

According to a particular aspect, the gas selective filters according to the invention have CO2 permeance of about 200 to about 50’000 GPU, depending upon porosity in graphene and CO2 concentration in feed, for example from 2’500 to about 20’000 GPU, in particular from 8’000 to about 18’000 GPU (e.g. 10’000 GPU). According to a particular aspect, the gas selective filters according to the invention have CO2/N2 selectivity of about 20 to about 2’000 (e.g. 100) depending upon CO2 concentration in feed and pore size in graphene.

In an advantageous embodiment, the gas selective filters according to the invention have CO2/N2 from about 100 to about 2’000 (e.g. 400) at low CO2 feed (e.g. 0.5-1% CO2).

In an advantageous embodiment, the gas selective filters according to the invention have CO2/N2 from about 20 to about 200 (e.g. 100) at high CO2 feed (e.g. 10-20% CO2).

The remarkable observed CO2 permeance as well as the CO2/N2 selectivity of the gas selective filters according to the invention would allow their use as a valuable tool for CO2 capture from efflux gases (e.g. steel and cement industries) since this combination of performance will significantly reduce the energy usage and the cost of carbon capture.

The binding of CO2 with two-dimensional pores of graphene, containing pyridinic-N-substituted nanopores and amines has been fully supported by spectroscopy and directly observed by microscopy in the examples. It is believed that the association of CO2 with the pyridinic N group at the pore edge assists competitive transport of CO2 compared to N2, yielding to this advantageous combination of highly attractive CO2 permeance as well as high CO2/N2 selectivity.

The high permeance would reduce the needed membrane area for treating a given volume of gas mixture, thereby, will reduce the capital cost of the separation process. The high CO2/N2 selectivity is crucial for the low-feed-pressure separation application such as post-combustion capture.

The invention having been described, the following examples are presented by way of illustration, and not limitation.

EXAMPLES

Example 1: General method of preparation of a graphene membrane according to the invention

A method of the invention for the preparation of a gas selective filter is illustrated on Figure 1A & B and as detailed below. a) Providing a supported nanoporous single-layer graphene membrane comprising nanopores having a pore density in the range of 10 ¹⁰ cm' ² to 10 ¹³ cm' ² with van der Waals gap in the pore less than 1 nm functionalized with O-functional groups with a O/C ratio of 0.01- 0.2, said single-layer graphene membrane being supported on a mechanical support

A single-layer graphene membrane as described above can be provided either on a sacrificial or a non-sacrificial mechanical support. Typical non-sacrificial mechanical supports include polymeric supports sur as poly [1 -(trimethyl silyl)propyne (PTMSP), PTFE (polytetrafluoroethylene), known as TEFLON™, or nanoporous carbon film, carbon nanotube mesh or combination of above.

Typical sacrificial mechanical supports include an annealed copper foil onto which the single-layer graphene membrane has been deposited for example by CVD.

If the single-layer graphene membrane as described above is provided on a sacrificial support, the sacrificial support may be replaced by a mechanical-non sacrificial support through i) the coating of the single-layer graphene membrane on the sacrificial support with a support layer such as for example of polymeric hydrophobic layer and then ii) the removal of the sacrificial support by wetetching.

The supported nanoporous single-layer graphene membrane comprising nanopores functionalized with O-functional groups can be obtained by subjecting a nanoporous single-layer graphene membrane to an O3 treatment to incorporate O functional groups on the graphene lattice. b) Subjecting the supported nanoporous single-layer graphene membrane to N- functionalization of the graphene nanopores

The above supported nanoporous single-layer graphene membrane is then subjected to the N- functionalization of the graphene nanopores through the reaction of the O functional groups (e.g. epoxy groups) of the nanoporous single-layer graphene membrane with a source of nitrogen (e.g. ammonia or nitrogen plasma or small N containing molecules such as hydrazine, ethylene diamine, etc. For example, the N-functionalization of the nanopore is conducted under the action of gaseous ammonia. Gaseous ammonia can be provided in the form of gaseous ammonia or as a saturated ammonia vapour formed in a mixture of NH3 and methanol (e.g. 7N NBL/Methanol). For scaling up, pure gaseous ammonia can be used directly used from a gaseous tank for the N- functionalization. N-functionalization of membranes through a gas phase reaction can achieved in a large batch or roll to roll process such a described in (Bae et al., 2010 Nat. Nanotechnol., 5, 574 578).

The reaction can be conducted at temperatures from about 5 to 100°C, preferably around room temperature, from about 10 to about 80°C (e.g. at about 20°C).

The reaction can be conducted at an ammonia pressure between 10 mbar to about 50 bars, typically from about 1 bar to 10 bar.

The functionalization with N species is based on the following reaction route: NH3 reacts with epoxy groups on the graphene lattice forming primary amine (-NH2) via the ring-opening chemistry (Vacchi et al., 2016, supra). -NH2 forms carbamate (-NHCOO ’) and -NH3 ⁺ (eq 1 and 2) via zwitterion reaction with CO2 and H2O (Kolle et al., 2021, supra). The incorporation of pyridinic and graphitic N takes place via the reaction between the semiquinone group at the pore edge and NH3 (Wang et al., 2016, J. Phys. Chem., 120 (10), 5673-5681). In the case of mono-vacancy defects, semiquinone groups would be replaced by an N atom bonded with neighboring three carbon atoms (Fig. 2(f)). This leads to a doped N on the graphene lattice (graphitic N) which heals the mono-vacancy defects (Bigras et al., 2020, 2D Mater. Appl., 4 (1), 42). On the other hand, the pores larger than monovacancy should yield pyridinic N. A smaller shoulder at -403.1 eV is expected to be Py.CO2 (CO2 adsorbed at the pyridinic N site).

