A Zintl Cluster for Transition Metal-Free Catalysis: C=O Bond Reductions

The first fully characterized boron-functionalized heptaphosphide Zintl cluster, [(BBN)P7]2– ([1]2–), is synthesized by dehydrocoupling [HP7]2–. Dehydrocoupling is a previously unprecedented reaction pathway to functionalize Zintl clusters. [Na(18-c-6)]2[1] was employed as a transition metal-free catalyst for the hydroboration of aldehydes and ketones. Moreover, the greenhouse gas carbon dioxide (CO2) was efficiently and selectively reduced to methoxyborane. This work represents the first examples of Zintl catalysis where the transformation is transition metal-free and where the cluster is noninnocent.


■ INTRODUCTION
Methanol (CH 3 OH) is a clean fuel and a highly important raw material for chemical industries. 1−3 Over half of the world's methanol is upcycled into everyday products, including pharmaceuticals, adhesives, agrochemicals, and paints/coatings. Producing methanol from carbon dioxide (CO 2 ) has attracted global attention, because it converts a greenhouse gas into a resource that can re-enter the energy cycle. This approach is "two birds one stone" in contributing to global climate control and sustainable energy efforts. 4,5 On an industrial scale, the conversion of atmospheric CO 2 into methanol was first realized in 2012 by the George Olah Plant (Iceland), which produces up to 4500 m 3 of methanol per year. 6 The intrinsic stability of CO 2 means that catalysts are essential for efficient reduction, and these are often based on expensive metals. 7 The large-scale development and utilization of CO 2 reduction necessarily requires any process to be sustainable and cost-effective, and so identifying less expensive and sustainable alternatives to these metals is an important target.
Heterogeneous phosphorus-containing materials represent one possible alternative to transition metals and are currently being explored for CO 2 reduction. 8,9 The synthesis of welldefined molecular analogues of these phosphorus materials offers the opportunity for mechanistic investigation, and molecular clusters offer an important middle ground between molecules and bulk solids. Zintl clusters, in particular, can be thought of as molecular mimics for heterogeneous materials: 10 for example, the structure of [P 7 ] 3− can be viewed as a fragment of red phosphorus, 11−14 which is an inexpensive and abundant material, albeit a challenging one to study because of its poor solubility. In contrast, [P 7 ] cages, especially those that have been functionalized, are soluble in common laboratory solvents, offering a broader range of handles for in situ investigation. An improved understanding of reactivity patterns for [P 7 ] could be extended to corresponding reactions with materials based on the many allotropes of phosphorus.
Only a small number of catalytic applications involving Zintl-derived clusters have been reported, where the cluster typically acts as a spectator ligand, supporting active rhodium or iridium centers 10,15,16 (Figure 1) Catalysis of the reverse water−gas shift reaction by a Ru/Sn Zintl cluster, [Ru@Sn 9 ] 6− , has been reported in the recent literature, although the degree to which the cluster remains intact when deposited on a CeO 2 surface remains to be established. 17 The reaction chemistry of clusters based on the [P 7 ] framework is an emerging field, 11,12,18 but salt metathesis with group 14 electrophiles has already proven to be a powerful route to functionalizing the cluster. Further, in 2012, Goicoechea and co-workers reported the use of the protonated heptapnictide clusters [HPn 7 ] 2− (Pn = P, As) in hydropnictination reactions with carbodiimides and isocyanates. 19−21 They also reported that the [Pn 7 ] 3− (Pn = P, As) cluster reacted with alkynes to afford 1,2,3-tripnictolides, 22,23 while reaction with carbon monoxide afforded the [PCO] − anion. 24 In 2021, we reported that the trisilylated derivatives (R 3 Si) 3 P 7 (R = Me, Ph) captured 3 equiv of heteroallene and underwent subsequent small-molecule exchange reactions. 25 Herein we report that Zintl clusters based on the [P 7 ] architecture are catalytically competent for borohydride reductions. Specifically, we prepare the transition metal-free functionalized heptaphosphide Zintl cluster shown in Figure 1 and establish its ability to catalyze the reduction of C�O bonds. Our initial focus is on the reduction of organic aldehydes and ketones, where the progress of reactions can readily be monitored via NMR spectroscopy, and then move on to heteroallenes where two double bonds are present. This then leads naturally to a survey of CO 2 hydroboration, where complete selectivity to methoxyborane is observed. This Zintl catalyst displays turnover numbers, turnover frequencies, recyclability, and selectivity that are competitive with main group catalysts reported in the literature, under mild conditions. ■ RESULTS AND DISCUSSION

Synthesis of Catalyst and Stoichiometric Studies.
