Counter‐Intuitive Gas‐Phase Reactivities of [V₂]⁺ and [V₂O]⁺ towards CO₂ Reduction: Insight from Electronic Structure Calculations

Abstract

[V₂O]⁺ remains “invisible” in the thermal gas‐phase reaction of bare [V₂]⁺ with CO₂ giving rise to [V₂O₂]⁺; this is because the [V₂O]⁺ intermediate is being consumed more than 230 times faster than it is generated. However, the fleeting existence of [V₂O]⁺ and its involvement in the [V₂]⁺ → [V₂O₂]⁺ chemistry are demonstrated by a cross‐over labeling experiment with a 1:1 mixture of C¹⁶O₂/C¹⁸O₂, generating the product ions [V₂ ¹⁶O₂]⁺, [V₂ ¹⁶O¹⁸O]⁺, and [V₂ ¹⁸O₂]⁺ in a 1:2:1 ratio. Density functional theory (DFT) calculations help to understand the remarkable and unexpected reactivity differences of [V₂]⁺ versus [V₂O]⁺ towards CO₂.

The fleeting existence of [V₂O]⁺ as a reactive intermediate in the oxidation of [V₂]⁺ by CO₂ to form [V₂O₂]⁺ has been identified; the counter‐intuitive oxidation behavior of [V₂]⁺ versus [V₂O]⁺ and the quite enormous reactivity differences have been traced back to the supportive role of the bridging oxygen atom in [V₂O]⁺.

Rolf Huisgen has shaped physical organic chemistry in a nearly unparalleled fashion.1 His extensive, seminal work on, for example, 1,3‐dipolar cycloadditions1b forms a pillar of contemporary mechanistic organic chemistry, not to speak of their role in the synthesis of heterocyclic compounds. By using the then available experimental repertoire, the Huisgen school unraveled the finest mechanistic details of numerous chemical processes, and their work gives testament for chemist generations to come.2

A familiar kinetic scheme of a multi‐step process is depicted in Equation 1, and one problem here concerns the identification of “intermediate” [B]. This holds true in particular when the rate constants k ₁ and k ₂ differ significantly, that is, k ₂ ≫ k ₁.

A typical example is provided by the gas‐phase chemistry of the [Pd]⁺/CH₃I couple which gives rise to the formation of [Pd(CH₂I)]⁺.3 In this case, the product ion is not formed by the entropically favored direct cleavage of the C−H bond of CH₃I.3a Rather, in the first step a short lived [Pd(CH₃)]⁺ intermediate is generated, which reacts further with CH₃I in a very fast metathesis process to form [Pd(CH₂I)]⁺ and CH₄.3b Recently, in the context of combined experimental/computational studies on the timely topic of CO₂ activation,4 we came across a similar, unexpected situation in the thermal gas‐phase reactions of bare diatomic [V₂]⁺ with CO₂.

The cluster [V₂]⁺ was formed by supersonic expansion of helium into a vanadium plasma generated by laser ablation/ionization of a vanadium disk using a Nd:YAG laser, operating at 532 nm inside the external cluster source of a Fourier‐transform ion cyclotron resonance (FT‐ICR) mass spectrometer as described previously (for details, see the Supporting Information).5 A fraction of the ion population was then guided by a static ion optical system into the ICR cell. Next, in a sequence of pulses, argon was admitted to the ICR cell such that the ions collide on average about 1×10⁵ times with argon. This procedure ensures thermalization of hot ions and quenching of excited electronic states. After mass‐selection of the thermalized [V₂]⁺ ions, they were reacted with CO₂ or C¹⁸O₂ at a constant pressure low enough to ensure single collision conditions. The elementary compositions of the charged species have been confirmed by exact mass measurements using high resolution mass spectrometry.

