UV/Vis Spectroscopy of Copper Formate Clusters: Insight into Metal‐Ligand Photochemistry

Abstract The electronic structure and photochemistry of copper formate clusters, CuI 2(HCO2)3 − and CuII n(HCO2)2n+1 −, n≤8, are investigated in the gas phase by using UV/Vis spectroscopy in combination with quantum chemical calculations. A clear difference in the spectra of clusters with CuI and CuII copper ions is observed. For the CuI species, transitions between copper d and s/p orbitals are recorded. For stoichiometric CuII formate clusters, the spectra are dominated by copper d–d transitions and charge‐transfer excitations from formate to the vacant copper d orbital. Calculations reveal the existence of several energetically low‐lying isomers, and the energetic position of the electronic transitions depends strongly on the specific isomer. The oxidation state of the copper centers governs the photochemistry. In CuII(HCO2)3 −, fast internal conversion into the electronic ground state is observed, leading to statistical dissociation; for charge‐transfer excitations, specific excited‐state reaction channels are observed in addition, such as formyloxyl radical loss. In CuI 2(HCO2)3 −, the system relaxes to a local minimum on an excited‐state potential‐energy surface and might undergo fluorescence or reach a conical intersection to the ground state; in both cases, this provides substantial energy for statistical decomposition. Alternatively, a CuII(HCO2)3Cu0− biradical structure is formed in the excited state, which gives rise to the photochemical loss of a neutral copper atom.


Introduction
With global warming becoming an ever-increasing problem, efficient activation and transformation of carbon dioxide becomes ad esirable optionf or carbon capture and usage (CCU). Transformation into formic acida nd methanol are very attractive candidates. [1][2][3] Formic acid is regarded as the simplest stable form of activated carbon dioxide and is widely investigated. [4][5][6][7] In these transformations, formate is considered as a key intermediate for the hydrogenation of carbon dioxide. [2,4] Furthermore, the highly selectived ecomposition of formic acid could play ak ey role in hydrogen storagea pplications. [8,9] Copper-based catalysts are highly active and applieda sacarboxylationa gent. [10,11] Additionally,t hey are already widely researched and used in industry within methanols ynthesis. [12][13][14][15] However,t he nature of the active sites of catalytic materials and the respective electronic configurationremainhotly debated. [16,17] For am echanistic understanding,g as-phase investigations of well-defined, size-selectedm odel systems, in combination with quantum chemical calculations,a re useful. [18][19][20] Additionally,g as-phase approaches can be translatedt hrough size-selectedc luster depositiontos tudy surface effects within heterogeneous catalysis. [21,22] In the gas phase, it was shownt hat the activation of methane to methanolm ight take place on a CuO + center. [23] Al 2 CuO 5 + complexesh ave been found to selectively convert methane into CH 2 O. [24,25] Furthermore, it was revealed that hydrated doublyc harged copperi ons, Cu 2 + (H 2 O) n ,u nderwent charges eparation until ac riticals izeo f n = 8. [26][27][28] Additionally,g as-phase experiments and quantum chemicalc alculations have been applieds uccessfully to understand the mechanismsb ehind the activationo fc arbon dioxide on copperh ydrides, [29][30][31][32][33] with copper formate clustersa sp articularly suitable model systems. [31,33] Recently,w es howedt hat the formationo ff ormic acid within copperf ormate clusterst ook place in the gas phase in ar edox reaction through proton-coupled electron transfer for clusters with two copperc enters, whereas it proceeded through hydrogen atom transfer with as ingle copperc enter. [31] Furthermore, we observed that the size of the cluster and the copper oxidation state of + II played ak ey role in the reaction of formate towards formic acid, emphasizing the importance of electronic structures for catalysis. LargerC u II formate anions have been found to fragment in ac ascadet owards clusters with one or two copperc enters, upon which redox reactions can be observed, followed by decarboxylation towards Cu I hydrides as the predominant process. [33] The formate anion in variouse nvironments has been investigated, in detail,b y means of infrared spectroscopy and theoryd ue to anomalies within the vibrational spectrum with intense anharmonicities, [34][35][36][37][38] and its electronic structure has been studied in solution. [39] In the catalytically relevant case of formate ligated copper, structurald ata on coppers urfaces can be found, [40][41][42] whereas data on the electronic structure is still sparse. The electronic structure of small, neutral copperc lusters was investigated in an eon matrix by using spectroscopy techniques within the UV/Vis range. [43][44][45] In well-definedg as-phasee xperiments,t he electronic structure of al igated CuO + core thought to be responsible for many reactions in nature within coppercontaining enzymes has been characterized, [46] along with detailed calculations on the bareC uO + ion. [47] In the closely related Cu(NO 3 ) 3 À anion,l igand-to-metal charge-transfer (LMCT) transitions and transitions within the nitrate ligand wereo bserved in the UV region. [48] Excitation causes the evaporation of an eutral NO 3 radical, similart oc ollision-induced dissociation (CID) experiments, which indicates efficient internal conversion to the electronic ground state. [48][49][50] Herein, we presenta ni nvestigation into the electronic structure of copper formatec lusters in the gas phase by using UV/ Vis spectroscopy along with quantum chemical calculations. We also explore selected photochemical reactionp athways.

