In-situ synthesis of hydrogen peroxide in tandem with selective oxidation reactions: A mini-review

One-pot combination of the in-situ generation of hydrogen peroxide (H2O2) with selective oxidation reactions enhances the environmental and economic attractiveness of the use of H2O2 as green oxidant at industrial scale. This tandem reaction improves not only the safety aspects associated to H2O2 storage and transportation but also reduces the capital and operation costs. This review covers the progress in the field focussing on the areas of catalysis development, mechanistic understanding and engineering


Introduction
Hydrogen peroxide (H 2 O 2 ) is currently presented as an attractive alternative to the industrial use of oxidants such as tBuOOH, N 2 O, NaClO, chromate or permanganate [1,2]. These stoichiometric oxidants produce large amounts of waste by-products requiring additional purification and separation steps. In contrast, hydrogen peroxide is a highly active and selective oxidation agent for many http oxidation reactions with a high active oxygen content (47 wt%) producing only water as a co-product [3]. Additionally, hydrogen peroxide normally presents a higher oxidation potential than molecular oxygen, being efficient in a wide range of reaction conditions both acidic and basic. An important breakthrough in this field took place in 2008 with the commercialisation of the epoxidation of propylene with H 2 O 2 to obtain propylene oxide (HPPO process) using titanium silicalite (TS-1) as a catalyst. Currently there are two HPPO-process plants jointly developed by BASF and Dow in Belgium and Thailand which finished their start-up in 2008 and 2011, respectively [4,5].
Currently, hydrogen peroxide is industrially produced via the anthraquinone route. Simplistically, in this process, an alkyl anthraquinone is used as a carrier which is firstly hydrogenated and later oxidised, releasing hydrogen peroxide [3]. One of the main advantages of this system is its capability of producing highly concentrated H 2 O 2 which is normally marketed with concentrations ranging from 35 to 50 wt% or even above, however, the transportation and storage of H 2 O 2 in concentrated form presents associated safety and economic issues. Currently, there is great scientific and industrial interest in the direct synthesis of hydrogen peroxide from molecular hydrogen and oxygen as a way of resolving its current challenges and potentially reducing its price [6,7]. Additionally, despite the relatively low concentration of H 2 O 2 achieved during its direct synthesis compared to the high concentration solutions via the anthraquinone route, it aligns with the requirement for oxidation reactions (3-5 wt%).
Most of the current research in the area of direct synthesis of H 2 O 2 as an oxidising agent focuses on the development of highly active and selective noble metal catalysts, with emphasis on gold and gold-palladium due their high yield stability [6,8,9]. Despite recent progress in the area [7], its large-scale application is mainly limited by its parallel decomposition into H 2 and water on the metal surface, decreasing the hydrogen efficiency of the process compromising its economic feasibility. Different strategies are currently being explored to overcome this problem [6], being the one-pot direct synthesis and utilisation of H 2 O 2 one of the most elegant approaches [10][11][12]. In this way, in-situ generated hydrogen peroxide from molecular oxygen and hydrogen is simultaneously used for oxidation without the requirement of intermediate purification and/or separation steps. This integration of processes presents obvious advantages towards the reduction of the capital and operational costs which are expected to further the enhancement of the economic attractiveness of the use of hydrogen peroxide at industrial scale. Further safety benefits are gained by the lack of transportation and storage of hydrogen peroxide. Additionally, a controlled feeding of the desired concentrations of hydrogen peroxide in the system can be naturally achieved and tuned depending on the oxidation requirements. Considering these advantages, it is not surprising that the study of oxidation reactions using in-situ generated hydrogen peroxide has always been of great scientific interest, especially within the green chemistry movement. Consequently, the number of articles devoted to this tandem reaction has exponentially increased since the first studies in the 90s with most of the early progress covered in a review article by Clerici et al. in the late 90s [10]. Since then, significant progress has taken place in the areas of catalyst development, mechanistic understanding and engineering approaches such as the application of membrane reactors which are reviewed here. Focus has been directed towards specific systems involving epoxidation of propylene oxide, the hydroxylation of benzene, oxygenate synthesis from methane and Fenton processes due to their industrial potential. This review also highlights some of the current limitations in this tandem system that must be considered for future developments such the lower overall yields to oxidation products compared to the use of pre-synthesised hydrogen peroxide, the high flammability of H 2 /O 2 mixtures and H 2 efficiency.

