Gas-phase reactivity of hexanuclear molybdenum [Mo₆I₁₄]²⁻ and its photo-fragments towards O₂ and CO₂: combined mass spectrometry (MSⁿ) and quantum chemical simulations

Abstract

The optical properties, photo-decomposition and photo-reactivity of [Mo₆I₁₄]²⁻ clusters and their associated photo-products are examined in the gas phase. The UV-visible optical spectrum of the electron-rich [Mo₆I₁₄]²⁻ cluster is measured and the wavelength dependence of the principal fragments is interpreted on the basis of the quantum chemistry (QC) simulated spectrum and detailed transitions. A special focus of attention is the photo-generated [Mo₆I₁₄]⁻ and photo-specific radical [Mo₆I₁₃]²⁻˙ as well as halogen-depleted clusters [Mo₆I_n]⁻ with n < 14. The fragmentation and reaction energetics of all species are examined with a combination of mass spectrometry experiments, ion–molecule reactions with O₂ and CO₂ under irradiation and QC simulations. Reactions with O₂ are shown to be very exergonic with up to 5 eV release when O₂ is dissociated on the cluster. Experimental evidence for the photo-decomposition of CO₂ into CO and O on [Mo₆I_n]⁻ ions is provided together with detailed QC energetics.

---

Introduction

Atomically precise metal nanoclusters display unique reactivity that cannot be inferred from either atomic or bulk matter. Their properties have been widely reviewed.¹ Among them, hexanuclear Mo₆ clusters are promising photocatalysts, reported for hydrogen evolution² and CO₂ photo-reduction.³ Their robustness and stability make them sustainable alternatives to dye-based systems. Synthesized at high temperature and stabilized by halogen ligands (X), they form [{Mo₆Xⁱ₈}X^a₆]²⁻ building blocks, where face-capping (Xⁱ) and terminal (X^a) ligands control optical and redox properties.^4,5 Ligand exchange enables integration in diverse matrices⁴ or immobilization onto surfaces.^6–8 [Mo₆X₁₄]²⁻ complexes exhibit broad UV-vis absorption and strong red/NIR emission^5,9 making them attractive, less costly photo-sensitizers and catalysts. However, their precise reaction mechanisms, particularly under irradiation, remain poorly understood.

Mass spectrometry (MS) offers a complementary approach to solution studies by generating, isolating, and probing reactive intermediates at the molecular scale. Optical spectroscopy of mass-selected ions, using both UV-vis¹⁰ and IR,¹¹ provides insight into transient species relevant for reactions in solution¹² in laser-ablation-generated gas-phase clusters. For metal and metal–oxide clusters, this approach has elucidated CO₂ coordination modes and proposed mechanisms for cluster-assisted reduction reactions.^11,13–20

Although Mo₆ nanoclusters are established photocatalysts,^2,3 their gas-phase photo-reactivity is less explored. Warneke et al. investigated electrosprayed [Mo₆X₁₄]²⁻ (X = halogen): photoelectron spectra revealed electron binding energies²¹ while collision-induced dissociation (CID) showed halogen-radical loss and secondary reactivity with O₂.²²

Here, we extend these pioneering studies to the photo-decomposition and photo-reactivity of Mo₆ clusters in the gas phase, focusing on [Mo₆I₁₄]²⁻. Specifically, we ask the following: (i) do laser-induced dissociation (LID) pathways differ from CID pathways? (ii) What is the stability of photofragments? (iii) How do they react with O₂ or CO₂? (iv) How do binding sites, coordination, and radical/electron-acceptor character govern reactivity?

Experiments are complemented by quantum-chemical (QC) simulations to characterize excitations, photofragment structures, stabilities, and reactivity with O₂ and CO₂. Ion–molecule reactions in an ion trap further provide mechanistic insight into these well-defined species.^23–25

Experimental and theoretical methods

Synthesis

Cs₂[{Mo₆Brⁱ₈}Br^a₆] and Cs₂[{Mo₆Iⁱ₈}I^a₆] were synthesized as described in ref. 26. Binary halides Mo₆Br₁₂ and Mo₆I₁₂ were first prepared using solid-state reactions of elemental precursors. Stoichiometric mixtures of CsX and Mo₆X₁₂ were then sealed in quartz ampoules and heated to 850 °C (Br) or 700 °C (I). The resulting Cs₂[{Mo₆Xⁱ₈}X^a₆] complexes are soluble in water and organics, where labile X^a ligands can be exchanged with inorganic or organic groups (L). Cs⁺ counterbalances the cluster charge.

Gas-phase experiments

All multi-stage (MSn) experiments were carried out using a modified a Thermo LTQ Velos ion trap mass spectrometer. [Mo₆I₁₄]²⁻ was generated by electrospray ionization (ESI) of Cs₂Mo₆I₁₄ (1 μM, acetonitrile). CID of [Mo₆I₁₄]²⁻ and product ions was performed in-trap and energy values are reported in instrument specific normalized collisional energy (NCE) units.²⁷ Single isotope peaks were isolated when possible; above m/z 2000, isolation windows included adjacent isotopes due to resolution limits. Assignments for fragment ions in Fig. 1, 6 and 11 are detailed in the SI Tables S1 and S2.

Fig. 1 (top) Photo-fragmentation spectrum (LID) at 260 nm and (bottom) collision-induced dissociation (CID) of a single isotopic peak (m/z 1175.2) of the precursor ion (6,14,0)²⁻ denoted by a star (see the nomenclature section of fragments annotations). Both the oxidized (electron-photo-detached) (6,14,0)¹⁻ mono-anion species at m/z 2350.5 and the (6,13,0)²⁻˙ dianionic species at m/z 1111.5 resulting from the I˙ photo-cleavage are photo-specifically formed. Primary fragments are denoted with thick blue arrows. See SI Table S1 for details on assignments.

Optical spectroscopy

Optical action (fragmentation) spectra were obtained by irradiating mass-selected [Mo₆I₁₄]²⁻ with single OPO laser pulses (200–600 nm, 2 nm steps). Laser power was tuned to keep fragmentation <1% at reference wavelengths (260, 340, and 540 nm). For each point, 200 mass spectra were averaged. Fragmentation ratios were normalized to laser power × wavelength and smoothed to yield fragmentation yields (details in SI Complement S3).

Photo-reactivity and ion–molecule reactions

[Mo₆I₁₄]²⁻ precursors were mass-selected and irradiated (260 nm, 1 mJ, 5 ns pulses) to generate [Mo₆I₁₃]⁻, [Mo₆I₁₃]²⁻˙, or [Mo₆I₁₄]⁻˙, which were isolated and monitored during MS³. Helium buffer gas was seeded with O₂ or CO₂ (5000 ppm; estimated 840/67 and 910/73 collisions per second, respectively; see the SI). Reaction times were varied up to 10 s before MS acquisition. For photochemical studies, [Mo₆I₉]⁻ was trapped in CO₂ and irradiated with a 488 nm CW laser (80 mW).

Simulations

Density functional theory (DFT) calculations were carried out using the Amsterdam Modeling Suite (AMS 2022.105)²⁸ driver with the Amsterdam Density Functional (ADF) engine.^29,30 The revPBE exchange-and-correlation functional^31,32 was employed to optimize all geometrical structures, to calculate electronic, vibrational and electronic excitation properties. Dispersion correction with the electronegativity equilibrium model developed by the Grimme group³³ was added. Due to the presence of heavy elements in the considered herein systems, the scalar relativistic effects have been included by using first-order perturbation theory from a ZORA Hamiltonian.³⁴ The all-electron quadruple-ζ Slater-type orbital basis set with the four polarization function (QZ4P) basis set implemented in the ADF was used. Vertical excitation energies were obtained using Tamm–Dancoff approximation (TDA-DFT).³⁵ Considering that the calculated systems are (doubly)-negatively charged, the conductor-like screening model³⁶ (COSMO) with a dielectric constant of 1.45 and the solvent radius of 2.36 Å (mimicking N₂) was applied to ensure that the energy of the occupied molecular orbitals are negative (see SI Complements S4). Vibrational frequency calculations were performed to ensure that the optimized structures are minima of the potential energy surface. The basis set superposition error was taken into account via the counterpoise correction while calculating the vertical bond dissociation energies.³⁷

Nomenclature/terminology

Here, “cluster” denotes the Mo₆ octahedral unit; together with eight inner halogens it forms the “cluster core.” The full motif, with apical ligands, is the “Mo₆ complex.” Product ions are labeled by triplets (#Mo, #I, and #O) followed by charge. For example, [Mo₆I₁₄]²⁻ and [Mo₆I₁₀O₂]⁻˙ are denoted as (6,14,0)²⁻ and (6,10,2)¹⁻˙.

