The knowledge in homogeneous catalysis is very high, due to the conceptual advance of molecular organometallic chemistry. Typically, reports in homogeneous catalysis provide not only information on the catalytic performances (activity, selectivity and life time), but also, in most cases, a detailed mechanistic understanding of the catalytic system. The actual elementary steps of the reaction, directly derived from the principles and the investigations of organometallic chemistry, are usually described. This knowledge allows a predictive approach of these systems, mainly based on the fact that it is possible to have only one well-defined catalytic species in the system. Unfortunately, from an industrial point of view, homogeneous catalysis suffers from many disadvantages and very often heterogeneous systems are preferred even if they are ill-defined and less active. The development of better catalysts in heterogeneous catalysis has always relied on empirical considerations since it is difficult to characterize the really active sites on the surfaces, as the so-called ‘active sites’ are usually in small number(s). At the present time, the number of accepted ‘elementary steps’ is still limited to a few examples mostly demonstrated by means of surface science, and the predictive approach, based on molecular concepts, is rare. The concept of surface organometallic chemistry has been developed as a possible answer to this problem. Its main objective is the creation on a support (which can be an oxide, a clay, a polymer etc.), of organometallic fragments which will be well-defined and uniform along the entire surface. These species will be characterized by all available physico-chemical methods in order to have a description of the coordination sphere around the metal as precise as possible, as for homogeneous complexes. This strategy, initially proposed by the group of J. M. Basset, has also been developed by other groups (see Table 1.1 for some examples) and has been the subject of numerous reviews and books.^1–9  In most cases the support was silica (flame silica, porous silica or mesoporous silica) but there are some examples using other oxides such as alumina or magnesia (Table 1.1).

The grafted organometallic complexes can then be modified by using the classical rules of organometallic chemistry leading to species which are potentially active in catalysis. In these compounds the support can be a mono-, di- or tripodal ligand. Recently, a new dimension was added to this chemistry by performing, prior the reaction with the organometallic complex, a reaction replacing the grafting sites by other species such as N–H,¹⁰  Si–H,¹¹  or phenol groups.¹² 

As a consequence, it is now possible to prepare, on a surface, many organometallic complexes where the electronic and steric effects can be tuned easily. This knowledge now allows researchers to have a relatively predictive approach where the starting point is not the organometallic complex but a catalytic reaction. First, a catalytic cycle is proposed on the basis of the classical rules of organometallic chemistry. Second, a grafted organometallic complex which is one species involved in the postulated catalytic cycle is prepared. Third, the catalytic reaction is performed. Depending on the results the ligands around the metal are modified or, if the reaction does not proceed by the proposed catalytic cycle, another mechanism has to be proposed and tested.

We will describe here this surface chemistry via some examples mostly from work done in our laboratory. Our purpose will be to give the rules which will govern the reactivity of the organometallic complexes with the surface and not to compile a complete list of what can be made. We will first describe the grafting sites on the surface as this point will govern the reactivity of organometallic compounds. We will then describe the reaction of these sites with organometallic complexes and how the resulting species can be transformed into active sites before giving some examples in catalysis.

