Polyethylenimine covalently grafted on mesostructured porous silica for CO₂ capture

Abstract

Polyethylenimine (PEI) has been grafted on a 2D hexagonal mesostructured porous silica of MCM-41 type (LUS silica) using 3-glycidoxypropyltrimethoxysilane (GTMS) as a grafting agent to develop sorbents for CO₂ capture. The advantage of this tether is to create ethanolamine units upon reaction of the epoxy group with the amine functions of PEI. Two synthetic routes have been explored: (1) reaction of GTMS and PEI and then grafting on a calcined MCM-41 silica (M-1), and (2) grafting of GTMS on the silica and then reaction with PEI (M-2). In both cases, the grafted solids are well structured according to the XRD patterns. The amounts of glycidoxypropylsilane (GS) and PEI are 14 and 9 wt%, and 21 and 16 wt%, respectively, for samples M-1 and M-2. The CO₂ adsorption capacity of both materials has been tested at 303 K and 101 kPa and compared to a bare LUS silica sample impregnated with 25 wt% PEI (M-3-25). Samples M-1 and M-2 containing ethanolamine groups show higher CO₂ adsorption capacities, with loading of about 150 and 134 mgCO₂ / gPEI (36 and 43 mgCO₂ / g-sorbent), respectively, while the CO₂ adsorption capacity was about 55 mgCO₂ / gPEI (14 mgCO₂ / g-sorbent) for the impregnated solid M-3-25.

---

Introduction

The rapid rise of temperature on Earth observed since the beginning of the industrial age has yielded a noticeable increase of the sea level, enlargement of deserts, modifications of infectious disease vectors and climatic perturbations. Global warming is arguably attributed to the increase of the CO₂ concentration in the troposphere (from 280 ppm in pre-industrial times to ∼400 ppm in 2005) due to anthropogenic CO₂ emissions, mostly driven by fossil fuel combustion.¹

CO₂ recycling or remediation is now recognized as a societal priority. Technically, it involves three steps: capture from the emission source, transportation and long-term sequestration or reuse when possible.² Apart from research efforts focusing on various solid adsorbents such as zeolites, pillared clays, activated alumina and carbons, hydrotalcites and metal–organic frameworks,^3,4 loading amines on porous materials also appears a promising strategy for sorbent design. Ordered mesostructured silicas (OMS) are interesting candidates for this aim on the basis of their large surface areas (∼1000 m² g⁻¹) and rich surface chemistry ascribed to surface silanol groups.

The simplest method for incorporating amines on mesoporous silica is by direct wet impregnation from a solution of an amine polymer (e.g., polyethyleneimine or PEI and tetraethylene-pentamine or TEPA) and further solvent evaporation. The amine polymer is stabilized by formation of hydrogen bonds with surface silanol groups, which facilitates its distribution throughout the pore volume. This approach, generating the so-called ‘molecular baskets’,^5–11 is however limited due to the fact that a significant amount of amine groups can be sterically hindered within the pore volume and form agglomerates between particles, both aspects affecting negatively CO₂ diffusion within the amine-impregnated porous matrix.

A second method for incorporating amine groups is by direct co-condensation during the synthesis.^12,13 In this one-pot synthesis approach, an aminosilane, the surfactant (cationic or anionic) and the silica precursor are mixed together and aged to allow hydrolysis and condensation of the silica precursor. This method offers the important advantages of mitigating pore blocking and providing homogeneous amine distributions,^14–17 translating into higher CO₂ loadings (see for instance the comparative studies reported by Yokoi et al.¹⁸ and Wang et al.¹⁹). Nevertheless, the degree of mesoscopic order of the solids decreases with the aminosilane concentration in the reaction mixture, leading ultimately to totally disordered solids. This method also requires the use of extraction techniques for surfactant removal to preserve the organic moieties.

Finally, amines can also be covalently grafted on a silica support avoiding surface polymerization. Unlike impregnated silicas, covalently tethered amine adsorbents avoid amine leaching, offering high regenerability through repeated adsorption/desorption cycles. A first possibility to graft amines on silica is via a two-step process involving a functionalisation step with alkylhalide groups followed by a S_N2 reaction with an amine precursor.²⁰Amine groups can also be directly grafted in one-pot synthesis via condensation of primary, secondary or tertiary aminoalkyltrialkoxysilanes with surface silanol groups.^17,21–26 The final amine loading is limited both by the surface area, the number of accessible silanol groups on the support, the alkyl chain and the synthesis conditions. The incorporation of amine groups on mesoporous silicas usually involves a reduction of the pore size (and consequently the surface area). Nonetheless, the pore size can be increased through post-synthetic incorporation of pore-expanding agents (e.g., dimethyldecylamine, DMDA), tuning the pore size in MCM-41 materials in the range 3.5-20 nm.²⁷ The presence of partially occluded non-ionic surfactant (e.g., polypropylene oxide) in amine-impregnated silicas has been reported to promote the CO₂ adsorption capacity by dispersing the guest species, as well as by stabilizing CO₂ intermediates via specific interactions between CO₂, amine and hydroxyl groups belonging to the surfactant.⁸

This study is devoted to the formulation of synthetic strategies for controlling the location of PEI on a 2D hexagonal mesoporous silica with MCM-41 structure (LUS silica). A tether (3-glycidoxypropyltrimethoxysilane, GTMS) has been used for the first time to favour irreversible PEI anchoring and inhibit its migration along the porous surface, as an attempt to promote gas diffusion. This synthetic approach also presents the advantage of forming ethanolamine units by ring opening of epoxy groups in GTMS after reaction with the primary amine groups of PEI (see Fig. 1), mimicking the conventional solvent used for chemical absorption in the Girbotol process.