Results and discussion

QC, CID vs. LID of the precursor [Mo₆I₁₄]²⁻

Primary fragments and photo-specificity. CID of the face-capped (6,14,0)²⁻ ion (single isotopic peak selected at m/z 1175) yields exclusively the singly charged (6,13,0)¹⁻ ion at m/z 2223 (see Fig. 1/bottom) by loss of iodide (I⁻, with mass 126.9 Da). This is consistent with prior CID observations by Warneke and coworkers.²² They additionally observed secondary losses of I˙, although they remarked in their multi-instrument approach that “CID substantially varies among mass spectrometers”. By contrast, the photo-activation (at 260 nm in Fig. 1/top) of (6,14,0)²⁻ yields a much richer variety of ions. Among the photo-product ions, it is possible to identify three primary ions (thick blue arrows in Fig. 1 and Scheme 1): first, the electron detachment product (6,14,0)¹⁻ at m/z 2350 (Scheme 1, eqn (1)); second iodide (I⁻) loss product (6,13,0)¹⁻ at m/z 2223, similar to CID (Scheme 1, eqn (2)); and lastly the radical iodine (I˙) loss product (6,13,0)²⁻˙ at m/z 1112 (Scheme 1, eqn (3)).

Scheme 1 Three independent photofragments are initially generated by irradiation (blue arrows): loss of an electron (eqn (1)), loss of iodide (I⁻, eqn (2)) and loss of iodine (I˙, eqn (3)). Clusters depleted in iodine ((6,12,0)¹⁻˙ to (6,7,0)¹⁻) are also observed and may arise as secondary fragments from further losses of I₂ (black arrows) or I˙ (red arrows, attested by CID, see Fig. S1 and S2). Loss of I⁻ is also observed in CID of (6,13,0)²⁻˙ (red dotted arrow, see Fig. S3).

The precursor (6,14,0)²⁻ was first investigated via QC calculations with an initial geometry optimization step following the procedure described in the Computational details section. The molecular orbital (MO) energy diagram of (6,14,0)²⁻ is given in Fig. 2 together with the (spin)orbital energy diagram of the primary photo-fragments (6,14,0)¹⁻˙, (6,13,0)¹⁻ and (6,13,0)²⁻˙. The frontier MOs are plotted in Fig. S4. The (6,14,0)¹⁻˙ radical moiety arises from the depopulation of one electron of the highest occupied molecular orbital (HOMO) which is mainly iodide in character (61%). The cluster core and the Mo–I^a bonds are moderately affected by this one-electron oxidation: only a slight elongation of one apex of the Mo₆ octahedra from 2.678 Å to 2.733 Å Mo^apex–Mo^basal distance, together with a slight shortening of the Mo–I^a distance from 2.881 Å to 2.834 Å. The resulting spin-density is spread over the Mo₆ cluster (Fig. S5), forming an oxidized metal Mo₆ cluster bearing a formal +11 charge, thus Mo^1.83+. In contrast, the iodine loss (−I˙) leads to (6,13,0)²⁻˙ where most of the unpaired electron density (0.9) is localized on the iodine-free apical site and the spin density forms a lobe pointing toward the iodine vacancy (Fig. S5). In other words, simulations show that (6,13,0)²⁻˙ is a radical species with the radical localized on the lacunar molybdenum, leaving unaffected the oxidation state of the metal Mo₆ cluster. Lastly, the electronic structure of the fragment (6,13,0)¹⁻ resulting from iodide loss (−I⁻) is characterized by the presence of an electron accepting MO (Fig. S5).

Fig. 2 (Spin) Orbital molecular energy diagram of [Mo₆I₁₄]²⁻, [Mo₆I₁₄]¹⁻˙, [Mo₆I₁₃]²⁻˙, and [Mo₆I₁₃]⁻. For each system the molecular orbital or spin-orbital α (HOSO_α) with the same spatial extension and sign as the highest occupied molecular orbitals (HOMO) of the [Mo₆I₁₄]²⁻, were energetically aligned with it. LUMO = lowest unoccupied molecular orbital. LUSO = lowest unoccupied spin-orbital (α or β). Isocontours = ±0.02 (e Bohr³)^½. The plots of all frontier (spin) orbitals are given in Fig. S4.

From an energetics point of view, the non-adiabatic I˙⋯[Mo₆I₁₃]²⁻˙ iodine-cluster homolytic bond dissociation energy in the ground state was calculated at 3.547 eV while the I⁻⋯[Mo₆I₁₃]¹⁻ heterolytic apical bond dissociation energy was calculated at 0.819 eV. For the sake of comparison, the bromine and bromide bond dissociation energies in [Mo₆Br₁₄]²⁻ were also calculated and are respectively 4.276 eV and 2.523 eV. Thus, the iodinated cluster [Mo₆I₁₄]²⁻ is more susceptible than the brominated cluster [Mo₆Br₁₄]²⁻ (+0.729 eV) to yield the species (6,13,0)²⁻˙ by homolytic halogen bond cleavage. This is consistent with the non-observation of (6,13,0)²⁻˙ among the LID fragments of [Mo₆Br₁₄]²⁻ (see Fig. S6). However, in both iodinated and brominated cluster simulations, the heterolytic bond cleavage is very significantly favoured with regards to the homolytic cleavage (+2.728 eV and +1.753 eV, respectively). This is consistent with the experimental evidence that the loss of halogenide (I⁻/Br⁻) represents the major fragmentation feature observed in the gas phase in both CID and LID mass spectra. The calculated ionization potential is +2.755 eV, thus 0.792 eV more accessible than the homolytic cleavage I˙⋯[Mo₆I₁₃]²⁻˙. The QC computed energetics is thus in line with the absence of primary fragments (6,14,0)¹⁻˙ and (6,13,0)²⁻˙ following collisional activation (CID). It suggests that both electron loss (Scheme 1, eqn (1)) and iodine loss (Scheme 1, eqn (3)) are not thermally accessible from the ground state. In contrast, their observation following laser activation (LID) suggests that both phenomena may arise from processes involving electronic excited states.

Optical spectroscopy of [Mo₆I₁₄]²⁻ and wavelength dependence of the primary fragments

Fig. 3 displays the UV-vis optical spectrum of the precursor (6,14,0)²⁻ measured in the gas phase by “action” photo-fragmentation spectroscopy, after mass selection and isolation in the ion trap, overlaid with the simulated spectrum (TDA-DFT calculations, see Computational details). Transitions towards the three first quasi-degenerated excited states (S1, S2 and S3) are calculated at 580 nm in a vacuum and the onset of fragmentation in the visible range is measured accordingly above 560 nm. However, the density of states and transitions increases very significantly around 500 nm which gives rise to two bands respectively at 480 nm and 440 nm in the simulated spectrum convoluted with 20 nm Gaussian profiles (thin black dotted line in Fig. 3). The experimental spectrum accordingly presents two large, partially unresolved fragmentation bands in the visible range with maxima between 420–440 nm and 470–500 nm. In the UV range, one large band appears between 310 and 350 nm, which corresponds to a 20 nm blue shift with regards to the calculated UV band at 350 nm. Lower in the UV (higher photon energies), one major feature appears on the photo-fragmentation spectrum with a maximum around 270 nm, which matches the major feature from simulations. Overall, the good agreement between experiments and theory enables the use of simulations for a more in-depth interpretation of the fragmentation behaviour.

Fig. 3 Optical action (fragmentation) spectrum of [Mo₆I₁₄]²⁻ in the UV-visible range. The photo-fragmentation was recorded independently for three regions: <290 nm (blue), [290–390] nm (purple) and >410 nm (green). Laser irradiance and alignment conditions vary between regions so that their absolute intensities should not be compared. Thus, experimental data was scaled to ease the comparison with the simulated spectrum. Data were recorded with a 2 nm pitch. Circles on the experimental curves represent the position of the laser alignment for each spectral range. The simulated spectrum (thin black line) is calculated from 0.12 eV Gaussian convolution of the oscillator strength for each transition (red bars). A factor ×15 is applied to simulated data in the visible range (dotted line).

