Tailored Gas Adsorption Properties of Electrospun Carbon Nanofibers for Gas Separation and Storage

Abstract Carbon nanofibers (CNFs) derived from electrospun polyacrylonitrile (PAN) were investigated with respect to their gas adsorption properties. By employing CO2 adsorption measurements, it is shown that the adsorption capacity and selectivity of the fibers can be tailored by means of the applied carbonization temperature. General pore properties of the CNFs were identified by Ar adsorption measurements, whereas CO2 adsorption measurements provided information about the ultramicroporosity, adsorption energies, and adsorption capacities. Ideal adsorbed solution theory (IAST) selectivities under practically relevant conditions were determined by evaluation of single‐component data for N2 and CO2. Especially for low carbonization temperatures, the CNFs exhibit very good low‐pressure adsorption performance and excellent CO2/N2 IAST selectivities of 350 at 20 mbar and 132 at 1 bar, which are attributed to a molecular‐sieve effect in very narrow slit pores. These IAST selectivities are some of the highest values for carbon materials reported in the literature so far and the highest IAST selectivities for as‐prepared, non‐post‐treated carbon ever.


Introduction
The scientific significance and public perception of anthropogenic climate change are increasing significantly, [1] as the fraction of CO 2 in the atmosphere has reached 407 ppm and is still growing. To attenuate the devastatinge ffects of climate change, CO 2 emissions must be reduced by any means. Two feasible ways to do so are to avoid the consumption of fossil fuels such as crude oil and natural gas, and to capture and store CO 2 whenever possible. Although technologically and economically challenging, the latterm ay have benefits besides the reduction of CO 2 emissions. Once separated from other gases, CO 2 can chemically be transformed into value-added organic compounds such as methanol [2] and ethylene. [3] For this, electrochemical CO 2 reduction is particularly attractive due to its versatility [4] and the possibility to combine both CO 2 capture and utilization in one process. By transforming CO 2 into syn-thetic fuels or raw organic chemicals,ac losed carbon cycle leadingtoaC O 2 -neutral economy can be established.
With the aim of efficient CO 2 adsorption, [5] al arge number of different materials have been investigated, including metal-organic frameworks (MOFs), [6][7][8] zeolites, [9][10][11][12] carbons, [13][14][15][16][17] polymers, [18] and functionalized silica. [19] Among thesem aterials, carbonso ffer severala dvantages over zeolites and MOFs, since they are widely abundant, cheap, easy to prepare, and comparatively insensitive towards contaminants such as water. [5,20,21] Moreover, in contrast to polymers, carbonsa re mostly electrically conductive, [20] which enables their application as current collectors in CO 2 electrolyzersa sw ell. However,t wo aspects that are lacking with respect to present carbon materials are decent adsorption capacity andC O 2 selectivity at low pressure, which so far are significantly better for tailor-made MOFs. [5,21] Owing to the numerous advantages,al arge variety of carbons has been investigated in search of improvedC O 2 adsorption properties. [15][16][17] Many of these werebiomass-based, for example,c arbons that have been prepared from soy bean dreg, [22] coconut, [23] palm shell, [24] bamboo, [25] and many more. Biomass-based carbonsa re fairly easy to preparea nd cheap. However,t ailoring their properties is demanding, since the chemicalc ompositiona nd structure (i.e.,t he pore system)a re fixed by the precursor. [15] In contrast, carbonsm ay also be synthesized from polymers, which provideh igherc ontrollability of the chemical properties. This includes carbons obtained from melamine [26] or urea-based resins [27] as well as carbonized polypyrrole, [28] polyindole, [29] poly(vinylidene fluoride), [30] and polyacrylonitrile (PAN). [31][32][33] With the aim of maximum CO 2 adsorption capacity and selectivity,t wo main approaches are availablef or carbon materials. Firstly,n itrogen functional groups can be introduced into Carbon nanofibers (CNFs) derived from electrospun polyacrylonitrile (PAN) were investigatedw ith respectt ot heir gas adsorptionp roperties. By employing CO 2 adsorption measurements,i ti ss hown that the adsorption capacity and selectivity of the fibers can be tailoredb ym eans of the applied carbonization temperature. General pore properties of the CNFs were identified by Ar adsorption measurements, whereas CO 2 adsorptionm easurements provided information about the ultramicroporosity,a dsorption energies, and adsorption capacities. Ideal adsorbed solution theory (IAST) selectivities under practi-cally relevant conditions wered etermined by evaluation of single-component data for N 2 and CO 2 .E speciallyf or low carbonization temperatures, the CNFs exhibit very good low-pressure adsorption performance and excellent CO 2 /N 2 IAST selectivities of 350 at 20 mbar and 132 at 1bar,which are attributed to am olecular-sieve effect in very narrow slit pores. These IAST selectivities are some of the highest values for carbon materials reported in the literature so far and the highest IAST selectivities for as-prepared,non-post-treated carbon ever.
