Efficient Selective CO₂ Composite Sorbent from Amino Acid Ionic Liquids and Silica Gel: ²H NMR Spectroscopy Provides Insight on the CO₂-Binding Mechanism and in-Pore Microscopic Viscosity

Graphical Abstract

CO₂ sequestration by Amino Acid Ionic Liquids@SiO₂. A composite sorbent for CO₂ sequestration based on Amino Acid Ionic Liquids (AAILs) immobilized in mesoporous silica gel is proposed. The optimal AAIL loading is derived from kinetic experiments. The CO₂ binding mechanism and its effect on the AAILs microscopic viscosity has been investigated by ²H NMR spectroscopy, ab-initio calculations and sorption experiments.

Abstract

A promising CO₂ sorbent based on the [Emim][Gly] (1-ethyl-3-methylimidazolium glycinate)/silica gel composite has been studied. CO₂ sorption experiments have shown that the optimum loading of [Emim][Gly] ionic liquid in silica gel is 40 wt.%. The 2-step process CO₂ binding mechanism in [Emim][Gly] has been proposed based on the results of sorption experiments, ²H NMR spectroscopy and ab-initio calculations. The impact of CO₂ on the microscopic viscosity and the dynamical melting of ionic liquid has been thoroughly investigated. ²H NMR spectroscopy has revealed that CO₂ strongly binds cation and anion in [Emim][Gly], forcing them to move in a correlated fashion.

Introduction

The problem of the selective CO₂ capture, despite numerous efforts, still has no universal solution. Depending on the specific task, a permanent¹ and a milder, reversible² types of binding materials are regarded. From technological point of view, the reversible type is more convenient, as the active material can be reused. The ionic liquids, which chemically interact with CO₂, are preferable for capturing it from gaseous mixtures, such as flue gases.⁵ Among them, the amino acid ionic liquids (AAILs) attracted a particular interest, due to their high CO₂ sorption capacity, low desorption temperatures, strong resistance to oxidative degradation, and high biodegradability.⁶ AAILs are capable of absorbing CO₂ at 25–80 °C and can desorb captured CO₂ upon heating to 90–150 °C. Moreover, the AAILs based on imidazolium cations, such as [1-Ethyl-3-Methylimidazolium⁺][Glycinate⁻] ([Emim][Gly], Figure 1), were recently reported to be highly attractive systems for one-pot capture and catalytic conversion of CO₂.⁹

Figure 1

Structure of the 1-ethyl-3-methylimidazolium glycinate AAIL ([Emim][Gly]).

Yet, using bulk AAILs is not always practical for applications: partially due to the potential leakage but primarily due to the saturation issues. The main disadvantage of AAILs is their high viscosity resulting in low rates of CO₂ absorption and desorption. Moreover, AAILs suffer from drastic increase in viscosity upon the reaction with CO₂,¹¹ which is usually explained by the formation of complex hydrogen-bond networks.¹² A simple way to enhance the dynamics of the CO₂ capture process is dispersing AAILs in porous media. Several research groups demonstrated that using supported AAILs instead of pure ones allowed achieving much more rapid CO₂ absorption.¹³ Another advantage is that the AAIL confined into pores does no leak out neither during the CO₂ sorption or the regeneration process. Thus, the resulting composite material is more practical to handle compared to the liquid AAIL. The efficiency of such composite sorbents strongly depends on the choice of the porous matrix for the AAIL dispersion. In particular, silica-based mesoporous materials are a good choice of the support, since they have proven to be good dispersing agents for AAILs.^{13, 14, 17}

Such composite sorbent strategy has the advantage to be highly tunable in nature, however, poses multiple questions in regard to the molecular mechanism regulating the process. It is suggested that the main mechanism for CO₂ interaction with AAILs is similar to that of alkanolamines used for aqueous amine scrubbing, but in fact, the exact molecular mechanism in all applications even in the bulk phase is still under dispute.²³

Apart from the mechanism of CO₂ sorption by the dispersed AAIL, there are other fundamental questions on how the CO₂ affects the AAIL in the pores. Is the faster CO₂ kinetics due to the dispersion of the active ingredient (AAIL) and thus to the better accessibility of the amino groups, or is there more to it? What is the optimal percentage of pore filling by the AAIL for the most effective CO₂ sorption by the AAIL? How does CO₂ influence the stability and dynamics of the anion-cation pair and, consequently, the microscopic viscosity?

Several research groups demonstrated that using supported AAILs instead of pure ones allowed achieving much more rapid CO₂ absorption.^{13, 25} Different kinds of support materials have been utilized to obtain AAIL-based solid sorbents for CO₂ capture studies, including silica gel,^{13, 14, 17, 18, 26} ordered mesoporous silica materials,^{14, 19-22} alumina,^{16, 27} titania,¹⁴ activated carbon,^{22, 28} polymethylmethacrylate microspheres (PMMA),^{14, 15, 21, 29} and metal-organic frameworks.³² The analysis of literature data shows that the most effective supports for AAILs are mesoporous materials, e. g. commercially affordable silica gels.