Fig. 1 Molecular structure of PEI and GMTS.

Experimental

Materials

Polyethyleneimine (PEI, Aldrich, MW = 800 Da, 19 amine groups/molecules), 3-glycidoxypropyltrimethoxysilane (GTMS; 98% Abcr), hexadecyltrimethylammonium-p-toluene-sulfonate, (CTATos; > 99% Merck), Ludox HS-40 (40% SiO₂; Aldrich), sodium hydroxide (Acros), toluene (99%, Acros), cyclohexane (99%, SDS–Carlo–Erba), acetone (99%, Aldrich), ethanol (96%, Elvetec), methanol (99.9%, SDS–Carlo–Erba), and HCl 1 M were used as received. The sodium silicate solution was prepared as follows: Ludox (187 mL) was added to sodium hydroxide (32 g) in deionised water (800 mL) and stirred at 313 K until obtaining a clear solution.

LUS synthesis. A solution of CTATos (6.4 g) in deionised water (231 mL) was stirred at 333 K for 1 h. In parallel, the sodium silicate solution (160 mL) was stirred at 333 K for 1 h and added dropwise into the CTATos solution. The mixture was stirred at 333 K for 2 h and then statically heated in a 500-mL Teflon-lined stainless steel autoclave at 403 K for 20 h. The resulting white precipitate was filtered, washed with deionised water (3 × 100 mL) and dried overnight at 383 K, leading to 9 g of as-synthesized LUS-Asilica. Elemental Analysis (EA): C(30.4%), H(6.4%), N(1.7%), S(0.39%), residual weight at 1273 K (56.05%). Further details are provided in the ESI.†^28,29.

Surfactant removal. The removal of the surfactant present in the as-made silica was performed in two consecutive steps: (i) HCl treatment and (ii) calcination. HCl treatment: LUS-A (2 g) was placed in an erlenmeyer, and then ethanol (50 mL, 96%) and HCl (3.1 mL, 1 M) were added. The mixture was stirred at 313 K for 80 min. After filtration and washing with ethanol (3 × 20 mL) and acetone (2 × 15 mL), the solid was dried overnight at 353 K. LUS-H (1.1 g) was obtained. Calcination: LUS-H (0.34 g) was calcined in an horizontal oven at 823 K for 6 h under 30-mL(STP) min⁻¹ N₂ flow using a heating ramp of 10 K min⁻¹, obtaining 0.30 g of calcined solid LUS-C. EA of LUS-C: C(< 0.3%), N(< 0.1%), residual weight at 1273 K (96.6%). The BET surface area of LUS-C was 1290 m² g⁻¹ after vacuum treatment at 393 K for 9 h.

Synthesis of M-1. PEI (0.114 g) and GTMS (0.135 g) in methanol (1.50 g) were stirred at 333 K for 1 h in a covered flask. The PEI amount introduced to the flask corresponded to ca. 1 eq. of a monolayer covering the external surface of the calcined LUS-A silica, while the GTMS amount corresponded to ca. 4 eq. of PEI or 44% of the total surface. The obtained solution of modified PEI was then introduced into a Schlenk containing 0.25 g of calcined silica (LUS-C) previously dehydrated at 403 K under vacuum, and then cooled down to 333 K under Ar flow. The resulting mixture was stirred at 333 K for 15 min under Ar to favour grafting on the external surface and on pore mouths. The solvent was then evaporated under vacuum and the resulting solid was washed with methanol (40 mL). After filtration, the solid was dried at 333 K for 3 h, obtaining material M-1 (0.34 g). EA of M-1: C(12.6%), H(3.1%), N(4.7%), residual weight at 1273 K (64.6%).

Synthesis of M-1′. Material M-1′ was prepared as M-1. The only difference was the introduction of an additional step where the modified PEI was stirred at 295 K under Ar for 2 h before rising the temperature to 333 K to allow grafting onto LUS-C, yielding 0.39 g of solid M-1′. EA of M-1′: C(21.3%), H(4.8%), N(8.2%), Si(24.5%), residual weight at 1273 K (58.1%).

Synthesis of M-2. A solution of GTMS (0.135 g) in methanol (0.50 g) was stirred at 333 K for some minutes and then injected into a Schlenk containing 0.25 g of calcined silica (LUS-C) previously dehydrated at 403 K under vacuum, and then cooled down to 333 K under Ar flow. The resulting mixture was stirred at 333 K for 15 min under Ar. A solution of PEI (0.114 g) in methanol (1.0 g) previously heated at 333 K was then introduced into the reaction mixture and stirred at 333 K for 1 h. After solvent evaporation under vacuum, the resulting solid was washed with methanol (40 mL), filtered and dried at 333 K for 3 h, obtaining material M-2 (0.295 g). EA of M-2: C(19.7%), H(4.4%), N(7.8%), Si(25.0%), residual weight at 1273 K (56.8%).