the carbon matrix, [34] as ah igh nitrogen content is favorable owing to acid-base interactions between the adsorbent and CO 2 .T he number and nature of nitrogen based functional groups are usually adjusted by means of nitrogen-containing precursors or subsequentfunctionalization of the prepared carbons. However,t he impact of certain nitrogen-containing groups on the CO 2 adsorption properties is still am atter of debate. Whereas Li et al. consider pyrrolic groups most important for CO 2 adsorption, [35] Kim et al. found al arger influence of pyridinic groups. [33] Secondly,t he physicals urfacea rea may be enlarged by processes such as KOH etching and CO 2 activation to increase the number of bare interaction sites. [17] Shen et al. [36] used KOH to activate commercial PANf ibers, which had aB ET area of 0.24 m 2 g À1 .B yK OH activation, the BET area was enlargedt o 2231.24 m 2 g À1 ,w hich increased the CO 2 capacity by af actor of ten to 4.5 mmol g À1 . [36] Besides the adsorption capacity,t he selectivity is another important property,e specially for practical applications. For instance,i ncreasing the CO 2 /N 2 selectivity of at ypical material (CO 2 adsorption capacity: 3mmol g À1 )f rom 50 to 100 can mitigate the cost for CO 2 capture from flue gas by 20 %, that is, from 35 to 28 USD t À1 . [37] For high selectivity an arrow pore system rather than ah igh specific surface area is required to achieve am olecular-sieve effect. [21] Moreover,t he selectivity can further be enhanced by the introduction of nitrogen functionalities, which may contribute to high selectivity due to selective interactions with CO 2 as well. [21] Indeed, PAN-derived carbons have been under investigation for CO 2 adsorption before andh ave mostly been obtained from bulk polymer, [38,39] wet spinning, [31,35,36] or electrospinning. [33] However,m ost of these carbonsw ere post-treated by variousa ctivation processes to achievee nhanced CO 2 adsorption properties. Thus, surprisinglyl ittle information is available on CO 2 adsorption on unmodified PAN-derived carbon.
In this work, unmodified, electrospun, PAN-derived carbon nanofibers (CNFs) werei nvestigated as adsorbentsf or CO 2 , with the aim of deeper understanding of the adsorption processes on polymer-based carbons.P AN-derived CNFs are easy to prepare, even on al arge scale, and contain as ignificant amounto fn itrogen functionalities, if an appropriate carbonization temperature is applied. [40] In electrospinning, aP AN polymer solution is spun in ac ontrolled atmosphere under ah ighvoltage electric field for the preparation of ac arbon mat consisting of nonwoven fibers. Afterwards, the PANp olymer chains are stabilized and cross-linked in air and carbonized in argon to yield carbon fibers in the submicrometer range, that is, CNFs, which have av ery high surfacea rea. [41] The elemental composition and the pore properties of the as-prepared CNFs are finely adjustable by means of the carbonizationt emperature, and thus excellent low-pressurea dsorption capacity and selectivity towards CO 2 are attainable. Thus,t he approach to prepareahighly selective carbon material proposed herein particularly abstains from excessive post-treatmentb yk eeping the synthesis procedure simple and, therefore, as cost-efficient and scalable as possible. As ar esult,w er eport maximum CO 2 adsorption capacities of 1.5 mmol g À1 at 100 mbar and 2.8 mmol g À1 at 1bar.M oreover,b yt ailoring the ultramicropore system (micropores < 0.7 nm) [42] it is possible to achieve ideal adsorbed solution theory (IAST) selectivities of 350 at a low pressure of 20 mbar and 132 at 1bar.B oth values are very close to typical resultsf or MOFs, among the highestv alues for carbon materials reported so far,and the highest values for unmodified carbons.

Results and Discussion
Fiber morphology,structure and chemistry By electrospinning and carbonization at different temperatures rangingf rom 600 to 1100 8C, CNF mats were prepared from a 10 wt %P AN solution (see Experimental Section). In the SEM images (Figure 1a-d) the nonwovenf ibers show an even surface and no preferred orientation at all carbonization temperatures. The fiber diameters are rather uniform and decrease only slightly from 250 to 220 nm in the investigated range of carbonization temperatures. Moreover,t he TEM images in Figure 1e and fr eveal as lightly ordered carbon structure, which shows increased surface roughnessf or higher carbonization temperatures. Ad etailed in situ TEM analysis under vacuum of the effect of the carbonization temperature on the CNFs has been published elsewhere. [43] To analyze the chemicalc omposition of the CNFs, CHNO elemental analysisw as performed ( Table 1). The carbon content of the CNFs increases continuously for increasing carbonization temperature from 600 to 1100 8Cw ith al arger step from 77.6 wt %( 900 8C) to 91.5 wt %( 1000 8C). Simultaneously,t he nitrogen content of the carbonized fibers decreases almost linearly from 23 to 7wt%.T his correlationf airly matches the expectations based on the literature, in which 17 wt %n itrogen for 700 8C [44] and 5.8wt% nitrogen for 1000 8C [45] were reported for carbonized PAN. In contrastt oc arbon and nitrogen, the oxygen content remains almostc onstant between 10 and 12 wt %inthe temperature range from 600 to 900 8C, which indicatesh igher thermals tability of the remaining oxygen-containing functional groups.A bove 900 8Ct he oxygen content decreases to 0.8 wt %f or ac arbonization temperatureo f 1100 8C. It is expectedt hat the majority of the oxygen-containing groups in the cross-linked fibers already decomposed and released oxygen as water at lower temperatureso f3 00-400 8C. [45] Furthermore, the decreaseo ft he hydrogen content from 2.7 to 0.4 wt %i ndicated progress of the carbonization. In fact, at highert emperatures CÀHb onds are thermally destroyed and cross-linking processes between carbon atoms take place. [46] In additiont ot he elemental analysis, X-ray photoelectron spectroscopy (XPS) was performed to characterize the surface composition and the functional groups of the individualf iber mats (fors pectra and integration,s ee Figures S2 and S3 in the Supporting Information). Moreover,X PS also yields data regardingt he elemental composition( Table 2). On comparing the data from the CHNO method and XPS, it can generallyb e observedt hat the individual resultsshow similartrends regarding the carbon and nitrogen contents. Althoughc omparing the exact numbers is not possible, as XPS data is given in atom %a nd CHNO data in wt %, the differencei ssmall for light elements with similar atomicm ass. However,t he XPS data show highera bsolute values for carbon (76.5-95.1 atom %) and lower absolute values for nitrogen (19.7-3.1 atom %) and oxygen (3.8-1.8 atom %). In addition, the sudden jump in carbon content betweent he materials carbonized at 900 and 1000 8Ci sl ess pronouncedf or the XPS resultsa sc ompared with the findings obtained by the CHNO method. In contrast to the results provided by combustion elemental analysis (CHNO), the surface-oxygen fraction determined by XPS is far lower and does not exceed 4atom %. Furthermore, the oxygen contentd oes not show any trend dependingont he carbonization temperature.T his suggestst hat mosto ft he oxygen that was detected by the CHNO method is trapped inside the fibers and is not accessible to XPS, as this technique is surface-sensitive, limited to ap enetration depth of af ew nanometers, [47] and also depends on the angle of incidence. [48] However,t he difference between elemental analysis and XPS results can also be explained by the experimental procedure, as the sample may have contained adsorbed CO 2 or O 2 during the elemental analysis, whereas XPS is performed under ultrahigh vacuum. Furthermore, the indirect measurement procedure for the determination of oxygen in the elemental analysis( see Experimental Section) could causedeviations.
