Effective and Reversible Carbon Dioxide Insertion into Cerium Pyrazolates

Abstract The homoleptic pyrazolate complexes [CeIII 4(Me2pz)12] and [CeIV(Me2pz)4]2 quantitatively insert CO2 to give [CeIII 4(Me2pz⋅CO2)12] and [CeIV(Me2pz⋅CO2)4], respectively (Me2pz=3,5‐dimethylpyrazolato). This process is reversible for both complexes, as observed by in situ IR and NMR spectroscopy in solution and by TGA in the solid state. By adjusting the molar ratio, one molecule of CO2 per [CeIV(Me2pz)4] complex could be inserted to give trimetallic [Ce3(Me2pz)9(Me2pz⋅CO2)3(thf)]. Both the cerous and ceric insertion products catalyze the formation of cyclic carbonates from epoxides and CO2 under mild conditions. In the absence of epoxide, the ceric catalyst is prone to reduction by the co‐catalyst tetra‐n‐butylammonium bromide (TBAB).


Introduction
Inexorably rising CO 2 levels in the earths atmosphereand their consequential environmental impact-have spurred much interest in combating CO 2 build-up. [1] Capture technologies, such as carbon dioxide capture and storage (CCS) and direct air capture (DAC), [2] and CO 2 conversion into fuels or chemical feedstocks [3] appear promising. However, such tactics suffer from either a lack of appropriate storage and transportation of CO 2 , or overcoming the high activation barrier of CO 2 . [4] To date, the most effective sorbents for CCS/ DAC are alkali-metal/alkaline-earth metal hydroxide solutions, inorganic salts (e.g. alkali-metal carbonates), [1,2] or highsurface supported polyamines (max. CO 2 sorption capacity ca. 3 mmol g À1 at 1 bar) [5] and magnesium-based metal-organic frameworks (ca. 35 wt % or 8 mmol CO 2 g À1 at 1 bar). [6] Like alkaline-earth metals, rare-earth metals (Ln) feature a high affinity for carbon dioxide (cf. bastnaesite is the most important Ln III deposit in the Earths crust). Thus, metalorganic derivatives easily react with or insert CO 2 , as initially demonstrated by Bochkarev [7,8] and alkoxides [Ln(OnBu) 3 ] (Scheme 1 c,d). [9,10] Similar archetypes (including Ln III aryloxides) can also be used for chemical transformations, such as the catalytic conversion of a CO 2 /epoxide mixture into cyclic carbonates [11] or copolymers. [10a, 12, 13] However, highly reactive organo-rareearth-metal complexes such as alkyl [14] and hydride [10a, 15] (Scheme 1 a,b) or divalent derivatives [16] display irreversible CO 2 insertion or favor additional transformations through CO 2 post-activation (e.g. formation of CO, CO 3 2À , C 2 O 4 2À ). [16] Recently, cerium, the most abundant rare-earth element, has gained attention for CO 2 activation. [8,[17][18][19] For example, while the hydrogen-bonded Ce IV oxo complex [(L OEt ) 2 [17] ortho-NHC-substituted aryloxide Ce III complexes (NHC = N-heterocyclic carbene) insert CO 2 into the Ce À C NHC bond in a semireversible manner, and catalytically form propylene carbonate from propylene oxide. [18,19] Bulky cyclopentadienyl (Cp) derivatives (e.g. Ln-(C 5 Me 5 ) 3 ) were shown to accommodate CO 2 insertion in a unidirectional manner, thereby forming very stable carboxylato moieties through a h 5 -to-h 1 switch in the C 5 Me 5 coordination (cf. Scheme 1 a). [20] Pyrazolates (pz), on the other hand, are dinitrogen-derived Cp counterparts, where the putative N À CO 2 bond may tolerate a more reversible insertion process, as seen for other CO 2 -heteroatom bonds (Scheme 1 e). [21,22] As the tetravalent [Ce(Me 2 pz) 4 ] 2 complex was recently shown to undergo reversible insertion of ketones into the CeÀN bond, [23] we extended the study toward CO 2 . Quantitative insertion of CO 2 into the Ce À N(Me 2 pz) bond was observed for both tetravalent and trivalent cerium Me 2 pz complexes, and intriguingly the insertion process was found to be fully reversible. Scheme 1. Irreversible reaction of carbon dioxide with archetypal organo-rare-earth-metal complexes, with the exception of (e) as shown for (C 5 Me 5 ) 2 Sm(SePh)(thf). [22] Results and Discussion Carbon Dioxide Insertion into a Ceric Pyrazolate: Treatment of [Ce(Me 2 pz) 4 ] 2 (1) with excess CO 2 in either toluene or thf (under 1 bar CO 2 pressure) led to a color change from dark red to orange within 5 minutes (Scheme 2). Crystallization from concentrated toluene or thf solutions at À40 8C gave orange crystals of [Ce(Me 2 pz·CO 2 ) 4 ] with either toluene (2·toluene, 54 %) or thf (2·thf, 64 %) within the lattice. Discounting the lattice solvent, this accounts for about 25 wt % CO 2 or 5.7 mmol CO 2 per gram.
