Reductive samarium (electro)catalysis enabled by Sm^III-alkoxide protonolysis

Abstract

Samarium diiodide (Sml₂) is a privileged, single-electron reductant deployed in diverse synthetic settings. However, generalizable methods for catalytic turnover remain elusive because of the well-known challenge associated with cleaving strong Sm^III–O bonds. Prior efforts have focused on the use of highly reactive oxophiles to enable catalyst turnover. However, such approaches give rise to complex catalyst speciation and intrinsically limit the synthetic scope. Herein, we leveraged a mild and selective protonolysis strategy to achieve samarium-catalyzed, intermolecular reductive cross-coupling of ketones and acrylates with broad scope. The modularity of our approach allows rational control of selectivity based on solvent, pK_a (where K_a is the acid dissociation constant), and the samarium coordination sphere and provides a basis for future developments in catalytic and electrocatalytic lanthanide chemistry.

Sm^II species are very versatile, single-electron reductants. Since its introduction to synthesis by Kagan and colleagues in 1977 (1), SmI₂ has become a privileged reagent (2). The Ln coordination sphere is highly sensitive to Lewis basic additives, which modulate both the Sm^III/II reduction potential and the steric profile of the reagent, enabling fine control of reactivity and stereoselectivity (3–5). This tunability is invaluable in natural product synthesis, in which stoichiometric Sm^II has been used to effect a variety of reductive transformations of carbonyl functional groups (see Fig. 1A for a representative example) (6–12). In contrast to alternative strong reductants, the compatibility of Sm^II species with Brønsted acids enables proton-coupled reduction reactions, including the conversion of N₂ to fixed-N products (Fig. 1A) (13–16).

Fig. 1. Motivations and challenges associated with reductive Sm catalysis.

(A) Utility of stoichiometric Sm^II reductants in diverse applications. (B) Targeted Sm-catalyzed cycle for ketyl-olefin coupling. (C) Representative Sm catalysis precedents. (D) Inverse relationship between Sm^III-ligand affinity and reduction potential. (E) Sm-catalyzed reductive cross-coupling of ketones and acrylates under mild chemical and electrochemical conditions as described herein.

Despite the value and versatility of Sm^II reductants, they are predominantly deployed (super)stoichiometrically. Additionally, SmI₂ typically must be used under dilute reaction conditions because the solubility of SmI₂ is <0.1 M in tetrahydrofuran (17). Therefore, SmI₂ is not desirable for use as a reagent in large-scale settings or in the early stages of multistep synthesis. However, these limitations could be overcome by the development of a robust and generalizable strategy to use Sm^II in catalytic quantities.

The reactivity of Sm^II is typically driven by the high oxophilicity or azaphilicity that is characteristic of the f elements (18). For example, although electron transfer from SmX₂ to ketone substrates is disfavored based on the comparison of outer sphere reduction potentials (Fig. 1B, step i; X = halide, ΔE° > 1 V for X = I), the strong coulombic interaction between Sm^III and the resulting ketyl radical anion drives such reactions forward (19). However, this stabilizing interaction presents the primary barrier to catalytic turnover (20). The cathodic reduction potentials of Sm^III(OR)_n species are prohibitively negative for desirable catalysis (21–23). Exchange of OR⁻ with X⁻ to generate more readily reduced SmX₃ species is an attractive approach for turnover (Fig. 1B, steps iii and iv), but mild, selective, and tunable methods for the cleavage of Sm–O bonds remain elusive.

The few reports that have attempted to address the challenge of reductive Sm catalysis used halosilanes (R₃SiX) as oxophiles to cleave alkoxides from Sm^III (Fig. 1C) (24–27). Such methods have not been widely adopted, possibly because the reagents required for turnover have limited substrate compatibility. For instance, whereas relatively mild chlorosilane reagents are capable of cleaving alkoxides from Sm^III, as exemplified by a pinacol-coupling reaction reported by Greeves and co-workers (Fig. 1C), chloride rapidly displaces iodide from the Sm coordination sphere (Fig. 1D) (26, 28). SmCl₃ is more difficult to reduce than SmI₃ and therefore requires a strong reductant such as Mg⁰. Corey and Zheng avoided the problem of halide scrambling in their Sm-catalyzed cross-coupling of ketones and acrylates by using Me₃SiOTf as an oxophile in combination with LiI (Fig. 1C); however, this required manual slow addition of Me₃SiOTf to mitigate parasitic consumption of the acrylate coupling partner (24). As a final point, halosilanes are not compatible with the protic additives ubiquitous in Sm^II chemistry (Fig. 1A).