The specific wavelength dependence of the primary photo-fragments is examined in Fig. 4 and allows several observations. First, the major feature is the clear predominance of the iodide loss over all other primary fragments. This fragment is the only one observed by collisional heating (CID), which retrospectively denotes for LID the importance of internal conversion as a relaxation path from the excited state towards the ground state. It indicates that a large part of the fragmentation might arise from sequential multiple photons’ absorption cycle and fragmentation in the ground state.³⁸

Fig. 4 Distribution of the three primary fragments relative intensities as a function of the wavelength. Boxplots represent the 1st and 3rd quartiles of the branching ratios corrected for the laser power profile for the loss of I⁻ (red) and loss of I˙ (blue) on the left axis; and for the e⁻ loss (purple) on the right axis. Thick lines are smoothed medians to guide the eye.

The second observation is that among the two bands in the visible region, the loss of iodine is only observed in the red-most band at 470–500 nm. Interestingly, natural transition orbital³⁹ (NTO) analysis of the most intense calculated transitions allows interpreting this asymmetry in terms of the nature of the excited states (see Fig. 5). The strongest transitions in the red-most band correspond to the excited states S₂₉ at 495 nm (with oscillator strength f = 0.023) and S₅₀ at 484 nm (with f = 0.020). The NTO analysis shows that only 20% of S₂₉ is described by a transition from the Mo–I^a Sigma bonds to non-bonding NTOs, which corresponds to a partial weakening of the metal–ligand bond. On the other hand, S₅₀ hardly affects the metal–ligand bonds but involves a major (72%) character of iodine electron density transfer from apical to inner sites. Thus, transition to S₅₀ can be described as the oxidation of the apical iodines (with strong ionic character in the ground state compared to Mo–Xⁱ bonds with strong covalent character) which is susceptible to give rise to radical iodine loss, as observed experimentally. Interestingly, the strongest transition in the second band around 440 nm corresponds to excited state S₉₂ at 439 nm (oscillator strength, f = 0.017). NTO analysis of S₉₂ shows a major loss of Mo–I^a π/d-antibonding character partially transferred to Mo–I^aσ/d-antibonding levels, but globally the iodine electron density is maintained. This band is thus, as observed experimentally, more susceptible to evolve through prompt dissociation of iodide (I⁻) compared to the bands at lower energy examined previously.

Fig. 5 Electron/hole natural transition orbital representations of the three lowest degenerated excited states, and of the excited states of [Mo₆I₁₄]²⁻ with oscillator strengths f > 0.007 (only one of the three degenerated states is represented in each case).

The third observation is that electron detachment leading to (6,14,0)¹⁻˙ is only observed below 320 nm. This is consistent with prior photoelectron spectroscopy experiments by Warneke et al.²¹ who reported adiabatic/vertical electron binding energies of 2.2–2.4 eV (560–515 nm) and repulsive Coulomb barriers of 1.6 eV, leading to a minimum of 3.8–4.0 eV (i.e. one photon of 310–340 nm) required to efficiently photo-detach an electron from (6,14,0)²⁻. Since the Mo₆ clusters are rigid and no major rearrangement occurs in the ionized state (hence the very small differences between adiabatic/vertical ionization energies), no major effect of the solution phase environment is expected on their ionization properties. As a result, the photo-detached species (either the iodinated but even more the brominated species which are not even formed in the gas phase) are unlikely to be formed in solutions irradiated by visible light and are, thus, unlikely to explain the reactivity observed in solution.³ Thus, it is likely that this photo-oxidized species is not involved as a reactive intermediate in the photo-catalytic cycle. However, since the reported electron binding energy is relatively low, within reach of a photon in the visible range, it is very likely that (6,14,0)²⁻ behaves as a visible-light activated photo-reducer in the excited state. Alternatively, since both iodide loss (−I⁻) and iodine loss (−I˙) are observed in the visible range, they might play a role as photo-prepared reactive intermediates and their respective reactivity should be investigated.

The last observation is that below 300 nm the electron, iodide and radical iodine losses present similar band shapes, pointing towards a low spectral selectivity in the UV. However, even if band shapes are similar, iodide loss dominates by a factor of 2 over the photo-specific radical iodine loss, which in turn dominates the electron photo-detachment by a factor of 20.

In summary, a photo-reducer excited state Mo₆ cluster might be formed with low-energy photons (≥560 nm) with a reduced propensity for fragmentation. The radical (6,13,0)²⁻˙ is optimally formed (relative to other fragments) around 500 nm. The oxidized species (6,14,0)¹⁻˙ is only formed below 320 nm and is never a major fragment. Overall, the path with I⁻ loss systematically dominates other fragmentation paths.

Secondary fragments (6,n,0)¹⁻

In addition to the three primary photo-fragment ions discussed above, several ion series are observed in the LID mass spectrum (see Fig. 6 and Scheme 1). A full description is given in SI Complements S2 with complete QC and MSⁿ experiments supporting assignments and discussions on the dissociation mechanisms. In brief, the first series corresponds to iodine depleted cluster complexes from (6,12,0)¹⁻ to (6,7,0)¹⁻. They may arise as secondary fragments from photo-specific sequential losses of I₂ (Scheme 1, black arrows) or from consecutive I˙ losses (Scheme 1, red arrows). In general, and although QC geometry optimization indicates that all I˙ loss reactions are largely endergonic (around 3 eV, see Fig. 7), iodine loss would be consistent with prior mass spectrometry results by Warneke et al.²² which revealed that upon collisional activation and after the initial loss of iodide I⁻, further losses of halogen radicals represent the main fragmentation channels.

Fig. 6 Photo-fragmentation spectrum (LID) at 260 nm of a single isotopic peak (m/z 1175.2) of the precursor ion (6,14,0)²⁻ denoted by a star (zoomed in LID spectrum shown in Fig. 1). The focus is made on secondary fragment ion series: (6,n,0)¹⁻ iodine-depleted ions (grouped in I₂ loss series), and (6,n,2k)¹⁻ reaction products with one, two and three O₂ molecules. Example of −I₂/+O₂/−I˙ inter-series connectivity is shown with curved black arrows. See SI Table S1 for details of assignments.

Fig. 7 Difference of Gibbs free energy at 298 K of sequential iodine loss and O₂ addition reactions starting from the most stable isomer of (6,12,0)¹⁻. These energies (in eV) correspond to the reaction leading to the most stable possible isomer (S = singlet state; more details in the SI). The red star indicates the Mo centre close to which the O₂ molecule was positioned to generate the (6,n,2)¹⁻ initial geometry. The dotted circles show which iodine atom was removed to yield the most stable (6,n − 1,0)¹⁻ and (6,n − 1,2)¹⁻ isomers. See SI Table S4 for all reaction energy changes.

In any case, whether it results from I₂ or from I˙ losses, the formation of such secondary fragments is energy demanding and requires excess energy in the primary fragments.

Interestingly, Warneke and coworkers have shown with ion mobility and computational simulations that the octahedral Mo₆ structure remains largely unchanged in the iodine-depleted Mo₆ complexes.²² This is consistent with prior ion mobility mass spectrometry results in our group which showed that the Mo₆ cluster size and shape are retained irrespective of the chemical nature of the ligands (halogenides or hydroxides),⁴⁰ which advocates for a very robust Mo₆ octahedral cluster structure. It is also what we calculate for the series (6,n,0)¹⁻ for n ranging from 14 to 5. It should nevertheless be emphasized that the symmetry of the Mo₆ core is importantly decreased. Indeed, in order to compensate for the electronic depletion due to the iodine departure, the metal–metal bonds involving the more I-depleted Mo atoms are importantly shortened: down to 2.391 Å in (6,5,0)¹⁻, compared to 2.678 Å in the parent (6,14,0)²⁻ cluster. The shorter Mo–Mo distance in I-depleted sites is in the order of magnitude of reported triple Mo–Mo bonds (while close to bond order 1 in the parent cluster).⁴¹

Reactivity of the photofragments with O₂

Another series of ions observed on the LID mass spectra corresponds to the addition of one or multiple O₂ on the complexes. While oxygenation products are absent for both (6,14,0)¹⁻˙ and (6,13,0)¹⁻ even at long reaction times (300 ms in Fig. S7 and S8), they are observed for (6,12,0)¹⁻˙ and further iodine-depleted complexes. Noteworthy, there are only trace levels of O₂ in the ion trap. However, the 10 Hz laser repetition rate together with the asynchronous mass spectrometer behaviour enforces on average a significant delay between the irradiation of selected precursors and the detection of the fragment ions. In our single laser pulse experiments, photo-product ions experience an average 60 ms reaction time post-irradiation, which is enough for several collisions to occur between ions and neutrals, even at trace levels below 1‰ in the background gas at the trap pressure (5 mTorr (ref. 42)). While the formation of H₂O adducts is not common but not atypical in mass spectrometry (since most of the ESI experiments use at least partially aqueous solvent mixtures and the addition of water often corresponds to an early sample solvation stage with stabilizing H-bonds possible with most organic samples), the addition of O₂ is on the contrary very unusual under ion trap conditions where very low oxygen trace levels are expected. Thus, the appearance of such O₂ adducts (see the orange series in Fig. 6 and peak list in Table S1) reveals a specific affinity between O₂ and the Mo₆ cluster-based complexes. Further series of ions displaying up to 3 O₂ additions were observed (red and green series in Fig. 6).