With this in mind, we decided to investigate the effect of CO₂ on physical and chemical properties of the AAIL confined in pores using a model system [Emim][Gly]@Silica gel by combining CO₂ sorption measurements, advanced solid state NMR experiments and DFT calculations. For this study, we have chosen the composite material with the optimum loading of [Emim][Gly] based on CO₂ sorption measurements. State-of-the-art ²H NMR experiments provided the ratio between the absorbed CO₂ and IL content suggesting a 2-step process for the CO₂ binding mechanism in [Emim][Gly] involving two amino groups from different anions. We have also shown that the absorbed CO₂ shifts the melting transition of the IL towards higher temperatures by almost 40 K. In the presence of CO₂, the ions cannot rotate independently anymore and rotate as ionic pairs as a whole, resulting in higher microscopic viscosity.

Results and Discussion

Composite Sorbent Preparation

In this work, the mesoporous silica gel Davisil grade 646 supplied by Grace was used as a support for the 1-ethyl-3-methylimidazolium glycinate AAIL ([Emim][Gly]) produced by Sigma Aldrich. The synthesis procedure was similar to that described in our previous work.²⁶ In short, the synthesis of composite materials was carried out by wet impregnation of the porous support with the solution of the AAIL in ethyl alcohol. Alcohol solutions were prepared with a mass content of AAIL of 20–60 wt.%. The impregnating solution was added to the silica gel in such an amount that the weights of the support and ethanol were equal. After impregnation, the material was dried in an oven at 50 °C to remove the solvent. Further heat treatment was carried out at 100 °C in an inert atmosphere.

The characteristics of the porous structure for the composite materials and the silica support are presented in Table S1. The isotherms of N₂ adsorption/desorption at 77 K and BJH pore size distributions for the silica support and composite materials are presented in the supplementary material (Figures S1 and S2). The silica gel and composite sorbents contain mesopores with a characteristic size in the range from 5 to 25 nm, while the average pore size (D_av) tends to increase with the AAIL weight content (Table S1). The sample with the AAIL content of 60 wt.% shows a deviation from the general trend because its pore volume and specific surface area are higher than those for the sample containing 50 wt.% of [Emim][Gly]. We attribute this effect to a less efficient penetration of the ionic liquid into the silica pores causing a part of the AAIL in the 60 wt.%[Emim][Gly]/SiO₂ sample to remain on the outer surface of silica gel particles.

Figure 2 shows SEM images recorded for the selected composite materials. The SEM images obtained for the 20 wt.%[Emim][Gly]/SiO₂ composite (Figure 2a) show a transparent particle of the composite, for which the primary microparticles of the silica gel are still visible. The particles of the 40 wt.%[Emim][Gly]/SiO₂ material are no longer transparent due to filling the mesopores with the AAIL (Figure 2b). SEM image for the composite material containing 60 wt.% of [Emim][Gly] that show the excessive spreading of the ionic liquid over the surface of the carbon tape (Figure 2c). It confirms that the major part of the ionic liquid in this material is located on the outer surface of the particles.

Figure 2

SEM images for a) 20 wt.%[Emim][Gly]/SiO₂; b) 40 wt.%[Emim][Gly]/SiO₂; c) 60 wt %[Emim][Gly]/SiO₂.

The CO₂ Sorption Kinetics

Figure 3 shows the kinetic curves of CO₂ sorption by the composite materials with different AAIL content and unmodified silica gel. The dynamic sorption characteristics of sorbents are largely determined by the AAIL weight content in the material. The silica support adsorbs a small amount of carbon dioxide and ultimately does not make a significant contribution to the final value of the CO₂ sorption capacity for the composite materials. For the samples containing from 20 to 40 wt.% of the AAIL, the sorption capacity reaches a stationary value after ca. 5 min. A further increase in the mass content of AAIL (≥50 wt.%) leads to a visible decrease in the rate of CO₂ sorption and less effective utilization of the ionic liquid. As a result, after 10 min of CO₂ sorption, the resulting difference in performance of the materials containing 40 and 50 wt.% [Emim][Gly] is rather insignificant. Moreover, the material containing 60 wt.% of [Emim][Gly] shows the lowest values of the dynamic CO₂ sorption capacity among the composites. Notably, the full saturation of silica gel pores corresponds to the loading of 56.5 wt.%. Thus, a part of the AAIL in this material is located not in the pores, but on the outer surface of the silica grains. This is confirmed by the SEM images recorded for the composite material containing 60 wt % of [Emim][Gly] (Figure 2c).

Figure 3

a) Kinetic curves of CO₂ sorption from the gas mixture containing 15 vol% CO₂ at 30 °C by the [Emim][Gly]/SiO₂ composite materials, bulk [Emim][Gly] and unmodified SiO₂ ; b) dynamic CO₂ sorption capacity (a_Σ) of the composite materials after 10 min of the CO₂ sorption process depending on the AAIL weight content; c) amount of sorbed CO₂ (in mol) per 1 mol of [Emim][Gly] (N) for the composite materials after 10 min of the CO₂ sorption process depending on the AAIL weight content.

By the end of the sorption step, the values of the molar ratio of CO₂:[Emim][Gly] obtained for composite materials with [Emim][Gly] content from 20 to 50 wt.% are more or less close to 0.5 mol of CO₂ per 1 mol of the AAIL (Figure 3c). This is generally in agreement with the assumed two-stage reaction mechanism (Equation (1) and (2)). For the sample containing 60 wt.% of [Emim][Gly], the molar ratio after 10 min is significantly lower than for other composite materials indicating that the sorption process was far from completion. The observed effect of slowing down the process of CO₂ sorption is apparently associated with a decrease in the dispersity of the ionic liquid when the silica pores are almost filled which leads to the hindered CO₂ diffusion in the liquid layer.