Inspired by advances made in molecular frustrated Lewis pair (FLP) chemistry, 26 [1]) as either the sodium or potassium salt, shown in Scheme 1, along with the elimination of H 2 : gas formation could be observed during these reactions. This dehydrocoupling chemistry has no precedent in current synthetic strategies toward functionalizing Zintl clusters. Salt metathesis reactions using BBNOTf (OTf = triflate), Cy 2 BI, or Cy 2 BOTf with [P 7 ] 3− did not result in functionalization of the cluster, but instead gave NMR spectra consistent with decomposition of the cluster.
Nuclear magnetic resonance (NMR) spectroscopy studies of [M(18-c-6)] 2 [1] revealed five resonances in the 31 P NMR spectrum, each exhibiting extensive P−P coupling, along with a single relatively sharp resonance in the 11 B NMR spectrum at 11.14 ppm (Supporting Information, Figures S2 and S4). These spectroscopic features are consistent with the [P 7 ] cage having a mirror plane and κ 2 -coordination of the BBN moiety to the cluster. Similar 31 P NMR spectroscopic features were reported for the structurally related [(Ph 2 In)P 7 ] 2− cluster. 35 Single-crystal X-ray diffraction (XRD) studies were performed on both sodium and potassium salts [Na(18-c-6)] 2 [1] and [K(18-c-6)] 2 [1]. In both cases, consistent with the spectroscopic data, the BBN moiety in [1] 2− is coordinated to the [P 7 ] fragment in a κ 2 -coordination mode ( Figure 2). The average B−P bond length of 2.072 Å in[1] 2− is somewhat longer than B−P bonds in typical borylphosphines (R 2 P−BR 2 , 1.889−1.953 Å) and considerably longer than those in typical phosphinoborenes (R 2 P�BR 2 , 1.762−1.857 Å), where P�B π bonding is significant. The sum of the angles around phosphorus is 276°, also typical of borylphosphines (284− 328°) rather than phosphinoborenes (359.8−328.3°). 36,37 All efforts to react [Na(18-c-6)] 2 [1] directly with H 2 were unsuccessful. However, stoichiometric reactions of [Na(18-c-6)] 2 [1] with carbonyls, including benzaldehyde and acetophenone, and heteroallenes such as phenyl isocyanate and CO 2 , resulted in an immediate color change from orange to red. Unfortunately, and despite multiple efforts, XRD diffraction quality crystals could not be obtained from any of these   reactions, but in situ 31 P and 11 B NMR spectroscopy confirmed the disappearance of [Na(18-c-6)] 2 [1], along with the formation of novel (asymmetric) 25 functionalized clusters that would be consistent with formation of a carbonyl or heteroallene adduct (see Supporting Information, Figures S126−S138). We note that others have previously isolated and crystallographically characterized adducts of carbonyls or heteroallenes with FLPs where the C−O bond bridges the B··· P gap. 38−41 The clear color change and disappearance of [Na(18-c-6)] 2 [1] in these carbonyl and heteroallene reactions but not with H 2 follows established reactivity patterns for borylphosphines 42 rather than phosphinoborenes. 43 Catalytic Reduction of Carbonyls.