Considering the challenge of Sm^III–OR turnover, we recognized that protonation would be a tunable approach to Sm-alkoxide cleavage. Here, we demonstrate rapid and reversible protonolysis of alkoxide ligands from Sm^III through judicious pairings of cationic Brønsted acids and halide donors. This transformation was leveraged to achieve Sm-catalyzed reductive cross-coupling of ketones and acrylates using Zn⁰ as a relatively mild source of reducing equivalents at the SmI₃/SmI₂ redox couple (Fig. 1E). Sm(OTf)₃ serves as a shelf-stable, commercially available Sm precursor, and the reactions can be conducted on a gram scale at 10-fold higher concentrations than is typically used when stoichiometric SmI₂ is used (fig. S3). The optimized conditions translate into a bona fide electrocatalytic system, distinct from prior systems in which electrochemically driven Sm^III/II turnover has been difficult to firmly establish (29–32). Finally, we provide a thermochemical analysis of the factors controlling the alkoxide protonolysis step as a basis for future developments in catalytic and electrocatalytic Sm chemistry.

We began our studies by using Sm(OⁱPr)₃ as a model of the Sm^III-alkoxide species generated under reductive coupling conditions with the goal of identifying a suitable proton donor and iodide source to generate redox-active SmI₃ and enable catalysis (Fig. 2A). We anticipated that successful conditions would meet the following requirements for alkoxide-iodide exchange at Sm^III: (i) the conjugate base of the acid should not outcompete coordination of I⁻ to Sm^III; (ii) the counterion should be chemically compatible with SmI₃/SmI₂ redox cycling (22, 33); and (iii) the pK_a (where K_a is the acid dissociation constant) of the acid in MeCN should be <19, guided by the benchmarked pK_a value of 19.9 for a cationic [Sm^III]⁺ complex with MeOH in MeCN (34).

Fig. 2. Protonolysis turnover strategy and reaction development.

(A) Proposed conversion of Sm-alkoxides to Sml₃. (B) CVs of 2 mM Sm(OⁱPr)₃ (black dashed trace) after the successive addition of 3 equiv each of Lil and LutHTFSI (magenta trace) and 3 equiv of LiOTf (green trace) overlaid with the CV of 2 mM Sm(OTf)₃ after the addition of 3 equiv of Lil (dashed blue trace) at 100 mV s⁻¹ on a glassy carbon working electrode in THF containing 0.1 M BMPipTFSI (where BMPip is 1-butyl-l-methylpiperidinium). All potentials are referenced to Fc^+/0 = ferrocenium/ferrocene. (C) Reaction optimization and control experiments conducted at 0.05 mmol scale. Listed concentrations correspond to the ketone substrate 1a. Yields for entries 1 to 12 were determined by ¹H nuclear magnetic resonance (¹H NMR) spectral integration using 1,3,5-trimethoxybenzene as an internal standard; entries 13 and 14 are isolated yields.

Thus, we investigated a panel of acids (baseH⁺) and iodide sources and identified lutidinium bis(trifluoromethylsulfonyl)imide(LutHTFSI) as meeting these criteria, as demonstrated by cyclic voltammetry (CV). The strongly donating alkoxide ligands of Sm(OⁱPr)₃ render it redox inactive in the tetrahydrofuran (THF) solvent window (fig. S33). However, after the addition of LutHTFSI (3.0 equiv; pK_a = 14.2 in MeCN) (35) and LiI (3 equiv) to Sm(OⁱPr)₃ in THF, a quasireversible wave centered at −1.44 V versus Fc^+/0 appeared in file CV (Fig. 2B, magenta trace), suggestive of SmI₃ generation. To verify this assignment, SmI₃ was generated through ion exchange between Sm(OTf)₃ and LiI under identical electrochemical conditions. This mixture also featured a quasireversible reduction centered at −1.47 V (Fig. 2B, dashed blue trace). We attribute the small offset in potential to OTf⁻ association (Fig. 2B, teal trace). A parallel spectrophotometric experiment confirmed the generation of SmI₃ from Sm(OⁱPr)₃ through protonolysis-iodide substitution (fig. S8).