Globally, the greater the depletion in iodine ligands, the more oxygenated the clusters become. Fig. 8 shows that the single O₂ addition onset appears after the loss of at least 2 iodine atoms from the Mo₆ complex (blue trace), but becomes systematic and the major cluster components after 4 iodine losses. Similar observations and onset of reactivity with O₂ were also reported by Warneke and coworkers²² under similar ion trap conditions.

Fig. 8 Evolution, as a function of the number of I removed from the precursor, of the complex oxygenation ratio (continuous line) calculated as the sum of all (6,k,2n)¹⁻ ion intensities (n ≥ 1) normalized by the sum of all (6,k,2n)¹⁻ ion intensities (n ≥ 0). The dashed red line corresponds to the ratio of multiple oxygenation (6,k,2n + 2)¹⁻ (n ≥ 1) normalized by (6,k,2n)¹⁻ (n ≥ 0). Data are averaged from the 4 experimental fragmentation ranges in the UV and visible ranges.

Lastly, the third primary photo-product ion (6,13,0)²⁻˙ was also observed to react with O₂, with (6,13,2)²⁻˙ becoming the major product ion after 250 ms reaction time (Fig. S9). Although both (6,13,0)²⁻˙ and (6,13,0)¹⁻ display one non-coordinated site, no sign of reactivity with O₂ is observed for the singly charged (6,13,0)¹⁻ (see Fig. 8 and Fig. S10). This is consistent with computations which reveal that the O₂ addition to (6,13,0)¹⁻ is not significantly favoured: only very slightly exergonic towards the most stable geometry and spin state of (6,13,2)¹⁻ (triplet state, ΔG_298K = −0.023 eV) and even weakly endergonic towards its most stable singlet state (ΔG_298K = 0.082 eV). Interestingly, our experimental results reveal that the O₂ affinity of the dianion (6,13,0)²⁻˙ is comparable to that of (6,12,0)¹⁻˙, which has two vacant sites with radical character: ∼10% O₂ adduct formation under the same experimental conditions, Fig. 9. This is also consistent with QC calculations of ΔG_298K for the O₂ addition reactions which are of the same order of magnitude, respectively −0.949 eV and −1.221 eV for (6,12,0)¹⁻˙ and (6,13,0)²⁻˙.

Fig. 9 Evolution of the precursor ion [Mo₆I₁₃]²⁻˙ survival yield after variable reaction times in the ion trap filled with various background gas compositions: He with unavoidable traces of O₂, 0.5% O₂ seeded in helium, 0.5% CO₂ seeded in helium. Boxplot boundaries represent the first and 3 quartiles; thick continuous and thin dashed red lines show the respective positions of the median and the mean for each reaction time dataset. Notches give impressions of the uncertainty windows estimated via a bootstrap approach. The dotted line corresponds to the 1/e level relevant for mono-exponential decays.

Interestingly, the addition of a second O₂ on the complexes is shifted by 2 iodine losses in Fig. 8 (red curve), which suggests that, beyond the onset, binding of O₂ may involve 2 coordination sites. The O₂ coordination modes on the various (6,n,0)¹⁻ parents and their relative energetics were examined closely by QC calculations. The (6,n − 1,0)¹⁻ parent isomers considered in the present work (n = 13–7) were generated by sequential single I˙ loss of the most stable (6,n,0)¹⁻ isomer starting from (6,13,0)¹⁻, considering all apical and inner positions. The reactivity of these (6,n,0)¹⁻ ions with O₂ was evaluated for all possible attack positions of the O₂ molecule identified on the basis of the spatial extension of the lowest unoccupied (spin)orbitals (LUMOs/LUSOs) and of the spin density (if applicable).

Warneke and collaborators reported QC calculations on the reaction of O₂ with (6,n,0)¹⁻ for n = 12 and the subsequent I˙ loss leading to (6,11,2)¹⁻.²² Our calculations on these specific systems are in full agreement with their results which show that O₂ binds one of the opposite vacant apical sites in η¹ coordination mode (via one oxygen atom) in the most stable (6,12,2)¹⁻ isomer. In the present work, we extend the study (6,n,0)¹⁻ with n = 11–7, testing all active Mo coordination sites. The geometry optimizations reveal that many O₂ coordination modes involving one or multiple Mo binding sites exist, leading to inner, face-capping or edge bridging modes involving one or the two oxygen atoms. Interestingly, the most stable (6,11,0)¹⁻ isomer possesses one vacant apical site and one adjacent inner vacant site. The most stable optimized (6,11,2)¹⁻ isomer has a structure in which O₂ is fully dissociated with one of the oxygen atoms in an apical position and the second oxygen atom in a μ₃ bridging, inner position (Fig. 7). The optimized (6,9,0)¹⁻ structure which has 4 free apical sites and one inner site available for coordination gives rise to the same O₂ full dissociation. For (6,10,0)¹⁻ and (6,8,0)¹⁻, the most stable structures possess respectively 3 and 5 vacant apical sites and one vacant inner site and the most stable O₂ adducts display η²O₂ coordination to an apical position. Lastly, the optimized (6,7,0)¹⁻ contains 5 vacant apical sites and 2 vacant inner sites and the optimized O₂ adduct yields full O₂ dissociation with one O in the apical position and the other oxygen atom in the μ₂ edge-bridging position. Note that the optimized O₂ adduct geometries, for all (6,n,2)¹⁻ species, result from the geometry optimization of systems including an O₂ molecule positioned close to reactive Mo sites on the (6,n,0)¹⁻ isomers.

From the energetics point of view, QC simulations show that the coordination of O₂ is systematically exergonic for all clusters with n ≤ 12, irrespective of the O₂ coordination mode on (6,n,0)¹⁻ (see energies of all calculated systems in Table S3 and reaction energy changes in Table S4). However, the extent of exergonicity strongly depends on the coordination mode of O₂ on the cluster. Typically, η¹O₂ releases less than 1 eV while η²O₂ releases 1.5–1.7 eV and O₂ dissociation releases considerably more energy: 4.8 to 5.6 eV. Similarly, QC calculations show that the magnetic character (non-zero spin value) of the parent iodine cluster is not a prerequisite to the O₂ addition. However, the spin state of the most stable parent cluster seems to affect the coordination mode of O₂, in particular its dissociation fate, and therefore the exergonicity of the reaction between the (6,n,0)¹⁻ clusters and O₂.

Interestingly, as soon as O₂ is coordinated more strongly than η¹ (i.e. for (6,n,2)¹⁻ with n ≤ 11), the exergonicity of O₂ additions may be large enough to induce further fragmentation, especially in a dilute gas phase environment where internal energy cannot diffuse away via collisions with the solvent. In particular, the large exergonicity may compensate for the global endergonicity of I˙ losses and yield another I˙ loss (see Fig. 7). Thus, O₂ binding (6,11,0)¹⁻ may favorably yield the (6,10,2)¹⁻ species by I˙ loss (2.678–4.828 = −2.150 eV) whereas O₂ binding (6,12,0)¹⁻ may not decay into (6,11,2)¹⁻ (2.839–0.949 = +1.890 eV). These QC calculations are consistent with experiments in both the current work and that of Warneke et al.²² where the O₂ addition onto (6,n,0)¹⁻ is generally followed by an I˙ loss for n ≤ 11. However, it is interesting to observe that the most stable structure yielded after O₂ addition followed by I˙ loss may be different from the structure resulting from I˙ loss followed by O₂ addition: the coordination of O₂ affects the preferential iodine departing site, which limits the O₂ coordination rearrangement in the final structure (see for instance (6,11,2)¹⁻ coming from (6,11,0)¹⁻ or (6,12,2)¹⁻ in Fig. 7).

