1 Introduction

Since the seminal publication of Kappes and Staley in 1981 on “Gas-Phase Oxidation by Transition-Metal Cations” [1], various aspects of this topical problem have been addressed in numerous reviews [2–25]. The enormous interest is due to the fact that gas-phase studies on ‘isolated’ reactants provide an ideal arena for probing experimentally the energetics and kinetics of a chemical reaction in an unperturbed environment at a strictly molecular level without being obscured by difficult-to-control or poorly understood solvation, aggregation, counterions and other effects, thus providing an opportunity to explore the concept of single-site catalysts directly [26–33]. Further, in these experiments reactive intermediates can be characterized in detail, mechanisms uncovered, and questions addressed on how factors such as cluster size and dimensionality, stoichiometry, oxidation state, degree of coordinative saturation, aggregation, or charge state affect the outcome of a chemical process. Active or single-sites in heterogeneous catalysis are usually rather ill-defined and often characterized by dangling bonds, kinks, steps, defects, or nano-sized particles; probing them experimentally is all but trivial [31, 32, 34] and their identification constitutes one of the intellectual cornerstones in contemporary catalysis. As ‘naked’ gas-phase species are, in general, much more reactive than their condensed-phase counterparts, these studies will, in principle, of course never account for the precise kinetic and mechanistic details which prevail at a surface or in the condensed phase. Yet, complemented by appropriate computational studies, gas-phase experiments have proved meaningful, on the ground that they permit a systematic approach to address the above mentioned questions and provide a conceptual framework. The DEGUSSA process, that is the platinum-mediated coupling of CH4 and NH3 to generate HCN [35], may serve as a good example. Mass-spectrometry based experiments [36, 37] suggested (i) the key role of CH2NH as a crucial transient, and (ii) the advantage of using a bimetallic rather than a pure platinum-based catalysts for the C–N coupling step in competition with undesired soot formation; the existence of CH2NH was later confirmed by in situ photoionization studies [38] and currently used catalysts contain silver–platinum alloys. Obviously, each and every information and insight that help to optimize or improve the often trial-and-error based strategies on catalyst developments [39] are highly welcome.

In this invited perspective, we focus on selected aspects of four gas-phase catalytic reactions all of which are mediated by ionic species under thermal conditions; they encompass (i) the coupling of carbon–carbon bonds, (ii) the CO → CO2 conversion at ambient conditions, (iii) the activation of hydrocarbons, and (iv) the selective oxidation of methanol to formaldehyde.

While we will refrain from describing the various experimental techniques (which are available from the references given), we will rather focus on the elucidation of the often intriguing mechanisms.

2 Metal-mediated Formation of Carbon–Carbon Bonds

Cyclooligomerizations of unsaturated hydrocarbons, in particular assembling them to form benzene, are versatile reactions for the synthesis of aromatic compounds [40]. Although these reactions are quite exothermic, they are usually hampered by large barriers if non-activated hydrocarbons are employed. Transition-metal complexes have been found to facilitate these processes in the condensed phase, and even single Ag, Rh, and Pt atoms supported on a MgO(001) surface were found to bring about acetylene trimerization at ambient conditions [41]. Also in the gas phase, certain ‘bare’ transition-metal cations M+ affect these cyclization processes, and the catalytic reactions are often accompanied by dehydrogenation steps. The most classical example of the stepwise route [42] correspond to the dehydrogenative gas-phase trimerization of C2H4 by atomic W+ [43], U+ [44], Fe+ [45–47], or Fe n + cluster [48–51]. The unique reactivity of the Fe4 + cluster, in comparison to other cluster sizes of iron or the complete absence of reactivity of Ni4 + towards C2H4 already illustrates the often-noted non-scalability of cluster properties—in fact, each atom counts [11]!

As depicted in Fig. 1, the oligomerization sequence commences with the formation of a cationic metal-ethyne complex. In the next, often rate-limiting step, the M(C2H2)+ intermediate brings about dehydrogenation of a further ethene molecule to produce M(C4H4)+; for some metal cations M+, for example U+, there is experimental evidence that this complex contains a C4 unit rather than two separate C2H2 ligands [44], while in the M(C4H4)+ complexes, generated by association of atomic Fe+ or Ni+ with C2H2, the preferred structure corresponds to M(C2H2)2 + [44, 52]. Addition of a third C2H4 molecule results in the formation of a metal-benzene complex as evidenced by numerous experimental findings. Although the process M(C4H4)+ + C2H4 → M(C6H6)+ + H2 is rather exothermic, the heat of reaction liberated is usually not sufficient to overcome the quite significant bond dissociation energy of M+–C6H6 and to release benzene from the catalyst; as a consequence, regeneration of the active catalyst M+ is not observed under strictly thermal conditions and can only be achieved by external energy supply in form of e.g. collisional induced dissociation (CID). Of course, in a ‘perfect’ catalytic cycle the catalyst should be regenerated in the reaction without additional supply of energy. This is conveniently achieved in gas-phase experiments by employing ‘high-energy’ reactants [42, 45, 46]. For example, substituting C2H2 for C2H4 as a reactant increases the exothermicity of the final step M(C4H4)+ + C2H2 → M+ + C6H6 by approximately the heat of dehydrogenation of C2H4, i.e. 42 kcal mol−1. As this additional reaction energy is stored completely in the isolated encounter complex, spontaneous detachment of C6H6 is possible. While these gas-phase studies have certainly uncovered interesting mechanistic details, there is no doubt that many important features, e.g. the particular role of spin states or that of the geometric structures of the catalytically active metal-ion clusters remain to be resolved as indicated by a comprehensive DFT study of the Fe n 0,+/(C2H2) m systems (n = 1–4; m = 2, 3) [47].

