Synthesis of Intramolecular P/Al‐Based Frustrated Lewis Pairs via Aluminum‐Tin‐Exchange and their Reactivity toward CO₂

Abstract

Frustrated Lewis pairs (FLPs) composed of acidic alane and basic phosphane functions, separated by a xanthene linker, can be prepared through the corresponding Me₃Sn derivative and methyl aluminum compounds with elimination of Me₄Sn. This way MeClAl‐, Cl₂Al‐ and (C₆F₅)₂Al‐ moieties could be introduced and the resulting FLPs are stabilized by a further equivalent of the alane precursors. In contact with the FLPs CO₂ is bound via the C atom at the phosphane functions and the two O atoms at the Al centers. The residues at the latter determine the binding strength. Hence, in case of MeClAl CO₂ capture occurs at higher pressure and under ambient conditions CO₂ is released again, while for Cl₂Al and (C₆F₅)₂Al CO₂ binding becomes irreversible. The results of DFT calculations rationalize these findings by the high thermodynamic stabilization in case of more electronegative residues, which concomitantly lead to higher barriers, and in case of (C₆F₅)₂Al further stabilization is achieved through a low reorganization energy.

Exchange of the players stannane and alane provides a viable route to intramolecular FLPs with large separations between Lewis acidic aluminum und Lewis basic phosphane functions. Neutral molecules occupy the space between the two moieties and are easily displaced by approaching substrates like CO₂, which is spanned in the reaction pocket ‐ particularly strongly if the residues at the Al centers correspond to C₆F₅.

Introduction

Intramolecular frustrated Lewis pairs (FLP) based on Lewis acidic boron centers have been extensively studied in recent years. ^[1] Representatives bearing an aluminum acceptor site, however, remain rare due to synthetic challenges that are associated with its high intrinsic Lewis acidity and the fragility of the Al−C bond. ^[2] So far, merely two synthetic strategies ^[3] to connect an Al^III center with the carbon atom of an organic linker scaffold have been followed, namely hydroalumination ^[4] and transmetalation from organolithium compounds[ ² , ⁵ ] (Scheme 1). Both perform well for close arrangements of Lewis acid and base such as geminal and vicinal patterns, but are only conditionally suitable for bringing the components into a larger spatial separation.

Scheme 1.

Synthetic approaches to form P/Al‐based frustrated Lewis pairs.

Our initial endeavors to synthesize an intramolecular P/Al‐based FLP on a xanthene backbone by lithiation and transmetalation resulted in the irreversible coordination of the utilized solvent (Et₂O, THF) to the Lewis acidic Al^III center. This led to interesting insights with regard to the functioning of such FLPs, exemplarily for the ring‐opening of THF, ^[6] but further applications were limited: the solvents are bound tightly and cannot be replaced by external substrates, so that the Lewis acid site is quenched. Consequently, we sought to develop an alternative route that leads to the corresponding solvent‐free xanthene‐linked Al/P FLP. Obviously, on such a route the aluminum‐carbon bond linking the Al site to the xanthene backbone has to be established in non‐coordinating solvents to allow for subsequent reactivity toward small molecules.

Eisch et al. and Gabbaï et al. independently reported on the exchange of trimethylstannyl groups bound to sp‐ or sp ² ‐hybridized carbon atoms by alkylaluminum chlorides in aliphatic or aromatic solvents (Scheme 2). ^[7] However, these two unique examples did not allow for conclusions as to whether the presence of a Lewis base (e. g. a xanthene‐bound diphenylphosphine moiety) might adversely affect the desired transmetalation. Beckmann et al. reported on the attempted synthesis of diphenylphosphinoacenaphthylaluminum dichloride starting from the corresponding tri‐n‐butylstannyl precursor and aluminum trichloride, but no conversion could be observed. ^[5g] Contrarily, the other group 13 E^III trichlorides (E=B, Ga, In, Tl) did show reactivity, suggesting that a phosphine function does not generally act as an inhibitor. Herein, we present our findings on the applicability of aluminum‐tin transmetalations in the synthesis of a xanthene‐linked Al/P‐based FLP and the utilization of the generated systems for CO₂ activation.

Scheme 2.

Aluminum‐tin exchange under formation of different by‐products as reported independently by Eisch et al. and Gabbaï et al.

Results and Discussion

The starting materials for our studies 4‐trimethylstannyl‐5‐diphenylphosphino‐9,9‐dimethylxanthene 1 and 4‐trimethylstannyl‐5‐diphenylphosphino‐2,7‐di‐tert‐butyl‐9,9‐dimethylxanthene 2 were synthesized from the respective 4‐iodo and 4‐bromo precursors. ^[8] Lithiation in THF and subsequent addition of trimethyltin chloride yielded in 1 and 2, which could be isolated in good yields of 80 % and 76 %, respectively (Scheme 3). Both compounds were fully characterized by ¹H, ¹³C, ³¹P and ¹¹⁹Sn NMR spectroscopy, single crystal XRD and elemental analysis. It is noteworthy, that the ¹H NMR spectra of 1 and 2 in C₆D₆ exhibit prominent signals at δ=−0.47 ppm and δ=−0.50 ppm, respectively, which originate from the trimethylstannyl moiety, distinctive due to their ¹¹⁹Sn and ¹¹⁷Sn satellites.