Experimental Section
Mass selected, isotopically enriched copper-63 formate cluster anions were produced through electrospray ionization (see ref. [31] for details) and investigated at room temperature in the gas phase with aB ruker APEX Qe 9.4 Tesla Fourier-transformation ion cyclotron resonance (FT-ICR) mass spectrometer,w hich is described in more detail elsewhere. [31,51,52] The trapped clusters were irradiated with an EKSPLA NT342B optical parametric oscillator,w hich provided pulsed laser light from l = 225 to 2600 nm. The obtained photodissociation spectra were corrected for fragmentation caused by blackbody infrared radiative dissociation (BIRD) [53][54][55][56][57][58] and collisions with the background gas. Photodissociation cross sections were calculated as described in detail before. [59] The laser power was measured after each mass spectrum was recorded and corrected for transmission through one additional CaF 2 prism and window. The beam profile was estimated as ah omogenous Gaussian beam with ad ivergence of 0.002 mrad and ab eam diameter of 0.005 m at the laser output, along with ac onstant output angle, leading to uncertainties in the absolute cross-section calculations. To account for changes in the laser beam profile and alignment upon switching between the different laser stages, ac orrection factor was applied at the signal to idler beam transition (> 710 nm), signal to UV stage (< 410 nm), and SF/SH beam transition (< 296 nm). Here, the cross section of the signal region was taken as ar eference because this was the most reliable value. For the modeling of the electronic ground-state properties of copper formate clusters, DFT calculations at the B3LYP/def2TZVP level were used for optimization, based on extensive benchmarking performed previously. [31] Al arge number of structures were optimized based on refs. [31] and [33].H erein, we considered several conformers with formate ligands that exhibited different binding motifs (monodentate, bidentate, and bridging bidentate) and orientations on one or more copper centers. The energetically most favorable isomers, with significant structural differences, were then used for further analysis. For the computational description of the excited states, equation of motion-coupled cluster singles and doubles (EOM-CCSD) calculations were performed for small clusters (< 150 electrons). The augcc-pVDZ basis set was used here because it provided results close to the triple-zeta basis set, see Table S5 in the Supporting Information. All excited states within 6eV, or at least 12 states (in the case of convergence issues), were considered. For structural optimization in the electronically excited states, the 6-31 + g* basis set was employed. Near conical intersections, the optimization started to oscillate. Therefore, reduced convergence criteria were used (predicted change in energy < 2 10 À5 hartree). To compute electronic transitions in larger clusters, time-dependent (TD) DFT with the BMK functional was applied, which was chosen based on its performance compared with EOM-CCSD for copper formate in oxidation state + II, see Table S5 in the Supporting Information. Orbitals participating in electronic transitions were analyzed by using the natural transition orbital (NTO) scheme at the TD-BMK/aug-cc-pVDZ//B3LYP/def2TZVP level of theory.A ll calculations were carried out with the Gaussian 16 program. [60] The reported reaction energies were zero-point corrected.

Results and Discussion
Absorption spectra In Figure 1, spectra of Cu I 2 (HCO) 3 À ,C u II (HCO) 3 À ,C u II 2 (HCO) 5 À , Cu II 3 (HCO) 7 À ,a nd Cu II 8 (HCO) 17 À are shown in the 0.9-5.5 eV region, along with the calculated electronic transitions for selected isomers. The Cu I system, Cu I 2 (HCO) 3 À ,e xhibits absorptions only in the UV region,s tartinga ta bout 4.0 eV.T he Cu II systemse xhibit two absorption bands in the 1.0-2.4 and 3.5-5.5 eV regions.The intensity of the absorption bands is comparable for all Cu II clusters, with the low-energy absorption being about five times less intense than that of the high-energyo ne.
