Controlling the strength of interaction between carbon dioxide and nitrogen-rich carbon materials by molecular design

Max Planck Institute of Colloids and Interfa Golm, Am Mühlenberg 1, D-14476 Potsda mpikg.mpg.de Chair of Theoretical Chemistry and Center f of Paderborn, Warburger Str. 100, D-33098 Fakultät II: Mathematik & Naturwissensc Universität Berlin, Hardenbergstr. 40, 1062 Institute of Chemistry, University of Potsd Potsdam, Germany † Electronic supplementary informa 10.1039/c9se00486f Cite this: Sustainable Energy Fuels, 2019, 3, 2819


Introduction
Due to the rapid increase of the CO 2 level in the atmosphere and the problems associated with this, selective CO 2 capture from air or from point sources of emission (e.g., coal-red power plants) is intensively studied. [1][2][3][4] Chemical absorption of CO 2 in dissolved amines ("scrubbing") is the industrially established technique for CO 2 capture from power plant exhaust as it is very selective due to covalent carbamate bond formation. Chemisorption of CO 2 on amines immobilized on solid surfaces can even be applied for selective CO 2 capture from air, that is, at very low partial pressure. 2 Due to the high binding enthalpy of CO 2 , these methods suffer from high energy penalty during regeneration. [5][6][7] Physical adsorption (physisorption) of CO 2 with lower binding enthalpy (in the range of van der Waals forces) is an alternative with less energy intensive regeneration. Great research efforts have been made recently and more and more nanomaterials for selective CO 2 capture have been proposed. [8][9][10] Typical physical adsorbents applied are nanoporous materials such as zeolites or zeolitic imidazole frameworks, [11][12][13][14][15] carbon-based materials, [16][17][18][19][20][21] porous organic frameworks, 22,23 and metal-organic-frameworks (MOFs). [24][25][26][27] Common methods to increase their affinity towards CO 2 (that is, the CO 2 adsorption enthalpy) include the introduction of specic interaction sites (e.g., basic nitrogen sites/or extra framework cations in porous carbons or amine MOFs), [28][29][30][31][32][33] as well as the tuning of the ultramicroporosity of the sorbents. 18,34,35 Some thermodynamic sweet spots have been found for different families of materials, oen inspired by mimicking nature's highly selective CO 2 capture mechanism in the ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco). 36 However, as long as high CO 2 /N 2 selectivity is the target, these attempts suffer from one or more of the following problems: (i) the quadrupole moment (13.4 Â 10 À40 C m 2 ) and polarizability (26.3 Â 10 À25 cm 3 ) of CO 2 are indeed higher than for N 2 (4.7 Â 10 À40 C m 2 and 17.6 Â 10 À25 cm 3 ), but it should be kept in mind that any attempt to increase CO 2 selectivity by introducing specic binding sites will also increase the binding enthalpy of the adsorbent towards N 2 ; (ii) every porous material with a high porosity for nitrogen will always suffer from signicant contribution of less discriminative van der Waals interaction; (iii) many of the record-holders in terms of selectivity for atmospheric CO 2 capture suffer from limited stability under wet conditions, are difficult to be synthesized, and oen contain metals.