Synthesis of M-3-X (X = 25, 50 and 75). A solution of PEI (0.33, 1.0 or 3 g) in methanol (4.0 g) was stirred at 333 K for 15 min and then injected into a Schlenk containing 1.0 g of calcined silica (LUS-C) previously dehydrated at 403 K under vacuum and then cooled down to room temperature. The PEI amounts correspond in each case to ca. 50, 100 and 150% of the internal volume of calcined silica. The resulting mixture was stirred at room temperature for 30 min under Ar. The solvent was then evaporated under vacuum obtaining materials M-3-25, M-3-50 and M-3-75, respectively. EA of M-3-25: C(12.7%), H(3.6%), N(7.77%), residual weight at 1273 K (68.20%); EA of M-3-50: C(26.8%), H(6.4%), N(16.0%), residual weight at 1273 K (44.9%); and EA of M-3-75: C(37.5%), H(8.8%), N(23.1%), residual weight at 1273 K (24.6%).

Synthesis of M-4. A solution of GTMS (0.135 g) in methanol (1.50 g) was stirred at 333 K for some minutes and then injected into a Schlenk containing 0.25 g of calcined silica (LUS-C) previously dehydrated at 403 K under vacuum and then cooled down to 333 K under Ar. The resulting mixture was stirred at 333 K for 15 min under Ar. After solvent evaporation, the solid was washed with methanol (40 mL), filtered and dried at 333 K for 3 h, obtaining material M-4 (0.194 g). EA of M-4: C(5.59%), H(1.55%), N(< 0.1%), Si(37.49%), residual weight at 1273 K (85.59%).

Analytical techniques

XRD: low angle X-ray powder diffraction experiments were carried out on a Bruker (Siemens) D5005 diffractometer using Cu-Kα (λ = 1.5418 Å) monochromatic radiation. FT-IR: Fourier-transformed infrared spectra were recorded from KBr pellets using a Mattson 3000 FT-IR spectrometer. ICP-AES: inductively coupled plasma atomic emission elementary analyses were carried out on a Perkin Elmer M1100 equipment with preliminary silica dissolution using HF acid. NMR: Solid ¹H and ¹³C-NMR analyses were performed on a Bruker DSX400 apparatus using TMS as reference. TGA measurements were collected from Al₂O₃ crucibles on a DTA-TG Netzsch STA 409 PC/PG instrument under 30 cm³(STP) min⁻¹ airflow in the temperature range 290–1273 K using a heating ramp of 10 K min⁻¹. TGA-MS analyses were performed on PtRh 10% crucibles using a Setaram, Setsys Evolution 12 TGA-TDA instrument coupled to a Pfeiffer Omnistar 2006 mass spectrometer (1–200 uma) under 30-cm³(STP) min⁻¹ air flow in the temperature range 290–973 K using a heating ramp of 10 K min⁻¹. N₂ sorption/desorption isotherms at 77.4 K were measured using a Micromeritics ASAP 2010M volumetric apparatus on solids outgassed overnight under vacuum (10⁻⁴ mbar) at 353–523 K (the BET specific surface was determined in the pressure range 0.05 < P/P₀ < 0.20, while the mean pore size was measured using the Kruk equation by implementing the pore volume and d₁₀₀ interplanar spacing³⁰). CO₂ sorption isotherms were measured at 303–333 K using a BELSORP-Hp apparatus on solids outgassed under vacuum (∼10⁻⁴ mbar) at 323 K for 12 h.

Results and discussion

XRD patterns

All the solid samples present a regular mesostructure irrespective of the PEI grafting method. Three peaks can be observed in all cases (Fig. 2 & Fig. S1, ESI†), which can be assigned to <100>, <110> and <200> reflections based on a hexagonal unit cell, indicating the presence of a 2D hexagonal channel array. The unit cella₀ value calculated from the d₁₀₀ spacing (a₀ = 2d₁₀₀/√3) is only slightly modified upon organic functionalisation (a₀ = 4.5–4.7 nm, Table 1) compared to the parent silicaLUS-C (a₀ = 4.6 nm). The thickness of the silica walls is about 1.6 nm.

Fig. 2 XRD patterns of samples LUS-C, M-1, M-2, M-3-25 and M-4.