According to the calculations, the onset of reaction with O₂ in Fig. 8 can be associated with the radical character of available binding sites on the cluster complex where O₂ may bind as μ₀η¹. This property is applicable for (6,n,0)¹⁻ species with n = 12, 10, 8 (which are, according to simulations, doublets with strong radical character) as well as for (6,13,0)²⁻˙. This is consistent with the increased reaction rate of O₂ with open-shell transition metal clusters described in the literature (see the very complete review by Fielecke¹¹), and has been explained by their lower electron binding energies correlated with a facilitated electron transfer towards O₂ to yield the superoxo η¹O₂ structure. Then, the sudden step in oxidation ratios measured after 4 iodine losses likely results from the increased O₂ affinity for the cluster with increased binding strengths permitted by the appearance of neighboring free sites that allow both η²O₂ binding (peroxo structures have also been described by Fielecke¹¹) but also, more importantly, O₂ dissociation that yields oxo structures (see Fig. 7). In addition to this increased affinity, the large exergonicity associated with O₂ dissociation is responsible for the (6,11,2)¹⁻ oxygenated cluster dissociation into (6,10,2)¹⁻, which explains the unexpectedly low and high apparent oxidation ratios measured for n = 11 and 10 in Fig. 8. More generally, the exergonic character of the η²O₂ binding, and even more markedly the exergonicity of the O₂ dissociation on the cluster, may be at the foundation of the extended dehalogenation and advanced oxygenation of the Mo₆ cluster complexes as observed experimentally in the present work (see Fig. 10 bottom panel). The structure and energetics of the fully oxygenated species extend beyond the scope of the present paper and will be described in more detail in a future communication.

Fig. 10 Ion–molecule reactions under irradiation (photo-IMR-MS²) spectra of precursor species (6,9,0)¹⁻. No collisional energy is applied but precursors are irradiated with a CW laser at 488 nm (80 mW) during the whole experiment, from the ion accumulation/injection time, through the 200 ms reaction time with the background gas and up to ion detection. Background gas composition: helium with O₂ traces (bottom) and He seeded with 0.5% CO₂ (top). (6,n,2k)¹⁻ adducts formed by reaction with O₂ are marked in black while (6,n,2k + 1)¹⁻ adducts specifically formed in the presence of CO₂ are marked in red. See SI Table S2 for details on assignments.

Interestingly, despite the large exergonic character of η²O₂ or O₂ dissociative binding, no fragmentation of the cluster core is observed experimentally. Actually, cluster core fragmentation was only observed after reaction of the dianionic (6,13,0)²⁻˙ with O₂ (Fig. S3 with CID and S9 without CID but 250 ms IMR with O₂ traces). In both cases, aside from a major peak corresponding to (6,13,2)²⁻˙ and the associated iodine loss product (6,12,2)²⁻, two major fragment ions are detected with complementary masses summing up to the oxygenated precursor ion and with isotopic patterns typical for respectively [MoI₂O₂]⁻ and [Mo₅I₁₁]⁻ (especially visible in Fig. S9). This points towards the Mo₆ core fragmentation from the (6,13,2)²⁻˙ species, i.e. after addition of O₂ on the precursor radical dianion. Additional cluster core fragments are observed at m/z 1652 and 1525 (Fig. S9 and S11) and tentatively assigned to (4,10,0)¹⁻ and (4,9,0)¹⁻, respectively. As a confirmation, longer reaction times (250 ms, Fig. S9) yield higher [MoI₂O₂]⁻ peak intensity and further core fragmentation than shorter reaction times (100 ms, Fig. S11 top panels). Also, when the trapping atmosphere is seeded with 0.5% O₂ (Fig. S11 middle panels), all reactions immediately shift towards oxygenated products (in particular, species (6,13,2)²⁻˙ becomes the most abundant dianion within less than 100 ms) and all ensuing secondary products massively increase, in particular the Mo₆ core degradation paths. Interestingly, MoI₂O₂ has been reported as neutral in the gas phase,⁴³ with the Mo^VI oxidation state. The observed species [MoI₂O₂]⁻ was never reported individually: oxidation state Mo^V is well known but predominantly exists under the form of metal dimers sharing e.g. an oxo-bridge.^44,45 The Mo^V is nevertheless a stable oxidation state, thus fragment [MoI₂O₂]⁻ is not unexpected. However, the [MoI₂O₂]⁻ fragment is only formed from the di-anionic, oxygenated (6,13,2)²⁻˙. The cluster complex fragmentation was never observed previously for the iodinated Mo₆X₁₄ mono- or di-anions, and was observed only for higher collision energies (between 80 and 150 V) for the chlorinated clusters.²² These elements point towards the specific instability of the Mo₆ core in (6,13,2)²⁻˙. This is consistent with QC calculations which show that the coordination of O₂ leads to the destabilization by ∼0.2 eV of the HOMO in (6,13,2)²˙. In addition, the analysis of the simulated geometry of (6,13,2)²⁻˙ (see Fig. S12) shows that O₂ binds at the apical free site in η² mode while an inner face-capping iodine decoordinates from the Mo which binds O₂ (Mo^O₂) to become edge bridging on the opposite 2 Mo of its original face. This is accompanied by the Mo^O₂–Mo bond weakening (elongation up to 2.828 Å) together with the elongation of the Mo–I distances.

Reactivity of the photofragments with CO₂

Primary photo-fragments and comparison between O₂ and CO₂. In view of the photo-catalytic character of the Mo₆ cluster complexes, the reactivity of the photo-specific fragments was investigated with CO₂. At 0.5% seeding percentage, the collision rate of CO₂ with the precursor dianions/monoanions is calculated as ∼910/455 collisions per second in the HPT (see the Methods section and more details in the SI, section Complements S1). Neither (6,14,0)¹⁻˙ nor (6,13,0)¹⁻ displays any sign of reaction. In the case of (6,13,0)²⁻˙, all three previously identified reaction paths are maintained: first, formation of the oxidized (6,13,2)²⁻˙ species, and subsequently loss of I˙/I⁻ or Mo₆ core dissociation (see Fig. S11, bottom panels). The relative intensities of these reaction products are nevertheless much lower than in He:O₂. For instance, after 100 ms reaction time, the precursor ion (6,13,0)²⁻˙ is still the major species detected in He:CO₂ (∼60% of the total ion signal on the full m/z range [305–2500]) while it has nearly completely reacted in He:O₂ (∼0.4% of the signal). This observation confirms that these paths are bound to the initial reaction with O₂ (the concentration of which falls back to trace levels in He:CO₂). The absence of any specific and new reaction product ion in the mass spectrum after the introduction of CO₂ discards the possibility of a direct chemical route between the precursor and CO₂. Since CO₂ is not itself a radical, the radical character (6,13,0)²⁻˙ must be set aside and it can only be concluded that it is not a potent enough reducing agent to directly reduce the very stable CO₂.

However, unexpectedly, the O₂-related fragments’ intensities rise a little compared to unseeded helium, which is reflected by a significantly lower survival yield in CO₂ than helium as reaction time increases. Fig. 9 displays the precursor relative intensity as a function of trapping time in helium with traces of O₂ (blue) vs. 0.5% CO₂ (green) or 0.5% O₂ (red). Under the rough but reasonable approximation of a first-order kinetics reaction of the precursor with O₂ (maintained at a constant concentration in the ion trap), the precursor survival yield ought to be modelled with a mono-exponential decay function. Thus, the reaction time constant can be approximated by the reaction time at which the precursor level is down to 1/e (Fig. 9 dotted line). In terms of reaction rate, the estimations evolve from 2.5 s⁻¹ (400 ms) to 5 s⁻¹ (200 ms) from unseeded helium with traces of O₂ to He:CO₂. Thus, the precursor disappears nearly twice as fast with 0.5% CO₂ than with only O₂ traces. This enhanced apparent reactivity of the precursor under CO₂ atmosphere, although with exactly identical fragments as with O₂, likely corresponds to a mere collisional activation effect stemming from the high instability of the precursor ion (very small activation energy required to induce reactions, see the above section, “Secondary fragments (6,n,0)¹⁻”) combined with the frequent collisions (910 s⁻¹) and higher collisional energy between the precursor ions and CO₂ (1 order of magnitude higher in reduced mass compared with helium). This combination is likely to mechanically increase the internal temperature of the precursor ions and result in higher reaction rates.