[M(18-c-6)] 2 [1] salts (M = Na, K) were then explored for their potential as catalysts in the reduction of carbonyls, a transformation that is widely employed by the pharmaceutical, agrochemical, polymer, and fine chemical industries. 44 Initially the reduction of benzaldehyde and benzophenone was targeted (Table 1). In tetrahydrofuran (THF) with 5 mol % [Na(18-c-6)] 2 [1] catalyst loading and the mild reductant pinacolborane (HBpin), the hydroboration of benzaldehyde (2a) and benzophenone (3a) was quickly achieved at room temperature (RT) to give the benzyloxyboranes 2b and 3b in high yields. Lowering the catalysts loading under the same conditions decreased the conversion, but it was found that a change of solvent to ortho-difluorobenzene (oDFB) improved the yield, giving near quantitative conversions even at 1 mol % catalyst loadings, under similar conditions. Hydrosilylation of the carbonyls using triethylsilane and triphenylsilane was also tested, but no reactions were observed. Changing the countercation from sodium to potassium did not affect catalyst performance. Control reactions confirmed that catalyst [M(18c-6)] 2 [1] was necessary and also that the unfunctionalized K 3 P 7 salt was completely inactive in this hydroboration. The [Na(18-c-6)] 2 [HP 7 ] salt was also tested as a precatalyst and displayed lower catalytic performance compared to [M(18-c-6)] 2 [1].
Next, the scope of ketone hydroborations was extended (Table 3). Substituting one phenyl group of benzophenone for a 2-pyridyl group (3a vs 14a) leads to a significantly longer reaction time (3 vs 18 h, respectively). In contrast, the hydroboration of di(pyridyl) ketone (15a) only required 30 min to give complete conversion to 15b. Coordination of the pyridyl group on the aldehyde to free HBpin could be a factor in determining the rate of these reactions; in support of this proposal, it was found that one of the pyridine nitrogens of 15b coordinates to the Bpin boron center in its crystal structure (Supporting Information, Figure S66). Similar to the hydroboration of benzophenone (3a), acetophenone (16a) and thiophene-2-carboxaldhyde (17a) gave complete conversion to 16b and 17b, respectively, after 30 min.
The incorporation of electron-withdrawing substituents resulted in longer reaction times, for example in the case of 1-(perfluorophenyl)ethan-1-one (18a) and acetylferrocene (19a), which required 72 h to give high conversions to 18b and 19b, respectively. The sterically bulky adamantanone (20a) gave 20b after 18 h. The alkyl-substituted ketone cyclobutyl methyl ketone (21a) resulted in 21b after 2 h. Introduction of ethynyl and alkenyl groups on the ketone (22a and 23a) resulted in the formation of byproducts, decreasing the conversion to the desired hydroborated product.
To investigate the chemoselectivity of carbonyl hydroborations, a competition reaction was carried out between benzaldehyde, 2a, and acetophenone, 16a, with 1 equiv of HBpin (Scheme 2). Exclusive hydroboration of benzaldehyde was observed under these conditions, but addition of a second equivalent of HBpin resulted in hydroboration of the acetophenone as well. Neither 2b or 16b was found to undergo deoxygenation in the presence of an excess of HBpin. The selectivity of reduction toward aldehydes in the presence of ketones is consistent with literature precedent 45−47 and with the data in Table 1, which shows higher degrees of conversion for 2a compared to 3a.