Example 2: Method of preparation of a graphene membrane according to the invention

A method of the invention for the preparation of a gas selective filter was carried out as described below. a) Providing a single-layer graphene membrane on a sacrificial support

A single-layer graphene film is synthesized on Cu foil by chemical vapor deposition (CVD) as follows. An annealed copper foil was used to synthesize single-layer graphene (SLG) via a low- pressure CVD process (Rezaei et al, 2020, J. Memb. Sci., 612, 118406). A copper foil (Strem Chemicals Inc., 99.9% purity, 50 pm) was placed into the furnace heated to l’000°C in a 700 torr CO2 atmosphere to remove organic contaminants. Following this, CO2 was purged out, and a mixture of H2 and Ar (1: 10 ratio of H2 to Ar) was introduced into the furnace at 700 torrs. The temperature of the furnace was elevated to 1075°C and was maintained for 4h once the pressure reached stable. Finally, the furnace was cooled down to 1’000 °C at a rate of 0.1 °C min' ¹ . The resulting copper foil can be obtained. Subsequently, 24 seem of CFU and 8 seem of H2 were introduced into the furnace with the system pressure at 460 mtorr. The flow of CFU was switched off after 30 mins reaction time, and the furnace cooled down naturally. The resulting single-layer graphene film was taken out from the furnace for further treatment. b) Forming nanopores in the single-layer graphene film on a sacrificial support by O3 etching Further, pore formation in the resulting single-layer graphene film is then carried out by two step process: high temperature O3 etching to incorporate pores, and low-temperature O3 exposure for further pore expansion which is a well-controlled Ch-millisecond etching process (Hsu et al., 2021, supra) to incorporate nanopores in the single-layer graphene film in a well-controlled manner.

Then, the obtained nanoporous single-layer graphene film was subjected to a further treatment at high temperature in an H2 atmosphere to smoothen Cu in view of the subsequent removal of the sacrificial support. Briefly, the as-synthesized single-layer graphene/Cu was placed into a home-made millisecond gasification reactor setup connected to a vacuum pump. The system was continuously pumped to remove the gases. The reactor was heated to 250°C with a 20 seem H2 flow. Then, Ar flow was swapped with H2 flow to maintain an inert environment. A controlled ozone flow (9 wt% O3 on a molar basis) was introduced into the millisecond gasification reactor chamber. After delivering the desired O3 condition (maximum pressure: 150 torr and etching time: 3s), the system was cooled down to room temperature naturally with Ar flow.

A single-layer graphene film with nanopores (3. Ox 10 ¹² cm’ ² with a van der Waals gap in the pore less than 1 nm) was obtained. Then, Ch-treated single-layer graphene film supported on Cu was subjected to a treatment at high temperature (e.g., 500°C) for 60 min in an H2 atmosphere. c) Subjecting the supported nanoporous single-layer graphene membrane to O- functionalization of the graphene nanopores

The supported nanoporous single-layer graphene membrane comprising is subjected to an O3 treatment for epoxy O-functionalization (e.g. epoxy functionalization) of the graphene lattices.

Briefly, the obtained single-layer graphene film supported on Cu was placed into the chamber with an Ar flow to purge out the residual air. Then, the Ar flow was switched off, and the chamber was filled with O3/O2 mixture flow (9 wt%) at 20°C. After the desired reaction time (0.5 - 2 hrs), the O3 flow was swapped with an Ar flow. Then, Ch-treated graphene on copper foil was removed from the chamber.

Oxidation of graphene in a controlled manner leads to the formation of O clusters surrounding the pores where the chemical identity of O is epoxy, ether, and semiquinone groups (Li et al., 2022 JACS Au, 2, 723-730). The latter is present solely at the pore edge restricted by the maximum number of covalent bonds for C atoms. A nanoporous single-layer graphene membrane comprising nanopores with a pore density in the range of IO ¹⁰ cm’ ² to 10 ¹³ cm’ ² with van der Waals gap in the pore less than 1 nm functionalized with epoxy groups with a molar O/C ratio in the range of 0.01- 0.2 was obtained wherein said single-layer graphene membrane is supported on a sacrificial mechanical support. In the obtained nanoporous single-layer graphene membrane, the nanopores are surrounded by O-functional groups which react with NH3 to form pyridinic and -NH2 groups around the pores as illustrated under Figure 2(f). d) Removing the sacrificial support to prepare a single-layer graphene film on a porous mechanical support layer

The single-layer graphene film on a sacrificial support was coated by a mechanical reinforcing layer and subjected to wet-etching to remove the sacrificial support. In short, graphene resting on the Cu foil was spin-coated (1’000 rpm for 30s, and then 2’000 rpm for 30s) with a gas permeable polymer (e.g. 1.25 wt% poly[l-(trimethylsilyl)propyne (PTMSP) solution in toluene). The resulting PTMSP film (-250 nm thick) was dried at room temperature for 12 h in the fume hood and a vacuum oven for another 12 h. Following this, the resulting PTMSP coated single-layer graphene film resting on Cu was floated on 1 M FeCh aqueous etching solution for 30 mins to remove the Cu. This leaves a floating graphene/PTMSP film on water which is then transferred onto a porous support as follows:

The resulting PTMSP coated single-layer graphene film without its sacrificial support was further transferred in a 1 M HC1 bath and a deionized water bath for 1 hr for each. The washed film was scooped up by porous tungsten support (hosting 5 pm laser-drilled pores) as earlier described (Huang et al., 2018, Nat. Commun., 9 (1), 2632). The resulting supported single-layer graphene membrane was dried on the tungsten support in the fume hood for at least 12 h before measurement. The O-functionalized nanoporous single-layer graphene membrane was obtained wherein said single-layer graphene membrane is supported on a non-sacrificial porous mechanical support. e) Subjecting the supported nanoporous single-layer graphene membrane to N-functionalization of the graphene nanopores

The obtained mechanically supported (e.g., porous tungsten support) nanoporous single-layer graphene membrane comprising nanopores functionalized with epoxy groups is subjected to a gasphase NFF functionalization reaction by placing the supported graphene membrane into a conical flask containing ammonia/methanol solution (7 N) and before sealing the flask, the system was pumped to remove the air inside the flask to ensure that the NH3 partial pressure reaches the desired concentration.