After demonstration of this Sm-alkoxide cleavage step, we explored the reductive coupling between 1,4-cyclohexanedione monoethylene acetal (1a) and acrylates (R = ^tBu, CH₂CF₃, Ph) to give spirocyclic γ-lactone 3a (Fig. 2C) (36). Sm(OTf)₃ was used as an inexpensive, commercially available, air-stable precatalyst. Although the CV studies used LiI as the iodide source, MgI₂ was found to be necessary for the synthetic transformation. When LiI or ⁿBu₄NI was used, the reactions did not change to the purple color indicative of SmI₂ in 2-methyltetrahydrofuran (2-MeTHF) (Fig. 2C, entry 2). Zn⁰ powder was selected as a mild terminal reductant. After an initial evaluation of acrylates (R = ^tBu, CH₂CF₃, Ph; fig. S2, entries 3 and 4), phenyl acrylate was found to perform best, furnishing γ-lactone 3a in quantitative yield under the optimal conditions [10 mol % Sm(OTf)₃, 3.0 equiv MgI₂, 1.1 equiv LutHTFSI, 3.0 equiv Zn⁰ in 2-MeTHF (0.05 M) at 18°C]. No product was observed in the absence of Sm(OTf)₃ (Fig. 2C, entry 6), but lowering the MgI₂ loading decreased the yield slightly (Fig. 2C, entry 7). When Gd(OTf)₃ was used as a redox-inactive Lewis acid substitute for Sm(OTf)₃, no product was formed, supporting Sm^III/II redox activity in catalysis (Fig. 2C, entry 8). Substituting Sm(OTf)₃ with Mg(OTf)₂ also did not furnish any product, ruling out the role of triflate in product formation (Fig. 2C, entry 9). Zn⁰ was required for product formation (Fig. 2C, entry 10), whereas omission of LutHTFSI resulted in low yield (Fig. 2C, entry 11). Finally, 2-MeTHF was superior to THF as a solvent (Fig. 2C, entry 12).

A practical advantage of the ability to use catalytic Sm for reductive transformations is that the reactions can be performed at higher substrate concentrations. Because of the poor solubility of SmI₂ in THF, solutions of this reagent are usually prepared at concentrations of 0.1 M or lower. Indeed, the average concentration of SmI₂-mediated reactions reported in the literature is 0.02 M with respect to ketone (fig. S3). Under these catalytic conditions, comparable yields of product 3a can be formed at a 10-fold higher concentration (0.20 M; Fig. 2C, entries 13 and 14), which to the best of our knowledge is the highest concentration reported for a reductive Sm transformation.

Substrate scope and demonstration of synthetic utility

The scope of the above reaction is consistent with that of prior investigations (36) using stoichiometric SmI₂. A variety of aliphatic and aromatic ketones performed well, giving the γ-lactone products in good to excellent yields (Fig. 3A). Common functional groups, such as silyl ethers (3h), esters (3j), aryl halides (3n to 3r), sulfonates (3w), and boronate esters (3x), were compatible under the reaction conditions. Aryl ketones bearing strong electron-withdrawing substituents (3aa) resulted in lower yield due to competitive pinacol coupling. This effect was even more pronounced with ^tBu-acrylate, which further slows the rate of Giese addition relative to 2 (fig. S10). A cyclohexanone substrate bearing an α-tethered unactivated olefin exclusively formed the spirocyclic γ-lactone (3i, 3:1 dr) without any evidence of 5-exo-trig cyclization. The pharmaceutically relevant heterocyclic building blocks tetrahydrothiopyran (3k) and tetrahydropyran (3l) were produced in synthetically useful yields.

Fig. 3. Scope of Sm-catalyzed reductive cross-coupling.

(A) Substrate scope of Sm-catalyzed reactions. Reactions were conducted on a 0.3 mmol scale. Isolated yields are reported unless otherwise specified. Yields in parentheses were determined by ¹H NMR analysis using 1,3,5-trimethoxybenzene as an internal standard due to the volatility of the product or instability to silica gel. *3.3 equiv LutHTFSI was used. (B) Intramolecular Sm catalysis with E- and Z-1ab.