Although the promise of kinetic size exclusion of N 2 (kinetic diameter: 0.36 nm) over CO 2 (kinetic diameter: 0.33 nm) to enhance CO 2 /N 2 selectivity of physically adsorbing materials is generally accepted 34,37 and although it is obvious that CO 2 /N 2 selectivity increases signicantly at lower available N 2 surface area, 8 a real molecular CO 2 sieving adsorbent has not yet been reported. The likely reason is that a decrease of the "pore size" of the adsorbents down to the molecular dimensions of CO 2 leads to a signicant activation barrier for adsorption, or in other words, CO 2 adsorption in such narrow spaces is kinetically hindered. The only way to overcome this limitation seems to be the development of adsorbent materials with higher chemical affinity towards CO 2 . It is not surprising that materials with "ionic character" (i.e., materials that contain Coulomb charges in their structure as does Rubisco) such as metalorganic materials and zeolites have already been successfully utilized for CO 2 /N 2 separation at outstanding selectivity. [38][39][40] In contrast to such crystalline materials with dened pore size, carbon-based materials instead show narrow pores with a distribution of sizes, but their mean size can in principle be sleeplessly adjusted. For the separation of small molecules like N 2 and CO 2 , carbon materials still lack the combination of narrow porosity and high polarizability towards CO 2 . We have recently reported that simple thermal condensation of hexaazatriphenylene-hexacarbonitrile (HAT-CN) molecules results in the formation of nitrogen-rich HAT-CN-derived carbon materials with high structural microporosity and a near-perfect C 2 N-type composition. Porosity and nitrogen content can be adjusted by the synthesis temperature. The high heteroatom doping level with well-dened structure motives introduced from the pre-organized precursor and the narrow micropores lead to zeolite-like adsorption properties of these C 2 N materials. 20,41 We further utilize this concept and realize a CO 2 /N 2 molecular sieve effect by only partially condensing HAT-CN at temperatures below 550 C. Synthesis at 525 C leads to nitrogen-rich carbon materials with porosity that is still partly accessible for reversible CO 2 adsorption but the material shows no notable uptake of N 2 . To the best of our knowledge, that is the rst time, a metal-free carbon-based material with such narrow pores has been used for CO 2 molecular sieving working on a reasonable timescale.

Materials synthesis
The synthesis of hexaazatriphenylene-hexacarbonitrile (HAT-CN) was carried out according to a previously described procedure. 42 Hexaketocyclohexane octahydrate (4 g, 12.6 mmol) and diaminomaleonitrile (10.88 g, 100.8 mmol) were reuxed in acetic acid (AcOH, 150 mL) for 2 h. The black suspension was ltered off while hot and washed with hot AcOH (3 Â 25 mL) resulting in a black solid. The solid was suspended in 30% HNO 3 (60 mL) and heated at 100 C for 3 h. The hot dark brown suspension was poured into ice water (200 mL) and cooled overnight. The suspension was ltered and the solid was reuxed in MeCN (400 mL) for 2 h and was ltered. The ltrate was dried under vacuum to give an orange solid (2.4 g, yield 50%). 13  Nitrogen-rich carbons (C-HAT-CN-X materials) were prepared using HAT-CN (500 mg) for carbonization in a horizontal tubular furnace at different temperatures for 1 h under N 2 gas ow. The heating ramp was 2 C min À1 from room temperature to 80 C and 4 C min À1 from 80 C to the maximum temperature (500 C, 525 C, 550 C, or 700 C). The condensed materials are labelled as C-HAT-CN-X, where X represents the synthesis temperature.

Materials characterization
Prior to all physisorption measurements, the samples were degassed under vacuum at 200 C for 20 h. Ar, CO 2 , N 2 , and H 2 O vapor physisorption were performed on a Autosorb IQ apparatus (Quantachrome Instruments). Low pressure physisorption measurements were performed using Ar at À186 C with $20 mg of sample. The pore size distributions were calculated using the quenched solid density functional theory (QSDFT) method (adsorption branch kernel) for Ar adsorbed on carbon with cylindrical/sphere pore shape at 87 K, integrated into the ASiQwin 3.0 analysis soware. The specic surface areas (SSAs) were calculated using the multi-point Brunauer-Emmett-Teller (BET) model (p p 0 À1 ¼ 0.005-0.05). Total pore volumes (V total ) were determined at p p 0 À1 ¼ 0.95. CO 2 (0 C and 25 C), N 2 (À196 C, 0 C, and 25 C), and H 2 O (25 C) physisorption measurements were carried out with 40-50 mg of sample at the same instrument. The isosteric heat of CO 2 adsorption was obtained based on the Clausius-Clapeyron equation using an option integrated in the ASiQwin 3.0 analysis soware (Quantachrome Instruments). The determination of CO 2 over N 2 selectivity (S) (at 298 K; for N 2 /CO 2 ratio of 90/10) followed the ideal adsorption solution theory (IAST) method 43 and was calculated using following equation: where X is the molar ratio of CO 2 or N 2 in the adsorbed phase and Y is the molar ratio in the gas phase.