Organometallic complexes can be deposited on many supports, such as metals, zeolites, oxides or carbons. Depending on the nature, density and homogeneity of the reactive sites on the surface of these materials, different behaviors will be observed, leading to sometimes completely different catalytic applications. As an example we will describe in more detail silica, as it will be extensively used in the following, as it is the simplest support (see Table 1.2 for a non-exhaustive list of grafting sites on other supports). Silica can be considered as the simplest support as it contains only SiO₄ tetrahedra linked by ≡Si–O–Si≡ bridges. It can be found in various forms such as silica gel, flame silica or mesoporous silica (like MCM-41 or SBA-15). In all cases, at room temperature, the surface is covered by hydroxyl groups ≡Si–OH and siloxane bridges ≡Si–O–Si≡ in interaction with adsorbed water molecules. Upon heating under vacuum at ca. 150 °C all water molecules are desorbed and the infrared spectrum shows mainly, in the ν(O–H) domain, a very broad band between 3700 and 3500 cm⁻¹ attributed to ≡Si–OH groups linked via hydrogen bonds. Upon heating at higher temperature, condensation between two neighboring hydroxyl groups occurs, leading to the evolution of water molecules and formation of ≡Si–O–Si≡ bridges. As a consequence, the intensity of the broad infrared band decreases and a new sharp band, attributed to isolated silanol groups, appears at ca. 3750 cm⁻¹. After heating at 500 °C only isolated silanols are present. Their amount can be determined by chemical methods (such as by their reactivity with CH₃Li) or physical techniques (such as quantitative solid-state ¹H MAS NMR). The two values can be different as the first method will only quantify the hydroxyl groups accessible to the reagents while the second one will give an estimation of the total number of hydroxyl groups. In the case of microporous solids the difference can be very important. In the case of flame silica, which is non-porous, the two methods lead to an OH density of ca. 1.4 OH nm⁻².^41,42  Upon heating under vacuum at 700 °C the OH density decreases to ca. 0.7 OH nm⁻². For such low values, one can reasonably suppose that the hydroxyl groups are far away from each other and so well-defined grafted organometallic isolated species will be expected upon reaction with these hydroxyl groups. This is the key point of surface organometallic chemistry. However, this view is not fully realistic even if it is sufficient in most cases; sometimes a more precise description of the silica support is needed for explaining the experimental data. First of all, there are not only monohydroxyl ≡Si–OH groups on the surface but also dihydroxyl ones Si(OH)₂, as evidenced by solid-state ²⁹Si MAS NMR. These Si(OH)₂ groups give infrared bands at the same position as the ≡Si–OH ones and so they cannot be distinguished by this method. On the samples heated at high temperature the ²⁹Si MAS NMR spectra become broader, preventing the separation of the different Si(OSi)_4−x(OH)_x (x=0–2) groups and so the presence of some Si(OH)₂ species cannot be excluded even if their amount is probably very low. Studies by ¹H double quanta MAS NMR are more informative. This method allows the observation of protons pairs which are separated by less than ca. 5 Å. As a consequence, not only pairs of protons involved in hydrogen bonds are seen but also protons at a slightly higher distance. An intense signal is observed for pairs of isolated protons even on silica treated at 700 °C (for which the OH density is 0.7 OH nm⁻²), probing unambiguously that the repartition of the hydroxyl groups is not fully homogeneous.⁴³  Additional experiments by triple quanta ¹H MAS NMR or spin counting were performed.^43,44  These experiments showed that some hydroxyl groups are present as nests of three silanols located on a cycle containing six silicon atoms as those observed on the (111) face of cristobalite. In that case, the distance between the hydrogen atoms will be ca. 3 Å. One, two or three cycles can be adjacent.

Finally, upon heating at a very high temperature (ca. 1000 °C), highly strained cycles containing two silicon atoms and two oxygen atoms are formed by condensation between two adjacent silanols. These cycles are highly reactive even if their amount is low (0.14 nm⁻² on silica dehydroxylated at 1000 °C while the amount of residual hydroxyl groups is 0.4 nm⁻²).

Scheme 1.1 summarizes the variety of species which are present on dehydroxylated silica as deduced from these studies. Depending on the nature of the silica (non-porous flame silica, mesoporous silica, etc.) the relative amount of these species will be different, leading then in some cases to different reaction products.

As shown above, even if silica can be considered as the simplest support, there is not only one species on its surface. For other oxides such as alumina the situation is more complicated and more than five different types of hydroxyl groups can be observed, with all their combinations, without taking into account the Lewis acid sites. The most complex support is probably carbon, as its surface contains a lot of functional groups covering all the fields of organic chemistry. This complexity of the surface support will have a consequence on the number of species which will be obtained upon reaction with organometallic complexes as the strength of the bond between the metal and the surface will be more or less strong. It is for this reason that preliminary studies are always made on silica and mainly on silica dehydroxylated at relatively high temperature (500 or 700 °C).

As we have seen above, the active sites of the support are mainly (if carbon is excluded) hydroxyl groups; only their distribution and strength depend on the treatment and on the oxide under study. The formation of a chemical bond between the organometallic fragment and the solid will then pass, in most cases, through a reaction with these hydroxyl groups. We will describe here what will happen and by comparison of various supports and organometallic compounds how the reaction proceeds.