Catalytic Reduction of CO 2 . Isocyanates (RN�C�O) are, of course, structurally related to the greenhouse gas CO 2 (O�C�O), and so the successful deoxygenation of isocyanates to methyl amines motivated further investigations into the catalytic reduction of CO 2 . Molecular main group catalysts to convert CO 2 to products such as formic acid (HCOOH), methanol (CH 3 OH), methane (CH 4 ), and carbon monoxide (CO) have been previously reported, 48 perhaps the best known of which are the FLP catalysts used to hydroborate or hydrosilylate CO 2 . 29,49 Triethylsilane and triphenylsilane were tested in the catalytic reduction of CO 2 and gave no conversion ( Table 5), confirming that, as was the case for the carbonyl compounds, silanes are not suitable reducing agents to drive the reduction of CO 2 using[Na(18-c-6)] 2 [1] as a catalyst. Next, 3 equiv of HBpin with 3.33 mol % [Na(18-c-6)] 2 [1] was pressurized with 1 atm of CO 2 at RT and found to give a mixture of products. In good agreement with literature reported chemical shifts, 50 the formic acid (  Journal of the American Chemical Society pubs.acs.org/JACS Article were screened (see Supporting Information Section 5.1). Catecholborane (HBcat) and borane dimethylsulfide (BH 3 · SMe 2 ) did not result in detectable amounts of product, but when the HBBN dimer, which has pinned back aliphatic groups and an accessible hydride, is used as the reductant, the conversion is 95% with a product distribution of 1:9 formylborane (28b):methoxyborane (28d). Control experiments confirmed that the naked [P 7 ] 3− clusters (as K or K(18c-6) salts) or HBBN dimer as catalyst is not independently catalytically active (Table 5, entries 7−9). The tris-functionalized P 7 cluster (Me 3 Si) 3 P 7 was also prepared 34 and found to be catalytically inactive (Table 5, entry 10). These controls confirm that the BBN moiety and P 7 cluster of [1] 2− cooperate to enable catalytic activity. The effect of varying reaction conditions with the HBBN dimer as the reductant is summarized in Table 6. Reducing the catalyst loading from 3.33 mol % to 0.33 mol % increased the selectivity toward methoxyborane (MeOBBN, 28d) to >99%, while the introduction of toluene as a cosolvent improved solubility of (HBBN) 2 and thus increased the turnover frequency (TOF). Increasing the temperature from RT to 50°C further increased the TOF to 300 while lowering the catalyst loading to 0.01 mol %, giving the maximum turnover number (TON) of 9800. Cluster decomposition is observed above 50°C, limiting catalyst screening conditions at high temperatures. The maximum TON and TOF ( In addition to the excellent TON observed, catalyst [Na(18c-6)] 2 [1] was found to be surprisingly robust: after complete catalytic hydroboration of CO 2 to methoxyborane (28d) no decomposition of [Na(18-c-6)] 2 [1] was detected by NMR spectroscopy. To investigate whether catalyst [Na(18-c-6)] 2 [1] could be recycled, the reaction sample was reloaded with 3 equiv of HBBN and 1 atm of CO 2 . Again, at RT, complete conversion to 28d was observed overnight. Catalyst [Na(18-c-6)] 2 [1] was recycled a total of seven times with no loss of catalyst activity, consistent with living catalysis. After these cycles, water hydrolysis of methoxyborane gave methanol in complete conversion: 0.62 mmol of methanol was produced using 0.9 μmol of catalyst (a 689-fold excess of methanol).
The CO 2 hydroboration with HBBN dimer was monitored by 1 H NMR spectroscopy with 0.33 mol % [Na(18-c-6)] 2 [1] at room temperature and 50°C (Table 6, entries 2, 3). In both cases, intermediate CH 2 (OBBN) 2 (28c) quickly formed and then was consumed (Figure 3). Meanwhile 28d is formed gradually throughout the reaction. In line with literature reports, the intermediate formyl-BBN (28b) was not detected under these conditions and is presumed to be significantly more reactive toward hydroboration compared to CO 2 , 28c, and 28d.
Stoichiometric reduction of CO 2 using equimolar [Na(18-c-6)] 2 [1] and HBBN exclusively gave HC(O)OBBN (28b). Further, reaction of [Na(18-c-6)] 2 [1] with 2 equiv of HBBN under a CO 2 atmosphere again exclusively gave 28b, indicating that CO 2 is not trapped and reduced to 28b then further reduced to 28c and 28d while intact on the catalyst. Rather, the evidence suggests that 28b is released after one addition of borane but then re-enters the catalytic cycle to form 28c and 28d.   Mechanistic Investigations. In order to probe the mechanistic landscape for the reduction of C�O functional groups, we have performed DFT calculations on the simplest model substrate, formaldehyde ( Figure 4). The optimized structure of [1] 2− reproduces the crystallographic data with good accuracy: the optimized B−P bond lengths are 2.07 Å (vs 2.052(8) and 2.082(7) Å in Figure 2), while P1−P2 and P1− P3 are 2.21 Å vs crystallographic values of 2.213(2) and 2.216(2) Å, respectively. In its equilibrium structure, the boron center in [1] 2− is saturated by two B−P bonds, but a wider survey of the potential energy surface reveals a second shallow local minimum only 15.1 kcal/mol above the equilibrium structure, where the BBN fragment is coordinated to only one of the two phosphorus centers (isomer [1′] 2− , Figure 4). In this case the single B−P bond length is 1.89 Å, precisely in the range for borylphosphines.