Fig. 1
figure 1

Dehydrogenative oligomerization of C2H4 and formation of benzene by consecutive gas-phase ion-molecule reactions (adapted from Ref. [42])

Full size image

An entirely different type of carbon–carbon coupling has been reported for the oxidative dimerization of methane, Eq. (1); this large-scale conversion is conventionally performed in a heterogeneous process at temperatures above 650 °C [53, 54] and the challenge is to suggest a catalyst that operates under more benign conditions.

$$ 2{\text{CH}}_{4} + {\text{ O}}_{2} \to {\text{C}}_{2} {\text{H}}_{4} + \, 2{\text{H}}_{2} {\text{O}} $$
(1)

While some intriguing mechanistic aspects of the rate limiting C–H bond activation step in the metal-oxide mediated methane coupling will be described in Section 4, here we mention a few notable cluster-size and temperature effects which were reported recently by Lang et al. [55–57]. In contrast to atomic ground-state Au+ (1 S 0), which is unreactive towards CH4 at ambient conditions [58], isolated Au2 + clusters bring about C–C coupling of methane to yield ethene in full thermal catalytic cycles, and for this particular cluster size the chemoselectivity in terms of product formation depends crucially on the temperature and the absence or presence of oxygen. For the metal clusters of palladium and platinum, both varying in size, a much lower selectivity has been reported [56, 59–62], and for the Au2 +/CH4/O2 system, detailed experimental investigations, complemented by first-principle simulations, revealed the coupled catalytic cycles shown in Fig. 2.

Fig. 2
figure 2

a Relative intensities of the products Au2(C2H4)+ and Au2(C3H8O2)+ as a function of temperature in the system Au2 +/CH4 (p = 0.05 Pa)/O2 (p = 0.10 Pa); b coupled catalytic cycles for the temperature-tunable formations of CH2O and C2H4 from CH4 (adapted from Ref. [57])

Full size image

In the absence of O2 or at higher temperature, at which O2 does not readily adsorb on Au2 +, the mass-selected Au2 + cluster reacts with a first CH4 molecule to yield collisionally stabilized Au2(CH4)+; C–H bond activation and dehydrogenation do not take place but require the adsorption of a second methane molecule to form Au2(C2H4)+ and 2H2. Obviously, these processes are the outcome of a co-operative action of both ligands. Oxidative coupling is observed only at temperatures >250 K, and the energy-demanding release of C2H4 from Au2(C2H4)+ requires both higher temperatures (~300 K) as well as the adsorption of yet another molecule of CH4. It is this very step that closes cycle II and regenerates the active catalyst Au2(CH4)+. In the presence of O2 and at lower temperature (210 K) considerable changes in the product distribution take place. Two new, oxygen-containing products, Au2(CH4)2O2 + and Au2(C3H8O2)+, are formed at the expense of Au2(CH4)+ and Au2(C2H4)+. The kinetic analyses in combination with labeling experiments and computational studies suggest the catalytic formation of formaldehyde according to cycle I and Eq. (2). As shown recently in a different context, a stoichiometric, direct conversion of CH4 to CH2O at room temperature can also be achieved by using Al2O3 + [63].

$$ 2{\text{CH}}_{4} + \, 2{\text{O}}_{2} \to 2{\text{CH}}_{2} {\text{O }} + \, 2{\text{H}}_{2} {\text{O}} $$
(2)

