Edward I.
Brewer
,
Alice E.
Green
,
Alexander S.
Gentleman
,
Peter W.
Beardsmore
,
Philip A. J.
Pearcy
,
Gabriele
Meizyte
,
Jack
Pickering
and
Stuart R.
Mackenzie
*
Department of Chemistry, University of Oxford, Physical and Theoretical Chemistry Laboratory, South Parks Road, Oxford, OX1 3QZ, UK. E-mail: stuart.mackenzie@chem.ox.ac.uk
First published on 7th September 2022
We report a combined experimental and computational study of carbon dioxide activation at gas-phase Ho+ and HoO+ centres. Infrared action spectra of Ho(CO2)n+ and [HoO(CO2)n]+ ion–molecule complexes have been recorded in the spectral region 1700–2400 cm−1 and assigned by comparison with simulated spectra of energetically low-lying structures determined by density functional theory. Little by way of activation is observed in Ho(CO2)n+ complexes with CO2 binding end-on to the Ho+ ion. By contrast, all [HoO(CO2)n]+ complexes n ≥ 3 show unambiguous evidence for formation of a carbonate radical anion moiety, . The signature of this structure, a new vibrational band observed around 1840 cm−1 for n = 3, continues to red-shift monotonically with each successive CO2 ligand binding with net charge transfer from the ligand rather than the metal centre.
CO2 has been shown to bind molecularly to charged metal species in at least four characteristic binding motifs: M(η1–OCO), M(η1–CO2), M[η2–(C,O)O], and M[η2–(O,O)C], representing varying degrees of molecular activation.3 The natural limit of activation, dissociative binding (or insertion reaction), leads to formation of the metal oxide and a CO molecule.4–11 Metal anions have a greater capability to activate and dissociate CO2 compared with their cationic counterparts as they are able to donate electron density into the antibonding LUMO of CO2 molecules. This is reflected in the wider range of binding motifs for M−(CO2) species.12–22 As well as activating and dissociating CO2, metal anions have been shown to promote dimerization of CO2 forming oxalate-like structures in which two of the CO2 ligands form a single chelating ligand.18 Size-selected metal clusters provide a way to tune the degree of activation with CO2. Only larger Con− (n > 8) clusters bind CO2 at all5 whilst CO2 binds molecularly to small Ptn− clusters (n < 5) but dissociatively on larger ones.6
Metal cations necessarily donate electron density less effectively than anions and M+(CO2) binding is usually M(η1–OCO),4,23–28 dictated by the dominant charge–quadrupole interactions and σ-donation. The typical spectroscopic signature of such binding is a weak blue shift in the CO2 asymmetric stretch around 2349 cm−1.29 There are, however, notable exceptions to this η1 binding; Jiang and coworkers observed M[η2-(C,O)O] motifs in M+(CO2)7 complexes (where M = V, Cr and Mn) resulting from charge transfer from the additional ligands,26 a phenomenon also observed in other systems, such as [MgCO2(H2O)n]+ complexes.30 In rare cases, metal cations have also been shown to lead to oxalate-type structures.7 Room temperature reactivity experiments have shown that some early transition metal cations undergo direct O-atom transfer reactions with CO2.31 CO2 reacts with Ta+ cations under single-collision conditions to form TaO+ which can, in turn, react with a further CO2 to produce TaO2+.32 Direct spectroscopic evidence of dissociatively bound CO2 was observed in Ti+(CO2)n (n = 3–7), V+(CO2)5+ and inferred on Ni+(CO2)n+ (n = 1–12) complexes.7–10 Little work has been performed on CO2 reactivity with naked metal cluster cations although Tan+ (n = 1–16) clusters show size-selective reactivity towards CO2.11
Metal oxides can, in principle, provide the possibility for more interesting and extensive chemistry to occur. Duncan and co-workers have investigated structures of NiO2(CO2)n+ complexes and only found evidence of molecularly bound CO24 as did Lang and coworkers in MnxOy(CO2)z+.33 We found similar binding in TaO2(CO2)n+ and NbO2(CO2)n+.34 Jiang and co-workers, however, found evidence for {CO3} moiety formation in infrared studies of YO(CO2)n+ (n ≥ 4) complexes.35
Here we report an infrared photodissociation (IR-PD) study of gas-phase Ho(CO2)n+ and HoO(CO2)n+ ion–molecule complexes in order to elucidate the binding behaviour of CO2 to Ho+ and HoO+ ions. Like most lanthanide metals, holmium typically adopts a +3 oxidation state in solution, the result of (relatively) facile removal of electrons from outer orbitals and larger enthalpies of hydration of higher charge states. Based on their ability to stabilize higher oxidation states, Ho+ cations were selected for this work in order to investigate whether they might activate CO2 more effectively than transition metal cations.
