Quantifying the efficiency of CO₂ capture by Lewis pairs

A microfluidic strategy is used to assess the relative efficiency and thermodynamic parameters of CO₂ binding by three Lewis acid/base combinations.

Abstract

A microfluidic strategy has been used for the time- and labour-efficient evaluation of the relative efficiency and thermodynamic parameters of CO₂ binding by three Lewis acid/base combinations, where efficiency is based on the amount of CO₂ taken up per binding unit in solution. Neither tBu₃P nor B(C₆F₅)₃ were independently effective at CO₂ capture, and the combination of the imidazolin-2-ylidenamino-substituted phosphine (NIiPr)₃P and B(C₆F₅)₃ was equally ineffective. Nonetheless, an archetypal frustrated Lewis pair (FLP) comprised of tBu₃P and B(C₆F₅)₃ was shown to bind CO₂ more efficiently than either the FLP derived from tetramethylpiperidine (TMP) and B(C₆F₅)₃ or the highly basic phosphine (NIiPr)₃P. Moreover, the proposed microfluidic platform was used to elucidate the thermodynamic parameters for these reactions.

Introduction

Anthropogenic carbon dioxide (CO₂) emissions continue to climb to unprecedented levels and have played a key role in global climate change.¹ This worldwide issue has prompted many researchers to explore a wide variety of approaches to both reduce CO₂ emissions and lower CO₂ concentrations in the atmosphere. Efforts targeting the use of CO₂ as a C₁ chemical feedstock for conversion to formic acid, carbon monoxide,² or reusable fuels such as methane or methanol,³ have prompted many studies targeting new catalyst development.⁴ Although these developments offer the potential for disruptive technologies, it is important to note that the capture of CO₂ will be an integral component of any such advancement. A variety of approaches have been explored to capture CO₂ including the use of zeolites, silica gels, aluminas, and activated carbons,⁵ as well as sophisticated metal–organic frameworks (MOFs).⁶

Investigations of the reactions of CO₂ with main group reagents have included a variety of amines,⁷ alkanolamines⁸ amidines, guanidines,⁹ and N-heterocyclic carbenes (NHCs).¹⁰ A decade ago, the use of frustrated Lewis pairs (FLPs) to capture CO₂ emerged with the report of Stephan, Erker, and coworkers who described intramolecular and intermolecular B/P-based FLPs for the capture of CO₂.¹¹ Since then, a wide variety of B/N,¹² B/P,¹³ Al/P,¹⁴ and Si/P^12a,15 systems have been shown to capture or effect stoichiometric or catalytic reduction of CO₂. In a very recent development, Dielmann and coworkers described the synthesis of highly basic phosphines, generated by the inclusion of imidazolin-2-ylidenamino substituents.¹⁶ These are the first phosphines to be shown to sequester CO₂ in the absence of the Lewis acid necessary to form an FLP.¹⁷

Although a number of FLP and main group systems have been shown to capture CO₂,^13c the ability to quantitatively compare the efficiency of such systems remains experimentally challenging. Standard batch-scale characterization methods for reactions at the CO₂ gas–liquid interface suffer from long reaction times and are often diffusion controlled.¹⁸ Recently, Kumacheva and coworkers developed a microfluidic (MF) platform for the study of gas/liquid reactions.¹⁹ The MF methodology was validated for the well-studied CO₂ reaction with amine^19b and used small amounts of reagents thus providing fast and cost-efficient access to thermodynamic data for gas/liquid reactions (10–15 min per experiment).

In Fig. 1, a gas and a reagent solution are supplied to two inlets of a MF reactor. At a Y-junction, the gaseous stream breaks up in a periodic manner to generate uniformly sized gas plugs that are separated by liquid segments (slugs). As alternating gas plugs and solution slugs flow through the MF channel, the dissolution of the gas and its reaction with reagents in the solution results in a decrease in the volume of gas plugs with time (or the distance from the Y-junction). Analysis of digitized images of the gas plugs allows for the quantification of gas consumption using the ideal gas law.^19a,20 After a particular time (directly related to distance in the MF reactor), the dissolved reagents and the gas reach equilibrium, and the gaseous plug volume remains constant. This enables the determination of the equilibrium constant of the reaction, and a study of the reaction at different temperatures enables assessment of the thermodynamic parameters, ΔG°, ΔH°, and ΔS°. The validity of this methodology was demonstrated with the study of the sequestration of CO₂ by the FLP, ClB(C₆F₅)₂/tBu₃P.^19a,20

Fig. 1. (a) Schematic depiction of the MF gas/liquid device. (b) Magnified view of the outlined region shown in (a), which shows the shrinkage of the gas plugs as they flow through the channel.