First of all, it is necessary to choose an organometallic compound with a M′–X bond for which the reaction M–OH+[M′]–X→M–O–[M′]+HX will be favored thermodynamically, M–OH being a hydroxyl group of the support. Many complexes can be chosen, such as chlorides or alkoxy derivatives. However, in these two cases the evolved hydrogen chloride or alcohols can further react with hydroxyl groups or M–O–M bridges of the support and so modify its properties. A typical example is the reaction at room temperature of tantalum methoxide Ta(OMe)₅ with silica dehydroxylated at high temperature.^45  13C CP-MAS NMR of the resulting material shows clearly two signals for methoxy species. One of them can be attributed to a methoxy group on tantalum as expected and a second to a methoxy group on the silica support (such species can be synthesized by treatment at relatively high temperature of silica with methanol). These silica methoxy groups are formed by reaction of evolved methanol with a siloxane bridge or a silanol group. As this reaction does not proceed at room temperature in the absence of the tantalum complex, the metal plays the role of catalyst for this reaction. The consequence is that the grafting reaction will not be clean and the starting treatment of the support for the creation of isolated grafting sites will not be efficient as new potential grafting sites will be created during the reaction. It is for this reason that the organometallic complexes which will be chosen must lead to inert X–H species. The best choice is to have evolution of alkanes which cannot be activated by these supports. For this purpose the ligands around the metal will be alkyl, alkylidene or alkylidyne groups.

Very often homoleptic organometallic complexes are chosen as they will lead to only one surface complex, all ligands being equivalent. One problem is that, kinetically, the reaction will be slow, compared for example with that achieved with alkoxy compounds, due to the fact that the first step, the physisorption on the support, will not be favored, the interaction between hydroxyl groups and alkyl groups not being strong. To overcome this problem the complex can be sublimed on the support, avoiding the use of a solvent, but this method can be used only when it has a sufficient vapor pressure and sometimes the sublimation is accompanied by a partial decomposition. In all cases the observed reaction can be simply written as:

When using homoleptic complexes and silica dehydroxylated at high temperature, well-defined species are obtained which are uniform over all the solid. This strategy has been applied to a lot of metals from the left (Ti, Zr, Hf) to the right (Sn, Ge) of the periodic table (see, for example, Basset et al.⁴⁶ ). In all cases the same result was obtained but the reaction did not proceed at the same rate: For metals of the left, such as Ti or Zr, the grafting reaction occurred easily at room temperature while for metals of the right, such as Sn, it occurred only at high temperature (ca. 180 °C). This is related to a different reaction mechanism as evidenced by the use of other supports with a stronger acidity such as cloverite,⁴⁷  Y zeolite⁴⁸  or heteropolyacids.⁴⁹  Heteropolyacids such as H₃PW₁₂O₄₀ are molecular compounds more acidic than sulfuric acid. Surprisingly, they do not react with alkyl complexes of Ti or Zr while they react at room temperature (and even below) with tin complexes. These results can be understood as follows. For metal complexes of the left of the periodic table, which are highly electron deficient with empty d orbitals, the grafting reaction occurs via an attack of the M–C bond by the oxygen atom of the hydroxyl group followed by evolution of the alkane. So the first step is the formation of the bond with the surface (Scheme 1.2). For metal complexes of the right, which have a high electronic density, the first step is an attack by the proton, leading first to the evolution of alkane and then to the formation of the metal–support bond. As increasing the acidity decreases the oxygen nucleophilicity this explains the different reactivity as a function of the acidity of the support. A consequence is that it is not possible to graft metal alkyl complexes of the left of the periodic table on highly acidic supports. Another consequence is that it is not possible to graft platinum methyl complexes on silica via breaking of the Pt–Me bond as these compounds are not stable thermally.

Another consequence of this mechanism is that it can be possible to graft metal complexes of the right of the periodic table by use of acidic species as catalysts: The grafting reaction of tetramethyl tin occurs at room temperature on H₃PW₁₂O₄₀ supported on silica but the amount of evolved methane exceeds by at least one order of magnitude the number of protons of the polyacid and can only be explained by a migration on the surface. The mechanism (Scheme 1.3) passes through an attack of the tin complex by the acidic proton of the heteropolyacid followed by a migration of the grafted tin species on the surface and restoration of the acidic proton.