There is no significant shift in energy observed between systems with one or two copper centers. We can therefore conclude that the bonding interaction between two neighboring copper centers in stoichiometric copper(II) formate clusters is marginal. Instead, the oxidation state of the copper centerh as the largest influence on the spectra.I nt he UV spectrum of larger clusters, the electronic transitions appear with alesspronounced maximum and occur deeper within the UV region for the largest case, Cu II 8 (HCO) 17 À . The calculated EOM-CCSD electronic transition energies are shown in Figure 1a andb .For larger clusters (Figure 1c and d), TD-BMK calculations were employed. The calculations reveal that the orientation and binding motif of the formate ligands heavily influence the position and oscillator strengtho ft he bands, as evidenced by the calculated electronic transitions of selected structures. For Cu I compounds and low-lyings tates in Cu II species, av ery good agreement with the experiments is reached.T he higher-lying states in Cu II compounds, on the other hand, are shifteds ystematically to highere nergies, see, for example, the Cu II (HCO) 3 À spectrum in Figure 1b.H ere, multireference effects potentially play ar ole. Moreover,d ynamic effects can furthera ffect the absorption spectra. Considerable energy and intensitys hifts, as wella ss ubstantial peak broadening, can be expected due to, for example, internal rotation of the formate group or partial opening of ab identate ligand to am onodentate structure. Zero-point energye ffects, [61] estimated as 0.2-0.3 eV,m ight also play ar ole. Overall, the agreement between experiment and theory is well within the expected range for transition-metalc ompounds.
In contrastt ot he strong effect of the formate ligands on the transitions, the change in the number of copper centers from one to two does not significantly change the position of the transitions, in agreement with experiment. This is also consistent with the previously described extremelyw eak antiferro-magnetic coupling of the unpaired electrons in Cu II 2 (HCO) 5 À and the large CuÀCu distance in copper formate clusters. [31,33] The observed absorption band structures can be rationalized by the electronic configuration of the copperi on (Scheme 1). In oxidation state + I, ac opper ion has af ull ds hell, whereas there are only nine de lectrons for the oxidation state + II. This deeply affects the character of possible excitations.
Copper(I) only affords excitations from the filled do rbital into the empty s/p shells, which corresponds to the band starting at about 4eV( see Figure 1a). The hole in the ds hell of Cu II ,however,makes low-lying d-d transitions and LMCT transitions from the negatively charged formate ligands possible, Figure 1. The photodissociation cross section, s,o fa )Cu I 2 (HCO 2 ) 3 À ,b )Cu II (HCO 2 ) 3 À ,c )Cu II 2 (HCO 2 ) 5 À ,d)Cu II 3 (HCO 2 ) 7 À ,and e) Cu II 8 (HCO 2 ) 17 À ,together with the calculated oscillators trength, f,o ft he electronic transitionsfor selected isomers. Relative energy of isomers is given in eV in the samecoloru sed in the bar graph of the oscillatorstrengthw ithin eachcorresponding subsection. The level of theory is EOM-CCSD/aug-cc-pVDZ//B3LYP/def2TZVP for Cu I 2 (HCO) 3 À and Cu II (HCO) 3 À ;TD-BMK/aug-cc-pVDZ//B3LYP/def2TZVP is used for Cu II 2 (HCO) 5 À and Cu II 3 (HCO) 7 À .Experimental cross sections below 2.75eVare multiplied by a factoro f5or 3, and theoretical f is multiplied by 50 for better legibility. which correspond to the absorption bands at 1-2 and 3.5-5.5 eV,r espectively (see Figure1b-e). The d-s/p transitions are substantially blueshifted for Cu II due to the higher number of ligands, which result in more restricted space for the target 4s/po rbitals, and therefore, lie outside the range of our laser system.