Fig. 2 shows a typical time-of-flight mass spectrum of the species generated in the source. The distribution is dominated by the target Ho(CO2)n+ and [HoOm(CO2)n]+ complexes and their Ar-tagged equivalents. Here, we employ square brackets, [X], to indicate complexes for which different chemical structures are plausible. The Ho(CO2)n+ complex signal falls rapidly with increasing n with little signal apparent by n > 4. By contrast, strong [HoO(CO2)n]+ signals are observed well beyond n = 15. The holmium target, as with many metals, contains enough surface oxide to generate the oxide complexes without any additional oxidant though we cannot rule out the possibility that Ho+ ions and/or small Ho(CO2)n+ complexes react with the CO2 itself to generate the HoO+ species.
Two variants of infrared dissociation experiments have been performed. Employing the in-line arrangement in Fig. 1c, target complexes are Ar-tagged, permitting the IR-PD spectra of a wide range of complexes to be recorded simultaneously in depletion of the parent [HoOm(CO2)n–Ar]+ signal:25,34
[HoOm(CO2)n–Ar]+ + hνIR → [HoOm(CO2)n]+ + Ar. | (1) |
In what follows, such spectra, recorded in depletion mode, are denoted [–Ar] and shown as blue data points. These spectra have been normalised to the largest fractional depletion signal in the spectrum.
We have recently added the ability to mass select the parent ion with the addition of a quadrupole mass filter as shown in Fig. 1b.36 This configuration has been used to record spectra for mass-selected ion–molecule complexes individually via loss of either a CO2 ligand or an Ar messenger depending on complex size:
[HoOm(CO2)n]+ + hνIR → [HoOm(CO2)n−1]+ + CO2, | (2) |
[HoOm(CO2)n–Ar]+ + hνIR → [HoOm(CO2)n]+ + Ar, | (3) |
Tuneable infrared radiation between 1650 and 2450 cm−1 is provided by a pulsed optical parametric oscillator/optical parametric amplifier (OPO/OPA, LaserVision, employing an AgGaSe2 crystal). Infrared action spectra are recorded either in parent ion depletion or daughter ion enhancement by comparing mass spectra recorded with and without infrared excitation as a function of infrared wavenumber. The data reported as cross-sections, σ where
(4) |
Ar-tag depletion experiments (process (1)) provide an effective way to record a consistent series of spectra for all complexes simultaneously. However, the signal-to-noise ratio of such spectra is often poor. Spectra recorded in fragment ion channels, processes (2) and (3), have better signal to noise provided a single photon is sufficient to cause ligand loss – usually the case for all but the very smallest complexes.
In order to aid the interpretation of the infrared spectra obtained, energetically low-lying structural isomers have been calculated using density functional theory (DFT) using the Gaussian 16 program40 in combination with a modified Kick3 algorithm.41,42 The UB3P86/Def2-TZVP functional/basis set was used.43,44 Other basis set and functional combinations were tested but minimal differences were found except for the relative energy ordering of some low-lying isomers and predicted vibrational band positions. Our interest here lies in better understanding the vibrational spectra recorded and thus we focus on the geometrical structures. We have not undertaken full spin–orbit calculations and structures are labelled according to their spin quantum number within a Russell–Saunders coupling scheme. Calculated vibrational frequencies have been scaled by a factor of 0.96 to better match with experiment, this number determined from the ratio of the experimental CO2 asymmetric stretch fundamental band at 2349 cm−1,29 and that calculated, 2441 cm−1. Line spectra have been convoluted with a Lorentzian function with a full width at half-maximum (FWHM) of 8 cm−1 to aid comparison with experimental spectra.
Our search of the potential energy surfaces reveals different distinct binding motifs with characteristic vibrational spectra with which to compare the experimental spectra. For example, Fig. 4 shows several structures identified for the Ho(CO2)2+ complex. The global minimum structure, MI, is a quintet spin state, linear, (D∝h) structure with one strong infrared allowed band. A similar triplet state structure lies 0.1 eV higher in energy. Two additional quintet structures, Dis and Dim, lie 0.11 eV higher in energy but no experimental evidence for either of these structures is obtained. Instead, the single spectral feature, near 2350 cm−1, is assigned to structure, MI. Substantial rearrangements and/or bond breaking would be required to form structures Dim and Dis and it is likely that formation of these is kinetically hindered.