Thermal response measurements were performed at 1 bar and 298 K on an optical calorimeter (InfraSORP). 44 For the CO 2 adsorption/desorption measurements, $30 mg of the C-HAT-CN-X materials were placed in the sample holder, and purged under N 2 ow of 72 sccm until a constant sample temperature was observed. Then, the sample was exposed to a CO 2 ow of 72 sccm for 100 seconds. Finally, N 2 ow was directed to the sample cell again for 200 s causing a decrease in temperature due to desorption of the test gas. The peak areas (A) of the thermal response curves have been integrated using OriginPro2015 soware. The peak area was divided by the sample mass used for the measurements and is also normalized to the total pore volume.
The CO 2 adsorption cycling experiments were performed using similar gas ow and time-programmed conditions. All X-ray powder diffraction (XRPD) patterns were collected on a Bruker D8 Advanced diffractometer equipped with a scintillation counter detector using Cu Ka radiation (l ¼ 0.1518 nm) in the 2q range from 5 to 70 with a step size of 0.03 and counting time 1 s per step. Possible impurity phases were checked by comparing XRPD patterns with those in the PDF4 database (powder diffraction le, ICDD, release 2016).
C/H/N Elemental analysis (EA) was accomplished as combustion analysis using a Vario Micro device. The oxygen content of the samples was calculated from the residual amount of the EA which has not been detected as C, H, or N.
Fourier-transform infrared (FTIR) spectra were recorded on a Varian 1000 spectrometer with an attenuated total reectance setup.
Thermogravimetric analysis (TGA) measurements were performed using a thermo microbalance TG 209 F1 Libra (Netzsch, Selb, Germany). A platinum crucible was used for the measurement of 10 AE 1 mg of samples in a N 2 ow of 10 mL min À1 and a purge ow of 10 mL min À1 . Additional 5 mL min À1 O 2 ow has been used for the measurements in synthetic air. The samples were heated to 1000 C with a heating rate of 5 C min À1 . The data was recorded and analysed by the Proteus (6.0.0) and Quadstar (7.03) soware package.
Solid-state nuclear magnetic resonance (ssNMR) spectra were recorded with a Bruker Avance 400 MHz spectrometer operating at 100.56 MHz for 13 C. 13 C{ 1 H} magic angle spinning (MAS) ssNMR experiments were carried out at a spinning rate of 10 kHz using 4 mm MAS HX double resonance probe. The 13 C p/2 pulse length were 3.0 ms. Two pulse phase modulation (TPPM) heteronuclear dipolar decoupling was used during acquisition and a recycle delay of 20 s was implemented. All 13 C { 1 H} spectra are referenced to external TMS at 0 ppm using solid adamantine as a secondary reference.