The result of the reaction of [V₂]⁺ with CO₂ at a partial pressure of 1.0×10⁻⁷ mbar and a reaction time of 10 s inside the ion trap of the FT‐ICR machine is shown in Figure 1 b. In addition to the signal of the starting material [V₂]⁺ (signal A, Figure 1), a signal D appears, which has been assigned the formula [V₂O₂]⁺ by exact mass measurements; also, very small amounts of atomic [V]⁺ are present (signal B). Formally, the formation of [V₂O₂]⁺ corresponds to the loss of one carbon atom. If C¹⁸O₂ acts as a reaction partner, signal D from Figure 1 b is shifted by four mass units (Figure 1 c; signal E). Clearly, oxygen atom transfer (OAT) from CO₂ to the vanadium cluster takes place. However, since a carbon atom represents an extremely poor leaving group, the suggestion was raised as to whether the formation of [V₂O₂]⁺ from the [V₂]⁺/CO₂ couple is the result of a multi‐step process, although the spectra in Figure 1 b,c do not exhibit any clue for the involvement of conceivable intermediates. The assumption for the operation of a multi‐step process is corroborated by a crossover‐labeling experiment using a 1:1 mixture of C¹⁶O₂ and C¹⁸O₂: The appearance of [V₂ ¹⁶O¹⁸O]⁺ (signal F in Figure 1 d) together with the other two isotopomers in a 1:2:1 ratio verifies this hypothesis. According to Equations (2) and (3) and in line with Figure 1 d, one has to conclude that [V₂]⁺ reacts consecutively with two CO₂ molecules via an OAT by the release of one CO molecule each time; the monoxide [V₂O]⁺ serves as a short‐lived intermediate (Scheme 1). Further oxidation processes according to Equation (4) in Scheme 1 do not occur at ambient temperature.

Figure 1.

Representative mass spectra: a)–e) are for the thermal reactions of [V₂]⁺ with a) Ar at 1.0×10⁻⁷ mbar, b) C¹⁶O₂ at 1.0×10⁻⁷ mbar, c) C¹⁸O₂ at 1.0×10⁻⁷ mbar, and d) a 1:1 mixture of C¹⁶O₂ and C¹⁸O₂ at 2.0×10⁻⁷ mbar after a reaction time of 10 s, respectively; e) C¹⁸O₂ at 1.0×10⁻⁷ mbar and a reaction time of 1s; f) and g) represent the thermal reactions of [V₂O]⁺ with f) Ar at 4.0×10⁻⁹ mbar, g) C¹⁸O₂ at 4.0×10⁻⁹ mbar, after a reaction time of 2s, respectively. All x‐axes are scaled in m/z, and the y‐axes are normalized relative ion abundances.

Scheme 1.

Multi‐step activation of CO₂ by [V₂]⁺ with [V₂O]⁺ serving as a fleeting intermediate.

In the [V₂]⁺/CO₂ couple, the supposed intermediate [V₂O]⁺ can only be detected experimentally as an extremely weak signal under highly optimized conditions of both pressure and reaction time (signal G in Figure 1 e). However and fortunately, [V₂O]⁺ can be synthesized independently and in sufficient quantities in the external ion source by adding traces of O₂ to the helium buffer gas (Figure 1 f). As shown in Figure 1 g, the so‐formed [V₂ ¹⁶O]⁺ indeed reacts with C¹⁸O₂ to produce [V₂ ¹⁶O¹⁸O]⁺ by the expulsion of C¹⁸O. The collision efficiency6 φ of that reaction is close to 100 %. The high reactivity of [V₂O]⁺ explains why, [V₂O]⁺ from the reaction of [V₂]⁺ with CO₂, remains elusive. As soon as the monoxide has formed within the ICR cell, it immediately reacts with another CO₂ molecule. The high reactivity of the [V₂O]⁺ intermediate, generated from the [V₂]⁺/CO₂ couple, is further enhanced due to the substantial exothermicity of the process (see below) and the fact that, owing to the absence of a heat bath, part if not most of the excess energy is stored in the ion by partitioning the energy released between [V₂O]⁺ and CO. To distinguish possible reactions of the charged particles with omnipresent background gases from those of the selected substrate, spectra with argon were recorded as well. As shown in Figure 1 a,f side reactions with background gases like trace amounts of oxygen are limited to the generation of small quantities of products typical for reactions of [V₂]⁺ with O₂,9e which are more pronounced in [V₂O]⁺, since O₂ has been added to the He buffer gas.