If for the elements of group 4 (Ti, Zr, Hf) it is possible to prepare metal complexes such as M(–CH₂–C(CH₃)₃)₄ without hydrogen atoms in β position of the metal (which can lead to side reactions, see below), it is not possible to obtain the corresponding species for those of groups 5 and 6. Indeed the steric hindrance around the metal is so high that α-H abstraction occurs, leading to the evolution of alkane and the formation of a metal alkylidene (for group 5) and even, after a second α-H abstraction, of a metal alkylidyne (or a dialkylidene). The resulting complexes are now heteroleptic as they contain two types of ligands which could lead, a priori, to two types of surface complexes. However, only one surface species is obtained, for example ≡Si–O–Ta[CH₂–C(CH₃)₃]₂[CH–C(CH₃)₃)] in the case of tantalum.^45,50  Two mechanisms can be proposed for the grafting reaction: (i) selective reaction of the metal alkyl moiety as for the alkyl complexes or (ii) addition of the oxygen atom on the carbene moiety followed by α-H abstraction and neopentane evolution (Scheme 1.4). The second mechanism was first proved by use of deuterated silica.⁵¹  Indeed if the reaction proceeds via reaction with the alkyl moiety, all neopentane should be monodeuterated while in the second case only one fourth will be deuterated, as observed experimentally. Later the pentacoordinated intermediate was observed by solid-state ¹³C NMR and its molecular analogue was synthesized confirming the above mechanism.⁵² 

In the case of alkylidynes complexes (for example W[CH₂–C(CH₃)₃]₃[≡CH–C(CH₃)₃)] quite the same reactivity is observed with the formation of surface alkylidynes species.⁵³  Another example is the rhenium complex Re[CH₂–C(CH₃)₃]₂[CH–C(CH₃)₃)][≡CH–C(CH₃)₃)] which contains the three different ligands and whose reaction product with silica dehydroxylated at high temperature has been fully characterized by various physico-chemical methods and DFT calculations (Figure 1.1).⁵⁴  For this surface complex the presence of an agostic interaction between the metal and the hydrogen of the carbyne moiety could be proved by solid-state NMR. This illustrates the precision which can be attained in the characterization of a surface complex.

More recently, additional data on the nature of the bond between the metal and the silica support were obtained by ¹⁷O solid-state NMR.⁵⁵  Not only the oxygen atom of the ≡Si–O–M bridge could be identified but also the presence of ≡Si–O–Si≡ bridges in small interaction with the metal was observed giving new insights into the structure of the grafted organometallic species and their interaction with the silica carrier.

This chemistry has then been extended to other complexes containing ‘inert’ ligands in combination with alkyl or alkylidene ones in view of the synthesis, for example, of olefin metathesis catalysts. This is the case of a tungsten imido complex based on the Schrock catalyst.^56,57  Another example is the synthesis of tungsten oxo species on the silica surface which can be considered as models of the industrial WO₃/SiO₂ metathesis catalyst.⁵⁸ 

As described above, this reaction leads to the formation of a surface complex linked to the support by one bond. However, this species can further react with other hydroxyl groups if they are in close proximity. This can be observed in the case of silica dehydroxylated at moderate temperature (200 °C, for example) or in mesoporous silicas where the distribution of hydroxyl groups is not the same than in flame silica. Quite the same reaction as that described above will then occur, leading to a new evolution of alkane and formation of a digrafted species, one alkyl ligand being replaced by a ≡Si–O one. However, the question is: Is it possible to prepare cleanly a digrafted organometallic complex on the silica surface? Some people have claimed it to be so, but in most cases it is not possible as silica dehydroxylated at low temperature contains also isolated silanols as proved by infrared spectroscopy. A proof of this assumption was obtained in the case of the grafting reaction of the zirconium complex Zr[CH₂–C(CH₃)₃]₄ on silica dehydroxylated at various temperatures. For this purpose the reactivity of the surface complex with trimethylphosphine was studied (Scheme 1.5).⁵⁹  Indeed it had been reported that alkane could evolve, via α-H abstraction, upon reaction of alkyl complexes with a phosphine.⁶⁰  The monografted complex does not react with trimethylphosphine at room temperature probably due to a steric hindrance preventing the coordination of phosphorus on zirconium. This point can be checked simply by studying the reactivity of the phosphine with the complex synthesized on highly dehydroxylated silica. In contrast the digrafted complex reacts with evolution of neopentane. As the tri-grafted complex has only one neopentyl ligand, it cannot undergo a α-H abstraction. By knowing the amount of neopentane evolved during this reaction and that during the grafting step, the relative proportions of the mono-, di- and tri-grafted species can be determined. The results are shown on Figure 1.2. Clearly it is not possible to prepare the digrafted complex alone, even if on silica dehydroxylated at 300 °C it represents more than 70% of the surface species. It is always present with the two other complexes. Only the monografted complex can be obtained as a pure species.

