Exploring Molecular Dynamics of Adsorbed CO2 Species in Amine-Modified Porous Silica by Solid-State NMR Relaxation.

Previous studies on CO2 adsorbents have mainly addressed the identification and quantification of adsorbed CO2 species in amine-modified porous materials. Investigation of molecular motion of CO2 species in confinement has not been explored in depth yet. This work entails a comprehensive study of molecular dynamics of the different CO2 species chemi- and physisorbed at amine-modified silica materials through the determination of the rotating frame spin-lattice relaxation times (T 1ρ) by solid-state NMR. Rotational correlation times (τC) were also estimated using spin relaxation models based on the Bloch, Wangsness, and Redfield and the Bloembergen-Purcell-Pound theories. As expected, the τC values for the two physisorbed CO2 species are considerably shorter (32 and 20 μs) than for the three identified chemisorbed CO2 species (162, 62, and 123 μs). The differences in molecular dynamics between the different chemisorbed species correlate well with the structures previously proposed. In the case of the physisorbed CO2 species, the τC values of the CO2 species displaying faster molecular dynamics falls in the range of viscous liquids, whereas the species presenting slower dynamics exhibit T 1ρ and τC values compatible with a CO2 layer of weakly interacting molecules with the silica surface. The values for chemical shift anisotropy (CSA) and 1H-13C heteronuclear dipolar couplings have also been estimated from T 1ρ measurements, for each adsorbed CO2 species. The CSA tensor parameters obtained from fitting the relaxation data agree with the experimentally measured CSA values, thus showing that the theories are well suited to study CO2 dynamics in silica surfaces.

Introduction

Global warming and associated climate change are a major concern for scientists, politicians, and general public, becoming one of the most important challenges of humankind in the 21st century. Different reports and studies from Intergovernmental Panel on Climate Change, among others, have identified CO2 emissions to be primary responsible for global warming.[1−3] Among many of the identified sources, flue gas emissions have a major role contributing to the rising of CO2 levels in the atmosphere.[4,5] Therefore, its scavenging from post-combustion gases combined with other greenhouse gas removal strategies is compulsory to curtail CO2 emissions.[6,7]

Different materials have been proposed, with amine-modified porous silica (AMPS) sorbents emerging as a promising alternative to the use of liquid amines for CO2 capture.[8−13] These materials show advantageous features for CO2 capture in post-combustion applications, such as high selectivity and capture capacity toward CO2 at low partial pressures even in the presence of moisture. In order to improve the design and optimization of these materials, gathering knowledge on the nature and dynamics of physi- and chemisorbed CO2 species formed is crucial.[14] Several studies have been published in recent years addressing the structure of the different types of CO2-amine adducts formed in amine-modified silicas (AMPS),[15−26] mainly performed by Fourier transform infrared (FT-IR) spectroscopy[15−18] and nuclear magnetic resonance (NMR).[20−30] However, the unambiguous assignment of CO2 species under certain conditions is still a challenge wherein the tandem use of solid-state NMR and density functional theory (DFT) has provided some of the most prolific results.[21−24] These contributions confirmed the formation of chemisorbed CO2 species of several types in CO2-adsorbed materials containing distinct amine loadings, in which the carbamic acid and alkylammonium carbamate are the predominant CO2 species. Recently, our group was able to identify, at least, three new physisorbed CO2 species in SBA-15 mesoporous silica functionalized with a primary amine, 3-aminopropyltriethoxysilane (APTES@SBA-15) combining relaxation and quantitative NMR.[31]

Although the study of CO2 speciation in AMPS has been reported,[18,20,23,24] information on CO2 molecular dynamics properties is still scarce. FT-IR or X-ray diffraction methods are useful to investigate structural features but have limitations when seeking to understand the molecular dynamics of adsorbates confined in porous systems. Our group has recently investigated CO2 dynamics in silicas for the first time by NMR relaxation.[31] Solid-state NMR relaxation techniques have proven to be one of the most valuable approaches, particularly in probing the dynamics of biomolecular systems,[32−37] to observe fluctuations of local environments in adsorbed molecules, often yielding information about local interactions and the motional behavior of molecular moieties.[32,33,38−44] For amorphous materials (like silicas) and non-rigid solids, the correlation times of molecular motions measured by NMR are expected to be in the range of the μs–ms timescale.[39] Therefore, the dynamics of adsorbed CO2 in AMPS may be investigated by rotating frame spin–lattice relaxation (T1ρ) measurements. Recent theoretical developments provide the analytical treatment of T1ρ as a function of the sample rotation frequency to estimate actual correlation times and amplitudes of motion in solids, particularly in the study of slow protein dynamics.[35,37,45,46]