Table 1 Textural properties of mesoporous materials deduced from N₂ sorption isotherms at 77.4 K

Material	a 0 ± 0.1 (nm)^a	d 100 HWHH ± 0.01 (2θ/°)^a^b	d p ± 0.01 (nm)^c	S BET ± 5% (m^2 g^−1)	V p ± 5% (cm^3 g^−1)^d	Accessible volume (%)^e	Inaccessible volume (%)^e
a determined from XRD measurements from (100) peaks. b half with at half height for the d100 (in 2θ). c mean pore size computed using the Kruk equation (in parentheses, values measured using BJH). d cumulated pore volume at P/P0 = 0.98–0.99 e %Accessible volume computed as Vp/0.76 × 100 (0.76 cm^3 g^−1 is the pore volume of LUS-C at P/P0 ∼0.7), and %Inaccessible volume computed from the difference 100−%Accessible vol.−%vol. PEI + GS using a density of 1 g cm^−3 for loaded PEI and GS (weight loadings indicated in Table 2).
LUS-C	4.6	0.34	3.9 (3.6)	904	0.81	100	0
M-1	4.3	0.42	1.5	80	0.054	7	63
M-1′	4.5	0.35	1.4	53	0.046	6	41
M-2	4.5	0.32	1.3	42	0.04	5	46
M-3-25	4.5	0.37	2.7	426	0.22	29	39
M-3-50	4.7	0.34	0.8	6.6	0.011	2	34
M-3-75	4.5	0.29	—	—	—	0	8
M-4	4.5	0.30	3.4 (2.7)	600	0.50	66	18

---

Material M-1 exhibits a lower a₀ value, i.e. 4.3 nm, consistent with a shrinkage of the structure. This observation, already reported for PEI-impregnated mesoporous silicas ⁷, is likely due to the modification of the surface tension upon direct interaction of PEI oligomers with silanol groups through hydrogen bonding. In addition, the half width at half peak height (HWHH) for the d₁₀₀ reflection (Table 1) is larger for sample M-1 (0.42°) than for the other materials (HWHH = 0.30–0.35°). This difference may arise from a higher heterogeneous dispersion of PEI–GTMS units inside the silica channels for this sample. A narrower peak for M-1′ is consistent with a better dispersion of PEI–GTMS as a result of the pre-mixing step at 295 K before grafting (see above). Another striking observation concerns the intensity of the <100> peak, which depends on how the pores are filled by PEI. The intensity is expected to decrease when PEI completely fills the channels, decreasing the XRD contrast. However, a partial filling by a polymer layer should maintain a relatively high contrast. The evolution of the <100> peak intensity should therefore be correlated to the effective organic loading. This point will be addressed later.

The amount of glycidoxypropylsilyl (GS) and PEI groups on the solids measured from chemical and TGA analyses (Fig. 3 & Table 2), indicate that sample M-1 presents the lowest PEI and tether loadings (14 and 9 wt%, respectively) compared to samples M-1′ (25 and 15 wt%) and M-2 (21 and 16 wt%). These results suggest that both PEI and tether diffusion in the channel network is more difficult to perform in a one-pot reaction than the sequential route, resulting in a higher loading when the reactants are left in contact for 2 h before heating for sequential grafting. In addition, using the sequential route, similar PEI and GS loadings (21 and 16 wt%, respectively) are achieved in material M-2 (Table 2). Finally, the impregnated sample M-3-25 shows a PEI loading similar to that obtained for sample M-2, i.e. the 24 wt% (Table 2). M-3-25 sample will be considered hereinafter as a reference to establish a proper comparison with PEI-grafted sorbents.

Fig. 3 TGA (left) and DTG (right) patterns on samples M-1, M-1′, M-2, M-3-25, M-3-50, M-3-75 and M-4.

Table 2 PEI and GS loadings measured by TGA and CO₂ adsorption capacities for the different materials prepared in this study

Material	PEI ^a wt% ± 1	GS^a wt% ± 1	CO2 adsorption capacity (101 kPa)^b		Henry's constant (× 10^2)^b^c
			(mgCO2 / g-sorbent)	(mgCO2 / g-PEI)	(mol g^−1kPa^−1)	(mol gPEI^−1kPa^−1)
a PEI: polyethylenimine, GS: 3-glycidoxypropylsilane. b measured at 303 K. c values computed from the derivate of the adsorption isotherm at P→0 using spline fitting functions.
LUS-C	—	—	23	—	0.601	—
M-1	14	9	36	150	2.70	11.2
M-1′	25	15	35	80	2.69	6.14
M-2	21	16	43	134	4.22	13.2
M-3-25	24	—	14	55	2.66	10.5
M-3-50	49	—	43	86	8.13	14.3
M-3-75	70	—	77	102	1262	1672
M-4	—	12	15	—	0.458	—

---

FT-IR spectra

To further characterize the nature of the organic moieties in PEI-functionalised silicas, some FT-IR characterization tests were carried out on both PEI-impregnated and PEI-grafted samples. Upon PEI impregnation in LUS-C, new IR bands appear at about 1066, 1132, 1309, 1365, 1465, 1560, 2823 et 2939 cm⁻¹), the intensity increasing linearly with the polymer loading (M-3-X, X = 25, 50, 75, Fig. S2, ESI†). These bands are also present in the spectra obtained on PEI-grafted samples and can be attributed to the IR fingerprint of the raw PEI polymer (Fig. 4). The two bands appearing at about 2900 cm⁻¹ and in the range 1400–1600 cm⁻¹ can be unambiguously attributed, respectively, to stretching vibrations of C–H and C–N bonds, i.e. ν(C–H) and ν(C–N).

Fig. 4 FT-IR spectra in the range 350–1850 cm⁻¹ (on top) and DRIFT spectra in the range 2400–4000 cm⁻¹ (on bottom) on samples LUS-C, M-1, M-1′, M-2, M-3-25 and M-4.

