Synthesis and Characterization of Covalent Triazine Frameworks Based on 4,4‘-(Phenazine-5,10-diyl)dibenzonitrile and its Application in CO2/CH4 Separation

1. Introduction

Porous materials contain interconnected pores which can have different length scales from micro- (< 2 nm), meso- (2–50) nm to macropores (>50 nm) [1,2]. Covalent triazine frameworks (CTFs) are micro-mesoporous organic polymers that are constructed from 1,3,5-triazine rings joined with linkers to ideally give two-dimensional networks with hexagonal openings (Figure 1) [3,4,5].

The nitrogen content and the porosity of the CTFs together with their thermal stability make them interesting materials for gas adsorption, storage and separation, pollutant removal, catalysis and sensing [6,7,8,9,10,11,12,13], including CO₂/N₂ and CO₂/CH₄ separation, both in the neat form [14,15,16,17,18] and as a filler for organic polymers in mixed-matrix membranes [19,20,21]. Since the inception of covalent triazine frameworks in 2008 [22], these materials were investigated for CO₂ adsorption [23,24,25,26,27,28,29,30,31,32].

One of the standard synthesis methods of CTFs is the ionothermal method, in which the nitrile monomer is heated with excess zinc chloride under vacuum or an inert atmosphere to a chosen temperature in the range between 400-900 °C [33]. In this reaction, the molten zinc chloride salt acts as a solvent, as a Lewis acid and as a porogen [34].

The surface area is not necessarily the determining and main property of CTFs for CO₂ adsorption. A study showed an inverse relation between BET surface area and the uptake of CO₂ [17]. This correlation can be understood by increase of the surface area with synthesis temperature and the concomitant decrease of the nitrogen content. When the CTFs are synthesized under relatively low temperatures (e.g., at 350 °C), the nitrogen loss is minimized but so is the surface area. Under higher synthesis temperatures (over 500 °C), significant nitrogen loss occurs, resulting in materials that transition towards high-surface area carbon structures with some residual nitrogen content. Often 400 °C is chosen as a compromise between good surface area and not too high nitrogen loss. For a high CO₂ uptake a high nitrogen content is aimed for [22].

In this work we use the dinitrile monomer 4,4'-(phenazine-5,10-diyl) dibenzonitrile (pBN) (Figure 1) to increase the nitrogen content for CO₂ adsorption and separation from CO₂/CH₄ mixtures. The CTFs from this monomer were synthesized with two ratios of ZnCl₂ to observe the impact of the salt ratio on the surface area and two temperatures to compare the effect of the temperature on the CO₂ and CH₄ adsorption.

2. Results and Discussion

New covalent triazine frameworks with the monomer 4,4'-(phenazine-5,10-diyl)dibenzonitrile (pBN) were synthesized by the ionothermal route with molten zinc chloride at two molar ZnCl₂:monomer ratios of 10:1 and 20:1 at three temperatures of 350 °C, 400 °C and 550 °C, all at the reaction time of 48 h. The CTF samples are coded as pBN-CTF-xx-yyy by giving the molar ZnCl₂:monomer ratio (xx = 10 or 20), followed by the reaction temperature (yyy = 350, 400 or 550 °C). The pBN-CTF products were obtained as black monoliths, as typically observed for CTFs [35]. The ZnCl₂ was tried to be removed through washing with acidified water as described in the literature for CTFs [35,36]. Product yields ranged from 68-92% (Table 1). Scanning electron microscopy images indicate the typical shard-like morphology of CTFs (Figure S1a-d, Supplementary Material).

In the infrared spectra the characteristic C-N stretching band of the triazine units was observed at 1384 and 1508 cm^-1 slightly shifted from the C-N breathing and stretching mode of a molecular triazine unit (1363 and 1511 cm^–1, respectively), in agreement with the infrared spectra of other CTFs [37,38]. At the same time the CN band of the monomer at 2227 cm^–1 has disappeared (Figure S2, Supplementary Material), signaling that the monomer was consumed during the polymerization.