3 Low-temperature, Catalytic Oxidation of CO

Catalytic conversion of harmful gases, produced in fossil-fuel combustion, such as CO or the oxides of nitrogen, into nitrogen and carbon dioxide, is of utmost importance both environmentally and economically. While these redox reactions are exothermic, for example Δr H = −87.3 kcal mol−1 for the process N2O + CO → N2 + CO2, they do not occur directly to any measurable extent at either room or elevated temperatures due to high barriers exceeding 47 kcal mol−1 for the N2O/CO couple [64]. Catalysts are required to reduce these barriers, and the first example of a homogeneous catalysis in the gas phase in which atomic transition-metal cations bring about efficient N2O reduction by CO was reported by Kappes and Staley as early as 1981 [1]. Later, numerous other atomic main-group and transition-metal cations have been tested as catalysts [64–68]. Out of 59 atomic cations investigated, 26 systems for the catalysis of O-atom transport were shown to lie within the ‘thermodynamic window of opportunity’ [11] defined by the oxygen affinities (OA) of N2 and CO, with OA(N2) = 40 and OA(CO) = 127 kcal mol−1. Catalytic activity, however, was observed with only ten atomic cations, namely Ca+, Fe+, Ge+, Sr+, Ba+, Os+, Ir+, Pt+, Eu+, and Y+. The remaining 16 cations, which meet the thermodynamic criteria for oxygen-atom transport (Cr+, Mn+, Co+, Ni+, Cu+, Se+, Mo+, Rn+, Rh+, Sn+, Te+, Re+, Pb+, Bi+, Tm+, and Lu+), reacted too slowly during either the formation of MO+ or its reduction by CO. As shown earlier [69], this is due to a kinetic barrier resulting from an inefficient, spin–orbit coupling mediated curve crossing that is required for the change in multiplicities [9].

Metal-mediated, catalytic gas-phase oxidation of CO by nitrogen oxides is not confined to N2O as demonstrated by Bohme and co-workers [70]. Also NO and NO2 can be reduced, and taken together, these three systems constitute rare examples of metal-cation catalyzed reductions of NO2, NO, and N2O coupled with the formation of an N–N bond during the termolecular reductive dimerization of NO. As shown in Fig. 3, overall NO2 is reduced by CO to N2 catalyzed efficiently by any of the three atomic metal cations M+ (M = Fe, Os, Ir).

Fig. 3
figure 3

Catalytic cycles for the room-temperature homogeneous reduction of nitrogen oxides by CO, mediated by the atomic transition-metal cations Fe+, Os+, and Ir+ (adapted from Ref. [70])

Full size image

In the context of ‘catalyst poisoning’, studies with platinum clusters revealed remarkable effects of both the cluster size and the charge state for the CO/N2O couple [71–73]. For example, for the Pt7 + cluster, the active species in the redox process are Pt7 +, Pt7O+, Pt7O2 +, and Pt7CO+ with a turnover number >500 in their thermal reaction with CO. Adsorption of more than one CO molecule to the Pt7 + cluster, however, completely quenches the catalytic activity, so that an elevated CO partial pressure has to be avoided [71]. Pronounced charge-state effects were reported for the Pt4 +/− clusters, which are known as the least reactive for the cationic and the most reactive one for anionic platinum clusters [60, 72, 74]. Also for the latter, the catalytic activity terminates as soon as two or more CO molecules are adsorbed on the cluster. The enormous reactivity differences for the anionic versus cationic Pt4 cluster ions have been addressed in theoretical studies. Some of the differences are due to geometrical features showing a near planar anion and a structurally distorted tetrahedral cation. The former provides significantly stronger bonds than Pt4 + with both reactants N2O and CO [72]. In addition, for the Pt4 +/CO/N2O system there are kinetic barriers for both the doublet and quartet spin states that prevent the reaction to occur under thermal conditions [73].

More recently, the redox-features of heteronuclear metal-oxide clusters were exploited to bring about catalytic oxidation of CO by N2O at room temperature [75], and the bimetallic oxide cluster couple AlVO3 +/AlVO4 + may serve as a good example. As shown in Fig. 4, AlVO4 + in the presence of CO is reduced to AlVO3 +, and if N2O is added, the reverse reaction occurs. Both processes are clean and proceed with efficiencies of 59 and 65 % relative to the collision rate, respectively. As no by-products are formed, the turnover number of the catalytic cycle is principally infinite but in reality limited by side reactions with background impurities e.g. hydrogen-atom abstraction from water or residual hydrocarbons [75, 76].