In the case of the GS tethered sample M-4, only one band at about 2900 cm⁻¹ can be clearly distinguished from the silica IR spectra ascribed to the stretching vibration of C–H bonds (Fig. S3, ESI†). This bands overlaps with that belonging to PEI after grafting for samples M-1, M-1′ and M-2. FT-IR does not appear therefore to be a specific technique to characterize GS groups, since no specific IR bands ascribed to epoxyde groups can be clearly visualized (this group is better characterized by ¹³C-RMN, see below). Note however that the band appearing at about 960 cm⁻¹ decreases after grafting for samples M-4, M-1 and M-1′, this observation being coherent with a reduction of the number of surface silanol groups. Moreover, this band does not completely vanish after PEI grafting for sample M-2, reflecting that GS groups are not detached in the sequential grafting route.

The complex band appearing between 1000 and 1300 cm⁻¹, corresponding to the asymmetric stretching vibration ν_a(Si–O) of SiO₄ tetrahedra, is slightly modified for samples M-1, M-1′ and M-2 compared to the parent LUS-C silica. The bands are red shifted and the shape changes, which might be ascribed to the presence strong hydrogen bonds between PEI and the silica surface. The band centred at 1650 cm⁻¹ can be attributed to the bending vibration of adsorbed water, δ(H₂O), decreasing upon GS grafting, i.e.M-4, since the surface becomes less hydrophilic due to a decrease of the number of silanol functions. However, upon PEI incorporation (M-2), this band recovers a marked intensity, since PEI favours water adsorption on the solid despite a reduction of the number of silanol groups.

¹³C-NMR

Fig. 5 shows the ¹³C-RMN spectra obtained on samples M-4 and M-2. The spectrum corresponding to sample M-4 confirms the attachment of GS groups on the silica walls, reproducing the whole collection of chemical displacements for the different C atoms obtained on GMTS (not shown). Note also that the peak centred at about 50 ppm, which should be more intense due to the presence of three methoxy groups, actually shows a lower intensity than for the other peaks, confirming the GMS attachment on the silica support. Fig. 5 shows the assignation of the different possible chemical displacements for the different C atoms in the grafted GS groups.

Fig. 5 ¹³C-NMR spectra on samples M-4 and M-2.

In the case of material M-2, one can remark the variation in the form and intensity of the peaks appearing at 56 and 73 ppm compared to those observed on sample M-4, both peaks appearing simultaneously in the GS and PEI¹³C-NMR spectra. The original peak appearing at 44 ppm in sample M-4 shows a practically negligible intensity after PEI grafting, while two new additional peaks appear at about 39 and 70 ppm, being ascribed, respectively, to CH₂–NH₂ and CH–OH units. These latter two peaks provide unambiguous evidence of the reaction between the grafted GS and PEI according to an oxacyclopropane ring opening mechanism described in Scheme S1, ESI.†

N₂ adsorption/desorption isotherms at 77.4 K

Fig. 6 plots the N₂ adsorption/desorption isotherms at 77.4 K on samples LUS-C, M-1, M-2, M-3-25, M-3-50 and M-4, while Table 1 collects the most significant textural properties of these solids. The genuine LUS-C support exhibits a Type IV isotherm pattern typical of mesoporous silicas with no hysteresis loop and possesses a large pore volume (0.81 cm³ g⁻¹) and a high specific surface (904 m² g⁻¹). Material M-4, consisting of grafted GS tether on silica (12 wt%), presents a net decrease of both the pore volume (0.50 cm³ g⁻¹) and the specific surface area (600 m² g⁻¹). The reduction of the phase transition pressure (P/P₀) from 0.32 for LUS-C to 0.28 for M-4 due to narrow-sized mesopore filling (or secondary micropore according to the IUPAC) due to significant interactions with the pore walls is compatible with a reduction of the mean pore size from 3.9 to 3.4 nm (see Table 2). The steep step in the adsorption isotherm of M-4 solid is consistent with an homogeneous distribution of the GS tether, most likely lying down the channel mouths.

Fig. 6 N₂ adsorption/desorption isotherms at 77.4 K on samples LUS-C, M-1, M-2, M-3-25 and M-4.

