Effects of Protonation, Hydroxylamination, and Hydrazination of g-C3N4 on the Performance of Matrimid®/g-C3N4 Membranes

One of the challenges to continue improving polymeric membranes properties involves the development of novel chemically modified fillers, such as nitrogen-rich 2-D nanomaterials. Graphitic carbon nitride (g-C3N4) has attracted significant interest as a new class of these fillers. Protonation is known to afford it desirable functionalities to form unique architectures for various applications. In the work presented herein, doping of Matrimid® with protonated g-C3N4 to yield Matrimid®/g-C3N4 mixed matrix membranes was found to improve gas separation by enhancing the selectivity for CO2/CH4 by up to 36.9% at 0.5 wt % filler doping. With a view to further enhancing the contribution of g-C3N4 to the performance of the composite membrane, oxygen plasma and hydrazine monohydrate treatments were also assayed as alternatives to protonation. Hydroxylamination by oxygen plasma treatment increased the selectivity for CO2/CH4 by up to 52.2% (at 2 wt % doping) and that for O2/N2 by up to 26.3% (at 0.5 wt % doping). Hydrazination led to lower enhancements in CO2/CH4 separation, by up to 11.4%. This study suggests that chemically-modified g-C3N4 may hold promise as an additive for modifying the surface of Matrimid® and other membranes.


Introduction
CO 2 is significantly present in mixtures where CH 4 is the major and valuable component. Its removal from CH 4 mixtures is particularly important in processes such as biogas upgrading or natural gas sweetening. Conventional industrial methods used for such CO 2 removal include adsorption [1], and water scrubbing and absorption [2] processes. Nonetheless, separation processes mediated by membranes can generally offer advantages over aforementioned techniques in terms of low capital cost, ease of processing, small footprint area, high energy efficiency, and ease of preparation and control [3,4].
Another closely related burgeoning field is that of oxygen production from air separation, since it can be used in many environmental applications, such as in the combustion enhancement of natural no. 7803-57-8; 98%) were purchased from Sigma-Aldrich (Munich, Germany). Reagents were used as received, without any further purification.

Synthesis of Graphitic Carbon Nitride (g-C 3 N 4 )
Graphitic carbon nitride was prepared by a modification of the thermal oxidation method [30], using melamine cyanurate as a starting material for the polymerization/condensation reaction [31]. The reagent was placed in a Vycor ® glass vial and heated at 650 • C for 4 h under aeration, with a cooling/heating ramp of 10 • C·min −1 . The resulting yellow powder was transferred to an agate mortar and ground. Subsequently, the samples were exposed to ultrasonic treatment for approximately 2 h using a Branson (FisherScientific, Hampton, NH, USA) Sonifier 450. Samples were dried in an oven at 105 • C for 24 h to remove moisture.

Synthesis of Protonated g-C 3 N 4
Due to the presence of abundant -C-N-bonds in the g-C 3 N 4 framework, it can be easily protonated by HCl, resulting in a modification of the surface charge, from a negatively to positively charged ( Figure 1a) [32]. The process consisted in adding 2 g of pristine g-C 3 N 4 sample to a 12 mL diluted 3% solution of HCl and treating it with ultrasound for 2 h. Samples were not subsequently washed. They were placed in an oven at 105 • C for approximately 24 h and were then ground. Dimethylacetamide (DMAc; CAS no. CAS: 127-19-5; ≥99%) and hydrazine monohydrate (CAS no. 7803-57-8; 98%) were purchased from Sigma-Aldrich (Munich, Germany). Reagents were used as received, without any further purification.

Synthesis of Graphitic Carbon Nitride (g-C3N4)
Graphitic carbon nitride was prepared by a modification of the thermal oxidation method [30], using melamine cyanurate as a starting material for the polymerization/condensation reaction [31]. The reagent was placed in a Vycor ® glass vial and heated at 650 °C for 4 h under aeration, with a cooling/heating ramp of 10 °C·min −1 . The resulting yellow powder was transferred to an agate mortar and ground. Subsequently, the samples were exposed to ultrasonic treatment for approximately 2 h using a Branson (FisherScientific, Hampton, NH, USA) Sonifier 450. Samples were dried in an oven at 105 °C for 24 h to remove moisture.