For the NMR experiments we chose the sample containing 40 wt.% of [Emim][Gly] as it reaches its saturation with CO₂ in short time so we can be sure that after the sorption step CO₂ is homogeneously distributed within the pores and all the AAIL ion pairs are interacting with the absorbed CO₂ in a similar manner.

Probing the CO₂ Binding Mechanism by ²H NMR:

Knowing the optimal composition and the conditions to ensure the homogeneous distribution of the confined CO₂, we are ready to shed some light on molecular mechanism behind its biding. To do so, we apply the solid state ²H NMR spectroscopy coupled with ab initio calculations. Why ²H NMR? In most reported cases, the microscopic evidence of the CO₂ binding is based on the different variant of the FT-IR spectroscopy.³⁴ However, even in conditions of a bulk IL treated with high CO₂ pressure, the spectroscopic fingerprints are not always straightforward for interpretation. When applied to the AAILs supported on silica gel, as shown on Figure S3, the respective lines become strongly broadened and the unambiguous interpretation becomes even more challenging.

The solid-state deuterium NMR, has an advantage of superior selectivity: its spectrum is determined by two measurable parameters: the quadrupole coupling constant DQCC, χ_D=(e²q_zzQ/h), and the asymmetry parameter, η=(q_xx-q_yy)/q_zz, which describes the principal elements q of the electric field gradient (EFG) tensor.³⁸ The DQCC is a measure of the magnitude of the main component of the electric field gradient tensor, q_zz, at the deuterium site while the asymmetry parameter provides information about the symmetry of the electric field gradient. Both spectral parameters are sensitive and raise the possibility to distinguish proton species with different chemical nature and hydrogen bonding state.⁴⁰ Possible mechanisms assume direct interaction of the CO₂ with protons from the anion's amino groups (NH₂) or with the acidic H(2) proton within the cations imidazolium ring (CD). Consequently, if labelled by deuterium, these sites should reflect the possible interaction with CO₂ by a change in the respective DQCC. At the same time, the DQCCs in all possible configurations can be computed by ab initio calculations⁴² which provides solid grounds for CO₂ coordination mechanism. Moreover, for a selected deuteron the EFG tensor orientation is attached to the orientation of the respective chemical bond, CD or ND in the present case. In such a case, the line shape change is determined by the type and rate of its molecular reorientation. A combination of the line shape and spin relaxation analysis gives a very detailed characterization of protic species composition and dynamics within the range of 10⁻⁴–10⁻¹¹ s,^{39, 47} thus providing unique and selective information on the organization of the ionic pairs and microscopic viscosity. These experimental evidences can be then supported by calculations allowing to select between different CO₂ coordination mechanisms.

Both discussed protons position in the cation and the anion can be exchanged to deuterium by mixing the [Emim][Gly] IL with heavy water with subsequent drying procedure. Thus, to prepare samples for ²H NMR measurements, the [Emim][Gly]/SiO₂ composite with optimal 40 wt.% loading was first prepared by the described above protocol and only then exposed to D₂O for 3 hours and subsequently dried overnight at 80 °C. The procedure was repeated 3 times to reach sufficient H/D exchange level. Part of the sample was transferred to the NMR glass cell and degassed under high vacuum conditions at 353 K for 6 hours. The CO₂-free sample (I) was then sealed by the micro torch flame while being kept in liquid N₂. Another part (m=0.2 g) was saturated with CO₂ in a flow reactor at 30 °C using a gas mixture containing 15 % CO₂ and 85 % Ar. The duration of sorption was 1 h. After the material was ready, it was transferred into a 5 mm glass cell connected to a vacuum line. The sample was carefully degassed at vacuum conditions while being kept under liquid nitrogen to preserve the samples composition. After the degassing the CO₂-loaded sample (II) glass cell neck was sealed off by a micro torch, while still keeping the sample under liquid N₂.

The comparison of sample I and II aims to catch the CO₂ influence on the confined IL. However, before making it we need first exclude the possible influence of the porous matrix on the ²H NMR line shape composition: the silica gel contains Si-OH groups on the inner surface of the pores and can be as well visible in the spectrum after the deuteration procedure.⁴⁸ Estimations based on a previous study of a silica gel¹⁴ show, that for the loading of 40 wt.% the overall quantity of the Si-OH groups is at least 20 times smaller compared to the deuterium contained in the confined ionic pairs. Hence, even if present, these species should not pose any disturbances in the ²H NMR spectrum. However, we still need to prove, that our deuteration scheme is valid and that we can regularly detect signals from both cation and anion. To do so we first inspect ²H NMR pattern collected from the sample of the bulk [Emim][Gly] IL treated similarly as sample I.