The saturated ammonia vapor reacted with the supported nanoporous graphene at 20°C or 80°C with desired reaction time, for example for about 24 h at 20°C under about 0.8 bar and for about 1.5 h at 80°C under about 3.6 bar. The N-functionalization of the graphene membrane proceeds with a progressive replacement of the oxygen atoms by a nitrogen atom at the edge of the nanopores. In this step, pyridinic groups are incorporated at the pore edges. In addition, -NH2 groups are also incorporated.

After the reaction time, the supported N-functionalized graphene membrane was placed into the vacuum for 12 hrs before heating up to 150°C in the gas permeation setup to remove the residual solvent. The resulting membrane was cooled down to the desired temperature for measurement. The resulting membranes can be then then loaded in membrane module for gas permeation studies as described in Example 2 wherein the membrane module for the membranes is composed of a quatre-inch Swagelok VCR fitting where a leak-tight metal-to-metal seal was achieved.

Example 3: Characterization of the obtained N-functionalized graphene membrane

The obtained N-functionalized graphene membrane was characterized by the following techniques: N-functionalized graphene samples were measured by X-ray photoelectron spectroscopy (XPS) to confirm the presence of the N-functional groups and the analysis was conducted on used O- functionalized single-layer graphene as a reference. After confirming the absence of N-functional groups on the graphene surface, O-functionalized single-layer graphene was exposed in NFF vapor for N-functionalization. The presence of N-functional groups on N-functionalized single-layer graphene was confirmed by XPS. and on functionalized membranes obtained in a method of the invention, i.e., O-functionalized graphene (step a4) or N-functionalized graphene membranes (step b) deposited on the tungsten foil using the monochromated Ka line of an aluminum X-ray source (1486.6 eV) with the analyzer set at pass energy of 20 eV.

The single-layer graphene membranes were prepared as described in Example 1 (O-functionalized graphene steps a to a4 and N-functionalized graphene steps a to b from Figure 1) and transferred as membrane fabrication (Cu was removed by etching in the FeCh solution, and membranes were transferred to the porous Tungsten support). Before the XPS measurement, graphene samples membranes of the invention were rinsed in toluene (three batches of toluene solution for 2 hrs each) and acetone solvent to remove the PTMSP reinforcement layer. The sample was then heated in Ar at 150°C before loading into the XPS chamber to remove the residual solvent. During the measurements, the samples were electrically grounded to the sample stage and measured.

Adsorption-desorption measurements were done in a Near Ambient Pressure XPS (NAP -XPS). The fresh N-functionalized graphene sample was placed into the NAP -XPS chamber, and were measured before the desorption experiment (heating to 150°C) to measure the initial spectrum. Following this, N-functionalized graphene sample was heated to 150°C under vacuum (10‘ ⁹ mbar) to desorb the CO2 and H2O. The desorbed N-functionalized graphene sample was measured for the desorption spectrum.

Then, the measurement was carried out again for the so-obtained desorbed samples. Subsequently, 20 mbar of CO2 was introduced into the NAP -XPS chamber at 30°C for CO2 adsorption with the desired time (for 30 mins), and then before each measurement, XPS chamber was evacuated, and samples were allowed to rest at 30°C for 1 h. Following this, corresponding spectra were collected. Further desorption was conducted by heating the samples inside ultra-high vacuum chamber (UHV) at desired temperatures for 1 hr. Then, the spectra were measured. All the fitting of spectra was performed via CasaXPS software.

Characterization of NHs -treated sample with X-ray photoelectron spectroscopy (XPS) revealed the emergence of a broad Nls peak (Fig. 2b). Deconvolution of a high-resolution Nls spectrum (Fig. 2b) confirmed the incorporation of pyridinic N (398.1 eV) (Li et al., 2009, J. Am. Chem. Soc., 131 (43), 15939 15944) in graphene. Graphitic N and primary amine (-NH2) were also detected at 401.7 eV (He et al., 2014, Angew. Chemie Int. Ed., 53 (36), 9503-9507), and 399.5 eV (Compton et al, 2010, Adv. Mater., 22 (8), 892-896), respectively. Functional groups formed by the sorption of atmospheric CO2 on pyridinic N and -NH2 sites were also detected. These groups are CO2- associated pyridinic N (referred hereafter as Py CO2, -403.1 eV) (He et al., 2014, supra) carbamate (-NHCOO', -400.3 eV) (Compton et al., 2010, supra) and ammonium (-NH3 ⁺ , -401.7 eV) (Navaee et al., 2015, RSC Adv., 5 (74), 59874-59880). The 400.3 eV peak can also be assigned to pyrrolic N (Tian et al., 2020, Nat. Commun., 11 (1), 388), however, a mass balance on the total density of - NH2 and its derivatives (carbamates and ammonium) during sorption and desorption indicated that the amount of pyrrolic N could be neglected since pyrrolic N involves a five-membered ring (5- MR) whereas graphene lattice has a honeycomb structure made of a six-membered ring (6-MR).

The incorporation of N in the graphene lattice was also confirmed in the energy-dispersive X-ray spectroscopy (EDS, Fig. 4a). A quantitative analysis of EDS mapping indicates that N in the lattice constitutes -9 % of the total atoms (C and N) which is in agreement with the XPS data. Gasification of the cluster by the beam was not observed, in contrast to the rapid gasification of oxygen clusters in graphene, indicating higher stability of the N-functional groups over the O-functional groups (Fig- 4(f))

Raman mapping was carried out to probe the incorporation of pyridinic N in the graphene lattice. Heteroatom doping is known to alter the Fermi level and electron carrier density of the graphene lattice and can be studied by Raman spectroscopy where with an increasing dopant density, the intensity ratio for the 2D peak to the G peak (ED/IG) decreases and the G peak position, WG, undergoes a shift (Froehlicher et al., 2015, Phys. Rev. B, 91, 205413).