The CHN elemental combustion analysis reveals the typical nitrogen loss which increases with higher temperature (Table S1). The nitrogen loss is due to a partial aromatic nitrile decomposition into HCN, CN radicals, NH₃ and other species from the synthesis at the temperature of several hundred °C [3,5,36,40,41,42,43,44,45,46]. It can be seen that the C/N ratio is increased while the C/H ratio does not change much and stays close to the theoretical ratio, thereby indicating the primary loss of nitrogen-rich species. These results correlate with the general observation that an increase in temperature in ionothermal CTF synthesis leads to enhanced carbonization of the samples [5,12,40,43]. To avoid the nitrogen loss, we have also performed the pBN-CTF formation at 350 °C. It became evident, however, that the surface area, porosity and gas sorption of the samples at 350 °C varied greatly from batch to batch and among different probes from the same batch (Section S4, Figure S5, Figure S6, Table S2). Thus, samples at 350 °C were inhomogeneous and could not be reproducibly synthesized. Therefore, results from the reaction temperature at 350 °C were not included in the discussion here in the main text.

The remaining difference of the combined weight percentage of C, H and N to 100% amounts to ~20% and is usually explained by residual ZnCl₂ or by the adsorption of water upon sample handling. Energy-dispersive X-ray spectroscopy (EDX) gave a consistent amount of both Zn (~4.5 wt%) and Cl (7-10 wt%, Table S1). The ZnCl₂ which is embedded in the pores of the CTFs is difficult to remove even by extended washing, as some of the pores may no longer be accessible. The still remaining difference of ~5 wt% was shown to be due to the adsorption of moisture from air in the porous CTF. We have recently verified that CTFs are hygroscopic, with a water uptake of up to 0.12–0.20 g g^–1 (equivalent to 11–17 wt%) at 50–60% air humidity (that is P/P₀ ≈ 0.5–0.6) when handled or stored under ambient air [15,43,47].

Powder X-ray diffractograms (PXRDs) yield only broad reflexes without any clear signature of (001) reflections for parallel two-dimensional sheets in eclipsed stacking (Figure S3), which indicates a very amorphous structure because of defects in the idealized hexagonal sheets with possibly partial interpenetration or three-dimensional framework arrangements.

The nitrogen sorption isotherms of the CTFs in Figure 2 show all a pronounced adsorption step at P/P₀ < 0.05 corresponding to gas sorption in micropores (pores < 2nm, Figure 3, see Figure S5a for the 350 °C samples). The adsorption isotherms at 400 °C are largely of Type Ib, indicative of materials with micropore size distributions over a broader range and narrow mesopores (pores > 2 nm, Figure 3a,b, Table 2) [1]. There is a H4 hysteresis, where the hysteresis loop closes only at very low relative pressure P/P₀. Such H4 loops are found among others with micro-mesoporous carbons [1,2]. For 550 °C, the adsorption isotherm of the 10-550 sample appears to be a Type I and IV combination. The adsorption branch has a "knee" at P/P₀ ~0.4 and the saturation plateau, which is a typical feature of Type IV isotherms, is then reached at high P/P₀. Type IV isotherms are given by largely mesoporous adsorbents (Figure 3c,d, Table 2). The isotherm at 550 °C has a hysteresis loop of Type H2b which is associated with pore blocking in a wide range of pore neck widths. The N₂ adsorption isotherm of the sample 20-550 can be assigned as a mixture of Type I and Type II isotherms. The nitrogen uptake does not saturate towards P/P₀ = 1 which is due to a Type II branch. Type II indicates macropores (pores > 50 nm) which can also be caused by the voids between the particles. The isotherm has a H3 hysteresis loop that correlate with macropores that are not filled with pore condensate [1,2].

The specific surface areas were obtained from the Brunauer–Emmett–Teller (BET) model over the pressure range of P/P₀ ≈ 0.01-0.07. Generally, the surface areas are higher for the 10:1 than for the 20:1 ZnCl₂: monomer ratios, giving 809-1460 m² g^-1 for the former and only 348-950 m² g^-1 for the latter (Table 2).

