Release of Formic Acid from Copper Formate: Hydride, Proton‐Coupled Electron and Hydrogen Atom Transfer All Play their Role

Abstract Although the mechanism for the transformation of carbon dioxide to formate with copper hydride is well understood, it is not clear how formic acid is ultimately released. Herein, we show how formic acid is formed in the decomposition of the copper formate clusters Cu(II)(HCOO)3 − and Cu(II)2(HCOO)5 −. Infrared irradiation resonant with the antisymmetric C−O stretching mode activates the cluster, resulting in the release of formic acid and carbon dioxide. For the binary cluster, electronic structure calculations indicate that CO2 is eliminated first, through hydride transfer from formate to copper. Formic acid is released via proton‐coupled electron transfer (PCET) to a second formate ligand, evidenced by close to zero partial charge and spin density at the hydrogen atom in the transition state. Concomitantly, the two copper centers are reduced from Cu(II) to Cu(I). Depending on the detailed situation, either PCET or hydrogen atom transfer (HAT) takes place.

À and Cu(II) 2 (HCOO) 5 À . Infrared irradiation resonant with the antisymmetric CÀ O stretching mode activates the cluster, resulting in the release of formic acid and carbon dioxide. For the binary cluster, electronic structure calculations indicate that CO 2 is eliminated first, through hydride transfer from formate to copper. Formic acid is released via proton-coupled electron transfer (PCET) to a second formate ligand, evidenced by close to zero partial charge and spin density at the hydrogen atom in the transition state. Concomitantly, the two copper centers are reduced from Cu(II) to Cu(I). Depending on the detailed situation, either PCET or hydrogen atom transfer (HAT) takes place.
Carbon dioxide reduction is the key step in power to fuel applications. [1] The simplest reduced form of CO 2 is formic acid, which has received wide attention in electrochemical carbon dioxide activation, [2] not only as energy carrier and commodity, but also as a promising hydrogen storage material. [3] Gas-phase models for the formation of formic acid from CO 2 have focused on hydrogenation at the carbon atom, either by radical abstraction in the reaction of CO 2 À (H 2 O) n with thiols [4] or by hydride transfer from metal hydrides to CO 2 , [5][6][7] in both cases leading to formate. The mechanism of the release of neutral formic acid, however, is not known. We show here that formic acid release from a binary copper formate cluster proceeds by hydride transfer from formate to a metal center, followed by proton-coupled electron transfer (PCET), with concomitant reduction of two copper centers from Cu(II) to Cu(I). One of the formate ligands is sacrificed to provide the hydride intermediate, resulting in the net reaction (1).