Besides the elemental composition,X PS measurements also provide information on the functional groups on the material surface. Especially nitrogen-containing functional groups are of interestf or CO 2 adsorption owing to their basic nature [20,27,49] and their positive influence on weak hydrogen-bonding interactions between CO 2 and hydrogen on the carbon surface. [22] Accordingly,t he fractions of different nitrogen species were evaluated in dependence on the applied carbonization temperature. Figure 2a depicts the XPS data normalized to the nitrogenp eak area, and Figure2bs hows the same data multiplied by the total nitrogen content from Table 2, which provide the overall contribution of the nitrogen-containing functional groups to the atomics urface composition. In Figure 2a,q uaternary and pyridinic nitrogen are the dominant species. In contrast, pyrrolic nitrogen as the third potentially important speciesi sa lmosta bsent. For ac omparatively low carbonization temperatureo f6 00 8C, the numberso fq uaternary and pyridinic groups are almost equal, with as mall advantage for the latter.H owever,w ith increasing carbonization temperature, the ratio of quaternary nitrogen to pyridinic nitrogen increases, with ar atio of approximately 4:1a t1 100 8C. This can be explained by the fact that pyridinic nitrogen in carbonized PANi s transformed into quaternary nitrogen at high temperatures. [50] At the same time, the fraction of pyrrolic nitrogen increases for increasing carbonization temperatures from 600 to 800 8Ca nd, besides as mall variation, remains constant and always less than 5% for carbonization temperatures above 800 8C.
Looking at the fraction of nitrogen moieties in the total number of surfacea toms in Figure2br eveals that, even thoughq uaternary nitrogen becomes more dominant, the absolute number of both quaternary and pyridinic nitrogen species continuously decreases from about 10 atom %t ol ess than 2atom %w ith increasing carbonization temperature. Ap ossible reason is the decomposition of functionalg roups to N 2 , HCN, and NH 3 . [50,51] Furthermore, the number of pyrrolic nitrogen moieties shows am aximum at carbonization temperatures of 800 and 900 8C, but their fraction relative to the total number of surface atoms remains below 1atom %a nd therefore most probablyd oes not play as ignificantr ole in CO 2 adsorption.
Ar adsorption,s urfacearea, and pore structure Besides functional groups,t he pore system of am ateriali s considered to be an important factor for CO 2 adsorption, [49] since al arger surfacea rea increases the number of potential adsorption sites. Furthermore, narrow pores are highly beneficial to low-pressure adsorption, as the adsorptionp otentials of opposing pore walls overlap. [42,52] To investigate the pore structure of the CNFs, static Ar adsorption measurements were performed at 87 K. The resulting isotherms are shown in Figure 3. For clarity,t he resultsf or the materials carbonized at 600 and 700, and 800-1100 8Ca re shown separately in Figure 3a and b, and can be explained as follows. On carbonizing at 600 and 700 8C, the materials exhibit type Ii sotherms, [42] which are typical of microporous adsorbents. However,t he two isotherms are not in equilibrium, as the desorption branches do not meet the adsorption branch againa tl ow relative pressures, even thought he isotherms were measured with the highest technically possible equilibration parameters. This pseudo-irreversibility of the Ar adsorption implies kinetic hindrance, which is known to occur when ultramicropores smaller than 0.45 nm are present. [53] The isotherms both show as imilar total adsorbeda mount of Ar of 4mmol g À1 at ambient pressure (1000 mbar). The only significant difference between the two materials is the slope of the isotherma tr elative pressures below 0.1. Here, the much steeper slope of the materialc arbonized at 700 8Ci ndicates narrower micropores or ah igher adsorption energy. In contrastt ot he under-equilibrated isotherms of the materials carbonized at 600 and 700 8C, those carbonized at 800 8C and above exhibit aw ell-equilibrated type II isotherm without any hysteresis ( Figure 3b), which is expected for nonporous materials or materials that do not have pores accessible to Ar. Furthermore, whereas the shape remains very similarf or all materials, the overall adsorbed amount of Ar at 1000mbar slightly increases from 0.6 to 0.9 mmol g À1 ,w hich is most probably due to slightly decreasing fiber diameter with increasing carbonization temperature.