Fig. 4
figure 4

Fourier-transform ion-cyclotron resonance (FT-ICR) mass spectra showing the thermal reactions of a AlVO4 + with CO (t = 3 s) and c AlVO3 + with N2O (t = 2 s); the pressures of CO and N2O in each case are 8 × 10−7 Pa. The relative intensities of AlVO4 + and AlVO3 + with increasing reaction times are shown in “b” and “d”, respectively (adapted from Ref. [75])

Full size image

Insight in the actual mechanism and in particular the question of the active site in the heteronuclear AlVO4 + cluster is provided by DFT calculations [75]. As shown in Fig. 5, the uncatalyzed reaction via transition state TS1 is much too high in energy to play a role at ambient conditions. In contrast, the catalytic conversion, which takes place at the doublet ground state of AlVO4 +, commences by an initial, barrier-free binding of the carbon atom of CO to the radical oxygen atom of the Al–Ot • moiety to generate intermediate 1. This species is formed with an internal energy of 71.2 kcal mol−1 below the entrance channel; as the energy in an ‘isolated’ system cannot be dissipated to a heat bath, liberation of CO2 occurs spontaneously requiring only 28.9 kcal mol−1. The catalytic cycle is completed by re-oxidation of AlVO3 + with N2O; as shown in Fig. 5, this reaction is also straightforward without barriers exceeding the energy of the entrance channel. Interestingly, this highly efficient catalytic cycle of a redox couple cannot be promoted by the non-radical terminal oxygen atom of the V = Ot moiety of AlVO4 +. Computational findings demonstrate that this pathway is kinetically and thermochemically much less favorable than the one commencing at the Al–Ot • unit (Fig. 6). Thus, the combined experimental/computational study reveals the existence and operation of an ‘active site’ already in a rather small heteronuclear cluster. The particular and crucial role of oxygen-centered radicals in various other bond-activation processes will be addressed in more detail in Section 4 [77]. With regard to CO oxidation, it may suffice to mention that also cationic clusters of the general composition (ZrO2) n + (n = 2–5) as well as the anionic systems Zr n O2n−1 − (n = 1–4) bring about room-temperature catalytic oxidation of CO. Once again, it is a highly localized terminal oxygen radical-center which acts as the active site [16, 78].

Fig. 5
figure 5

Potential-energy surfaces (B3LYP/TZVP) for the oxidation of CO by N2O in the absence (red line) and the presence of AlVO4 + (blue/green lines). The relative energies ΔE are given in kcal mol−1 and corrected for zero point energy. The blue and green profiles correspond to the reaction of AlVO4 + with CO and of AlVO3 + with N2O, respectively. TS transition structure; R = CO + N2O + AlVO4 +; P = CO2 + N2 + AlVO4 + (adapted from Ref. [75])

Full size image
Fig. 6
figure 6

The reaction [OtV(μ-O)2AlOt]+ + CO → [V(μ-O)2AlOt]+ + CO2 (blue lines) versus [OtV(μ-O)2AlOt]+ + CO → [OtV(μ-O)2Al]+ + CO2 (red line) (adapted from Ref. [75])

Full size image

In the context of catalytic, low-temperature CO oxidation, experimental and computational studies of free gold clusters occupy a central position in the literature [10–12, 25]. This is due to several factors: (1) Generally, the reactivity of a heterogeneous process is a complex convolution of the properties of metal cluster and those of the support. Therefore, the investigations of free, gas-phase clusters may help to reveal the intrinsic chemical features of an, e.g. nano-cluster catalyst. (2) Highly dispersed gold particles supported on metal oxides bring about low-temperature CO oxidation [79]; the catalytic activity correlates with the degree of dispersion, and Au8 clusters bound to oxygen-vacancy F center defects on Mg(001) were found to be the smallest clusters to mediate this reaction at low temperature [80]. (3) The reactivity of free gold cluster towards molecular oxygen, which is rightly considered as the ideal terminal oxidant, depends crucially on the charge state and the cluster size. While cationic gold clusters are completely inert toward O2, Au n − clusters react at room temperature and exhibit a notable odd/even alternation. For example, only cluster anions containing an even number of gold atoms (resulting in an odd number of valence electrons) were found to adsorb one O2 molecule [10, 81–83]; this reactivity pattern corresponds with the odd/even variations of the vertical detachment energy showing minima for Au n − (n = 4, 6, 8, …) [84]. Thus, the charge and size dependent electronic structures of the gold clusters fundamentally affect the chemical reactions with adsorbate molecules, and it was suggested that the interplay between gas-phase cluster physics and surface chemistry is a promising strategy to uncover “mechanisms of elementary steps in nanocatalysis” [85].

Next, some remarkable aspects pertinent to cooperative effects in the oxidation of CO with O2 will be presented. For excellent reviews on related topics, see Ref. [8, 10, 12, 16, 22, 25]. In the context of Au-mediated catalytic CO oxidation by O2 notable effects have been observed when the gold clusters are exposed to both reactants, either simultaneously or sequentially. Although the same rules pertaining to individual CO or O2 adsorption continue to apply, the pre-adsorption of one reactant on a cluster may lead to an increased reactivity of the cluster to the other reactant. Thus, rather than competitive co-adsorption, the rare phenomenon of cooperative co-adsorption prevails. For example, experiments with mass-selected Au6 − [86] (Fig. 7), or Au2 − [87, 88] have demonstrated that this cooperative co-adsorption gives rise to the evaporation of CO2 in a truly catalytic cycle at room temperature or below.