In the case of PEI-impregnated samples, a mere impregnation of 24 wt% PEI on LUS-C leads to material M-3-25 with a neatly smaller pore volume of 0.22 cm³ g⁻¹ and a specific surface area divided by two (S_BET = 426 m³ g⁻¹). In the case of samples M-3-50 and M-3-75, with 49 and 70 wt% PEI, respectively, the pore volume and the surface area become practically suppressed (≤ 0.01 cm³ g⁻¹ and ≤ 7 m² g⁻¹, see Table 1 & 2, isotherms not provided). This result suggests PEI migration into the channels and complete filling beyond 50 wt% PEI loading. Note that at 24 wt% PEI loading, i.e. half of the loading necessary to achieve full pore filling, the pore volume has decreased ca. 4 times instead of 2 compared to the original pore volume of LUS-C (0.81 cm³ g⁻¹). A more accurate evaluation may be obtained assuming that the bare silica support, LUS-C, keeps its structure upon impregnation and that all PEI fills exactly the pore channels in M-3-50 (i.e. no PEI occupies external positions). Accordingly, the volume occupancy by PEI in M-3-25 would be 32%, leaving a 29% of pore volume accessible to N₂ at 77.4 K and the remaining 39% inaccessible, even at P/P₀ ∼1 (see Table 1). These values have been computed from the pore volume measured at P/P₀ ∼0.7 on the LUS-C sample (about 0.76 cm³ g⁻¹) and taking a PEI density of 1 g cm⁻³. Assuming again full pore occupancy by PEI in sample M-3-75, 30% of the PEI would lay inside the pores, while 70% would be located outside the mesoporous grains. Finally, the step at P/P₀ = 0.21 in sample M-3-25 shows that the pore diameter (∼2.7 nm) is rather regular and that PEI forms a rather homogeneous ∼0.8-nm thick layer on the channel surface accessible to N₂. The apparent density of loaded PEI in samples M-3-25 and M-3-50 computed as the ratio weight loading of PEI and pore volume reduction (see Table 1 & 2) is, respectively, about 0.44 and 0.65 g cm⁻³ (lower than the value of 1 g cm⁻³ used in the estimation of the inaccessible volume to N₂).

In the following, the modification of the porosity upon the introduction of both GS tether and PEI will be discussed. Fig. 7 illustrates the different proposed PEI coverage schemes that can be achieved on these materials. In material M-1, where the tether is covalently linked to PEI before grafting to LUS-C, the isotherm still shows a Type IV isotherm pattern with two steps on the adsorption branch for P/P₀ about 0.20 and 0.90. Desorption occurs with a shift towards higher adsorbed volumes all along the desorption branch leading to a marked hysteresis loop (see Fig. 6). Note also a first hysteresis loop starting at P/P₀ > 0.85. Below this value, the adsorption and desorption branches match together with a quasi-constant upward shift and exhibit roughly the same step height. This result suggests that N₂retention does not occur in the channels, but rather within intergrain voids partially blocked with PEI. Note also that the plateau in range P/P₀ = 0.2–0.7 is larger than that observed for LUS-C. Assuming a loaded GS + PEI density about 1 g cm⁻³, the estimated GS + PEI volume occupancy in sample M-1 is about 30% with accessible and inaccessible volumes of about 7% and 63%, respectively (see Table 1). Knowing that PEI is not totally localized inside the silica channels, accessible and inaccessible volumes should be regarded as slightly overestimated.

Fig. 7 PEI coverage schemes inside silica channels and on the external surface for samples M-1, M-1′, M-2, M-3-25 and M-3-50.

Material M-1′, differing from M-1 only by a pre-mixing step before heating to favour PEI migration within the channels during grafting, also shows a Type IV isotherm pattern with a marked hysteresis loop due to N₂retention at higher pressures. The main difference resides on the higher N₂ sorbed volume at high pressures and on a higher desorption pressure (0.94), leading to a less pronounced hysteresis loop. This result is consistent with a lower amount of PEI located within intergrain voids than for material M-1, showing larger windows through which N₂ can desorb. The volume occupancy also suggests a large internal volume fraction inaccessible to N₂ in sample M-1′ (about 41%, see Table 1). In both samples M-1 and M-1′, PEI tends to cover uniformly the channel surface, leading to a similar apparent pore size of 2.1 nm. This apparent size is similar to that measured for the impregnated material M-3-25. Such agreement might be attributed to the formation of a 0.8-nm thick PEI monolayer. This hypothesis can also be supported by the marked reduction of the <100> peak intensity in the XRD pattern for both materials (see Fig. 1).

Material M-2 corresponds to another situation where PEI is grafted on material M-4, namely after grafting the GS tether on LUS-C. The resulting N₂ adsorption/desorption isotherms on material M-2 are similar to those observed on sample M-1′, with similar plateau heights and upward shifts of the desorption branch. Nonetheless, the N₂ adsorption capacity at P/P₀ > 0.94 is smaller (−30%) and pore filling begins at a higher pressure (P/P₀ ∼0.26). These observations suggest that, unlike sample M-1′, PEI locates more preferentially within intergrain voids leaving a large internal volume fraction inaccessible to N₂ (at least 46%, see Table 1). Contrary to that material, the condensation step in sample M-2 is completed at a higher pressure (P/P₀ ∼0.29, apparent pore size about 2.5 nm). This size is close to that measured for sample M-4 with the surface only covered by the tether (2.7 nm). This observation suggests a virtual absence of PEI coverage on the inner accessible volume of the sample. This latter case also suggests that PEI reaction on the grafted tether leads to broader PEI segregation on the external surface.

Finally, note the upward shift of the desorption branch down to the lowest pressure for materials M-1, M-1′ and M-2, suggesting N₂ forcing into the silica channels pores at P/P₀ > 0.8 and further retention when the pressure decreases. This behaviour is most likely due to a slow N₂ diffusion within the PEI plugging barrier in 1D channels. Note however that the amount of N₂ retained during desorption represents less than 7% of the N₂ adsorbed in the accessible volume.