Our analysiso fo rbitals participating in electronic transitions for Cu I 2 (HCO 2 ) 3 À ,C u II (HCO 2 ) 3 À ,a nd Cu II 2 (HCO 2 ) 5 À confirms these qualitative arguments (Figure 2). For Cu I 2 (HCO 2 ) 3 À ,atransition from copper d-type orbitals into the unoccupied copper s/p orbitals is observed. In the ground state of Cu I 2 (HCO 2 ) 3 À ,t he 4s/p orbitals of the two copper ions are unoccupied. The fully filled do rbitals have an onbonding character for isomer 1, whereas isomer 2e xhibits aC u ÀCu s bond. This means av ery Scheme1.Low-lying electronic transitionsw ithin copper formate complexes for Cu I and Cu II . Figure 2. NTOs for selected electronic transitions in a) Cu I 2 (HCO 2 ) 3 À ,b)Cu II (HCO 2 ) 3 À ,and c) Cu II 2 (HCO 2 ) 5 À calculated at the TD-BMK/aug-cc-pVDZ//B3LYP/ def2TZVP level of theory. CuÀCu bondsa re shownonly to guide the eye;f or isomer 1o fC u I and Cu II ,t he Cu-Cu interaction is weak. The shown orbitals are NTOs, that is, they correspondtot he initial and final electron orbitals within given excitations. For Cu I 2 (HCO 2 ) 3 À ,two isomers are shown. For nearly degenerate transitions, orbitalso fd ifferent symmetry mix, so only the most important component is shown. limited bonding interaction betweent he two copperc enters in S 0 for isomer 1, whereas isomer 2h as ab onding interaction, which is reflected in the bond length of d(CuÀCu) = 3.3 and 2.5 in isomers 1a nd 2, respectively.C omparing the isomers, one can see why excitation energies within Cu I 2 (HCO 2 ) 3 À are so sensitivet ot he orientation of the formate ligands. In isomer 1, two so rbitals form ab onding s(CuÀCu) orbital that is compressedb yt he presence of formate ligands, which extends into the space along the CuÀCu axis. The single electron is thus confined to ar elativelyn arrow space, which resultsi na less favorable electronic state. In isomer 2, the s/p orbitals extend freely along the CuÀCu axis, shifting the first excitation energy by about À1.0 eV.The target orbitals are almost degenerate, irrespectiveo ft heir phase, showing the independence of both copperi ons;s ee also below for ad iscussion on the resulting photodynamics.
For Cu II (HCO 2 ) 3 À (Figure 2b), four d-dt ransitions compose the first absorption band, exhibiting al ow oscillator strength because they are symmetry forbidden in Cu II .I nt he UV region, the singly occupied do rbital can accept an electronf rom the nonbonding p orbitalo ft he negativelyc harged formate ligands.T his allows for comparatively intenseL MCT transitions. Dependingo nt he orientation of the ligands, the electronic orbitals are more or less confined by the formate ligand, leading again to different excitation energies and oscillator strengths. This also intuitively explainsw hy the open isomer of Cu II (HCO 2 ) 3 À ,w ith only monodentate formate ligands (see Figure 1) has energetically lower-lying transitions because more space is available for the target orbital. Additionally,t he overlap of the ligand orbitals with the copper center is expected to change upon ligand rotation. The observed character of the transitions in Cu II (HCO 2 ) 3 À is in good agreement with the previously studied Cu II (NO 3 ) 3 À anion, for which TDDFTc alculations predict d-d and LMCT transitions, along with energetically higher transitions on the nitrate ligand. [48] The singly occupied do rbitals of the copper centers in Cu II 2 (HCO 2 ) 5 À are independent. This is reflectedi nt he first two participating orbitalsofeach absorption band illustrated in Figure 2c,w hich are energetically nearly degenerate and differ only by the position of the orbital on the right-or left-hand sides of the cluster. Furthermore, the structure has ar elatively long bond length of d(CuÀCu) = 3.0 .

