Platinum group metals (PGMs) have played a critical role in the development of modern homogeneous catalysis, shaping the landscape of catalyzed industrial processes, from bulk to commodity chemicals. Despite the unquestionable success of PGM-based catalysts, research on catalysts based on Earth-abundant metals (EAM) has received great interest in recent times. This renaissance of EAM-based catalysis is mainly due to concerns over sustainability, as they seem a better option than their PGM analogues owing to their global availability, low cost, and relatively low toxicity and environmental impact [1]. Particularly, 3d transition metals allow for the preparation of inexpensive and efficient catalysts from readily available sources [2,3,4,5,6,7] and, therefore, may pave the way to the development of more sustainable processes. Moreover, 3d metals do not behave simply as mimics of their 4d and 5d counterparts; on the contrary, they present a prolific reactivity mainly due to their ability to perform one-electron redox processes and straightforward ligand reorganization. To some extent, the development of well-defined 3d metal-based catalysts has been hampered by the rather unpredictable behavior of these complexes; however, the use of appropriate ligands that tame their reactivity may offer the possibility to unravel new reactions and mechanisms.

The large scale required for the implementation of global energy transformations has prompted the scientific community to aim at the development of catalysts based on abundant metals for these types of reactions. In this context, formic acid (FA) has been proposed as a promising hydrogen carrier because of several key properties. Among them, we can highlight its volumetric capacity and energy density (53 g H₂/L, 1.77 kW·h/L, respectively), which are markedly higher than those of compressed hydrogen (23 g H₂/L, 0.767 kW·h/L, respectively, at 35 MPa and 27 °C) [8,9]. In addition, FA presents low toxicity and its dehydrogenation often affords CO-free H₂, which is crucial for fuel cell applications [10,11]. Up to the present time, many efforts have been devoted to the development of homogeneous catalysts for formic acid dehydrogenation (FADH) based on PGM metals; however, their EAM counterparts have been less explored—although remarkable advances have been reported in recent times.

Catalysts for FADH often present participative ligands that permit metal-ligand cooperation [12]. These bifunctional catalysts have been described as containing ligands able to act as a proton shuttle that may assist with key steps of the catalytic cycle; namely, the protonation of a hydride ligand to generate H₂ or the decarboxylation of formate [13,14,15,16,17,18].

2. Transition Metal-Based Catalysts

2.1. Cobalt Homogeneous Catalysts

The first example of a cobalt catalyst for the dehydrogenation of FA was described in 1993 by Onishi [26]. Complex [CoH{PPh(OEt)₂}₄] catalyzes the decomposition of HCOOH into CO₂ and H₂ in THF at 30 °C under Pyrex-filtered photoirradiation with a high-pressure mercury lamp and in the absence of a base, reaching a turnover number of 3 after 6 h. The proposed reaction mechanism entails the protonation of [CoH{PPh(OEt)₂}₄] by HCOOH to give [CoH₂{PPh(OEt)₂}₄][OOCH] as preactivation step. Photoirradiation triggers the release of dihydrogen with concomitant formation of [Co(OOCH){PPh(OEt)₂}₄]. The last step involves the photoirradiated decarboxylation of the formate ligand to regenerate the active species (Scheme 1).

Scheme 1. Catalytic cycle proposed for the dehydrogenation of FA with [CoH{PPh(OEt)₂}₄].

More recently, Beller and coworkers explored the activity of a series of cobalt complexes in the catalytic dehydrogenation of FA [27]. Initially, Co(II) complexes that had been previously reported to show activity in hydrogenation reactions were tested under various conditions, but no activity was detected [28]. The fact that the active species in these processes have been proposed to comprise Co(I) metal centers encouraged the authors to explore the performance of related Co(I) complexes 1–3 (Figure 1). These complexes were isolated and tested as catalysts in a HCOOH/DMOA mixture 10/11 (DMOA = N,N-dimethyloctylamine) at 80 °C. 1 and 2 are barely active; however, 3 showed good activities, allowing total conversion after 90 min (TON = 2260). Complex 3 also proved active in aqueous solutions employing formate as base. It is worth highlighting the extreme sensitivity of this system toward electronic/steric effects, since the exchange of the ⁱPr substituents of the phosphane moieties by Ph groups dramatically improves the performance of the catalyst.