²H NMR for the Neat [Emim][Gly]

The corresponding ²H NMR line shapes of the bulk neat [Emim][Gly] IL are summarized in Figures 4 and S4. Between 123–243 K the line shape consists of a broad anisotropic spectrum composed from two main signals static signals 1 (Q₁=166 kHz, η₁=0.08) and 2 (Q₂=212 kHz, η₂=0.12) with relative populations ratio P₁/P₂≈2 : 1. The DQCC for 1 is close to values reported for the deuterons in the amino groups in urea⁴⁹ and metal-organic framework with amino-functionalized linkers.⁵²₂ Considering this and the expected stichometry 2 : 1, we can conclude that the second signal 2 reflects the deuteron of the CD group (2-H ring position) of the [Emim] cation. Notably, this is the first direct⁵³ measurements of the quadrupolar coupling parameters of the acidic CD(2) group in the imidazolium-based ionic liquids. Interestingly, this value is notably larger compared to the values measured for imidazolium bound as linker in the ZIF-62 metal-organic framework (MOF), where both CD(2), CD(4) and CD(5) positions are characterized by a QCC value of 184 kHz.⁵⁵ Considering, that CD(4) and CD(5) position in the methylimidazolium in a similar by composition ZIF-8 MOF⁵⁶ are characterized by a QCC of 203 kHz we can state that for imidazolium-based systems the coordination and functionalization of the cation should play an important role. For the ionic liquid [Emim][SCN] we obtained QCC values for the CD(2), CD(4) and CD(5) positions at the imidazolium ring ranging between 180 and 210 kHz in reasonable agreement with what is observed here. The QCC values were derived from a combined experimental and theoretical approach in the bulk liquid phase of the ionic liquid. No specific interaction with the MOF framework had to be considered.⁵⁹

Figure 4

²H NMR line shape of neat [Emim][Gly] ionic liquid as function of temperature: (a) experimental; (b) simulated; (c–d) simulation deconvolution. Here P_anis is relative population of ion pairs still in fully immobilized state and P_iso the fraction of species which are already mobile.

At 253 K, in addition to the anisotropic pattern, an isotropic signal appears, which indicates the gradual dynamical melting occurring in the glassy IL.⁶ At 258 K the spectrum is already purely isotropic and composed of two signals with distinct linewidth. Based on population ratio 2 : 1 we attribute the narrower signal to the anion's amino group, while the broader to the cation's CD group. Above room temperature these signals become narrow enough to be fully resolved due to different chemical shifts, which is expected as CH and NH₂ groups of [Emim][Gly] are well resolved in high resolution ¹H NMR.⁶ Hence we can conclude that both molecular sites of interest are deuterated and ²H NMR line shapes are able to differentiate them.

²H NMR on [Emim][Gly]/Silica Gel Composite

The ²H NMR line shapes of the [Emim][Gly]/Silica gel composite (Sample I) summarized in Figure 5 and S5–S7. The experimental line shape evolution follows the behavior of the bulk IL showing a low temperature region with only static anisotropic signals (T<203 K), intermediate region of coexistence of static and mobile signals (203–263 K) and a fully dynamically melted region with only mobile isotropic signal (T≥273 K). The anisotropic pattern between 143–253 K is composed of two signals, 1 (Q₁=170 kHz, η₁=0.09) and 2 (Q₂=220 kHz, η₂=0.16) with relative populations ration P₁/P₂≈2 : 1. The ratio ‘between the two spectra components and their DQCC's allows us to unambiguously attribute signal 1 with anion ND₂ groups and signal 2 with cations CD(2) groups. As expected, the Si-OD groups of the porous host are not observable in the present conditions due to much lower population. While the amino groups QCCs remain almost unchanged, the CD groups interaction constant increases by 8 kHz relative to the bulk state. Previously, the ab initio calculations of the ionic pair suggested⁶⁰ that the regarded H(2) ring proton is involved into the hydrogen bonding with the anion's oxygen. Hence, the DQCC reflects the strength of the hydrogen bonding within the pair⁴¹ and a larger interaction constant signals that its binding strength is weaker. The present result evidences, that when confined in pores, the ions in the [Emim][Gly] IL form a “soft” pair in contrast with the bulk case scenario.

Figure 5

²H NMR line shape of [Emim][Gly]/Silica gel composite as function of temperature: (a) experimental; (b) simulated; (c–f) simulation deconvolution. Here P_iso is the relative population of ion pairs in the mobile state; the fraction of species which are still frozen P_anis=1P_iso. P₁ is the fraction of the signal from the ND₂ groups of the anion, P₂ – fraction of the signal from the CD groups of the cation.

The dynamical melting is also affected by the confinement as the mobile fraction appears already at 203 K. Notably the melting occurs within the range of ~60 K with dynamical melting temperature of T_m^I=273 K, which is much larger than in the bulk phase. This is in line with the recent reports⁶¹ suggesting a gradual core-to-walls melting model of the ionic liquids confined in pores. The important conclusion from this observation is that when the regarded IL is dispersed within silica gel pores, the ions mobility is greatly facilitated and present even at much lower temperatures compared to the bulk case.

Notably, even in the dynamically heterogeneous state, the mobile fraction P_iso is composed of two signals with distinct line widths and relative populations P₁/P₂≈2 : 1. The co-existence of the isotropic signals can be tracked up to the 403 K. This imply, that the cation and the anion within dispersed IL, at least while the isotropic reorientation is regarded, move independently from each other. This evidence strongly supports the transition to the soft-pair model for the IL confined in silica gel pores.