Photochemistry
In Figure 3, we analyze the photochemical fragments of copperf ormate clustersf or Cu I 2 (HCO 2 ) 3 À ,C u II (HCO 2 ) 3 À ,a nd Cu II 2 (HCO 2 ) 5 À ,a sr epresentative examples. The predicted reaction channels for the most intense fragments of the three species are listed in Table 1, togetherw ith their experimental appearance energies andc alculated reaction energies. In the spectrum of Cu I 2 (HCO 2 ) 3 À ,w eo bserve Cu I 2 H 3 À ,C u II (HCO 2 ) 2 H À , Cu I (HCO 2 ) 2 À ,a nd Cu I (HCO 2 )H À ,r eactions (1 a)-(1 d), with roughly constant branching ratio acrosst he absorption band. Reaction (1 a) includes three decarboxylation steps, requiring 2.29 eV,w hich is the same decomposition pathway as that observed in infrared multiple photon dissociation (IRMPD) experiments. [33] With UV photons, the clusters can also disproportionate by losing an eutral copper atom and decarboxylate to Cu II (HCO 2 ) 2 H À ,r eaction (1 b).A lternatively,t hey could lose a neutralc opper formate unit and partially decarboxylate to form Cu I (HCO 2 ) 2 À or Cu I (HCO 2 )H À ,r eactions (1 c) or (1 d), respectively.R eactions (1 b)-(1 d) have not been observed upon IR heating, [33] which suggeststhat they require an electronically excited state. The calculated reactione nergies are consistently lower than the experimental appearance energies, which confirms that dissociation is energeticallyf easible with as ingle photon.
For Cu II (HCO 2 ) 3 À and Cu II 2 (HCO 2 ) 5 À ,f ragment branching ratios depend considerably on wavelength (Figure3ba nd c). In the spectrum of Cu II (HCO 2 ) 3 À ,a ll observed fragments, namely, Cu II (HCO 2 ) 2 H À ,C u I (HCO 2 ) 2 À ,C u I (HCO 2 )H À ,a nd Cu 0 (HCO 2 ) À ,c an be explained by previously observed reactions (2 a)-(2 d')u pon IR heating. [31,33] This points towards af ast internal conversion of the photon energy,f ollowed by decomposition in the ground state. The changing branching ratio could be explained by the amount of available energy,a llowing more sequential fragmentation stepsf or higherphoton energies. For the low-energy band of Cu II 2 (HCO 2 ) 5 À ,m ainly the formation of Cu I 2 (HCO 2 ) 3 À ,C u I 2 (HCO 2 ) 2 H À ,a nd Cu II (HCO 2 ) 3 À is observed, with appearance energies in the range of 1.0-1.2 eV, see reactions (3 a)-(3 c) in Table 1. Additionally,C u II (HCO 2 ) 2 H À and Cu I (HCO 2 ) 2 À are formed through reactions (3 d) and (3 e), albeit with low intensity.F or energetic reasons, two photons are required for reactions (3 d) and (3 e), indicating that these ions are secondary products of Cu II (HCO 2 ) 3 À .R eactions (3 a)-(3 e) have all been observed previouslyi nIRMPDe xperiments, [31,33] which suggests that internal conversion precedes dissociation. Interestingly,t he absorption band starts with the lowest barrier fragment, the formation of formic acid with eliminationo fC O 2 (reaction (3 a)). SubsequentC O 2 elimination from the product is energeticallyf easible, and becomes more important with increasing photon energy (reaction (3 b)). The entropically favored product, Cu II (HCO 2 ) 3 À ,r eaction(3c), requires slightly higherp hoton energies, in line with experiments,a nd becomes dominant at the high-energye nd of the absorption band.
In the UV absorption band in the 3.3-5.5 eV region, Cu II (HCO 2 ) 2 H À is the dominant fragment, followed by Cu I (HCO 2 ) 2 À (reactions(3d)a nd (3 e)). Both reactions require more than 2.4 eV. [31] Products appearing at the beginning of the fragmentation cascade of Cu II 2 (HCO 2 ) 5 À are only observed Figure 3. The total dissociation cross section upon laser irradiation of a) Cu I 2 (HCO 2 ) 3 À ,b )C u II (HCO 2 ) 3 À ,a nd c) Cu II 2 (HCO 2 ) 5 À ,a long with selected partial cross sections of the most importantd issociation pathways. Lines representt he respective running average over five data points.The valuess hown in b) andc)a re enlarged by afactoro f5below 2.75eVfor convenience.
in trace amounts, such as Cu II (HCO 2 ) 3 À and Cu I 2 (HCO 2 ) 3 À .O nly one fragment is observed that is not part of the IRMPD cascade, namely,C u I/II 2 (HCO 2 ) 4 À ,r eaction (3 f), which is comparable in intensity to Cu I (HCO 2 ) 2 À .F ormation of the formyloxyl radical requires an energy of 2.08 eV.N otably,H CO 2 might furtherd ecompose into Ha nd CO 2 ,w ith ar eactione nergy of À0.30 eV and ab arrier of 0.16 eV.A lternatively,aHOCO unit might be formed with ar eactione nergy of À0.36 eV.G iven that the UV absorption band is due to aL MCT transition, elimination of the neutralf ormyloxyl radical seems entirely plausible because the alternatives requirem ore extensive rearrangements. The HCO 2 loss channel after LMCT would also be feasible for Cu II (HCO 2 ) 3 À through reaction (2 e).