Fig. 7
figure 7

Cooperative, thermal catalytic oxidations of CO to CO2 in the presence of O2 by the cluster anion Au6 − (Au yellow, C black, O red). The free Au6 − ion in its equilibrium structure (I) adsorbs O2 in its superoxide form (II); subsequent co-adsorption of CO may initially form an Au6CO3 − species (III), which rearranges to the stable CO3 − adsorbate (IV); elimination of CO2 yields the Au6O− species (VI), from which a second CO2 molecule may be released and regenerate the Au6 − catalyst. For the sake of clarity, the Au6 − structure is depicted as retaining the same structure throughout the whole cycle (adapted from Ref. [86])

Full size image

A possible explanation for this enhancement of co-adsorption activity occurring in an Eley–Rideal mechanism is that the first adsorbate affects the electronic structure of the cluster thus causing it to appear electronically different to the second approaching molecule. Accordingly, CO binds much more tightly to neutral Au n than to Au n − (n = 2, 4, 6, …). Consequently, an Au cluster anion with a preadsorbed, one-electron acceptor O2 molecule will appear to be neutral to the approaching CO molecule because of the charge transfer that takes place from the Au n − cluster to the antibonding 2π* orbital of the O2 adsorbate. The analogy to the surface-catalyzed oxidation [89] of CO becomes clear in that the excess electron in Au n − is crucial for the reaction to occur, and the neutral supported clusters acquire this electron by charge transfer from the support. In the gas phase, a turnover frequency of approximately 100 CO2 molecules per Au atom per second has been estimated [86] for the reaction catalyzed by Au n − (n = 10). This efficiency is two(!) orders of magnitude greater than that observed for the commercial gold catalyst. Similar, temperature-dependent cooperative effects were reported for the Au3 −/CO/O2 system. While Au3 − was found to be inert toward O2 in the temperature regime 100–200 K, pre-adsorption of CO resulted in a charge transfer from the metal cluster’s HOMO into the 2π* antibonding orbital of CO [12]; this is accompanied with an isomerization of the Au3 − cluster from a linear to a triangular geometry. As the latter exhibits a significantly lower electron detachment energy, charge transfer to O2 is possible resulting in the experimentally observed co-adsorption products Au3(CO)(O2)2 − [90].

Even cationic gold clusters which, in general, are inert toward molecular oxygen [81, 83, 91], can be activated by pre-adsorption of molecular hydrogen [91]. Molecular binding of H2 in for example Au4(H2)4 + brings about charge transfer from the H2 ligands to the Au4 + core thus enabling the cluster to coadsorb O2 by donation of 0.14 e to the adsorbed O2 molecule. Similar effects were observed for Au n + (n = 2, 16) [91], as well as for preoxidized Pd n + clusters (n = 2–7) [92] or the oxides of both cationic and anionic gold cluster ions [93–95]. Once more, these (and other) examples clearly demonstrate that for the chemistry and physics of small cluster systems the motto holds true that “each atom counts!” [11].

4 Oxygen-centered Radicals as Active Sites in Catalytic Hydrocarbon Activation

Oxygen-centered radicals have been proposed to be responsible for the selective, large-scale heterogeneous oxidation of quite a few chemical compounds [96–98], and doping metal oxides, such as MgO, with e.g. lithium to generate radical oxygen centers in bulk metal oxides may serve as an example [99]. As shown in the previous Section, the gas-phase metal-oxide mediated conversion CO → CO2 is strongly affected by the presence or absence of these active sites. Here, the focus will be on two other elementary processes, i.e. (i) hydrogen-atom transfer (HAT) from methane, Eq. (3) [77, 100, 101], and (ii) oxygen-atom transfer (OAT) from metal-oxide clusters to ethene or ethyne, Eq. (4).

$$ {\text{CH}}_{4} + {\text{ MO }} \to {\text{CH}}_{3} + {\text{ MO}} - {\text{H}} $$
(3)
$$ \{ {\text{M}}_{n} {\text{O}}_{m} \}^{ + / - } + {\text{ C}}_{ 2} {\text{H}}_{ 2 , 4} \to \{ {\text{M}}_{n} {\text{O}}_{m - 1} \}^{ + / - } + {\text{ C}}_{ 2} {\text{H}}_{ 2 , 4} {\text{O}} $$
(4)

Regarding the mechanistic details of the gas-phase HAT reaction, two variants have been reported. The direct HAT process is operative predominantly for cationic open-shell oxide clusters with metal centers in relatively high oxidation states and with coordination numbers that prevent the indirect pathway from occurring. Examples showing this pattern include the non-metal system SO2 •+ [102], as well as the metal-containing clusters Ce2O4 •+ [103], V n P 4−n O10 •+ (n = 0, 2–4) [104–106], (Al2O3) n •+ (n = 3–5) [107], VAlO4 •+ [76], or (V2O5) n (SiO2) m •+ (n = 1, 2; m = 1–4) [108]. The polynuclear cluster V4O10 •+ was studied in great detail [104], showing that the rather efficient reaction proceeds barrier-free without the formation of a long-lived intermediate (Fig. 8).