Figure 1. Depiction of complexes 1–3.

The high air-sensitivity of 3 prompted the authors to evaluate the activity of the in situ generated species. 3 was formed by the reduction of the parent Co(II) complex, [Co(Cl)₂(PNP)] (4) (PNP = bis(2-(diphenylphosphanyl)ethyl)amine), with NaBEt₃H in toluene at room temperature. Subsequently, the mixture thus generated was added to the dehydrogenation system previously charged with a water solution of FA and potassium formate at 60 °C. The use of 1 equivalent of NaBEt₃H resulted in lower catalytic activities than those obtained for the isolated complex, probably due to the incomplete reduction of 4; conversely, the addition of 2 equivalents led to slightly better results than those achieved by 3. In this regard, the authors proved that the activity of 3 with 2 equivalents of KOOCH is the same as that obtained for 4 with 3 equivalents of KOOCH, which suggests a common active species in this case.

Optimization of the reaction conditions permitted to lower the amount of KOOCH to 1 equivalent relative to FA. However, the best results were obtained at 80 °C, with a 1:4 HCOOH:KOOCH ratio and 3 as catalyst, leading to a TON value of 7166 after 70 h.

The reaction mechanism was substantiated by DFT calculations using 3 as catalyst. All the species included in the Gibbs free energy profile have triplet ground states, being the singlet states too high in energy and, therefore, discarded. The formation of the active species (I1) takes place via substitution of the chloro ligand in 3 by formate, which coordinates to the metal center by one of the O atoms while the other stablishes a hydrogen bond interaction with the proton of the amine. Subsequently, dissociation of the Co–O bond allows the formation of intermediate I2, thus allowing hydride abstraction and concomitant formation of I3 and a molecule of CO₂. The last step brings about the formation of H₂ and the regeneration of I1 via TSI3-I1, which entails hydride protonation by a molecule of FA stabilized by a hydrogen bond between the carbonyl and N–H moieties (Scheme 2). The direct protonation of the hydride by the N–H moiety or by “free” FA were discarded owing to their high energy barriers.

Scheme 2. Catalytic cycle proposed for the dehydrogenation of FA with 3. The symbol “‡” denotes a transition state.

The latest example of a cobalt complex that catalyzes the dehydrogenation of FA has been reported by Cantat and coworkers [29]. This catalyst, analogously to Beller’s (vide supra), also features an N-H moiety in the vicinity of the metal center; however, in this case no base or additives are required to achieve catalytic activity. Initially, the authors prepared Co(II) complexes 5 and 6 (Figure 2), which were reacted with five equivalents HCOOH at 150 °C, thus leading to quantitative conversion of HCOOH into CO₂ and H₂. In addition, the formation of diamagnetic species 7 (Figure 2) was detected by ¹H and ³¹P{¹H} NMR, although its isolation was not possible by this method. The Co(I) complex 7 was prepared and successfully isolated by reaction of 5 and 6 with 3 equivalents of trimethylsilylformate at 80 °C for 16 h.

Figure 2. Depiction of complexes 5–7.

The catalytic activity of complexes 5–7 (0.5 mol%) was evaluated at 80 °C using a solution of FA in 1,4-dioxane. Complex 5 showed no activity under these conditions, while 6 reached a TON value of 160 after 20 min, and 7 achieved full conversion in 10 min, which amounts to a TON of 200 with a TOF of 67 min⁻¹. A TON value of 458 was achieved after 4 h by means of consecutive cycles under the reaction conditions described above. It is noteworthy that 7 also showed activity in other solvents, such as toluene or water, in the absence of additives or base, although lower TON and TOF values were recorded. Computational studies on this system revealed a reaction pathway, according to which the dangling amine moiety at the PPP ligand plays a major role, allowing outer-sphere interactions with the substrate (HCOOH) by means of hydrogen bonding (Scheme 3). The initial step entails the isomerization of 7 to give intermediate I4, where the hydride and carbonyl ligands exchange positions relative to 7. Subsequently, protonation of the hydride in I4 affords a molecule of H₂ and I5. Subsequently, although β-hydride elimination is a possible reaction pathway, the authors found that hydride abstraction aided by a molecule of FA, via intermediates I6 and I7, was a lower energy pathway. Thus, CO₂ release renders intermediate I4, closing the catalytic cycle.