Synthesis of Protonated g-C3N4
Due to the presence of abundant -C-N-bonds in the g-C3N4 framework, it can be easily protonated by HCl, resulting in a modification of the surface charge, from a negatively to positively charged ( Figure 1a) [32]. The process consisted in adding 2 g of pristine g-C3N4 sample to a 12 mL diluted 3% solution of HCl and treating it with ultrasound for 2 h. Samples were not subsequently washed. They were placed in an oven at 105 °C for approximately 24 h and were then ground. 2.2.3. Modifications of g-C3N4 with Oxygen Plasma and Hydrazine Monohydrate g-C3N4 is a laminar material with non-oxidized aromatic regions and aliphatic regions containing phenolic, carboxyl, and oxygen epoxide groups that make it hydrophilic in aqueous media. This behavior can be modified through treatment with oxygen plasma or with hydrazine monohydrate.
In the former approach, interfacial forces are improved via plasma treatment, an ecologically benign process that does not produce liquid chemical residues. When the plasma is used with oxygen gas, it dissociates in order to generate oxygen-containing radicals [33], which incorporate -OH, as 2.2.3. Modifications of g-C 3 N 4 with Oxygen Plasma and Hydrazine Monohydrate g-C 3 N 4 is a laminar material with non-oxidized aromatic regions and aliphatic regions containing phenolic, carboxyl, and oxygen epoxide groups that make it hydrophilic in aqueous media. This behavior can be modified through treatment with oxygen plasma or with hydrazine monohydrate.
In the former approach, interfacial forces are improved via plasma treatment, an ecologically benign process that does not produce liquid chemical residues. When the plasma is used with oxygen gas, it dissociates in order to generate oxygen-containing radicals [33], which incorporate -OH, as hydroxylamine groups, into the surface of g-C 3 N 4 , improving its dispersibility [34] (Figure 1b). The oxygen plasma treatment was conducted using a Harrick Plasma (Ithaca, NY, USA) PDC-002 apparatus. 150 mg of g-C 3 N 4 were placed on a Petri dish on a quartz sample holder within the plasma cavity. The operating power remained at 10.2 W, the chamber pressure was 300 Torr with an O 2 flow of 35 cm 3 ·min −1 , and the treatment time was 90 min.
In the later approach, diazanyl group modified g-C 3 N 4 -NHNH 2 (Figure 1c) was obtained according to the procedure described by Chen et al. [35]: 1 g of as-synthesized g-C 3 N 4 was mixed with 20 mL water and 4 mL hydrazine hydrate, followed by stirring at 80 • C for 40 min.

Preparation of the Matrimid/g-C 3 N 4 Mixed Matrix Membranes
The Matrimid/g-C 3 N 4 MMMs were prepared from a Matrimid commercial polymer solution doped with 0.5 wt % or 2 wt % g-C 3 N 4 loading in 5 mL of DMAc solvent at adequate concentration (6% w/v) to achieve a reasonable viscosity. Working at such g-C 3 N 4 loadings, previous work showed that measurable effects were obtained as compared with pristine Matrimid membranes, while much higher loadings led to an inhomogeneous distribution of g-C 3 N 4 across the membrane film, thus compromising the mechanical stability of the resulting membranes. After stirring until complete homogenization, all solutions were sonicated for 5 min at an amplitude of 60% in a sonicator with a work/rest period of 20 s and 10 s. Next, membranes were prepared by the casting method following the protocol described by Recio et al. [36], using a glass plate at 25 • C. The resulting films were dried at 60 • C for 24 h to complete the removal of the solvent and were then placed in a vacuum oven, where they remained at 120 • C for 4 hours before raising the temperature to 180 • C for another 4-6 h. Finally, the membranes were separated from the glass plates and their thickness was determined using a Fischer (Sindelfingen, Germany) Dualscope MP0R. All thicknesses were in the 40-60 µm range. Three samples of each formulation were prepared, and at least two pieces of each membrane were tested. Their results were averaged, with a sample to sample variability in the usual range for laboratory prepared membranes (5-10%).