Photodynamics
To gain some insight into the photodynamics of Cu I 2 (HCO 2 ) 3 À , we calculated excited-state potential-energy curves along the vector connecting the minimao fg round and first excited states by using internal coordinates. Figure 4a,b and Figure S6 in the Supporting Information show the resultsf or isomers1 and 2, respectively.A fter excitation,b oth isomersr elaxt oa local minimum on the excited-state potential-energy surface. In isomer 1, a3de lectron is promoted to a s(CuÀCu) orbital, which is half-filled in S 1 .D uring geometry optimization,t he two monodentate formate ligands rotate out-of-plane and the CuÀCu distance decreases from 3.3 to 2.3 ,w hich indicates a strong CuÀCu interaction due to the occupation of ab onding orbitali nS 1 .T he resulting CuÀCu distance is close to the bond length of 2.35 in the Cu 2 À ion. [62] The S 1 minimum lies 1.8 eV below the S 1 energy in the FC point. Relaxation through fluorescencei nto the ground state would be substantially redshifted with respect to the excitation, since the S 1 minimum lies only 1.3 eV above the respective ground-statestructure. Consequently, at otal of (hvÀ1.3 eV) would be availablea fter fluorescence as vibrational energy to support the fragmentation cascade leadingt oC u 2 H 3 À ,w hich requires 2.29 eV. [33] Alternatively, the system could reachaconicali ntersection beyond the S 1 local minimuma tw hich the extrapolated ground-a nd excitedstate potential-energy curves nearly coincide, and the entire photon energy would be available for fragmentation on the ground-state potential-energy surface. The photodynamics thus rationalize the formation of Cu I 2 H 3 À ,a sp reviously observed in the IRMPD cascade of Cu I 2 (HCO 2 ) 3 À . [33] No direct mechanism, however,i sc onceivable for the elimination of a coppera tom or copperf ormate from the excited state for reactions (1 b)-(1 d) from isomer 1.
In isomer 2, the target NTO has an antibonding character, with ap ossible nonbonding contribution at the FC point. In the S 1 state, the system relaxest oalocal minimum with a slightly elongated CuÀCu bond ( Figure S6 in the Supporting Information), which does not seem to be relevant for the observed photochemistry.W ithin reach,h owever,t here is ac onical intersection seamt hat is, at most, 0.1 eV above the excitation energy of S 1 in the FC point. This is easily surpassed with the thermal energy contentofthe cluster ion before excitation, in addition to surplus energy from excitation into highers tates in the experiment. Passing throught his conicali ntersection, the system reaches am inimum with pronounced biradical character (see Figure 4b), with an electron transferred between the two copperc enters. The biradical target structure in Figure 4b was obtained as am inimum in triplet spin multiplicity (CCSD/6-31 + g*) because geometry optimization of the relevant singlets tate at the EOM-CCSD/6-31 + g* level was less feasible. The excited state after the conical intersection can thus be described as Cu II (HCO 2 ) 3 Cu 0À ,w ith the excited state manifold separated by only about 0.6 eV from the groundstate surface (EOM-CCSD/6-31 + g*). To reach the biradical minimum, one bridging formate ligand is transferred towards the Cu II center, along with an elongationo ft he CuÀCu bond length to 2.7 .F urthermore, the CuÀOb onds on the bridging formatel igandsa re elongated on the Cu 0 center, providing space for its singly occupied so rbital. In the ground electronic state, the Cu I oxidation state for both copper atoms is maintained.I ti sc onceivable that an eutral Cu 0 atom can be released from this structure at ac ost of 3.18 eV,w ith respect to the FC point. With the remaining energy,t he Cu II (HCO 2 ) 3 À structure can then sequentially decarboxylate for reaction(1b), or even afterwards lose ah ydrogen radicala nd another CO 2 (see reactions (2 b) and (2 c)) to give rise to the Cu II (HCO 2 ) 2 H À , Cu I (HCO 2 ) 2 À ,a nd Cu I (HCO 2 )H À fragments. An internal conversion to the ground state becomes accessible through ac onical intersection close to this local minimum ( Figure 4b), allowing statistical dissociation on the ground-state potential-energy surfaceinr eaction (1a).