α-H abstraction can proceed in the case of sterically hindered complexes and leads to the evolution of alkane and the formation of a surface alkylidene which can be active in olefin metathesis. This reaction is also observed upon treatment of a digrafted zirconium with trimethyl phosphine (see above)⁵⁹  or by heating a digrafted chromium complex.⁶¹ 

Another reaction which can proceed is the β-H abstraction. This reaction is observed, for example, during the thermolysis of tin complexes supported on silica and leads to the evolution of alkene and the formation of a tin(ii) species.⁶²  This reaction is also very important in catalysis, for depolymerization and hydrogenolysis or metathesis of alkanes. This point will be discussed in more details in the catalysis part.

γ-H abstraction has been proposed to occur during the thermolysis of some tetraneopentyl complexes of group 4.⁶³  This reaction leads to the evolution of neopentane and the formation of a metallacyclobutane. Even if it is relatively uncommon, it has been observed during the thermolysis of neopentyl hafnium complexes supported on silica.⁶⁴ 

When looking at all above surface complexes only those containing the alkylidene ligand can be used in catalysis, for olefin metathesis. However, highly active species for the activation of alkanes can be prepared by hydrogenolysis. Treatment under hydrogen at moderate temperature (ca. 150 °C or below) of the alkyl, alkylidene and alkylidyne surface complexes leads to the formation of surface hydrides which are highly electron deficient and so potentially highly active catalysts. The expected reaction can simply be written by replacing one alkyl ligand by one hydrogen, one alkylidene one by two hydrogens and one alkylidyne one by three hydrogens with evolution of the corresponding alkane. However, the resulting species are so electron deficient and so reactive that they will react with neighboring siloxane bridges ≡Si–O–Si≡ leading to the formation of M–O–Si≡ bonds with the surface and the formation of a Si–H group. Thermodynamically this reaction is highly favored. The result is a digrafted metal hydride which can further react with another siloxane bridge leading to the formation of a tri-grafted metal hydride. This is the case for group 4 metal complexes which will lead to the formation of a mixture of mono- and dihydrides on the surface (Scheme 1.6).^65,66  The best characterization method for these species is two-dimensional (2D) double quanta ¹H MAS NMR which allows the observation of an autocorrelation peak for the dihydride (Figure 1.3). The formation of these two species is probably related to the heterogeneity of the silanol distribution on the solid as explained in the first part, the presence of nests preventing the formation of the third metal–surface bond. Indeed, the restructuration of the surface is so high that the hydride does not have a siloxane group in close proximity.

For elements of group 5 (mainly tantalum) the reaction is more complicated as this element can exist as tantalum(iii) and tantalum(v). As a consequence a reductive elimination of hydrogen can be observed and the main species is a digrafted tantalum(iii) monohydride which can also be in interaction with a siloxane bridge as shown by EXAFS.⁶⁷  In presence of hydrogen this tantalum(iii) monohydride can coordinate one H₂ molecule via an oxidative addition, leading to a digrafted tantalum(v) trihydride in equilibrium with the monohydride (Scheme 1.7).

In the case of group 6 complexes, no hydride is obtained on silica and only the formation of large particles by sintering is observed. In contrast, when the alkylidyne complex of tungsten is grafted on alumina, a well-defined and highly active hydride is formed.⁶⁸  This is related to the strength of the bond between the metal and the surface and is well known in heterogeneous catalysis (platinum on alumina is used industrially instead of platinum on silica).

The supported alkyl (or hydride) complexes can also react with many molecules, such as alcohols or amines leading to the formation of various derivatives which are more stable than their precursors and can be used in reactions involving for example oxygen derivatives (epoxidation of olefins,⁶⁹  deperoxidation of cyclohexyl peroxide,⁴⁵  etc.(. Depending on the amount of alcohol or amine which will be used and on the reaction conditions, different compounds can be obtained. In contrast to a direct grafting of the corresponding complexes, alkoxy derivatives for example, this method allows the obtention of well-defined and well-dispersed species on the surface, as the silica–metal bond had been created previously. One typical case is that of Ta(OMe)₅ which is a dimer. When treating the alkyl–alkylidene surface complex with methanol a monomer is obtained while reaction of the alkoxy derivative leads to the formation of a dimer, as shown by EXAFS. This can be of interest for applications in catalysis as in some cases dimers are expected to be the active sites. Depending on the molecule a lot of functions can be introduced around the metal. If the reaction is made on the complex prepared by reaction with a silica dehydroxylated at high temperature, this new species will have only one bond with the surface. If the starting silica had been dehydroxylated at low temperature, the main species will have two bonds with the surface and in the case of group 4 elements if the reaction is made with the hydride the main resulting species will have three bonds with the surface. However, and as for the direct grafting reaction of the alkoxy derivatives, alkoxy groups are also formed on the surface (see above). This point can be important in catalysis as it can modify the sorption properties of the solid and so can affect the kinetics of the reaction.