Characterization Methods
Vibrational information of the dopant and the Matrimid/g-C 3 N 4 MMMs was retrieved by using a Thermo Scientific (Waltham, MA, USA) Nicolet iS50 Fourier-Transform Infrared (FTIR) spectrometer. 8-10 mm Ø samples were cut from the films and their transmittance was measured at room temperature in a scanning range between 410 and 4000 cm −1 , with a 1 cm −1 spectral resolution and 64 scans.
The morphology of the as-prepared samples was examined using field emission scanning electron microscopy (FESEM), with a FEI (Hillsboro, OR, USA) QUANTA 200F device.
Thermogravimetric analysis (TG) was used to determine the thermal and/or oxidative stability of the materials. TG experiments were performed with a Mettler Toledo (Columbus, OH, USA) DMA/SDTA 861 device, heating the samples from 50 • C to 850 • C at a heating rate of 10 • C·min −1 , under a N 2 flow of 20 cm 3 ·min −1 .

Gas Separation Performance Measurement
The pure gas permeation tests were performed on an isochoric (constant volume, variable pressure) permeation system to measure the permeability properties of the membranes. Five pure gas species (He, N 2 , O 2 , CH 4 , and CO 2 ) were applied as test gases, and the permeability and perm-selectivity of Matrimid/g-C 3 N 4 MMMs were evaluated. Experiments were carried out at 35 • C using a constant feed pressure of 3 bar. To assure constant flow of each gas tested, desired pressure was maintained during 1 h and then flow measurements were taken by using a gas flow meter.
The permeability of a given membrane was calculated from Equation (1) where P is the permeability in barrer (1 barrer = (10 −10 cm 3 (STP) × cm)/(cm 2 × s × cmHg)), Q is the volumetric flow rate [cm 3 (STP)/s], L is the membrane thickness [cm], ∆P is the pressure difference between two sides of the membrane [cmHg], and A is the effective membrane area [cm 2 ]. The perm-selectivity α A/B is defined as the ratio of the permeability coefficients of two gases A and B (P A and P B ), according to Equation (2)

ATR-FTIR Spectra of the Modified g-C 3 N 4 Fillers
The spectra of the protonated g-C 3 N 4 samples revealed similar characteristic features to those found in the as-prepared g-C 3 N 4 , thus confirming that the structural integrity of g-C 3 N 4 remained intact after protonation ( Figure 2). Strong absorption bands could be observed in the 1200-1650 cm −1 range, arising from the skeletal stretching of C-N heterocycles and comprising both trigonal (N-(C)3) (full condensation) and bridging C-NH-C units (partial condensation) [37]. This would point at a successful development of an extended C-N-C network. The broad band ranging from 3000 to 3700 cm −1 was assigned to N-H and O-H stretching, due to the free amino groups and adsorbed hydroxyl species, respectively, whereas the sharp band at ca. 805 cm −1 was originated from the breathing vibration of tri-s-triazine units [38]. where P is the permeability in barrer (1 barrer = (10 −10 cm 3 (STP) × cm)/(cm 2 × s × cmHg)), Q is the volumetric flow rate [cm 3 (STP)/s], L is the membrane thickness [cm], ΔP is the pressure difference between two sides of the membrane [cmHg], and A is the effective membrane area [cm 2 ]. The permselectivity αA/B is defined as the ratio of the permeability coefficients of two gases A and B (PA and PB), according to Equation (2)

