IR multiple photon dissociation spectroscopy of MO₂⁺ (M = V, Nb, Ta)

Transition metals and their oxides are important compounds in industrial catalysts of oxidation–reduction reactions because they show a wide variety of coordination modes and bonding due to their ability to adopt different oxidation states.¹ Unfortunately, the way catalysts enhance the reaction rate is often not well understood, and their improvement could benefit from mechanistic insight at the molecular level. Small clusters can act as model systems for catalytically active sites by virtue of their well-defined size and elemental composition. By reacting them in the gas phase with molecules and sampling the products with mass-spectrometric techniques, a detailed understanding of the energetics of metal–molecule interactions can be obtained.^2–4 

Optical spectroscopy allows for the complementary structure determination of such products, information that cannot be obtained from mass-spectrometry alone. Robust infrared (IR) spectroscopic techniques have been developed to get spectroscopic information, often relying on the mass-spectrometric detection of resonant IR induced fragmentation. For this purpose, metal clusters are often complexed with so-called messenger species that are relatively weakly bound and that are consequently relatively easily eliminated upon resonant absorption. Thus far, IR spectra for transition metal oxide clusters have been obtained using such “tagging” techniques, either using atoms or molecules specifically introduced to perform this task, such as Ar, He, or H₂,^5–7 or by the presence of intact dioxygen, physisorbed to the cluster surface, as a remainder from the aggregation process.^8–12 IR fragmentation of metal oxide clusters, eliminating either atomic O or metal oxide fragments, has only been observed by Asmis and co-workers for anionic M_xO_y⁻.^13,14 To investigate products resulting from reactions with metal and metal-oxide clusters that go beyond the formation of the entrance complex, more elevated temperatures are frequently required; these temperatures typically do not allow for complexation with weakly bound messengers. One solution is to transfer the products to a low-temperature ion trap to tag them prior to spectral characterization. Approaches without the need of tagging are double-resonance techniques, e.g., IR–UV spectroscopy,^15,16 or resonant IR irradiation followed by irradiation using an off-resonant, intense CO₂ laser to enhance fragmentation.^17,18 Another one is to rely on direct fragmentation with intense IR light, which could be problematic for ions with high binding energies.

To explore the latter route, we have recently connected a laser vaporization source to the Fourier-Transform Ion Cyclotron Resonance (FT-ICR) mass spectrometer coupled to the Free Electron Laser for Intra-Cavity Experiments (FELICE) beamline. In this instrument, cluster ions are collected and stored in an intermediate trap, where reactions between clusters and reactant molecules can take place. After a certain collection and/or reaction time, clusters and reaction products are transferred to one of the FT-ICR cells, where they can be mass-selected and spectroscopically characterized using intense IR light.¹⁹ Although, due to the necessity to absorb multiple photons to induce dissociation, IR multiple-photon dissociation (IRMPD) spectra are not equivalent to linear absorption spectra, they are sufficiently close that they can reliably serve for structure determination.^20,21

Here, we present the first IRMPD spectra of laser ablated metal oxide species using this instrument, taking VO₂⁺, NbO₂⁺, and TaO₂⁺ as examples for strongly bound triatomic species. To the best of our knowledge, these are the smallest covalently bound cationic systems that have been spectroscopically characterized via direct vibrational excitation over their complete fingerprint region. The MO₂⁺ systems under study are not only small, hampering rapid intramolecular vibrational redistribution, a crucial element in efficient IR excitation; they are also very stable, raising the bar even further for direct dissociation in the IR. Theoretical work showed that the geometric structure of the MO₂⁺ cations is of C_2v symmetry in the ¹A₁ electronic ground state,^22–24 with somewhat conflicting binding energies of 3.91 eV and 4.39 eV for VO₂⁺, and 5.49 eV for NbO₂⁺.^23,24 Experimentally, the [OV⁺–O] bond energy was established using collision induced dissociation experiments as 3.51 ± 0.36 eV,²⁵ whereas guided ion beam (GIB) experiments found a value of 3.06 ± 0.40 eV.²⁶ Further GIB studies reported experimental bond energies for [ONb⁺–O] and [OTa⁺–O] of 5.71 ± 0.17 eV and 6.08 ± 0.12 eV, respectively.^27,28