Raman characterization was carried out on pristine and N-functionalized graphene membranes transferred onto the tungsten. The sample transfer and preparation are the same as the preparation for XPS samples. The measurement was carried out by either single-point data collection or mapping by using Renishaw micro-Raman spectroscope equipped with a blue laser (XL = 457 nm, EL = 2.71 eV). Analysis of the Raman data was carried out using MATLAB. For the calculation of the D, G, and 2D peak height and position, the background was substracted from the Raman data using the least-squares curve fitting tool.

Raman spectrum modeling

Functionalization and doping on graphene lattice can change the Fermi level (electron carrierdensity) (Beams et al., 2015, J. Phys. Condens. Matter, 27 (8), 083002). Decreased (increased) Fermi level led p-type (n-type) doping affects the position of G-peak and 2D-peak and the intensity of the 2D peak. Therefore, the changes in the peaks are the indicators of doping concentration dependence. Here are the principles of the Raman shift.

1. G peak shifts to the right with the doping concentration increase.

2. The intensity of the 2D peak is suppressed with the increase of doping concentration. However, G peak is insensitive to it. As a result, FD/IG decreases with a higher concentration of electron carrier density.

To model the doping concentration by the Raman spectrum, graphene was transferred to tungsten, and a blue laser (457nm) was carried out for the Raman mapping spectrum measurement. At least 50 points were collected from each sample. The G and D peaks properties, such as intensity, peak position, and FWHM (full width half maximum), were fitted with a Gaussian curve after the background of spectrums were subtracted.

The blue shift of the 2D peak indicated that Os-treated graphene and N-functionalized graphene are p-type doping (Froehlicher et al., 2015, Phys. Rev. B, 91, 205413). For the p-type doping, the G peak shift can be used for calculating the doping concentration regardless of the presence of defects on the graphene lattice ⁿ . The Fermi level of p-type doped graphene can be estimated by the eq. 1

E _F = -18Am _c - 83 (1)

Where E _F is expressed in meV for the change of Fermi level due to functionalization and Am _c , the shift of G peak, in cm’ ¹ .

The m _G of pristine graphene, Os-treated graphene, and N-functionalized graphene (20 and 80 °C) are 1584, 1588, 1588, and 1592 respectively. The obtained E _F are -155.6, -155.6, and -226.0 meV for Os-treated graphene, and N-functionalized graphene (20 and 80 °C). The electron-carrier density (m' ³ ) was converted from E _F (Table S3) by eq 2.

Where n is electron-carrier density (cm’ ³ ), v _F is fermi velocity (10 ⁶ m/s), h is Planck’s constant.

As the thickness of graphene is 0.35 nm, the electron-carrier density of graphene (cm’ ² ) can be obtained. XPS measurement showed that N-functionalized graphene treated at 20°C and 80°C host 9 % and 20 % of N. We note that each graphitic N-doped contributes to -0.42 mobile carriers to the graphene lattice. Assuming that N-functional groups should contribute a similar value to mobile carriers, N-functionalized graphene at 20°C (80°C) should host 1.2X 10 ¹⁴ (2.2X 10 ¹⁴ ) cm’ ² of mobile carriers. Also, we also noticed that the presence of O-functional groups on the graphene lattice also contributed to the electron-carrier density. As a result, the results from the Raman spectrum are in a similar order to the value calculated from the XPS results.

I2D/IG ratio decreased from over 2 (pristine lattice) to less than 1 for the NHs -treated graphene (Fig. 5d) and this was concomitant with a blue shift in WG. The shift increased when the concentration of pyridinic N was raised, i.e., by increasing NH3 reaction temperature to 80°C. These observations confirm that pyridinic N led to the /^-doping of graphene lattice. Based on the dependence of Fermi level on WG, an electron carrier density of 2.9X 10 ¹⁴ and 5.3X 10 ¹⁴ cm’ ² was estimated for NH3 treatment at 20 and 80°C. The results from Raman are in a similar order to the value as XPS results.

An aberration-corrected high resolution transmission electron microscope (AC-HRTEM) analysis was performed in a double-corrected Titan Themis 60-300 (FEI) equipped with a Wein-type monochromator. The single-layer graphene specimens were prepared on silicon nitride or gold TEM grids hosted in 0.6 - 1.2 pm laser-drilled holes. The graphene was transferred to the microscope grid by using paraffin as a support layer. Following this, paraffin was removed from the grid by dissolving in heptane. The grid was annealed at 900°C in a CO2 atmosphere to remove hydrocarbon contaminants from the atmosphere. Following this, the single-layer graphene on TEM grid was treated with O3 at 66 °C for 10 mins. Following this, heating at 200°C Ar was carried out to create pores on the graphene lattice. The O-functionalization was conducted on the graphene to introduce O-functional groups. Then, N-functionalization was carried out to incorporate N- functional groups on the pore edges.

An 80 keV incident electron beam was used to inhibit the knock-on damage during acquisition. The high-pass and Gaussian filters were applied to AC-HRTEM images to reduce noise and improve contrast. The composition was obtained using energy-dispersive spectroscopy (EDS) at 80 kV.

The Scanning Tunneling Microscopy (STM) imaging experiment was conducted in a low- temperature STM (CreaTec Fischer & Co. GmbH). To avoid the oxidation of Cu during the ammonia treatment, HOPG was used instead of using graphene on Cu. The O3 and NH3 treatment is the same as above for graphene treatment. Before loading into the STM chamber, HOPG was heated to 150°C in an Ar environment to remove the surface contamination. The imaging condition was at 77 and 4.2 K. Prior to the imaging, the STM probe was prepared by cutting a commercial Pt/Ir wire (Pt: 90 wt % and diameter of 0.25 mm; Alfa Aesar). The following STM image analysis was used in the WSXM software (Horcas et al., 2007, Rev. Sci. Instrum. 2007, 78 (1), 13705').

FEI Teneo (field emission 0.2-30 keV) scanning electron microscope, operating with an acceleration voltage of 1 - 5 kV and a working distance of 2.5 - 4.0 mm, was used for observing the cross-sectional morphologies of the membrane. Three samples were imaged for the average and standard deviation of the film thickness.