ATR-FTIR Spectra of the Modified g-C3N4 Fillers
The spectra of the protonated g-C3N4 samples revealed similar characteristic features to those found in the as-prepared g-C3N4, thus confirming that the structural integrity of g-C3N4 remained intact after protonation ( Figure 2). Strong absorption bands could be observed in the 1200-1650 cm −1 range, arising from the skeletal stretching of C-N heterocycles and comprising both trigonal (N-(C)3) (full condensation) and bridging C-NH-C units (partial condensation) [37]. This would point at a successful development of an extended C-N-C network. The broad band ranging from 3000 to 3700 cm −1 was assigned to N-H and O-H stretching, due to the free amino groups and adsorbed hydroxyl species, respectively, whereas the sharp band at ca. 805 cm −1 was originated from the breathing vibration of tri-s-triazine units [38]. After oxygen plasma treatment, all the characteristic bands were retained ( Figure 2), featuring bands at 800 cm −1 (assigned to the tri-s-triazine ring); and at 1200, 1309, 1395, 1455, 1533, and 1627 cm −1 (related to C-NH-C and N-(C) stretching modes). The peaks located at 1010 cm −1 , 1079 cm −1 , and above 3000 cm −1 corresponded to signals from N-OH groups. These results indicated that the oxygen plasma treatment led to the incorporation of N-OH on the surface of the C3N4.
The treatment of g-C3N4 with hydrazine monohydrate ( Figure 2) appeared to decrease the signal intensity of all peaks with respect to the original untreated g-C3N4 sample or to that treated with oxygen plasma. After oxygen plasma treatment, all the characteristic bands were retained (Figure 2), featuring bands at 800 cm −1 (assigned to the tri-s-triazine ring); and at 1200, 1309, 1395, 1455, 1533, and 1627 cm −1 (related to C-NH-C and N-(C) stretching modes). The peaks located at 1010 cm −1 , 1079 cm −1 , and above 3000 cm −1 corresponded to signals from N-OH groups. These results indicated that the oxygen plasma treatment led to the incorporation of N-OH on the surface of the C 3 N 4 .
The treatment of g-C 3 N 4 with hydrazine monohydrate (Figure 2) appeared to decrease the signal intensity of all peaks with respect to the original untreated g-C 3 N 4 sample or to that treated with oxygen plasma.
In relation to the ATR-FTIR spectra of MMMs doped with oxygen plasma and hydrazine treated g-C3N4, also depicted in Figure 3, it is worth noting that no bands associated with the N-OH or -NHNH2 groups of g-C3N4 (which would result from those treatments) were observed. This may be due to low dopant doses, and would be in agreement with the findings of Chen et al. [35], who reported no remarkable differences between pristine g-C3N4 and g-C3N4-NHNH2. On the other hand, the band at 1606 cm −1 , present in the pristine matrix, was found to disappear upon doping in both cases.

SEM Analysis
SEM was used to study the MMMs cross-section morphology. The micrographs of the neat Matrimid membrane (not shown) featured a smooth surface without obvious imperfections, in good agreement with those reported, for instance, by Ebadi Amooghin et al. [39].
Representative micrographs of Matrimid/g-C3N4 MMMs with different g-C3N4 loadings and sonication times are shown in Figure 4a-c. At low loadings (0.5 wt %, Figure 4a,b) and sonication times (30 min), the g-C3N4 dopant was typically well dispersed. However, at higher loadings (e.g., 10 wt %, Figure 4c) and moderate sonication times (4 h), several large clusters of g-C3N4 could be found. Differences in the nature of the dopant, i.e., protonated vs. non-protonated g-C 3 N 4 , mainly affected the band at 1672 cm −1 (amide C=O stretching), which, after protonation, was shifted to 1663 cm −1 . This finding is important because it suggests that the carbonyl groups in Matrimid can act as electron acceptors from -NH + units from exfoliated protonated g-C 3 N 4 , resulting in a material assembled through electrostatic interactions (this matter will be discussed in detail in Section 3.4).
In relation to the ATR-FTIR spectra of MMMs doped with oxygen plasma and hydrazine treated g-C 3 N 4 , also depicted in Figure 3, it is worth noting that no bands associated with the N-OH or -NHNH 2 groups of g-C 3 N 4 (which would result from those treatments) were observed. This may be due to low dopant doses, and would be in agreement with the findings of Chen et al. [35], who reported no remarkable differences between pristine g-C 3 N 4 and g-C 3 N 4 -NHNH 2 . On the other hand, the band at 1606 cm −1 , present in the pristine matrix, was found to disappear upon doping in both cases.