Scheme 3. Catalytic cycle proposed for the dehydrogenation of FA with 7 in the absence of base.

Remarkably, in contrast with the mechanism described above for 3 and 7, the decarboxylation of formate without the assistance of a cooperative mechanism has been recently described for a cobalt complex [30].

2.2. Iron Homogeneous Catalysts

In 2010 Beller, Ludwig and coworkers explored the activity of a wide range of iron-carbonyl complexes in the presence of various ligands and visible-light irradiation [31]. [Fe(CO)₅], [Fe(CO)₃(COT)], [Fe₂(CO)₉], [Fe(CO)₃(BDA)] and [Fe₃(CO)₁₂], (COT = cyclooctatetraene; BDA = benzylideneacetone) were tested at 40 °C under light irradiation with a 300 W xenon lamp (filtering the UV portion with a hot mirror) in a DMF solution of HCOOH and Et₃N (5:2). [Fe₃(CO)₁₂] with 3 equivalents of PPh₃ and tpy (terpyridine) was found to be the best performing catalyst, closely followed by [Fe₂(CO)₉] and [Fe(CO)₃(COT)] with analogous ligand ratios (Fe/PPh₃/tpy 1:1:1). It is worth noting that when these ligands were used independently, a drastic drop of the catalytic activity was observed.

Experiments performed by successively turning on and off the light source showed that catalytic activity was only observed under irradiation, while dark periods revealed no gas evolution, which led the authors to conclude that the reaction was photoassisted instead of simply photoinduced.

Fine tuning of the ligand system was carried out by testing a broad variety of P- and N-ligands, obtaining the best activities with PPh₃ and 6,6″-substituted terpyridine ligands. Namely, 6,6″-(bromo)-2,2″:6′,2″-terpyridine and 6,6″-(phenyl)-2,2″:6′,2″-terpyridine delivered the highest TOF value (200 h⁻¹). Although less active, phenanthroline ligands led to higher catalyst stability.

Based on NMR, IR and DFT studies the authors proposed [Fe(H)(CO)₃(PPh₃)]^– (I8) as the active species, which forms by irradiation of [Fe(CO)₃(PPh₃)₂] in the presence of FA. The benefit of using N-ligands is unclear since no complex containing these ligands was identified by means of stoichiometric experiments. Nonetheless, the authors claim a stabilizing role after phosphane dissociation, because the loss of a CO ligand was observed in the presence terpyridine, which may be a catalyst deactivation pathway. Once I8 is generated, protonation of its hydride ligand by FA leads to formation of the dihydrogen complex I9. Subsequently, phosphane coordination and H₂ release renders I10. Finally, β-hydride elimination closes the catalytic cycle delivering a molecule of CO₂ (Scheme 4).

Scheme 4. Catalytic cycle proposed for the dehydrogenation of FA with [Fe(CO)₃(PPh₃)₂] in situ generated from the mixture [Fe₃(CO)₁₂]/PPh₃/tpy.

Shortly after, Beller, Ludwig and Laurenczy developed an Fe(II) catalyst based on the tetradentate ligand P(CH₂CH₂PPh₂)₃ (PP₃), which was prepared in situ employing ratios Fe/ligand 1:1 and 1:2, with the latter showing better activities [32]. Remarkably, the reaction works without the need for base or additives, and using an environmentally benign solvent, namely, propylene carbonate. Moreover, high pressures of CO₂ and H₂ (up to 30 and 20 bar, respectively) showed no significant effects on the activity of the catalyst. The presence of air or the traces of water present in the solvent or FA did not affect the performance of the catalyst, but addition of chloride to the system resulted in a sharp reduction of the catalytic activity.