In this article, we take inspiration on the NMR relaxation studies demonstrated in protein dynamics and perform for the first time T1ρ measurements to explore the dynamics of physi- and chemisorbed CO2 species formed at the silica surface, applying the theoretical model proposed by Kurbanov et al.[45] This formalism estimates average 13C–1H distances and 13C chemical shift anisotropy (CSA) parameters. The study of T1ρ allows assessing the correlation times for each 13CO2 species, thus providing a molecular level insight into the dynamics of each CO2 species formed in AMPS materials.

Materials and Methods

Material Preparation

SBA-15 was synthesized according to a procedure reported previously by our group.[20] First, (EO)20(PO)70(EO)20 copolymer (4.0 g; Aldrich) was dissolved in a 1.6 M solution of HCl (126 cm3). Next, tetraethyl orthosilicate (9.1 cm3; Aldrich) was added to this solution with constant stirring. The solution was then stirred at 40 °C for 20 h and subsequently heated at 100 °C for 24 h, under static conditions. Afterward, the solution was filtered, and the obtained solid was washed with deionized water and dried in an oven at 40 °C. The solid was calcined at 550 °C for 5 h with a heating ramp of 1 °C/min. The resulting SBA-15 product was stored in a desiccator for further use.

The calcined SBA-15 was functionalized with a primary amine, APTES (Sigma-Aldrich, purity > 98%). 2 g of SBA-15 was introduced in a closed reflux apparatus connected to a vacuum line and heated to 150 °C for 2 h. After cooling, nitrogen was introduced into the system prior to the opening of the reflux apparatus, and SBA-15 was refluxed with 100 cm3 of dry toluene (Alfa Aesar, 99.8%) containing 9 mmol of APTES for 24 h in a nitrogen atmosphere. The resulting material (APTES@SBA-15) was purified by Soxhlet extraction with dry toluene, to remove the unreacted amino-alkoxysilane, and finally dried under vacuum, at 120 °C for 24 h.

13CO2 Sorption Procedure

The sorption apparatus comprises a laboratory-made high-vacuum line, connected to a turbomolecular pumping station (HiCube 80, Pfeiffer Vacuum), capable of vacuum greater than 10–2 Pa. A borosilicate glass cell was connected to the vacuum line and served as an enclosure for an NMR rotor to allow degassing and heating zirconia NMR rotors up to 300 °C under high vacuum. The heating was performed with a laboratory-made oven connected to a power controller (Eurotherm 3116), and the temperature was measured with a thermocouple. The desired gas was introduced into the system from the canister connected to the vacuum line and the cell. The pressure inside the cell was measured with a capacitance transducer (MKS instruments, Baratron 722B).

All samples of APTES@SBA-15 were packed in zirconia NMR rotors, enclosed into the sorption apparatus, and dried by degassing and heating (150 °C, 3 h, ramp of 2.5 °C/min) under vacuum. After cooling down under vacuum, 13CO2 (Cortecnet, 99 atom % 13C; <3 atom % 18O) was introduced into the system at a partial pressure of 770 Torr and allowed to equilibrate for 4.5 h. Finally, the NMR rotor was closed inside the cell, and only then, the cell was opened to remove the rotor for NMR measurements.

Solid-State NMR Measurements

All 13C NMR spectra were acquired on a Bruker Avance III 400 spectrometer operating at B0 field of 9.4 T, with a 13C Larmor frequency of 100.6 MHz. All experiments were performed on a double-resonance 4 mm Bruker magic-angle-spinning (MAS) probe at a MAS frequency of 10 kHz and under room temperature conditions. Samples were packed into ZrO2 rotors with Kel-F caps. 13C chemical shifts are quoted in parts per million (ppm) from α-glycine (secondary reference, C=O at 176.03 ppm).