Fig. 8
figure 8

MD simulation showing the evolution of the potential energy (in kcal mol−1) and the relevant bond lengths (in Å) (green V, red O) for the thermal reaction of V4O10 •+ with methane. The energy is shown in black, d(C–H) in blue, d(O–H) in red, and d(V–C) in green. The fluctuations after 450 fs result from vibrational motions, mainly of the OH group. The blue isosurface indicates the spin density within the respective intermediate (adapted from Ref. [77])

Full size image

The indirect, metal-mediated HAT is generally limited to small, often diatomic metal oxides, such as MnO+ [109], FeO+ [110], MgO+ [111], PbO+ [112], CuO+ [113], SnO+ [114], GeO+ [114], CaO+[115], SrO+ [115], or BaO+ [115]. These systems have a vacant coordination site at the metal atom; thus, an encounter complex [CH4···M–O•]+ as well as an intermediate [CH3–M–OH]+ are generated. The MgO•+/CH4 couple serves as a good example, Fig. 9 [111]. The initially formed encounter complex has enough internal energy to rearrange the hydrocarbon part towards the reactive oxo site at which HAT occurs. Subsequently, in a metal-controlled fashion, the methyl group returns back to give the linear [CH3–Mg–OH]+ intermediate, from which CH3 • is expelled. In general, while direct HAT resembles reaction patterns that prevail at surfaces, the indirect variant is closer to enzyme-mediated homolytic C–H bond activation.

Fig. 9
figure 9

Potential-energy surfaces (in kcal mol−1) for the reaction of MgO+ with CH4 calculated at the MP2/6-311 + G(2d,2p) level of theory; selected bond lengths are given in Å. The encircled structures depict the rearrangements occurring along the reaction coordinate (adapted from Ref. [111])

Full size image

The crucial role of a high spin density at an oxygen atom to which the hydrogen is transferred, as explained in a quite general way by state-of-the-art quantum chemical calculations [101] and verified by numerous experimental studies [77], is nicely demonstrated by main-group aluminum-oxide clusters, which bring about efficient C–H bond scission of methane at room temperature [107]: Only those clusters having an even number of aluminum atoms (Al2O3) n •+ (n = 3–5) are reactive, for example Al8O12 •+; in contrast, clusters with an odd number of aluminum atoms do not react at all with CH4, as for example Al7O11 +. As shown in Fig. 10, in the doublet ground-state of Al8O12 •+ the spin is exclusively localized at one terminal oxygen atom, while in the triplet ground-state of Al7O11 + the spin is distributed among four bridging oxygen atoms. The consequences of these electronic features for the reactions of the two clusters with CH4 become obvious upon inspection of the respective potential-energy surfaces (Fig. 11). For the Al8O12 •+/CH4 couple a direct, barrier-free HAT results; in contrast, HAT to an oxygen atom of Al7O11 + is not favored kinetically, and the significant barrier for the HAT results from the promotion energy to prepare a state which is capable to homolytically cleave the C–H bond [77, 101]. Efficient intracluster spin-transfer in homonuclear P4O10 •+ and V4O10 •+ and barriers associated with this “preparation” step for the heteronuclear V3PO10 •+ system have been suggested as origin of the quite different reaction efficiencies of these two classes of structurally related clusters in their HAT reactions with CH4 [116].

Fig. 10
figure 10

Lowest-lying structures of doublet ground-state of Al8O12 •+ (a) and triplet ground-state of Al7O11 + (b), derived from DFT/UB2LYP calculations (gray Al, red O). The spin density is indicated by the blue isosurface (adapted from Ref. [77])

Full size image
Fig. 11
figure 11

Potential-energy surfaces for the reactions of Al8O12 •+ (a) and Al7O11 + (b) with CH4 and the associated structures of the intermediates and transition structures, obtained at the UB3LYP/TZVP level of theory. Relative energies, corrected for zero-point energy contributions, are given in kcal mol−1 (adapted from Ref. [77])

Full size image

With regard to OAT to C2H2,4, Eq. (4), combined experimental/computational studies on the (ZrO2) n + (n = 1–4)/C2H2,4 systems [16, 78] also demonstrated the particular role that oxygen-centered radicals play. All these clusters exhibit a high spin density at a terminal oxygen atom and they bring about OAT-reactivity. As shown for the couple ZrO2 +/C2H4 (Fig. 12), the reaction commences with the formation of a C–O bond to be followed by an intramolecular hydrogen migration. It is this very step, that is crucial for the eventual release of acetaldehyde. The catalytic cycle is closed by an efficient reoxidation of ZrO+ with N2O; a similar mechanistic scenario holds true for the reaction of these cluster cations with C2H2 to generate ketene (CH2CO). Finally, in line with condensed-phase studies [117], gas-phase oxidation of C2H4 with various vanadium-oxide cluster cations also gives rise to CH3CHO [118].