From the Ar adsorption isotherms, BET areas were calculated ( Table 3). The materials carbonized at 600 and 700 8Ch ave BET areas of approximately 250 m 2 g À1 ,w hichi sc omparatively high, as they were not activated by any additional reactant. Nevertheless, the value is low compared to those of chemically activated carbons, which easily exceed values of 1500 m 2 g À1 . [31,36,38,39] However,both results should only be considered asr ough estimates, since the recorded isothermsw ere not fully equilibrated. In contrast, the materialsc arbonized at (a) Materials carbonized at 600 and 700 8C(C600 and C700). (b) Materials carbonized at 800-1100 8C( C800-C1100).In( b) the desorption branches are not shownfor the sake of clarity,but can be found in Figure S4. 800-1100 8Ch ave surface areas of approximately 15 m 2 g À1 , which is close to the expectedv alue of 9m 2 g À1 for nonporous fibers with ad iametero f2 00 nm. Overall, the surfacea reas show as light trend to higher values for increasing carbonization temperatures,a so bserved for the adsorbed gas volumes at ambient pressure.  Figure S6). Figure 4b shows the cumulative pore size distributions, which were obtained from the isotherms shown in Figure 4a.
In Figure 4a the individual CO 2 adsorption isotherms show marked changes in shape depending on the appliedcarbonization temperature.T he materials carbonized at 600-850 8C show as teep increaseo ft he adsorbed amount of CO 2 at low pressures, leading to 0.85 mmol g À1 and 1.3 mmolg À1 at pressures as low as 25 and 75 mbar,r espectively,f or 600 8C, which is ar emarkable, practically relevantr esultf or carbon. In fact, excellent low-pressurep erformance is ak ey requirement for gas separation applications, for example, CO 2 separation from flue gas with al ow relative pressure of 8-13 %f or CO 2 .( Note: unlike the Ar measurements, the materials carbonized at 600 8Ca nd 700 8Cw ere evaluated in equilibration due to the higher measurement temperature used for CO 2 adsorption. [54] ) At higher pressures, the isotherm slope for the materialc arbonizeda t6 00 8Cd ecreases and resultsi nafinal amount of 2.7 mmol g À1 at 1bar.I nc ontrast, the materialc arbonized at 900 8Cs hows as maller initial slope of the CO 2 adsorption isotherm, but only as lightly reduced amount of adsorbed CO 2 of 2.4 mmol g À1 at 1bar.A th igherc arbonizationt emperatures of 950-1100 8Ct he isotherms tend to becomea lmost linear and the total adsorbed amount of CO 2 at 1bar remains below 1.0 mmol g À1 .F urthermore, the adsorption kinetics for materi-als that were carbonized between 950a nd 1100 8Ca ppear to become slower,s ince the difference between adsorption and desorption branch increases with increasing carbonization temperature (see Figure S6).
For an improved overview of the individual adsorptionproperties, the adsorbed amountso fC O 2 at certain pressures are given in Ta ble S2 in comparison with those of somec ommercial carbons. At apressure of 50 mbar,the adsorbed amount of CO 2 for the materialc arbonized at 600 8Ci sm ore than 50 times higher than that of the material carbonized at 1000 8C. However,a ta mbient pressure, the amount of ad-sorbedC O 2 is only about five times higher. Nevertheless,i n contrastt ot he commercial carbons chosen for comparison, materials carbonized at temperatures between 600 and 875 8C show excellent low-pressure adsorption performance, regardless of their comparatively low BET area. Black Pearls 2000, for example, has aB ET surfacea rea greater than 1500 m 2 g À1 and adsorbs 4.3mmol g À1 CO 2 at ambient pressure, but its CO 2 uptake at 50mbar is only about half of those of the materials carbonized at 600-875 8C. Furthermore, Super Pn onporous carbon and graphene platelets with am edium scaled BET area of 250 m 2 g À1 also do not reach the adsorbed amounts of the investigated CNFs, neither at low nor at ambient pressure. Thus, the electrospun CNFs offer superior adsorption properties for CO 2 at low pressures in comparison with other carbons (see also Table S1), which favor application in separating CO 2 from gas mixtures with low CO 2 concentrations.  To elucidate the origin of the superior adsorption properties, pore size distributions were derivedf rom the individualC O 2 isothermsb ys tandard Monte Carlo (MC) calculations. The cumulative pore size distributions are shown in Figure 4b.T he materials carbonized at 600-850 8Ca ppear to exhibit as ignificant amounto fp ores that are smaller than the calculation limit of 0.35 nm. This is indicated by the fact that the cumulative pore volumec urves do not start at 0cm 3 g À1 ,b ut slightly above at 0.05 cm 3 g À1 .
Overall, the pore volume of ultramicropores smaller than 0.40 nm decreases with increasing carbonizationt emperature. Moreover,l arger ultramicropores are present as well, reaching at otal volumeo f0 .1 cm 3 g À1 for the materials carbonized at 600-850 8C. In contrast, at ac arbonizationt emperature higher than 950 8Ct he materials no longer have as ignificant measurable specific ultramicropore volume, but still show some supermicroporosity (0.7-2.0 nm). [42] The slit-pore width of 0.35 nm that was observed fort he materials carbonized at 600-850 8C, is very close to the interlayer distance in graphitic (0.3354 nm) or turbostratic carbon (0.34 nm), [55] whichi mplies that the observedp orosity in this range can be attributed to interlayer spaces of the carbon. In addition, the observed width range of the ultramicropores is not only close to the interlayer distance of graphite, but also matches the kinetic diameters of technically relevant gases. Indeed, CO 2 has ak inetic diameter of 0.330 nm, [56] whereas N 2 and Ar are slightly larger (0.364 [56] and0 .340 nm, [54] respectively). Therefore, am olecular-sieve effect appears to be areasonable explanation for the excellent CO 2 adsorption capability comparedw ith Ar.