The rate constants k ₄([V₂]⁺/CO₂) and k ₅([V₂O]⁺/CO₂) for reactions shown in Equations (2) and (3) in Scheme 1 were measured and they amount to 3.8×10⁻¹² (efficiency φ=0.5 %) and 8.8×10⁻¹⁰ cm³ molecule⁻¹ s⁻¹ (φ=1.15), respectively; these numbers translate to a ratio of k ₅/k ₄>230. Owing to an uncertainty in the determination of the absolute pressure of CO₂, an error of ±30 % is associated with the rate measurements.5b For the relative rate constants, however, the error is much smaller, typically around ±5 %.

The considerable reactivity difference of [V₂]⁺ versus [V₂O]⁺ is unexpected and counter‐intuitive on the ground that the reactions in Equations (2) and (3) constitute redox processes in which electrons are transferred from the metal‐containing cluster ions to the CO₂ molecule. Therefore, one would expect that [V₂]⁺ reduces CO₂ to CO more readily than the already oxidized and electron‐depleted [V₂O]⁺. Consequently, k ₄ should be larger than k ₅; this is not the case. Why does the reactivity of [V₂]⁺ and [V₂O]⁺ not reflect their respective oxidation states? Further, does [V₂O]⁺ react mechanistically the same way with CO₂ as [V₂]⁺? These questions have been addressed by density functional theory (DFT) calculations.

Earlier theoretical work had already shown that a description of vanadium dimers constitutes a notoriously difficult case to be dealt with by DFT calculations.7 This dilemma holds true as well for [V₂O₂]⁺.8 While for [V₂O₂]⁺ the gas‐phase IR spectrum suggests a planar, four‐membered V‐O‐V‐O ring structure of C _s symmetry,8b an unambiguous assignment of the electronic ground state (²A′ or ⁶A′) remained unresolved at the DFT levels employed.8b, 8c The problems are compounded by the observation that the gas‐phase chemistry of vanadium ions9 is often marred by a two‐ or multi‐state reactivity scenario.10 As to the bare [V₂]⁺ cluster, due to its multireference character7, 11 we applied the ZORA‐NEVPT2(9e,12o)/ZORA‐def2‐QZVPP Scheme (Table S1 in the Supporting Information).12 According to these high‐level calculations, the ground state of [V₂]⁺ corresponds to ⁴Σ_g ⁻. This assignment agrees well with experimental data and other, less sophisticated calculations.13 The lowest excited state of [V₂]⁺ is a doublet (²Δ) and exceeds the ground state by only about 3 kJ mol⁻¹ (Table S2). Therefore, it is very likely that under the given experimental conditions a mixture of two electronic states, ^2,4[V₂]⁺, is present, in which the quartet may dominate. We mention in passing that low‐energy excited states of the neutral V₂ dimer were identified by Hübner and Himmel using matrix isolation spectroscopy.11 To construct a simplified potential energy surface (PES) for the [V₂]⁺/[V₂O]⁺/CO₂ systems, extensive quantum chemical calculations were performed at the ZORA‐M06L/ZORA‐def2‐QZVPP//M06L/def2‐TZVP level of theory. Electronic structure calculations using the multi‐reference ZORA‐NEVPT2 scheme are not feasible technically. Rather, the M06L functional was chosen based on the exhaustive testing carried out by Truhlar and co‐workers.7 This functional performed best also among all the density functionals we benchmarked (see Tables S2, S6 and S7). In addition, M06L delivers fairly good results for the electronic states of [V₂O]⁺ and [V₂O₂]⁺ (see Tables S3 and S4).

Figure 2 displays the PESs of the most favorable pathways for the reduction of CO₂ with vanadium ions having doublet spin multiplicity as well as selected structural parameters of stationary points along the reaction coordinate. This particular spin state was chosen on the ground that along the whole transformation [V₂]⁺ → [V₂O₂]⁺ the spin multiplicity was conserved. It should be mentioned at this point, however, that because of the strongly correlated electrons, the energetic and geometric properties of some stationary points with different spin multiplicities are located fairly close to each other. Therefore, it is quite possible that there is a superposition of the doublet with the quartet state and vice versa, facilitating a multi‐state reactivity scenario pending on the efficiency of spin‐orbit coupling (see below for a further comment).10 However, our aim is not to provide a quantitatively correct description of the whole PESs; rather, we aim at offering a qualitative understanding of the amazing reactivity behavior in the [V₂]⁺/[V₂O]⁺/CO₂ systems. As the reactivity pattern proved to be quite complex, herein we limit ourselves to the main results. For further details, see Figures S2 and S3.