Spectroscopically, VO₂⁺ was studied by Asmis and co-workers who recorded IR spectra employing the FELIX free-electron laser via photodissociation of the VO₂⁺·He complex; without tagging, no fragmentation signal was observed for VO₂⁺.^6,29 They observed bands at 990 cm⁻¹ and 1017 cm⁻¹ and assigned them to the antisymmetric and symmetric stretch vibrations, respectively.^6,29 The VO₂⁺ bending mode was not observed.²⁹ NbO₂⁺ and TaO₂⁺ were never spectroscopically characterized in the gas phase. Zhou and Andrews investigated the reaction of niobium and tantalum with oxygen and recorded the IR spectra of reaction products deposited in a matrix, obtaining the signatures of NbO₂⁺ and TaO₂⁺.²² Gong, Zhou, and Andrews summarized the spectroscopic properties and theoretical studies on metal oxides in an extensive review.³⁰ Here, we will demonstrate that the brightness of the IR laser in the FELICE intracavity arrangement is sufficient to fragment these strongly bound metal oxide trimers, allowing for the observation of all fundamental modes.

The new experimental setup is schematically shown in Fig. 1. Clusters are generated in a newly installed laser vaporization source.^31,32 This ion source can produce a variety of cluster ions such as metal and metal oxide clusters,^31,33 hydrated metal clusters,^34,35 or metal free (hydrated) cluster ions.^36,37 A metal target disk is irradiated by a frequency doubled Nd:YAG laser with a repetition rate of 30 Hz, which forms a plasma cloud above the target. The plasma cloud is cooled by a helium pulse released by a piezo-valve with a backing pressure of 16 bars. The mixture of helium and ions formed travels through a 6 cm long channel, before it adiabatically expands into vacuum. Ions are then transferred through a 2 mm diameter skimmer located 3 cm downstream, a gate valve separating the source from the rest of the instrument, and a set of segmented cylindrical electrodes operating in an Einzel lens configuration, into a quadrupole bender that allows us to select this source or a sublimation source with electron impact ionization.¹⁹ Ions are further transferred through a radiofrequency (RF) linear quadrupole in guiding mode, before entering a collision cell containing a 1 · 10⁻² mbar pressure of Ar and a linear quadrupole ion trap with rectangular electrodes, segmented in three parts. After collection over ∼0.6 s, all ions are released from the trap by lowering the end cap voltage and transferred to one of the four storage and detection cells of the FT-ICR. Under the conditions employed, we formed metal ions with different numbers of oxygen atoms. After trapping all ions in one of the ICR cells, MO₂⁺ ions were mass-selected by ejecting unwanted species using a stored-waveform inverse Fourier-transform (SWIFT) pulse.³⁸ MO₂⁺ ions are then irradiated using one or several macropulses of FELICE. The trapped sample is mass-analyzed after irradiation.

FELICE provides IR radiation in the 100 cm⁻¹–2000 cm⁻¹ spectral range in 5 or 10 macropulses/s, where every macropulse consists of a ∼10 μs pulse train of micropulses spaced 1 ns apart. The spectral bandwidth is near transform limited and is set to ∼0.6% full-width at half-maximum (FWHM) of the central frequency. The four storage and detect cells are placed 100 mm away from each other, which is slightly more than the Rayleigh length of 82 mm. The first cell (cell 1) coincides with the FELICE focus, and the IR fluence is thus reduced by a factor of 2.3 for each further cell. During the experiments, the macropulse energy was between 0.9 J and 1.4 J for both the low and high frequency spectral ranges. The laser fluence in cell 1 (in focus, low-frequency range) is then 140 J/cm² and about 10 J/cm² in cell 4 (out-of-focus, high frequency range).

Upon resonant IR excitation, atomic O is eliminated resulting in the appearance of MO⁺ ions. The IRMPD yield is then calculated as ln(1 + I_MO/I_MO2), with I_MO and I_MO2 being the ion intensity recorded for MO⁺ and MO₂⁺, respectively.