In the present work, this innovative MF approach has been applied to compare the efficiency of CO₂ sequestration in reactions of three Lewis acid–base combinations with CO₂. The prototypical FLP tBu₃P/B(C₆F₅)₃, as well as the FLP derived from tetramethylpiperidine (TMP)/B(C₆F₅)₃, were investigated. In addition, the extremely basic imidazolin-2-ylidenamino-substituted phosphine (NIiPr)₃P was investigated alone and in combination with B(C₆F₅)₃. Although each of these systems is known to bind CO₂ (Scheme 1),^11,12b,17b the present MF study provides qualitative and quantitative comparisons of this CO₂ binding. Such data afford insights that are important for the design of main group systems for CO₂ capture.

Scheme 1. Reactions of Lewis acid–base combinations with CO₂. NR = no reaction.

Results and discussion

The established MF protocol^19a was used to determine the thermodynamic parameters associated with the reaction of CO₂ with combinations of the Lewis acid B(C₆F₅)₃ and one of the Lewis bases, tBu₃P, TMP, or (NIiPr)₃P. Bromobenzene was selected as a suitable solvent due to its low volatility and the solubility of the reagents and corresponding CO₂ adducts. In an initial reference experiment, physical dissolution of CO₂ gas in bromobenzene was characterized by the temporal variation in the concentration of physically dissolved CO₂ (i.e., [CO₂]_dissolved) by analysing the digitized dimensions of alternating slugs of solvent and plugs of CO₂ flowing through the MF reactor. By monitoring the decrease in CO₂ plug volume and applying the ideal gas law^19a (eqn (1) and (2), see ESI, Fig. S1 and S2†), the number of moles of CO₂ transferred from the gas plug to the adjacent liquid slug at time t, n _CO2(t), was determined. The equilibrium concentration of physically dissolved CO₂ (C _tot) was reached after approximately 2 s (Fig. 2).

1

2

n _CO2(t): moles of CO₂ in the plug at time t; P: pressure; V _p(t): volume of CO₂ at time t, R: gas constant (8.314 J mol^–1 K^–1); T: temperature; V _s(t): volume of the liquid solution at time t.

Fig. 2. Variation in total concentration of CO₂ transferred at 293 K from gas plugs to reagent solution slugs plotted as a function of time. The gaps in the data from 1.5 s to 2.0 s result from the exclusion of microchannel bends outside the region of interest (see ESI, Fig. S1†). (a) Plots for the FLP derived from tBu₃P and B(C₆F₅)₃. (b) Plots for the FLP derived from TMP and B(C₆F₅)₃. (c) Plots for the combination of (NIiPr)₃P and B(C₆F₅)₃. For C₆H₅Br alone (), C _tot = [CO₂]_dissolved. Each experimental point represents the average of three experiments conducted under identical conditions, where 300 images were acquired for each experiments with a minimum of 4000 CO₂ plugs.

The addition of either B(C₆F₅)₃, or tBu₃P independently to the solvent had little effect on the equilibrium concentration of CO₂ in the liquid slugs, beyond the dissolution of CO₂ in bromobenzene (Fig. 2a). However, combining B(C₆F₅)₃ and tBu₃P in solution resulted in increased CO₂ uptake, which further increased with elevated FLP concentration. These observations are consistent with the known inability of the individual components to capture CO₂, and the established efficacy with which CO₂ is captured by this FLP. These observations are also consistent with our earlier MF study of CO₂ capture by the related ClB(C₆F₅)₂/tBu₃P FLP.^19a

Investigation of the FLP derived from B(C₆F₅)₃ and TMP revealed that TMP alone in solution is able to sequester CO₂ (Fig. 2b, ), consistent with the known ability of secondary amines to reversibly bind CO₂.^12b It is noteworthy, however, that the concurrent presence of B(C₆F₅)₃ in solution results in a significantly enhanced CO₂ uptake. Again, increasing concentration of the FLP results in increased CO₂ sequestration.

In sharp contrast to tBu₃P, increasing concentrations of (NIiPr)₃P led to increasing capture of CO₂ (Fig. 3). These observations are consistent with the work of Dielmann and coworkers^17b who have demonstrated the ability of imidazolin-2-ylidenamino-substituted phosphines to bind CO₂. In further contrast, addition of B(C₆F₅)₃ to solutions of (NIiPr)₃P inhibited CO₂ uptake beyond the physical dissolution of CO₂ into the solvent ( vs. , Fig. 2c). This result indicates an irreversible reaction of (NIiPr)₃P with B(C₆F₅)₃. Monitoring of this reaction by NMR spectroscopy supports the formation of several products, including the zwitterionic product (NIiPr)₃PC₆F₄BF(C₆F₅)₂ as the major species (Scheme 1, see ESI†). Analogous products have been previously reported for sterically encumbered, basic phosphines.²¹ Presumably, the highly basic nature of the phosphine (NIiPr)₃P prompts this reactivity with B(C₆F₅)₃ and precludes capture of CO₂.