Within the photochemistry of Cu II (HCO 2 ) 3 À ,a ll products already appear in the established IRMPD cascade, sharing as equentialr eactionp athway (see reactions (2 a)-(2 c)) or branching from it in smaller amounts (see reactions (2 d) and (2 d')). [31] This meanst hat they are mostl ikely formed in the electronic ground state. It implies that conicali ntersections are available to connect the electronically excited states to the ground state. In Figure 4c and d, we follow relevant excited states along the vector connecting the FC point with D 0 /D 1 and D 4 / D 5 conical intersections, again using the internal coordinates. Excitation into the first excited state, D 1 (Figure 4c), leads the system barrierlessly into aD 0 /D 1 conical intersection, with the bidentate ligand rotatinga roundi ts CÀHa xis. The trajectory is monotonically downhill, and the full excitation energy is converted into vibrational degrees of freedom, which explains the observed fragmentation channels for the d-d transitions.
For the excitation of Cu II (HCO 2 ) 3 À into the lowest lying state of the second absorption band, D 5 ,t he bidentate formate ligand starts dissociating on the way towards aD 4 /D 5 conical intersection, which is reached again without any barrier (Figure 4d). Here, the dissociating formate ligand has ar elatively long CuÀOb ond length of 2.5 ,c ompared with the typical bond lengths of 1.9 and2 .1 for mono-and bidentate formate ligands, respectively,i ng round-state Cu II (HCO 2 ) 3 À .T his behavior can be explained by the charge-transfer character of the LMCT excitation ( Figure 2). After an electron is transferred from HCO 2 À to Cu II ,t he now-neutralH CO 2 unit is only weakly bound to the cluster. In the vicinity of the conical intersection, the structure of the remaining Cu I (HCO 2 ) 2 À unit resembles the most stable isomer of the product ion, Cu I (HCO 2 ) 2 À .A fter passing the D 4 /D 5 conical intersection, the ion might either follow the dissociation pathway and evaporate an eutral HCO 2 radical or eventually reach the ground state, as discussed above. HCO 2 dissociation from Cu II (HCO 2 ) 3 À in reaction (2 e) is in good agreement with the experimentalo bservation of Cu I (HCO 2 ) 2 À , which exhibits the highest branching ratio in the LMCT band in Figure 3b.H owever,H CO 2 loss for Cu II (HCO 2 ) 3 À cannot be distinguished from sequential dissociation of CO 2 and Hi nr eaction (2 b), as observedi nt he IR experiment. [33] On the contrary,t he formation of Cu 2 (HCO 2 ) 4 À from Cu II 2 (HCO 2 ) 5 À in Figure 3c is direct evidencet hat this reaction takes place on the excited-state potential-energy surface, since the fragment is not observed in IRMPD and Sustained off-reso-nance irradiation (SORI) CID experiments of Cu II 2 (HCO 2 ) 5 À . [31] All other fragments were also observed in our recent IRMPD experiments. [31,33] This suggests, once more, that larger Cu II formate clusters can also reach the electronic ground state through conicali ntersectionsa nd redistribute the full photon energy in the system, similar to Cu II (HCO 2 ) 3 À . Due to this rapid internal conversion, the competitive formation of Cu II (HCO 2 ) 3 À and Cu I 2 (HCO 2 ) 3 À from Cu II 2 (HCO 2 ) 5 À ,w ithin the d-d band in Figure 3c,p rovidesf urther insighti nto its ground-state reactivity.D ecarboxylation of Cu II 2 (HCO 2 ) 5 À ,f ollowed by the formation of formic acid, features at ight transition state. The competitive evaporation of neutral copperd iformate Cu II (HCO 2 ) 2 can be expected to be entropically favored. However,f ormic acid formation through reactions (3 a) Figure 4. Excited-state potential-energy surfaces, along with simplified photochemicalreaction coordinates.Interpolation betweent he Franck-Condon (FC) point towards a) al ocal minimum within the first excited state of Cu I 2 (HCO 2 ) 3 À for isomer 1and b) al ocalm inimum of the lowest lying triplet state in isomer 2. Additionally,N TOs from the groundt othe first exciteds tate, along with the corresponding structures in the FC pointand minimum, are shown. Interpolationt owards c) D 0 /D 1 and d) D 4 /D 5 conical intersectionsi nCu II (HCO 2 ) 3 À ,along with the associated structures. The exciteds tates are calculated at the EOM-CCSD/6-31 + g* level with the FC pointoptimized at the B3LYP/def2TZVP level of theory.The triplet minimum in isomer2of Cu I 2 (HCO 2 ) 3 À is optimized by using CCSD/6-31 + g*. The NTOs were calculated at the TD-BMK/aug-cc-pVDZ leveloft heory. and (3 b) requires 0.89 eV,c ompared with 1.36 eV for reaction (3 c), as calculated at the B3LYP/def2TZVPl evel. [31] This explains why the channel producing Cu I 2 (HCO 2 ) 3 À ,a long with its sequential decarboxylation product, Cu I 2 (HCO 2 ) 2 H À ,i so bserved with high selectivity at the start of the absorption band at 1.0 eV,a tw hich the entropically favored evaporation is not yet accessible. Oncee nough energy is introduced into the ion, the evaporation of neutralc opperf ormate becomes the predominant channel at about 1.5 eV.T he production of formic acid can be expected to be highly temperature dependent in catalytic processes, due to the high energy dependence between the two competing reaction channels.