Fig. 12
figure 12

B3LYP-derived PES for the reaction of ZrO2 •+ with C2H4 and snapshots based on MD simulations. Relative energies are given in kcal mol−1, and the time scale for the reaction in fs (adapted from Ref. [78])

Full size image

In contrast to the cationic \( \left( {{\text{ZrO}}_{ 2} } \right)^{ + }_{n} \left( {n = \, 1 - 4} \right) \) clusters, in the reactions of C2H4 and C2H2 with the anionic clusters \( {\text{Zr}}_{n} {\text{O}}^{ - }_{2n + 1} \left( {n = 1 - 4} \right) \), association clearly dominates over OAT [16]. The reason is due to the fact that in the anionic systems on electrostatic grounds the nucleophilic hydrocarbons associate with the less coordinated, more electrophilic zirconium atom of e.g. Zr2O5 − rather than to the oxygen radical center being located on the opposite side of the cluster (Fig. 13). Dissociation of the strong zirconium-carbon bond and migration of a peripheral oxygen-atom to a bridging position between the two zirconium centers requires an energy well above the entrance channel, as shown in Fig. 14 for the Zr2O5 −/C2H2 system, and is thus not accessible at ambient conditions. Consequently, oxidation of the olefin is prevented.

Fig. 13
figure 13

Molecular electrostatic potentials for a Zr2O5 − and b Zr2O4 + (adapted from Ref. [16])

Full size image
Fig. 14
figure 14

B3LYP-derived PES for the reaction of Zr2O5 − with C2H2. Relative energies are given in kcal mol−1 (adapted from Ref. [16])

Full size image

5 Mechanistic Aspects of Catalytic CH3OH → CH2O Conversion

In the mechanistic understanding of the industrially important CH3OH → CH2O oxidation, some of the relevant questions center around the following topics: (i) In the initial step (Fig. 15), does a metal-based mediator [M] induce preferentially a cleavage of the stronger O–H bond (102.4 kcal mol−1) or the weaker C–H bond (91.7 kcal mol−1) of CH3OH, and (ii) for a given sequence of events, which of the two hydrogen-transfer steps constitutes the rate-limiting one? As these questions have been dealt with quite comprehensively in a recent review [20], here we will mention briefly only a few examples which demonstrate the rather unique role that the metal species and the ligands attached to them play in this seemingly simple reaction.

Fig. 15
figure 15

Pathways for the metal-mediated methanol–formaldehyde conversion

Full size image

As shown in Fig. 16 and supported by additional experiments as well as extensive DFT-based calculations [119, 120], electrospray-ionization (ESI) of methanolic solutions of MX2 (M = Fe, Co, Ni; X = Br, I) brings about exclusive activation of the O–H bond for iron to produce Fe(OCH3)+ while the nickel precursor specifically cleaves the C–H bond resulting in the formation of Ni(CH2OH)+. For the cobalt system, one encounters an intermediate situation with a slight preference for the generation of the methoxy complex Co(OCH3)+ in competition with generating Co(CH2OH)+.

Fig. 16
figure 16

Partial ESI mass spectra of the Fe, Co, and Ni halides MX2 (X = Br, I) dissolved in a CH3OH/H2O, b CD3OH/H2O, and c CH3OD/D2O (adapted from Ref. [119])

Full size image

This metal-dependent selectivity of O–H versus C–H bond activation of CH3OH has its origin in the genesis by which the precursor species are formed. For iron, in the initial step a Fe(OCH3)(CH3OH) n + (n ≤ 8) cluster is generated via solvolysis of FeX2 by the nucleophilic solvent CH3OH. For the co-generation of isomeric [Co,C,H3,O]+, two pathways have been identified. The one, resulting in the Co(OCH3)+ complex, is analogous to that for the iron system starting from Co(OCH3)(CH3OH) n + (n = ≤8). However, this precursor, in competition with sequential CH3OH evaporation, undergoes loss of CH2O to generate Co(H)(CH3OH)+. This intermediate, in a spin-allowed elimination involving the Co–H bond and a hydrogen atom from the methyl group of the CH3OH ligand, then decomposes to H2 and Co(CH2OH)+. For the exclusive generation of Ni(CH2OH)+, two pathways are operative, both involving NiX(CH3OH)+ (X = H, Br) as precursors; in the subsequent evaporation of HX, based on labeling experiments, the hydrogen atom originates specifically from the methyl group of CH3OH.

