Introduction

Curbing CO2 emissions with effective sequestration is among one of the major contemporary environmental and technological challenges1,2. Organic amines in solution or tethered to high-surface area supports are commonly used for the absorption of CO23,4,5. Although the amines impart intrinsic selectivity to these systems, a major drawback is that the solvent and support add unproductive weight to the absorbent. The CO2 absorption capacity (g CO2/g absorbent) is maximized when the amount of solvent and support is minimized. Furthermore, removing anything that is not amine (especially solvent) would reduce the energy demand for regeneration (CO2 desorption). It would be ideal to exploit the intrinsic selectivity of amine-bearing materials by using the lowest amount of support to maximize uptake and avoid the energy demand of solvent heating. Polyethyleneimine (PEI) has a high potential for CO2 absorption since it has one amine group every two carbon atoms. However, high molecular weight branched PEIs absorb CO2 extremely slowly because of their high viscosity. PEIs have been applied in combination with various supports6,7,8,9,10 and in particular very high PEI loadings were attained using silica foam and mesoporous capsules as supports8,9. Alternatively higher PEI loadings (CO2 absorption capacities) could be obtained by cross-linking PEI. The mass of cross-linker used to convert the PEI to an effective CO2 absorbent could be lower than that of an actual support. The cross-linking of PEI with bi-functional reagents is employed in gene transfection using biodegradable bridges to improve transfection efficacy and reduce citotoxicity11,12. However, there is no known work in the cross-linking of PEI for enhanced CO2 capture. Herein, we show how the loading and CO2 absorption performance of PEI can be maximized using C60 as PEI cross-linker. In particular, the resulting composite material, PEI-C60, has excellent CO2 absorption capacity at high temperature and very high selectivity both in the presence of N2 and CH4.

Results and discussion

Initially, PEI-C60 was synthesized with different molecular weights (600, 1,800, 10,000 and 25,000 Da), showing that the CO2 capture performance improved with increasing molecular weight (Supplementary Information). Thus all results presented herein are related to branched PEI 25,000 Da. The preparation of PEI-C60 is readily scalable. A brown precipitate of PEI-C60 is formed upon mixing solutions of PEI and C60 dissolved in chloroform and toluene, respectively, in the presence of NEt3. PEI-C60 is insoluble in water, ethanol and chloroform. The PEI/C60 weight ratio was measured as 73/27 using thermogravimetric analysis, corresponding to a C/N weight ratio of ca. 74/26. This value is comparable to the results from X-ray photoelectron spectroscopy (XPS), 70/30 and elemental analysis, 72/28, which corresponds to about 10 molecules of C60 per molecule of PEI (Mw = 25,000 Da). Covalent functionalization of C60 (rather than physical wrapping) is confirmed by NMR spectroscopy, vide infra. Since branched PEI has many primary amines, it is likely that PEI would react with several nanocarbon molecules resulting in a highly interconnected network. The surface area of PEI-C60 was measured in the order of about 2.7-2.9 m2/g (Supplementary Information).

It should be noted that PEI-C60 behaves completely differently to a physical mixture or its components. PEI is a viscous material, while PEI-C60 is a non-sticky porous composite capable of absorption. The absorption of CO2 on solid C60, 0.002 g/g (g CO2/g absorbent) at 1 bar13, is dramatically lower than that of PEI-C60 (0.2 g/g at the same pressure). PEI-C60 also shows a greater absorption of CO2 than PEI-SWNTs (single walled carbon nanotubes) (0.09 g/g)14 in agreement with a higher loading of PEI on C60 (75% w/w) compared to SWNTs (50% w/w). Additionally, the PEI:C molar ratio of PEI-C60 (1:695) is larger than that of PEI-SWNTs (1:2065) showing that C60 can accommodate more PEI molecules than SWNTs despite its much smaller aspect ratio. The hydrophilic PEI segregates from the hydrophobic surface of the SWNT14, in contrast C60 appears fully internalized in the PEI matrix as inter- and/or intra-molecular cross-linker. The incorporation of hydrophobic centres in the PEI media would possibly force the hydrophilic amine moieties to point toward the surface of the material making the reactions with CO2 more effective. PEI/G-silica (PEI-impregnated graphene-porous silica sheets) also exhibited high absorption capacity, 0.19 g/g15, due, as we speculate, to a comparable hydrophobic-hydrophilic enhancing effect.

The performance of PEI-C60 equates or outperforms those of metal organic frameworks (MOFs) particularly at higher temperatures16,17. Importantly, the performance of PEI-C60 at low pressure and high temperature is better than that of Mg-MOF-74, Mg2(1,4-dioxido-2,5-benzenedicarboxylate), which has exceptionally high CO2 capacity at very low CO2 pressure18,19. The absorption capacity of PEI-C60 is more than twice that of Mg-MOF-74 in single-component CO2. Moreover, PEI-C60 is extremely selective.