The 13C cross-polarization MAS (CPMAS) spectrum, as shown in Figure , was acquired under the following experimental conditions: the 1H π/2 pulse length was set to 3.0 μs corresponding to a radio frequency (rf) of ∼83 kHz; the CP step was performed with a contact time (CT1) of 3000 μs using a 50–100% RAMP shape pulse in the 1H channel and using a 55 kHz square shape pulse on the 13C channel; the recycling delay (D1) was 7.5 s. During the acquisition, SPINAL-64 decoupling was employed at a rf-field strength of 70 kHz. The total number of scans was 256. The multiple cross polarization (multiCP) sequence, as shown in Figure S1, used a total of n = 6 CP blocks, fulfilling the Hartmann–Hahn condition at the rf field strengths of 55 kHz in the 13C channel and 48 kHz in the 1H channel with a contact time (CT2) of 15 μs. The 1H rf field strength was ramped from 90 to 100%. The D1 and the inter-CP-blocks delay (D2) were, respectively, 7.5 and 3 s. During the acquisition, SPINAL-64 was applied at a rf-field strength of 80 kHz. The number of scans was 256.

Figure 1

(a) Schematic molecular representations correspond to the three different chemisorbed species (A–C) found in APTES@SBA-15 silica loaded with 13CO2 (P = 770 Torr) as proposed by previous studies.[22,23] The drawing scheme on the right is a simplified picture of a CO2-filled pore of the silica after adsorption, as proposed by our recent work.[31] The different physisorbed species D, E, and F are represented in black, red, and orange colors, respectively. The chemisorbed species are represented by the simplified structures in white. (b) 13C CPMAS (purple) and MultiCP (blue) NMR spectra of dry APTES@SBA-15 loaded with 13CO2 at p = 770 Torr. The intensity of the resonance associated to the physisorbed species is low in CPMAS but clearly polarized by MultiCP, contributing to the peak at around 125 ppm.

13C T1ρ times were measured using the NMR experiments, as shown in Figure S1. These two different approaches were used depending on the type of adsorbed CO2 species. The conventional method (Figure S1a) was based on CP-MAS[47] for chemisorbed species and a modified version of multiple cross polarization (multiCP,[48]Figure S1b) for the physisorbed fraction. The labels used for the chemisorbed (A–C) and physisorbed (D–F) species are the same as reported in our previous work.[20,22−24,31] Species F was found to exhibit relaxation parameters (T1) compatible with a highly dynamic environment confined in a porous media.[31]

To measure the 13C T1ρ after generating 13C magnetization by CPMAS (species A, B, C, and D) or multiCP (species D and E), the 13C spins were locked in the x,y-plane of the rotating frame for a variable time (τ) (Table S2) by applying a locking pulse. Under these conditions, the total locked magnetization in the transverse plane (Mτ) decays exponentially with a specific time constant T1ρ influenced by the modulation of the transverse relaxation Hamiltonian induced by molecular motions. The remaining Mτ magnetization after different locking-field τ durations is fitted by the equationwhere M0, is the initial locked magnetization in the transverse plane at thermal equilibrium of each 13CO2 species i, n is the total number of 13CO2 species, and T1ρ, is the spin–lattice relaxation time in the rotating frame of each 13CO2 species i. During the application of the locking field, 1H decoupling was not applied to avoid interferences on the relaxation mechanism. In our study, to facilitate the analysis, eq was linearized as

The T1ρ values (Table S3) were extracted by fitting the linearized eq to the experimental data (Figures S4–S9). The estimated fitted errors are below 10%.

The overlapping resonances of the chemisorbed 13CO2 species centered at 164, 160, and 154 ppm were deconvoluted using software ssNake 1.1,[49] keeping constant the width and position of the peaks within rows of the same pseudo two-dimensional experiments for T1ρ measurements (Figure S10).