Scheme 3.

Synthetic routes to FLPs 3 and 4 via aluminum‐tin exchange starting from the corresponding stannanes 1 and 2.

With these starting materials in hand, the Al−Sn exchange was addressed. While Gabbaï et al. had observed a Me−Al/Sn−C metathesis to produce Me₄Sn, Eisch et al. reported a Cl−Al/Sn−C metathesis yielding in Me₃SnCl as the by‐product of the exchange. Therefore, dimethylaluminum chloride was selected as the reagent, as this enabled the system to react via any of the two types of pathways. Treatment of 1 and 2, respectively, with one equivalent of Me₂AlCl in C₆D₆ led to the slow formation of a new species as indicated by ³¹P and ¹H NMR spectroscopy. In the ³¹P NMR spectra this new product resonates at a chemical shift of δ=−16.7 ppm for the conversion of 1 and of δ=−16.3 ppm for the conversion of 2, respectively. Both ¹H NMR spectra featured an intense signal at δ=0.00 ppm regardless of the chosen xanthene precursor, which was thus attributed to the by‐product. The chemical shift of this resonance in combination with its integral suggests Me₄Sn as secondary product and therefore a reactivity resembling the one reported by Gabbaï and coworkers.

However, unlike in case of the reported examples, the conversion stopped at 50 % even after heating the reaction mixture to 80 °C for several hours indicating a 1 : 2 stoichiometry of the transformation. Indeed, when two equivalents of Me₂AlCl were added, the precursors 1 and 2, respectively, as judged by NMR spectroscopy, were fully consumed and selectively converted into single products, which as outlined below were identified as ^HXant(PPh₂−AlMe₂Cl−AlMeCl), 3, and ^tBuXant(PPh₂−AlMe₂Cl−AlMeCl), 4. The high selectivity of the reactions allowed for simple workup by evaporating the solvents and Me₄Sn to give 3 and 4 quantitatively in satisfactory purity (Scheme 3). Both compounds are sensitive toward air and moisture and degrade slowly at room temperature to multiple unidentified species.

Attempts to grow single crystals of 3 and 4 suitable for X‐ray diffraction analysis were performed by slow diffusion of n‐hexane into saturated solutions of the compounds in toluene. The molecular structures determined for the resulting crystals are shown in Figure 1. Interestingly, quite different solid state structures were found in spite of the close resemblance of the precursors 1 and 2. In both cases, the anticipated formation of an Al−C bond at the xanthene scaffold has been achieved successfully and the concomitant elimination of Me₄Sn, as inferred from the ¹H NMR spectra, is further confirmed, as the Al‐center bound to the xanthene is surrounded by one methyl and one chlorido group. The structure of 3 is monomeric and charge‐neutral as the second equivalent of Me₂AlCl that was required to reach full conversion is bound in an FLP type fashion in‐between the Lewis basic phosphine and the Lewis acidic xanthene‐bound alane sites. By contrast, the structure determined for crystals grown from solutions of 4, corresponding to a salt 4’, can be rationalized as follows: Two xanthene molecules, each equipped with a Ph₂P and an Al(Me)Cl unit, have dimerized via intermolecular mutual contacts between the acid and base sites. One of the chloride ligands from the resulting molecule has been abstracted by two equivalents of the Me₂AlCl precursor leading to the anion [Me₄Cl₃Al₂]⁻ (the latter is disordered in the crystal, containing varying ratios of Me/Cl). The remaining chlorido ligand in the corresponding cation thus formed enters a bridging position between both aluminum centers.

Figure 1.

Molecular structures of one independent molecule of the asymmetric unit of 3 (left) and the complex cation of 4’ (right) as determined by X‐ray diffraction analysis. H atoms omitted for clarity; thermal ellipsoids drawn at the 50 % probability level. Due to disorder isotropic refinement was conducted on the phenyl and tert‐butyl groups of the cation and the methyl and chloride groups in the anion in 4’. Selected bond lengths (Å) and angles (°): (for 3) P1‐Al1 2.467(3), Al1‐C2 1.957(7), Al1‐C3 1.924(8), Al1‐Cl1 2.252(2), Al2‐Cl1 2.301(2), Al2‐Cl2 2.162(3), Al2‐C1 1.971(5), P1‐Al1‐Cl1 99.15(9), Al1‐Cl1‐Al2 120.1(1); (for 4’) P1‐Al2 2.474(2), P2‐Al3 2.471(2), Al2‐Cl5 2.237(2), Al3‐Cl5 2.250(2), Al2‐C72 .1.952(5), Al3‐C40 1.951(7), P1‐Al2‐Cl5 95.95(7), P2‐Al3‐Cl5 95.17(7), Al2‐Cl5‐Al3 118.79(8). The structural parameters of the four independent molecules of the asymmetric unit of 3 are nearly identical; slight deviations are noted for the Al−Cl−Al bridge for which the bond lengths measure between 2.240–2.264 Å and 2.295–2.314 Å, respectively.