SEM Analysis
SEM was used to study the MMMs cross-section morphology. The micrographs of the neat Matrimid membrane (not shown) featured a smooth surface without obvious imperfections, in good agreement with those reported, for instance, by Ebadi Amooghin et al. [39].
Representative micrographs of Matrimid/g-C 3 N 4 MMMs with different g-C 3 N 4 loadings and sonication times are shown in Figure 4a-c. At low loadings (0.5 wt %, Figure 4a,b) and sonication times (30 min), the g-C 3 N 4 dopant was typically well dispersed. However, at higher loadings (e.g., 10 wt %,  Figure 4c) and moderate sonication times (4 h), several large clusters of g-C 3 N 4 could be found. These clusters ranged in size from 20 to 50 nm. This would be in agreement with the findings of Tian et al. [23], who evaluated the improvement of gas separation performance of mixed matrix membranes composed of PIM-1 doped with different g-C 3 N 4 loads as compared to pure PIM-1 membranes. They found that, using CHCl 3 as a solvent and after sonication for 12 h, the g-C 3 N 4 dopant was well dispersed throughout the matrix at low loading levels (0.5-1 wt %), and also observed the presence of partial agglomeration for higher filler contents (>1 wt %). a-cage morphology (i.e., when there are interfacial voids larger than the penetrating molecules, so a by-pass around the particles is produced, enhancing the permeability and reducing the apparent selectivity [41]) at each loading.
SEM analysis was also conducted upon Matrimid doping with 0.5 wt % of g-C3N4 treated with oxygen plasma (Figure 4d,e) or with hydrazine monohydrate (Figure 4f). In the hydroxylaminated g-C3N4-doped samples, one could observe a crater-like pattern in which the eye of each crater was formed by nanosized fillers, similar to that obtained for protonated g-C3N4, albeit more evident. The filler also appeared to be uniformly distributed along the polymer matrix. According to Ebadi Amooghin et al. [39], such crater-like morphology of MMMs arises from an adequate compatibility existing between two phases, possibly due to the interfacial stress concentrations because of the polymer matrix/filler consecutive debonding. Thus, this deformation of the polymer matrix was again indicative of a strong interaction of polymer chains and filler. On the other hand, the MMMs doped with hydrazinolyzed g-C3N4 did not show a significant modification of the smooth aspect of pristine Matrimid, thus suggesting a weaker interaction.  It is also worth noting that, in contrast with the neat Matrimid membrane, a scalloped morphology was observed for the MMMs doped with protonated g-C 3 N 4 (Figure 4a,b). As noted by Venna et al. [40] (and references therein), it may be attributed to the formation of elongated polymer segments with increased plastic deformation of the polymer and can be regarded as an indication of a good interaction between the polymer and the filler, further supported by the absence of a sieve-in-a-cage morphology (i.e., when there are interfacial voids larger than the penetrating molecules, so a by-pass around the particles is produced, enhancing the permeability and reducing the apparent selectivity [41]) at each loading.
SEM analysis was also conducted upon Matrimid doping with 0.5 wt % of g-C 3 N 4 treated with oxygen plasma (Figure 4d,e) or with hydrazine monohydrate (Figure 4f). In the hydroxylaminated g-C 3 N 4 -doped samples, one could observe a crater-like pattern in which the eye of each crater was formed by nanosized fillers, similar to that obtained for protonated g-C 3 N 4 , albeit more evident. The filler also appeared to be uniformly distributed along the polymer matrix. According to Ebadi Amooghin et al. [39], such crater-like morphology of MMMs arises from an adequate compatibility existing between two phases, possibly due to the interfacial stress concentrations because of the polymer matrix/filler consecutive debonding. Thus, this deformation of the polymer matrix was again indicative of a strong interaction of polymer chains and filler. On the other hand, the MMMs doped with hydrazinolyzed g-C 3 N 4 did not show a significant modification of the smooth aspect of pristine Matrimid, thus suggesting a weaker interaction.