In their recent total synthesis of (+)-euphorikanin A, Carreira and co-workers demonstrated that the diastereoselectivity of an intramolecular SmI₂-mediated lactonization is dictated by the E/Z geometry of the acrylate (6). To determine whether this is also true under the catalytic conditions, an analogous pair of intramolecular reductive lactonizations were performed with E- and Z-1ab (Fig. 3B). Using stoichiometric conditions otherwise identical to those of Carreria et al., the cis product was favored using the Z-olefin, whereas the trans product was favored with the E-olefin. The inversion in diastereoselectivity was observed using the catalytic system, albeit with slightly diminished dr. The slight erosion in dr might result from competing Mg²⁺ ion coordination to the acrylate.

Demonstration of electrocatalysis

Although Zn⁰ is well suited to SmI₃/SmI₂ turnover, it is not suitable for generating Sm^II species with substantially more negative reduction potentials (37). Electrochemical methods in which the applied potential can be matched to the Sm^III/II reduction potential are thus appealing. However, Sm-mediated electrocatalysis is poorly developed. These reactions can suffer from competing reactivity mediated by the oxophile or metal cations generated at the sacrificial anodes; in some cases, the use of an Sm metal electrode was reported to be necessary (27, 29–32). We sought to address these challenges by developing well-defined electrocatalysis using the Sm-alkoxide protonolysis strategy discussed above.

The CV of SmI₃ generated by combining Sm (OTf)₃ and MgI₂ in 2-MeTHF features a quasi-reversible wave centered at −1.55 V (Fig. 4A, black trace). The CV of SmI₃ with ketone 1a, acrylate 2, and MgI₂ (Fig. 4A, magenta trace) exhibits an irreversible wave that is double the current intensity of the 1e⁻ reduction of SmI₃. This response, which is also observed with the aromatic ketone substrate 1m (Fig. 4B, magenta trace), is consistent with net Sm-mediated 2e⁻ reductive coupling of the ketone and acrylate to yield a γ-alkoxy-enolate species (Fig. 4C, iii or iv).

Fig. 4. Development of Sm electrocatalysis.

All CVs were collected on a glassy carbon working electrode in 2-MeTHF containing 0.2 M BMPyTFSI (where BMPy is 1-butyl-1-methylpyrrolidinium) at 25 mV s⁻¹. (A) CVs of 2 mM Sm(OTf)₃ and Mgl₂ (25 equiv, black trace) after the addition of substrates 1a (10 equiv, solid light blue trace), 2 (20 equiv, dashed light blue trace), their combination (magenta), and the acid LutHTFSI (20 equiv, green trace) overlaid with the CV of the substrates, acid, and Mgl₂ in the absence of Sm (light green trace). (B) CVs of 2 mM Sm(OTf)₃ and Mgl₂ (25 equiv, black trace) after the addition of substrates 1m (10 equiv, light blue trace), both 1m and 2 (10 and 20 equiv, magenta trace), and the acid LutHTFSI (20 equiv, green trace) overlaid with the CV of the substrates, acid, and Mgl₂ in the absence of Sm (light green trace). (C) Plausible mechanistic pathways accounting for CV responses. (D) CVs of 2 mM Sm(OTf)₃ and ⁿBu₄NI (50 equiv, black trace) after the addition of the substrates 1a and 2 (10 and 40 equiv, respectively, blue trace), followed by titration of Mg(TFSI)₂ (light blue-magenta traces). (E) CPE conditions (0.1 mmol scale). Yields were determined by ¹H NMR analysis using 1,3,5-trimethoxybenzene as an internal standard.

Further addition of LutHTFSI in the presence of both ketone and acrylate gave rise to S-shaped multielectron waves at the potential of SmI₃ reduction (Fig. 4, A and B, green traces), indicative of electrocatalytic turnover. Control experiments confirmed that none of the individual reaction components (see the supplementary materials), nor their combination in the absence of Sm (Fig. 4, A and B, light green traces), were responsible for the current at −1.5 V.

We also investigated the use of ⁿBu₄NI in place of MgI₂ to generate SmI₃ (Fig. 4D, black trace). In this case, addition of 1a and 2 resulted in an irreversible wave with less enhancement in current relative to when MgI₂ was used (Fig. 4D, light blue trace, versus Fig. 4A, magenta trace). The full 2e⁻ current was regained on titration of Mg(TFSI)₂ (Fig. 4D, light blue-magenta traces), suggesting that the second electron transfer to the presumed radical intermediate (ii) at the electrode is facilitated by Mg²⁺ (38). In most stoichiometric SmI₂ reductions, every electron transferred to substrate also generates an equivalent of Lewis acidic Sm^III; with low concentrations of Sm, however, Mg²⁺ may alternatively stabilize alkoxide intermediates.