While the solid state structure of 3 rationalizes that a second equivalent of Me₂AlCl is required, the structure of 4’ does not, as the additional alane molecule is not incorporated in the FLP entity but rather acts as chloride scavenger and thus is not recognizably needed for a full conversion of the precursor 2. Therefore, the structure found for 4’ may well not correspond to the primary product of the reaction but might form only during the crystallization process. This hypothesis is underpinned by the occurrence of a single resonance for the dimethyl groups at the xanthene scaffold in the ¹H NMR spectrum of the reaction mixture, from which 4’ was crystallized, suggesting a σ mirror plane along the xanthene plane of the product. DOSY experiments confirmed the presence of monomeric 4 in solution (see Supporting Information), and thus, a similar structure in solution as found for 3 is proposed.

Another common feature of both structures is that each aluminum and phosphorus atom is tetracoordinated, which raised the question whether the compounds show reactivity toward small molecules as the electrophilicity of the alane and the nucleophilicity of the phosphine appear quenched. Similar structures to 3 which also exhibit P−Al−X−Al structural motifs (with X=H) have been investigated by Uhl et al., ^[4b] however, their reactivity toward substrates has not been reported. ^[9] To test whether 3 and 4, respectively, can act as FLP, i. e. whether the Me₂AlCl unit in 3 is bound weakly enough to act as a replaceable space holder, solutions of 3 and 4 in toluene‐d ₈ were treated with an excess of THF, since the THF ring‐opening reaction on a xanthene linked Al/P FLP has proved possible in our previous work. ^[6] In the ³¹P NMR spectra recorded after both reactions a single new resonance was found at a chemical shift of δ=−18.8 ppm (for 3) and δ=−16.8 ppm (for 4), consistent with the coordination of a THF molecule to the xanthene‐bound aluminum center. Heating the reaction mixtures to 80 °C for three days resulted in complete transformation of the intermediates to the THF ring‐opened products resonating at δ=26.5 ppm (^HXant(PPh₂(CH₂)₄O(AlMe₂Cl)AlMeCl), 5, originating from 3) and δ=27.0 ppm ( ^tBuXant(PPh₂(CH₂)₄O(AlMe₂Cl)AlMeCl), 6, originating from 4), respectively, in the ³¹P NMR spectra (Scheme 4). In both ¹H NMR spectra two resonances with an integral of three each can be observed belonging to the Me₂C unit in the xanthene backbone, which indicates an asymmetric structure of the products. Furthermore, the ¹H NMR spectra show that in contrast to the ring‐opened compounds reported earlier, 5 and 6 feature a second alane moiety coordinating to the oxygen atom of the 4‐oxybutyl bridge. For product 6 this was confirmed by single crystal X‐ray diffraction (Figure 2).

Scheme 4.

THF cleavage by the FLPs 3 and 4 to give the betainic compounds 5 and 6.

Figure 2.

Molecular structure of 6 as determined by X‐ray diffraction analysis. H atoms omitted for clarity; thermal ellipsoids drawn at the 50 % probability level. Selected bond lengths (Å) and angles (°): P2‐C17 1.798(2), Al1‐O2 1.879(2), Al2‐O2 1.846(2), O2‐C20 1.467(3), Al1‐O2‐Al2 119.74(8), C20‐O2‐Al1 121.4(1), C20‐O2‐Al2 114.2(1).

Next, we addressed the question whether the FLPs 3 and 4 are also reactive toward a less donating and more inert substrate like CO₂. Therefore, a degassed solution of 3 and 4, respectively, in toluene was treated with 2 bar ¹³CO₂ (Scheme 5). After 30 minutes at room temperature reaction control by ³¹P NMR spectroscopy indicated full conversion of the starting compound to a product exhibiting a downfield‐shifted broad doublet in both cases. The chemical shifts of δ=12 ppm (for 3) and δ=14 ppm (for 4) are in line with the chemical shifts of other reported CO₂ adducts of P/Al‐based FLPs.[ ^4a , ^5a , ¹⁰ ] When the excess pressure of ¹³CO₂ was released, the rise of bubbles from the solution indicated a reversible process, which was confirmed after evaporating and redissolving the residues: ³¹P NMR spectra revealed that the substrates 3 and 4, respectively, were fully recovered. However, under CO₂ overpressure the product of the reaction of 3 with CO₂ could be crystallized from a supersaturated solution over time, giving single crystals of ^HXant(PPh₂CO₂(AlMe₂Cl)AlMeCl), 7⋅toluene, with sufficient stability and suitable quality for X‐ray diffraction analysis.

Scheme 5.

Reversible capture of CO₂ by 3 forming 7.