Propylene epoxidation to propylene oxide
The selective oxidation of hydrocarbons is one of the most important processes for producing oxygenates, which are used as key intermediates in organic synthesis. One example of a high value oxygenate is propylene oxide (hereafter denoted as PO) which is mainly used as a feedstock for producing urethane polyols and propylene glycol.
The selective oxidation of propylene with molecular oxygen has been demonstrated to be very difficult to carry out as low yields and selectivity to PO are obtained. Thus, most of the industrial processes for PO production use clorohydrin or hydroperoxides in a two-stage process, which involves additional separation and/or purification units that increase the overall cost of the process. Recently a new process called hydrogen peroxide to propylene oxide (HPPO process) based on the selective oxidation of propylene with hydrogen peroxide has been implemented in Belgium and Thailand [4,5]. The HPPO process presents significant advantages from an economical and environmental point of view over former systems based on clorohydrin or tert-butyl peroxide such as the lower purification costs, the decrease in wastewater and the required energy input.
An interesting alternative to the HPPO process is the synthesis of PO by selective oxidation with in-situ generated H 2 O 2 from oxygen and hydrogen, which can be undertaken in both liquid or vapour phase. Most of the studies in this area involve the use of metal nanoparticles supported on titanium-based materials as catalytic systems. Gold attracts most of the attention, however, there are also studies based on silver or other precious metals such as platinum or palladium. It is generally agreed that the role of the metallic species is mainly to produce hydrogen peroxide from H 2 and O 2 which is then converted into hydroperoxide-like species at the Ti sites of the support [8,[13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28]. However, the metallic species can also be involved in the epoxidation reaction itself. An IR spectroscopic study [29] showed that the presence of gold nanoparticles on titania also catalyses the reaction between propylene and titania sites forming a bidentate propoxy species, similar to that formed by prolylene adsorbed on titania. The gold nanoparticles also catalysed the consecutive oxidation of the bidentate species to form carbonate/carboxylate species, which could be the cause of catalyst deactivation as discussed below. In the presence of H 2 and O 2 , the intermediate bidentate propoxy species can desorb from the catalyst surface forming peroxide species over gold as the rate determining step [22,29,30]. Therefore, the mechanism of epoxidation of propylene over metallic-based catalysts can be simplistically described by the following steps [31,32]: (i) Propylene adsorbs on the titania-based support to produce bidentate propoxy species. (ii) Simultaneously, hydrogen and oxygen produce hydroperoxide-like species (OOH or H 2 O 2 ) on the metallic surface and (iii) The peroxide species aid the desorption of the bidentate propoxy species from the catalyst, producing propylene oxide and water and restoring the support to its original state. Alternatively, these bidentate propoxy species could be further oxidised to produce strongly adsorbed carboxylate species causing the loss of the activity of the catalyst. Despite the synergetic role of the support and the metal nanoparticles in the tandem reaction, the effect of both is going to be considered separately for the sake of structure and clarity, however, one should consider both simultaneously when designing selective and active catalytic systems.