In the mechanistic scheme shown in Figure 4 Some support for the presence of I4 in solution comes from monitoring of the hydroboration of 5a, 7a, 11a, 14a, and 17a (see Supporting Information, Section 6.4). 1 H NMR studies on the hydroboration of 5a, 7a, and 11a reveal the presence of a second-order AB spin system with similar chemical shifts and coupling as the hydroborated products (selected spectra, Figure 5a). The AB spin system is consistent with the presence of a pair of enantiotopic protons displaying a 10−12 Hz geminal coupling, an assertion that is further supported by 1 H correlated NMR spectroscopy (COSY) studies (Supporting Information, Figure S142). 1 H diffusion-ordered NMR spectroscopy (DOSY) studies on the hydroboration of 7a (Supporting Information, Figure S143) confirmed that the species giving rise to the AB spin system has the lowest diffusion coefficient, indicating that it is the largest species present in the reaction mixture, in line with the proposed structure for I4 given in Figure 4. In contrast, the analogues of I4 detected from the hydroboration of the ketones 14a and 17a, where only a single proton is present, display similar resonances but without further geminal coupling (selected spectra, Figure 5b). The characteristic resonances for I4 are observed only during the reaction and disappear when the reaction is complete. Moreover, addition of 500 equiv of 7b to in situ generated I2 did not result in formation of I4, confirming the irreversibility of the I4 → I2 step.
In contrast, a direct spectroscopic signature of I2 could not be identified in any of the reactions with aldehydes or ketones. However, we note that insertion of carbonyl groups into P−B bonds has precedent in the work by Fontaine and Ramos, where C−O bonds in the crystallographically characterized products are typically in the range 1.39−1.40 Å. 57 The optimized structure of I2 (Supporting Information, Figure  S165) shows a similarly activated C−O bond (C−O = 1.37 Å, B−O = 1.47 Å), along with a short P−C bond length of 1.89 Å. Moreover, during stoichiometric CO 2 capture experiments with [Na(18-c-6)] 2 [1] using 13 C-labeled CO 2 , we observe a doublet at 194.13 ppm with 1 J CP = 49 Hz in the 13 C{ 1 H} NMR spectrum. A C−P coupling constant of this magnitude is qualitatively consistent with the presence of a direct C−P bond as in I2 (or indeed I4), and the DFT-computed value (wB97XD/def2-TZVP) in the analogue of I2 with CO 2 (see Supporting Information, Figure S170) rather than formaldehyde as substrate is 41 Hz.
