Dehydrogenation of Formic Acid to CO 2 and H 2 by Manganese(I)–Complex: Theoretical Insights for Green and Sustainable Route

: In this work, a detailed computational study on a recently synthetized Mn(I)-dependent complex [( tBu PNNOP)Mn(CO) 2 ] + is reported. This species promotes the dehydrogenation of formic acid to carbon dioxide and hydrogen. The here proposed catalytic cycle proceeds through the formation of stabilized adduct between [( tBu PNNOP tBu )Mn(CO) 2 ] + and formate and the progressive release of CO 2 and H 2 , mediated by the presence of trimethylamine. In order to evaluate the inﬂuence of the environment on the catalytic activity, different solvents have been taken into account. The computed barriers and the geometrical parameters account well for the available experimental data, conﬁrming the robustness of the complex and reproducing its good catalytic performance. Outcomes from the present investigation can stimulate further experimental works in the design of new more efﬁcient catalysts devoted to H 2 production.


Introduction
The increasing need for new and sustainable energetic resources represents one of the most important challenges characterizing the current century [1,2]. Indeed, fossil fuels, gas, coal and nuclear energy are still widely used, but environmentally dangerous, energy sources [2][3][4][5]. The intensive use of fossil fuels, for example, has been directly linked to the increasing level of CO 2 and greenhouse gas emissions, dramatically influencing the climate changes [3].
For these reasons, in the last fifty years, the interest devoted to possible "green" alternatives, like the use of sunlight-, wind-and water-based energies [6], have been increased, but, despite their promising efficiency, different technical issues are related to them and in particular to the storage of energy vectors on large scale [7].
One of the possible solutions is represented by the so-called sustainable hydrogen economy [8][9][10][11]. In this route, indeed, the electricity is converted in a secondary chemical energy carrier that can be used on demand [12][13][14][15][16]. The combustion of H 2 in the presence of O 2 , in devices like fuel cells, formally produces electricity and H 2 O, a green product.
On the other hand, the H 2 is not present on earth and it can be obtained/stored from/in organic compounds, like methanol and formic acid (FA) [17][18][19]. In particular, since its chemical-physical properties and its involvement in chemical industries and biomass production, the FA is believed a promising species for the hydrogen economy [1,2,20,21]. The dehydrogenation of FA, which generates CO 2 and H 2 , (Scheme 1) is usually mediated by metal-containing catalysts. In detail, the most widely adopted species contain Pd, Pt, Ir, Rh and Ru (see refs [22][23][24][25][26] as recent examples), with turnover frequencies (TOFs) variating from 10 2 to 10 6 h −1 [27]. Since all these species contain precious metals, nowadays one of the main goals is synthesizing complexes containing equally efficient metal ions belonging to the first row of transition series, which are notoriously cheaper and more abundant in the Earth's crust.
Promising solutions can arise by Fe-based catalysts showing turnover values comparable to those containing precious metals, but still demanding an improvement of their catalytic efficiency [28][29][30]. In addition, these species require the presence of a general base and Lewis acid, like Li + , to explicate the catalysis [29].
Very recently, for a new synthesized complex of Mn(I)-based complex, fast catalyzed dehydrogenation of formic acid has been described [31]. The organometallic compound, having the structure [( tbu PNNOPMn(CO)2] + (2), with tbu PNNOP = 2,6-(di-tert-butylphosphinito)(di-tert-butylphosphinamine)pyridine), exhibited substantially higher catalytic activity with respect to the unique known Mn-dependent complex acting on the formic acid [32], with TOFs improved to the value of 8500 h −1 [31]. The reaction requires the presence of a proton-acceptor system, but intriguingly the catalysis was not affected by the nature of the base [31].
In addition, the contribution of the hybrid tbu PNNOP ligand has been highlighted, since attempts with tbu PONOP ( tBu PONOP =2,6-bis(di-tert-butylphosphinito)pyridine, 4) and tbu PNNNP ( tBu PNNNP = 2,6-bis (di-tert-butylphosphinamine)pyridine, 5), two ligands tested on a previous Ru-based organometallic complex and depicted in Scheme 2 [33], did not provide remarkable catalytic activity [31]. Scheme 2. The three ligands (on the top) and the three related complexes (on the bottom). The name of the complexes has been retained according to the experimental work [31].