Figure 2.

Schematic potential energy surfaces (ΔH _298K in kJ mol⁻¹) obtained at the ZORA‐M06L/ZORA‐def2‐QZVPP//M06L/def2‐TZVP level of theory for the reactions of a) ²[V₂]⁺ and b) ²[V₂O]⁺ with CO₂, respectively. Key structures with selected geometric parameters and symmetries are also provided. Bond lengths [Å]; angles [°]. The relative energies of ^2,4[V₂]⁺ were obtained at the ZORA‐NEVPT2(9e,12o)/ZORA‐def2‐QZVPP level of theory. For details, see the Supporting Information.

As displayed in Figure 2 a, at the doublet surface the exothermic OAT14 in the system [V₂]⁺/CO₂ commences with the barrier‐free formation of an encounter complex 1 a by coordinating the incoming ligand to only one of the two vanadium atoms in an end‐on (η ¹) mode. Subsequently, intermediate 2 a is generated by surmounting transition state TS1 a, which is accompanied by a substantial elongation of the V−V bond from about 1.6 Å to 2.5 Å; this bond length does not change that much further in the remaining steps. In contrast, significant structural reorganization happens for CO₂ at these stages. The initially linear substrate gets bent with an O‐C‐O angle of 141° in 2 a, and the C−O bond length increases from 1.17 to 1.27 Å; clearly, CO₂ is already markedly activated in this initial step. Next, the complete scission of one C−O bond of the CO₂ moiety occurs along the route 2 a → TS2 a → 3 a. The reaction is completed by the release of CO to generate ²[V₂O]⁺. The energetically submerged transition state TS1 a, as part of the reduction sequence of the first CO₂ molecule, is with 29 kJ mol⁻¹ located below the entrance asymptote and represents the rate‐limiting step of the entire reaction.

As to the exothermic reaction of [V₂O]⁺ with CO₂ and as displayed in Figure 2 b, ²[V₂O]⁺ is approached by carbon dioxide as well, to then also form an η ¹ encounter complex 1 b; subsequently, this complex has sufficient internal energy to easily surmount transition state TS1 b and generate 2 b. Next, the complete cleavage of the activated C−O bond takes place along the sequence 2 b → TS2 b → 3 b, and release of CO from 3 b provides ²[V₂O₂]⁺. In contrast to the [V₂]⁺/CO₂ couple, here it is transition state TS2 b (−48 kJ mol⁻¹) in which the C−O bond of the CO₂ molecule is broken and which represents the rate‐limiting step in the reduction of CO₂ to CO.

At first glance, the reductions of the two CO₂ molecules by [V₂]⁺ and [V₂O]⁺ do not seem to differ too much in terms of structural features of the intermediates. So, why do they exhibit quite pronounced reactivity differences? As will be shown, these features can be attributed to both intrinsic structural and electronic properties. As suggested by a reviewer, one reason may be due to the spin situation. For the [V₂O]⁺/CO₂ couple the OAT is both exothermic and spin‐allowed (Figure 2 b); the whole reaction is confined to the doublet surface. This must not necessarily be the case for the [V₂]⁺/CO₂ system. While for electronically excited ²[V₂]⁺ the reaction with CO₂ is also both exothermic and spin‐allowed (Figure 2 a), for the ground state ⁴[V₂]⁺ a quartet → doublet spin change has to occur along the overall exothermic reaction coordinate (Table S5). If this electronic reorganization is inefficient, as has been reported for the analogous exothermic [V]⁺/CO₂ conversion,9f as well as for the [V]⁺/N₂O couple,9a the reaction of [V₂]⁺ with CO₂ will be slowed down as well. Further, while thermochemical arguments on the endo‐ versus exothermicity of a reaction,7 based only on DFT calculations, may be considered with caution, the insensitivity of the [V₂]⁺/CO₂ reactivity to intentional excitation of the [V₂]⁺ projectile, strongly suggests that process (2) is exothermic and that its much reduced reactivity is not due to a thermodynamic impediment.