Fig. 3. Variation in equilibrium concentration of CO₂ (C _reacted) plotted as a function of initial reagent concentration at T = 293 K (repeated in triplicate, analysing 300 images with a range of 4000–7000 plugs of CO₂).

The CO₂ uptake caused directly by chemical reaction, C _reacted, was determined by subtracting the [CO₂]_dissolved (for the CO₂–bromobenzene system) from the total equilibrium uptake of CO₂, C_tot, for each reagent. A plot of C _reacted against reagent concentration illustrates the relative efficacy of the reaction of CO₂ with the Lewis acid, Lewis base, or Lewis acid–base combination (Fig. 3).

Using the concentration data allows determination of the equilibrium constants (K _eq) for each system. In eqn (3), [CO₂ adduct] is equal to C _reacted (determined from Fig. 3), the [Lewis acid] and [Lewis base] are calculated directly by subtracting C _reacted from the initial reagent concentrations, and [CO₂]_dissolved is the equilibrium concentration of CO₂ dissolved in bromobenzene (Fig. 2). In this fashion, the room temperature equilibrium constants for CO₂ binding for tBu₃P/B(C₆F₅)₃, TMP/B(C₆F₅)₃, and (NIiPr)₃P were determined and are collected in Table 1. The data shown in Fig. 3 reveal that the FLP systems derived from tBu₃P or TMP with B(C₆F₅)₃ are more efficient at CO₂ capture by 29% and 16%, respectively, than the highly basic phosphine, (NIiPr)₃P. In this context, efficiency is taken to be the amount of CO₂, per binding unit, sequestered by the binding reagents. These data also illustrate that the FLP tBu₃P/B(C₆F₅)₃ is 11% more efficient at CO₂ uptake than the FLP derived from TMP/B(C₆F₅)₃. Thus, although previously reported NMR experiments demonstrated the ability of these systems to bind CO₂, the present MF methodology provides a fast, efficient and high-throughput platform for quantitative ranking of the ability of these systems to bind CO₂ at ambient temperature.

Table 1. Thermodynamic parameters for CO₂ capture determined by the MF method ^a .

Reagents	K eq (293 K)	ΔH 293 b kJ mol–1	ΔS 293 J mol–1 K–1	ΔG 293 kJ mol–1
tBu3P/B(C6F5)2Cl19a	223 M–2	–39.3	–89.3	–14.8
tBu3P/B(C6F5)3	517 M–2	–100.0	–289.3	–15.2
TMP/B(C6F5)3	267 M–2	–73.8	–205.4	–13.6
(NIiPr)3P	4158 M–1	–29.1	–30.8	–20.0

^aAdditional values for 273 K, 283 K, 303 K, and 313 K are deposited in the ESI (Table S2–S4).

^bThe value for ΔH is determined by the corresponding slope in Fig. 5.

The reactions of tBu₃P/B(C₆F₅)₃, TMP/B(C₆F₅)₃, and (NIiPr)₃P with CO₂ were also studied in the temperature range from 273 to 313 K. A plot of the amount of CO₂ captured (C _reacted) versus the concentration of either the FLPs or (NIiPr)₃P was monotonic and linear. Due to the exothermic nature of the reactions,^19a,20 as the reaction systems were cooled, the degree of CO₂ capture was enhanced at each of the concentrations of reagents (Fig. 4). Using the values of K _eq at different temperatures, the corresponding Gibbs free energy for each reaction can be obtained (eqn (4)). The use of a van't Hoff plot (Fig. 5; eqn (5)) allows the determination of the corresponding enthalpy (ΔH°) and entropy (ΔS°) values (Table 1). The linearity of the association between ln(K _eq) and 1/T indicates that enthalpy does not change appreciably within the temperature range investigated. It is noted that the ΔH° of the reaction of B(C₆F₅)₃, tBu₃P, and CO₂, –100.0 kJ mol^–1, is in excellent agreement with the value of –100.4 kJ mol^–1 obtained calorimetrically by Autrey and coworkers.²²

3

ΔG° = –RT ln(K_eq) 4

5

Fig. 4. Variation in the equilibrium concentration of captured CO₂, plotted as a function of initial FLP (B(C₆F₅)₃ and tBu₃P) concentration at T = 273 K, 283 K, 293 K, 303 K, and 313 K (repeated in triplicate analysing 300 images with a range of 4000–7000 plugs of CO₂) (see ESI† for plots for the other systems).