Details of the electron microscopy characterization of the samples and computational details are given in the ESI. †

Structure of the HAT-CN-derived carbon materials
Thermal condensation of HAT-CN by elimination of the nitrile groups in the temperature range of 500-700 C leads to the formation of nitrogen-rich carbon materials with a C 2 N-like stoichiometry (Fig. S1 †). C-HAT-CN-500 and C-HAT-CN-525 show nitrogen contents of 31-34 at%, as determined by elemental analysis (EA) ( Table 1). The signicant amount of hydrogen and oxygen detected in these measurements likely arises from water adsorbed on the samples. The C/N atomic ratio of C-HAT-CN-500 (1.55) is only marginally higher than in HAT-CN (1.53). This indicates that no major condensation of the nitrile groups occurs at this temperature. Aer condensation at 525 C and 550 C, the C/N ratios increase to 1.60 and 1.67, respectively. Finally, at 700 C, the ratio reached 1.96 and a nearly perfect C 2 N-type material is obtained. According to our previous ndings, condensation of the HAT-CN precursor molecule with a composition of C 18 N 12 releases cyanogen (C 2 N 2 ) as the condensation byproduct. It can thus be assumed that the condensation proceeds according to the general equation C 18 N 12(s) / C 12 N 6(s) + 3C 2 N 2(g) . In agreement with thermal analysis of HAT-CN under inert atmosphere (Fig. S2 †), there seems to be a sharp onset point of this condensation reaction located at 500 C or slightly below. The yield of nitrogen-rich carbon material aer condensation at 500 C is 70% and drops signicantly to 51% aer thermal treatment at 525 C and to 33% aer reaction at 700 C, which is in good accordance with the TGA experiment. These elemental compositions and yields do not only reveal that HAT-CN is a very well-suited precursor molecule for the synthesis of C 2 N materials, but also that the degree of condensation (and with it the chemical and textural properties of the resulting materials) is precisely adjustable via the condensation temperature because the nitrogen content decreases at higher condensation degree due to the release of cyanogen. FTIR (Fig. 1a) and XRPD (Fig. 1b) measurements further indicate a distinct structural change between 500 C and 550 C, resulting from condensation of the HAT-CN monomer. The XRD pattern of C-HAT-CN-500 is still in nearly perfect accordance to the pattern of the untreated HAT-CN. The crystal structure is still intact aer heating to 525 C, but there is already a notable contribution of defects in the HAT structure, as can be seen by the obvious distortion of the baseline underneath the characteristic peaks in the 2q range from 10 to 30 . The contribution of amorphous scattering intensity at very low 2q values can be indicative for the development of narrow pores. This is not observed for the HAT-CN and increases in intensity from C-HAT-CN-500 to C-HAT-CN-525. In accordance with our recent study, C-HAT-CN-550 and C-HAT-CN-700 have an amorphous structure as it is typical for high surface-area carbons. The (002) peak (representing graphitic stacking in the structure) has low intensity and can be found in the 2q range from 25 to 30 . Its maximum shis to a slightly lower angle with increasing condensation temperature. FTIR spectra of the HAT-CN show sharp and narrow peaks, which disappear or get broader with increasing condensation temperature. The intense peak at 1339 cm À1 originates from C-N stretching vibrations. The C]N and C]C vibrations in the aromatic ring system appear at 1708 cm À1 and 1560 cm À1 , respectively. The peak at 2240 cm À1 is attributed to the nitrile C^N groups. This peak does not disappear as the carbonization temperature increases. This might be related to the nitrile groups located on the edges of the carbon network. In accordance with EA and XRPD measurements, the spectra show that there is a continuous increase of the condensation degree with increasing synthesis temperature.
The chemical environment of carbon atoms in HAT-CN and HAT-CN-derived carbon materials is investigated by 13 C{ 1 H} magic angle spinning (MAS) solid-state (ss) NMR experiments (Fig. 1c). Three characteristic peaks for HAT-CN are present. The peak at $132 ppm (marked with an orange circle) corresponds to aromatic carbon atoms in the inner benzene ring of the HAT-CN molecule. The peak at $139 ppm (green circle) corresponds to the pyrazinic carbon, and the one at $110 ppm (blue circle) to the nitrile carbon. 45 As expected for a well crystallized organic molecule, the peaks for HAT-CN are sharp and well separated. The condensation of the organic molecule towards a porous amorphous network clearly leads to broadening of the peaks which gets more pronounced at increasing temperature. The peak of the nitrile carbons remains for the C-HAT-CN-700, but weakens. This is due to the surface terminations at grain boundaries of the porous network particles, as it was already shown in the FTIR spectra.