Several Fe-hydride complexes featuring the PP₃ ligand were tested, with [Fe(H)(PP₃)]BF₄ (8) and [Fe(H)(H₂)(PP₃)]BF₄ (9) showing better catalytic activities than the catalyst generated in situ from [Fe(H₂O)₆][BF₄]₂/PP₃ (1:1)—the TON_2h values reported for these systems were 745, 727 and 562, respectively. The activities of the systems [Fe(H)(PP₃)]BF₄/PP₃ (1:1) and [Fe(H₂O)₆][BF₄]₂/PP₃ (1:2) are very similar, with TON_2h values of 1227 and 1279, respectively. DFT and experimental studies on the reaction mechanism led the authors to propose two competing catalytic cycles, A and B in Scheme 5. In the case of cycle A, the first step involves the protonation of the hydride ligand in 8 to afford I11 with concomitant release of H₂. I11, which is in equilibrium with an analogous species featuring a monodentate formate, undergoes β-hydride elimination to yield I12. Finally, CO₂ dissociation closes the cycle. Regarding cycle B, coordination of formate to 8 leads to the formation of I13, which renders I14 upon protonation of the hydride and β-hydride elimination of the formate, thus generating a molecule of CO₂. Finally, H₂ dissociation restarts the cycle.

Scheme 5. Catalytic cycle proposed for the dehydrogenation of FA using complex 8 as catalyst.

Complexes 10 and 11 (Scheme 6) can catalyze the dehydrogenation of FA in the presence of a 50 mol% Et₃N at 80 °C in dioxane, allowing a TOF of 739 h⁻¹ and a TON of 994 after 2.5 h for 11b [33]. The authors showed that the use of a Lewis acid (10 mol%) instead of Et₃N gives rise to excellent catalytic activities; namely, LiBF₄ renders outstanding TOF and TON values (196,728 h⁻¹ and 983,642, respectively)—note that in the absence of additives the system proved barely active. Mechanistic studies suggest that Et₃N is not directly involved in the catalytic cycle, since all the steps of the mechanism may take place in the absence of base. The improved activities in the presence of Et₃N and Lewis acids, such as LiBF₄, hint at a similar role in the reaction mechanism. The authors propose that the decarboxylation of 11, which is the rate liming step according to kinetic experiments, takes place via hydride abstraction assisted by a Lewis acid (instead of β-hydride elimination). Rearrangement of the formate ligand yields an H-bound intermediate, which is stabilized by hydrogen bonding between the carboxylate moiety and the Lewis acid, or, in the case of Et₃N, the Brønsted acid [Et₃NH]⁺ that forms in the presence of excess FA (Scheme 7). This lowers the activation energy of the decarboxylation step, thus improving the reaction rates.

Scheme 6. Catalytic cycle proposed for the dehydrogenation of FA using 10 or 11 as catalysts.

Scheme 7. Decarboxylation step assisted by a Lewis acid or [Et₃NH]⁺.

The overall reaction mechanism entails the formation of the active species, 11, by protonation of the amide in 10 with FA and coordination of the thus formed formate. Subsequently, decarboxylation of 11 affords dihydride I15, which is protonated by FA to afford I16. Finally, release of H₂ from I16 regenerates the active species 11.

The tridentate PNP ligand and the ancillary ligands about the iron center were modified to optimize the activity and stability of the catalyst [34]. However, under analogous conditions to those described above for 10 or 11 (0.1 mol% catalyst loading and 50 mol% Et₃N in dioxane at 80 °C) no significant improvement was achieved with catalysts 12–14 (Figure 3). In fact, only 13, which is closely related to 11, produced comparable results.

Figure 3. Depiction of catalysts 12–14.