The solid state structure of 7 reveals the anticipated activation of CO₂ at the binding site spanned by the xanthene‐bound alane and phosphine functions (Figure 3). While the phosphine expectedly binds to the carbon atom and the xanthene‐bound aluminum atom to one of the oxygen atoms, the second oxygen atom coordinates to the additional dimethylaluminum chloride unit. This causes the two C−O bonds to be lengthened to a similar extent (d _C14‐O3=1.249(2) Å vs. d _C14‐O2=1.241(1) Å) as the formal negative charge is delocalized over the C(OAl)₂ moiety, comparable to the CO₂ adduct of the intermolecular Mes₃P/AlX₃ (X=Cl, Br) FLP by Stephan and coworkers. ^[10a] The O−C−O angle is 128.35(10)°; the sum of angles around the carbon atom is 360°, rendering the P−CO₂ unit trigonal planar.

Figure 3.

Molecular structure of 7⋅toluene as determined by X‐ray diffraction analysis. H atoms and co‐crystallized solvent omitted for clarity; thermal ellipsoids drawn at the 50 % probability level. Selected bond lengths (Å) and angles (°): P1‐C14 1.883(1), Al1‐O3 1.865(1), Al2‐O2 1.8837(9), C14‐O2 1.241(1), C14‐O3 1.249(2), O2‐C14‐O3 128.3(1), P1‐C14‐O2 111.95(8), P1‐C14‐O3 119.71(8).

We were now interested whether the equilibrium of the CO₂ capture can be shifted toward CO₂ fixation by increasing the Lewis acidity of the aluminum acceptor sites. Therefore, we also tested the precursors trimethylaluminum, methylaluminum dichloride and aluminum trichloride regarding their suitability for the aluminum‐tin exchange reaction in contact with 1. We found that the tendency of the alanes with the general formula Me_xAlCl_3‐x (x=0–3) to react with 1 (0.5 equiv.) correlates with their Lewis acidity, i. e. while trimethylaluminum behaves inert even during several days at 100 °C, dimethylaluminum chloride (see above) and methylaluminum dichloride react selectively and are fully consumed after two days / one hour at room temperature, respectively; aluminum trichloride shows a highly unselective reactivity even at low temperatures. The product of the reaction of 1 with two equivalents MeAlCl₂ ^HXant(PPh₂−AlMeCl₂−AlCl₂), 8, is formed by elimination of Me₄Sn as indicated by ¹H NMR spectroscopy, and the chemical shift of δ=−19.1 ppm in the ³¹P NMR spectrum suggests an analogous structure as found for 3 (Scheme 6). Like 3 and 4, compound 8 is highly sensitive toward moisture and oxygen and degrades at room temperature.

Scheme 6.

Synthetic routes to more Lewis acidic xanthene‐bound aluminum acceptor sites with chlorido (8) or pentafluorophenyl (9) substituents.

To attach a highly Lewis acidic diarylaluminum site on the xanthene‐backbone, the reactivity of 1 toward MeAl(C₆F₅)₂ was tested. Since the MeAl(C₆F₅)₂ bears two different Al−C bonds (Al−C(sp ³ ) and Al−C(sp ² )) that in principle can be transferred to the trimethyltin group producing either Me₄Sn or Me₃Sn(C₆F₅), two different reaction courses were conceivable, with the one leading to Al(C₆F₅)₂ sites obviously being the more desirable.

Analogous to the aforementioned syntheses, two equivalents of freshly prepared MeAl(C₆F₅)₂ were required to achieve full conversion of the precursor 1. After one hour at room temperature the ¹H NMR spectrum of the reaction mixture in toluene‐d ₈ showed both the formation of Me₄Sn (δ=0.00 ppm) and Me₃Sn(C₆F₅) (δ=0.21 ppm) with an integral ratio of 4 : 1 (i. e. a Me₄Sn/Me₃Sn(C₆F₅) ratio of 3 : 1), suggesting a notable preference of C(sp ³ )‐Al/Sn−C over C(sp ² )‐Al/Sn−C metathesis. The presence of Me₃Sn(C₆F₅) could be substantiated, as after evaporating and dissolving the mixture in CDCl₃ the ¹H NMR and ¹⁹F NMR spectra showed the resonances reported for Me₃Sn(C₆F₅). ^[11] The ¹⁹F NMR spectrum of the reaction mixture exhibits four signal sets belonging to pentafluorophenyl groups, in line with a mixture of products containing C₆F₅ units. In the ³¹P NMR spectrum two prominent resonances with chemical shifts of δ=−7.7 and −9.0 ppm (integral ratio of roughly 1 : 2) were found, indicating full consumption of the precursor 1 (δ=−19.6 ppm) and the occurrence of two main xanthene‐based products. Altogether, these spectroscopic findings may be rationalized as follows: The observation of Me₄Sn in the ¹H NMR spectrum suggests the successful attachment of the Al(C₆F₅)₂ unit to the xanthene backbone, while the occurrence of Me₃Sn(C₆F₅) is likely due to the introduction of a Al(C₆F₅)Me unit to the linker scaffold. In both cases, the binding of the second equivalent of MeAl(C₆F₅)₂ is expected, so that the reaction mixture likely contains ^HXant(PPh₂−AlMe(C₆F₅)₂‐Al(C₆F₅)Me) as well as the desired product ^HXant(PPh₂‐AlMe(C₆F₅)₂−Al(C₆F₅)₂), 9 (Scheme 6), in addition to the accompanying tin compounds. Attempts to separate this complicated mixture and further characterize the products of interest failed.