Besides the atomic construction, the porosity of the HATderived carbon materials is also controllable by the synthesis temperature. In our recent study on HAT-CN-derived C 2 Nmaterials, it has been shown that (despite the absence of any additional porogen) condensation at 550 C or above leads to the formation of structural microporosity. 41 The Ar physisorption (À186 C) isotherm (Fig. 2a) shows that the C-HAT-CN-700 material with near-ideal C 2 N-stoichiometry has indeed a high specic BET surface area and a signicant total pore volume of 785 m 2 g À1 and 0.31 cm 3 g À1 , respectively. Starting from the HAT-CN with no notable surface area, there is a stepwise increase of the porosity detected with Ar at À186 C with increasing condensation temperature. Whereas the samples C-HAT-CN-500 and C-HAT-CN-525 still do not show any notable uptake of Ar, condensation at 550 C forms a material with  a BET surface area of 627 m 2 g À1 and a total pore volume of 0.24 cm 3 g À1 . The lager pore volume of C-HAT-CN-700 is the consequence of further increasing pore size with increasing degree of condensation (Fig. S3 †). The micropore volume detected for C-HAT-CN-700 with Ar physisorption at À186 C (0.30 cm 3 g À1 ) is comparable to the one detected with N 2 physisorption at À196 C (0.27 cm 3 g À1 , Fig. 2b). Notably, the difference for C-HAT-CN-550 is much larger. In this material, Ar physisorption analysis detects a micropore volume of 0.24 cm 3 g À1 , whereas nitrogen physisorption analysis only results in a micropore volume of 0.20 cm 3 g À1 . This indicates accessibility limitations for the slightly larger N 2 molecules with higher quadrupole moment at lower measurement temperature into the very small micropores of C-HAT-CN-550. In other words, these materials are possible molecular sieves, not only for CO 2 /N 2 adsorption, but also for Ar/N 2 adsorption. The preferred adsorption is based on a kinetic size exclusion on the expense of the (even stronger adsorbing) nitrogen molecules, that is, a (kinetic) molecular sieving effect. The uptakes of argon and nitrogen at relative pressures close to 1 bar are due to adsorption in the larger pores, which is in line with the TEM investigations as discussed below.
In order to utilize these materials for selective CO 2 /N 2 capture based on molecular sieving, a pore size close to the dimensions of the CO 2 molecules, but smaller than for N 2 , is required. On the other hand, the internal surface of the materials should be highly polarizing in order to have a high adsorption affinity towards CO 2 . H 2 O vapor (a molecule with even higher adsorption enthalpy than CO 2 and thus a wellsuited probe for the presence of polar adsorption sites) physisorption isotherms of the samples show an extremely polar character of the pore walls in all HAT-CN-derived materials, as indicated by the signicant water uptake at low relative pressures (Fig. S4 †). The adsorbed volumes follow the available pore volumes detected with Ar physisorption. It should be noted that the strong interaction with H 2 O will of course remain a problem for CO 2 separation with these materials under practical conditions.
High-resolution transmission electron microscopy (HRTEM) as well as annular dark-eld scanning transmission electron microscopy (ADF-STEM) images of the condensed HAT-CN materials ( Fig. 3 and S5 †) show the presence of an amorphous, covalent, carbonaceous microstructure without distinct stacking of layers in the C-HAT-CN-550 and C-HAT-CN-700, which is typical for highly microporous carbon materials and in line with our previous ndings. 41 This highly disordered structure likely results from the rapid condensation due to elimination of the nitrile groups of HAT-CN. C-HAT-CN-500 and C-HAT-CN-525 show an amorphous carbon microstructure in the TEM images. In combination with the XRPD results (Fig. 1b), this indicates that the condensed HAT-CN materials consist of crystalline and amorphous constituents. The C-HAT-CN materials seem to have a core-shell-like structure with a distinct difference between the constructions of the inner and outer areas of the individual disk-shaped particles (Fig. S5 †). While the inner areas consist of a 3-dimensional network of agglomerated small grains with larger pores in between, the double-walled shell has a rather dense appearance.