Estimation of the Correlation Times of Adsorbed 13CO2 Species

To extract the motional correlation times (τ) of physi- and chemisorbed species, T1ρ was measured as a function of the rf locking field strength (ω1). For each measurement, the 13C rf field strength during spin-lock was carefully calibrated using a nutation experiment in glycine. The values of ω1 used for the correlation time study of each adsorbed component were chosen around the condition ω1≈ γBL, where BL is the local magnetic field amplitude driving the relaxation mechanism in the rotating frame. The main interaction driving the T1ρ relaxation is CSA since homonuclear and heteronuclear dipolar couplings are negligible for physisorbed CO2 species E and F due to their high mobility. Therefore, for E and F species, Bloembergen–Purcell–Pound theory (BPP theory)[50] can be applied. The used ω1 frequencies for each CO2 species are listed in Table S2.

The dependence of R1ρ with respect to τc for the fast components E and F falls within the fast motional regime (τc ≪ T2). In this case, the expression for R1ρ is given bywhere γ is the gyromagnetic ratio of 13C, δ is the reduced anisotropy δ = δ – (1/3) (δ + δ + δ), and the rest of the parameters are the same as defined above. For species E, the data obtained at different ω (Table S3) was fitted with eq to extract τc. In the case of species F, weak locking field pulses to measure T1ρ are needed to assess its fast motional regime, which is currently not possible with our NMR spectrometer.

For the chemisorbed CO2 species A, B, C, and D their dynamical behavior falls within the intermediate case (τc ≈ T2). Considering the 13C nuclei of these species are isolated, that is, without covalently bonded protons and that they involve sp/sp2-hybridized carbons with considerable CSA contribution, the most suitable model to explain the 13C relaxation of these species combines CSA and 13C–1H heteronuclear dipolar coupling contributions. The expression used to fit the experimental data is derived from the Bloch, Wangsness, and Redfield theory (BWR theory)[51,52] accounting for the considerations from Kurbanov et al.,[45] in which the MAS frequency and the rf offset contributions are included. In fact, the MAS dependence of R1ρ (1/T1ρ) was crucial to obtain good fittings of our experimental data, which provide molecular dynamic processes in the range of microseconds to milliseconds.[33,38,47,48]

The 13C relaxation rate associated to CO2 chemisorbed species, considering an off-resonant ω1 locking-field and MAS frequency (ωr), is given by

Equation is the sum of the relaxation rates under CSA and heteronuclear dipolar mechanisms. According to Rovó,[45] the R1ρ(off)CSA is given bywithand R1ρ(on)CSA written as

δ is the reduced CSA, ωC is the resonant frequency of the 13C spin in question, and θρ is the off-resonance angle (the angle between the B0 and B1 fields). J(ω) is the spectral density functions where ω can be replaced by different frequency arguments.

Analogous expressions for the heteronuclear dipolar relaxation mechanism can be derivedwithandwhere bIS is the 13C–1H dipolar coupling constant and ωH is the 1H resonant frequency.

Considering the simplest model with a bond vector motion with only one correlation time and an axially symmetric CSA tensor, J(ω) both for the dipolar and CSA relaxation mechanisms can be written as

S2 is the generalized order parameter describing the amplitude of the motion, and it satisfies the condition 0 ≤ S2 ≤ 1, where S2 = 0 and S2 = 1 represent the fully disordered and completely rigid states, respectively.[32,45] In the present work, considering the amorphous character of the silica as fully disordered, condition S2 = 0 has been used. We have used these equations to fit the experimental T1ρ dependence on the ωI spin-lock field. For the case of A, B, and C, the off-resonance (eq ) was used to fit their T1ρ evolution with ωI, while for species D, data were fitted considering only the on-resonance part. As a result, correlation times and bIS and δ values are extracted for these four 13CO2 species.

Results and Discussion

CO2 Speciation after Adsorption in APTES@SBA-15

Figure shows the 13C CPMAS (purple) and multiCP (blue) spectra of APTES@SBA-15. Both spectra display four peaks, corresponding to the chemisorbed and physisorbed CO2 species, which have been addressed in previous studies.[15,16,20,22−24,29,31,53] Nevertheless, we provide a brief overview of their assignment for the sake of simplicity. The peak at ∼125 ppm arises from physisorbed 13CO2 at the silica surface, corresponding to confined CO2 in the pore. The full characterization of the three physisorbed CO2 species, dubbed D, E, and F, contributing to this resonance (125 ppm) has been performed previously using quantitative CSA and T1 relaxation analyses.[31] The other three resonances at ∼154 ppm (species A), ∼160 ppm (species B), and ∼164 ppm (species C) are assigned according to our previous studies,[20−24,31] that is, to carbamic acid species (A, B) and ammonium carbamate species (C). In the next sections, we employ T1ρ to assess correlation times of each CO2 species, which are more facile to monitor for an accurate estimation of the dynamics of the adsorbed CO2 species.