Substrate coupling could be initiated either by ketone reduction or acrylate reduction (Fig. 4C) (39). A “ketone-first” mechanism is likely operative with aromatic ketones such as 1m. In the CV of SmI₃ and 1m alone, the Sm^II reoxidation feature completely disappeared (Fig. 4B, blue trace), indicating that irreversible reduction of the aromatic ketone by SmI₂ is rapid under electrochemical conditions. By contrast, the SmI₃/SmI₂ wave remained reversible in the presence of either (but not both) the aliphatic ketone 1a or acrylate 2 (Fig. 4A, solid and dashed light blue traces, respectively). These data indicate that the initial electron transfer step to form i or v is slow and/or uphill with these substrates (40), as is typical for the reduction of unactivated carbonyl substrates by SmI₂ (2, 41). However, reduced and homocoupled products of both aliphatic ketones and acrylate 2 were observed when each substrate was subjected to the standard Zn⁰-driven catalytic conditions in the absence of the respective cross-coupling partner, suggesting that SmI₂ is competent for reduction of both substrates (see section 6.6 of the supplementary materials). Indeed, CVs of SmI₃ lost reversibility (figs. S27 and S31) at increased concentrations of 1a and 2. Although irreversible consumption of SmI₂ was more rapid with 2 than with 1a in this regime, electroanalytical studies suggest that the observed kinetics with 2 are an aggregate of electron transfer and homocoupling rates (see section 8.2 of the supplementary materials). Without direct access to the relative rates of initial aliphatic ketone versus acrylate reduction, both “acrylate-first” and “ketone-first” mechanisms must be considered viable.

Having gained an understanding of the reduction events through electroanalytical studies, we investigated electrocatalytic formation of lactone 3a. We used constant potential electrolysis (CPE) to avoid the electrode-mediated hydrogen evolution reaction (HER) with LutHTFSI (HER onsets at ~−1.7 V under these conditions; fig. S21). Oxidation of Hantzsch ester (HEH₂) was selected as a well-behaved counter-reaction. CPE of ketone 1a and acrylate 2 with Sm(OTf)₃, LutHTFSI, and MgI₂ at an applied potential of −1.55 V (carbon cloth cathode; two-compartment cell) furnished the cross-coupled lactone 3a in 75% yield at 75% Faradaic efficiency (Fig. 4E). HE was produced quantitatively. Under the same conditions, phenyl-substituted lactone 3m was prepared in 85% yield. With this more activated substrate, current attributable to ketone reduction was observed in the absence of Sm(OTf)₃ at −1.65 V; however, under these conditions, only the pinacol product 4m was formed. This finding highlights the role of Sm in favoring lactone formation over possible competing processes.

Thermochemistry and outlook

We also investigated the factors influencing the key proton-transfer step in Sm^III–OR reactivation. The equilibrium of the Sm^III alkoxide protonolysis and ligand substitution can be decomposed into a thermochemical cycle of five components (Fig. 5A). Net protonolysis is favored by (i) a weaker affinity of the alkoxide for Sm^III, (ii) a higher pK_a of the corresponding alcohol, (iii) a stronger Brønsted acid (baseH⁺), (iv) a relatively weak affinity of the halide for its corresponding countercation M⁺, and (v) a stronger affinity of the halide for Sm^III. The last three components are readily decoupled through independent variation of the acid, the halide donor, and the identity of the halide, enabling rational control of the net alkoxide cleavage step.

Fig. 5. Factors controlling Sm^III-OR protonolysis.

All CVs were collected on a glassy carbon working electrode with 0.1 M BMPipTFSI as the supporting electrolyte. (A) Thermochemical cycle describing Sm^III-OR protonolysis. (B) CVs demonstrating reversibility of Sm(OⁱPr)₃ (2 mM) protonolysis and iodide substitution with CoIHTFSI and Lil at 100 mV s⁻¹ in THF. (C) CVs demonstrating the sensitivity of net protonolysis and iodide substitution to the acid pK_a and availability of iodide at 25 mV s⁻¹ in 2-MeTHF. (D) Product distribution of Sm-catalyzed reductive cross-coupling of acetophenone and ^tBu-acrylate as a function of dielectric strength and acid pK_a. Diameters of circles correlate to yield of product (see figs. S40 to S44 for values). (E) CVs demonstrating the sensitivity of net protonolysis and ligand substitution to the identity of the substituting ligand at 25 mV s⁻¹ in THF.