CO 2 /N 2 molecular sieving with HAT-CN-derived carbon materials
The narrow pore size of the HAT-CN-derived carbons can be rather precisely adjusted by the synthesis temperature, and the materials generally have a very polar surface resulting from their high nitrogen content. On the one site, these pyrazinic nitrogen sites have basic and electron-donating properties which are particularly strong in the C 2 N structure and able to strongly bind to the electrophilic carbon atom in CO 2 . A synergistic effect can be achieved by the similar presence of electron decient carbon atoms in the C 2 N which can strongly rebind to the nucleophilic oxygen atoms in CO 2 . These properties render the materials attractive candidates for selective adsorption of CO 2 in the presence of N 2 . CO 2 and N 2 physisorption isotherms of the HAT-CN-derived carbon materials at 0 C and 25 C show a maximum IAST CO 2 /N 2 selectivity of 121 for the C-HAT-CN-525 (Table 2 and Fig. 4). In this material, the pores are already large enough to adsorb a signicant amount of CO 2 , but are not yet too large to take up high volumes of N 2 . At higher synthesis temperatures, CO 2 uptakes increase further as a result of the increasing pore volume, but the CO 2 /N 2 selectivity decreases. At both temperatures, the selectivity of C-HAT-CN-550 with smaller pores is still higher than for C-HAT-CN-700. This conrms that the use of materials with a high N 2 surface area is not suitable for selective CO 2 capture, as long as high selectivity is required because of the increasing contribution of less discriminative van der Waals interactions. 8 In accordance to the nanostructure and nitrogen content, C-HAT-CN-525 shows a signicantly higher isosteric heat (Q st ) of CO 2 adsorption (Table 2 and Fig. S6 †) of $52 kJ mol À1 as compared to C-HAT-CN-550 ($36 kJ mol À1 ) and C-HAT-CN-700 ($42 kJ mol À1 ). It should be mentioned that these numbers may have a relatively high experimental error but such a high Q st is the result of the high nitrogen content, as well as the pore size, which is perfectly tuned for the molecular encapsulation of CO 2 . Typically, isosteric heats of adsorption, as well as IAST CO 2 /N 2 selectivity in this range are only achievable with metal-containing materials including coulomb charges. 8,24,33 In this way, C-HAT-CN-525 indeed combines the necessary requirements for highly selective CO 2 capture, which are molecular sieving and an atomic construction tuned for strong CO 2 binding (i.e., separation of the electron density between carbon and nitrogen atoms) without the need for any metal.
Thermal response measurements of CO 2 adsorption with the InfraSORP technology 44 (Fig. 5a) provide further information about CO 2 adsorption capacities and kinetics of the HAT-CNderived materials. In line with the volumetric measurements, the highest temperature peak areas are present for C-HAT-CN-700 and C-HAT-CN-550 due to their large micropores, and thus highest CO 2 uptake at 1 bar. Regarding their very low Ar and N 2 accessible microporosity, the signicant temperature signals of C-HAT-CN-525 and C-HAT-CN-500 are remarkable. According to their higher Q st values their normalized peak areas (divided by mass and the total pore volumes determined with Ar physisorption) are signicantly higher as compared to the materials obtained by HAT-CN condensation at higher temperatures (Fig. 5b). The mass-related peak areas for adsorption and desorption are nearly similar for one and the same material, indicating no occurrence of irreversible binding of CO 2 . Compared to a microporous carbide-derived carbon material (TiC-CDC, synthesized at 800 C), 44 the presence of specic CO 2 binding sites is further indicated by the minor asymmetry in the adsorption and desorption temperature signals in the cases of the C-HAT-CN materials ( Fig. 5c and 5d). The CO 2 desorption process by ushing with N 2 causes a slightly lower peak temperature decrease than the adsorption of CO 2 . This is typical for materials with narrow pores and abundant nitrogen sites, and thus not observable for the heteroatom-free TiC-CDC with slightly larger micropores. The latter is showing rather symmetrical and "sharp" temperature signals in CO 2 adsorption.
Despite the high CO 2 adsorption enthalpy of C-HAT-CN-525, the binding is fully reversible for 80 cycles by simple ushing with N 2 (Fig. S7 †). This means, that this molecular sieving concept is applicable to simple pressure swing adsorption without loss of CO 2 separation efficiency over time.