T1ρ of CO2 Species Formed in 13CO2-Adsorbed APTES@SBA-15

The measured T1ρ values of CO2 species A–F are listed in Table . The first clear observation is that T1ρ values for physisorbed species D–F are considerably longer than the values for chemisorbed species A–C. This is expected since the mobility of the former species is higher than the latter ones. The motional averaging of relaxation mechanisms (mainly heteronuclear dipolar couplings and CSA) causes long T1ρ relaxation times for the species D–F. Particularly, long T1ρ is obtained for species E and F (Table ), which according to our previous work[31] are confined CO2 phases with liquid-like and gas-like T1 relaxation rates. CO2 species D is engaged in H-bonds with the chemisorbed CO2 layer formed on the silica surface.[20,23,31] These interactions are sufficient to restrict the mobility of this species opening routes for a more efficient spin relaxation through dipolar and CSA mechanisms. CO2 species F undergoes faster dynamics compared to CO2 species D and E; thus, a much longer T1ρ value is expected for this species (Table ).

Table 1

T1ρ, τc, 13C–1H Dipolar Coupling Constant (bIS) and the Reduced CSA (δ) Values at ω1 = 30 kHz for each CO2 Species Adsorbed in APTES@SBA-15 after Exposure to 770 Torr of 13CO2

13CO2 species	T1ρ/ms	τc/μs	bIS/Hz	δ/ppm
A	4.7(0.10)	162(32)	6949(325)	71.90(12.95)
B	1.1(0.11)	62(2)	15269(258)	74.60(13.43)
C	1.4(0.03)	123(1)	22132(1087)	49.30(8.89)
D	8.1(0.13)	32(4)	3962(169)	57.70(10.38)
E	9.4(0.90)	20(4)		0.20(0.03)
F	44.0(5.00)

As for the chemisorbed CO2 species A–C, larger T1ρ differences are found between A and B/C though. These discrepancies can be explained invoking the strength of interactions and the different molecular dynamics. Because species B and C refer to paired amines,[20,22,23] the density of coupled 1H spins surrounding these species is higher. The magnitude of the heteronuclear dipolar couplings of B and C is presumably larger than that in the case of species A, offering a more efficient 13C relaxation route for B and C.

In addition, A interacts with silanol groups at the silica surface through hydrogen bonds, leading to a strong restriction in motion, which therefore translates in a higher τ value as it will be discussed ahead.

Correlation Times (τc) of the Adsorbed CO2 Species in APTES@SBA-15

Relaxation rates (R1ρ) were evaluated by acquiring T1ρ as a function of the spin-lock field. The resulting curves were fitted (Figure ) using the relaxation models described in Section . In the case of the chemisorbed CO2 species A, B, and C, the data were fitted using eq . For the physisorbed CO2 species D, the on-resonance part of eq was employed, whereas for the physisorbed species E, eq was used to fit the data. Detailed parameters obtained from data fittings including estimated correlation times for each CO2 species can be found in Table and Figure . In the case of CO2 species F, the correlation times could not be retrieved due to the difficulties in measuring T1ρ at different ω1 spin-lock field values. Particularly, when the spin-lock field values are several orders of magnitude higher than the size of the interaction driving the relaxation mechanism, or when the locking field is weak, it renders strongly inhomogeneous pulses.

Figure 2

Plots of the T1ρ values for chemisorbed and physisorbed species A, B, C, D, and E as a function of the spin-lock field ω1. The line represents the best fit, using eqs and 2.4 for species E and A, B, C, and D, respectively. From this fittings τc, bIS and δ values are obtained. The values are presented in Table , Figure , and Table S3.