Analysis of CO2 adsorption and desorption on N-functionalized graphene

Interestingly, the adsorption of CO2 on the pores could be visualized by imaging by LTSTM (Fig. 3b). For this, highly-oriented pyrolytic graphite (HOPG) substrate is used for the Os-treatment to avoid the rough surface led by oxidation. Then, N-functionalization was carried out on Os-treated graphene for obtaining N-functionalized HOPG. The resulting pyridinic-N-substituted pores were incorporated in a highly-oriented pyrolytic graphite (HOPG) substrate. The substrate was exposed to atmospheric air for 10 minutes and then was inserted in LTSTM at 4 K for imaging. Several clusters were observed as bright features in the image (Fig. 3b). These clusters were 2-3 nm in size and had a density of 3.0X 10 ¹² cm' ² , comparable to the size and density of clusters on the oxidized lattice before pyridine substitution. These clusters are essential pores surrounded by oxygen functional groups (superstructure of cyclic epoxy trimer chain), formed by the oxidation of the lattice (Li et al., 2022, supra). A portion of the epoxy groups in the clusters is expected to be converted to -NH2 by reacting with NH3. Interestingly, while a donut morphology was observed for clusters before the NH3 treatment, consistent with the existence of a pore in the center of the cluster, the donut morphology was not observed after the NH3 treatment (Fig. 3b). A likely explanation is that the pore could be occupied with CO2 which then masks the pore during imaging. To confirm this, the specimen was heated to 150°C to desorb CO2. Indeed, donut-shaped clusters hosting a cavity appeared after heating. To confirm whether this change in morphology was due to reversible CO2 desorption or due to irreversible cluster gasification, the cluster was exposed again to atmospheric air for 10 minutes, and imaged the sample again. Indeed, the resulting clusters appeared to be filled (Fig. 3b). The emptying and filling of the cluster upon heating and exposing to the atmosphere, respectively, could be easily visualized by the three-dimensional height profile (Fig. 3c). The cluster area is around 9% of the total area. As the percentage of N and O is higher than this, it is likely that the clusters are formed by stacked functional groups.

The reversible adsorption and desorption of CO2 on the pyridinic N substituted pores could also be demonstrated inside the UHV chamber of a near-ambient pressure XPS (NAP-XPS, Fig. 2d). CO2 adsorption was carried out by exposing the specimen to 20 mbar CO2 at 30°C for 30 mins. The desorption experiment was carried out by heating the specimen in UHV at 150°C for 1 h. The repeated adsorption and desorption cycles reveal that CO2 sorption was quantitatively reversible (Fig. 3d-e) No capacity loss was observed. The total number of pyridinic N and -NH2 sites were conserved. The increase (decrease) in the density of Py.CCh during adsorption (desorption) was quantitatively similar to the decrease (increase) in the density of pyridinic N (Fig. 3f). This trend was also true for -NH2 and its derivatives (-NHCOO' and -NH3 ⁺ ). Due to the absence of water in the UHV chamber, we did not observe the formation of HCOs'.

The XPS data indicated that N sites comprised 9% of the C sites in graphene which amounts to a N density of 3.8X 10 ¹⁴ cm' ² . This is consistent with the fact that the clusters occupied approximately one-tenth the graphene area (based on STM images). Further mass balance indicated that pyridinic N account for 13.3% of the total N sites, which makes pyridinic site density of 5.OX 1O ¹³ cm' ² . Given a pore (cluster) density of 3.0X 10 ¹² cm' ² , this corresponds to an average of 16 pyridinic N per pore.

The incorporation of N in the graphene lattice can be explained as following. The pyridinic and graphitic N are generated by the reaction of NH3 with the semiquinone groups at the pore edge (Fig. 2(e)). The -NH2 groups are generated by the ring-opening chemistry of NH3 with the epoxy groups.

Pyridinic N serves as a Lewis base and complexes with CO2. The -NH2 sites chemisorb CO2 forming derivatives (reactions 1 and 2, Fig. 2a). Briefly, two neighboring -NH2 groups associate with CO2 to form -NHCOO' and -NH3 ⁺ (reaction 1). In the presence of adsorbed water, HCCh' is also formed (reaction 2). Indeed, Cis XPS spectrum of the sample revealed the appearance of a shoulder at 289.0 eV which can be assigned to HCCh' (Lao et al., 2019, Angew. Chemie Int. Ed., 58 (16), 5432- -5437; Bezerra et al., 2014, Appl. Surf. Sci., 314, 314 321).

2(-NH ₂ ) + CO ₂ -NHCOO' + -NH ₃ ⁺ (1)

-NH2 + CO ₂ + H ₂ O HCO ₃ ' + -NH ₃ ⁺ (2)

CO2 associated with pyridinic N sites could be desorbed by heating to 150°C for 60 min in the ultrahigh vacuum (UHV) chamber of XPS. This was indicated by a complete loss of the Py.CO2 peak in the Nls spectrum (Fig. 2). The gain in the density of pyridinic N site was quantitatively consistent with the loss of Py.CO2 (Fig. 2e). Heating also led to a decrease in the intensity of - NHCOO' and -NH3 ⁺ peaks in Nls and HCOs' peak in Cis spectra. Again, the gain in the density of the -NH2 sites was quantitatively consistent with the loss of -NHCOO' and -NH3 ⁺ sites. These observations point to complete desorption of CO2 from pyridinic N sites at 150°C which also accompanies a partial recovery of chemisorbed CO2 at the -NH2 sites. This observation is consistent with physisorption and chemisorption with pyridinic N, and -NH2 sites, respectively.

Desorption for removing CO2 and H2O

Following the desorption, the intensities of -NHCOO', -NHC/graphitic N, and Py.CCh were analyzed. CO2 were reduced from 27.3, 46.5, and 4.9 to 18.8, 35.7, and 0.0%, respectively; meanwhile, the peak of -NH2 and pyridinic N showed a reverse trend (Fig. 2d). Also, the Nls peak shifted toward the right, indicating the alteration of chemical bonding. Those indicated that desorption removed part of CO2 and H2O from the graphene surface.