The effect of the support
The catalytic support has two main roles in the tandem in-situ generated H 2 O 2 and epoxidation of propylene. It contains the titanium active sites for the adsorption of propylene and its further epoxidation and it is responsible of the dispersion of the metal species. Additionally, depending on its structure, the support can also play a role in the catalyst deactivation. The presence of Ti sites in tetrahedral-coordination is crucial for catalytic activity in detriment to those Ti sites in octahedral coordination. Tetrahedral coordinated Ti sites are found not only to be favourable for attaining small gold nanoparticles and consequently a high metal dispersion [33] but also to be involved in the adsorption of propylene [14,31,34]. Early studies on the effect of the support showed that isolated and/or highly dispersed tetrahedral titanium sites can be obtained in either mesoporous titanosilicate (TiO 2 -SiO 2 ) or microporous titanosilicate (TS-1). Well dispersed titanium-containing supports result in catalysts more resistant to deactivation and able to obtain slightly higher PO yields in spite of the low propylene conversion values achieved (Table 1) [35]. However, these tetrahedral sites together with neighbouring surface hydroxyl groups (Ti-OH, Si-OH) are found to be responsible of the readsorption of PO [36], which further causes catalyst deactivation by producing bidentate propoxy species, carbonates, carboxylates, and oligomers on the catalyst surface [13][14][15][16][17]30,[35][36][37][38]. Accordingly, it was observed that deactivation with time on stream occurs even when nonporous silica was used as a major carrier of Au catalysts [39,40], indicating that both the surface properties and the hydrophobic character rather than steric structure of the Ti-containing supports are related to the deactivation. Further confirmation of the effect of silanol sites (Si-OH) in the catalyst deactivation [17,38,39,[41][42][43] was given by Qi et al. [44] who observed by IR and HRTEM analyses that the interaction between titanium oxides and commercial non-porous silica can be intensified to form more isolated TiO 4 units with tetrahedral coordination with an increase in heat treatment temperature, diminishing catalyst deactivation. Likewise, the deactivation of Au/TiO 2 -SiO 2 catalysts can be controlled by adding trimethylamine in the reaction feed [41,45], since in this way, silanol sites could be masked. Similarly, Nijhuis et al. [46] observed that the presence of water in the feed gas stream can also suppress the catalyst deactivation by lowering the PO concentration adsorbed on the titania through competitive adsorption, however, an associated small decrease in the catalytic activity is also observed as a side effect.
Consequently, in order to obtain highly active and stable catalysts, it is important to optimise the Ti/Si ratio in the support to get a compromise between tetrahedral Ti dispersion and deactivation caused by silanol groups. Interestingly, a volcano-type effect of the Ti/Si ratio on the propylene conversion is observed while the PO selectivity decreases linearly with the Ti/Si ratio (Fig. 1).
The superior performance in terms of propene conversion, PO selectivity and H 2 efficiency in comparison with the Au catalysts supported on conventional Ti-containing mesoporous materials were reported by Yang et al. [43], who studied the performance of Au/Ti-HMS (titanium-containing hexagonal mesoporous silica). This fact was ascribed to the homogeneous distribution and uniformity of titanium species in combination with the accessible pore structure. Similarly, Xu et al. [48] used a hybrid material composed of a core TS-1 zeolite particle surrounded by a shell of mesoporous silica-gold nanoparticles which is not only more robust against catalytic deactivation but also more selective to PO than Au/TS-1 catalysts.
As previously stated, the support is also partially responsible (in addition to the gold loading method) for the dispersion level of the metal nanoparticles. Consequently, few studies have used high-surface area mesoporous supports for the tandem system. Indeed, gold nanoparticles on Ti-modified mesoporous supports such as MCM-41, MCM-48, TUD-1 and SBA-15 has been observed to improved PO yields [37,47,[49][50][51][52][53][54]. For example, Kalvachev et al. [52] studied the vapour-phase epoxidation of propylene over Au deposited on Ti-MCM-41 confirming the improved performance of mesoporous supports with respect to TiO 2 -SiO 2 mixed oxides. The performance of gold particles deposited on the Ti-MCM-41 and Ti-MCM-48 supports were compared by Uphade et al. [37], who observed a better performance for the Ti-MCM-48. This fact was ascribed to its three-dimensional pore system. At the same time, Qi et al. [39] found that Au/Ti-MCM-41 type amorphous support is one order of magnitude more active than Au/Ti-MCM-48 in terms of both the PO formation rate per unit weight of Au and of hydrogen efficiency. Both studies are not directly comparable as different reaction conditions are used. Sacaliuc et al. [54] synthesised Au/Ti-SBA-15 using two different Ti-SBA-15 supports synthesised by Ti-grafting and direct synthesis concluding that those materials obtained by Ti grafting had greater catalytic activity, which was ascribed to an increase in the amount and dispersion of Ti within the mesoporous silica support. Consequently, it influences the final Au nanoparticle size, emphasising the advantage of using high surface area silica supports for this reaction.