Encouraged by the results' novelty, which bring Mn into the family of earth-abundant metals catalyzing FA dehydrogenation [31], a detailed mechanistic study on the catalysis mediated by 2 in the framework of density functional theory (DFT) has been per-Scheme 1. Conversion of formic acid in CO 2 and H 2 .
In detail, the most widely adopted species contain Pd, Pt, Ir, Rh and Ru (see refs [22][23][24][25][26] as recent examples), with turnover frequencies (TOFs) variating from 10 2 to 10 6 h −1 [27]. Since all these species contain precious metals, nowadays one of the main goals is synthesizing complexes containing equally efficient metal ions belonging to the first row of transition series, which are notoriously cheaper and more abundant in the Earth's crust.
Promising solutions can arise by Fe-based catalysts showing turnover values comparable to those containing precious metals, but still demanding an improvement of their catalytic efficiency [28][29][30]. In addition, these species require the presence of a general base and Lewis acid, like Li + , to explicate the catalysis [29].
Very recently, for a new synthesized complex of Mn(I)-based complex, fast catalyzed dehydrogenation of formic acid has been described [31]. The organometallic compound, having the structure [( tbu PNNOPMn(CO) 2 ] + (2), with tbu PNNOP = 2,6-(di-tertbutylphosphinito)(di-tert-butylphosphinamine)pyridine), exhibited substantially higher catalytic activity with respect to the unique known Mn-dependent complex acting on the formic acid [32], with TOFs improved to the value of 8500 h −1 [31]. The reaction requires the presence of a proton-acceptor system, but intriguingly the catalysis was not affected by the nature of the base [31].
Promising solutions can arise by Fe-based catalysts showing turnover values comparable to those containing precious metals, but still demanding an improvement of their catalytic efficiency [28][29][30]. In addition, these species require the presence of a general base and Lewis acid, like Li + , to explicate the catalysis [29].
Very recently, for a new synthesized complex of Mn(I)-based complex, fast catalyzed dehydrogenation of formic acid has been described [31]. The organometallic compound, having the structure [( tbu PNNOPMn(CO)2] + (2), with tbu PNNOP = 2,6-(di-tert-butylphosphinito)(di-tert-butylphosphinamine)pyridine), exhibited substantially higher catalytic activity with respect to the unique known Mn-dependent complex acting on the formic acid [32], with TOFs improved to the value of 8500 h −1 [31]. The reaction requires the presence of a proton-acceptor system, but intriguingly the catalysis was not affected by the nature of the base [31].
In addition, the contribution of the hybrid tbu PNNOP ligand has been highlighted, since attempts with tbu PONOP ( tBu PONOP =2,6-bis(di-tert-butylphosphinito)pyridine, 4) and tbu PNNNP ( tBu PNNNP = 2,6-bis (di-tert-butylphosphinamine)pyridine, 5), two ligands tested on a previous Ru-based organometallic complex and depicted in Scheme 2 [33], did not provide remarkable catalytic activity [31]. Encouraged by the results' novelty, which bring Mn into the family of earth-abundant metals catalyzing FA dehydrogenation [31], a detailed mechanistic study on the catalysis mediated by 2 in the framework of density functional theory (DFT) has been per-Scheme 2. The three ligands (on the top) and the three related complexes (on the bottom). The name of the complexes has been retained according to the experimental work [31].
Encouraged by the results' novelty, which bring Mn into the family of earth-abundant metals catalyzing FA dehydrogenation [31], a detailed mechanistic study on the catalysis mediated by 2 in the framework of density functional theory (DFT) has been performed, with the aim of shedding light on the possible reaction mechanism and to characterize the intermediates and transitions states intercepted along the related potential energy surface (PES).
The adducts with formic acid of two other manganese ligands (complex 4, bu PNNOP and complex 5, Bu PNNNP of Scheme 2) have been considered. In addition, different solvents have been taken into account to evaluate their effect on the rate limiting step.
We hope that our results on the detailed catalytic mechanism of 2 can increase the knowledge of such a promising complex, helping the rationalization of experiments design of more efficient catalysts.