In other CTF synthesis, e.g., with the tetra(4-cyanophenyl)ethylene monomer ZnCl₂:monomer ratios of 10:1 and 20:1 were compared, with the former giving a more than two-fold higher surface area (2235 vs. 784 m² g^–1) [48,49]. Thus, a ratio of 10:1 seems optimal for many ionothermal CTF syntheses. In the following, we will therefore only discuss the results for the 10:1 molar ratio, that is, the pBN-CTF-10 series. In agreement with other CTF works, the surface area of the sample synthesized at 400 °C is lower than that at 550 °C (Table 2, Figure 2) as generally the surface areas and total pore volumes for the resulting products increase with temperature [43]. The surface areas of the pBN-CTFs are comparable with other CTFs with longer linkers, e.g., terphenyl prepared by Kuhn et al. with a surface area of 975 m² g^-1 [22], or even 2,8-dicyano-6H,12H-5,11-methanodibenzo[1,5] diazocine that was synthesized by Wang et al. with 612 m² g^-1 [50].

Using NL-DFT calculations with a slit-pore model on the N₂ adsorption isotherms, the pore widths and distribution as well as the total and micropore volume can be estimated (Figure 3, Figure S5, Table 2) [36]. The pore width distribution diagrams for the CTF-400s indicate pronounced maxima in the micropore region (<2 nm) and pore sizes larger than 2 nm in a broad distribution. At 400 °C, the micropore volume encompasses more than 50% of the total pore volume, that is V_micro/V_tot values are above 0.50, (Table 2). At 550 °C the total pore volume more than doubles in comparison to 400 °C and pore width distributions exhibit more p broader contribution of mesopores between 2 and ~5 nm (Figure 3c) and beyond (Figure 3d), such that V_micro/V_tot drops below 35% (Table 1). Noteworthy, the material with the highest surface area pBN-CTF-10-550 has the lowest micropore volume fraction V_micro/V_tot.

The pore size distribution (PSD) from N₂ sorption at 77 K is generally limited to pores between ~1 to ~40 nm. Macropores (> 50 nm) are not accounted anymore by N₂ sorption. For pores smaller than 1 nm (10 Å), the size and distribution need to be obtained from CO₂ gas adsorption data, because for N₂ sorption at 77 K the diffusion of the molecules into micropores smaller than 1 nm is very slow, hence would require very long N₂ adsorption measurements for equilibration of the adsorption isotherms, which cannot be assured. To avoid erroneous PSD results from the N₂ adsorption analysis, CO₂ adsorption analysis can be used (Figure 4a,b). The saturation pressure of CO₂ at 10 °C is ~4480 kPa (~33450 Torr), so that a low relative pressure which is necessary for the micropore analysis is achieved in the range of moderate absolute pressures [51]. The micropore analysis with CO₂ at 283 K instead of N₂ at 77 K allows for a faster equilibration and access of even smaller pores as the kinetic diameter of CO₂ is only 3.30 Å versus 3.64 Å for N₂. The NL-DFT analysis of the CO₂ adsorption isotherms of the pBN-CTF-10s with the "CO₂ on carbon based slit-pore" model yield similar corrugated pore size distribution curves for the CTFs below 1 nm with pronounced maxima between 0.5-0.9 nm and at ~0.85 nm (Figure 2). The surface area was also calculated using the CO₂ adsorption at 195 K (values can be seen in Table 3 and isotherms in Figure S6) which gave a smaller surface area that that of the N₂ counterpart, corresponding to Literature where similar material was investigated [48,52,53,54,55].

