Benchmark HCN-transfer reactions without zeolite
Since the first report on the transfer hydrocyanation reaction in 2016, Ni, Pd and Co catalysts have been developed5,19,20,25, 28–31. Lewis acid-mediated systems generate hydrocyanated alkenes with anti-Markovnikov selectivity, and are the most active since both oxidative addition and reductive elimination can be facilitated by the Lewis acid32. With Ni(COD)2 or NiCl2, DPEphos and AlMe2Cl as catalytic system, isovaleronitrile and butyronitrile can be used as HCN-donors, producing isobutene and propene as volatile and thermodynamically stable co-products5,30. In our initial catalyst screening, Pd(OAc)2 combined with Xphos and AlMe2Cl proved to be a better catalyst system for small nitrile donors, affording yields up to 25% with propionitrile as HCN-donor/solvent (Supplementary tables S2-S4). In an open system, with ethylene evaporation, the yield slightly improves to 30%. Although fairly poor, this result affirms the feasibility of using propionitrile as HCN-donor, and simultaneously highlights the strict thermodynamic limitations under which the standard reaction operates (Supplementary tables S5-S6).
Benchmark HCHO-transfer reactions without zeolite
The transfer hydroformylation reaction was first reported by Dong in 2015, although earlier studies by Brookhart laid out the prospect for such a reaction7,33. A Rh/Xantphos catalyst is used in cooperation with a benzoate counter-ion, which acts as a proton-shuttle facilitating de- and reprotonation of the Rh-complex. The system was initially applied exclusively to the dehydroformylation of valuable aldehyde feedstocks using norbornadiene as a sacrificial HCHO-acceptor. Later, the same catalyst system was used in a forward transfer hydroformylation of alkynes with butyraldehyde as sacrificial HCHO-donor13. However, as the alkyne needs to be used in excess, the practicality with regard to industrial application is fairly limited. Theoretical studies reaffirm the apparent need for an activated acceptor (e.g. strained olefin or alkyne) to drive the reaction to completion21. Thus, we aimed to challenge this notion via our zeolite-based equilibrium-shifting concept. As benchmark, a model reaction using 4-phenyl-1-butene as unactivated HCHO-acceptor was investigated, indicating that yields up to 47% could be obtained when a small aldehyde donor is used as solvent (e.g. propanal, butanal, 3-methylbutanal, pentanal; Supplementary table S7). Through further tweaking of the carboxylate counter-ion, the reaction conditions, and the use of acetic acid as additive, the yield could be improved to 55% (Supplementary table S8). Interestingly, styrenes could also be used as acceptors, but they require slightly higher reaction temperatures and work better with N-Xantphos as ligand (Supplementary table S9).
Zeolites as equilibrium-shifting agents
With this frame of reference, we started out exploring the validity of our equilibrium shifting concept by adding microporous zeolite catalysts. We initially aimed at solids containing transition metals or acid sites, as these could lead to the selective production of internal olefin dimers and trimers34–41. For each transfer reaction, an extensive list of zeolites with various pore sizes and acidity was screened (Supplementary tables S10 and S13). In the HCN-transfer reaction with isovaleronitrile as donor and 4-tert-butylstyrene as acceptor, medium-pore zeolites (with 10-membered rings; 10-MR) with high Si/Al ratio (≥ 100) such as H-ZSM-5 (MFI), H-ZSM-48 (*MRE) and H-ZSM-11 (MEL) clearly increased the yield of the HCN-transfer reaction compared to the zeolite-free benchmark. In contrast, the addition of zeolites with more Al or with large pore sizes (e.g. 12-MR) resulted in a decrease of the yield. A straightforward explanation is found in the sensitivity and small kinetic diameter of the AlMe2Cl Lewis acid; low Si/Al zeolites contain larger amounts of hydroxyl groups which react with the electrophilic Lewis acid. Indeed, inductively-coupled plasma (ICP) and 27Al magic-angle spinning NMR measurements show up to 3-fold increases in the zeolite Al-contents (Supplementary table S24 and fig. 4d). As the Lewis acid is needed for the HCN-transfer reaction, its consumption deteriorates the reaction performance (Supplementary table S11). To our delight, the commercially available H-ZSM-5 with Si/Al = 140 and 10-MR, hits the sweet spot between minimizing unwanted Lewis acid consumption and maximizing the shape-selective conversion of the alkene co-product (fig. 2a). The reaction yield with multiple small and medium-sized (≤C6) nitrile donors shows significant improvements with increasing amounts of H-ZSM-5 (Supplementary fig. S3). Most notably, with propionitrile as HCN-donor and solvent, the reaction with zeolite achieves nearly quantitative conversion (yield = 96%), compared to merely 30% without zeolite in an open system. When larger nitrile donors are used (e.g. 3-phenylpropionitrile), the zeolite has little to no effect, which hints at shape-selectivity. Kinetic experiments clearly visualize the beneficial effect of H-ZSM-5, showing that the reaction immediately stalls when the zeolite is removed at any point (Supplementary fig. S14).