HCOO
Copper formate has been chosen since copper is a versatile carboxylation catalyst, [8] and numerous gas phase studies have addressed various aspects of copper catalysis. [5,6,9] Surfacedeposited size-selected copper clusters have even been shown to catalyze the conversion of methane to methanol. [10] Gas phase models of CO 2 functionalization include studies with CO 2 À (H 2 O) n as well as Grignard-type metal complexes. [11] Copper formate clusters Cu n (HCOO) m À are generated by electrospray ionization and stored in the cell of a Fouriertransform ion cyclotron resonance (FT-ICR) mass spectrometer. Following mass selection, they are gently activated by irradiation with an infrared OPO system tuned to the antisymmetric CÀ O stretching vibration of the formate ligand around 1670 cm À 1 . Figure 1a shows mass spectra of mass-selected CuðIIÞ 2 ðHCOOÞ À 5 clusters after different irradiation times. The majority of the clusters reacts by loss of formally 2 HCOO * units, most likely in the form of formic acid and carbon dioxide, reaction (2), which according to our calculations lies 4.7 eV lower than loss of two formyloxyl radicals (at the B3LYP/ We performed extensive quantum chemical calculations to understand the mechanism of reaction (2). Since the binary CuðIIÞ clusters are difficult to describe, we first performed benchmark calculations against reliable thermochemical data on Cu þ ðH 2 OÞ n and CuOH + (H 2 O) n from the Armentrout laboratory, [12] as well as single-point CCSD calculations, see the Supporting Information (SI). The B3LYP and BMK functionals overall performed a little better than M06L, and reproduced reasonably the experimental data. The potential energy surface (PES) of the decomposition of Cu(II) 2 (HCOO) 5 À is shown in Figure 2a, calculated at the B3LYP/def2TZVP level of theory, with the corresponding BMK/def2TZVP values given in parentheses.
Two paths A and B for formic acid formation, reaction (2), were found, overall exothermic by À 0.62 eV at the B3LYP/ def2TZVP level. As detailed in Figure 2b, Path A proceeds via TS1, TS2 and TS3, with 0.89 eV for TS2 as the highest barrier while Path B goes through TS4 and TS5, the latter lying at 1.36 eV. Both pathways compete with cluster decomposition, reaction (3), at 1.36 eV. Further decarboxylation, reaction (4), proceeds through TS6 at 0.37 eV relative to the starting point, followed by a rearrangement through TS7 to reach the CuðIÞ 2 ðHCOOÞ 2 H À product with C 2v symmetry. Overall, the B3LYP surface is relatively flat while the BMK surface shows more pronounced barriers. The discrepancy comes mostly from the binary Cu(II) species, the relative energies along the Cu(I) part of the PES are quite similar for both functionals. This indicates possible problems of the single-reference methods to describe the interaction of two open-shell Cu(II) centers.
While the benchmark calculations favor neither BMK nor B3LYP, only the B3LYP barriers along pathway A are consistent with experiment. The branching ratio of the decomposition reaction (3) is only 8 %, ruling out Path B with a tight transition state for both functionals. The same argument applies to TS3 calculated with BMK, lying higher than the dissociation asymptote, reaction (3). With the high-lying TS3 in BMK, one would expect to observe the decarboxylation intermediate CuðIIÞ 2 ðHCOOÞ 4 H À , but no trace of the respective ion is visible at a noise level of 0.025 %, see Figure S1. We thus conclude that reaction (2) proceeds along pathway A as calculated with the B3LYP functional, without stabilization of the intermediate. This is plausible, given the overall exothermicity of the reaction and the relative energies of the transition states TS1 � TS2 > TS3.
To classify the mechanisms for the two steps, we analyzed the partial charges within the CHELPG [14] charge analysis and Mulliken spin densities along Path A, Table 1. The reactant CuðIIÞ 2 ðHCOOÞ À 5 is a singlet biradical, and only an unrestricted (UDFT) calculation yields a stable wave function. However, the spins at the two copper centers are not significantly antiferromagnetically coupled since the singlet is only 0.02 eV lower in energy than the triplet wave function (see Table S6). In TS1, the spin density stays mostly on the copper centers, and the hydrogen atom carries a significant negative charge of À 0.27 e, identifying this step as a hydride transfer. Similar values are obtained for TS2. In contrast, hardly any spin density is found on neither Cu nor H atoms in TS3. Together with the negligible partial charge of À 0.07 e on the H atom, the mechanism leading to formic acid is identified as proton-coupled electron transfer (PCET). [15,16] In the products, spin density has disappeared, consistent with a closed-shell Cu(I) 2 (HCOO) 3 À product ion.
Formic acid release via TS3 ultimately requires hydride, formate and two Cu(II) centers to accept the surplus electrons. The barrier from the intermediate via TS3 is only 0.40 eV, and is readily surpassed at room temperature. Reaction (1) can thus be rewritten as (1'), emphasizing these requirements.
The decomposition of Cu(II)(HCOO) 3 À provides additional insight into possible reactions at copper centers with relevance for the catalytic conversion of CO 2 and hydrogen storage. Figure 1b shows IRMPD mass spectra of this species. The dominant pathway is decarboxylation, reaction (5). In the next step, the product Cu(II)(HCOO) 2 H À loses a hydrogen atom, reaction (6), along with reduction of the copper center. Direct dissociation of HCOO from Cu(II)(HCOO) 3 À to form CuðIÞðHCOOÞ À 2 can be ruled out: Computationally, it requires 0.82 eV more energy compared to reaching TS9. In addition, if the laser is detuned to be off-resonance for Cu(II)(HCOO) 2 H À while still being on-resonance for Cu(II)(HCOO) 3 À , only Cu(II) (HCOO) 2 H À is formed. This confirms that Cu(II)(HCOO) 3 À does not directly decompose to Cu(I)(HCOO) 2 À .
Reaction (5) is very interesting since the reverse reaction is formate formation, which so far was not observed with Cu(II) hydrides in the gas phase. O'Hair and co-workers demonstrated it with Cu(I) species [6] while He and co-workers used Cu 2 H 2 À . [5] With the reduced copper species, formate formation proceeds nearly barrierless, as corroborated in our calculated PES for the reverse of reaction (8), proceeding through TS12 in Figure 2c. With Cu(II), the reverse barrier of reaction (5) via TS9 is 0.43 eV, which does not seem prohibitive in a real-life catalytic environment. Mechanistically, all formate decarboxylation reactions It is quite intriguing to note that the reactions (7), (7') and (9), in which Cu(II) is reduced, proceed via hydrogen atom transfer (HAT). The spin density in the transition state, displayed Figure S7, reveals that the transferred hydrogen atom is carrying a significant fraction of the spin density drawn from the Cu(II) center, with partial charges of + 0.01 e � q � À 0.23 e. The latter value, in comparison with the partial charges reported above for hydride transfer, however, illustrates that partial charge does not always allow to distinguish hydride from hydrogen atom transfer while spin density is a clear indicator.
In summary, we have established a mechanism for formic acid formation via PCET from two formate ligands at two Cu(II) centers. One formate ligand is sacrificed to provide the hydride required for formic acid release. HAT is operative in the decomposition of mononuclear copper formate complexes, including release of molecular hydrogen. These mechanisms may be operative at the active sites of heterogeneous copperhydride based catalysts.

Experimental Section
The clusters are prepared using electrospray ionization and transferred into the cell of a 9.4 T Fourier-transformation ion cyclotron resonance mass spectrometer which is explained in more detail elsewhere. [17] Here, ions are mass selected and can be irradiated in the wavelength range of 4476-12000 nm by a tunable optical parametric oscillator. To trigger decomposition, the ions are vibrationally excited through photon absorption of the antisymmetric CÀ O stretching vibration around 1670 cm À 1 and the fragmentation is investigated as a function of irradiation time. Further details are given in the SI. For the modelling of copper formate clusters, density functional theory (DFT) with the B3LYP and BMK functionals is used along with the def2TZVP basis set for geometry optimization. The choice of the functionals is based on benchmarking against the coupled cluster method, see Tables S1-4 in the SI. The wave function stability was tested for all calculated species. Transition states are verified through the intrinsic reaction coordinate (IRC) calculations and charge analysis is performed using the CHELPG scheme, with the Cu radius of 1.4 Å. [14] T 1 diagnostics at the CCSD/def2TZVP//B3LYP/def2TZVP level was performed (see Table S5), showing that values for ions with both one and two copper centers lie below 0.04, suggesting that single-reference calculations might be of reasonable accuracy in most cases. [18] However, for structures containing two Cu(II) centers, CCSD calculations had convergence issues, and no T 1 value could be obtained. All calculations are carried out in Gaussian 09. [19] All energies are reported with zero-point corrections .  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57