Volumetric CO₂ and CH₄ adsorption studies resulted in the isotherms depicted in Figure 4. At 283 K and 293 K the pBN-CTF-10 materials show similar CO₂ sorption isotherm curvatures which did not level off much at 1 bar but still have rather positive slope which indicates the uptake at 1 bar is far from saturation. At 195 K the CO₂ uptake at 1 bar differentiates considerably for the pBN-CTF-10-400 and 10-550 material (Table 3, Figure S7), increasing nearly two-fold from pBN-CTF-10-400 with 175 cm³ g^–1 to pBN-CTF-10-550 with 320 cm³ g^–1. This increase correlates with the increase in surface area and pore volume from the 400 °C to the 550 °C material in Table 2.

Comparing the pBN-CTFs from this work to other CTFs with linkers equal or longer than a biphenyl unit it can be seen that the pBN-CTFs can compete very well in terms of CO₂ uptake (Figure 5, Table S4).

Covalent triazine frameworks are widely investigated for CO₂/N₂ and CO₂/CH₄ separation. The ideal adsorbed solution theory (IAST) can give an indication of the selectivity in different gas mixtures at a given pressure or for a given gas mixture at different pressures. The only criterion that IAST requires is that both gases should have an equal spreading pressure at the given temperature [56]. The IAST selectivity is derived from the single gas adsorption isotherms and was calculated here on the bases of fitting the adsorption isotherms with the Freundlich-Langmuir adsorption model and the parameters that resulted from the isotherm fitting (Table S5) were used in calculating the selectivity (Table 3).

IAST underscores the selectivity for CO₂ over CH₄ for the pBN-CTF-10-400 material, as seen already in the higher uptake of CO₂ over CH₄ at the same temperature (Table 3). At 283 K the slightly pressure and composition dependent CO₂/CH₄ selectivity for pBN-CTF-10-440 varies between 7-22. It decreases with pressure and increases with increasing CH₄ fraction (Figure S11). At 293 K the CO₂/CH₄ selectivity for pBN-CTF-10-440 stays rather constant between 0.01-0.8 CH₄ molar fraction. The preference for CO₂ can be explained from the pore structure and the interaction strength between the gas molecules and the framework. pBN-CTF-10-400 has a good microporosity and nitrogen content. Micropores favor CO₂ adsorption due to its smaller kinetic diameter (3.3 Å) compared to CH₄ (3.8 Å), while nitrogen functionalities enhance CO₂ affinity through dipole-quadrupole interactions.

From a measurement of gas adsorption at two temperatures with ∆T = 10 to 20 °C the enthalpy (∆H) or heat of gas adsorption (Q_ads = –∆H) can be obtained [13,57]. Near zero coverage the heat of adsorption for CO₂ is remarkably high in comparison to other CTFs with 79 kJ mol^–1 for pBN-CTF-10-400 (Table S6). In Figure 6 the isosteric heat of adsorption was plotted against the amount of CO₂ and CH₄ adsorbed by the frameworks. pBN-CTF-10-400 has a higher microporosity and higher nitrogen content than pBN-CTF-10-550. This relates to a higher affinity for CO₂ than for CH₄ because (as just noted) CO₂ with its smaller kinetic diameter (3.3 Å) compared to CH₄ (3.8 Å) can occupy smaller micropores, and the CO₂ quadrupole can interact with the dipole of nitrogen functionalities. Micropores generally allow for multi-site or "wall···guest···wall" interactions between guest molecules and the inner pore surface. The large decrease in the isosteric heat of adsorption from near zero adsorbed amount to ~0.5 mmol g^–1 adsorbed amount seen in Figure 6 for both gases is due to the initial filling of the very small or ultra-micropores with a diameter in the dimension of the adsorbate molecule with wall-to-wall interactions and the occupation of the nitrogen atom sites which also have higher adsorption energies. Noteworthy, the CO₂ heat of adsorption values of pBN-CTF-10-550 drop below the liquefaction heat of CO₂ of 17 kJ mol^–1 when the adsorbed amount surpasses 1.1 mmol g^–1 while the heat of adsorption of pBN-CTF-10-400 stays well above the liquefaction heat of CO₂.

