Structural Properties of Gas-Phase Molybdenum Oxide Clusters [Mo4O13]2−, [HMo4O13]−, and [CH3Mo4O13]− Studied by Collision-Induced Dissociation

Molybdenum oxide-based catalysts are widely used for the ammoxidation of toluene, methanation of CO, or hydrodeoxygenation. As a first step towards a gas-phase model system, we investigate here structural properties of mass-selected [Mo4O13]2−, [HMo4O13]−, and [CH3Mo4O13]− by a combination of collision-induced dissociation (CID) experiments and quantum chemical calculations. According to calculations, the common structural motif is an eight-membered ring composed of four MoO2 units and four O atoms. The 13th O atom is located above the center of the ring and connects two to four Mo centers. For [Mo4O13]2− and [HMo4O13]−, dissociation requires opening or rearrangement of the ring structure, which is quite facile for the doubly charged [Mo4O13]2−, but energetically more demanding for [HMo4O13]−. In the latter case, the hydrogen atom is found to stay preferentially with the negatively charged fragments [HMo2O7]− or [HMoO4]−. The doubly charged species [Mo4O13]2− loses one MoO3 unit at low energies while Coulomb explosion into the complementary fragments [Mo2O6]− and [Mo2O7]− dominates at elevated collision energies. [CH3Mo4O13]− affords rearrangements of the methyl group with low barriers, preferentially eliminating formaldehyde, while the ring structure remains intact. [CH3Mo4O13]− also reacts efficiently with water, leading to methanol or formaldehyde elimination. Electronic supplementary material The online version of this article (10.1007/s13361-019-02294-4) contains supplementary material, which is available to authorized users.

Combined with computational methods, i.e., DFT calculations, the studies by Raghavachari, Jarrold, and coworkers map elementary steps with potential relevance for hydrogen evolution reaction (HER) catalysis on molybdenum oxide and sulfide [28,29]. However, HER with molybdenum oxides lead to highly oxidized, mostly saturated clusters whereby it is necessary to reduce the cluster in a second step to complete the catalytic cycle [29]. Ray [29]. Due to this multistep process, subsequent barriers of up to 2 eV are present [29]. Later experimental investigations, however, indicate that these barriers can be overcome along the reaction path, providing indirect evidence for the theoretically found catalytic cycle [17]. On the experimental side, Topolski et al. directly investigated the reactivity of water with molybdenum sulfides and oxosulfides and verified energetic as well as structural similarities of these reactions with the corresponding oxides [28]. Only for molybdenum sulfides, however, barrier-free hydrogen producing reaction pathways could be found. For the oxosulfide species, the molybdenum atom bound to oxygen was calculated to be the reactive site for hydrogen evolution. A fundamental understanding of the relevant processes may facilitate targeted HER-catalyst optimization. To this end, we have recently studied the reactivity of molybdenum clusters with dimethyl disulfide [30]  ] − forms in the gas phase using mass spectrometry. In particular, we investigate the dissociation pathways of the clusters in the hexapole collision cell of a Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR MS), which is well suited for studying clusters in the gas phase [30,[32][33][34][35][36] as well as investigating catalytic cycles [32,34,35,37,38]. We show that the precursor ions have closed-ring structures and that the stability of protonated [HMo 4 O 13 ] − against collisions is strongly enhanced compared to the pure [Mo 4 O 13 ] 2− cluster. This could be attributed to hydrogen roaming prior to any dissociation, as suggested by theoretical calculations. Furthermore, we observe formaldehyde elimination by collisional activation of methylated [CH 3 Mo 4 O 13 ] − , which is a key step in catalytic methanol oxidation, as shown previously by Waters et al. [26].

Experimental and Computational Details
The experimental setup consists of a Bruker Apex Qe FT-ICR Mass Spectrometer, which is equipped with a 9.4 T superconducting magnet and a combined electrospray ionization (ESI) and matrix-assisted laser desorption ionization (MALDI) Apollo II Dual ESI/MALDI ion source. The instrument features a hexapol collision cell for collision-induced dissociation experiments. See a previous publication for details [39].