The CO2 absorption performance of PEI-C60 in the low pressure range (Fig. 1) shows uptake increasing with temperature with a CO2 absorption capacity (at 1 bar) going from 0.15 to 0.20 g/g at 70 and 90°C, respectively. This is in dramatic contrast with MOFs whose uptake decreases above room temperature. More significant is the uptake at lower pressures. The CO2 uptake of PEI-C60 surpasses that of Mg-MOF-74 (Fig. 1, dashed curve A) for pressures below 0.7 bar at 90°C. Mg-MOF-74 is one of the best MOFs for the adsorption of CO2 at low pressure and high temperature due to the presence of unsaturated Mg sites that have strong affinity to CO216,19,20. As a comparison, without such binding sites the performance of MOF-17720, Zn4O(1,3,5-benzenetribenzoate)2, is poorer (Fig. 1, dashed curve B). In the case of PEI-C60, the amine groups appear to have a higher affinity (reactivity) toward CO2 making PEI-C60 an excellent material for the capture of CO2 at very low pressures (already significant at 0.05 bar) and relatively high temperatures (70–90°C) both ideal properties for the absorption of CO2 from flue gases21.

Figure 1
figure 1

CO2 uptake of PEI-C60.

Comparison of the CO2 uptake of PEI-C60, Mg-MOF-74 and MOF-177 in the low pressure range. The CO2 uptake of PEI-C60 was measured at 70, 80 and 90°C. The dashed curves are for the absorption of CO2 on (A) Mg-MOF-74 and (B) MOF-177 both at 90°C. PEI-C60 outperforms Mg-MOF-74 in the capture of CO2 at low pressure: the uptake of PEI-C60 is twice as much as that of Mg-MOF-74 at 0.1 bar and 90°C. Furthermore, the CO2 absorption capacity of PEI-C60 increases with increasing temperature in striking contrast with MOFs which capacity decreases with increasing temperature.

The absorption performance of PEI-C60 in CO2, CH4 and N2 at various pressures is compared in Fig. 2. PEI-C60 does not absorb any measurable amount of CH4 at pressures up to 50 bar. Practically no N2 is absorbed in the range from 0 to 1 bar. Whereas, PEI-C60 reaches almost its full CO2 absorption capacity, 0.2 g/g, at 1 bar. This very high selectivity has two practical implications. One is related to the capture of CO2 from flue gas21, assuming that the partial pressures of CO2 and N2 in the flue gas are 0.15 and 0.75 bar, respectively, PEI-C60 would absorb about 0.15 g/g of CO2 and 0.0005 g/g N2 at 90°C (inset of Fig. 2). This compares to 0.14 g/g of CO2 and 0.002 g/g of N2 at 25°C for mmen-Mg2(dobpcd), mmen = N,N'-dimethylethylenediamine and dobpdc4− = 4,4'-dioxido-3,3'-biphenyldicarboxylate, an amine-functionalized expanded MOF-74 structure22. Thus, PEI-C60 has promise for capturing CO2 from N2-rich hot flue gases.

Figure 2
figure 2

Selectivity of CO2 uptake of PEI-C60.

Single component CO2, CH4 and N2 uptakes of PEI-C60 at 90°C in the high pressure range. The dashed lines indicate the corresponding uptakes at the typical pressures of natural gas (5 bar CO2 and 45 bar CH4) and flue gas (0.15 bar CO2 and 0.75 N2) in the inset. The outstanding selectivity of PEI-C60 is particularly evident for natural gas where PEI-C60 reaches full CO2 capacity at 5 bar with no significant amount of absorbed CH4 at 45 bar.

A second major implication is important for natural gas sweetening23,24. Of particular interest is CO2 removal at the wellhead where the gas is typically at >50°C. To date physical adsorbent like activated carbons, zeolites and MOFs have not been able to replace in large-scale amine scrubbing solutions because of their lack of selectivity toward the capture of CO2. Selectivity is essential for natural gas sweetening because if the absorbent captures both CO2 and CH4 an extra step must be added in order to recover that part of final product captured in the absorbent. This is not necessary with PEI-C60, the removal of CO2 from natural gas at 50 bar, roughly made of 5 bar CO2 and 45 bar CH4, would give maximum absorption capacity for CO2 of 0.2 g/g at 5 bar and no measurable absorption for CH4 at 45 bar (Fig. 2). This compares to 0.35 g/g for CO2 and 0.1 g/g for CH4 at the same pressures and 70°C for Mg-MOF-74, which would require the recovery of large amounts of CH4 captured by the absorbent following the CO2 removal step25.