CO2 adsorption-desorption cycles

A right direction in the zwitterion reaction (reactions 1 and 2) is expected. The restoration was observed after CO2 desorption. The quantitative analysis (Fig. 2d) indicates that the amount of - NHCOO', -NHs graphitic N, and Py.CO2 were increased to 26.8, 43.6, and 7.4% after CO2 adsorption. Contrarily, the peak of -NH2 and pyridinic N decreased to 13.3 and 8.9%. Interestingly, the results showed consistency with the stoichiometry of reactions 1 and 2. Moreover, the change percentage of Py.CO2 equaled pyridinic N after CO2 adsorption. It evidently indicated that the peak at -403.1 eV is Py.CO ₂ .

N-functionalized graphene was heated to 150°C under vacuum (10‘ ⁹ mbar) for desorption. An XPS measurement was carried out after the temperature was cooled down to 30°C. The N-functionalized graphene showed a reversible trend. The percentage of -NHCOO', -NHsVgraphitic N, and Py.CO2 were decreased to 18.9, 35.5, and 0.0%. On the other hand, the peak of -NH2 and pyridinic N increased to 29 and 16.6%. The second time CO2 adsorption-desorption cycle was carried out. The results are similar to the 1 ^st cycle CO2-adsorption cycle. The reversible trend showed that N- functionalized graphene followed the reversible zwitterion reaction during adsorption-desorption. Also, it also showed the robustness of XPS fitting.

The overall chemisorbed CO2 coverage can be obtained from the total CCh-adsorbed N functional groups. The amount of -NHCOO' is 26.9%, -NHC is 43.5, and Py.CO2 is 7.4. As a result, the amount of adsorbed CO2 is 26.9% and 16.6% via reaction 1 and 2, respectively. Considering CO2 adsorption on pyridinic N is 7.4%, the total CO2 adsorption involves 50.9% of the N atoms. Given that N atom constitutes 9% of the N-functionalized lattice, and the carbon density is 3.8X 10 ¹⁵ cm' ² , a CO2 coverage of 1.7X 10 ¹⁴ molecule cm' ² is obtained. Temperature-dependent study on further desorption

N-functional groups are tunable by direct heating the N-functionalized graphene sample. In the previous section, heating N-functionalized graphene can desorb H2O and CO2 from the graphene lattice. The signals from -NHCOO' and -NH3 ⁺ gradually decreased with increasing temperature.

The desorption provides information on the amount of graphitic N and -NH3 ⁺ . As the percentage of -NH ₃ ⁺ has to be greater than or equal to the percentage of -NHCOO' due to the two zwitterion reactions, the amount of graphitic N (< 4.0) can be subtracted from the results at 150°C.

Effect of NH3 treatment and its concentration

To understand the effect of pyridinic N, NH3 treatment was directly carried-out on a nanoporous single-layer graphene membrane supported on a 250-nm-thick polyfl- (trimethylsilyl)propyne] (PTMSP) according to the invention. In this manner, the changes in the gas transport properties could be probed. As a control, NH3-treatment for standalone PTMSP film was carried out. No change in the gas transport behavior of the PTMSP film was observed.

In contrast, for a given graphene membrane, in the gas permeation experiment carried out as described below, a sharp increase in CO2/N2 permeance accompanied by a decrease in CO2 permeance was observed after the NH3 treatment. The increase in selectivity followed treatment time, with longer treatment leading to a larger improvement, consistent with increasing concentration of pyridinic groups with increasing reaction time. When gas permeation data was collected as a function of kinetic diameter, a drop in permeance for all gases (He, H2, CO2, O2, N2, and CH4) which were probed was observed. However, the drop in N2 permeance was much larger. As a result, attractive H2/N2, CO2/N2, and O2/N2 selectivities could be realized accompanying large permeances (Fig. 5b and c). A likely explanation for decreasing permeance and increasing gas pair selectivity is shrinkage of pore size, driven by chemisorbed (on -NH2 sites) or physisorbed CO2 (on pyridinic N) near the pore mouth. The measurement of gas transport as a function of permeation temperature indicated activated transport for all gases except for CO2 given the strong adsorption effect of CO2 as discussed before.

A 7 N of NH3/CH3OH solution was used for NH3 treatment. Vapor-liquid equilibrium (VLE) was used to estimate the concentration of NH3 in the vapor phase. The total pressure of NH3/CH3OH (P) and the molar ratio of NH3 in the gas phase (yi) were extracted from the curve fitting with literature data for the different treatment temperatures (20 and 80°C) (Feng et al. 199, J. Chem. Eng. Data, 44 (3), 401 -404; Schafer et al., 2007, J. Chem. Eng. Data, 52 (5), 1653 1659). The points of P versus molar composition of NH3 in the liquid phase (xi) were fitted with a polynomial equation with degree 5. The points of yi versus xi were fitted with a linear equation. Following this, P and yi were obtained by the known xi.

In the 20°C VLE system, the molar NH3 concentration (yi) at the gas phase is 0.94, and P is 0.089 MPa in the 7N NH3/CH3OH solution. On the other hand, yi is 0.77 and P is 0.47 MPa in the 80°C VLE system. The higher temperature (80°C) with higher pressure promoted the N doping, leading to a higher total N percentage of N-functionalized graphene (9 % and 20 % at 20°C and 80°C. Furthermore, the percentage of pyridinic N increased to 50.3% (80°C) compared to 15.9% of the sample treated at 20°C.

Gas permeation

Single component and mixed gas permeation tests were carried out in a homemade permeation module. The permeation setup was composed of the mass flow controllers (MFCs) and mass spectrometer (MS, Hiden Analytical, HPR-20). A calibration on MFCs and MS was performed within a 5% error. The membrane module for the membranes is composed of a quatre-inch Swagelok VCR fitting where a leak-tight metal-to-metal seal was achieved. For all the measurements, the pressure differences between the permeation side and feed side were maintained at 1 bar, and Ar (15 seem, 1 bar) was used as the sweep gas to carry the permeate gas to the MS. For the mixture test, 20, 50, 80% of CO2 on a molar basis in N2 and CFU mixture were used on the feed side. The saturated water vapor was introduced into the system with CO2/N2 mixture (20/80%). The typical gas permeation experiment was tracked in real-time using the MS. Then, the performance of the membrane was calculated based on the extracted steady-state results and reported.