CO₂ adsorption isotherms

Fig. 8 plots the anhydrous CO₂ adsorption isotherms of solids LUS-C, M-1, M-1′, M-2, M-3-25 and M-4. Two adsorption patterns can be clearly distinguished: (1) physical adsorption on surface silanol groups for samples LUS-C and M-4, where the CO₂ loading shows a quasi linear trend with pressure, and (2) physical adsorption on amine groups for the rest of samples (grafted or impregnated with PEI), showing a characteristic Type I (pseudo-Langmuir) adsorption pattern indicative of stronger interactions between acidic CO₂ and alkaline amine surface groups.

Fig. 8 CO₂ adsorption isotherms at 303 K for PEI-grafted samples LUS-C, M-1, M-1′, M-2, M-3-25 and M-4.

Physical adsorption on surface silanol groups provide higher CO₂ adsorption capacities than for amine groups, especially in the case of sample M-4, but at the expense of a lower CO₂ affinity (lower Henry's constants). Although the samples grafted or impregnated with PEI show much lower CO₂ adsorption capacities due to their reduced pore volumes, their affinity to CO₂ becomes promoted (higher Henry's constants) due to the formation of stronger bonds between CO₂ and PEI or grafted ethanolamine groups, probably through the formation of surface carbamates.

In the case of impregnated M-3-50 and M-3-75 samples, despite the different impregnated PEI amounts, the CO₂ adsorption capacities remain comparable (see Table 2 and Fig. 9). This observation suggests that the adsorption capacity of PEI-impregnated solids does not depend much on the PEI loading. Note however the sharp adsorption trends obtained at low pressures, reflecting the formation of strong bonds between CO₂ and amine groups belonging to PEI.

Fig. 9 CO₂ adsorption isotherms at 303 K for PEI-impregnated samples M-3-25, M-3-50 and M-3-75.

Table 2 compares the adsorption capacities per gram of adsorbent and per gram of PEI at 303 K and 101 kPa obtained on the different materials prepared in this study. Higher CO₂ adsorption capacities are obtained on systems containing ethanolamine units: 150 and 134 mgCO₂ / gPEI (36 and 43 mgCO₂ / g-sorbent), respectively, for samples M-1 and M-2 compared to 55 mgCO₂ / gPEI (14 mgCO₂ / g-sorbent) for the impregnated sample M-3-25. Sample M-2 shows the highest CO₂ adsorption capacity per gram of adsorbent, but displays a lower capacity per gram of PEI than for sample M-1 containing more external PEI. Although PEI-impregnated samples M-3-50 and M-3-75 show high CO₂ adsorption capacities per gram of adsorbent, this capacity decreases when computed per gram of PEI due to their extremely low accessible volume for gas diffusion. The surface is only slightly grafted by PEI in sample M-3-50 and the amount of CO₂ adsorbed per gram of PEI is rather modest. In the case of sample M-3-75, about 30% more PEI is grafted on the external surface, which explains the different CO₂ capacities between this sample and M-3-50.

More detailed CO₂ isotherms have been obtained in the temperature range 303–343 K for sample M-1 showing the highest affinity of CO₂ to PEI (see Fig. S4 & S5, ESI†). At low pressures (<100 kPa), all the materials show similar curves at the three temperatures, the adsorption capacities only being slightly promoted at 323 K. Beyond this pressure, the materials show markedly higher adsorption capacities at 323 K than at 343 and 303 K; the isotherms show a quasi-linear increase of the adsorption capacity, similar to that observed for sample M-4 owing to the presence of free surface silanol groups. This hybrid behaviour has also been observed by Sayari & co-workers³¹ on monoamine-grafted PE-MCM-41 and modelled using a binary Langmuir–Freundlich isotherm equation including two contributions accounting, respectively, for physical adsorption on the bare support and chemical adsorption on amines. These trends suggest that CO₂ adsorption on silanol groups is affected by the accessibility of CO₂ to the active centres, most probably due to partial pore blocking by PEI.

Comparing the CO₂ adsorption behaviour of materials M-1′ and M-3-25, the former shows higher adsorption capacities at 303 K, but lower at 323 K. Although both materials possess a similar PEI loading (25% and 24%, respectively, see Table 2), the N₂ adsorption isotherms at 77.4 K reveal that PEI does not occupy the same positions in both materials. Indeed, in sample M-1′, a higher PEI amount is expected to be located on the external surface blocking channel mouths, inhibiting CO₂ diffusion into the channels at 303 K and in turn its accessibility to the active centres. In contrast, the same PEI amount is distributed on the whole external and internal surface for sample M-3-25.

Conclusions

We have demonstrated the design and engineering of mesostructured silicas for CO₂ capture based on the generation of supported ethanolamine groups. Sample M-1, consisting on one-pot PEI-GS grafting, shows the most promising adsorption capacities computed as mgCO₂ / gPEI due to a better dispersion of the surface amine groups on the internal surface of the mesoporous channel network. Sample M-1′ is less active for CO₂ adsorption owing to a more heterogeneous dispersion of PEI-GS units, providing amine groups with different chemical nature and in its turn different chemical affinity to CO₂, as well as more diffusional restrictions.