²H NMR on (0.5CO₂-[EMIm⁺][Gly⁻])@Silica Gel Composite

Generally, it was proposed, that the main mechanism for CO₂ interaction with AAILs is similar to that of alkanolamines used for aqueous amine scrubbing.¹³ Such anionic mechanism is expressed by Equations (1) and 2:

$$ (1)

$$ (2)

In these consecutive reactions, 1 mole of the amino group reacts with 1 mole of CO₂ to form a carbamic acid (reaction 1), and then the generated carbamic acid further reacts with 1 mol of another amino group to form a carbamate anion (reaction 2). For AAILs with a relatively small cation, e. g. tetramethylammonium¹⁴ or 1-ethyl-3-methylimidazolium^{15, 29}, usually 2 mol of AAIL absorb 1 mol of CO₂ upon reaction completion, which is in accordance with the proposed mechanism (reactions 1 and 2). However, the reaction stoichiometry close to 1 : 1 was observed for CO₂ absorption by AAILs with a large trihexyl(tetradecyl)phosphonium cation ([P66614][AA]), so it was suggested that only carbamic acid is formed in this case (reaction 1).⁶⁴ Moreover, Yang et al. have recently demonstrated that CO₂ absorption by phosphonium AAILs proceeded through both 1 : 2 and 1 : 1 reaction mechanisms and that the observed molar ratio depended on the type of the cation and amino acid anion.²³

For the particular case of the [Emim][Gly] ionic liquid an additional cationic mechanism was proposed: the [Emim] cation's ring proton H(2) is highly acidic⁶⁵ and is able to bind CO₂ by forming a carboxyl group. Such cationic mechanism is expressed by the Equation 3:

$$ (3)

Among various AAILs, 1-ethyl-3-methylimidazolium glycinate ([Emim][Gly]) showed one of the best gravimetric CO₂ absorption capacities.^{15, 27} Therefore, it was chosen as an active component for this study. Considering the tunability of “the AAIL in pores” system, it is crucial for the material design to understand the fundamental mechanism of CO₂ capture in the AAIL-filled pores. The results obtained for some [Emim][Gly]-based composite materials indicate that 1 mol of this AAIL stoichiometrically sorbs 0.5 mol CO₂.¹⁵ So, it is suggested that CO₂ interacts with [Emim][Gly] following the two-stage mechanism (Equation (1–2)). However, mechanistic studies which could prove the formation the carbamate anion as the final product of the chemical reaction between the AAIL and CO₂ are present only in bulk AAILs mixtures of [Emim][Gly] and [Emim][Ala]: by using high resolution ¹³C NMR authors have shown the presence of specific chemical shift change pointing the presence of both carbamates and carbamic acid.¹¹ However, the case of pure [Emim][Gly] remained is still not clarified. Similarly, the ILs trapped into the porous material were not studied, in part, because the high-resolution NMR in such conditions is much more challenging to perform.

At this stage, we have enough information to analyze what happens when the dispersed IL is loaded by CO₂. The respective ²H NMR line shapes of the (0.5CO₂-[Emim][Gly])/Silica gel composite (Sample II) are summarized in Figures 6 and S8–S10.

Figure 6

²H NMR line shape of (0.5 CO₂[Emim][Gly])/Silica gel composite as function of temperature: (a) experimental; (b) simulated; (c–f) simulation deconvolution. Here P_iso is the relative population of ion pairs in the mobile state; the fraction of species which are still frozen P_anis=1–P_iso. P₁ is the fraction of the signal from the ND₂ groups of the anion, P₂ – fraction of the signal from the CD groups of the cation.

Qualitatively the temperature evolution of the Sample II line shape mimics the CO₂-free case with static, dynamically heterogeneous, and liquid-like phases. The key aspect is however in the exact numbers: the low-temperature region (T<203 K) shows two static signals 1 (Q₁=167 kHz, η₁=0.09) and 2 (Q₂=220 kHz, η₂=0.15) with relative populations ration P₁/P₂≈2 : 1. In fact, the experimental patterns for samples I and II at 143 K are almost identical. This means, that the interaction of the IL's ions with CO₂ does not change the quadrupolar interaction constants and the nature of the chemical bonds of the labelled sites. This fact rules out the cationic mechanism (3) of CO₂ capture by forming of the carboxyl group with cation, since the deuteron's QCC in the newly formed CD+CO₂=C−COOD group must be different from a QCC of the hydrogen bonded CD group, as the unbonded −COOD has the QCC of 250 kHz. For the same reasons, it rules out the 1-step anionic mechanism (1) ND₂+CO₂=ND-COOD. Thus, only the 2-step mechanism (2) can be valid, i. e. ND₂(1)+ND₂(2)+CO₂=ND(1)-COO⁻+ND₃⁺(2). This fully supports the conclusions based on the reaction stoichiometry. This conclusion is however based on the assumption that the resulting ND(1) and ND₃⁺(2) fragments remain static in the regarded temperature range and have bonds electronic structures similar to the initial amino groups. The formed assumption is easily validated, as the mobility of the amino fragments is well known to be hindered by electronic structure of the C−N bond⁶⁶: in fact, even in relatively unrestricted state of a guest-free amino-functionalized MIL-53 porous metal-organic framework,⁵² the -ND₂ groups within the functionalized linkers remain fully immobile up to circa 400 K due to the high activation barrier. The second assumption can be directly proven by the ²H NMR investigation of the bulk glycine⁶⁷: in bulk state, the solid glycine forms a zwitterionic pair with ND₂(1)+ND₂(2)->ND⁻(1)+ND₃⁺(2) as hydrogen bonded protonic species, which mimic the 2-step mechanism (2). At 230 K these species were shows to be fully static on the ²H NMR time scale with both ND⁻(1) and ND₃⁺(2) deuterons QCC of ~174 kHz, i. e. very close to our expectations. Moreover, the barrier for the ND₃⁺ rotation was reported to be ~23 Kj mol⁻¹ and the specific to C₃ axial rotation line shapes changes were observed only above 253 K. This explains, why we are not able to observe this in our experiments -at 253 K the whole line shape is already averaged by the isotropic reorientation of the ion.