Results and Discussion
Before proceeding with the mechanistic study, it was mandatory to determine the spin multiplicity. For this purpose, optimizations of 2 catalyst in different spin states (2S + 1 = 1, 3, 5) have been performed. The singlet state (2S + 1 = 1) resulted in being the most stable one in agreement with d 6 configuration of Mn(I) species, as analogously determined in a recent study concerning the Mn(I)-mediated catalysis of organic nitriles to amides [34], while the triplet and quintet lay at 12.1 kcal/mol and 18.7 kcal/mol, respectively, with respect to the singlet spin state.
Small spin contamination (<6%, see Table S1), monitored as described in other works [35,36], has been observed with <S 2 > equal to 2.10 and 6.29 for triplet and quintet (2.00 and 6.00 after annihilation, respectively) and the single state has been selected for mechanistic investigation. Consistently with the X-ray structure of substrate-free complex (CCDC 1848774) [31], in the case of singlet state, good agreement with crystallographic distances between Mn(I) and atoms of its coordination sphere were found, with small deviation (<0.05 Å) compared with the experimental counterpart (see Figure S1). In the case of higher spin states, an increasing shift with respect to the X-ray structure has been observed (see Figure S1), confirming that the singlet spin state represents the suitable species for the study of catalytic mechanism.
The proposed reaction mechanism is reported in Scheme 3 and the related free energy profile is presented in Figure 1. After the formation of the complex-FA − adduct, the rearrangement of the substrate takes place, in which the bond with the metal switches from the oxygen to the hydrogen of the substrate. Successively, the reaction mechanism can be divided in two phases, as follows: In the first (Equation (1)), the CO 2 (the first product) is released with concerted formation of H-Mn bond. In the second phase (Equation (2)), due to the presence of Me 3 NH + species, the H 2 (the second product) is formed and subsequently released. Every phase consisted of a multistep process. The final mechanism included the formic acid in its deprotonated form (FA) produced by the trimethylamine acting as proton acceptor, since all the attempts performed on neutral species failed.
The first part of the study is related to the formation of adduct between complex 2 and formate (FA).  In the first (Equation (1)), the CO2 (the first product) is released with concerted formation of H-Mn bond. In the second phase (Equation (2)), due to the presence of Me3NH + species, the H2 (the second product) is formed and subsequently released. Every phase consisted of a multistep process. The final mechanism included the formic acid in its deprotonated form (FA) produced by the trimethylamine acting as proton acceptor, since all the attempts performed on neutral species failed.
The first part of the study is related to the formation of adduct between complex 2 and formate (FA). Two different conformers have been characterized with the formate bound to the Mn(I) in both Mn-O − (2:FA) and Mn-H (2:FA') fashions as shown in the optimized structures illustrated in Figure 2. The 2:FA' is higher in energy (8.3 kcal/mol) with respect to 2:FA, but, in spite of this, it resulted in being more reactive with the C-H distance of 1.29 Å and the OĈO angle of 140°, if compared to the corresponding ones in 2:FA (1.11 Å and 129°, respectively).
The interconversion between the Mn-O − and Mn-H takes place with an energy barrier of 15.5 kcal/mol across the transition state TS rot (see Figure 1). In the corresponding optimized geometry (see Figure 2), the formate acts as a bidentate ligand owing to the forming Mn-H 2.22 Å and the breaking Mn-O-bonds 2.81 Å, thus offering an epta-coordinated manganese. DFT calculations on the dehydrogenation of formic acid promoted by a pincer-supported iron catalyst evidenced analogous behavior proposing an energy barrier of 21.4 kcal/mol for the switching from the Fe-O − to the Fe-H conformer [29].