Studies on Thermal Stability of the Matrimid/g-C 3 N 4 Membranes
The thermal stability of the Matrimid/g-C 3 N 4 membranes was evaluated in comparison to pristine Matrimid ( Figure 5). The thermograms showed that all the studied membranes were stable up to 470 • C and that only a small weight loss was observed between 200 • C and 250 • C (see Table 1), which could be attributed to residual solvent evaporation (the boiling point of DMAc is 202-204 • C). The inflection point for this first stage occurred at 351 • C for pristine Matrimid, and at lower temperatures for the MMMs membranes: 295 • C for Matrimid loaded with protonated g-C 3 N 4 , 247 • C for Matrimid doped with oxygen plasm-treated g-C 3 N 4 , and 287 • C for Matrimid with hydrazine-treated g-C 3 N 4 . Above 450 • C, a decomposition in two stages took place, with inflection points at around 517 • C and 591 • C (for pristine Matrimid they appeared at 518 • C and 580 • C, respectively). This second weight loss could be ascribed to the evolution of CO, CO 2 , and CH 4 from the cleavage of the benzene ring pattern of the Matrimid matrix. Degradation of the Matrimid/g-C 3 N 4 composites ended at 850 • C after a brief carbonization process.

Studies on Thermal Stability of the Matrimid/g-C3N4 Membranes
The thermal stability of the Matrimid/g-C3N4 membranes was evaluated in comparison to pristine Matrimid ( Figure 5). The thermograms showed that all the studied membranes were stable up to 470 °C and that only a small weight loss was observed between 200 °C and 250 °C (see Table 1), which could be attributed to residual solvent evaporation (the boiling point of DMAc is 202-204 °C ). The inflection point for this first stage occurred at 351 °C for pristine Matrimid, and at lower temperatures for the MMMs membranes: 295 °C for Matrimid loaded with protonated g-C3N4, 247 °C for Matrimid doped with oxygen plasm-treated g-C3N4, and 287 °C for Matrimid with hydrazinetreated g-C3N4. Above 450 °C , a decomposition in two stages took place, with inflection points at around 517 °C and 591 °C (for pristine Matrimid they appeared at 518 °C and 580 °C , respectively). This second weight loss could be ascribed to the evolution of CO, CO2, and CH4 from the cleavage of the benzene ring pattern of the Matrimid matrix. Degradation of the Matrimid/g-C3N4 composites ended at 850 °C after a brief carbonization process.  Taking in account the delayed TG inflection points for MMMs vs. neat Matrimid at around 591 °C (588 °C for Matrimid/hydrazinated g-C3N4; 589 °C for Matrimid/hydroxylaminated g-C3N4 and 593 °C for Matrimid/protonated g-C3N4 vs. 580 °C for pristine Matrimid, as shown in Figure 5), it may be inferred that thermally stable structures were created in the membrane matrix after treatment with either g-C3N4(N-OH) or g-C3N4(N-H + ).
Doping of the Matrimid MMMs with protonated g-C3N particulates at 0.5 wt % increased the CO2/CH4 selectivity by 36.9% (from 36.3 to 49.6) as compared to the neat Matrimid membrane, Figure 5. TG (solid line, y-axis on the left side of the graph) and DTG (dashed line, y-axis on the right side of the graph) curves for neat Matrimid and Matrimid-based membranes doped with 0.5 wt % of protonated g-C 3 N 4 , hydroxylaminated g-C 3 N 4 , and hydrazinated g-C 3 N 4 . The inset shows the TG/DTG curves for the untreated g-C 3 N 4 filler. Table 1. TG curves features for neat Matrimid, the g-C 3 N 4 filler and the three MMMs under study  Taking in account the delayed TG inflection points for MMMs vs. neat Matrimid at around 591 • C (588 • C for Matrimid/hydrazinated g-C 3 N 4 ; 589 • C for Matrimid/hydroxylaminated g-C 3 N 4 and 593 • C for Matrimid/protonated g-C 3 N 4 vs. 580 • C for pristine Matrimid, as shown in Figure 5), it may be inferred that thermally stable structures were created in the membrane matrix after treatment with either g-C 3 N 4 (N-OH) or g-C 3 N 4 (N-H + ).