Calculating the QCC of the ND₃⁺(2) deuterons in a glycine zwitterion embedded in an overall neutral complex with a carbamate anion and two emim cations provide a value of Q=171 kHz, very close to the measured data of about 173 kHz. The structure is shown in the SI along with the procedure deriving the DQCC values from calculated electric field gradients. Thus, we can conclude that the ²H NMR line shape analysis result unambiguously support the 2-step mechanism of the CO₂ binding.

While the spectral composition is not showing a direct impact of the CO₂ presence, the dynamical melting does: the isotropic signal still appears already at 203 K but the full transition to an isotropic line shape occurs only at T_m^II=323 K. So, the dynamical melting spans over 120 degrees and is shifted to higher temperatures, which indicates that the CO₂ presence makes the ionic liquid more rigid with less mobile ions. This can be illustrated by plotting the relative population of the mobile and static signal versus temperature (see Figure 7): in addition to the complete melting temperature (T_m), the 50 %-melting point is also shifted up in temperature due to the presence of CO₂, from T₀₅^I=261 K to T₀₅^II=298 K.

Figure 7

Temperature dependence of the relative population of the mobile P_iso(□) and static P_anis=1-P_iso(▪) component of the ²H NMR spectrum for Sample I (in red) and Sample II (in dark blue).

The CO₂ presence also directly affects the ions dynamics: in contrast with the CO₂-free case, the mobile spectral component is now composed by the isotropic signals with almost identical line widths. Analysis of the line widths temperature dependences for both samples using a simple isotropic reorientation model^{39, 68} (k_iso=k_iso0 exp[−E_iso/RT]) yields the following picture: without CO₂ (Figure 8a) the cation (CD) reorientation (E_iso=18 kJ mol⁻¹; k_iso0/(2π)=5×10⁸ Hz) has almost twice larger activation barrier than the anions (ND₂) rotation (E_iso=9 kJ mol⁻¹; k_iso0/(2π)=2×10⁷ Hz). Here, the rotation rate is related to the rotation correlation time as τ_C=1/(k 2π). In the presence of CO₂ (Figure 8b) both cation and anion are characterized by similar rotation rates (E_iso=19 kJ mol⁻¹; k_iso0/(2π)=6×10⁸ Hz), close to the values observed for the cations rotation in the neat IL. Thus, in the presence of CO₂ the confined IL ionic pair moves like a whole. This means, that the interaction with CO₂ counterbalances the confinement effect and the ions in the [Emim][Gly] IL form again a “hard” pair as in the bulk case scenario which renders the IL much more viscous and hinders the transition to the mobile phase.

Figure 8

²H NMR spectra line widths temperature dependences in the Arrhenius coordinates for mobile species without (I) and with (II) CO₂: (a) I – cations (CD groups, black ○) and anions (ND₂ groups, red ○); (b) II – cations (CD groups, black •) and anions (ND₂ groups, red •). The dashed lines show the linear approximation

²H NMR T₁,T₂ Relaxation Time Analysis

Despite the line width temperature dependences already proved that CO₂ affects molecular mobility of the confined ILs, this method provides information only on slow time scale, a more general approach is to probe the spin-lattice (T₁) and spin-spin (T₂) relaxation times temperature dependences of the mobile (isotropic) signal. In the general case, the two deuterium labels should provide two signals with distinct T₁ and T₂. For both samples I and II the signals strongly overlap and are not resolved due to rather broad line widths. However, even in such a case it is possible to derive individual relaxation times if the two signals are characterized by notably different T₁ relaxation times (the so-called T₁-filtration). Experimental analysis showed that for both samples I and II there is only one T₁ relaxation time. The T₁ relaxation time reflects primarily fast (τ_c~(ω_o 2π)⁻¹~10⁻⁹ s) local motions,^{39, 68} which means that in all cases on the fast time scale the cation and the anion move synchronously. The T₂ relaxation time reflect slower, primarily isotropic motions. For CO₂-loaded sample the cation and anion signals fully overlap, hence we can conclude that the T₂ relaxation times are also the same for the two signals. For the CO₂-free sample the two signals have different line widths, different isotropic dynamics and hence expected to have different T₂ relaxation times. The relaxation time is measured and central part of the isotropic spectrum where the most dominant contribution is of the signal from the ND₂ groups of the anion. Hence for Sample I the measured T₂ relaxation time reflects the dynamics of the anion. For the case of simplicity, we further regard the discussed relaxation times of the ND₂ groups of the anion. The corresponding experimental T₁,T₂ relaxation times temperature dependence curves are shown in Figure 9.

Figure 9

²H NMR spin relaxation temperature dependences for [Emim][Gly]/Silica gel composite mobile state: (a) Sample I, without CO₂ – experimental T₁ (▵) and T₂ (▿); (b) Sample II, with CO₂ – experimental T₁ (▴) and T₂ (▾). Simulations are given by solid lines. All experimental data was measured at 61.4 MHz resonance frequency.