Remarkably, the reaction rate in 0.5% O₂ can be estimated to be >200 s⁻¹ (< 5 ms), i.e. ≥80 times faster than in pure helium, with only traces of O₂. Considering a constant reaction rate per collision, the trace level of O₂ in the ion trap with unseeded helium can thus be estimated to be less than 0.5%/80, i.e. ∼0.005%. Since the collision rate between (6,13,0)²⁻˙ and O₂ in He:O₂ is ∼840 s⁻¹, the reaction probability per collision with O₂ can be calculated and its lower boundary amounts to 0.24, which is considerable (1 of 4 collisions at least is reactive) and definitely points at O₂ additions as a major limitation to the radical photofragment (6,13,0)²⁻˙ lifetime.

Case of the reactivity of highly dehalogenated cluster complexes. Beyond the case of the primary photofragments, the reactivity of under-coordinated Mo₆ cluster-based complexes was also assessed. First, the major observation is that throughout experiments with CO₂, the major ions are identical to what is observed in helium with traces of O₂: (6,n,2k)¹⁻ fragments with variable numbers of halogen and dioxygen adducts (Fig. S13 and S14 top). In other words, no major effect is observed due to the presence of CO₂ in the trapping gas. However, two significant observations are made among lower intensity peaks: first, several peaks are detected with unusually large width and asymmetrical shape and, second, several ions are detected with an odd number of oxygen atoms. Both observations, detailed below, are specific to the presence of CO₂.

It is striking that in the presence of CO₂, no CO₂ adduct is apparently formed with cluster complexes displaying multiple binding sites available. However, several peaks are detected with unusually large width and asymmetrical shape at ∼42 Th above (6,n,k)¹⁻ species. For instance, in the CID spectrum of (6,5,6)¹⁻ (Fig. S14), such deformed peaks are detected at m/z 999.8 and 1110.1, 42 Th above ions (6,3,7)¹⁻ and (6,2,8)¹⁻. The width and shape of these peaks recall peak fronting, which has been reported in ion traps, specifically for fragile ions during their ejection towards the detectors.^46,47 During resonant ejection, collisions with the buffer gas may energize the ions: in the case of fragile ions, part of the ions can fragment before ions reach the detector, which results in deformed peaks with inaccurate mass assignments. By analogy, this suggests that in our experiment the ions detected with such profiles are very fragile/metastable. In addition, the relative amplitude of the fronting and the global mass assignment of these fragile ions depend on the survival of the precursor ions during the ion detection which depends on the scan speed. Yost and co-workers have reported up to 1.09 Th shifts towards low m/z for precursor anions with mass 314 Da.⁴⁷ Since the ions detected in the present study are significantly higher in m/z (towards 1000 Th), and the mass resolution of the spectra is significantly higher than in Yost and co-workers’ study, it is likely that the scan speed is also significantly slower. Under these conditions, it is expected that the fragile ions’ mass shift is higher than that reported previously. Thus, the actual m/z value of the very fragile ions detected in the present study is probably significantly more than 1.09 Th above their measured position. In other words, the hypothesis of short-lived, metastable CO₂ adducts (+44 Da) that do not survive detection is very plausible despite the inaccurate mass difference. Despite the fact that these potential CO₂ adducts have only been observed from oxygenated cluster (Fig. S14) which would suggest the η¹- or η²-carbonato coordination mode of CO₂ on a metal-oxo site,¹¹ peak fronting is most probably a sign of weakly physisorbed CO₂ with only little or no electron transfer from the cluster.

Alternatively, CO₂ might bind to the cluster and dissociate rapidly into CO and O and yield, after neutral CO loss, a product anion with an odd-number of oxygen atoms. Such CO₂ activation with CO loss has been observed previously on anionic pure metal clusters,¹¹ but also on highly reduced transition metal oxide clusters including molybdenum oxide clusters¹⁶ as well as on niobium clusters^17,19,20 and metal carbides.¹⁸ Interestingly, while O₂ addition only yields species with an even number of oxygen atoms, the dissociation of CO₂ might yield species with an odd number of oxygen atoms. In the present study, Mo₆ species with odd numbers of oxygen atoms were observed exclusively in the presence of CO₂ in the trap. Thus, (6,7,5)¹⁻ is observed readily after in-source CID of the precursor (6,14,0)²⁻ (Fig. S13) while (6,2,9)¹⁻ and (6,3,7)¹⁻ are formed after CID of the mass-selected precursor (6,5,6)¹⁻ (Fig. S14). These odd numbers of oxygen species are particularly present when ion–molecule reactions with CO₂ are performed under CW irradiation. Fig. 10, top panel, displays an MS² spectrum corresponding to reaction for 200 ms of precursor (6,9,0)¹⁻ with CO₂ (0.5%) under constant irradiation with a 488 nm CW laser. The following odd numbers of oxygen species are then detected: (6,0,11)¹⁻, (6,1,9)¹⁻, (6,2,7)¹⁻, (6,3,7)¹⁻, (6,4,3)¹⁻, (6,4,5)¹⁻, (6,5,3)¹⁻, (6,5,5)¹⁻, (6,6,3)¹⁻, (6,7,1)¹⁻, and (6,8,1)¹⁻. Thus, these species seem specifically related to the presence of CO₂ in the ion trap and significantly favoured by irradiation, which is evocative of light induced electron transfer (LIET) from the cluster to CO₂. Remarkably, such LIET seem to be favoured in adducts where no charge transfer occurs in the ground state,⁴⁸ which would be the case for transient, weakly bound physisorbed CO₂.

Computational studies were performed considering sequential addition of CO₂ and loss of CO to the most stable isomer of ion (6,8,0)¹⁻ observed in Fig. 10. The Gibbs free energies of reaction are given in Fig. 11 together with the geometrical representation of the different isomers. Interestingly, the CO₂ binding is exergonic in all configurations that were found after considering all active binding sites (−0.148 eV ≤ ΔG^298K ≥ −1.17 eV). The CO₂ binding energy remains modest compared to O₂ (3–5 eV, see the prior section) and in any case lower than the Mo–I binding energy, which explains that CO₂ adducts cannot be stabilized by I˙ losses as for O₂ adducts and may dissociate rapidly in the gas phase. QC calculations show that CO₂ coordination modes are diverse, similar to what is reported in the literature,¹¹ with the carbon atoms or the oxygen atom(s) interacting with one or two molybdenum atoms. Simulations on the non-oxygenated (6,8,0)¹⁻˙ show that CO₂ binding can occur under multiple configurations (Fig. 11), and is associated with various degrees of electron transfer yielding bound CO₂⁻ and attested by the appearance of an angle in the otherwise linear neutral CO₂: η¹-C̲O₂ (isomers D and F), η¹-O̲CO (isomer E), η²-O̲C̲O (isomers B and C), and μ₂κ³(η²-O̲C̲O, η²-OC̲O̲) (isomer A). The most stable adduct, labelled A in Fig. 11, shows a bent configuration of CO₂ where each “CO” branch interacts with a Mo center in η² manner. The Gibbs free energies associated with CO loss from these structures are all endergonic with slightly larger absolute energy values than the ΔG^298K CO₂ binding. The energy balance for the successive CO₂ binding, dissociation and CO release is endergonic by ∼0.35 eV for most of the configurations (except F) if the reactions are barrierless. This may explain the added efficiency of odd oxygen species under light irradiation: the absorption of a photon may promote the precursor ion (or the short-lived CO₂ adduct) into a reactive electronically excited state and, more generally, provide the system with the necessary additional energy to enable the dissociation of CO₂.

Fig. 11 Difference of Gibbs free energy at 298 K for the sequential addition of CO₂ and loss of CO to the most stable isomer of (6,8,0)¹⁻. All possible coordination de-coordination sites are considered. See SI Table S4 for all reaction energy changes.