An overview of the pore volumes and other textural properties such as the micropore surface area S micro obtained by MC and Dubinin-Radushkevich (DR) calculations is given in Ta ble 4 in comparison with those of commercial carbon materials.F or carbonization temperatures of 600-900 8Ct he CNFs exhibit mi-cropores urfacea reas of approximately 600 m 2 g À1 (MC), which decreaset oa bout 150 m 2 g À1 for ac arbonization temperature of 1000 8Ca nd above. The DR results support the MC values. However, the DR values are about 25 %l ower,w hich is suggestedt ob ed ue to limitations of the DR equationr egarding the heterogeneity of surfacec hemistry or texture. [57] Furthermore, it is notablet hat even materialsw ith av ery low BET area of 15 m 2 g À1 show as ignificant micropore surfacea rea of more than 600 m 2 g À1 (e.g.,a t8 00 8C), which is, again,ahint that the ultramicroporosity of these materials is not accessible to Ar. Similar to the trend of the micropores urfacea rea,t he overall pore volume (MC) decreases from 0.190 cm 3 g À1 (900 8C) to 0.07 cm 3 g À1 (1000 8C), which is far lesss ignificant than the decrease in the adsorbed amount of CO 2 at low pressures mentioned above. However,t his trend complies better with the overall adsorbed amounts of CO 2 at 1bar,w hich was found to be five times higher forac arbonization temperature of 600 8C than for 1000 8C. This observation can be explained by the fact that larger pores are only filled at higher pressures. The DR microporev olumes are comparable to those obtained by MC calculations.I nterestingly,t he decrease in pore volumea t9 00 8C is sharper (0.195 cm 3 g À1 at 875 8Ct o0 .06 cm 3 g À1 at 925 8C) and slightly shifted towards lower carbonization temperatures when determined by DR, althoughboth data sets were derived from the same raw data.
Black Pearls 2000 has am icropore surface area and am icropore volumet hat are twice as high as those of the material carbonized at 600 8C( Ta ble 4). The latter is in very good agreement with the doubled CO 2 loading at ambient pressure. On the other hand, Super Pa nd graphene platelets have lower values, which correspond to their CO 2 adsorption capacity at 1bar as well.
Besides micropores urface area and micropore volume, the DR method allows one to obtain adsorption energies for CO 2 , which are also listed in Table 4. For ac arbonization tempera- ture of 600 8C, the adsorption energy is 36.0 kJ mol À1 and constantly decreases with increasingc arbonization temperature to 31.9 kJ mol À1 for the materialc arbonized at 875 8C. Above 875 8C, the adsorption energy decreases to 23.2 kJ mol À1 for 925 8Ca nd remains fairly constanta ta pproximately 20 kJ mol À1 for higherc arbonization temperatures. For further investigation, the isosteric heats of adsorption were calculated as well. The heats of adsorption are shown in dependenceo nt he carbonization temperature and CO 2 loading in FigureS8. For the materials carbonized at 600-800 8Ct he isosterich eats of adsorptiona re approximately 40 kJ mol À1 for small amounts of CO 2 and 25 kJ mol À1 for high CO 2 loadings. For carbonization temperatures of 1000a nd 1100 8Ct he isosteric heat decreases to 10 and5kJ mol À1 ,r espectively.A ni ncreasei na dsorption energy can be causede ither by as tronger chemicali nteraction between functional groups or by an overlap of pore-wall potentials in narrow pores. However,s ince the number of functional groups and the number of ultramicropores increase with decreasing carbonization temperature, clear separation of the two effectsisn ot possible. By relating Ar and CO 2 adsorption measurements with each other,i ti sp ossible to calculate as urfacea ffinity towards CO 2 , that is, by dividing the adsorbed amounto fC O 2 by the BET surfacea rea. On evaluating the surfacea ffinities towards CO 2 , it can clearly be observed in Figure 5t hat CO 2 adsorption is highly favoredi nacarbonization temperature range of 800 8C to 900 8C. For these materials, the calculated surfacea ffinities reach extremelyh igh values of 0.2 mmol m À2 and drop by one order of magnitude for ac arbonization temperature of 975 8C and above,d ue to the significantly lower CO 2 adsorption capacity.F or the materials carbonized at 600-700 8Ct he surface affinitiesa re lower than 0.03 mmol m À2 ,d ue to the larger accessibles urface areas, whichw ere, however,o btainedf rom under-equilibratedi sotherms.
Comparing the results in Figure 5w ithl iterature data reveals that the investigated CNFs have remarkable properties. In fact, most carbonss uggested for CO 2 adsorption shows ignificantly lower surface affinities, typically less than 0.005 mmol m À2 ,d ue to the very high BET areas of the corresponding materials (Table S1). Ta king the very narrow pore widths into account, the extraordinary surface affinity implies that the superior ad-sorptionproperties at low pressures and the very high selectivity towards CO 2 can be attributed to am olecular-sieve effect. For carbonization temperatures of 600 and 700 8C, the narrowest microporesa re accessible to both Ar atoms and CO 2 molecules. On carbonizing between 800 and 875 8C, larger Ar atoms can no longerp enetrate the shrinking pores, whereas CO 2 adsorptioni ss till possible. Furthermore, on further increasingt he carbonization temperature, the micropores become too narrow to allow either CO 2 or Ar adsorption, and the adsorption capacity decreases drastically.T os ubstantiate the hypothesis that am olecular-sieve effect is am ajor driver of the observed adsorption behavior of Ar and CO 2 ,a dditional measurements on commercial zeolite molecular sieves with defined pore structure were conducted. These measurements along with ashort interpretation are shown in Figure S7.