In both cases, the relaxation curves follow similar pattern: the T₁ curve shows a clear minimum (marked a on Figure 9a), while the T₂ curve is monotonously increasing with temperature (marked b on Figure 9a) and the two curves are not merging, neither have the same slope. This means that in addition to the slower isotropic reorientation motion the labelled species are involved into at least one fast anisotropic motion. In order to compare the two samples a proper dynamical model is needed. A simple dynamical model based on an isotropic rotation (k_iso) and an anisotropic motion modelled as uniaxial rotation about an axis (k₁) fails to properly describe both the T₁ and T₂ curves: In addition to the motional parameters, the relaxation times fitting depends on the quadrupolar coupling constant⁵⁴ introduced above. The DQCC for the both labelled species were already derived during the line shape analysis, i. e. for the ND₂ QCC can be taken as Q_ND2=166 kHz. For hydrogen bonded deuterons, the QCC can increase in the liquid state compared to a solid case due to certain weakening of the hydrogen bonds.⁵⁴ Yet, for the amino groups, the hydrogen bonding is usually weak and can be assumed as constant within the whole temperature range. The T₁ relaxation curve shows a clear minimum and for a single anisotropic motion the position of the minimum on the vertical axis is governed purely by the values of the QCC. For Q_ND2=166 kHz the relaxation time in the minimum has be T₁^M~4×10⁻³ s, while experimentally it is notably larger T₁^M~10⁻² s. To fit such a minimum, we must assume an unrealistically small DQCC ~120 kHz. As we have recently shown,^{54, 69} such an effect indicates the presence of an additional, faster anisotropic motion. This motion will be reflected in the relaxation curves as an additional minimum at lower temperatures (marked c on Figure 9a). Such motion is usually associated with faster local libration of the deuterated fragment and is modeled here as an additional restricted libration in a cone with rate k₂.⁶⁹ Since the regarded T₁ reflects the dynamics of the whole pair, a reasonable physical interpretation of these motions is the following: the faster motion reflects the strongly restricted local anisotropic libration of the pair, while the slower motion reflects its large amplitude anisotropic rotation, which involves stronger interaction with the neighboring ion pairs. In other words, the k₁-motion reflects the mechanism of the microscopic viscosity on the fast time scale. Such a model provides a perfect description of the experimental data set. The model parameters for both samples are summarized in Table 1 and the schematic representation of the respective motions on Figure 10.

Table 1. Dynamical parameters of the [Emim][Gly] mobility in mobile isotropic state used for the simulation of spin relaxation curves.

	Sample I CO2-free	Sample II CO2-saturated
E1, kJ mol^−1	17	33
k10/(2π), Hz	7×10^10	0.6×10^13
ϑ1, degrees	50^○	50^○
E2, kJ mol^−1	17	17
k20/(2π), Hz	3×10^12	0.8×10^12
ϑ2, degrees	22.7^○	22.7^○
Eiso, kJ mol^−1	10	20
kiso0/(2π), Hz	2.2×10^7	6.2×10^8

• The estimated accuracy for activation barriers is 10 %, for the preexponential factors - 50 %.

Figure 10

(a) Schematic representation of the motional model used for spin relaxation fitting. (b) In the neat IL isotropic rotation of cation and anion are independent. (c) In the presence of CO₂ ionic pair rotates isotropically as a whole.

The fitting result shows that the kinetic parameters for the isotropic reorientation of the anion follow the estimations based on the line widths temperature dependences. When CO₂ is introduced in the system the barrier for this motion increases two-fold and as discussed above is characteristic for both the anion and the cation. The fast local libration is apparently not so sensitive to the CO₂ presence; however, the anisotropic rotation of the whole pair is sensitive to CO₂ and shows an increase in activation barrier of almost two times. In other words, when CO₂ is present, all motions that involve the dynamics of the whole pair become slower due to two-fold increase in the activation barrier. This is a direct observation of the increased microscopic viscosity of the AAIL due to the CO₂ absorption.

On a molecular level, the reasons behind such behavior reflect the mechanism of CO₂ binding: when CO₂ binds to the ionic pair, the formal kinetic radius of the pair is not increasing notably, and in the case of the mechanisms (1) and (3), we should not expect any notable change in mobility. However, for the 2-step mechanism (2) CO₂ binding involves now two anions, or rather two ion pairs, which rotate in a correlated fashion. In such a case, the effective kinetic size of the rotating complex is twice as large, and it natural to expect an increase in the activation barrier for the motion on both the fast and slower time scales.

Lastly, it is worth discussing the presence of the isotropic rotation: for molecular liquids, like water or other ionic liquids, it is expected that in the liquid state, when the isotropic rotation is fast, T₁~T₂. In such a case, the respective rate constant should reflect the molecular mechanism behind the microscopic viscosity. However, in our case, the pore confinement induced a deviation from this scheme as T₂ is drastically lower than T₁, and the respective motion is characterized by ~10⁴ times longer characteristic times. Considering that the activation barrier of this motion is three times smaller than for the local anisotropic rotations, we conclude, that k_iso is not a local tumbling of an individual ion of pair, but rather the result of a correlated collective motion, that effectively results in an isotropic reorientation pattern. Such an instance is not surprising for molecular systems in confinement⁷⁰ and indicates that the ionic liquid layer is partially ordered by the silica gel pores surface. This results in a bi-modal viscosity: the fast one associates with local anisotropic motions, and the slower one associates with collective isotropic mixing of multiple pairs. To our knowledge, despite numerous computational prediction of such behavior,^{63, 71} this is the first direct experimental observation and characterization of the complex microscopic viscosity for ionic liquids confined in porous matrices.