The permeance, J, of gas i was calculated by equation.

J = Q/(A - >t) (3)

Where Q is the molar flow rate of gas i across the membrane, A is the active membrane area, and is transmembrane pressure difference for the component i. The ideal selectivity of two gases, i and I, was calculated by dividing the permeance of gas i by the permeance of gas I. For the mixture gas permeation tests, the separation factor of two gases, i and I, was calculated using equation (4).

To understand the application of pyridinic-N-substituted graphene membranes for post-combustion capture, CO2/N2 mixtures were probed over varying concentrations of CO2 (20, 50, 80 vol%) in N2 at 30°C. Attractive separation performance was obtained in all cases with CO2/N2 separation factor increasing with increased CO2 composition, likely from competitive adsorption of CO2 over N2. For feed with 20% CO2, we consistently observed a separation factor in the range of 45-61. Moreover, N-functionalized graphene with CO2 permeance near 10000 GPU combined with CO2/N2 separation factor in the range of 23 - 85 (Fig. 5e). For example, a CO2/N2 separation factor of 52 accompanying CO2 permeance of 12420 GPU could be achieved. Elevated feed temperature (60°C) also yielded attractive performance with a CO2/N2 separation factor of 22 accompanying CO2 permeance of 11800 GPU.

The membranes showed stable performance for several weeks. After 19.5 days of continuous test under 20% of CO2/N2 mixture at 30°C, the CO2 permeance of the membrane gradually decreased with an increasing CO2/N2 separation factor. However, the membrane performance could be completely restored by simple heating at 150°C for 30 mins. This indicates that the decrease in flux is likely due to contamination-led pore blockage event where contaminations could be successfully desorbed by heating at 150°C. A 20% (vol/vol) CO2/N2 mixed gas with saturated water vapor at 2 bars in the feed was used for mimicking flue gas conditions. Under these conditions, the membrane yielded CO2 permeance of 17700 GPU and CO2/N2 separation factor of 31 at 30°C. The membrane was also tested at 60°C under mixed gas with saturated water vapor. Under these conditions CO2 permeance of 10000 GPU and CO2/N2 separation factor of 25 could be obtained. The membrane displayed stable performance for several weeks (Fig. 5f). XPS measurement confirmed that N- functional groups on the graphene lattice were retained after two cycles of heating and CO2 adsorption in the in-situ XPS. Membrane stored in the lab for more than a year (445 days) displaying similar performance membrane (CO2 concentration of 20%, feed pressure = 2 bar, T = 30°C (Table 1).

Table 1

CO N

^{2 2} CO /N

Condition Permeance Permeance „ . ^{2 2}

, „ _n . . , , . Separation F actor CrrU) (CrrU)

As-synthesized 11330 430 26.0

After 445 days 14890 650 23.0

To further understand the effect of CO2 sorption on membrane performance, the saturation of pyridinic N sites as a function of CO2 concentration was performed by counting the density of Py.CCh complex using XPS. A sharp increase in the loading of pyridinic N sites at low CO2 pressure, with coverage close to 45% at CO2 partial pressure of 20 mbar was observed. A full site saturation (pyridinic N and -NH2) was observed with CO2 pressure of 1 bar. Fitting this data with a Langmuir single-site adsorption isotherm, we obtained an equilibrium constant, K _eq , of 4.4 x 10’ ⁴ Pa' ¹ (Fig. 5a). The sharp loading of pyridinic N sites at low CO2 pressure is also reflected in the increasing CO2 permeance as a function of CO2 feed pressure (Fig. 6b). CO2 permeance rose sharply when the pressure of CO2 was less than 0.2 bar. The permeance at low pressure (12 mbar) was several-fold higher than that at 1 bar. This can be explained by the gas transport model as detailed below. The number of gas molecules adsorbed on the pore edges determines the translocation rate and permeance. Based on Langmuir single-site isotherm, the number of CO2 adsorbed on pyridinic-N-substituted nanopores is the function of pressure. As a result, the pressuredependent CO2 transport model can be derived.

Calculation of the activation energy for gas transport across the nanopore

Gas transport across graphene nanopores takes place by the activated transport mechanism. By fitting the temperature-dependent gas flux with an Arrhenius relationship, the apparent activation energy, E _act-app , was extracted.

Where d ₀ is the pre-exponential factor for the Arrhenius term, C _o is the pore density, E _act-app is the sum of the activation energy (E _act ) for gas molecules to translocate the pores and the adsorption energy (E _sur ) of gas molecules on the graphene lattice. The gas permeation data at 30, 60, and 100°C of N-functionalized graphene membranes were calculated.

Gas transport model through pyridinic N-substituted nanopores.

The transportation of gas molecules across the graphene membrane follows the translocation mechanism. Gas transportation is assumed to be first-order kinetics, that is, p _ trans pore / | .

P ' ⁷

Where k _trans is the translocation coefficient for gas molecules, P is the permeance per pore, N _pore is the number of gas molecules at the pore mouth, and p is the pressure differences between feed and permeate.

The number of CO2 adsorption on pyridinic N substituted nanopores can be expressed by the Langmuir single-site isotherm. = aK _eq p

P ^ore _{1 +Keqp}

Where a is the constant coefficient, and K _eq is the equilibrium constant.

The combination of Eq.l and 2 gives the CO2 permeance as below: By fitting the pressure-dependent gas permeance with Eq. 3, the equilibrium constant can be extracted (Fig. 5b). The resulting K _eq is 5.1 x 10' ⁴ pa' ¹ .