The group of Gonsalvi developed a series of iron complexes containing rac- and meso-1,1,4,7,10,10-hexaphenyl-1,4,7,10-tetraphosphadecane (rac-p4 and meso-P4) [35]. The preformed complex cis-α-[Fe(CH₃CN)₂(rac-P4)](BF₄)₂ (15) was unsuccessfully tested as catalyst in the dehydrogenation of FA. However, in situ generated catalysts from the reaction of the iron(II) salt [Fe(H₂O)₆][BF₄]₂ with the ligands rac-P4 or meso-P4 showed good activities in propylene carbonate at temperatures between 40 and 60 °C under isobaric conditions (1 atm). The best activities were obtained for 2:1 and 4:1 excess of rac-P4, reaching TON and TOF values of 6061 and 1737 h⁻¹, respectively, for the latter. The reaction of rac-P4 with [Fe(H₂O)₆][BF₄]₂ affords complex 16, which was proposed as precatalyst.

Mechanistic studies based on NMR experiments led the authors to propose the catalytic cycle depicted in Scheme 8. The first step involves the formation of the κ²-formate intermediate I17, which is in equilibrium with its I18. β-hydride elimination from I19 affords the hydride intermediate I20. Finally, protonation of the hydride regenerates I17 and releases H₂. It is noteworthy that under isochoric conditions the authors observe the formation of a carbonyl complex (17), [FeH(CO)(rac-P4)]BF₄, which they suggest forms from the hydride species I20 by reaction with CO—generated by the competitive decomposition of HCOOH into CO and H₂O.

Scheme 8. Catalytic cycle proposed for the dehydrogenation of FA using 16 under isobaric conditions.

2.3. Manganese Homogeneous Catalysts

In 2016, Tondreu and Boncella reported that the Mn(I) complex [Mn(PNP)(CO)₂] (18; P = PⁱPr₂), which is isoelectronic with the Fe(II) species 10 (Scheme 6), reacts with FA to give the 1,2-addition product [Mn(PN^HP)(CO)₂] (19) (Scheme 9). Complex 19 proved to be an active catalyst precursor for the catalytic decomposition of FA in 1,4-dioxane. The authors observed that the addition of LiBF₄ to the reaction mixture inhibits the catalytic activity. The low activity of this manganese-based FA-dehydrogenation catalytic system, in addition to the poor selectivity—formation of CO was observed—, makes it unsuitable for fuel cell applications [36].

Scheme 9. Preparation of the first example of manganese-based FA dehydrogenation catalyst [Mn(PN^HP)(CO)₂] (19).

Some years later, the strategy of using the non-innocent N-H moieties in PNP ligands was applied to prepare the species [Mn(^tBuPN^HNOP)(CO)₂]Br (20). The deprotonation of the cationic species 20 with NaN(SiMe₃)₂ yields complex [Mn(^tBuPNNOP)(CO)₂] (21) (Scheme 10), which is an effective catalyst precursor for the dehydrogenation of FA to H₂ and CO₂ using chlorobenzene as solvent and Et₃N as base. The highest catalytic activity for this system (TOF = 8500 h⁻¹), which is comparable to some obtained using catalytic systems based on noble metals, was achieved using one equivalent of Et₃N to FA at 80 °C [37].

Scheme 10. Preparation of the manganese-based FA dehydrogenation catalyst [Mn(PNNOP)(CO)₂] (21).

Examples of phosphine free manganese catalysts based on bidentate Mn(I)-κ²-N-N species, recently reported by Beller et al., proved active for the selective dehydrogenation of FA in a mixture of H₂O (9 mL) and triglyme (4 mL) in presence of KOH (1.1 equiv. to FA) at 92.5 °C. The nature of the bidentate ligand influences the catalyst activity (Figure 4) [38,39]. However, the activity of Mn(I)-κ²-N-N species (TOF = 188–625 h⁻¹) is lower than that reported for 21.

Figure 4. Examples of Mn(I)-κ²-N-N catalyst precursors.

Britto and Jaccob have recently reported a mechanistic study at the DFT level on the 24-catalyzed dehydrogenation of FA. They found that the formation of the Mn-formate intermediate 28 from 24 follows a dissociative mechanism (Scheme 11). The dissociation of the Br^– ion creates a vacant site, which is occupied by a molecule of solvent (H₂O). The dissociative exchange of water by the formate ion leads to intermediate 28. The energy barrier of the activation process corresponds to TS26-28 (23.5 kcal mol⁻¹) [40].