Indeed, there are other factors to be considered to explain the much higher reactivity of the [V₂O]⁺/CO₂ couple. Let us begin by looking at the charge distribution in the ionic actors. According to Figure 3 a, the reactive vanadium atoms in ²[V₂O]⁺ carry more positive charge as compared to ²[V₂]⁺. It is therefore expected that [V₂O]⁺ will act as a worse electron donor for the reduction of CO₂ compared to [V₂]⁺.15 However, this is not the whole truth. In fact, the presence of the bridging oxygen atom in [V₂O]⁺ actually plays a supportive role in the CO₂ activation which overcompensates the unfavorable charge situation.

Figure 3.

a) NBO‐calculated charges at the vanadium atoms coordinated to CO₂ and b) development of the V−V bond length [Å] in the reactions of ²[V₂]⁺ and ²[V₂O]⁺ with CO₂.

First, the presence of the oxygen atom induces a structural rearrangement of the active site, thus, providing a “preorganized” structure, which reduces substantially the energy demand to reach transition state TS1 b; this effect is well‐known from enzymatic processes.16 The calculated Wiberg Bond Index (WBI)17 of 4.5 indicates a rather strong and short V−V bond for [V₂]⁺ (1.63 Å); a much smaller value of the WBI of 1.2 and an elongated V−V bond of 2.53 Å holds true for 2 a (Figure 2 and Figure 3 b). As a consequence, stretching of the V−V bond in [V₂]⁺ in the step 1 a → 2 a is quite energy‐demanding (Figure 2 a). In contrast, the corresponding transformation 1 b to 2 b in [V₂O]⁺ requires much less energy as the V−V bond is already sufficiently elongated due to the presence of the oxygen bridge. Thus, [V₂O]⁺ possesses a “prepared” structure which greatly enhances the reduction of CO₂. In the remaining stages of the reaction there is no further dramatic change in the properties of the V−V bond, as indicated by the rather small oscillations of the WBI's around a value of 1. The moderate movements of the two vanadium atoms relative to each other do not require that much energy; rather, it is this structural change of the “active” site which assists in lowering TS1 b as compared with TS1 a.

There is yet another aspect to be considered. In ²[V₂O]⁺ the presence of an oxygen atom raises the energy level of the electron‐donating d‐orbitals. As shown in Figure 4, in TS1 a of [V₂]⁺ the overlap between the lobes of the doubly‐occupied πd_xz/d_xz‐orbital of the metal center and the vacant π*‐orbital of the carbon atom of CO₂ has yet to be established. In contrast, for the early transition state TS1 b, the doubly‐occupied πd_xz/d_xz‐orbital is already delocalized quite extensively to the carbon atom of CO₂, and the root cause for this difference can be traced back to orbital energy differences. The introduction of an oxygen bridge raises the πd_xz/d_xz‐orbital energy from −10.5 eV in [V₂]⁺ to −9.5 eV in [V₂O]⁺; this greatly eases the transfer of the d‐electrons into the empty π*‐orbital of CO₂ thus favoring the reduction of [V₂O]⁺.

Figure 4.

The frontier orbitals of the transition states of TS1 a,b from the doublet surface of the CO₂ activation. Arrows and rounded rectangles indicate the overlap regions between the doubly occupied π‐electron‐donating orbital of the metals and the lowest empty orbital of CO₂.

As shown repeatedly,16, 18 the presence or absence of “prepared states” are crucial in controlling the structural, electronic and energetic features of transition states. While there exist quite a number of examples for this scenario in the area of gas‐phase C−H bond activation,19 to our knowledge the present study describes for the first time this effect in the context of CO₂ activation.

In conclusion, in the present experimental/computational work we report and explain quite a few unexpected observations:

1. A fleeting [V₂O]⁺ intermediate is involved in the multistep oxidation of [V₂]⁺ by CO₂.

2. The counter‐intuitive oxidation behavior of [V₂]⁺ versus [V₂O]⁺ and the quite enormous reactivity differences of the two cluster ions have been traced back to the supportive role of the oxygen bridge in [V₂O]⁺. This cluster oxide, in contrast to bare [V₂]⁺, exhibits structural and electronic features which aid in the activation of CO₂. While thermochemical aspects are not likely to cause the relative “inertness” of [V₂]⁺, an inefficient quartet → doublet spin change in the [V₂]⁺/CO₂ couple may well matter in slowing down its reactivity.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

J. Li, C. Geng, T. Weiske, H. Schwarz, Angew. Chem. Int. Ed. 2020, 59, 12308.