All the above synthesis methods allow only the preparation of neutral surface complexes linked to the surface by a covalent bond. However, in some cases, the catalytic species is a cation, for example metallocene species in the case of olefin polymerization or chromium cations for the trimerization of ethylene. Classically, this cation is formed by reaction of a neutral complex with a Lewis acid which must have a non-coordinating ability after its transformation into the corresponding anion. With regard to these conditions B(C₆F₅)₃ was found to be a very good candidate.

When applying this methodology to the design of single site heterogeneous catalysts on a surface such as silica, two different strategies may be used:

1. To graft first the Lewis acid co-catalyst directly to the surface in order to create a covalent bond with the surface and then to react this supported Lewis acid with the molecular complex in order to achieve the preparation of a ‘floating’ cationic complex.

2. To graft first a precursor complex directly with the silica surface in order to create a covalently bonded complex such as described above and in a second step react this surface complex with a Lewis co-catalyst which will abstract an alkyl group from the surface complex, leading to a cationic species covalently bonded to the surface.

These two methods will lead to different surface species the coordination sphere around the metal containing or not a covalent bond with the surface.

We will first describe one example of the first method with B(C₆F₅)₃ as a Lewis acid (Scheme 1.8) but this strategy can be applied to other compounds. When B(C₆F₅)₃ is contacted with silica dehydroxylated at high temperature no reaction is observed. However, in presence of a tertiary amine like NEt₂Ph a grafting reaction occurs with formation of HNEt₂Ph⁺ and the anionic grafted species [Si–O–B(C₆F₅)₃]⁻ which has been fully characterized.⁷⁰  The tertiary amine acts as an activator of the inert silanol moiety and allows then the Lewis acid to coordinate the oxygen. Reaction of this surface ionic pair with a metallocene like Cp*ZrMe₃ leads to the evolution of one methane molecule and the formation of the surface species [≡Si–O–B(C₆F₅)₃]⁻·[Cp*(Net₂Ph)ZrMe₂]⁺ which has been fully characterized and is active in ethylene polymerization without the need of a cocatalyst such as methylaluminoxane (MAO).⁷¹ 

The other strategy (reaction of the organometallic complex with the silica support followed by abstraction of a methyl group by the borane compound) can also be illustrated by the same Cp*ZrMe₃ complex. Reaction of this complex with silica dehydroxylated at high temperature leads to the formation of the well-defined species ≡Si–O–ZrCp*Me₂ (Scheme 1.9) which can further react with B(C₆F₅)₃ without the need of a tertiary amine. The expected reaction (transfer of a methyl group from zirconium to boron) is observed but it is not the major species on the surface. The reason is that, as in the case of hydrogenolysis, this species is very reactive (note that in contrast to that prepared by the first strategy it does not have a coordinated amine). It can further react with a siloxane bridge via a methyl transfer to the surface leading to a digrafted cation which has loss all its methyl ligands and so is be inactive in catalysis.⁷²  The catalyst contains the two species, resulting in an activity lower than expected.

We will give some examples of reactions catalyzed by the above complexes with the aim to show the advantages of the synthesis of catalysts by surface organometallic chemistry.

As shown above the hydrides are prepared by treatment under hydrogen of the surface alkyl complexes. However, instead of the expected neopentane, only methane (and ethane in the case of group 4 elements) is observed in the gaseous phase. This reaction is also observed when an alkane is contacted with the hydride in presence of hydrogen. The key step is a β-alkyl transfer, quite similar to the β-hydrogen transfer but which occurs more rarely (Scheme 1.10). The driving force for the reaction is the hydrogenation of the olefin which will shift the equilibrium to the right. Indeed, in absence of hydrogen no reaction proceeds and the opposite reaction can be performed: the catalyst polymerizes ethylene via the classical mechanism of insertion in the Zr–C bond.