In the transfer hydroformylation reaction with small, branched aldehydes as donors (isobutanal, 3-methylbutanal) and 4-phenyl-1-butene as acceptor, the yield clearly increases with the addition of acid zeolites, of which H-ferrierite (FER, Si/Al = 9) is most effective (Supplementary table S13). For instance, the HCHO-transfer reaction yield with 3-methylbutanal as donor could be increased from 55–80% by adding 100 mg H-FER (Fig. 2b). As most of the remaining 4-phenyl-1-butene is hydrogenated, the equilibrium appears to be fully shifted to the right. For comparison, the same reaction without zeolite in an open system with evaporation affords no significant improvement (Fig. 2b). With linear aldehydes as donors (propanal, butanal, pentanal), the yield decreases in the presence of acid zeolites (Supplementary table S13), likely due to the formation of aldehyde self-condensation products, which appear to poison the HCHO-transfer catalyst (Supplementary figs. S18-S19). With branched aldehydes, minimal to even no condensation products were observed. Moreover, these aldehydes provide a convenient tandem reaction pathway by shape-selectively reacting with their olefin co-product over an acid zeolite catalyst in a Prins reaction42–44.
Mapping the zeolite reaction networks
In the HCN-transfer reaction with H-ZSM-5, the zeolite transforms the olefin co-products derived from medium-sized HCN-donors (C4-C6) to a plethora of higher olefins, presumably via dimerization and trimerization reactions (propene reacts to 2-methyl-2-pentene and other hexenes; isobutene reacts to diisobutylene isomers etc; Fig. 3 and Supplementary table S12). The common thread for this group of donors is the formation of mono-, di- or tri-substituted alkene co-products, which are highly reactive when exposed to acidic zeolites45–50. Control reactions with the neat olefin co-products unequivocally confirmed the oligomerization activity of the acid zeolite (Supplementary part 7). In the presence of AlMe2Cl, the isobutene and propene conversions are noticeably higher, suggesting that the Lewis acid might generate additional acidity in the zeolite (Supplementary figs. S24-S25)51–53. With propionitrile as a donor, highly selective dimerization of ethylene was observed (trans-2-butene > 1-butene > cis-2-butene, Fig. 3). Moreover, hydrocyanated butenes, i.e. 2-methylbutanenitrile and valeronitrile, appeared in significant quantities. In contrast to substituted alkenes, ethylene is not dimerized over an acid zeolite at mild reaction temperatures and low ethylene concentrations (Supplementary figs. S26 and S28)40,41. This implies that ethylene must be converted inside the zeolite by a different mechanism than the other olefins. Control experiments showed that with the sole addition of Pd(OAc)2 and H-ZSM-5 to the reaction mixture, butenes were formed from ethylene at 100°C in high selectivities (C4 > 96%) (Supplementary figs. S26-S29). This firmly indicates that the Pd(II)-loaded zeolite is the active ethylene dimerization catalyst.