Microporosity versussurfacefunctionality
Besides microporosity, surfacef unctionality is as econd main factor influencing the CO 2 adsorption capacity of carbon materials [34] that can explain an increase in adsorptione nergy and surfacea ffinity.T herefore, it is discussed in competition with the effect of pore shrinkageb elow.T ot his end, Figure 6a shows the micropore volumef or pores narrower than 0.4 nm for all materials given in Table 4. Figure 6a reveals that the ultramicropore volume behaves similarly to thec arbonizationtemperature-dependent surface affinity shown in Figure5.F or materials carbonized at 600 and 700 8C, the volumeo fu ltrami-  cropores smaller than 0.4 nm is 0.05 cm 3 g À1 ,w hereas materials carbonized at 800-875 8Ce xhibit 0.04 cm 3 g À1 .T hen, in ac arbonization temperature range of only 100 8C, the micropore volumed rops to 0cm 3 g À1 for ac arbonization temperature of 950 8Cand beyond. Furthermore, Figure 6b correlates the adsorbed amounto f CO 2 at 50 mbar with the volume of micropores smaller than 0.4 nm. Thus, it can be concludedt hat the volumeo fu ltramicropores smallert han 0.4 nm is am ajor driver for the adsorbed amount of CO 2 at 50 mbar.I ndeed, whereas am aterialw ith an ultramicropore (< 0.4 nm) volume of 0cm 3 g À1 adsorbs almost no CO 2 ,amaterialw ith an ultramicropore (< 0.4 nm) volume of 0.05 cm 3 g À1 already leads to 1.1 mmol g À1 of adsorbed CO 2 at 50 mbar.T he correlation between ultramicropore (< 0.4 nm) volumea nd adsorbed amount of CO 2 is almost linear.T his observation is as trong hint that, at the given pressure of 50 mbar,the ultramicropores are mainly responsible for the adsorptiono fC O 2 ,r ather than the open surface, which is in accordancew ith results previously described in the literature. [58][59][60][61] To study the correlation between nitrogen functional groups and the CO 2 uptake, the adsorbed amount of CO 2 at 50 mbar is plotted asafunction of the fraction of surfacenitrogen functional groups determined by XPS in Figure 7. From Figure 7i t can be deduced that there is ar estrained correlation between pyridinic and quaternary groups and the adsorbed amount of CO 2 .Itappears that in the range of carbonization temperatures between 700 and 1000 8C, both groups facilitateC O 2 adsorption. This is confirmed by the correlation with the nitrogen content determined by elemental analysis( Ta ble 1), since the adsorbed amounto fC O 2 is higher for the materials with a higher nitrogen content. Nevertheless, the correlation between adsorbed amount of CO 2 at 50 mbar andt he nitrogen functional groups is less clear than that between the adsorbed amount of CO 2 and the ultramicropore volume. In contrast to pyridinica nd quaternary moieties, the amount of pyrrolic nitrogen is too small to show asignificant effect on CO 2 adsorption, and no clear trend is visible. Similar results for the relation between CO 2 adsorption and nitrogen content have been reported by Zhang et al. for polyaniline-based carbons. [58] They described ap ositive influence of nitrogen on CO 2 adsorption, albeit with significant scattering of data points.
In general, it is difficult to distinguish between the effect of microporosity and nitrogen surface groupso nt he CO 2 adsorption performance. Both parameters change continuouslyo ver the studied range of carbonization temperatures;t hus, the influence of both might overlap and result in misleadingc orrelations, although previouss tudies found as ynergetic effect of microporosity and Nd oping. [62] In addition, the influences of microporosity andf unctional groups cannot be discussed independently,s incei ti sn ot possible to decrease the nitrogen content without influencing the microporosity for the same polymer and same preparation procedure, respectively.A sa reason, we assumet hat nitrogen atoms act as structure-disturbing heteroatoms, the decreasing number of which with increasingc arbonization temperature resultsi nl ess disturbance and, therefore, shrinkage of the carbon interlayer distance (i.e., narrower micropores).
All in all, from ac ombinationo fm icropore and surfacegroup analyses, it can be deduced that the influence of ultramicropores appearst ob em ore significant than the effect of the nitrogen moieties. In fact, depending the carbonization temperature, it appears that slit pores with av ariable width occur between the carbon layersa nd govern the gas adsorption, as is schematically depicted in Figure 8. The materials carbonized at 600 and 700 8Cc an adsorb both CO 2 and Ar in rather large amounts between the carbon layers ( Figure 8e). However,t he extremelys low adsorption of Ar indicates that the pore width is very close to the limit that is accessible to Ar atoms, which have al arger kinetic diameter than CO 2 .F or carbonization temperatures of 800-875 8Ct he carbon interlayer spaces become too small to be penetrated by Ar,b ut they can still adsorb significant amountso fC O 2 ,a nd this leads to excellent surfacea ffinity towards CO 2 (Figure 8f). When the CNFs are carbonized at even higher temperatures (900-1100 8C), the coherence of the carbon layers becomes too strong, and even CO 2 can no longer be adsorbed between the carbon layers ( Figure 8g). Moreover,t he narrowness of the carbon-layer slit pores also leads to excellent low-pressure adsorption capability due to the overlap of pore-wallp otentials. The shrinkage of ultramicropore width is further supported by the fact that the pseudo-irreversibility due to kinetic restrictions of CO 2 adsorption, which is visible in CO 2 sorptioni sotherms ( Figure S6), increasesw ith increasing carbonizationt emperature. If as tronger interaction between functional groups and CO 2 molecules were responsible for the pseudo-irreversibility,t he relation would be inverse.