Cluster ions in the gas phase are obtained via electrospray ionization (ESI) of a 3 mM solution of isotopically enriched ammonium heptamolybdate (NH 4 ) 6  Research Chemicals) and left to dry under air until colorless crystals were obtained. Subsequently, a stock solution was prepared by dissolving the crystals in 8 mL water-methanol mixture (1:1). Since molybdenum oxide clusters with natural isotope distribution have many isotopologues, we used enriched 92 MoO 3 to disentangle the corresponding mass spectra and to improve the signal-to-noise ratio (see Figure S1). Each precursor ion is mass selected with the quadrupole mass filter and trapped in the hexapole collision cell (p ≈ 10 −3 mbar) for a collision time of t coll = 0.8 s. During this time, the ions are excited by a radio frequency signal of variable amplitude, which allows for a qualitative tuning of the collision energies. The fragments are then cooled and trapped by subsequent collisions with the argon buffer gas. Although high-purity argon (99.999%) is used, traces of water from the ESI solution or from air in the gas inlet system can lead to reactive collisions, as discussed in detail below. The collision energy of the trapped ions with the background gas correlates with the bias voltage V c of the collision cell. After fragmentation, the ions are electrostatically guided into the ICR cell, where a highresolution mass spectrum is acquired. By scanning the collision voltage V c and recording a mass spectrum for each set value, breakdown curves of the ion of interest can be recorded. Unfortunately, quantitative threshold energies cannot be derived from this measurement. It is worth mentioning that the instrument used for the employed CID technique is different from other types of multistage MS instruments like, e.g., the Guided Ion Beam studies of Armentrout used for energy resolved CID [40]. Since low-mass ions are discriminated in our setup due to the time-of-flight between the hexapole collision cell and the ICR cell, only ions with m/z > 57.75 were detected. Only fragment ions reaching a relative intensity of at least 0.1% in any CID spectrum were considered in the analysis.
In order to comprehend the observed dissociation pathways, quantum chemical computations are necessary. We model the structure of the precursor ions and dissociation products and, based on relative energies, assess the most probable reaction pathways. The respective dissociation energies provide the lower limit of the energy barrier along the dissociation coordinate. For charge reduction reactions, i.e., dissociation of [Mo 4 O 13 ] 2− into two singly charged fragments, the reverse Coulomb barrier leads to a significant increase of the overall barrier towards dissociation [41,42]. To gain further insight into the structural flexibility of the clusters, we also performed transition state calculations for transfer reactions of H and CH 3 on the Mo 4 O 13 ion.
Density functional theory (DFT) calculations were performed for all observed ions and their neutral counterparts. For each molecule, different initial structures were chosen based on chemical intuition and previous studies [15,24,29,[43][44][45][46]. The two lowest spin multiplicities were considered (higher spin multiplicity states were considered for selected structures to verify that they lie higher in energy). For most ions, the lowest multiplicity state turned out to be the most stable one, which is in contrast to other calculations for small Cr group oxides with high oxygen to metal ratio [13]. Although the ωB97XD functional was reported to provide reliable results for second-row transition metals [47], we used additionally the B3LYP and the M06 functionals to estimate the error of the calculated energies and possible structure variation; the def2TZVP basis set was used. Benchmarking with respect to decomposition of small molybdenum oxide and sulfide clusters shows that B3LYP provides the best results among the applied functionals (Table S3). Benchmarking calculations with respect to the vertical electron detachment energies of [Mo 2 O 7 ] − and [Mo 2 O 7 ] 2− , however, suggest the M06 functional to be the most reliable (see Table S4). Stability of the wave function was tested within all calculations. Transition states were identified by frequency and intrinsic reaction coordinate (IRC) calculations. Charge analysis was performed using the CHELPG method [48], with the radius of Mo chosen to be 1.7 Å. The reported energies include zero-point corrections. Calculations were performed using the Gaussian program [49]. It should be noted that in the collision cell, the collision rate is on the order of 10 5 s −1 , i.e., very minor amounts of water are sufficient to cause these signals.  Table 1. Comparing the energies calculated using different functionals leads to a mean standard deviation of 0.22 eV, and the relative order of the dissociation channels is independent of the functional. The given values constitute the energy difference of the lowest local minima found for the precursor ion and the fragments. Therefore, the energies represent also a lower limit of the energy barrier to   O 13 ] − cluster and its selected fragments, with spin multiplicity given in superscript next to each isomer. Calculated at the ωB97XD/def2TZVP level of theory, relative energy is given in eV overcome during fragmentation. For singly charged fragments, however, the reverse Coulomb barrier must be added, since Coulomb explosion is associated with significant kinetic energy release [41,42,[50][51][52]. We estimated the barrier based on a point-charge separation of 4.1 Å at the transition state of Coulomb explosion, which corresponds to the molecule diameter. This leads to a potential energy of 3.5 eV, representing a rough approximation for the reverse Coulomb barrier, which must be added to the energy of the fragments at infinite separation.