Figure 3

Plot bars comparing the different parameters extracted from T1ρ analysis among the different formed 13CO2 species. Top left: correlation times (τc) showing the different dynamics between species. Top right: reduced CSA (δ). This parameter is related with the degree of confinement, mobility, and interaction with the silica surface of the different species. Bottom left: heteronuclear dipolar coupling constant (bIS) extracted for the different species. The relaxation model used for species E does not consider heteronuclear dipolar coupling as the relaxation mechanism; therefore, bIS is absent for E. This parameter provides information about the chemical environment and proximity of chemicals groups from the silica surface to the different CO2 species. Bottom right: the average distance 13C–1H (⟨rCH⟩) in Angstroms (Å) for the adsorbed CO2 species. The values are calculated from the fitted bIS . The color scheme is as follows: green, blue, cyan, black, and red for the A, B, C, D, and E species, respectively.

As expected, the shortest τc is obtained for the physisorbed CO2 species D (32 μs) and E (20 μs). The obtained values agree with the mobile character of these species as they are weakly interacting inside the pores. However, the obtained τc values are much longer than the values reported in the literature for CO2 in liquid and gas states (typically in the range of picoseconds to nanoseconds).[54−56] These discrepancies in correlation times between bulk and confined CO2 reveal the dramatic effect on the dynamics of gases adsorbed in porous materials.[57,58] The τc value for D is slightly longer than for E, most likely due to the effect of H-bond interactions hindering the motion of CO2 species D.

As for chemisorbed species A, B, and C, their τc values are 2–8 times longer than that for the physisorbed D and E species, thus highlighting the differences in mobility between both fractions. The comparison of the τc values among the different chemisorbed species reveals striking differences in their molecular dynamics. Species A exhibits the longest τc (162 μs), which is in good agreement with rigid structures reported previously,[22,23] where the formation of either carbamic acid species stabilized by H-bonds with neighboring silanol groups or silylpropylcarbamate species are proposed. The formation of both species implies a strong interaction with the silica surface and, therefore, strong molecular rigidity that translates into long correlation times as found in the present work. Species C (τc = 123 μs) is more rigid than B (τc = 62 μs) since the former might be further stabilized by the formation of H-bond with neighbor silanol groups (Figure ).

NMR Interactions Obtained from R1ρ Analysis

The curve fitting of relaxation rates allows extracting important parameters such as bIS and δ, which provide further structural information regarding the different CO2 species (Table ). This information is complementary to the determined correlation times further aiding to investigate the CO2 structure in confined spaces.

The fitted δ values obtained for all the chemisorbed CO2 species (A: δ = 71.9 ppm, B: δ = 74.6 ppm, and C: δ = 49.3 ppm) are in good agreement with previously measured values reported elsewhere[22] (A: δ = 78 ppm, B: δ = 72 ppm, and C: δ = 52 ppm). The similarity in the δ values for A and B reflects the fact that both possess similar chemical structures (carbamic acid). However, differences in molecular dynamics and structural flexibility exhibited by both species impose different modulation of the CSA relaxation pathways, giving rise to distinct τc values. For C, the estimated δ value is very different compared to both species A and B. This is expected as the CO2 species C contains a carbamate ion pair, instead of a carbamic acid moiety. τc values for species C and A show the same order of magnitude as both species are engaged in additional interactions supposedly with nearby silanol groups from the silica surface.[22,23] In the case of species D, the estimated δ was 57.7 ppm, which matches approximately the value measured previously (δ = 57.0 ppm) by Vieira et al[31] through 13C CSA MAS NMR. A δ = 334.5 ppm has been measured for pure CO2 gas at low temperatures (10–20 K), under vacuum (2–4 × 10–6 Torr)[59] and has been used by other studies to estimate the τc of CO2 confined in microporous solids.[60] The discrepancy between the δ value measured for CO2 species D (a gaseous like CO2 species judging from the τc values) in the present work and the previously reported value[60] can be ascribed to different reasons. First, in our work, we are dealing with mesoporous materials instead of microporous. Second, the temperature and pressure conditions in both studies are different. Finally, the importance of the intermolecular interactions established between the CO2 molecules with the silica surface in AMPS[31] may have an impact.[31] The estimated δ value for the physisorbed CO2 species E was 0.2 ppm (Table ), using eq and considering the CSA interaction as the main relaxation source.