Surprisingly, when an excess of THF is added to the reaction mixture in toluene‐d ₈, the ³¹P NMR spectrum exhibits a sharp and intense signal at δ=−20.6 ppm that accounts for >90 % of P‐based products according to ³¹P NMR spectroscopy. This observation leads to the conclusion that the species found in the abovementioned reaction mixture are part of an equilibrium, and through shifting of this equilibrium a uniform reaction with THF can be reached. The resonance found at δ=−20.6 ppm was attributed to the reported intermediate ^HXant(PPh₂)(Al(C₆F₅)₂(THF)) and this assignment was further substantiated through the analytical data of the product of the following ring‐opening reaction, matching the ones reported for ^HXant(PPh₂(CH₂)₄OAl(C₆F₅)₂), 10 (Scheme 7). ^[6] In contrast to 6, the second equivalent of MeAl(C₆F₅)₂ does not coordinate to the oxygen atom of the 4‐oxybutyl bridge (see above) in 10, probably due to steric hindrance, and is washed away during the work‐up procedure.

Scheme 7.

Reactivity of 9 toward THF forming reported compound 10.

The reactivity of the two freshly synthesized FLPs 8 and 9 toward CO₂ was investigated using the same conditions as for 3. Solutions of 8 and 9, respectively, in toluene‐d ₈ readily reacted when exposed to 2 bar CO₂ as indicated by their ³¹P NMR spectra (Scheme 8). In case of 8, the new signal, attributed to the product ^HXant(PPh₂CO₂(AlMeCl₂)AlCl₂), 11, is downfield shifted (δ=15 ppm) and very broad, similar to the observations made for 7. After the reaction of 9 a prominent new broad resonance with a chemical shift of δ=10.6 ppm appeared. It is noteworthy that ‐ as in case of the THF reactions ‐ the original two resonances (δ=−7.7 and −9.0 ppm) found for the reaction mixture after the synthesis of 9, vanish after exposure to CO₂ and the new signal assigned to the product ^HXant(PPh₂CO₂(Al(C₆F₅)₂Me)Al(C₆F₅)₂), 12, at δ=10.6 ppm accounts for the major share of resonating phosphorus nuclei.

Scheme 8.

Irreversible capture of CO₂ by 8 and 9, respectively.

³¹P NMR spectroscopic investigations revealed that evaporation of the solutions of 11 and 12 did not lead to the release of CO₂. Thus, as anticipated, the attachment of the electron withdrawing chloride (for 11) and pentafluorophenyl (for 12) substituents, which enhance the Lewis acidity of the aluminum acceptor sites, shifts the equilibrium of the CO₂ capture reaction strongly to the right side, so that, unlike in case of 3 and 4, they are not reversible.

A supersaturated solution of 11, from which single crystals slowly formed, was prepared by treating a concentrated solution of 8 with CO₂. Single crystals of 12 were grown by vapor diffusion of n‐hexane into the reaction mixture of 12 in toluene. The solid state structures determined by X‐ray diffraction analysis display the same connectivity as the structure of 7⋅toluene, i. e. the CO₂ molecule is bound between the phosphine and alane moiety attached to the xanthene scaffold and the additional second equivalent of the applied organoaluminum reagent (Figure 4).

Figure 4.

Molecular structures of 11⋅toluene (left) and 12 (top down view, right) as determined by X‐ray diffraction analysis. H atoms and co‐crystallized solvent omitted for clarity; thermal ellipsoids drawn at the 50 % probability level. Note, that in 11 the methyl group of C15 is disordered with a chloride. Selected bond lengths (Å) and angles (°): (for 11) P1‐C14 1.884(2), Al1‐O2 1.840(1), Al2‐O3 1.853(1), C14‐O3 1.243(2), C14‐O2 1.252(2), O2‐C14‐O3 127.3(1); (for 12) P1‐C14 1.882(2), Al1‐O3 1.860(1), Al2‐O2 1.898(1), C14‐O2 1.239(2), C14‐O3 1.251(2), O2‐C14‐O3 128.3(2).

The CO₂ molecule is clearly activated as its C−O bonds are lengthened in both systems 11 and 12 compared to free gaseous CO₂ (d _C‐O=1.1632 Å). Both C−O bonds are stretched to a similar extent and the P−CO₂ unit shows a trigonal planar structure, arguing for a delocalization of the negative charge over the C(OAl)₂ moiety. However, the structural parameters of 11 and 12 do not differ significantly from the ones exhibited by 7, and there is no distinctive feature with regards to the CO₂ binding that would explain the irreversibility of CO₂ binding in case of 11 and 12.