This reaction is of no interest for small alkanes but can become important when applied to polyolefins, which can be considered as alkanes with a very long chain. For example, it can be applied to the hydrogenolysis of waxes. Indeed Fischer–Tropsch synthesis leads to the formation of such side products which cannot be directly valorized industrially. Degradation by use of a group 4 hydride allows their direct transformation into diesel fuel and gasoline.^73,74 

Another interesting example is the transformation of polymers into more valuable compounds. For example, polystyrene can react with the zirconium hydride in presence of hydrogen. Three reactions are observed: (i) breaking of a carbon–carbon bond of the chain, (ii) breaking of a carbon–phenyl bond leading to the evolution of benzene, and (iii) hydrogenation of the phenyl group of the polymer. Depending on the conditions the last two reactions can be prominent, leading to the formation of a new styrene/ethylene/vinyl cyclohexane terpolymer with interesting physical properties.⁷⁵ 

In the case of tantalum and tungsten hydrides only methane is evolved during the hydrogenolysis of the corresponding surface alkyl complexes. The above mechanism involving a β-alkyl transfer is not sufficient to explain these results and it is necessary to take into account another mechanism for the cleavage of the C–C bond of ethane. This mechanism is an α-alkyl transfer, which will lead to the formation of a surface alkylidene (Scheme 1.10). Combination of the α- and β-alkyl transfer then allows an explanation of this cleavage, which occurs via the olefin metathesis reaction, evolving methane and propane from two ethane molecules (Scheme 1.11).⁷⁶  Propane is then transformed by the classical mechanism of hydrogenolysis into methane and ethane which can further react. Finally, only methane is obtained. The resulting reaction, transformation of two ethane molecules into one methane molecule and one propane molecule, is called alkane metathesis by analogy with what is observed for olefins. It can be decomposed into three steps: dehydrogenation of the alkane, olefin metathesis and, finally, hydrogenation of the resulting olefins.

As shown above, the tantalum and tungsten hydrides can react with alkanes leading to the formation of alkylidene species which are catalysts for olefin metathesis. However, the tungsten hydride can also perform the ethylene dimerization into butene. The consequence is that this system can directly transform ethylene into propene via the following mechanism (Scheme 1.12).⁷⁷  Ethylene is first dimerized into but-1-ene which is isomerized via a classical β-H abstraction mechanism. Then the cross-metathesis of ethylene and butane leads propene. These three different reactions occur on the same site.

As shown above, olefin metathesis catalysts derived from those developed by Schrock and colleagues can be supported on silica.⁵⁷  The resulting materials can be very active for olefin metathesis. For example the system shown in Figure 1.4 can have turnover numbers higher than 20 000 after 20 h.

Recently, more active systems based on tungsten oxo complexes were developed and patented.⁷⁸  Typically these catalysts are based on the grafting reaction of W(O)F[CH₂C(CH₃)₃]₃ on an oxide support and lead to very high activities in the metathesis of propene.

Selective trimerization of ethylene to 1-hexene has been one of the most exciting developments of olefin catalysis in the past decades. A supported catalyst for the ethylene oligomerization has been designed very recently by a molecular approach.⁷⁹  It operates efficiently without any co-catalyst and delivers 1-hexene in good selectivity. This system has been obtained by grafting reaction of TaCl₂Me₃ on dehydroxylated silica, leading to the (≡SiO)Ta^vCl₂Me₂ precatalyst. During the initiation step ethylene reduces this complex, allowing then the generation of the key metallacycle intermediates and thus the selective production of the requested 1-hexene. DFT calculations have explained this unexpected reactivity and why other tantalum compounds are not active.⁸⁰ 

Such reactions involve oxygen or oxygenated molecules and so cannot be catalyzed by alkyl or hydride complexes such as those depicted above. However, by reaction of these species with alcohols it is possible to obtain new well-defined surface complexes which can be used for these reactions. Mono-, di- and tri-grafted titanium complexes can then be obtained which are active systems for the epoxidation of oct-1-ene. A study of these different systems and of their transformation during the catalytic reaction has allowed an understanding of the behavior of the industrial epoxidation catalyst developed by Shell, which deactivates rapidly at the beginning before becoming stable upon time.⁸¹ 