Consistent with Le Chatelier’s principle (42), the amount of redox-active SmI₃ after the reaction of Sm(OⁱPr)₃ with LiI and ColHTFSI decreased with the addition of LiTFSI and collidine (as reflected in the CVs in Fig. 5B). The initial current intensity was restored by the addition of ColHTFSI and ⁿBu₄NI. The influence of the pK_a of the acid (baseH⁺) was demonstrated by using ⁿBu₄NI as the iodide source and collecting CV data with a panel of acids spanning pK_a values of ~10 to 19. The redox activity of the system, which presumably reflects the position of the equilibrium between Sm(OⁱPr)₃ and SmI₃, decreased as the pK_a of the baseH⁺ increased (Fig. 5C). Addition of Mg cation [e.g., Mg(TFSI)₂], which has a stronger affinity for I⁻ than does ⁿBu₄N⁺, shifted the equilibrium toward Sm(OⁱPr)₃. As a result, stronger acids were required under these conditions to completely restore redox activity (Fig. 5C, compare the green versus red traces).

This relationship points to the potential breadth of the parameter space accessible for optimization of Sm-catalyzed reactions involving different substrates, intermediates, and desired products. As an illustrative example, depending on the acid and solvent used for the Sm-catalyzed coupling between 1m and ^tBu-acrylate, different products were observed (Fig. 5D). The cross-coupled products 3m and 5m were favored with high-pK_a acids in solvents such as THF and 2-MeTHF. A pronounced selectivity for the lactone product 3m over its acyclic counterpart 5m was observed in 2-MeTHF (see the supplementary materials for details). The pinacol and reduction products 4m and 6m were more prevalent with low-pK_a acids, particularly when strongly coordinating solvents such as acetonitrile or dimethoxyethane were used. This difference may be due to early protonolysis of the Sm^III-ketyl intermediate to release the neutral ketyl radical, which might rapidly dimerize or undergo reduction before productive addition to acrylate can occur (28).

The ability to tune the Sm^III/II redox potential by using additives with SmI₂ is an enabling feature of this reagent. For example, the addition of Br⁻ generates the stronger reductant, SmBr₂ (43); however, for redox cycling, an acid must be used that will protonate the Sm^III-alkoxide but will not undergo HER at the required potential for SmBr₃ reduction (−1.9 V). LutHTFSI is incompatible with such a strongly reducing potential. However, as the affinity of the incoming ligand for Sm^III increases, alkoxide cleavage becomes possible with a higher-pK_a acid. The weaker acid triethylammonium (Et₃NH⁺, TFSI counteranion) met the needed criteria to enable redox cycling of SmBr₃, giving rise to protonolysis of Sm(OⁱPr)₃ in combination with LiBr to generate SmBr₃ at a potential positive of the acid’s HER background (Fig. 5E, magenta trace). Similarly, the addition of the Lewis basic donor N-methylpyrrolidinone resulted in a cathodic shift to the Sm^III/II couple and enhanced alkoxide protonolysis with the intermediate acid BnMe₂NHTFSI (Fig. 5E, yellow trace).

These results lay the groundwork for a more generalized approach to reductive Sm catalysis and electrocatalysis under different redox regimes. Catalyst design based on the incorporation of supporting ligands is of high interest, particularly with respect to developments in asymmetric Sm catalysis. Whereas the ligand environment influences the Sm^III/II reduction potential, the pK_a, of the acid can enable rational optimization to favor a desired Sm-catalyzed coupling over competing HER. Successive proton and electron transfer to [Sm^III–OR] species is also ideal for the regeneration of [Sm^III–O(R)H] species, which can serve as potent net hydrogen atom donors. In sum, the straightforward Sm–O protonolysis strategy described herein is anticipated to enable diverse catalytic transformations, including the extension to other rare earth elements as catalysts.

Data and materials availability:

All data are available in the main text or the supplementary materials.