The decomposition trends stay roughlyt he same for larger Cu II formate clusters (see Figures S4 and S5 and Ta ble S4 in the Supporting Information) and depend mainly on the absorption band. For the d-d transitions in Cu II n (HCO 2 ) 2n + 1 À , n > 2, predominantly the evaporation of neutralc opper formatei so bserved, whereas, for n = 1o r2 ,d ecarboxylation happens as the first step, as also observed in IRMPD experiments. [31,33] In the UV band, the dissociation of af ormyloxyl radicalo ccurs for Cu II 3 (HCO 2 ) 7 À .H owever,f or al arge copperf ormate cluster, Cu II 8 (HCO 2 ) 17 À ,n of ormyloxyl radical dissociation is observed. Instead, only evaporation of neutral copper formatec lusters, similarly to IRMPD, [33] is recorded, whichi sc onsistentw ith rapid internal conversion followed by statistical decomposition in the electronic ground state.

Conclusion
The electronic structure of copper formatec lusters has been investigated in the gas phase by using UV/Vis spectroscopy between 0.9 and 5.5 eV.F or copperc enters in oxidation state + I, d-s/p transitions on the copper centers have been observed within the UV region. Copper formate with copperc enters in oxidation state + II exhibit copperd -d transitions in the visible/near-infrared range, along with intense LMCT transitions, which involve the formate ligands and the singly occupied do rbitalo nthe copper centerint he UV region.
The number of copper atoms in Cu II n (HCO 2 ) 2 n + 1 À does not change the observed electronict ransitions, and thus, provides evidencef or the absence of copper-copper bonding, with a singly occupied do rbitala te ach copperc enter.T he orientation of the formate ligands, however,p lays ak ey role in the spectrum, heavily shifting the intensity and energetic position of the transition.
The photochemistry of Cu I 2 (HCO 2 ) 3 À depends heavilyo nt he ground-state structure. Isomer 1, with one bridging formate ligand,m ost likely dissociates statisticallyi nt he electronic ground state, either after fluorescencef rom al ocal minimum in S 1 or by reaching ac onical intersection in the vicinity of the S 1 local minimum. For the paddlewheels tructure of isomer 2, CuÀCu charge transfer initiatest he photochemical loss of a neutralc oppera tom, which competes with statistical dissociation in the electronic ground state. Copperf ormate clusters in oxidation state + II are able to relax ultrafast back to the ground state through conicali ntersections in both excitation bands, and the redistributede nergy is availablef or statistical dissociation. However,i nt he UV region, LMCT makes photochemicalf ormationo ff ormyloxyl radicals possible. For large copperf ormate clusters, no formyloxyl radical loss is observed, which indicates that the photon energy is redistributed into vibrational degrees of freedom.T hrough the efficient energyr edistribution by conical intersections, the selectivity in the formationo ff ormic acid can be tuned via the photon energy in the d-d transition of Cu II 2 (HCO 2 ) 5 À .H igher energiesf avor entropicallyp referred dissociation pathways with loose transition states, such as the evaporation of Cu II (HCO 2 ) 2 .Alarge temperature dependence in the formic acid yield is therefore expected under catalytic conditions.