Comparisons of experimental and simulated comparisons for other complex sizes are available in the ESI.† It is clear that CO2 only binds to the bare Ho+ ion in a simple M(η1–OCO) fashion for n = 1–4 consistent with many transition metal cations investigated previously.4,23–28 This simple electrostatic binding with negligible back donation from the cation accounts for the lack of CO2 activation.
By contrast with the spectra of the Ho(CO2)n+ complexes, most [HoO(CO2)n]+ spectra exhibit two spectral features. All spectra exhibit features in the region of the CO2 asymmetric stretch (2350–2400 cm−1) which are assigned to simple molecular M(η1–OCO) binding. The better resolved spectra recorded in the [–Ar] depletion channel (right hand panel, Fig. 5) exhibit clear spectral structure for n = 2–6 and, as for the Ho(CO2)n+ spectra, these Ar–tagged peaks are blue shifted by around 10–15 cm−1. For n ≥ 4 complexes, additional features are observed in the 1850–1760 cm−1 in the CO2 loss channel of mass–selected complexes. This same feature is also observed in the Ar loss channel for HoO(CO2)3–Ar+. Jiang and coworkers assigned similar features in the spectrum of YO(CO2)n+ complexes, to a C–O stretch in a {CO3}δ− moiety bound in a M(η2–O,O(CO)) fashion35 and we have investigated the possibility of this type of structure in our [HoO(CO2)n]+ complexes.
The computational potential energy surface search reveals two distinct structural motifs for each cluster size, the relative energies of which are shown in Fig. 6 for the [HoO(CO2)n]+ (n =1–4) complexes. For each complex an energetically low-lying Ho(CO3)(CO2)n−1+ type structure is calculated. These are labelled C isomers, in which C represents the presence of a carbonate radical anion like moiety . In these structures one CO2 molecule reacts with the HoO+ forming a C2v Ho(CO3)+ sub-structure in which two O-atoms bind to the metal ion (M(η2–O,O(CO))) as observed by Jiang and co-workers in the YO+ analogues. Zhao et al. labelled this a “carbonate structure” but, according to our calculations, the charge on the {CO3} group never exceeds −1 so we prefer a “carbonate radical anion” description. Our calculations predict this structure to be the lowest energy isomer for all n >1. This Ho(CO3)+ forms a core structure to which we calculate additional CO2 molecules bind at the Ho+ centre in a M(η1–OCO) configuration.
The second type of structure found is a conventional HoO(CO2)n+ type structure in which all CO2 molecules are molecularly-bound to the Ho end of HoO+ in M(η1–OCO) fashion. These structures (M isomers) exhibit spectral bands only in the 2350–2400 cm−1 region.
In all [HoO(CO2)n]+ (n = 1–4) complexes, both M and C isomers are strongly bound, by 0.9–1.5 eV, relative to their respective dissociation limits. In n = 1, the molecularly bound isomer, M, is found to be more stable than the carbonate radical anion, C form whereas for n = 2, both structures are found to be essentially isoenergetic. For all complexes n > 2, the C isomer is markedly lower in energy than the M isomer making the presence of the Ho(CO3)+ moiety the likely dominant structural motif.
We have identified plausible barriers to the M ↔ C interconversion in the case of n = 1–3. According to Jiang,35 the formation of proceeds via a 2 + 2 cycloaddition transition state from the molecularly-bound form. This requires a nucleophilic attack upon the carbon atom in CO2 by the electron-rich oxygen in HoO+. The transition state energy is stabilized (relative to the M isomer minimum) with increasing complex size, as the charge donated by each successive CO2 ligand is localised on the oxide, but in all cases remains submerged below any relevant dissociation threshold. We have little direct information on the internal energy distribution of our complexes but, due to the nature of the molecular beam expansion, it is not uncommon for us to see entrance channel complexes trapped behind submerged barriers on potential energy surface.
The calculated structures and charge distributions of both n = 1 isomers are shown in Fig. 6(b and c) and for larger complexes in the ESI.† For the n = 1 C isomer, significant charge is donated from the Ho atom (leaving it +1.50 e, compared with +1.41 e in the HoO+ ion) to the CO3 moiety. The terminal CO bond length, at 1.17 Å, is almost unchanged from free CO2 but the other two C–O bonds are markedly elongated at 1.38 Å reflecting significant activation. In the M isomer the CO2 ligand donates electron density, reducing the Ho partial charge slightly to +1.25 e but with minimal effect on structure of the CO2 moiety.