Conclusions

The impact of CO₂ on the physicochemical properties of [Emim][Gly] IL incorporated in silica gel has been investigated. Sorption experiments have shown that the optimal loading of IL in silica gel is 40 wt.%. At this optimal loading ²H NMR experiments have been conducted. The ratio between the absorbed CO₂ and IL content indicates at the two-step process CO₂ binding mechanism in [Emim][Gly] involving two amino groups from different anions. This mechanism is supported by the ²H NMR spectroscopy accompanied by DFT calculations. We have shown that CO₂ shifts the melting transition of IL by 40 K to the higher temperature. It occurs due to the binding of ions by CO₂. As a result, ions cannot rotate independently as in the CO₂-free case. Instead, ion pair rotates as a whole, leading to the larger effective kinetic size of the rotating complex and higher microscopic viscosity.

Experimental Section

Study of CO₂ Sorption

To measure the values of the CO₂ dynamic sorption capacity of the materials, a SETARAM SENSYS TG DSC device for synchronous analysis was used, containing a symmetric balance and a Calvet-type differential scanning calorimeter.

A weighed portion of the sorbent (∼50 mg) was loaded into an aluminum crucible, which was placed in the measuring channel of the calorimetric unit. Alundum was loaded into the reference crucible. The pretreatment of the sample involved exposing it to a helium flow for 4 hours at 100 °C, followed by cooling to 30 °C.

The sorption of CO₂ was carried out at 30 °C from a mixture flow of 15 vol % CO₂ in helium. The flow rate of the gas mixture in the measuring and reference channels was 15 mL/min. The dynamic CO₂ sorption capacity (a, mg(CO₂)/g) was determined as follows:

$$ (3)

where Δm(t) is the weight gain of the sample due to CO₂ ad-/absorption, m₀ is the mass of the dry test sample after the pretreatment.

In addition, we calculated the molar ratio of the sorbed CO₂ to the AAIL in the composite material (N, mol(CO₂)/mol(AAIL)):

$$ (4)

where M(AAIL) and M(CO₂) are the molar masses of [Emim][Gly] и CO₂ (185 and 44 g/mol, correspondingly), and w(AAIL) is weight content of the AAIL in the composite material [wt %].

²H NMR Experiment

The Solid state ²H NMR experiments were performed at Larmor frequency ω_z/2π=61.42 MHz on a Bruker Avance-400 spectrometer, using a high-power probe with 5 mm horizontal solenoid coil. All ²H NMR spectra were obtained by Fourier transformation of quadrature-detected phase-cycled quadrupole echo arising in the pulse sequence (90°_x–τ₁–90°_y–τ₂–acquisition – t), where τ₁=20 μs, τ₂=21 μs and t is a repetition time of the sequence during the accumulation of the NMR signal. The duration of the π/2 pulse was 2.1 μs. Spectra were typically obtained with 50 – 20000 scans with repetition time ranging from 1 to 30 seconds.

N₂ Adsorption Measurements

The parameters of the porous structure for the silica support and composite sorbents were determined from nitrogen adsorption-desorption isotherms measured at 77 K using a Nova 1200e Surface Area and Pore Size Analyzer. Before the measurements, the samples were degassed at 100 °C under vacuum for at least 3 h.

The surface area (S_BET) for the samples was determined using the Brunauer-Emmett-Teller (BET) method. The total pore volume (V_tot) was determined by the amount of adsorbed N₂ at P/P₀=0.99. The average pore diameter (D_av) of the materials was calculated on the assumption that the pores are cylindrical: D_av=4V_tot/S_BET. The pore size distribution for the samples was calculated using the Barrett-Joyner-Halenda (BJH) method.

Scanning Electron Microscopy

The morphology of the silica support and composite sorbents was studied using a field-emission scanning electron microscope (FE-SEM, Regulus-8230, Hitachi, Japan) operating at 30 kV. Before the measurements, granules of a sample were placed on a double-sided carbon tape mounted on a sample holder.

FT-IR Measurements

ATR FT-IR spectra were recorded on an Agilent Cary 600 (Agilent Technologies, Santa Clara, CA, USA) spectrometer equipped with a Gladi ATR attachment (PIKE Technologies, Madison, WI, USA). 40 wt %[Emim][Gly]/SiO₂ sample before CO₂ sorption was heated to 80 °C in Ar flow for 4 h.

DFT Calculation Details

The structures including educts and products of reaction (2) were calculated at the B3LYP-D3/6-31+G* level of theory.⁷⁴ We have used the well-balanced, but small 6-31+G* basis set. It includes polarization as well as diffuse functions. We have shown that this approach is suitable for calculating neutral and ionic hydrogen-bonded clusters. The 6-31+G* basis set is also chosen for better comparison with earlier studies of molecular and ionic clusters.^{59, 78} For the optimized structures, we also calculated the electric field gradients for the determination of the deuteron QCC.

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Acknowledgments

This work was supported by the Ministry of Science and Higher Education of the Russian Federation within the governmental assignment for Boreskov Institute of Catalysis (project no. FWUR-2024-0032 and FWUR-2024-0036). R.L. thanks the Deutsche Forschungsgemeinschaft (DFG) under Grant LU 506/17-1 (Project Number 470038970) for support of this work. Open Access funding enabled and organized by Projekt DEAL.

Conflict of Interests

The authors declare no conflict of interest.

Open Research

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.