Since the kinetic diameters of many technically relevant gases are in the range of slit-pore width and this carbon interlayer distance can be adjusted by choosing an appropriate carbonization temperature, it appears possible to tailor the material for many gas separation applications.

IAST selectivity calculations
Ac ommon application for CO 2 adsorbents is the separation of CO 2 from flue gas, which contains8 -13 %C O 2 and 71-73 %N 2 .
To evaluate the performance of the CNFs with respect to this application,a dditional gas adsorption measurements with N 2 Figure 7. Influence of the nitrogen functional groups on the CO 2 adsorption capacity on electrospun, PANd erived carbon nanofibers at 0.05 bar.
were performed at 273 K. The adsorption isotherms of N 2 and CO 2 were simulated by using aT óth model (Table S3), which providesa na ffinity constant for the interaction between adsorptive and adsorbent. The much lower affinity constant for N 2 adsorption than for CO 2 adsorption indicates ah ighs electivity towards CO 2 .F urthermore, the affinity constantf or CO 2 decreases with increasing carbonization temperature, which is ah int that the selectivitym ay decrease as well.
By combining the Tóth fit results of CO 2 and N 2 adsorption obtaineda t2 73 K, it is possible to calculate adsorption selectivities by the IAST method. The resulting pressure-dependent selectivities for materials carbonized between 600 and 1100 8C are plotted in Figure9.F or the materialc arbonized at 600 8C, the calculated IASTselectivity at 20 mbar is as high as 350. At higher pressures the selectivity drops and reaches saturation at 1bar at astill very high value of 132, which is among the highest values reported for carbon materials in the literature so far. When carbonized at higher temperatures, the carbonf ibers exhibit similarb ehavior of the IAST selectivity,b ut with lower values. For example, the IAST selectivity for ac arbonization temperature of 700 8Ca tl ow pressures is as high as 250, whereas for 800 8Ct he materials hows as electivity of 180 at the same pressure. The plateau values at ambient pressure are high (130 and 80, respectively). In strongc ontrast to this, the materials carbonized at 1000 and 1100 8Ce xhibit IAST selectivities of less than 5.
The excellent adsorption selectivity towards CO 2 forc arbonization temperatures of 600 and 700 8C, can be explained by a molecular-sieve effect, whichi ss chematically depicted in Figure 8. Whereas CO 2 can penetrate the carbon interlayer spaces, N 2 molecules are excluded by their size, depending on the carbonizationt emperature of the adsorbent. In fact, the narrowu ltramicropore volume, whichi so nly accessible to CO 2 , shrinksw ith increasing carbonization temperature, whereas the outer fiber surface, which is accessible to both gases,d oes  Owing to the decreasing carbon interlayerdistance, the adsorption of individual gas moleculesi sp revented by size exclusion, that is, by am olecular-sieveeffect. (e) Both Ar and CO 2 fit into the ultramicropores in betweent he carbonl ayers. (f) Ar is excluded from adsorptioni nthe slit pores owingt ot he decreased pore width for carbonization at 800 8C. (g) Both Ar and CO 2 are excluded from ultramicropore adsorption due to afurtherd ecrease in ultramicropore width. not change significantly.H ence, the selectivity towards CO 2 decreasesw ith increasing carbonization temperature. Furthermore, the narrowest pores with high adsorptions electivity are filled at comparatively low pressures owing to the overlap of the adsorption potentials of the pore walls, which leads to an increasei na dsorption energy,w hich may even be enhanced by selectivei nteraction of nitrogen functional groups andC O 2 . With increasing pressure, the rather unselective bare surfacei s covered, andt his results in as ignificantlyh igher selectivity at lower pressures. As imilar trend for the pressure dependence of the IAST selectivity has been observed by Kim et al., [63] whereas Zhang et al. [64] and Wu et al. [65] found ac ontrary result of drastically increasing selectivity with higherp ressures. For additional comparison, the selectivities of previously reported carbonsc an be found in Table S3. However,n oneo ft he nonpost-treated carbons reaches the extraordinarily high adsorption selectivity of the currently investigatedc arbons. The latter are only matched by tailor-made MOFs and very few posttreated carbons, [21,66] which potentially enables the practical application of the electrospun CNFs in gas separation processes with feed gases having relativelys mall CO 2 fraction.

Conclusions
The gas adsorption properties of PAN-derived CNFs have been investigated.P AN fibers were prepared by electrospinning and cross-linking at 250 8C, and carbonized at varioust emperatures ranging from 600 to 1100 8C. In this temperature range three differentt emperature regimes influencing the gas adsorption properties have been identified. From 600 to 700 8C, the resulting carbon materials can adsorb CO 2 and Ar in large amounts. However,A ra dsorption appears to be very slow,o wing to kinetic hindrance. From 800 to 875 8C, the adsorption of CO 2 is still very high, whereas the Ar adsorption capacity decreases drastically.F or carbonization temperatures of 900 8Ca nd above,both CO 2 and Ar adsorption become very low,o wing to the significantly advancing carbonization of the investigated nanofibers.O nt he basis of calculated micropore size distributions, the gas adsorption properties of the CNFs are assumed to be highly dependent on the carbon interlayerd istance, which is af unctiono ft he carbonization temperature and controls the access of gas molecules with different kinetic diameters, such as Ar,N 2 ,a nd CO 2 .O wing to the narrows lit-pore width provided by the carbon-layer interspaces, the CNFs that were carbonizedb elow 900 8Co ffer superior low-pressure CO 2 adsorption capabilitiesa nd extremely high IAST selectivities of up to 350. Furthermore, the influence of functional groups on the CO 2 adsorption properties was found to be less important than that of the carbon-layer slit pores, which act as am olecular sieve for CO 2 .Apositivei nfluence of nitrogen functional groups was found as well. The molecular-sieve effect and the interaction between functional groupsa nd CO 2 make the PANderived CNFs promising materials for gas adsorption and separation applications, especially for low pressures or dilute gas streamso fC O 2 such as flue gas and might be tailored even further, beyond the currently investigated gases and gas mixtures.