The molecular structure of 2·toluene revealed an 8coordinate cerium(IV) center with four k 2 (N,O)-coordinating Me 2 pz·CO 2 ligands (Figure 1), in contrast to the k 2 (O,O) modes in carboxylates and related carbamates. The Ce À N and Ce À O bond lengths average 2.528 and 2.255 , respectively, thus matching the values found in the benzophenoneinserted product [Ce(Me 2 pz) 2 (pdpm) 2 ] (Ce1ÀN1 2.564 , Ce1ÀO1 2.173 ; pdpm = (3,5-dimethylpyrazol-1-yl)diphenylmethanolate). [23] Other homoleptic Ce IV complexes, [Ce-(L) 4 ] (with L as a donor-functionalized alkoxy ligand engaged in a 5-membered chelate to cerium), also have similar Ce À O bond lengths as those in 2, thus highlighting a common chelating coordination motif. [24,25] Seemingly, no delocalization occurs across the O=CÀO fragment, which exhibits average CÀO bond lengths of 1.207 (terminal) and 1.291 (bridging). Support for the localization of the CÀO double bond came from DRIFTS measurements of 2·toluene and 2·thf, which showed the presence of a strong absorption band at ñ = 1732 and 1718 cm À1 , respectively, for the CO stretching of the C À O double bond as well as a strong absorption band at ñ = 1336 cm À1 for the CO stretching of the C À O single bond.
The structure was also supported by NMR spectroscopic measurements. The 1 H NMR spectrum recorded in [D 8 ]toluene at ambient temperature revealed two distinct methyl group environments for all the pyrazolato ligands, indicative of ligand asymmetry and complete consumption of [Ce(Me 2 pz) 4 ] 2 . The 13 C signal of the inserted CO 2 was detected at d = 149.9 ppm, a region where pyrazolate N-CO 2 R signals are expected. [26] 1 H NMR measurements on 2·thf in [D 8 ]THF at ambient temperature showed a mixture of products, which could not be assigned. Cooling the solution to À40 8C under 1 bar CO 2 pressure led to a color change from red to orange and both the 1 H and 13 C NMR spectra recorded at À40 8C showed similar signals as 2·toluene in [D 8 ]toluene.
Variable-temperature (VT) NMR studies of 2·toluene and 2·thf in [D 8 ]toluene and [D 8 ]THF were conducted to investigate the reversibility of CO 2 insertion (Scheme 3). In [D 8 ]toluene, the formation of a new species at 40 8C was revealed and no further liberation of CO 2 was observed even after heating above 60 8C (see Figure S6 in the Supporting Information). The 1 H NMR spectrum recorded at 40 8C shows two sets of signals for different Me 2 pz moieties in a 1:1 ratio, which suggests the formation of putative compound [Ce- Recooling the solution did not reform 2·toluene quantitatively, likely because some of the liberated CO 2 was no longer within the reaction medium, but the addition of fresh CO 2 quantitatively reformed 2·toluene. The [D 8 ]THF VT NMR experiment of compound 2·thf showed a different CO 2deinsertion behavior (see Figure S10). As a consequence of competitive thf coordination, displacement of CO 2 starts at 10 8C and is complete at 60 8C, with formation of [Ce(Me 2 pz) 4 -(thf)]. As seen in the experiment in [D 8 ]toluene, this reaction is fully reversible by recooling the sample and subsequently   introducing CO 2 . Additionally, in situ IR measurements were performed at 60 8C, which showed complete loss of inserted CO 2 and formation of free CO 2 ( Figure 2 and see also Figure S58).