To fathom the difference in reversibility, quantum chemical calculations using density functional theory (DFT) were performed. Before addressing the calculated thermochemical and mechanistic data for the reaction of the FLPs 3, 8, and 9 with CO₂ in solution in detail, a closer look is taken at the computed (gas phase) binding energies ΔE_bnd of CO₂ in the corresponding reaction products 7, 11, and 12, as they naturally constitute an essential factor of the reversibility or irreversibility of CO₂ capture. As can be seen from Table 1, ΔE_bnd of CO₂ in 7 and 11 is comparable in magnitude (−27 and −36 kJ mol⁻¹), although the difference of about 9 kJ mol⁻¹ is a clear sign of the effect of the stronger Lewis acidic sites in 11 compared to 7. In contrast to this notable but relatively small difference, a pronounced stability increase (by 73 kJ mol⁻¹ relative to 11) in binding energy is seen for the Al(C₆F₅)₂ system 12.

Table 1.

Results of EDA‐NOCV analyses for the FLP:CO₂ bonding interactions in compounds 7, 11, and 12, as well as Wiberg bond indices (WBIs) for selected bonds and NPA charges (q) for the CO₂ fragment. Energy values (in kJ mol⁻¹) were computed at the BP86‐D3(BJ)/TZ2P level of theory.^[a]

						WBI[c,d]
compound	ΔEbnd [b]	ΔEprep(FLP)	ΔEprep(CO2)	ΔEprep(total)	ΔEint	P−C(O2)	Al−O(CO)	q(CO2)[c]
7	‐27	237	216	454	−481	0.775	0.260	−0.77
11	−36	255	208	463	−499	0.786	0.259	−0.79
12	−109	192	207	399	−508	0.782	0.261	−0.78
	contributions to ΔEint [a]
	ΔVelstat [e]	ΔEPauli	ΔEsteric [f]	ΔEorb [e]	ΔEdisp [e]
7	−891 (43.3 %)	1575	684	−1103 (53.6 %)	−62 (3.0 %)
11	−950 (43.6 %)	1680	730	−1165 (53.4 %)	−65 (3.0 %)
12	−922 (42.8 %)	1645	722	−1155 (53.7 %)	−75 (3.5 %)
	contributions to ΔEorb [g]
	ΔE(ρ1)	ΔE(ρ2)	ΔE(ρ3)	ΔE(ρresidual)
7	−741 (67.2 %)	−96 (8.7 %)	−88 (8.0 %)	−178 (16.1 %)
11	−781 (67.1 %)	−108 (9.3 %)	−98 (8.4 %)	−178 (15.3 %)
12	−777 (67.3 %)	−112 (9.7 %)	−92 (8.0 %)	−175 (15.1 %)

[a] Energy‐decomposition analyses with natural orbitals for chemical valence (EDA‐NOCV) at MARIJ‐BP86‐D3(BJ)/def2‐TZVP/COSMO(toluene) optimized structures. [b] Calculated reaction energy (BP86‐D3(BJ)/TZ2P) for the activation of CO₂ by the FLPs 3, 8, 9: FLP+CO₂→FLP‐CO₂. [c] Wiberg bond indices and natural charges computed at BP86/def2‐TZVP level. [d] The Wiberg bond indices for the two Al−O(CO) bonds in the molecules are given as average value. [e] The values in parentheses give the percentage contribution to the total attractive interactions. [f] ΔE_steric=ΔV_elstat+ΔE_Pauli. [g] The values in parentheses give the percentage contribution to the total orbital interaction ΔE_Orb.

To obtain more fundamental insights into these numbers, the intrinsic bonding situation was analyzed in detail based on EDA‐NOCV calculations, and the results are also summarized in Table 1.

As expected, the orbital term ΔE_orb in this type of adducts is dominated by three main orbital interactions (about 84–85 %), which comprise the P−CO₂ coordination covalent bond between the Lewis basic phosphorus atom and the activated CO₂ carbon atom, and two (CO)O−Al coordination covalent bonds resulting from interaction of the oxygen atoms with the aluminum centers. Plots of the corresponding deformation densities Δρ are depicted in Figure 5 for compound 7 as an example. By far the largest contribution to ΔE_orb, about 67 % in all cases, stems from to the P−CO₂ interaction, while each of the (CO)O−Al interactions contribute about 8–10 %. This result agrees well with the calculated Wiberg bond indices (WBIs), which indicate a considerably higher covalent character for the P−CO₂ bonds. The covalency of these bonds and, in contrast, ionicity of the (CO)O−Al interactions is also clearly reflected in the corresponding differences in the calculated NPA charges for the involved atoms (Table S1 in the Supporting Information). As can be seen from Table 1, the relative contributions for ΔE_orb stay almost constant across all adducts. However, the absolute values are notably larger for 11 compared to 7, resulting in a larger orbital contribution term for the former. In contrast, the presence of pentafluorophenyl substituted aluminum centers in 12 does not result in a further enhancement of all the pairwise interactions compared to the situation in 11. Instead, an overall slight decrease of ΔE_orb can be noted.