Theoretical vibrational frequencies and dissociation energies of the MO₂⁺ species are calculated using Gaussian 16 with the coupled-cluster method with single, double, and perturbative triple excitations [CCSD(T)] and def2QZVPPD basis set, utilizing the electron core potential approximation for Nb and Ta.^39,40 The respective IR absorption intensities are calculated at the B3LYP/def2QZVPPD level. All binding energies include the zero-point vibrational energy. To compare calculated linear vibrational spectra to experimental spectra, harmonic frequencies are scaled employing a uniform scaling factor of 0.95, and rovibrational spectra are simulated using the PGOPHER program.⁴¹ For this, each vibrational transition is assumed to be a pure a-, b-, or c-type transition, and rotational constants are taken from the CCSD(T) optimized structures. The vibrational temperature is assumed to be 300 K, and the individual rovibrational transitions are convoluted with a 5 cm⁻¹ FWHM Gaussian line shape function.

Figure 2 shows the recorded IRMPD spectra for VO₂⁺, NbO₂⁺, and TaO₂⁺. All spectra are recorded via dissociation of the parent species, MO₂⁺, leading to the formation of MO⁺. Spectra are initially recorded in cell 4 (cell center 30 cm out of focus) employing two to five FELICE macropulses. This leads for all M (=V, Nb, Ta) to the observation of a single, relatively broad band in the 800 cm⁻¹–1100 cm⁻¹ spectral region, sometimes with an asymmetric line shape hinting at the presence of a second band. No further bands are observed at lower frequencies under these irradiation conditions. Therefore, in the region of 250 cm⁻¹–500 cm⁻¹, spectra are recorded in cell 1, coinciding with the FELICE focus, using 5–20 macropulses. Here, for each metal, a well-structured band is observed.

Triatomic molecules of the C_2v symmetry group have three normal modes: the symmetric MO₂⁺ stretching vibration (ν₁, a₁ irreducible representation), the MO₂⁺ bending vibration (ν₂, a₁), and the anti-symmetric MO₂⁺ stretching vibrations (ν₃, b₂), all of which are dipole allowed. Based on calculations, the bands close to 1000 cm⁻¹ can safely be assigned to a combination of the stretching modes ν₁ and ν₃, whereas the bands close to 400 cm⁻¹ are due to the ν₂ bending vibration. An overview of calculated and observed vibrational frequencies is given in Table I. The optimized structures of the MO₂⁺ ions investigated in this study are shown in Fig. 3, and bond dissociation energies are given in Table II. The metal centers in the MO₂⁺ cations exhibit an electronic structure that is close to a d³ configuration (for details, see Table S3). For completeness, we included calculations for the triplet and quintet states of each ion (Table S2). Even though they are significantly higher in energy, we cannot exclude their presence due to potential incomplete quenching in the ion trap. As a consequence, their significantly shifted vibrational frequencies could betray their presence. Although the same holds for long-living electronically excited states, calculations for them are considerably more complex and deemed out of the scope of this contribution.

For VO₂⁺ [Fig. 2(a)], we observe a broad band around 955 cm⁻¹. According to our CCSD(T) calculations, there is an intense band at 981 cm⁻¹, which corresponds to the ν₃ asymmetric stretch vibration. The ν₁ symmetric stretch vibration is predicted at 990 cm⁻¹ with a much lower calculated intensity. Experimentally, we cannot resolve these individual bands, and we find a lower frequency than expected from calculations. Asmis et al., using He as a tag, observed resonances at 990 cm⁻¹ and 1017 cm⁻¹ with relative intensities of 3:1.²⁹ The resonances observed in our study thus appear at significantly lower frequencies.

In the 250 cm⁻¹–600 cm⁻¹ region, a broad band with maxima at 398 cm⁻¹ and 433 cm⁻¹ is observed. Having already assigned two out of three fundamental modes, we expect only a single vibration in this region. The calculations place the ν₂ bending mode at 408 cm⁻¹. As we clearly observe two subpeaks, there must be an underlying effect that splits this single vibrational mode.