Results and Discussion
The most stable isomer of [Mo 4 O 13 ] 2− at the ωB97XD level has a cyclic structure with one bridging oxygen atom, see isomer Ia in Figure 2, in agreement with the structure found on the BP86/TZP level by Vilà-Nadal et al. [24]. Two isomers (Ib, Ic) with more open structures are predicted to lie within 0.5 eV. At the B3LYP level, calculations even predict the open structure Ib to be the most stable isomer. The observed loss of MoO 3 , reaction (1.1) in Table 1, already at nominally V c = 0 V, indicates that this moiety is weakly bound and requires ringopening prior to dissociation. When the energy of the system increases during the multi-collision CID in the hexapole collision cell, isomerization to Ib or Ic prior to dissociation is very likely. Entropy will favor the more flexible, lower-lying isomer Ib, which can be expected to dominate at elevated energies. All employed DFT functionals (ωB97XD, B3LYP, M06) predict the lowest energy structure for [Mo 3 2). Alternatively, an intact Mo 2 O 6 molecule may be lost, reaction (1.2′), but this neutral product would not explain the high appearance energy of this ion.
Most fragments are singly charged, and assignment to specific dissociation channels is not straightforward in every case. During Coulomb explosion, low-mass-charged fragments gain most of the kinetic energy release and may be lost from the hexapole trap. Loss of fragments due to their high kinetic energy is also the most plausible reason for the increasing relative intensity of the precursor ion from 15 to 30 V collision voltage. Electron detachment may occur from hot doubly charged fragments, possibly with the contribution of tunneling [53]. Both effects disturb the quantitative evaluation of the ion signal and may explain the apparent increase of the precursor ion signal at higher energies.  Figure 1a requires about 4 eV (Table 1). Compared to our recent work on thiomolybdates [31], the observed dissociation patterns differ considerably. In the case of thiomolybdates, S 2 ligands are present on the cluster surface, which are lost preferentially. In the present case, no O 2 moieties are present on the cluster surface (see Figure 2). Dissociation of MoO 3 as a thermodynamically stable entity takes place instead. Figures 1b and  S2. There are only four fragments for the protonated form, and only two of them exceed 1% relative intensity at collision voltages up to 20 V, where about 30% of the precursor ions dissociate. Upon comparing the dissociation patterns of singly and doubly charged ions, one has to consider that doubly charged ions experience twice the acceleration of singly charged ions of the same mass in the electric field. Even then, however, this indicates a higher stability of the protonated ions, as also observed for thiomolybdates [31].

Breakdown curves of [HMo 4 O 13 ] − are shown in
The observed fragments together with plausible neutral products are listed in Table 2. The two protonated fragments, [HMo 2 O 7 ] − and [HMoO 4 ] − , are dominant and arise at low collision energies, reactions (2.1) and (2.2). According to the calculations, reaction (2.2) requires slightly less energy than (2.1), with Mo 3 O 9 formed as a neutral product. In the experiment, the breakdown curves of (2.1) and (2.2) behave parallel, which is consistent with their similar threshold energy. However, reaction (2.2) with the lower threshold is significantly less probable. This can be explained by the required reorganization of the [HMo 4 O 13 ] − precursor. As shown in Figure S3.33, the lowest energy structure of Mo 3 O 9 features a six-membered ring of alternating Mo and O atoms, while dissociation from a ringopened precursor ion would result in isomer II of Mo 3 O 9 , which lies 1.38 eV higher in energy. Consequently, reaction  Figure S3.21, which lie 1.3-1.4 eV above the minimum. The ring-opened structure VI, which affords dissociation into the minimum geometry of [HMo 2 O 7 ] − , lies at 0.61 eV and is thus much more populated.