Doping of the Matrimid MMMs with protonated g-C 3 N particulates at 0.5 wt % increased the CO 2 /CH 4 selectivity by 36.9% (from 36.3 to 49.6) as compared to the neat Matrimid membrane, although the O 2 /N 2 selectivity decreased by 5.7% (from 7.0 to 6.6). Doping at 2 wt % led to a lower increase in the CO 2 /CH 4 selectivity (by 16.5%, from 36.3 to 42.2), but the O 2 /N 2 selectivity was preserved ( Table 2). The behavior for the MMMs doped with oxygen plasma-treated g-C 3 N 4 treatment as a function of the filler doping values was the opposite: doping at 2 wt % led to a larger increase in the CO 2 /CH 4 selectivity (by 52.2%, from 36.3 to 55.2) than doping at 0.5 wt % (4.4% increase). On the other hand, doping at 0.5 wt % led to a significant enhancement of O 2 /N 2 selectivity (26.3% increase, from 7.0 to 8.9), while the increase for 2 wt % doping was only 2.7%.
In relation to the hydrazine treatment, it enhanced CO 2 /CH 4 selectivities (by 7.5% and by 11.4% for 0.5 wt % and 2 wt % doping, respectively), while O 2 /N 2 separation values remained similar to those of the Matrimid matrix (0.7% and −3.7% variation).
The results in the two latter treatments would be in agreement with observations by other authors, such as Tian et al. [23], who synthesized MMMs with a PIM matrix doped with g-C 3 N 4 , and an increase in the permeability values was observed as the filler concentration was increased to reach >1 wt %.
The enhancement of gas separation in the Matrimid membranes after doping with modified g-C 3 N 4 may be referred either to an increase of hydrogen bonding in the Matrimid/g-C 3 N 4 composite structure or to cross-linking modifications in Matrimid mediated by g-C 3 N 4 .
The first hypothesis suggests that the good results obtained for Matrimid MMMs doped with protonated g-C 3 N 4 would be referred to, on the one hand, their exfoliated nature and the associated surface gain for interaction, and, on the other hand, to the presence of -NH + groups capable of bonding, through hydrogen bonding, with the carbonyl groups of the Matrimid matrix. The benzene rings of both g-C 3 N 4 and Matrimid would lend themselves to the formation of "π-staking" (Figure 6a).
In the case of g-C 3 N 4 treated with oxygen plasma, designed to provide hydroxylamine N-OH groups that would bond with the carbonyl groups of Matrimid, it is unclear whether interactions between filler and matrix would be mediated by subsequent hydrogen bonding between such groups. There is also the possibility of formation of charge transfer between the -OH (behaving as electron acceptor group) and the aromatic nucleus in Matrimid (behaving as an electron donor group), as depicted in Figure 6b. In the case of Matrimid/hydrazinated g-C3N4 MMM, its behavior can be associated to a potential cross-linking effect of the g-C3N4(NH-NH2) filler. Addition of H2N-R-NH2 fillers has been reported to lead to cross-linking modification of polyimides [49], increasing chain packing and inhibiting the intra-segmental and inter-segmental mobilities of the matrix, resulting in higher gas selectivity [50]. For instance, p-phenylenediamine (PPD) was reported to be a good cross-linking agent for Matrimid, showing a remarkable enhancement of selectivity for O2/N2 as compared to the unmodified membrane (from 6.2 to 10.0 due to the affinity of nitrogen-containing molecules towards oxygen) [13]. Thus, the hydrazinated g-C3N4 discussed herein could be regarded as an alternative to PPD, albeit the associated enhancement of gas selectivity for O2/N2 would be lower.