For NbO₂⁺ [Fig. 2(b)], we observe a broad, asymmetric band in the 800 cm⁻¹–1100 cm⁻¹ region with a maximum at 921 cm⁻¹. This band is thus red shifted compared to VO₂⁺, and its shape suggests more clearly the presence of multiple peaks. CCSD(T) calculations predict the asymmetric stretch vibration at 928 cm⁻¹ and the symmetric stretch vibration with a lower intensity at 970 cm⁻¹. The asymmetric band observed could thus be explained as a superposition of a strong band at 921 cm⁻¹ and a weaker band at the high-frequency side. Other calculations (see Table I) predict the asymmetric stretch at 973 cm⁻¹ (scaled) or at 981 cm⁻¹ and 999 cm⁻¹ (unscaled), and the symmetric stretch at 1015 cm⁻¹ (scaled) or 1021 cm⁻¹ and 1034 cm⁻¹ (unscaled).^8,22,24 Matrix-isolation spectroscopy of deposited NbO₂⁺ yielded frequencies of 938 cm⁻¹ and 989 cm⁻¹,²² matching the current work quite well.

In the 250 cm⁻¹–600 cm⁻¹ region, we observe for NbO₂⁺ a similar band structure as for VO₂⁺. With maxima at 370 cm⁻¹ and 409 cm⁻¹, the separation between the bands is only 4 cm⁻¹ larger than for VO₂⁺, but the two peaks are more clearly separated. The calculated value of 361 cm⁻¹ is consistent with the previously reported (unscaled) values of 380 cm⁻¹ and 390 cm⁻¹.^22,24

Finally, for TaO₂⁺ [Fig. 2(c)], we observe a spectrum comparable to those of VO₂⁺ and NbO₂⁺. The band in the 800 cm⁻¹–1100 cm⁻¹ region has a maximum at 921 cm⁻¹. Our calculations predict the asymmetric and symmetric stretch vibrations at 908 cm⁻¹ and 964 cm⁻¹, which is lower than the previously reported unscaled values of 969 cm⁻¹–996 cm⁻¹ and 1019 cm⁻¹–1054 cm⁻¹, respectively.^22,42 Zhou and Andrews reported the experimental values of 938 cm⁻¹ and 992 cm⁻¹ for the matrix isolated clusters.²² 

The purported bending mode is observed in the 350 cm⁻¹ region, again with a remarkable substructure, and shifted toward the red compared to VO₂⁺ and NbO₂⁺. The low-frequency maximum at 354 cm⁻¹ is close to the calculated frequency of 348 cm⁻¹ for the bending mode, whereas the flatness of the band around 400 cm⁻¹ suggests the saturation of the high-frequency maximum. Other calculations expect the bending mode at 380 cm⁻¹ and 383 cm⁻¹ (unscaled).^22,42

If the only fundamental mode expected below 500 cm⁻¹ is the MO₂⁺ bending mode, the question arises what the marked substructure, observed for all M, is. One possibility for this is the appearance of vibrational hot bands, reflecting a population in the ν₂ = 1 state present prior to irradiation. We performed anharmonic calculations at the UB3LYP/def2TZVPP level of theory to estimate the frequency shift of such hot bands. For the bending mode, the anharmonicity found was only very small, with a shift of less than 1 cm⁻¹ for the ν₂ = 1 → ν₂ = 2 with respect to the ν₂ = 0 → ν₂ = 1 transition. Clearly, this is too small a shift to explain the observed band structure.

A second possibility is the rotational substructure of the vibrational bands. Due to their size, the rotational constants of MO₂⁺ are appreciable (e.g., the rotational A constant for VO₂⁺ of 0.938 cm⁻¹). Due to the room temperature environment of the ion trap where the MO₂⁺ ions are collected prior to transfer to the ICR cell, a substantial population of J, K states exists, which leads to a significant broadening of the bands observed.⁴³ To verify this, we simulate the rovibrational envelopes for each transition using the PGOPHER software package,⁴¹ employing for both the lower and upper vibrational states the same rotational constants. The resulting rovibrational spectra are depicted in Fig. 2 as the red traces. It is evident that the ν₃ asymmetric stretch mode for all metals, an a-type transition, shows a significantly narrower rovibrational envelope than the ν₂ bending vibration (b-type). The latter does exhibit a clear, symmetric double-peak structure that could potentially explain the observed structure, although the simulated splitting between the two subpeaks is too low for each of the MO₂⁺ systems studied. For VO₂⁺, we observe a splitting of 35 cm⁻¹, while simulations predict only 20 cm⁻¹ between both maxima, and this is even more pronounced for NbO₂⁺ (39 cm⁻¹ observed and 16 cm⁻¹ predicted, respectively), and TaO₂⁺ (43 cm⁻¹ and 15 cm⁻¹). A significant change in the assumed rotational temperature does not considerably alter the spacing between the peaks; it does affect the intensity distribution, but the experimental asymmetric line shape is not approximated much better. The small peak around 300 cm⁻¹ is interpreted as part of the rotational substructure. The low energy tail in this region is very sensitive to the temperature and rotational constants used for the simulation (see Fig. S1). Electronically excited states and hot bands may also contribute to the broad spectrum as the calculated absorptions shift to lower energies, although the calculated vibrational frequencies for other spin states (see Table S2) appear to rule out a direct assignment.