Overall, the population of the precursor prior to dissociation decides which product ion is formed. At higher energies, fragments with the same stoichiometry are formed, but the charge stays at the former neutral product, reactions (2.3) and (2.4). [Mo 3 O 9 ] − is formed with an overall relative intensity below 0.2% (see Figure S2) with the simultaneous decrease of the [HMoO 4 ] − signal at V c = 19 V. Therefore, it seems entirely plausible that Mo 3 O 9 is formed in reaction (2.2), and at higher energy has a slim chance to scavenge the electron, in line with the reorganization of [HMo 4 O 13 ] − discussed above.
The most stable structure of the protonated form (isomer Va in Figure 3) is similar to the lowest energy structure of [Mo 4 O 13 ] 2− (isomer Ia), with the proton bound to the central oxygen atom. This is consistent with the central oxygen atom in Ia carrying the most negative partial charge (− 0.97 e) as calculated within the CHELPG population analysis. For the five most stable isomers found, only the protonation site is different. The flexibility in protonation could be the reason for the lower fragmentation yield observed, as the cluster is able to use the available energy to mobilize the proton, without disrupting the Mo-O cage. The energy difference between different protonation sites lies within 0.5 eV; ωB97XD/ def2TZVP transition state calculations between isomers Va,b,e predict that 0.9-1.4 eV is necessary for intramolecular proton transfer, Table 3. Thus, hydrogen roaming would take place before any dissociation.  ] − fragment is formed at slightly higher energies and quickly becomes the dominant product, with relative intensity up to 75%. The corresponding dissociation energy of CH 3 MoO 4 was calculated to be relatively high, about 4 eV for reaction (3.5) in Table 4 Figure S2), but with two orders of magnitude lower intensity. This ion most likely arises from collisions with O 2 , reaction (3.6) or (3.6′), with the identical neutrals formed as in reaction (3.5) Figure S2), but about a factor of 25 reduced intensity.
Analogous to ( Table 3). Therefore, dissociation involving activation of the methyl group is more likely than methyl transfer, explaining the different fragmentation patterns of [CH 3  Overall, the only detected carbon containing fragment ion was [CH 3 Mo 2 O 7 ] − . This indicates that also the neutral counterparts might not contain carbon groups, but, for instance, dissociate into formaldehyde and a molybdenum oxide cluster with an additional H atom, which is reflected in the energetics of the product channels.

Conclusions
From electrospray ionization of heptamolybdate dissolved in water-methanol (1:1), we obtained predominantly [Mo 4  ] − supported by theoretical calculations on all observed fragment ions and neutral counterparts. In agreement with literature, our calculations predict a similar ring structure for all three precursor ions, with a central oxygen atom as an energetically favorable location for protonation [24]. The calculations show that in [HMo 4 O 13 ] − , or rather [Mo 4 O 12 (OH)] − , the proton becomes highly mobile prior to dissociation as the barriers were calculated to be around 1 eV. These intramolecular proton transfers are relevant for water activation at molybdenum oxide. In contrast to similar experiments with thiomolybdates [31], where pure S x (x = 2-6) and H x S y (x = 1-3, y = 2-6) neutral fragments dominate due to the presence of disulfide ligands, the investigated oxomolybdates do not possess dioxo ligands. Therefore, no pure oxygen dissociations were observed. In contrast, [CH 3 Mo 4 O 13 ] − , or rather [Mo 4 O 12 (OCH 3 )] − , is more stable with the methyl group bound to one of the bridging oxygen atoms. Here, the presence of several isomers is expected, as the six lowest lying isomers are within 0.5 eV relative energy. Furthermore, we found that the most effective dissociation pathway for [Mo 4 O 12 (OCH 3