In line with the HCN-transfer reaction with isovaleronitrile and H-ZSM-5, small amounts of diisobutylene and related isomers could be detected after the HCHO-transfer reaction with 3-methylbutanal as donor and H-FER as zeolite. However, larger amounts of additional products were observed, such as unsaturated C9H16 hydrocarbons (e.g. 2,6-dimethyl-2,4-heptadiene) and a ketone (2,6-dimethyl-4-heptanone) (Fig. 3 and Supplementary table S17). These products are formed via the Prins reaction between 3-methylbutanal and isobutene, initially leading to an α,β-unsaturated alcohol intermediate which can undergo dehydration to a diene, or isomerization to the enol followed by keto-enol tautomerization, resulting in a ketone43. Moreover, a significant fraction of the isobutene-derived products appears to remain inside the zeolite pores, probably due to adsorption and diffusion constraints, or in the form of coke (Supplementary fig. S34).
Fine-tuning the zeolite catalyst for different donor/acceptor combinations
The selection and design of the zeolites as equilibrium-shifting agents can be rationalized by maximizing their reactivity with the olefin co-products (Fig. 3). For example, decorating H-ZSM-5 with Ni centers leads to a more active propene dimerization catalyst, resulting in a larger equilibrium-shifting effect in the HCN-transfer reaction (Supplementary figs. S4 and S25). When using H-ZSM-48 (*MRE, Si/Al = 100), which exhibits a 1D pore channel structure, the formation of large, bulky olefins is possibly retarded54–57. As a result, H-ZSM-48 displays a larger equilibrium-shifting effect with nitrile donors that produce branched olefins, which are prone to formation of bulky oligomers (isocapronitrile, isovaleronitrile), compared to the H-ZSM-5 zeolite with its more spacious 3D pore channel structure (Supplementary fig. S5 and table S10). A similar effect may explain why H-FER outperforms other 10 MR zeolites in the HCHO-transfer reaction, as the ferrierite channels are more constrained, limiting e.g. 3-methylbutanal self-condensation. However, when the HCHO-transfer is conducted with styrenes as substrates at higher temperatures, H-ZSM-5 performs better than H-FER (Supplementary table S16). This could be the result of a kinetic effect, whereby the faster isobutene conversion in H-ZSM-5 can better keep up with the faster styrene conversion (compared to the slower 4-phenyl-1-butene reaction), thus maximizing the synergistic effect of the tandem reaction.
In-depth study of the propionitrile/H-ZSM-5 system
The interplay between the homogeneous catalyst and the zeolite was investigated using the HCN-transfer reaction with propionitrile and H-ZSM-5 as model system, as it gave the most impressive gains in yield (up to 80%). To disentangle the kinetic and thermodynamic reaction characteristics, the kinetics and the effect of increased zeolite loadings were monitored separately. The time profile displays a transient appearance of butenes followed by C5 nitriles, while the hydrocyanated styrene products are formed from the onset of the reaction in quasi first-order kinetics (Fig. 4a,b). As expected, the zeolite has little influence on the kinetics, but clearly shifts the plateau of the styrene nitrile products to a higher level (Supplementary figs. S13 and S30). This equilibrium-shifting effect can be related to a change of the molar reaction free energy, due to the coupling of the transfer hydrocyanation reaction with the ensuing dimerization. This ‘D(DrG°)’ can be calculated from the reaction data (Supplementary part 8.7.1). Theoretical DFT calculations validate the empirical values (Supplementary part 8.7.2), which are sufficiently large (D(DrG°) ⁓ -30 kJ/mol) to justify the experimental increases in yield. Although the zeolite’s beneficial effect is grounded in thermodynamics, the kinetics of the coupled reaction must still be favorable for fully valorizing the equilibrium-shifting effect. With propionitrile, the reaction is initially moderately fast, but thanks to the continuous conversion of ethylene by the zeolite, the formation of the hydrocyanated product is continued at the same rate, well beyond the styrene conversion level that would be the thermodynamic limit in the absence of the zeolite (compare the dark blue and dotted grey curves in Fig. 4a). With the other, more reactive nitrile donors, the conversion quickly reaches the normal equilibrium level, which in the presence of the zeolite, gradually progresses up to higher levels (Supplementary figs. S11-S17).