Synthesis of CNFs
All chemicals were used as received without further purification. The CNFs were prepared by electrospinning of as olution containing 10 wt %P AN in DMF.I natypical synthesis, DMF (72 g, 99.8 %, VWR Chemicals, Germany) was added to PAN( 8g, M w = 150 000, BOC Science, USA). The mixture was stirred at ambient temperature for 3d until ac lear solution was obtained. The solution was electrospun in an electrospinning device equipped with ar otating drum collector (IME Medical Electrospinning, The Netherlands) under constant climatic conditions of 25 8Ca nd 30 %r elative humidity in horizontal orientation. The solution was supplied by as yringe pump with af low rate of 2.4 mL h À1 and pumped through a spinning needle of 0.8 mm inner diameter.T he needle was moved laterally on an automated spinneret in ar ange of AE 60 mm from the central position with as peed of 20 mm s À1 and at urn delay of 500 ms. The acceleration voltage was 25 kV and the needle-to-collector distance was 150 mm. The rotating drum collector had ad iameter of 60 mm and ar otation speed of 1500 rpm. Electrospinning was performed for 6h with ac orresponding solution volume of 14.4 mL.
After electrospinning, the resulting PANf iber mat was cut into pieces and dried in air for 1h at 150 8C. Then, the PANp olymer chains in the fibers were cross-linked in air at 250 8Cf or 15 h. Subsequently,t he fibers were carbonized for 3h in Ar atmosphere at constant temperatures ranging from 600 to 1100 8C. The heating rate of the tube furnace was 300 Kh À1 ,a nd the cooling rate was 200 Kh À1 .

Material characterization
XPS measurements were performed with aP hi5000 VersaProbe II (ULVAC-Phi Inc.,USA). For the individual measurements, monochromatic 1.486 keV Al Ka radiation was applied. Peak analysis was performed by using CasaXPS with Shirley-background and instrumentspecific corrections. The spectra were calibrated at the C1ss ignal to 284.4 eV.
For the elemental analysis av ario EL cube elemental analyzer (Elementar,G ermany) was employed. 2mgs amples of each fiber material were burned in CHN mode, and 10 mg samples in Om ode. In CHN mode, the samples were burned and the combustion products were separated and detected. In Om ode the samples were treated in reductive atmosphere, in which O-containing fragments were converted to and detected as CO. This process was performed three times in both modes for each sample. For the sample of the material that was carbonized at 900 8C, polyethylene was added for better combustion.
SEM investigations were performed with aQ uanta FEG 650 microscope (FEI, USA). For each image, an acceleration voltage of 20 kV was used in combination with an Everhart-Thornley detector.F or the measurements, small strips of the materials were applied to the sample holder by using ac opper band for additional fixation and improved electrical conductivity.
TEM images were obtained with aT itan instrument (FEI, USA). The samples were prepared by ultrasound-mediated dispersion of the CNFs in ethanol.
Gas adsorption measurements were performed with an Autosorb iQ 2i nstrument (Quantachrome, USA), which was equipped with a cryocooler (CTI-Cryogenics, USA). The samples were prepared by ChemSusChem 2020, 13,3180 -3191 www.chemsuschem.org 2020 The Authors. Publishedb yWiley-VCH Verlag GmbH &Co. KGaA, Weinheim cutting fiber mats into strips of 1 5mm, 50-100 mg of which were transferred to ag lass sample tube. Subsequently,t he samples were degassed for 8hunder vacuum at 300 8C. For exact determination of the sample weight, the sample tubes were weighed three times in the empty state without sample. After loading with as ample the degassing process was performed and the filled tubes were weighed again three times. From the difference of the average of both masses, the sample weight was determined.
Gas adsorption measurements were performed with Ar (5.2, Air Liquide, France) at 87 Kf or ag eneral pore analysis. CO 2 (4.5, Air Liquide, France) adsorption measurements were performed at 273 Kt oi nvestigate microporosity and to evaluate CO 2 adsorption energies and selectivities. The isotherms were evaluated with regard to the surface area by the BET method. [67] Data evaluation according to the DR equation [57] was performed in the relative pressure range from 0.002 to 0.005 with a b value of 0.39. The pore size distributions were obtained by using the simulation methods provided by the measurement software Quantachrome ASiQWin 5. For Ar at 87 KaQSDFT equilibrium model (quenched solid state density functional theory,A ro nc arbon, slit pores) was used, whereas for CO 2 at 273 Ka nM Cm odel (Monte Carlo, CO 2 on carbon, slit pores) was employed. Selectivity calculations according to IAST were performed with the 3Psim software (3P instruments, Germany). Adsorption isotherms for N 2 (5.2, Air Liquide) were measured at 273 K. The resulting isotherms as well as the CO 2 adsorption isotherms were interpolated by using aT óth isotherm model. [68]