Recently, it was observed that not only the nature of the metal, but also the ligand L for a given metal M matters with regard to the course of competitive C–H versus O–H bond activation (Schlangen M, unpublished results). For example, the system Ni(OH)(CD3OH)+ gives rise to the formation of H2O/HDO in a ratio 33:1, for the electronically related complex Ni(Cl)(CD3OH)+ the ratio HCl/DCl drops to 2:1, for Ni(Br)(CD3OH)+ HBr/DBr loss amounts to only <0.05, and, finally, the celebrated Ni(H)(OH)+ species [121, 122] in its reaction with CD3OH undergoes exclusive elimination of HD, thus pointing to clean activation of the methyl C–D bond (Schlangen M, unpublished results). Clearly, these puzzling experimental findings constitute a challenge for computational chemistry to account for a highly metal- and ligand-dependent behavior.

Among the many examples of genuine catalytic cycles in the gas-phase oxidation of methanol [8, 11, 19, 20, 123–127], the system depicted in Fig. 17 is of particular mechanistic interest.

Fig. 17
figure 17

a, b Gas-phase catalytic cycles for the oxidation of methanol to formaldehyde. Reaction ② links Mo2O6(OCH3)− and Mo2O5(OH)− and appears in both cycles, which differ in the sequence of the reactions with CH3NO2 and CH3OH (adapted from Ref. [128])

Full size image

For both cycles the anionic complex Mo2(O6)(OCHR2)− (R = H, alkyl) serves as central intermediate [128], and three elementary steps matter: (1) condensation of the complex with the alcohols R2CHOH and elimination of H2O to produce an alkoxo-bound cluster; (2) oxidation of the alkoxo ligand and its liberation as an aldehyde or a ketone in a step which is rate-limiting and requires the supply of external energy through collision-induced dissociation; (3) regeneration of the catalyst by oxidation with nitromethane. The second cycle is similar, but differs in the order of the reaction with the alcohol and the use of nitromethane as the terminal oxidant.

The crucial role of the binuclear metal center in these redox processes was assessed by examination of the relative reactivities of the mononuclear MO3(OH)− and binuclear M2O6(OH)− complexes (M = Cr, Mo, W). The molybdenum and tungsten binuclear centers (M = Mo, W) were reactive towards alcohols, but the chromium complex was not; this finding is consistent with the order of basicity of the hydroxo ligand in these anionic complexes. However, the tungsten complex W2O6(OCHR2)− prefers a redox-neutral elimination of an alkene rather than oxidation of the alkoxo ligand to form an aldehyde or a ketone. This observation is in keeping with the oxidizing power of the anions. Interestingly, each of the mononuclear anions MO3(OH)− (M = Cr, Mo, W) was inert to reaction with methanol, which highlights the importance of the second MO3 unit in the catalytic cycles. Clearly, only the bimolybdate center has the appropriate balance of electronic properties that allows it to participate in each of the three steps; these gas-phase studies with well-defined cluster anions correspond to the unique role of molybdenum(VI) trioxide (MoO3) in the industrial oxidation of methanol to formaldehyde at 300–400 °C [129].

6 Miscellaneous

In addition to the topics addressed in this perspective there are numerous other examples for using gas-phase experiments with ‘isolated’ reagents as models for mimicking catalytic reactions in the condensed phase, and they include inter alia: (1) the mechanistic understanding of the Cytochrome P-450 mediated C–H bond oxygenation [130–133] based on a detailed analysis of the most simple system, that is FeO+/H2 [134, 135], (2) the relationship between the rich gas-phase chemistry of bare PtO2 + [66] and the extraordinary features exhibited by high-valent platinum oxides [136], (3) the gas-phase CH4 → CH3OH or C6H6 → C6H5OH conversions in fully thermal catalytic cycles [137, 138], (4) the efficient catalytic gas-phase dehydration of acetic acid to ketene [139], or (5) the elegant experimental/computational gas-phase investigation on the reactions of bare Ag2O+ with olefins which, in many ways, revealed crucial details of the large-scale heterogeneous olefin epoxidation [140].

There is indeed good reason to argue that an integrated approach employing the whole arsenal of seemingly esoteric gas-phase work in conjunction with appropriate computational studies will help to bridge the gap between chemistry and physics conducted at a strictly atomic level in the gas phase [8, 11, 16, 20, 22, 25] and the most complex behavior that prevails at surfaces [31, 32, 34, 89] or in solution [141, 142] and, at long last, may thus provide insight in the nature of active sites in catalysis.