A molecular picture of the active sites working in a concerted fashion was revealed via in situ and ex situ X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) measurements (Fig. 4c,e,f). Ex situ XANES and EXAFS spectra of zeolites preloaded with Pd(OAc)2 in toluene are consistent with a mononuclear square planar Pd(II)-zeolite species coordinated with 4 oxygen atoms; 2 from acetate and 2 from the zeolite (Fig. 4e and Supplementary figs. S36-S37 and table S25). Even when an equimolar amount of the reducing phosphine ligand Xphos is added, cationic Pd(II)-zeolite species are still formed inside the zeolite (Supplementary fig. S38). Zeolites recovered after an HCN-transfer reaction or an ethylene dimerization reaction show strikingly similar spectra, which can be assigned to subnanometric Pd(0)-carbide species (Supplementary fig. S40 and table S26)58. The EXAFS spectra of in situ XAS measurements of the HCN-transfer reaction are dominated by Pd-P contributions (Fig. 4f), originating from the catalyst complex Pd(0)(Xphos)2 (i.e. the resting state), along with smaller contributions of Pd-Pd interactions (Supplementary fig. S41 and table S27). When H-ZSM-5 is added to the reaction, Pd-O species can be detected before, after and during the reaction, highlighting the presence of zeolite-bound Pd(II) in the ethylene dimerization catalyst (Supplementary fig. S42 and table S28). Moreover, the contribution of Pd-O species becomes more dominant when the Pd:Xphos ratio is increased to 2:1 (as indicated by the arrow in Fig. 4f), since more Pd centers get to migrate inside the zeolite due to the lower amount of Xphos in the solution. ICP-data confirm this, showing a three-fold increase in the zeolite Pd wt.% at a Pd:Xphos ratio of 2:1 (Supplementary table S24).
Finally, modifications of the zeolite’s structure and acid characterics were studied with Powder X-ray diffraction (PXRD) and 27Al MAS NMR. No loss of crystallinity could be observed via PXRD, confirming that the zeolite structure stays intact (Supplementary fig. S33). The 27Al MAS NMR spectrum of parent H-ZSM-5 shows a single peak at 55 ppm corresponding to tetrahedrally coordinated framework aluminium (Fig. 4d)59. After contact with AlMe2Cl at room temperature or after a transfer hydrocyanation reaction, new signals appear at 5 and 35 ppm, with the former being dominant. Notably, no signature shifts of Al-CH3 or Al-Cl bonds could be observed. The signal at 5 ppm corresponds with extraframework aluminium that is octahedrally coordinated, while the one at 35 ppm can be assigned to pentahedral aluminium species53. Thus, AlMe2Cl appears to be hydrolysed inside the zeolite, which further corroborates the need for high Si/Al zeolites and slightly higher Lewis acid loadings. Residual silanol groups (Supplementary fig. S46) and water molecules associated with the zeolite’s acid sites may facilitate the hydrolysis process53.
Substrate scope, upscaling and recycling
To assess the shape-selective nature of our concept, alkene substrates with varying kinetic diameters were employed (Fig. 5a and Supplementary fig. S23). With 1-octene as substrate, the zeolite-assisted HCN-transfer reaction showed high conversion of 1-octene, but low selectivity to the nitrile products, indicating octene diffusion inside the zeolite. Selectivity towards nitriles increased with the kinetic diameter of the corresponding alkene, thus firmly asserting the shape-selective nature of the system. With propionitrile as simple and cost-efficient HCN-donor, the scope of the zeolite-assisted transfer hydrocyanation was explored (Fig. 5b). Various alkyl- and aryl-substituted styrenes (1–8) consistently show high yields (> 85%), with remarkable increases of ~ 70% compared to the reactions without zeolite.