The presence of significant concentrations of I4 would, in principle, also opens up the possibility of a third cycle, I2 (Figure 4, top, center), where the total exothermicity of −50.4 kcal/mol is distributed evenly across the borohydride binding, formaldehyde binding, and product release steps (−12.3, −25.8, and −12.4 kcal/mol, respectively). In their recent study of CO 2 reduction using Ph 2 PCH 2 CH 2 BBN, Ramos et al. have argued that trapping of CO 2 by the B/P FLP leads to formation of a formaldehyde adduct analogous to I2, 57 which, in fact, acts as the active catalyst in the dominant CO 2 reduction cycle. By analogy, [1] 2− would then be a precursor, lying outside the main cycle, while I2 is the active catalyst. Musgrave and coworkers have highlighted the "anticatalytic" role of the Lewis base in frustrated Lewis pairs: 59 the Lewis base, in binding to the nucleophilic site (the carbonyl carbon here), reduces its susceptibility to subsequent attack by the reducing agent. In the present context, the very strong binding of formaldehyde to [1] 2− (−33.2 kcal/mol) will reduce its susceptibility to attack by borohydride, and so the somewhat weaker binding of the second aldehyde (I3 + H 2 CO → I4, −25.8 kcal/mol) may accelerate its subsequent reduction by borohydride. The computed energetics, combined with the spectroscopic evidence for the presence of I2 and I4 in solution and the literature precedent for species like I2 to act as catalysts, leads us to propose that this cycle is likely to dominate the reaction. At this point, however, we offer an important caveat, that the binding of the carbonyl group is strongly dependent on the size of the R and R′ groups: the −33.2 kcal/mol computed for the binding of H 2 CO to [1] 2− is reduced to −25.2 kcal/mol for benzaldehyde (PhCHO). The energetics in Figure 4 for the simplest model aldehyde, formaldehyde, therefore offer an overview of potential intermediates on the mechanistic landscape, but should not be taken to apply quantitatively to any of the experimental data summarized in Tables 1−3. Stoichiometric studies between [Na(18-c-6)] 2 [1], acetophenone, and HBpin confirmed (Scheme 3) compound 16b to be the major product, but its BBN analogue 16b′ was also detected as a minor product in a 95:5 ratio. One possible source of the minor isomer is the methoxy-bridged intermediate, I3 in Figure 4, where the oxygen is almost symmetrically bonded to both boron centers (B−O = 1.59 and 1.60 Å to the BBN and [B] groups, respectively). Cleavage of the two almost equivalent B−O bonds could then lead to either 16b or 16b′. At the wB97XD/def2-TZVP level, decomposition to give 16b′ (+ [B]−P 7 ) was found to be less favorable by 15.1 kcal/mol (ΔG = −16.2 kcal/mol) than to 16b (+ [1] 2− ), qualitatively consistent with the observed product distribution. The formation of 16b′ in the experiment does, however, indicate a 5% degradation of the catalyst under these conditions. Of course, a cluster functionalized with Bpin ([(Bpin)P 7 ] 2− ) could also be catalytically active, following equivalent pathways to those shown in Figure 4, and a catalytic role for [(Bpin)P 7 ] 2− would be consistent with our description of the [Na(18-c-6)] 2 [HP 7 ] salt as a precatalyst in Table 1. We have not, however, been able to isolate or independently synthesize this species and test its catalytic performance, despite multiple attempts to do so. It is noteworthy that catalyst degradation via this route is not an issue when the HBBN dimer is used as a reducing agent, as it is in our experiments on CO 2 hydroboration, simply because the analogue of I3 is then symmetric. The mechanism of the CO 2 and heteroallene reduction reactions will be the subject of a further study, but it seems likely that, in their initial stages, at least, they will share many common features with the formaldehyde reaction shown in Figure 4. The geometries of the analogues of I2 with CO 2 , HNCNH, and HC�CCHO shown in Supporting Information, Figure S170, certainly show no significant differences in the binding mode of the substrate to [1] 2− .

■ CONCLUSION
In conclusion, a boron-functionalized group 15 Zintl cluster, [(BBN)P 7 ] 2− ([1] 2− ), is prepared and found to be competent in the catalytic hydroboration of C�O bonds. Aldehydes, ketones, carbodiimides, isocyanates, and CO 2 are all reduced, and, in the case of CO 2 hydroboration, based on catalyst performance alone, [Na(18-c-6)] 2 [1] is competitive with other main group catalysts. Further, high selectivity to methoxyborane under mild conditions and catalyst recycling is established. This work represents the first application of a Zintl cluster in transition metal-free catalysis and establishes clearly that the cluster itself can play an active role in the catalytic cycle beyond that of a spectator ligand. ■ ASSOCIATED CONTENT
The general information, experimental procedures, characterization data, and computational details (PDF)