Starting from 2:FA', the CO2 release takes place by overcoming the barrier of 8.4 kcal/mol represented by TS1 in Figure 3. In particular, the H lies at 1.76 Å from the Mn(I) and at 1.67 Å from the C, while the value of OĈO angle (152°) increased preparing for the next carbon dioxide release. The related imaginary frequency (236i cm −1 ) concerns the C-H stretching. In the intermediate 2(H) the H-Mn bond resulted in being definitely formed, as confirmed by the distance of 1.60 Å (see Figure S2). From this species starts the second phase of the dehydrogenation of formic acid leading to the formation of H2. To do this, a protonated species, most likely the conjugated acid of the base involved in the catalysis, must come into play acting as proton donor. As confirmed by experimental evidences [31], its presence is fundamental for the advancing of the reaction, being deputed to the deprotonation of the formic acid. Moreover, Tondreau et al. in their kinetic studies advised that "the identity of the base does not significantly alter catalytic results" [31], so in our investigation the protonated triethylamine used in the experiment has been replaced by the protonated trimethylamine (Me3NH + or TMA) in order to save computational cost.  In the intermediate 2(H) the H-Mn bond resulted in being definitely formed, as confirmed by the distance of 1.60 Å (see Figure S2). From this species starts the second phase of the dehydrogenation of formic acid leading to the formation of H2. To do this, a protonated species, most likely the conjugated acid of the base involved in the catalysis, must come into play acting as proton donor. As confirmed by experimental evidences [31], its presence is fundamental for the advancing of the reaction, being deputed to the deprotonation of the formic acid. Moreover, Tondreau et al. in their kinetic studies advised that "the identity of the base does not significantly alter catalytic results" [31], so in our investigation the protonated triethylamine used in the experiment has been replaced by the protonated trimethylamine (Me3NH + or TMA) in order to save computational cost.  Indeed, the proton shift, which leads to the formation of the H2 product, is described in the TS2, where the forming Hbase-H bond assumes a value of 1.05 Å vs. that of 1.45 Å found in 2(H):Me3NH + of Figure S2 while the Mn-H results to be 1.71 Å. These findings evidence the important role of base also in the formation of the second product, since its nature as a proton-donor species. TS2 lies at 11.8 kcal/mol related to the 2(H):Me3NH + intermediate but at 7.4 kcal/mol with respect to 2:FA (see Figure 1). In the 2(H2) species, The interconversion between the Mn-O − and Mn-H takes place with an energy barrier of 15.5 kcal/mol across the transition state TS rot (see Figure 1). In the corresponding optimized geometry (see Figure 2), the formate acts as a bidentate ligand owing to the forming Mn-H 2.22 Å and the breaking Mn-O-bonds 2.81 Å, thus offering an epta-coordinated manganese. DFT calculations on the dehydrogenation of formic acid promoted by a pincersupported iron catalyst evidenced analogous behavior proposing an energy barrier of 21.4 kcal/mol for the switching from the Fe-O − to the Fe-H conformer [29].
Starting from 2:FA', the CO 2 release takes place by overcoming the barrier of 8.4 kcal/ mol represented by TS1 in Figure 3. In particular, the H lies at 1.76 Å from the Mn(I) and at 1.67 Å from the C, while the value of OĈO angle (152 • ) increased preparing for the next carbon dioxide release. The related imaginary frequency (236i cm −1 ) concerns the C-H stretching.
Catalysts 2021, 11, x FOR PEER REVIEW 6 of 10 the molecular hydrogen is directly linked to the Mn(I). The H2 σ bond length is 0.77 Å and the distance of H2 from metallic center is equal to 1.91 Å, see Figure S2. The TS1 describing the release of CO2 represents the highest relative barrier, presenting an energy 1.2 and 9.3 kcal/mol higher than those of the TSrot and TS2, respectively. The value of 16.7 kcal/mol, in addition, is in good agreement with the available kinetic data proposing a kcat value of 2.4 s −1 , converted to a ΔG # = 17 kcal/mol adopting Eyring's equation [31]. As can be also noted from the analysis of the energy profile, the kinetic of reversible reaction, proceeding from 2(H2) → 2:FA is slower. Indeed, for this reaction an energy barrier of 21.1 kcal/mol resulted for the step 2(H):Me3NH + -TS1, 4.4 kcal/mol higher than that calculated for the 2:FA → 2(H2) pathway.