The absorption of CO2 by PEI-C60 from mixtures with CH4 and simulated natural gas at atmospheric pressure (Fig. 3) is about 0.15 g/g (after 60 min.), while CH4 is not absorbed. In the case of the two 10% CO2 mixtures, balanced with CH4 alone or CH4, ethane and propane, the two absorption curves are almost identical. The amount of CO2 captured in this case is about 0.08 g/g after 60 min. exposure, more than 50% of what is absorbed in single-component CO2, 0.15 g/g. This is a further evidence of the high affinity of PEI-C60 toward CO2, in fact the capture performance of PEI-C60 is five times better than that expected from a simple proportionality between absorption and dilution factor, i.e., 50% of the maximum capacity from a 10% diluted CO2.

Figure 3
figure 3

CO2 uptake from simulated natural gas mixtures.

CO2 uptake of PEI-C60 from single- and multi-component mixtures of CO2, CH4, C2H6 and C3H8 at 1 atm and 90°C. The decrease of mass in CH4 is likely due to a progressive drying of the absorbent. The superposition of the absorption curves in 10% CO2/90% CH4 and 10% CO2/90% CH4/3% C2H6/2% C3H8 further indicates the lack of interaction of PEI-C60 with hydrocarbons.

The performance of PEI-C60 was also analysed at atmospheric pressure with thermogravimetric analysis using dry and wet CO2. A total uptake of about 0.21 g/g CO2 was measured and confirmed with elemental analysis showing that moisture in the feeding gas does not affect the CO2 capture performance. Moreover, PEI-C60 has a relatively low temperature of regeneration (<90°C) when compared to the amine scrubbing processes (120–130°C), in agreement with what we previously observed with other PEI-modified nanocarbons14,26. PEI-C60 is relatively stable upon cycling maintaining more than 60% of its starting absorption capacity after 100 absorption/desorption cycles at 90°C (Supplementary Information).

The chemical species formed upon absorption of CO2 in PEI-C60 were analysed using nuclear magnetic resonance (NMR). The 13C NMR spectra (Fig. 4a) do not allow a definitive differentiation of the carbamate carbonyl signal from the bicarbonate carbonyl signal that may be present. Two bands are present in all 13C CP-MAS NMR spectra: one with a peak maximum at 50 ppm (sp3 carbons of PEI) and a weaker band with a peak maximum at about 150 ppm (sp2 carbons of functionalized C60). The former band has a shoulder at about 75 ppm consistent with the presence of sp3 nitrogen-substituted carbon atoms on C60, as seen for the sidewall functionalization of SWNTs27,28. A third sharper signal (164 ppm) is also evident in the spectra of samples exposed to wet or dry CO2 (but not in the spectrum for wet N2). Since this signal can be attributed to carbonate and/or carbamate species, we cannot readily determine the relative contributions of these two species in PEI-C60 conditioned in CO2. Fortunately, 15N CP-MAS NMR presents a much more secure way to determine the presence of carbamate in the presence of bicarbonate. The 15N CP-MAS NMR spectra of the PEI-C60 conditioned in N2 and dry CO2 are given in Fig. 4(b). In the 15N spectrum recorded after conditioning in dry CO2, the band at about -347 ppm can reasonably be assigned to PEI amine nitrogen environments, while the signal at about -297 ppm can reasonably be assigned to PEI-NH-COO carbamate species. In the sample conditioned in N2, the only appreciable signal, after more than 80,000 scans, was that of the PEI amine nitrogens. The XPS characterization of PEI-C60 conditioned in wet CO2 also supports the formation of bicarbonate and/or carbamate species (Supplementary Information).

Figure 4
figure 4

Solid state NMR of PEI-C60.

Solid state (a) 1H-13C and (b) 1H- 15N CP-MAS NMR characterization of PEI-C60 after conditioning in wet CO2, dry CO2 or N2.

With PEI-C60, we introduce a new class of materials where specifically selected cross-linkers are used to convert amine-rich compounds into effective CO2 absorbents. The C60 cross-linker can be depicted as the final result of a progressive shrinkage of a carbon support where PEI increasingly loses contact with the scaffold, as this shrinks, to end suspended between single C60 anchoring points. In this way, the amount of support is minimized in order to maximize the amine content and the CO2 absorption capacity. This simple approach redefines the way we think about preparing CO2 absorbents from anchoring amine compounds to a support to making the amine materials self-supporting with the aid of cross-linkers. We propose that the hydrophobic nature of C60 is responsible for the externalization of the hydrophilic amine groups of PEI boosting the absorption performance of the polymer. Accordingly, other cross-linkers could improve this or other critical properties of the resulting composites to achieve further enhanced CO2 capture performance with associated reduction in the cost of materials. These new composites could allow for a more efficient capture of CO2 and, when integrated in sequestration and utilization technologies, for the containment of the adverse effects of CO2 on the environment.