Figure 5.

Contour plots (isosurface value ±0.005 a.u.) of NOCV deformation densities Δρ and associated orbital stabilization energies (ΔE(ρ), in kJ mol⁻¹) of the three dominant pairwise FLP:CO₂ orbital interactions in 7 (BP86‐D3(BJ)/TZ2P). Electronic charge flow is from red to blue.

The calculated total interaction energies ΔE_int for the investigated products of CO₂ activation span a rather narrow range of 27 kJ mol⁻¹. As anticipated, a successive increase of ΔE_int for the CO₂ and FLP fragment interactions is observed in the order 7 (−481 kJ mol⁻¹) < 11 (−499 kJ mol⁻¹) < 12 (−508 kJ mol⁻¹). This appears to be fully consistent with an enhanced FLP:CO₂ bonding as consequence of a paralleling increase in Lewis acidity of the organoaluminum acceptor sites. Partial substitution of methyl with chloride in the FLP, i. e., transition from 7 to 11, raises ΔE_int by about 18 kJ mol⁻¹. However, substitution with the even stronger electronegative C₆F₅ groups in 12 enhances the interaction by only 9 kJ mol⁻¹ relative to 11. A breakdown of the interaction energies into their individual contributions shows that almost all terms, attractive and repulsive, are significantly larger for the AlCl₂‐FLP 11 compared to 7. The increase in orbital interaction ΔE_orb is consistent with a higher degree of covalent bonding in 11, which is partly compensated by the slightly larger repulsive sterics term ΔE_steric (=ΔV_elstat+ΔE_Pauli). Owing to the similarity in structure and composition, the dispersive interaction contribution is only slightly larger in 11 (about 3 kJ mol⁻¹). In contrast, compared to 11 the repulsive and most of the attractive terms are smaller again in the C₆F₅‐substituted 12. The notable exception is the ΔE_disp. contribution, which is by 10 kJ mol⁻¹ larger. The overall increase of ΔE_int in 12 thus essentially originates from enhanced dispersive interactions, not as a consequence of stronger orbital and weaker sterics contributions. We note also that electrostatic contributions remain relatively similar for all three systems, consistent with the similar NPA charges of the CO₂ fragment (Table 1).

While the total interaction energies match the trend of the computed binding energies, the differences in ΔE_int and ΔE_bnd show that the former remarkably accounts for only about 12 % (ΔΔE_int=9 kJ mol⁻¹) of the observed increase in stability (ΔΔE_bnd=73 kJ mol⁻¹) when comparing 11 and 12. The origin of the particularly large binding energy of CO₂ in 12 is thus clearly a result of the significantly smaller reorganization energy ΔE_prep in this case, that is, the necessary structural changes in the reactant FLP 9 to accommodate a CO₂ molecule.

This smaller reorganization energy can be traced back to a considerably altered bonding situation regarding the Lewis acidic centers in the reactant structure of 9 (Figure 6) compared to 3 (Figure 1) and 8. While a chlorine atom acts as a strong linker between the aluminum sites in the latter compounds, inspection of the most favorable computed structure for 9 (Figure 6) reveals that neither a methyl nor a C₆F₅ group (symmetrically or asymmetrically) bridges the centers via a shared carbon atom, as might be expected. Instead, the xanthene‐bound aluminum acceptor exhibits an Al⋅⋅⋅F interaction with one of the pentafluorophenyl rings of the phosphorus‐coordinated MeAl(C₆F₅)₂ unit, showing a comparably long aluminum‐fluorine distance of 2.022 Å. At the same time, the Al−P distance in 9 is noticeably elongated (2.525 Å) compared to 3 and 8 (2.474 and 2.439 Å). The same holds true for the Al⋅⋅⋅Al spacing, which is naturally rather similar for 3 and 8 (2.365 and 2.330 Å), but significantly enlarged in case of 9 (4.646 Å). Due to the smaller size of the substituents at Al in 3 and 8, the second AlR₃ group can occupy a position much closer to the P⋅⋅⋅Al axis of the FLP core in these compounds. Hence, despite the higher Lewis acidity the computed parameters seem to indicate a weaker bound central MeAl(C₆F₅)₂ unit, adopting a comparably detached position. This in turn would explain the relatively small reorganization energy necessary to enable activation of CO₂.

Figure 6.

Calculated molecular structure of 9 (MARIJ‐BP86‐D3(BJ)/def2‐TZVP/COSMO(toluene)). Hydrogen atoms have been omitted for clarity.