Conclusions
Modified g-C3N4 was assessed as a filler for the doping of Matrimid matrices with a view to enhancing their gas separation properties. The resulting Matrimid/g-C3N4 MMMs were characterized through ATR-FTIR vibrational analysis, SEM microscopy, thermal analysis techniques, and gas perm-selectivity assays. Due to the strong interfacial interactions among the modified g-C3N4 and the Matrimid matrix, the hybrid nanocomposite membranes featured high swelling resistance and mechanical stability. In the assays presented herein, the doping of Matrimid with modified g-C3N4 led to an enhancement of CO2/CH4 selectivity-as compared to that of the pure Matrimid membrane-in all cases. In particular, MMMs doped with 0.5 wt % protonated g-C3N4 and with 2 wt % hydroxylaminated g-C3N4 showed the best CO2/CH4 separation performances (with an increase by 36.9% and 52.2% vs. neat Matrimid, respectively). On the other hand, doping with 0.5 wt % oxygen plasma-treated g-C3N4 improved the O2/N2 selectivity by 26.3%, as compared to pristine Matrimid. Thus, the chemical modification of g-C3N4 may hold promise as an efficient pathway toward the doping of Matrimid and other membranes.  In the case of Matrimid/hydrazinated g-C 3 N 4 MMM, its behavior can be associated to a potential cross-linking effect of the g-C 3 N 4 (NH-NH 2 ) filler. Addition of H 2 N-R-NH 2 fillers has been reported to lead to cross-linking modification of polyimides [49], increasing chain packing and inhibiting the intra-segmental and inter-segmental mobilities of the matrix, resulting in higher gas selectivity [50]. For instance, p-phenylenediamine (PPD) was reported to be a good cross-linking agent for Matrimid, showing a remarkable enhancement of selectivity for O 2 /N 2 as compared to the unmodified membrane (from 6.2 to 10.0 due to the affinity of nitrogen-containing molecules towards oxygen) [13]. Thus, the hydrazinated g-C 3 N 4 discussed herein could be regarded as an alternative to PPD, albeit the associated enhancement of gas selectivity for O 2 /N 2 would be lower.

Conclusions
Modified g-C 3 N 4 was assessed as a filler for the doping of Matrimid matrices with a view to enhancing their gas separation properties. The resulting Matrimid/g-C 3 N 4 MMMs were characterized through ATR-FTIR vibrational analysis, SEM microscopy, thermal analysis techniques, and gas perm-selectivity assays. Due to the strong interfacial interactions among the modified g-C 3 N 4 and the Matrimid matrix, the hybrid nanocomposite membranes featured high swelling resistance and mechanical stability. In the assays presented herein, the doping of Matrimid with modified g-C 3 N 4 led to an enhancement of CO 2 /CH 4 selectivity-as compared to that of the pure Matrimid membrane-in all cases. In particular, MMMs doped with 0.5 wt % protonated g-C 3 N 4 and with 2 wt % hydroxylaminated g-C 3 N 4 showed the best CO 2 /CH 4 separation performances (with an increase by 36.9% and 52.2% vs. neat Matrimid, respectively). On the other hand, doping with 0.5 wt % oxygen plasma-treated g-C 3 N 4 improved the O 2 /N 2 selectivity by 26.3%, as compared to pristine Matrimid. Thus, the chemical modification of g-C 3 N 4 may hold promise as an efficient pathway toward the doping of Matrimid and other membranes.