Another example is the asymmetric epoxidation of allyl alcohol. In solution this reaction is usually made by use of the Sharpless catalyst which is based on a titanium complex with a tartrate ligand. A system has been designed on the surface by taking into account that the coordination sphere of the metal must contain one bond with the surface, two bonds for coordination of the tartrate ligand and it must also accommodate the two reagents, allyl alcohol and the peroxide. As a consequence five bonds are needed around the metal and so a group 5 metal complex should be expected, such as tantalum. The coordination sphere around the metal can be built by (i) grafting reaction of the Ta[CH₂–C(CH₃)₃]₃[CH–C(CH₃)₃] complex on silica dehydroxylated at high temperature (formation of the surface–metal bond); (ii) reaction of the grafted complex with ethanol in order to replace the alkyl and alkylidene ligands by ethoxy ones; (iii) addition of the asymmetric ligand, (+)-diisopropyl tartrate (Scheme 1.13). The resulting system is active for the asymmetric epoxidation of allyl alcohol with results quite comparable to those achieved in homogeneous catalysis.⁶⁹ 

A third example is the deperoxidation of cyclohexyl hydroperoxide. An important way of synthesis of adipic acid developed by Rhodia passes through the oxidation of cyclohexane in three steps. In a first step cyclohexane is oxidized, by a classical Fenton mechanism, into a mixture of cyclohexyl peroxide, cyclohexanol and cyclohexanone. In a second step, this mixture is transformed into cyclohexanol and cyclohexanone by oxidation of the peroxide. Finally, the two oxygenated compounds are oxidized into adipic acid by nitric acid. Surface organometallic chemistry was used in order to find a heterogeneous system for the second step. Indeed this reaction is usually made in solution in presence of a chromium salt which is highly toxic. The aim of this study was mainly to obtain a purely heterogeneous system without any leaching even if it was not very active. A screening of all available systems showed that ethoxy complexes of tantalum grafted on silica, such as that depicted on Scheme 1.13, were active without any leaching for the deperoxidation of the industrial mixture. This system has been patented but unfortunately the activity was too low, preventing its industrial use.^45,82 

Ammonia is a very important chemical as its production reaches ca. 10⁸ tons per year. It is mainly produced by the Haber–Bosch process from nitrogen and hydrogen. The reaction proceeds at high pressure and relatively high temperature even if it is highly exothermic and so in conditions where the equilibrium is not shifted to the ammonia production. A lot of work has been done for a search of catalysts allowing this reaction to proceed in mild conditions and recently Schrock has proposed a catalytic cycle with a monometallic molybdenum complex.⁸³  However, up to now nothing had been made on a heterogeneous catalyst. It has been shown recently that the tantalum hydride complex can activate dinitrogen in presence of hydrogen leading to the formation of the amido imido complex of tantalum (≡SiO)₂Ta(NH)(NH₂).⁸⁴  The reaction proceeds at moderate temperature and pressure (250 °C and 0.5 bar) and is completely shifted to the production of this surface complex. Recent DFT studies have shown that the reaction proceeds via successive heterolytic cleavages of hydrogen.⁸⁵  Unfortunately ammonia cannot be desorbed from this complex which is probably a stable species as it can also be obtained by the reaction of ammonia with the tantalum hydride.⁸⁶  However, it should probably be possible to have a catalytic cycle by adding a reaction of the amido imido complex with an organic moiety. Further studies in this direction are in progress.

In the course of this short review we have shown that it is possible to prepare well-defined organometallic fragments on an oxide surface. Only one species can be obtained on the surface when choosing the right conditions. This surface complex can further react with various molecules, leading to active species for a lot of reactions including the activation of alkanes, as species unknown in solution can be stabilized by the surface. The nature and the structure of the support on which the organometallic fragment will be grafted will have an effect on the conditions where only one species is observed and on the presence of different species after subsequent reactions of the grafted complex. For example, flame silica will lead to monografted complexes after dehydroxylation at a lower temperature than a mesoporous silica. The heterogeneity of the hydroxyl distribution has no effect on the grafting reaction but is important for the synthesis of surface hydrides by hydrogenolysis, as two species will be obtained. In all cases this approach allows the obtention of one or only few surface species with a well-defined coordination sphere around the metal, allowing structure–activity relationships to be performed. This method can also allow a predictive approach for the design of catalysts for a given reaction to be obtained.