Fig. 5. Plot of ln(K _eq) vs. 1/T (T ranging from 273 K to 313 K).

To gain further insight, theoretical calculations were performed using Density Functional Theory (DFT) at the M11/6-311G(d,p) level of theory.²³ The optimized geometries of B(C₆F₅)₃, tBu₃P, CO₂, and tBu₃PCO₂B(C₆F₅)₃ were computed using the integral equation formalism variant of the polarizable continuum model (IEFPCM) to implicitly assess the effects of solvation by bromobenzene.²⁴ Frequency calculations confirmed that each structure was at a minimum on its potential energy surface and provided partition functions from which thermodynamic parameters were computed. The reaction enthalpy obtained at this level of theory, –176 kJ mol^–1, was significantly larger than the experimental value of –100.0 kJ mol^–1. To further investigate this discrepancy, the calculations were carried out at the B2PLYP-D3/6-311G(d,p) level of theory.²⁵ This double hybrid meta-GGA method includes an empirical long-range dispersion correction and performs well in the evaluation of main group thermochemistry.²⁶ A recent study comparing the ability of M11 and B2PLYP-D3 to evaluate the chemistry of compounds for which dispersive interactions are important found the latter to consistently outperformed the former.²⁷ The internal reaction energy for CO₂ capture by the B(C₆F₅)₃/tBu₃P FLP was computed to be –129 kJ mol^–1 at this double hybrid level of theory. This value more closely approaches the experimental reaction enthalpy. In addition, if this internal reaction energy is used in combination with the reaction entropy obtained at the M11/6-311G(d,p) level of theory, then the Gibbs free energy of the reaction computed at ambient temperature is –15.4 kJ mol^–1, in excellent agreement with the experimental value of –15.2 kJ mol^–1.²⁸

The experimentally determined thermodynamic parameters illustrate that CO₂ binding by the two presently studied FLP systems and the previously studied ClB(C₆F₅)₂/tBu₃P system^19a is less entropically favoured than that by (NIiPr)₃P, consistent with the three-component nature of the FLP reactions. On the other hand, CO₂ binding by the FLP systems is more enthalpically favoured than that by (NIiPr)₃P, consistent with the formation of two bonds in the FLP products versus only one bond in the reaction with (NIiPr)₃P. Given that the phosphine (NIiPr)₃P is among the strongest of nucleophiles known to independently bind CO₂, it is expected that this phosphine should be more efficient than simple amines, consistent with the presently described results with TMP. The present observations suggest that bidentate binding by an FLP improves the efficiency for CO₂ capture.

Of the three systems examined, the tBu₃P/B(C₆F₅)₃ FLP is most efficient at CO₂ capture at ambient temperatures; however the phosphine (NIiPr)₃P has the most negative ΔG ₂₉₃ among the systems studied. Among the FLP systems, tBu₃P/B(C₆F₅)₃ has a more exergonic ambient-temperature CO₂ binding reaction than TMP/B(C₆F₅)₃, but the thermodynamic parameters predict that, at elevated temperatures (347–365 K), the reaction of TMP/B(C₆F₅)₃ with CO₂ will be more exergonic. Nonetheless, these data infer that at room temperature the FLP derived from tBu₃P/B(C₆F₅)₃ binds CO₂ more effectively that the FLP derived from TMP/B(C₆F₅)₃.

Conclusions

The present study illustrates the power of the time- and labor-efficient MF platform for the qualitative and quantitative assessment of CO₂ binding by small molecules and for the determination of the thermodynamic parameters of these reactions. These data form a quantitative basis for comparison of CO₂ capture systems at different temperatures. The systems considered in this paper included FLPs, tBu₃P/B(C₆F₅)₃ and TMP/B(C₆F₅)₃, and the highly basic phosphine (NIiPr)₃P, which effectively span the range known for interactions of CO₂ with Lewis acids and bases. The data reveal the FLP derived from tBu₃P and B(C₆F₅)₃ to be the most efficient at capturing CO₂ at ambient temperature per equivalent of CO₂ binding unit. However (NIiPr)₃P offers a higher CO₂ content by mass. Certainly, these data do infer that further study of FLPs in CO₂ capture may uncover new systems that are more readily available (i.e. cheaper) and offer improved efficiency. To this end, we are continuing to employ this innovative MF methodology to assess and design new FLP systems for CO₂ capture and ultimately reduction. In addition, the study of other reactions that occur at a gas–liquid interface continues. The results of these studies will be reported in due course.