We also attempted to rationalize the effect of the pincer ligands 4 and 5 of Scheme 2 on the catalysis, focusing on their affinity (ΔHaff) for the substrate and to the fundamental formation of complex:substrate adduct that represents a crucial step for the proceeding of catalysis, as follows:
A possible explanation for this behavior arises from the analysis of the charge distribution in the three 2:FA, 4:FA and 5:FA adducts, reported in Figure S3.
In particular, an evident effect concerns the Mn(I) center that exhibits a more negative charge in presence of tbu PNNNP (−0.08 |e|) and of tbu PONOP (−0.05 |e|) than in presence of the tbu PNNOP. This could consequently affect the coordination of the negatively charged HCOO − with a reduced catalytic activity.
The ligand tbu PNNOP proposes the right balance between electron withdrawing and electron donating effect, represented respectively by the O and N atoms, to the pyridine ring. Moreover, from the maps of the electrostatic potential of the three ligands depicted in Figure S4 emerges as the presence of two oxygens in tbu PONOP leads to a marked decrease of negative charge on the aromatic ring contrarily to what occurs in tbu PNNNP, this behavior can have repercussions on their catalytic activity. In the intermediate 2(H) the H-Mn bond resulted in being definitely formed, as confirmed by the distance of 1.60 Å (see Figure S2). From this species starts the second phase of the dehydrogenation of formic acid leading to the formation of H 2. To do this, a protonated species, most likely the conjugated acid of the base involved in the catalysis, must come into play acting as proton donor. As confirmed by experimental evidences [31], its presence is fundamental for the advancing of the reaction, being deputed to the deprotonation of the formic acid. Moreover, Tondreau et al. in their kinetic studies advised that "the identity of the base does not significantly alter catalytic results" [31], so in our investigation the protonated triethylamine used in the experiment has been replaced by the protonated trimethylamine (Me 3 NH + or TMA) in order to save computational cost.  Figure S2) indicating its early activation.
Indeed, the proton shift, which leads to the formation of the H 2 product, is described in the TS2, where the forming H base -H bond assumes a value of 1.05 Å vs. that of 1.45 Å found in 2(H):Me 3 NH + of Figure S2 while the Mn-H results to be 1.71 Å. These findings evidence the important role of base also in the formation of the second product, since its nature as a proton-donor species. TS2 lies at 11.8 kcal/mol related to the 2(H):Me 3 NH + intermediate but at 7.4 kcal/mol with respect to 2:FA (see Figure 1). In the 2(H 2 ) species, the molecular hydrogen is directly linked to the Mn(I). The H 2 σ bond length is 0.77 Å and the distance of H 2 from metallic center is equal to 1.91 Å, see Figure S2.
The TS1 describing the release of CO 2 represents the highest relative barrier, presenting an energy 1.2 and 9.3 kcal/mol higher than those of the TSrot and TS2, respectively. The value of 16.7 kcal/mol, in addition, is in good agreement with the available kinetic data proposing a k cat value of 2.4 s −1 , converted to a ∆G # = 17 kcal/mol adopting Eyring's equation [31].
As can be also noted from the analysis of the energy profile, the kinetic of reversible reaction, proceeding from 2(H 2 ) → 2:FA is slower. Indeed, for this reaction an energy barrier of 21.1 kcal/mol resulted for the step 2(H):Me 3 NH + -TS1, 4.4 kcal/mol higher than that calculated for the 2:FA → 2(H 2 ) pathway.
We also attempted to rationalize the effect of the pincer ligands 4 and 5 of Scheme 2 on the catalysis, focusing on their affinity (∆H aff ) for the substrate and to the fundamental formation of complex:substrate adduct that represents a crucial step for the proceeding of catalysis, as follows:  [31].
A possible explanation for this behavior arises from the analysis of the charge distribution in the three 2:FA, 4:FA and 5:FA adducts, reported in Figure S3.
In particular, an evident effect concerns the Mn(I) center that exhibits a more negative charge in presence of tbu PNNNP (−0.08 |e|) and of tbu PONOP (−0.05 |e|) than in presence of the tbu PNNOP. This could consequently affect the coordination of the negatively charged HCOO − with a reduced catalytic activity.