Conversely, pBN-CTF-10-550 exhibits a higher Q_ads for CH₄ over the whole uptake range compared to pBN-CTF-10-400 and also a higher heat of adsorption for CH₄ than for CO₂ once the adsorbed CO₂ amount exceeded ~0.4 mmol g^–1 (compare Figure 6a and b). This can be explained by the more carbon-like non-polar nature of the pBN-CTF-10-550 material with less nitrogen content than pBN-CTF-10-400 which gives the former a relatively higher affinitiy to non-polar CH₄. As expected, the heat of adsorption near zero coverage is lower for CH₄ than for CO₂ for both CTFs (Table 3). Further literature comparison for CO₂ uptake and Q_ads⁰ for CO₂ in CTFs can be found in Table S4 and Table S6, respectively, in the Supplementary Material [15,19,26,27,59,60,61,62,63,64,65,66,67,68,69,70,71,72].

Thermogravimetric analysis showed that under air all samples started decomposing (with weight loss) at ~400 °C, including the pBN-CTF which was synthesized at 550 °C (Figure S13). This mass loss of the pBN-CTF-10-400 sample is complete below 700 °C with a residual mass of ~2.5 wt%. The mass loss of the pBN-CTF-10-550 continues to ~770 °C, leaving only ~0.2 wt%.

3. Conclusions

The molecule 4,4‘-(phenazine-5,10-diyl)dibenzonitrile with a long about 12 Å separation between the nitrile groups can be successfully transformed by ionothermal synthesis into a porous covalent triazine framework (pBN-CTF). The surface area increases expectedly with the synthesis temperature and ranges from 809 to 1460 m² g⁻¹ for the samples synthesized at 400 and 550 °C, respectively, with pores ranging between 1-4 nm. The pBN-CTF exhibited a good CO₂ uptake at 293 K, having a similar performance as benchmark materials like CTF-1, due to the relatively high micropore fraction that ranged between 35% for the 550 °C and 50% for the 400 ° samples. The new material showed significant difference and stark contrast in the adsorption of CO₂ over CH₄ for the potential separation with a selectivity that reaches 22. For further work on pBN-CTFs and other CTFs in general we plan to increase the nitrogen content of the formed framework through the addition of a nitrogen-rich compound such as melamine in order to introduce additional electron pair donors and, thereby, influence the adsorption properties. We will also check into the elongation of the phenyl group in pBN with a biphenyl group, giving the monomer 4’,4’’’-(phenazine-5,10-diyl)bis(([1,1’-biphenyl]-4-carbonitrile)).

4. Materials and Methods

4.1. Instrumentation

Fourier transform infrared spectroscopic measurements were done using a Bruker Tensor 37 (Bruker AXS, Karlsruhe, Germany) with KBr pellets in the range between 4000-500 cm^-1. For the N₂ sorption analysis a Quantachrome Autosorb-IQ-MP (Quantachrome, Boynton Beach, FL, USA) was used. The samples were degassed for 24 h at 120 °C before connecting to the device. The measurement was done at 77 K. The results were interpreted with the BET equation. The CO₂ sorption analysis was done with a Quantachrome Autosorb-IQ-MP (Quantachrome, Boynton Beach, FL, USA). The measurement temperature was 293, 283 and 195 K after activating (degassing) the samples under vacuum at 120 °C for 24 h. The temperature was held by virtue of a thermostated water bath (293 and 283 K) or with a cryodyne refrigerator model 8200 (195 K) (Janis, Woburn, Massachusetts, USA).

Thermogravimetric analysis was performed with a TG Tarsus 209 F3 (Netzsch, Selb, Germany). The samples were analyzed under synthetic air with a heating rate of 10 K/min from 25-900 °C. Powder X-ray diffraction patterns were recorded using a Bruker D2 phaser from Bruker (Bruker AXS, Karlsruhe, Germany) with Cu-Kα radiation, λ = 1.54182 Ǻ at 300 W, 30 kV, 10 mA. Nuclear magnetic resonance (¹H-NMR) spectra were collected with a Bruker Avance III-600-I (Bruker, Karlsruhe, Germany). The chemical shifts are given in ppm and are referenced to the residual proton signal of the deuterated solvent (7.26 ppm for CDCl₃, 7.16 for C₆D₆).