As shown experimentally, compounds 3 and 8 are both able to activate CO₂, but also constitute borderline cases for a reversible (3) or irreversible (8) binding of CO₂. In view of the comparably small difference in ΔE_bnd for 7 and 11 discussed above, the origin of reversible/irreversible CO₂ binding is apparently rooted in the corresponding barrier height for activation. To elucidate the basic kinetics behind this difference in reversibility, further DFT calculations at a more sophisticated level (see Computational Details) have been carried out to obtain the free energy profiles in toluene solution for the first steps of carbon dioxide activation by 3 and 8. The calculations reveal that a direct insertion of CO₂ into the Al−P bond exhibits a free energy barrier for both compounds which is too high to be overcome at room temperature (Δ_rG^ǂ=146 kJ mol⁻¹ (3)/166 kJ mol⁻¹ (8); TS₃ in Scheme S1 in the Supporting Information). Instead, the initial reaction does not necessarily involve CO₂ at all. As shown in Scheme 9, the first process step is likely a dissociation of the central aluminum unit from the Lewis basic diphenylphosphine site to form an intermediate (Int₁), which features a classical chlorine‐bridged dimeric structural motif of the Al centers. The corresponding transition states are much lower in free energy as compared to those of the direct CO₂ insertion (TS₁; Δ_rG^ǂ=54 kJ mol⁻¹ (3)/76 kJ mol⁻¹ (8)), rendering this process feasible at room temperature. Presumably, the energy loss due to the breaking of the P−Al bond is partly compensated by the concomitant formation of the second Al−Cl bridge, thus lowering the barrier. This opens up a cavity where CO₂ can more easily interact with both the Lewis acidic and the (now activated) basic site. To some extent this process shares similarities with the mechanism for CO₂ activation found for other (intermolecular) P/Al‐based FLPs in coordinating solvents. ^[12] The FLPs in this study however, by construction, exhibit an intramolecular coordination environment by means of the second AlR₃ unit that helps to facilitate the transformation, and thus does not have to rely on an external coordinating solvent. The rate‐determining step on the profile is the second transition state TS₂, in which the actual CO₂ activation takes place, and which leads to the formation of intermediate Int₂. The latter can subsequently rearrange into the respective final product 7 (Δ_rG=−22 kJ mol⁻¹) or 11 (Δ_rG=−45 kJ mol⁻¹). Similar to TS₁, FLP 8 exhibits a barrier for CO₂ activation which is about 20 kJ mol⁻¹ higher in free energy than in case of 3 (TS₂; Δ_rG^ǂ=73 kJ mol⁻¹ (3) / 94 kJ mol⁻¹ (8)). The general observation of higher barriers reflects the stronger Lewis acidic sites in FLP 8, which makes the cleavage of P−Al and Al−Cl bonds energetically more demanding. While the activation free energy that is required for TS₂ for both compounds is still much lower compared to a direct insertion, the transition state of 8 is likely energetically at the brink of a feasible room temperature process. Consequently, the substantial gain in thermodynamic stability upon formation of product 11 (Δ_rG=−45 kJ mol⁻¹) renders the reverse reaction impossible under these conditions. In contrast, and in line with the experimental findings, the reverse reaction is energetically still possible for product 7, as its required total activation free energy of Δ_rG^ǂ=95 kJ mol⁻¹ is similar in magnitude to TS₂ in the forward process of 8.

Scheme 9.

Free energy profile for the activation of CO₂ with compounds 3 (red lines) and 8 (blue lines), respectively. All reaction and activation free energies (Δ_rG_soln, in kJ mol⁻¹, 298.15 K, 0.1 MPa) in toluene solution are given relative to the separate reactants (MARIJ‐ωB97M‐V/def2‐TZVPP/COSMO‐RS(toluene)).

Conclusion

In conclusion, an aluminum‐tin exchange has been successfully utilized for the synthesis of an intramolecular P/Al‐based FLP on a xanthene scaffold. The large spatial separation between the donor and acceptor sites is unprecedented in the chemistry of unimolecular aluminum FLPs. Even though a second organoaluminum precursor molecule is incorporated in the reaction pocket, the compounds show reactivity toward small molecules such as THF and CO₂. The reversibility of the CO₂ capture at room temperature is dependent on the Lewis acidity of the aluminum sites, tuned by their substituents, and in turn by the thermodynamic stability of the products and the associated reaction barriers. Thermodynamically, the CO₂ adduct 11 is significantly more stable than 7 and the origin for irreversibility is the much larger barrier for the back reaction. In case of 12 there is an additional effect that strongly increases the energy for the binding of CO₂ as compared to 11, namely the significantly smaller reorganization energy due to the fact that the structure of 9 accommodates a CO₂ molecule with far less changes as compared to 8.

Our future efforts will focus on the synthesis of an intramolecular FLP comprising an aluminum acceptor and a phosphine donor site in a 1 : 1 stoichiometry, which would allow for further functionalization of the Al‐center by different substituents.

Experimental Section

Crystallographic data

Deposition Number(s) 2150565, 2150566, 2150567, 2150568, 2150569, 2150570, 2150571 and 2150572 contain(s) the supplementary crystallographic data for this paper. These data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe Access Structures service.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Dedicated to Prof. Holger Braunschweig on occasion of his 60^th birthday.

P. Federmann, R. Müller, F. Beckmann, C. Lau, B. Cula, M. Kaupp, C. Limberg, Chem. Eur. J. 2022, 28, e202200404.

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.