Calculation of the CO 2 adsorption state on C 2 N An atomistic model of the adsorption state of CO 2 on an idealized C-HAT-CN material with perfect C 2 N-type Table 2 Argon physisorption (À186 C) data summary, selectivity of CO 2 over N 2 (at 0 C and 25 C), and isosteric heat of adsorption (Q st ) of CO 2 of C-HAT-CN-500, C-HAT-CN-525, C-HAT-CN-550, and C-HAT-CN-700 Material SSA BET /m 2 g À1 SSA DFT /m 2 g À1 V Total /cm 3   stoichiometry and AAA-type layer stacking (Fig. 6a) has been obtained by dynamical simulated annealing 46,47 using the second-generation Car-Parrinello molecular dynamics approach of Kühne et al. (computational details are described in the ESI †). 48,49 We nd that the CO 2 molecules are oriented orthogonal to the C 2 N-planes occupying every second vacancy in a checkerboard-like fashion. This energy minimization arises from collective stabilization effects from the nitrogen atoms in the C 2 N layers surrounding the carbon atom in the CO 2 molecule and due to the electron acceptor properties of the carbon atoms from C 2 N in close proximity to the oxygen atoms in CO 2 . In simple words, all atoms of the gas molecule are surrounded in their electronically favourable environment. In the present supercell, which contains of 4 layers with 4 holes each, the insertion of 8 CO 2 molecules entails an adsorption energy of À231.5 kJ mol À1 that equals to À29 kJ mol À1 per molecule. This Q st value is slightly lower as compared to the experimental value determined for the C-HAT-CN-700 material. In accordance with our gas physisorption data, this indicates that the pore size in the experimental sample is lower than in the idealized C 2 N material used for the theoretical calculations with pores which would result from perfect AAA stacking. Another possible reason for the higher experimental Q st value could be the overlapping of the adsorption potentials from two pore walls in narrow cavities which is not taken into consideration by the theoretical calculations. In the corresponding nitrogen-free control system with C 3 H stoichiometry, all nitrogen atoms have been substituted by C-H groups. In this case, the adsorption energy is strictly positive being as high as +206 kJ mol À1 (Fig. 6b). No energetically favourable coordination of CO 2 takes place in this environment.

Conclusions
We have shown that the thermal condensation of HAT-CN is a process that produces extremely nitrogen-rich carbon materials under precise control over the molecular composition and nanostructure/porosity. At a condensation temperature of 525 C, pores start forming with a size that allows for CO 2 /N 2 molecular sieving. To the best of our knowledge, this is the rst time that a metal-free and non-organic material with an isosteric heat of CO 2 adsorption of up to 52 kJ mol À1 and a CO 2 /N 2 selectivity of more than 100 has been prepared.
These properties are mainly related to the combination of strong binding of CO 2 (provided by negatively polarized nitrogen atoms and positively polarized carbon atoms) as a result of the combination of the atomic construction with the narrow pore size of the C-HAT-CN materials. A dened pore size alone for CO 2 adsorption is useless as the capture will become too slow and a designed chemical construction combined with too large pores is also not benecial, as then the selectivity is limited by non-discriminative van der Waals interactions.
As it can be assumed that only a small part of the bulk volume present in C-HAT-CN-525 is really used for CO 2 binding, the logical next step would be the synthesis of a hierarchical C 2 N-type material in order to shorted mass transport pathways and thus to enhance the accessibility of the internal pore volume. This approach is currently applied in zeolite research 50,51 and will also help in case of such "organic zeolites". On the other hand, the insertion of extra-framework cations in these materials may be an approach to further increase the CO 2 adsorption enthalpy. The main focus of this work was to illuminate the origin of the strong interaction between CO 2 and nitrogen-rich carbon materials prepared by molecular design. Aer the general possibility of CO 2 molecular sieving with metal-free materials has been shown in this work, the applicability of such materials under real-world conditions, as well as the kinetics of CO 2 adsorption, has to be further investigated with mixed-gas breakthrough experiments under wet conditions.  Finally, this synthetic concept towards such zeolite-like carbon materials is potentially also applicable for the separation of other relevant gas mixtures.

Conflicts of interest
There are no conicts to declare.