Scheme 11. Energy profile for the formation of the Mn-formate active species 28 from 24.

Starting from the Mn-formate intermediate 28, the β-hydrogen elimination process to give the Mn-H species 29 has been found to be the rate-determining step of the catalytic cycle (19.0 kcal mol⁻¹). The generation of H₂ occurs by protonation of the Mn-hydride with a hydronium ion instead of FA (Scheme 12). This catalytic cycle is like that proposed by Beller’s group based on kinetic studies.

Scheme 12. Mechanism proposal for the 28-catalyzed dehydrogenation of FA.

3. Miscellaneous

3.1. Aluminum and Boron Homogeneous Catalysts

It is noteworthy that catalysts for the dehydrogenation of FA based on main-group elements have also been reported. Al-complex 33, which features an NNN participative ligand, proved active in a variety of solvents using Et₃N as an additive [47]. The best activities were observed in tetrahydrofurane at 65 °C using a 0.006 mol% of 33 and a HCOOH:Et₃N mixture (5:2), which bring about a TOF of 5200 h⁻¹ and a TON_1h of 2200. Detailed mechanistic investigation led the authors to propose the catalytic cycle depicted in Scheme 10. The pre-catalyst 33 reacts with FA to afford the resting state, 34, which presents a neutral NNN ligand formed by double protonation of the dianionic ligand in the parent complex. β-hydride elimination generates CO₂ and an Al-hydride intermediate (35). For this step, the authors postulate a transition state where the formate ligand dissociates to interact by hydrogen bond with the N-H moiety of the NNN ligand, thus allowing hydride abstraction by the Al-center (TS34-35). Finally, protonation of the hydride ligand by FA yields H₂ and regenerates the resting state (Scheme 17).

Scheme 17. Catalytic cycle proposed for the dehydrogenation of FA using 33 as catalyst. The symbol “‡” denotes a transition state.

Although boron is not a metal and the activities so far reported by catalysts based on this semimetal are modest, we believe that it is interesting to highlight the progress realized in this area because of the great promise for future development. Recent work by Cantat and coworkers describes the activity of dialkylborane derivatives in the presence of a HCOOH:Et₃N mixture (5:2) in various solvents at 130 °C. The best activity was obtained for [Cy₂B(OOCH)₂][Et₃NH] (35) in acetonitrile, allowing a TON of 100 after 26 h [48]. The authors proposed a catalytic cycle based on DFT studies that entails the formation of I21 by dissociation of formate, which allows the decarboxylation step to take place by hydride abstraction to give an active Cy₂BH intermediate that reacts with one formate ion to generate I22, which undergoes protonation of the hydride by FA to yield H₂ and I21 (Scheme 18).

Scheme 18. Catalytic cycle proposed for the dehydrogenation of FA using 32 as catalyst.

3.2. Heterodinuclear Homogeneous Catalysts

Heterodinuclear complexes featuring an Ir(III) center with a Co (36), Ni (37) or Cu (38) atom in the vicinity show a dramatic increase of their activity compared to the related mononuclear Ir(III) complex 39 (Figure 5) [49]. The highest synergistic effect was observed for Ni, with 36, 37 and 38 showing TOF values of 57, 72 and 26 h⁻¹, respectively. These TOF values sharply contrast with that of the mononuclear Ir(III) complex 39 (0.26 h⁻¹).

Figure 5. Depiction of complexes 36–39.

The authors proposed a reaction mechanism for 37 that entails the formation of a cationic aquocomplex (I23) that reversively coordinates a formate to give (I24). Subsequently, β-hydride elimination affords I25 upon coordination of a molecule of water. Finally, protonation of the hydride ligand at the Ir center by H₃O⁺, which is the rate limiting step, closes the catalytic cycle (Scheme 19).

Scheme 19. Catalytic cycle proposed for the dehydrogenation of FA using 37 as catalyst.