In the solid state, 2·toluene is stable for several weeks at À40 8C, but at ambient temperature it partially loses CO 2 over a few days or when it is exposed to vacuum, as indicated by a color change from orange to dark red. A thermogravimetric analysis (TGA), performed under a flow of Ar and heating the sample slowly from 28 8C to 250 8C, indicated an initial loss of mostly lattice toluene. The liberation of CO 2 and small amounts of lattice toluene was dominant between 55 and 95 8C, as revealed by a step of 21.92 % (theoretical proportion of CO 2 in 2·toluene 19.98 %). At 250 8C, only nonvolatile parts of 2·toluene remain, leaving a mass of 49.79 % of the initial weight (theoretical value 51.61 %; see Figure S59). Although the deinsertion of carbon dioxide was achieved in the solid state, bulk compound 1 did not insert any carbon dioxide when stored under 1 bar CO 2 pressure for three days. Moreover, compound 1 was hydrolyzed upon exposure to air within one hour (DRIFT spectrum, see Figure S57).
[Ce(Me 2 pz) 4 (thf)] (1-thf) was treated with stoichiometric amounts of CO 2 to generate the putative [Ce-(Me 2 pz) 2 (Me 2 pz·CO 2 ) 2 ] (Scheme 3). Although this species could not be isolated, it was possible to generate monoinserted [Ce 3 (Me 2 pz) 9 (Me 2 pz·CO 2 ) 3 (thf)] (3) in moderate yields of 46 % (Scheme 4). The crystal structure of ceric 3 shows a ring motif with two distinct 9-coordinate and one 10coordinate cerium atoms ( Figure 3). Although all the cerium centers are coordinated by three Me 2 pz ligands in an h 2 (N,N') fashion, Ce1 connects further to two oxygen atoms of neighboring Me 2 pz·CO 2 ligands as well as an additional thf molecule, 10-coordinate Ce2 is surrounded by two k 2 :(N,O)chelating Me 2 pz·CO 2 ligands, and Ce3 exhibits additional contacts to one k 2 :(N,O)-Me 2 pz·CO 2 ligand and an oxygen atom of a neighboring Me 2 pz·CO 2 ligand.
Each Me 2 pz·CO 2 ligand bridges between two cerium atoms. In contrast to homoleptic 2, all the oxygen atoms are engaged in cerium bonding, which implies delocalized OÀCÀ O bonds (av. CÀO, 1.247 ). The h 2 (N,N')-CeÀN(Me 2 pz) bond lengths are in the expected range. [24,27] According to VT NMR studies carried out in [D 8 ]toluene, the trimetallic entity 3 is retained in solution at low temperatures, with every dimethylpyrazolato ligand showing a distinct signal set in the proton NMR spectrum at À80 8C (see Figure S12). The signals for the protons of the bridging Me 2 pz·CO 2 ligands are shifted upfield compared to those of the terminal h 2 (N,N')-Me 2 pz ligands. As a consequence of the equilibrium between potential alternative oligomers formed in the presence of only one equivalent of CO 2 per cerium, the interpretation of the ambient-temperature NMR spectrum was difficult. This was already experienced for the insertion of benzophenone into the CeÀN(Me 2 pz) bond. [23] Upon heating to 90 8C, CO 2 was liberated and [Ce(Me 2 pz) 4 ] re-formed (see Figure S14). After cooling to ambient temperature, a partial reinsertion of CO 2 was observed, as was found in the VT NMR experiments on 2·toluene and 2·thf.
Carbon Dioxide Insertion into a Cerous Pyrazolate: To examine the role of the oxidation state of cerium (Ce IV versus Ce III ) and, therefore, the impact of its Lewis acidity, cerous donor-free [Ce 4 (Me 2 pz) 12 ] [28] (4) was used as a precursor for CO 2 insertion (Scheme 5). Remarkably, while retaining the    Figure 4). Compared to the six different coordination modes of the pyrazolato moieties in starting material 4, [28] the crystal structure of 5 revealed only three. Ce2, Ce3, and Ce4 form a nearly equilateral triangle bridged by m 3 -1k 2  indicate delocalization of the bridging Me 2 pz·CO 2 ligands, with two C À O bond lengths in the same region (1.235(9)-1.264(9) ) and rather localized C À O single (1. 26(1)-1.301-(9) ) and C À O double bonds (1.21(1)-1.229(9) ) for the capping and terminal Me 2 pz·CO 2 ligands. For further comparison, the cerous carbamate [Ce 4 (O 2 CNiPr 2 ) 12 ] features a lozenged arrangement of one 8-coordinate and three 7coordinate Ce III centers with CeÀO bond lengths in the range 2.322(7)-2.746(7) (no Ce-N interaction). [29] The latter complex was obtained from the reaction of CeCl 3 (DME) with HNiPr 2 and CO 2 (Scheme 1f). In accordance with the crystal structure of cluster 5, DRIFTS measurements show strong absorption bands for both the CÀO single bonds (ñ = 1250-1350 cm À1 ) and CÀO double bonds (ñ = 1600-1750 cm À1 ). 