4.2. Chemicals

Phenazine (99.86%) and 4-bromobenzonitrile (95%) was obtained from BLDpharm. Sodium dithionite (85%) was obtained from VWR chemicals. Palladium acetate (99.9%) and tri-tert-butyl phosphine (99%) was bought from Aldrich. All solvents were purchased from commercial suppliers with a minimum purity of 99.8%.

4.3. Synthesis of 5,10-dihydrophenazine

Following the literature [58] a phenazine (2.5 g, 13.87 mmol) solution in ethanol (30 mL) and a sodium dithionite (24.1 g, 137 mmol) solution in water (125 mL) are placed into a round-bottom flask and heated to reflux at 95 °C for 3 hours- Afterwards the flask was cooled to room temperature, the product is separated by filtration, washed three times with water (3 x 15mL), dried under vacuum (10^-3 mbar) and stored under nitrogen to avoid any oxidation. Yield 2.10 g, 85%.

¹H NMR (600 MHz, CDCl₃) δ 8.27 (dd, J = 6.8, 3.5 Hz, 1H), 7.86 (dd, J = 6.8, 3.4 Hz, 1H), 6.12 (s, 2H), 1.57 (s, 3H).

4.4. Synthesis of 4,4’-(phenazine-5,10-diyl)dibenzonitrile (pBN)

5,10-Dihydrophenazine (2 g, 11 mmol), 4-bromobenzonitrile (4.38 g, 24 mmol), and potassium carbonate (9.1 g, 65.8 mmol) are combined in degassed toluene (80 mL) in a round-bottom flask under N₂ atmosphere according to the literature [56]. To this mixture, palladium acetate (0.141 g, 0.62 mmol) and tri-tert-butyl phosphine (0.464 g, 2.30 mmol) dissolved in 10 mL toluene are added, the flask is then refluxed at 111 °C for 20 h. During cooling, water (30 mL) was added to the reaction mixture in order to stop the reaction. The product was extracted from the water phase using chloroform (200 mL). The separated organic phase was washed with brine (saturated aqueous NaCl solution) three times (3 x 30 mL) and was dried over magnesium sulphate (MgSO₄) for 15 min. The organic phase was then filtered and concentrated via a rotary evaporator to ~50 mL, after that 100 mL of n-hexane was added and cooled in an ice bath for 10 min. The separated product was filtered and dried in a vacuum oven (10^-3 mbar) at 60 °C.

¹H NMR (600 MHz, C₆D₆) δ 6.96 (d, J = 8.2 Hz, 4H), 6.71 (d, J = 8.2 Hz, 4H), 6.41 – 6.37 (m, 4H), 5.69 (dt, J = 7.9, 3.9 Hz, 4H).

4.5. pBN-CTF synthesis:

Inside the glove box a glass ampule with a Schlenk fitting was filled with (0.2 g, 0.5 mmol) pBN-2CN and 10 or 20 equivalents (0.680 g, 5 mmol or 1.3 g, 10 mmol) of anhydrous zinc chloride. Outside the glovebox, the ampule was evacuated and flame sealed and heated for 48 h in a tube furnace at the chosen temperature of 350, 400 or 550 °C. The 350 and 400 °C reactions were done in a normal borosilicate glass (Pyrex), for 550 °C a quartz glass ampule was used. After cooling the ampule to room temperature, the ampule was carefully opened with no sign of pressure built-up inside.

The reaction product was stirred in distilled water acidified with 0.5 mol/L hydrochloric acid (HCl) to pH = 4 (50 mL) for three days. The stirring was done vigorously to ensure the mechanical break-up of the black monolith to give fine particles. After additional stirring for 72 h in distilled water, the product is filtered and washed with the organic solvents chloroform, acetone, and methanol (30 mL each) in this order. After the washings, the product was dried in a vacuum (10^-3 mbar) oven at 60 °C for 24 h. Yields are listed in Table 1.