Also, Henry’s coefficient H _por e can be expressed as:

As a result, the CO2 permeance can be derived in Eq. 5

P ak _trans H _pore (5)

Permeance is proportional to the Henry coefficient, and Henry’s coefficient increases at the low- pressure region. This leads to increasing permeance of CO2 at the low-pressure region.

The CO2 adsorption on pyridinic N was modeled by the Langmuir single-site isotherm (Fig. 6a).

The CO2 coverage as a function of pressure is as below:

^N pore > n > K _eq p a ~ °pore ~ _1+Keqp W

Where 0 _pore is the CO2 coverage percentage.

The CO2 coverage on pyridinic N was extracted from XPS data measured at different CO2 partial pressure. K _eq (4.4 x 10' ⁴ pa' ¹ ) was extracted by fitting the experimental results with Langmuir single-site isotherm.

Following this, the experimental gas permeance data fitted well with the model. The resulting K _eq is 5.1 x I O' ⁴ Pa' ¹ which is consistent with the theoretical K _eq value above. It indicates that CO2 transporting through the pyridinic-N-substituted nanopores follows the pressure-dependent gas translocation model. As a result of increasing permeance at low CO2 concentration, extremely high performance could be obtained. One membrane yielded CO2 permeance of 50000 GPU combined with CO2/N2 separation factor of 30. Other membranes prepared with a lesser degree of oxidation led to CO2/N2 separation factor reaching 360 with CO2 permeance of 2010 GPU (Fig. 6c).

The method of the invention comprising the oxidation of graphene and NH3 treatment is scalable mainly because it involves treatment of graphene with gases, O3 for porosity incorporation and NH3 for incorporation of pyridinic groups. Centimeter-scale membranes were prepared using a mild oxidation approach which also showed attractive performance (CO2 permeance of 6570 GPU and CO2/N2 selectivity over 2000, Fig. 6d) when CO2 partial pressure in the feed gas was close to 1%.

The high permeance of CO2 at low CO2 concentration is highly attractive for capture from point emission sources such as building exhaust and coal-based power plants with CO2 concentration in the range of 0.2% to 7%. This is also attractive for capture from aluminum production and natural gas combined cycle where CO2 concentration in flue gas is 1-2%, and 3-4%, respectively (Wang et al., 2020, Frontiers in Energy Research, 15, 1-24).

A techno-economic analysis on the use of membranes with extremely dilute feeds (0.5 and 1%) suggests a capture penalty of 70 and 45 $/ton, respectively. Overall, those data shows that the N-functionalized nanoporous single-layer graphene membrane of the invention and the method of production thereof are useful for achieving membranes highly efficient carbon capture. The N-functionalization takes place preferentially near and on the gassieving pore edges by reacting with O-functional groups. The presence and content of N-functional groups (>10 ¹² cm' ² ) were confirmed by XPS, EDS, and Raman spectrum with good agreement. The filled and empty pore structure driven by the reversible CO2 adsorption-desorption on pyridinic-N- substituted nanopores was visualized by LTSTM. Those CO2 adsorbed on the pore edge and N- functional groups at the pore edge leading to blocking of N2 permeation.

The post-combustion capture performance from these membranes was compared to the state-of- the-art membranes, including nanoporous graphene without functionalization, graphene modified with amine rich polymer or ionic liquid, polymer, microporous polymers, and facilitated transport membranes (Fig. 5e) and Table 2 below which presents a comparison of the CO2/N2 mixture separation performance between the N-functionalized graphene membranes of the invention and other pore-engineered membranes under conditions of a CO2 percentage in the feed of 20%.

Table 2

* Indicates ideal selectivity.

20: Merkel et al., 2010, J. Memb. Set, 359 (1), 126-139,' 21: Wang et al., 2017, Angew. Chemie Int. Ed., 56 (45), 14246-14251; 22: Karunakaran, 2017, J. Mater. Chem. A, 5 (2), 649-656,' 23: Kim et al., 2013, Science (80-. ), 342 (6154), 91-95,' 24: Zhou et al., 2017, supra,' 25: Fu et al., 2018, Nat. Commun., 9 (1), 990' 26 Kim et al., 2013, J. Memb. Sci., 428, 218-224,' 27: Pang et al., 2020, J. Memb. Sci., 612, 118443;

28: Zhang et al., 2021, supra:, 29: Guo et al., 2020, Nano Lett., 20, 7995-8000 ; 30: Marius et al., 2022, supra,' 31: Yang et al., 2020, Chem, 6 (3), 631-645,' 32: Sutrisna et al., 2018, J. Mater. Chem. A, 6 (3), 918- 931,' 33: Sutrisna et al., 2017, J. Memb. Sci., 524, 266-279,' 34: Kim et al., 2018 J. Mater. Chem. A, 6 (17), 7668-7674,' 35: Dai et al., 2012, J. Memb. Sci., 401-402, 76-82; 36: Nikolaeva et al., 2017, J. Mater. Chem. A, 5 (37), 19808-19818; 37: Qiao et al., 2019, supra,' 38: Fu et al., 2016 Energy Environ. Sci., 9 (2), 434-440,' 39 Xie et al., 2018, Energy Environ. Sci., 11 (3), 544-550; 40: Li et al., 2013, J. Memb. Sci., 436, 121-131,' 41: Scofield et al., 2016, J. Memb. Sci., 499, 191-200,' 42: Hsu et al., 2021, supra.

The membranes of the invention not only yield an attractive post-combustion capture performance but achieves excellent performance in dilute CO2 sources. The pyridinic-N-substituted pores will pave a way for several applications that separate molecules based on chemical and charge affinity, including ion-ion separation (DuChanois etal., 2022, Sci. Adv., 8 (9), eabm9436; Qian et al., 2021, J. Am. Chem. Soc., 143 (13), 5080- -5090), desalination (Zhao et al., 2021, Nat. Mater, 20 (11), 1551-1558) and nanofiltration (Ling et al, 2022, Sci. Adv, 3 (4), el601939). Further, the fact that N-functionalized graphene membranes of the invention shows a similar carbon capture performance after 445 days indicates that a typical lifespan of 5 years can be easily achieved.