Conclusions

The photo-activation of the [Mo₆I₁₄]²⁻ cluster ions interestingly yields both the oxidized [Mo₆I₁₄]⁻˙ species and the electron-rich [Mo₆I₁₃]²⁻˙ radical clusters in addition to [Mo₆I₁₃]⁻. The latter is also produced by collisional heating (CID). The former two products are both photo-specific since neither is formed by collisional activation. The optical fragmentation spectrum of [Mo₆I₁₄]²⁻ displays features in both the UV and visible ranges, consistent with its QC simulated spectrum. Although the major fragments always include CID-type “thermal” fragments, both photo-specific fragments display remarkable wavelength dependence. The oxidized species [Mo₆I₁₄]⁻˙ is only significantly formed below 310–340 nm where photon energies exceed the sum of the repulsive coulombic barriers and electron binding energy. Although this species is only formed in the UV and ion–molecule collisions show no sign of reactivity with CO₂ or O₂, its mere presence is a sign that the cluster possesses strong reducing properties in the electronic excited state. Thus, it is possible that the photo-catalytic reaction takes place when the cluster is in an excited electronic state rather than with a ground state photo-product.

In contrast to [Mo₆I₁₄]⁻˙, the radical [Mo₆I₁₃]²⁻˙ clusters are formed along the first absorption band in the visible range ∼500 nm. Since the photo-reduction properties of these Mo₆ clusters have been reported upon irradiation in the visible range, it may seem plausible that [Mo₆I₁₃]²⁻˙ is a reactive intermediate in photo-catalytic cycles. However, ion–molecule collisions between CO₂ and the electron-rich [Mo₆I₁₃]²⁻˙ showed no sign of reactivity, which discards the possibility for the direct reduction of CO₂. It does not necessarily discard [Mo₆I₁₃]²⁻˙ as a reactive intermediate: its electron rich character corresponds to a non-reduced core surrounded by a radical site which could be driving radical reactions with alternative intermediate chemicals. Thus, ion–molecule reactions and QC calculations showed that [Mo₆I₁₃]²⁻˙ reactions with O₂ are very efficient and exergonic. However, QC and experiments also show that the oxygenation of [Mo₆I₁₃]²⁻˙ was largely responsible for the Mo₆ core destabilization and fragmentation, which could arguably be seen as catalyst poisoning.

Aside from the primary photo-products, many halogen-depleted clusters are observed experimentally via laser irradiation. Halogen depletion yields clusters with stronger metal–metal bonds compared to the original [Mo₆I₁₄]²⁻ species, reinforcing the metallic character of the cluster. These species also react very efficiently with O₂ and QC calculations show that these reactions are all exergonic, with associated ∼5 eV or more released upon O₂ dissociation. Binding of O₂ becomes quantitative with at least 3 vacant/radical sites, i.e. when apical-to-face-capping scrambling allows for bridging O₂ or dissociative binding configuration of the oxygens on the cluster. Lastly, ion–molecule reactions between these halogen-depleted clusters and CO₂ show, under constant irradiation, proof of CO₂ reduction into CO and O, with the release of CO and formation of the oxidized clusters. QC predicts that these reactions are slightly endergonic and may be favoured in the excited spin states, which could explain the specific role of photon excitation. The release of CO and formation of oxo complexes is in line with similar reactions on small metal clusters and points a priori towards Mo₆ acting as a sacrificial photo-reactant rather than a photocatalyst.

Author contributions

All authors contributed to the conceptualization; experimental investigations were performed by AT, LMA and ROH; computational investigations were performed by NT and KC; data curation was performed by AT, LMA, NT and KC. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

Cartesian coordinates for all computed structures are available at: https://doi.org/10.5281/zenodo.15519713. MS data are available anytime upon request to LMA.

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information includes details on (1) the collision rate calculations; (2) the secondary fragmentation; (3) the experimental optical spectroscopy method; (4) the COSMO correction. Tables reporting peak list, mass differences and assignments for all MS data shown are included. Tables reporting energies calculated for all species and reactions envisaged are included. Additional figures are included (optical data analysis, CID/LID-MS² spectra of various ions, representation of the frontier MOs and spin density for primary species, IMR-MS² spectra of various photo-products). See DOI: https://doi.org/10.1039/d5cp02611c.

Acknowledgements

The authors acknowledge the support of the French Agence Nationale de la Recherche (ANR), under grant ANR-20-CE05-0033 (project PHOTOCAT). LMA and RAJO acknowledge the French Centre National de la Recherche Scientifique (CNRS) for granting 1 month of invited professorship. RAJO thanks the University of Melbourne for providing a short study leave.