The series of PEI-grafted materials M-1, M-1′ and M-2 show a hybrid CO₂ adsorption behaviour between chemical and physical adsorption, the former ruling CO₂ adsorption at lower pressures through interaction with amines and the latter at higher pressures through interaction with residual surface silanol groups. Further studies are necessary to promote the CO₂ adsorption capacities, as well as to study the effect of moisture on the adsorption properties and stability of the materials.

References

1. IPCC Report on Climate Change, R. K. Pachauri and A. Reisinger (ed.), IPCC, Geneve, Switzerland, 2007.
2. IPCC Special Report on Carbon Dioxide Capture and Storage, O. D. B. Metz, H. de Coninck, M. Loos and L. Meyer (ed.), IPCC, Geneva, Switzerland, 2005.
3. S. Choi, J. H. Drese and C. W. Jones, ChemSusChem, 2009, 2, 796.
4. D. M. D'Alessandro, B. Smit and J. R. Long, Angew. Chem., Int. Ed., 2010, 49, 6058.
5. W. J. Son, J. S. Choi and W. S. Ahn, Microporous Mesoporous Mater., 2008, 113, 31.
6. X. C. Xu, C. S. Song, J. M. Andresen, B. G. Miller and A. W. Scaroni, Energy Fuels, 2002, 16, 1463.
7. X. C. Xu, C. S. Song, J. M. Andresen, B. G. Miller and A. W. Scaroni, Microporous Mesoporous Mater., 2003, 62, 29.
8. M. B. Yue, Y. Chun, Y. Cao and X. Dong, Adv. Funct. Mater., 2006, 16, 1717.
9. M. B. Yue, L. B. Sun, Y. Cao, Z. J. Wang, Y. Wang, Q. Yu and J. H. Zhu, Microporous Mesoporous Mater., 2008, 114, 74.
10. J. C. Hicks, J. H. Drese, D. J. Fauth, M. L. Gray, G. G. Qi and C. W. Jones, J. Am. Chem. Soc., 2008, 130, 2902–2903.
11. J. J. Wen, F. N. Gu, F. Wei, Y. Zhou, W. G. Lin, J. Yang, J. Y. Yang, Y. Wang, Z. G. Zoub and J. H. Zhu, J. Mater. Chem., 2010, 20, 2840.
12. G. Xomeritakis, C. Y. Tsai, Y. B. Jiang and C. J. Brinker, J. Membr. Sci., 2009, 341, 30.
13. A. Stein, B. J. Melde and R. C. Schroden, Adv. Mater., 2000, 12, 1403.
14. T. Yokoi, H. Yoshitake, T. Yamada, Y. Kubota and T. Tatsumi, J. Mater. Chem., 2006, 16, 1125.
15. L. Wang, L. Ma, A. Wang, Q. Liu and T. Zhang, Chin. J. Catal., 2007, 28, 805.
16. S. Huh S, J. W. Wiench, J.-C. Yoo, M. Pruski and Y. S. Lin, Chem. Mater., 2003, 15, 4247.
17. M. H. Lim and A. Stein, Chem. Mater., 1999, 11, 3285.
18. S.-N. Kim, W.-J. Son, J.-S. Choi and W.-S. Ahn, Microporous Mesoporous Mater., 2008, 115, 497.
19. T. Yokoi, H. Yoshitake and T. Tatsumi, J. Mater. Chem., 2004, 14, 951.
20. T. Tsuda and T. Fujiwara, J. Chem. Soc., Chem. Commun., 1992, 1659.
21. V. Zelenak, D. Halamova, L. Gaberova, E. Bloch and P. Llewellyn, Microporous Mesoporous Mater., 2008, 116, 358.
22. N. Hiyoshi, K. Yogo and T. Yashima, Microporous Mesoporous Mater., 2005, 84, 357.
23. P. J. E. Harlick and A. Sayari, Ind. Eng. Chem. Res., 2007, 46, 446.
24. G. P. Knowles, J. V. Graham, S. W. Delaney and A. L. Chaffee, Fuel Process. Technol., 2005, 86, 1435.
25. C. Knofel, J. Descarpentries, V. Benzaouia, S. Zelenak, S. Mornet, P. L. Llewellyn and V. Hornebecq, Microporous Mesoporous Mater., 2007, 99, 79.
26. J. Wei, J. Shi, H. Pan, W. Zhao, Q. Ye and Y. Shi, Microporous Mesoporous Mater., 2008, 116, 394.
27. A. Sayari, Y. Yang, M. Kruk and M. Jaroniec, J. Phys. Chem. B, 1999, 103, 3651.
28. L. Bonneviot, M. Morin and A. Badiei, 2001, WO 01/55 031A1.
29. P. Reinert, B. Garcia, C. Morin, A. Badiei, P. Perriat, O. Tillement, L. Bonneviot, in Nanotechnology in Mesostructured Materials, Elsevier Science Bv, Amsterdam, 2003, vol. 146, pp. 133–136.
30. M. Kruk and M. Jaroniec, Chem. Mater., 2001, 13, 3169.
31. R. Serna-Guerrero, Y. Belmabkhout and A. Sayari, Chem. Eng. J., 2010, 161, 173.

---

Footnote

† Electronic Supplementary Information (ESI) available. See DOI: 10.1039/c2ra01007k/

---