The largest zeolite-induced increases in yield were observed for α-methylstyrene (9) (13 → 93%) and trans-β-methylstyrene (10) (8 → 80%). The system is compatible with fluoro-, alkoxy- and silyl-substituents (12–16). Chlorostyrenes were initially problematic, presumably due to oxidative addition of the C-Cl bond by Pd(OAc)2/Xphos60. However, combining the Cl-tolerant Pd2dba3/CyJohnPhos61 with 0.3 wt.% of Pd-ZSM-5 zeolite furnished the nitrile products of 4-chlorostyrene (17) and 2-chlorostyrene (18) in good yields. Aliphatic cyclic and branched alkenes with C ≥ 8 resulted in good yields (19–22). Allylarenes were also effective, albeit with moderate linear regioselectivities (23–24). A fascinating example is 5-vinyl-2-norbornene (25), bearing both an activated double bond and an unactivated one (Fig. 5c). The reaction without zeolite in propionitrile solely functionalizes the activated double bond, enabled by the exergonic strain release in norbornene. However, with the zeolite, a significant fraction of dinitriles (31%) are also obtained, indicating sequential functionalization of the unactivated bond. Thus, for reactions of simple, non-strained alkenes, the Pd-zeolite-catalyzed ethylene dimerization balances the (energetic) scales, allowing efficient transfer hydrocyanation. To further illustrate the large-scale practicality of our concept, reactions were performed on a 7.5 mmol scale. The nitrile products of 9-vinylanthracene and 1,1-diphenylethylene, which are precursors of pharmaceuticals such as maprotiline (an antidepressant) and lercanidipine (an antihypertensive)62, could be obtained in good-to-excellent yield on a gram-scale (Fig. 5d). For simple styrenes, the HCN-transfer reaction with propionitrile works very well at temperatures as low as 80°C with relatively low catalyst loadings (Fig. 5d). Such a closed system setup, with practically no gas evolution or reflux, would be very attractive in a pharmaceutical setting.
Next, the zeolite-assisted transfer hydroformylation with 3-methylbutanal as HCHO-donor was applied to a range of olefins (Fig. 6a). Using the optimized system 4 mol% Rh/Xantphos, 20 mol% 4-methylsalicyclic acid, 3 eq. AcOH with 100 mg H-FER zeolite (‘condition A’), several terminal, unactivated alkenes containing free hydroxyl groups, ethers, esters, internal double bonds and an imidazolidine-derived scaffold could be converted to the corresponding aldehydes in good to excellent yields (57–87%) and regioselectivities (l:b > 94:6) (28–35). Additionally, electronically diverse olefins such as vinyl boronates and vinyl silanes are susceptible to the reaction (36,37). The effect of the zeolite on the yield fluctuates between 17–39%, resulting in net yields up to 87%, which makes our approach a viable alternative to traditional hydroformylation methods. Using a modified catalytic system consisting of 5 mol% Rh/N-Xantphos, 20 mol% benzo[b]furan-2-carboxylic acid and 100 mg H-ZSM-5 (Si/Al = 140) (‘condition B’), styrenes could be converted into the corresponding linear aldehydes in yields up to 94%. A diverse range of functional groups such as ethers, unprotected alcohols and carboxylic acids, chlorides, bromides, cyanides and trifluoromethyls, including a challenging α-CF3-styrene, were well tolerated (38–49). Furthermore, the flexibility of our method is demonstrated by a one-pot transfer hydroformylation-cyclization-dehydration sequence, converting Ts-protected allyl amine into Ts-protected 2-pyrroline (50) and sclareol into a hydroformylated manoyl oxide derivative (51) (Fig. 6b). Such schemes may incentivize the development of more complex synthetic pathways, building on the synergy between homogeneous metal complexes and solid acids as combined equilibrium-shifting agents and dehydration catalysts. Finally, the zeolite catalyst can be recycled via calcination, as the pores contain mostly organic material, justifying the use of higher zeolite amounts per reaction (Fig. 6c). From a process standpoint, this could open up possibilities for the use of packed bed reactors, enabling facile recovery of the stationary phase after reaction.