Notes and references

1. Z. Luo, A. W. Castleman and S. N. Khanna, Reactivity of Metal Clusters, Chem. Rev., 2016, 116, 14456–14492.
2. M. Feliz, M. Puche, P. Atienzar, P. Concepción, S. Cordier and Y. Molard, In Situ Generation of Active Molybdenum Octahedral Clusters for Photocatalytic Hydrogen Production from Water, ChemSusChem, 2016, 9, 1963–1971.
3. P. Kumar, S. Kumar, S. Cordier, S. Paofai, R. Boukherroub and S. L. Jain, Photoreduction of CO₂ to methanol with hexanuclear molybdenum [Mo₆Br₁₄]²⁻ cluster units under visible light irradiation, RSC Adv., 2014, 4, 10420.
4. S. Cordier, F. Grasset, Y. Molard, M. Amela-Cortes, R. Boukherroub, S. Ravaine, M. Mortier, N. Ohashi, N. Saito and H. Haneda, Inorganic Molybdenum Octahedral Nanosized Cluster Units, Versatile Functional Building Block for Nanoarchitectonics, J. Inorg. Organomet. Polym. Mater., 2015, 25, 189–204.
5. S. Akagi, S. Fujii and N. Kitamura, A study on the redox, spectroscopic, and photophysical characteristics of a series of octahedral hexamolybdenum(II) clusters: [{Mo₆X₈}Y₆]²⁻ (X,Y = Cl, Br, or I), Dalton Trans., 2018, 47, 1131–1139.
6. P. Kumar, H. P. Mungse, S. Cordier, R. Boukherroub, O. P. Khatri and S. L. Jain, Hexamolybdenum clusters supported on graphene oxide: Visible-light induced photocatalytic reduction of carbon dioxide into methanol, Carbon, 2015, 94, 91–100.
7. A. Barras, M. R. Das, R. R. Devarapalli, M. V. Shelke, S. Cordier, S. Szunerits and R. Boukherroub, One-pot synthesis of gold nanoparticle/molybdenum cluster/graphene oxide nanocomposite and its photocatalytic activity, Appl. Catal., B, 2013, 130–131, 270–276.
8. N. T. K. Nguyen, C. Lebastard, M. Wilmet, N. Dumait, A. Renaud, S. Cordier, N. Ohashi, T. Uchikoshi and F. Grasset, A review on functional nanoarchitectonics nanocomposites based on octahedral metal atom clusters (Nb6, Mo6, Ta6, W6, Re6): inorganic 0D and 2D powders and films, Sci. Technol. Adv. Mater., 2022, 23, 547–578.
9. D. G. Nocera and H. B. Gray, Electrochemical reduction of molybdenum(II) and tungsten(II) halide cluster ions. Electrogenerated chemiluminescence of tetradecachlorohexamolybdate(2-) ion, J. Am. Chem. Soc., 1984, 106, 824–825.
10. F. Tureček, UV-vis spectroscopy of gas-phase ions, Mass Spectrom. Rev., 2023, 42, 206–226.
11. A. Fielicke, Probing the binding and activation of small molecules by gas-phase transition metal clusters via IR spectroscopy, Chem. Soc. Rev., 2023, 52, 3778–3841.
12. J. Zelenka and J. Roithová, Mechanistic Investigation of Photochemical Reactions by Mass Spectrometry, ChemBioChem, 2020, 21, 2232–2240.
13. P. Armentrout, Reactions And Thermochemistry Of Small Transition Metal Cluster Ions, Annu. Rev. Phys. Chem., 2001, 52, 423–461.
14. L. G. Dodson, M. C. Thompson and J. M. Weber, Characterization of Intermediate Oxidation States in CO₂ Activation, Annu. Rev. Phys. Chem., 2018, 69, 231–252.
15. H. Schwarz, Metal-mediated activation of carbon dioxide in the gas phase: Mechanistic insight derived from a combined experimental/computational approach, Coord. Chem. Rev., 2017, 334, 112–123.
16. E. Hossain, D. W. Rothgeb and C. Chick Jarrold, CO₂ reduction by group 6 transition metal suboxide cluster anions, J. Chem. Phys., 2010, 133, 024305.
17. Y.-H. Zhang and J.-B. Ma, Consecutive Reduction of Five Carbon Dioxide Molecules by Gas-Phase Niobium Carbide Cluster Anions Nb₃C₄⁻: Unusual Mechanism for Enhanced Reactivity by the Carbon Ligands, J. Phys. Chem. A, 2024, 128, 2323–2329.
18. X.-Y. He, Y.-Z. Liu, S.-D. Wang, X. Lan, X.-N. Li and S.-G. He, Multiple CO₂ reduction mediated by heteronuclear metal carbide cluster anions RhTaC₂⁻, Dalton Trans., 2022, 51, 11491–11498.
19. L. Song, A. Eychmueller, R. J. St. Pierre and M. A. El-Sayed, Reaction of carbon dioxide with gaseous niobium and niobium oxide clusters, J. Phys. Chem., 1989, 93, 2485–2490.
20. H.-Y. Zhou, M. Wang, Y.-Q. Ding and J.-B. Ma, Nb₂BN₂⁻ cluster anions reduce four carbon dioxide molecules: reactivity enhancement by ligands, Dalton Trans., 2020, 49, 14081–14087.
21. R.-Z. Li, Q. Yuan, Z. Yang, E. Aprà, Z. Li, V. A. Azov, K. Kirakci, J. Warneke and X. Wang, Photoelectron spectroscopy of [Mo₆X₁₄]²⁻ dianions (X = Cl–I), J. Chem. Phys., 2019, 151, 194310.
22. P. Su, Z. Warneke, D. Volke, M. F. Espenship, H. Hu, S. Kawa, K. Kirakci, R. Hoffmann, J. Laskin, C. Wiebeler and J. Warneke, Gas Phase Reactivity of [Mo₆X₁₄]²⁻ Dianions (X = Cl – I), J. Am. Soc. Mass Spectrom., 2023, 34, 161–170.
23. S. Gronert, Mass Spectrometric Studies of Organic Ion/Molecule Reactions, Chem. Rev., 2001, 101, 329–360.
24. C. C. Bright, M. B. Prendergast, P. D. Kelly, J. P. Bezzina, S. J. Blanksby, G. da Silva and A. J. Trevitt, Highly efficient gas-phase reactivity of protonated pyridine radicals with propene, Phys. Chem. Chem. Phys., 2017, 19, 31072–31084.
25. R. A. J. O’Hair, Mass spectrometry based studies of gas phase metal catalyzed reactions, Int. J. Mass Spectrom., 2015, 377, 121–129.
26. K. Kirakci, S. Cordier and C. Perrin, Synthesis and Characterization of Cs₂Mo₆X₁₄ (X = Br or I) Hexamolybdenum Cluster Halides: Efficient Mo₆ Cluster Precursors for Solution Chemistry Syntheses, Z. Anorg. Allg. Chem., 2005, 631, 411–416.
27. L. L. Lopez, P. R. Tiller, M. W. Senko and J. C. Schwartz, Automated strategies for obtaining standardized collisionally induced dissociation spectra on a benchtop ion trap mass spectrometer, Rapid Commun. Mass Spectrom., 1999, 13, 663–668.
28. R. Rüger, M. Franchini, T. Trnka, A. Yakovlev, E. van Lenthe, P. Philipsen, T. van Vuren, B. Klumpers and T. Soini, AMS 2022.1, SCM, Theoretical Chemistry, Vrije Universiteit, Amsterdam, The Netherlands, 2022, https://www.scm.com.
29. G. te Velde, F. M. Bickelhaupt, E. J. Baerends, C. Fonseca Guerra, S. J. A. van Gisbergen, J. G. Snijders and T. Ziegler, Chemistry with ADF, J. Comput. Chem., 2001, 22, 931–967.
30. ADF 2022.1, SCM, Theoretical Chemistry, ADF 2022.1, SCM, Theor. Chem. Vrije Univ. Amsterdam, Netherlands, https://www.scm.com.
31. J. P. Perdew, K. Burke and M. Ernzerhof, Generalized Gradient Approximation Made Simple, Phys. Rev. Lett., 1996, 77, 3865–3868.
32. Y. Zhang and W. Yang, Comment on “Generalized Gradient Approximation Made Simple”, Phys. Rev. Lett., 1998, 80, 890.
33. E. Caldeweyher, S. Ehlert, A. Hansen, H. Neugebauer, S. Spicher, C. Bannwarth and S. Grimme, A generally applicable atomic-charge dependent London dispersion correction, J. Chem. Phys., 2019, 150, 154122.
34. E. van Lenthe, A. Ehlers and E.-J. Baerends, Geometry optimizations in the zero order regular approximation for relativistic effects, J. Chem. Phys., 1999, 110, 8943–8953.
35. S. Hirata and M. Head-Gordon, Time-dependent density functional theory within the Tamm–Dancoff approximation, Chem. Phys. Lett., 1999, 314, 291–299.
36. A. Klamt and V. Jonas, Treatment of the outlying charge in continuum solvation models, J. Chem. Phys., 1996, 105, 9972–9981.
37. S. F. Boys and F. Bernardi, The calculation of small molecular interactions by the differences of separate total energies. Some procedures with reduced errors, Mol. Phys., 1970, 19, 553–566.
38. M. Bouakil, A. J. Kulesza, S. Daly, L. MacAleese, R. Antoine and P. Dugourd, Visible Multiphoton Dissociation of Chromophore-Tagged Peptides, J. Am. Soc. Mass Spectrom., 2017, 28, 2181–2188.
39. J. M. Herbert, Visualizing and characterizing excited states from time-dependent density functional theory, Phys. Chem. Chem. Phys., 2024, 26, 3755–3794.
40. K. Harada, T. K. N. Nguyen, F. Grasset, C. Comby-Zerbino, L. MacAleese, F. Chirot, P. Dugourd, N. Dumait, S. Cordier, N. Ohashi, M. Matsuda and T. Uchikoshi, Light-dependent ionic-electronic conduction in an amorphous octahedral molybdenum cluster thin film, NPG Asia Mater., 2022, 14, 21.
41. M. H. Chisholm, Metal-metal bonds and metal-carbon bonds in the chemistry of molybdenum and tungsten alkoxides, Polyhedron, 1983, 2, 681–722.
42. T. P. Second, J. D. Blethrow, J. C. Schwartz, G. E. Merrihew, M. J. MacCoss, D. L. Swaney, J. D. Russell, J. J. Coon and V. Zabrouskov, Dual-pressure linear ion trap mass spectrometer improving the analysis of complex protein mixtures, Anal. Chem., 2009, 81, 7757–7765.
43. B. G. Ward and F. E. Stafford, Synthesis and structure of four- and five-coordinated gaseous oxohalides of molybdenum(VI) and tungsten(VI), Inorg. Chem., 1968, 7, 2569–2573.
44. E. I. Stiefel, Progress in Inorganic Chemistry, Prog. Inorg. Chem., 1977, 22, 1–223.
45. L. S. Villata, M. R. Féliz and A. L. Capparelli, Photochemical and catalytic properties of dimeric species of molybdenum(V), Coord. Chem. Rev., 2000, 196, 65–84.
46. J. P. Murphy and R. A. Yost, Origin of mass shifts in the quadrupole ion trap: dissociation of fragile ions observed with a hybrid ion trap/mass filter instrument, Rapid Commun. Mass Spectrom., 2000, 14, 270–273.
47. J. E. McClellan, J. P. Murphy, J. J. Mulholland and R. A. Yost, Effects of Fragile Ions on Mass Resolution and on Isolation for Tandem Mass Spectrometry in the Quadrupole Ion Trap Mass Spectrometer, Anal. Chem., 2002, 74, 402–412.
48. A. Göbel, A. Rubio and J. Lischner, Light-Induced Charge Transfer from Transition-Metal-Doped Aluminum Clusters to Carbon Dioxide, J. Phys. Chem. A, 2021, 125, 5878–5885.

---

Footnote

† Dedicated with respect and admiration to Professor Helmut Schwarz on the occasion of his award of the Wolf Prize in Chemistry, an important recognition for his outstanding work on exploring reactive species in the gas phase to solve fundamental problems in catalysis.

---