Methods

Materials

All materials were used as received. Fullerene C60 (99.5%) was purchased from Alpha Aesar, polyethyleneimine branched (PEI, Mw = 25,000 Da) and chloroform (≥99.8%) from Sigma Aldrich, toluene (99.98%) from OmniSolv EMD and triethylamine (99%) from Acros. Ar, N2 and CO2 high purity gases were all purchased from Matheson TRIGAS. Certified multi-component CO2 mixtures were obtained from Applied Gas, Inc.

Synthesis

PEI-C60 was prepared by adding a PEI/chloroform solution (1.00-1.20 g PEI in 35 mL CHCl3) to a C60/toluene solution (0.12 g C60 in 150 ml toluene with 6 mL NEt3) while rapidly stirring. A dark-brown PEI-C60 precipitate was formed and filtered on a 0.45 μm pore PTFE filter. The precipitate was washed with excess CHCl3 and transferred to a clean flask where 50 ml CHCl3 was added. The precipitate was bath sonicated for 10 min and again filtered and washed as before. The PEI-C60 precipitate was left drying in air overnight and collected as a clustery/rubbery brown solid.

Equipment

All low and high-pressure gas absorption isotherms were collected with a Setaram PCTPro volumetric apparatus using at least 100 mg of sample. The absorption isotherms at atmospheric pressure were collected with a TA Instrument SDT Q600 thermogravimetric apparatus using at least 5 mg of sample. In this case, the CO2 was used either in dry or wet form. Dry CO2 was prepared using a stainless still bubbler filled with 3 Å molecular sieves (vacuum dried at 250°C overnight) through which the CO2 was passed at room temperature and 50 psig. Wet CO2 was prepared using a stainless still bubbler filled with deionized water (bubbled with high-flow CO2 at atmospheric pressure and slow-flow CO2 at 50 psig for 1 h each) through which the CO2 was bubbled at room temperature and 2 psig. This last procedure was also used to prepare wet N2. All gases were at ambient pressure when in contact with the absorbent. All uptake values are given in the unit of g/g, i.e., in weight of CO2 (in g) per unit weight of absorbent (in g). A Cosctech ECS 4010 Nitrogen/Protein Analyzer was used for CHNO elemental analysis. Linear calibrations obtained with acetanilide standard were used for all elements (R2 > 0.999). The solid state 1H-13C and 1H-15N CP-MAS NMR spectra were obtained at room temperature using a Bruker AVANCE-III spectrometer (50.3 MHz 13C, 20.3 MHz 15N, 200.1 MHz 1H). Chemical shifts are reported relative to glycine defined as 176.46 ppm for the carbonyl carbon29 and -347.58 ppm for the nitrogen30. All of the 13C spectra were obtained with a 1ms contact time, 32.8ms FID with spinal64 decoupling, 5s relaxation delay and with 50 Hz (1 ppm) of line broadening applied to the FID. The number of scans varied: 20,512 with PEI-C60 conditioned in wet CO2 and 16,800 each in the case of N2 and dry CO2. All of the 15N spectra were obtained with a 3ms contact time, 20.5ms FID with spinal64 decoupling, 5s relaxation delay and with 20 Hz (1 ppm) of line broadening applied to the FID. The number of scans varied: 65,000 for dry CO2, 82,000 for N2 and 1648 for ammonium carbamate (ammonium signal at -358.1 ppm, carbamate signal at -300.9 ppm). The rotors were packed in a glove bag filled with the same gas used for the conditioning of the absorbent. The XPS spectra were acquired using a Physical Electronics PHI Quantera SXM equipped with an Al X-ray monochromatic source (Kα 1486.6 eV at 50.3 W) set at a 200 μm beam diameter and a 45° incident angle. The spectra were collected in ultra-high vacuum conditions (~10−9 Torr). The high-resolution spectra were deconvoluted into overlapping peaks using mixed Gaussian-Lorentzian curves after subtraction of a Shirley-type background. The SEM images of the PEI-C60 composites were collected using a FEI Quanta 400 ESEM at an accelerating voltage of 10-20 kV and high vacuum (< 5 × 10−9 Torr). The surface morphology and area of the PEI-C60 composites were characterized using a Veeco Nanoscope IIIA Atomic Force Microscope and a Quantachrome Autosorb-3B Surface Analyzer, respectively.