The resulting asymmetry resembles the one observed, for instance, for the bending vibration of SO₂ or O₃, two other molecules of C_2v symmetry.⁴⁴ We note that in the description of the SO₂ spectrum, the rotational constants for the excited state are different from those in the ground state.⁴⁵ Thus, in a final attempt to resolve the asymmetry and large splitting, we empirically varied the rotational constants in the vibrationally excited state and achieved a slightly better agreement with the observed structure of the bending mode as shown in Fig. 2 using the empirical rotational constants for the vibrationally excited state given in Table III. The variation in the rotational constants amounts to higher values than those reported for SO₂, but, for instance, for ketene, similar changes were reported.⁴⁶ We conclude that the rotational substructure of the vibrational bands can at least partially explain the structure observed. Further broadening effects due to the high intensity of the IR laser, required to overcome the several eV internal barrier toward fragmentation, likely contribute to the observed spectral bandwidths.

We have successfully coupled a laser vaporization source to the FT-ICR mass spectrometer installed at the second beamline of FELICE, the free-electron laser for intracavity experiments. Using this instrument, we recorded IRMPD spectra for MO₂⁺ (M = V, Nb, Ta) cations via the elimination of atomic oxygen, forming MO⁺. The binding energy to be overcome for this process to occur ranges from 3.1 eV to 3.5 eV for VO₂⁺, to 5.7 eV for NbO₂⁺, and to 6.1 eV for TaO₂⁺.^25–28 

For each species, the obtained spectra contain bands for both the ν₂ bending and ν₃ asymmetric stretching vibrations, whereas the ν₁ symmetric stretching vibration is likely obscured by the stronger ν₃. The observed frequencies for ν₂ and ν₃ are in reasonable agreement with CCSD(T) level harmonic calculations. The broadness of the bands is partially understood by taking into account the full rovibrational envelope for these small species with appreciable rotational constants. Further broadening is attributed to power broadening effects, required to overcome the high dissociation barriers toward O-elimination. The spectral shapes of the IRMPD spectra are more complex than a comparison with calculated, and linear absorption spectra can directly explain. A more thorough understanding of the spectra, the IR excitation mechanism, and the underlying structure of the multi-dimensional potential energy surface requires master equation modeling,^10,20,47 which we aim to pursue in our next study. The successful spectral characterization of these strongly bound species implies the introduced combination of cluster source and reactive ion trap, and FT-ICR holds great promise for the structural characterization of reaction products involving metal and metal-oxide clusters: IR spectroscopy (without the need for tagging) should be feasible for even the strongest covalently bound reaction products, whereas for more weakly bound species, a reduction in the micropulse repetition rate will allow flexibility in reducing the spectral broadening. Schemes that allow for flexible micropulse repetition rates are currently under development in the FELIX laboratory.

See the supplementary material for a benchmark of the used quantum chemical calculation methods, data on ions with other than singlet spin multiplicity, electronic configurations, and the Cartesian coordinates of the optimized structures.

The research leading to this result was supported by the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO) through the Materials for Sustainability Program (Grant No. 680.M4SF.028). We further gratefully acknowledge NWO for support of the FELIX Laboratory and for granting computational time at the CARTESIUS cluster of SURFsara in Amsterdam (NWO Rekentijd; Grant No. 2019.062). The project further received support from the project CALIPSOplus under Grant Agreement No. 730872 from the EU Framework Programme for Research and Innovation HORIZON 2020. The computational results presented here have been achieved, in part, using the LEO HPC infrastructure of the University of Innsbruck. We thank Arjan van Vliet for his substantial technical assistance.

The data that support the findings of this study are available from the corresponding authors upon reasonable request.