These potential energy surface calculations account qualitatively for the observations in Fig. 5. The lower wavenumber band, assigned to the C isomers appears persistently from n = 3 onwards by which stage the C isomer is calculated to lie significantly lower in energy than the M isomer. For the n = 3 complex itself, the band is clearly observed but only in –Ar loss from the tagged species, presumably because the CO2 binding energy exceeds the single photon energy. For similar reasons, we cannot exclude the presence of the carbonate radical anion for n = 1, 2 but there is no direct spectroscopic evidence for it.
It is noteworthy that CO2 loss is the only fragmentation channel observed in the untagged species. At no point was loss of CO3 observed for any species, this threshold lying at considerably higher energy.
In the CO2 loss spectra in Fig. 5 (red data points) the main bands around 2350 cm−1 narrows considerably between n = 1 and n = 4. The smaller complexes have higher ligand binding energies reflecting direct binding to the metal centre in the inner coordination shell. As a result more infrared photons are required to cause fragmentation. Beyond n = 4, CO2 molecules bind more weakly in a second solvation shell (see structures in the ESI†) and single photon absorption leads to facile ligand loss.
With this qualitative understanding of the action spectra on the basis of expected structures, Fig. 7 shows a comparison of experimental and simulated spectra for the [HoO(CO2)n]+ (n = 2–4) complexes based on the M and C isomers predicted. The main 2330–2380 cm−1 band is of limited use in identifying the structure as almost all calculated isomeric forms contain one or more molecularly-bound CO2 with ν3 fundamental band in this region (M(η1–OCO)). By contrast, the band between 1850 and 1900 cm−1 is a clear and unambiguous signature of the C isomer in each case and is clearly visible in the spectra of the n = 3, 4 complexes. This spectral band arises from the unique CO stretch in as shown by the vectors in Fig. 6c. As seen in Fig. 5, this band red shifts approximately linearly with increasing complex size from 1840 cm−1 (n = 3 complex) to 1760 cm−1 by n = 11 (see Fig. 8).
It is tempting to interpret two partially resolved peaks in the [–Ar] depletion spectrum (blue data points) of the [HoO(CO2)2]+ complex (at 2367 and 2378 cm−1, bottom panel, Fig. 7a) as the two components of the CO2 stretches in the M isomer (at 0.09 eV). This would be consistent with the absence of any band to the red, even in the –Ar loss daughter channel despite the calculated Ar binding of only 0.23 eV. It is dangerous, however, to draw conclusions from the absence of a band and we cannot rule out the possibility of both M and C isomers being present in the beam.
The –Ar loss spectrum (in green) of the tagged HoO(CO2)3+ complex (Fig. 7b) provides the first unambiguous evidence for the presence of the {CO3} moiety, its presence reproduced well in the simulated spectrum of the putative global minimum structure around 1850 cm−1. The calculated Ar binding energy to this complex is only 0.12 eV making it an ideal tag. Similarly, the two resolved features in the Ar depletion spectrum (blue) at 2364 cm−1 and 2382 cm−1 can be interpreted as a combination of both the M and C isomers with the latter not observed in the –CO2 daughter channel due to the higher CO2 binding energy (see Fig. 6). The signature of the {CO3} moiety is observed in the spectrum of all complexes n ≥ 4, though this doesn’t mean the presence of the M isomer can be ruled out. Several complexes (notably n = 4, 5, 6) exhibit detailed structure in the [–Ar] depletion spectra around the CO2 ν3 fundamental which could be interpreted as indicating the presence of multiple isomers and we have often seen such examples in our ablation source. Equally this structure could indicate different Ar binding sites and/or a reduction in symmetry. For complexes n ≥ 7, only one, narrow feature is seen in the [–Ar] depletion spectra suggesting a dominance of a single isomeric form.
The characteristic {CO3} stretch red-shifts as additional CO2 ligands bind (Fig. 5 and 8). Rather than reflecting increasing donation from the holmium atom, calculations show that this instead arises from increasing σ-donation from successive CO2 ligand addition. This serves to further activate the terminal CO bond, reduce the net charge on the Ho atom from 1.7 e (n = 1) to +1.3 e (n = 8) and reduce the barrier to M ↔ C isomer interconversion.
Footnote |
† Electronic supplementary information (ESI) available: Structural information on calculated isomeric forms; comparison of experimental and simulated spectra for additional complex sizes; potential energy surface details. See DOI: https://doi.org/10.1039/d2cp02862j |
This journal is © the Owner Societies 2022 |