1 H DOSY NMR measurements on 5 in [D 8 ]toluene, [D 8 ]THF, or a [D 8 ]toluene/3,3-dimethyl-1,2-butylene oxide mixture revealed distinct diffusion coefficients (see Figures S17-S20) for the solvents employed and only one additional peak corresponding to a much larger species but correlating with every other signal in the proton NMR spectra. Calculation of the molar mass of this compound ([D 8 ]toluene: M r = 1989 g mol À1 ; [D 8 ]THF: M r = 1643 g mol À1 ; [D 8 ]toluene + 3,3-dimethyl-1,2-butyleneoxide: M r = 2357 g mol À1 ) as a compact sphere-like molecule [30] suggests it exists as a tetrametallic (M r = 2242 g mol À1 ) or a non-monometallic species in solution. Treating [Ce(Me 2 pz) 3 (thf)] 2 with CO 2 in [D 8 ]THF gave the same NMR spectrum as that of 5·toluene, thus indicating the formation of a multimetallic compound also in donor solvents (see Figure S16). TGA of 5·toluene also showed an initial loss of toluene (cf. 2·toluene), followed by a pronounced step (21.39 % weight loss) in the range from 52 to 90 8C, consistent with the release of CO 2 (theoretical value: 16.75 %) and some lattice toluene ( Figure S60). At 250 8C, only the nonvolatile parts of 5·toluene remain and a total loss of 46.82 wt % compared to the starting material fits well with the theoretical value of 45.99 % for 10 molecules of toluene and 12 molecules of CO 2 eliminated from [Ce 4 (Me 2 pz·CO 2 ) 12 ]·10 toluene (5·toluene).
Catalytic Formation of Cyclic Carbonates from CO 2 and Oxiranes: Having established the efficiency and reversibility of CO 2 insertion into Ce À N(Me 2 pz) bonds, we were interested in any catalytic utilization. Accordingly, pyrazolate complexes 1 and 4 were probed as catalysts for the generation of cyclic carbonates from CO 2 and oxiranes. In the absence of CO 2 , compound 1 interacts with epoxides, as indicated by a noticeable shift in the 1 H NMR spectrum upon addition of one equivalent of 3,3-dimethyl-1,2-butylene oxide (see Figure S24). This is most likely due to formation of a donor adduct, which is considered a crucial step in Lewis-acidcatalyzed cycloaddition reactions. Even though such donor adducts were shown to be isolable (e.g. Tp tBu Ca(O-2,6-iPr 2 C 6 H 3 )·(PO); Tp tBu = tris(3-tBu-pyrazolyl)borato, PO = propylene oxide [31] ), the putative [Ce(Me 2 pz) 4 (PO)] (1-PO) could not be isolated. Tetra-n-butylammonium bromide (TBAB) was employed as a co-catalyst, since it was shown to promote the highest activities in such cycloaddition reactions. [11b, 32] The reaction was optimized for propylene oxide, which gave almost quantitative conversion after 24 h under mild conditions. Using 0.5 mol % 1 or 0.25 mol % 4 and  1 mol % TBAB without solvent at ambient temperature and 1 bar CO 2 pressure gave 93 % conversion for the tetravalent catalyst 1 and 98 % for its trivalent counterpart (Table 1, entries 1 and 3). The CO 2 -insertion complexes 2 and 5 displayed similar catalytic activity (entries 2 and 4). The conversion dropped drastically on increasing the steric bulk of the substituent on the epoxides. As a result, styrene oxide, and 3,3-dimethyl-1,2-butylene oxide showed only very low conversions (entry 21) for ceric 1 and almost no conversion in the case of trivalent catalyst 4 (entry 22). Moderate conversion was observed for 1,2-n-hexylene oxide with the tetravalent catalyst 1 (entry 13). Conducting the catalysis at higher temperature increased the TONs with both catalysts 1 and 4, and resulted in nearly quantitative conversion for both systems (entries 15 and 16). Without co-catalyst TBAB, 1 showed moderate catalytic activity at 90 8C (entry 17). In almost all cases, tetravalent 1 showed higher catalytic activity than cerous 4, which most likely results from the higher Lewis acidity of Ce IV versus Ce III . To further evaluate the catalytic reaction with catalyst 1 and propylene oxide, the TOFs at different stages of the catalysis were determined (see Table S1). After a short induction period, most likely corresponding to the insertion of CO 2 into 1, the TOF reached a maximum of 11 h À1 within the first 3 h. For comparison, a TOF of 155 h À1 was reported by Yao and coworkers when performing the reaction under 10 bar CO 2 pressure. [11a] Increasing the CO 2 pressure did not significantly affect the catalytic activity of compound 1 (entries 13 vs. 18 and 21 