In addition to the CSA interaction, the relaxation model for CO2 species A–D also involves the 1H–13C dipolar couplings as an extra relaxation pathway. Due to the proximity of these species to the 1H spins belonging to the grafted alkylamine chains, 13C spins from species A–D are thus strongly coupled with nearby protons. Therefore, the values of the 1H–13C dipolar couplings for the A-D species with 1H in their vicinity were estimated from the R1ρ curve fittings. For CO2 species A, a bIS of 6949 Hz was obtained, which corresponds to an average 1H–13C distance of 1.6 Å (Figure ). In the case of B and C, the estimated bIS values were 15,269 and 22,132 Hz, accounting for an average distance of 1.2 and 1.1 Å, respectively. These average 1H–13C distances support the proposed models (Figure ) for the chemisorbed species A, B, and C. B and C are species involving paired amines; therefore, their proton densities are higher compared to species A, which involves a single amine residue. The estimated bIS for species D is 3962 Hz (average 1H–13C distance of 2.2 Å), a much lower value compared to A, B, and C as this species is far more mobile (i.e., weaker dipolar coupling).

Conclusions

This work demonstrates that relaxation studies based on T1ρ measurements are a powerful tool to investigate the dynamics of chemi- and physisorbed CO2 species formed in AMPS. We observe that the three chemisorbed CO2 species (A, B, and C) possessing the highest rigidity give rise to the shortest T1ρ values (4.7, 1.1, and 1.4 ms); a CO2 species with higher flexibility corresponding to the weakly interacting physisorbed CO2 species D exhibits an intermediate T1ρ value (8.1 ms); finally, the highly dynamic physisorbed CO2 species E and F possess the longest T1ρ values (10.0 and 44.0 ms).

Furthermore, the dependence of the T1ρ with respect to the locking field was also studied allowing to retrieve correlation times (τc), which provide further information on the dynamics of CO2 confined in AMPS. The variations in τc values could eventually be correlated to the differences in desorption energy for the different species. We believe the knowledge about the dynamical properties of each CO2 species could then be used for the optimization of the regenerability of the sorbent material. The experimental data were fitted using either the BPP theory for the physisorbed CO2 species E or the BWR theory for the CO2 species A–D. This theoretical analysis also allowed us to extract bIS and δ values, providing further insights into CO2 dynamics and CO2 speciation. This analysis was unpractical for CO2 species F due to its fast dynamics, which requires weak rf fields beyond the instrumental capability of a typical solid-state NMR spectrometer. The extracted NMR parameters obtained from R1ρ curve fitting agree well with the ones obtained from previous studies combining solid-state NMR and DFT calculations,[22,23,31] thus consolidating our knowledge concerning the structure of CO2 adsorbed in porous materials.

The obtained τc values corresponding to physisorbed CO2 species D and E (32 and 20 μs) show typical molecular dynamics very close to a viscous liquid (Figure ), whereas chemisorbed species A, B, and C present a higher rigidity and therefore longer τc values (162, 62, and 123 μs), as typically observed in amorphous solids (Figure ). The dispersion in the τc values is also higher among chemisorbed than that in physisorbed CO2 species. The estimated τc value for A indicates that this species has the strongest rigidity, owing to its engagement in multiple H-bonds with silanol groups. Furthermore, we show that the T1ρ relaxation mechanism in the chemisorbed CO2 species is mainly driven by a combination of heteronuclear dipolar and CSA interactions. The largest bIS values estimated for CO2 species B and C suggest they are surrounded by a dense protonated environment. On the other hand, species A possesses a smaller bIS (6949 Hz) value compared to both B (15269 Hz) and C (22132 Hz), thus pointing toward the existence of a less protonated environment surrounding those CO2 species. The estimated bIS value for physisorbed species D (3962 Hz) is also compatible with CO2 molecules weakly interacting with the silica surface.

Figure 4

Dependence of T1ρ on τc, considering the dipole–dipole relaxation mechanism adapted from a study by Farrar et al.[38] Ordinate and abscissa values are only approximate. The dotted lines labeled with numbers represent the correlation times for different matter states: 1 non-viscous liquid, 2 viscous liquid, 3 non-rigid solid, and 4 rigid solid. The dashed lines are the correlation times obtained for species A (green), B (cyan), C (blue), D (black), and E (red).