The ligand tbu PNNOP proposes the right balance between electron withdrawing and electron donating effect, represented respectively by the O and N atoms, to the pyridine ring. Moreover, from the maps of the electrostatic potential of the three ligands depicted in Figure S4 emerges as the presence of two oxygens in tbu PONOP leads to a marked decrease of negative charge on the aromatic ring contrarily to what occurs in tbu PNNNP, this behavior can have repercussions on their catalytic activity.
Given the important role played in homogeneous catalysis by the solvent, with the aim of investigating its effect on the dehydrogenation of formic acid assisted by the 2 catalyst, different implicit solvents have been tested on the CO 2 release step (TS1), considering dielectric constant with gradually increasing values to simulate apolar, polar aprotic and polar protic solvents extensively used in organometallic reactions [37]. The results are collected in Figure 4. Given the important role played in homogeneous catalysis by the solvent, with the aim of investigating its effect on the dehydrogenation of formic acid assisted by the 2 catalyst, different implicit solvents have been tested on the CO2 release step (TS1), considering dielectric constant with gradually increasing values to simulate apolar, polar aprotic and polar protic solvents extensively used in organometallic reactions [37]. The results are collected in Figure 4.
An interesting behavior has been observed in the case of protic polar solvents. In the case of ethanol (ε = 24.8), methanol (ε = 32.6) and water (ε = 78.3), a remarkable increment of TS1 barrier resulted, with values of 3.2 kcal/mol, 4.2 kcal/mol and 3.8 kcal/mol, respectively, indicating that the choice of operative conditions can affect the efficiency of catalytic mechanism. Finally, for the apolar solvent cyclohexane (ε = 2.0), no relevant variations occurred.
An interesting behavior has been observed in the case of protic polar solvents. In the case of ethanol (ε = 24.8), methanol (ε = 32.6) and water (ε = 78.3), a remarkable increment of TS1 barrier resulted, with values of 3.2 kcal/mol, 4.2 kcal/mol and 3.8 kcal/mol, respectively, indicating that the choice of operative conditions can affect the efficiency of catalytic mechanism. Finally, for the apolar solvent cyclohexane (ε = 2.0), no relevant variations occurred.

Materials and Methods
All the calculations have been carried out with the Gaussian09 D.01 software package [38]. The available x-ray data (CCDC 1848774) has been selected as starting structure for the mechanistic study [31]. The geometries have been fully optimized without any physical constraints, adopting B3LYP functional [39,40] and 6-31+G(d,p) basis set for C, H, N, O and P atoms. For the Mn, instead, the SDD pseudopotential [41] and its related basis set has been chosen. In addition, the dispersions according to Grimme's scheme (D3-BJ) have been included in all the computations [42]. The strategy here presented has been successfully described and adopted in previous works on catalytic systems containing first row transition metals [34,[43][44][45][46][47]. Given the lack of information on the multiplicity of the catalytic system, preliminary investigation has been carried out considering different spin states, (2S + 1 = 1, 3, 5).
The final energies (∆G) reported on the PESs additionally include the most accurate electronic energy obtained from single point energy calculations employing the more extended 6-311+G(2d,2p) basis set, and the Gibbs free energy corrections, extrapolated from thermochemical analysis.

Conclusions
On the basis of a detailed computational investigation performed in solvent and taking into account the experimental information, a catalytic cycle for the dehydrogenation of formic acid promoted by a recently synthetized Mn(I)-dependent complex [( tBu PNNOP) Mn(CO) 2 ] + as catalyst has been proposed. The process consists of two elementary steps: release of CO 2 and production of H 2 . The proposed reaction mechanism well fits the experimental behaviors.
The CO 2 release represents the rate determining state requiring 16.7 kcal/mol. The pincer ligand with its steric hindrance ensures the occurrence of the catalytic process confirmed by the scaffold stability on all the intercepted stationary points.
Our calculations highlight that the presence of a base is fundamental for the dehydrogenation of formic acid, not only in the deprotonation of acid, but also during the formation of the second product, acting as proton-donor species.
The effect of the solvent on the catalytic process has been evaluated by taking into account apolar, polar aprotic and polar protic solvents simulated by opportune dielectric constant values.
We believe that the results from the present investigation can stimulate further experimental works in the design of new more efficient catalysts devoted to H 2 production.