vs. 23). However, a simultaneous increase of the temperature to 90 8C and the CO 2 pressure to 10 bar led to a marked improvement in the catalytic activity, resulting in TONs of up to 300 for the sterically demanding 3,3-dimethyl-1,2-butylene oxide (entries 24 and 26). The latter conditions were also applicable for the cycloaddition of CO 2 and cyclohexene oxide, an internal epoxide (entries [27][28][29]. Having optimized the reaction conditions, we determined the initial turnover frequencies for the different epoxides (entries 5, 12, 20, and 25). As expected, the TOFs increased as the steric of the substituents bulk decreased, ranging from 24 to 196 h À1 and giving almost quantitative conversion of propylene oxide after a reaction time of one hour (entry 5). Compared to the other catalyst systems based on rare-earth metals reported by Yao and co-workers (TOFs up to 440 h À1 ) or by Otero and co-workers (3167 h À1 ), our system shows only moderate catalytic activity under comparable conditions. [11a,b] The mechanism of the cycloaddition of CO 2 and epoxides using tetraalkylammonium salts as co-catalysts has been discussed in detail. [32,33] It is generally accepted that the epoxide is activated by coordination to a Lewis-acidic metal center followed by a nucleophilic ring-opening attack of the bromide to form a metal-alkoxy bond (Scheme 6). Subsequently, the alkoxide reacts with CO 2 and cyclizes to produce a cyclic carbonate. Hints that the mechanisms for the cycloaddition differ using ceric 1 or cerous 4 as the catalyst could be found when conducting the reactions with different amounts of catalyst loading (entries 7 and 9). This results in a change in the TONs for the tetravalent catalyst 1, whereas the TONs remained the same for trivalent complex 4. The occurrence of distinct reaction mechanisms is not surprising, as 1 is a monometallic complex while 4 is a tetrametallic compound in the solid state and in solution (for a more detailed possible mechanism see Scheme S1). Such a mechanism, involving multiple metal centers, was previously proposed for the bimetallic complex [Al(salen)] 2 O by North and co-workers. [32] However, any detailed information about the mechanism could not be retrieved from our catalyst system as the TOFs decreased enormously when the reactions were conducted in propylene carbonate (no significant conversion at ambient temperature after 24 h and ca. 10 % conversion after 20 h at 90 8C) or any other solvent, which makes kinetic studies unfeasible.  8 ]toluene under 1 bar CO 2 pressure. Although 2 is very stable in toluene solution under these conditions (no color change was observed after several days), it underwent reduction in the presence of TBAB within hours, as evidenced by decolorization of the orange solution. Most probably bromine is formed as an oxidation product; however, no brominated product could be detected in the reaction mixture (see Figure S23). The use of epoxides as a solvent seems to stabilize the tetravalent 2, as no paramagnetic signals were found in the 1 H NMR spectra of the catalytic reactions. The crystal structure of 6 revealed the same motif as seen in 2·toluene and 2·thf (see Figure S67). The 8-coordinate cerium center bears four k 2 (N,O) Me 2 pz·CO 2 ligands with elongated Ce À N and Ce À O bonds compared to ceric 2·toluene and 2·thf, as would be expected for a cerium(III) center. Cerous 6 displays poor catalytic activity compared to tetravalent 1 (Table 1, entry 19), thus underlining that it is a side product and not the active catalyst.

Conclusion
We have shown that carbon dioxide easily inserts into the CeÀN(Me 2 pz) bond of both ceric [Ce(Me 2 pz) 4 ] 2 and cerous [Ce 4 (Me 2 pz) 12 ] at an amount equivalent to 5.7 mmol CO 2 per gram complex and via the controlled activation of 12 molecules of CO 2 within one complex, respectively. The insertion process is reversible both in solution and in the solid state, with CO 2 desorption being complete at < 100 8C. Both trivalent and tetravalent cerium pyrazolate complexes are active catalysts for the cycloaddition of epoxides and carbon dioxide with TBAB as a co-catalyst under mild conditions. We are currently investigating the carbon dioxide capture performance of silica-grafted variants of Ce-pyrazolates, [34] and our findings might also stimulate research in the area of cerium-dipyrazolate-based CO 2 -"breathable"/expandable MOFs. [35,36]