The treatment of the support prior to metal loading has also been explored as a strategy to increase the conversion of propylene whilst keeping the selectivity to PO. Thus, Sacaliuc-Parvulescu et al.  [51] studied the effect of the ammonium treatment of the Ti-SBA-15 support on the catalyst performance, observing a 10 times increase in the PO formation rate. The presence of amine species favours the homogeneous distribution of gold nanoparticles favouring both the in-situ synthesis of hydroperoxo species and the posterior epoxidation. The same effect was observed over the ammonia-treated TS-1 support. In this case, the ammonia treatment results in a fourfold increase of the Au capture efficiency, leading to turnover numbers above 400 mol PO Au mol −1 h −1 at 200 • C [55].

The effect of the metal
As mentioned above, gold supported over Ti-based materials are the most studied systems in the propylene epoxidation reaction ( Table 1). The effect of gold size on the epoxidation of propylene has been scrutinised in numerous studies. Interestingly, Hayashi et al. [11] demonstrated that the reaction pathway can be tuned by carefully controlling the size of the gold particles. Gold particles larger than 2 nm produce PO whereas smaller sized gold particles turn into propane. Conversely, an optimum gold size of ∼1 nm is reported by Lee et al. [56]. In any case, the effect of gold size becomes more complex in the tandem in-situ hydrogen peroxide synthesis and its one-pot use for propylene epoxidation.
Metal nanoparticle size is well known to be directly related to the loading method used in catalyst preparation. The deposition-precipitation (DP) method was used to synthesise gold supported on a Ti-containing silicate mesoporous material (TUD-1) for the direct epoxidation of propylene with H 2 and O 2 mixtures [53]. It was observed that increasing the pH of the deposition of gold nanoparticles decreases not only the final gold loading and but also its particle size in the catalyst, and consequently increases the selectivity to PO. Hydrogen selectivity also increased with increasing pH of preparation as gold particles with sizes above 2 nm are found to be active in the decomposition of H 2 O 2 to water. A similar effect is obtained when using TS-1 as a support (Fig. 2). Low gold loadings and small gold particle sizes (around 1 nm) seems to be the most selective systems for H 2 efficiency and PO formation in the tandem H 2 O 2 synthesis-epoxidation reaction [56]. In general, a direct relationship between pH during the deposition-precipitation method and the PO turnover number is observed except at high pH values (>8.5) when the reactivity decreases. In these cases, the catalysts present partially oxidised Au species while those prepared at lower pH values contain mainly metallic gold. This observation is in agreement to that reported by Sinha et al. [57] for systems using a Ti-MCM-41 mesoporous support. They studied the influence of gold loading and speciation in the selectivity to PO concluding that the presence of metallic gold results in a higher epoxidation selectivity. A different alternative approach was developed by Zhan et al. [58,59] to deposit Au on TS-1. These authors applied a bio-reduction method to load Au using plant extract leading to an increase in propylene conversion and H 2 efficiency (48.9%).
The effect of different promoters in the catalytic system has also been studied by a series of authors. Wang et al. [60] observed that the impregnation with inorganic salts such as CsNO 3 over Au/Ti-MCM-48 catalysts could promote the epoxidation of propene to PO whereas KBr led to the production of propionaldehyde. Similarly,  [63] demonstrated the effect of the presence of alkali cations on promoting the epoxidation reaction to PO versus hydrogenation towards propane. The incorporation of cobalt during the preparation also increases the Au uptake of the support and increases its stability versus deactivation [64]. On the other hand, the presence of Ge on Au/TS-1 catalysts improves both propylene conversion and selectivity to PO due to a decrease of the activation energy of the epoxidation reaction caused by the ligand effect of Ge on the Ti active sites [65].
Although to a lesser extent, the effect of other precious metals apart from gold has been also reported over either titania or titania-silicate supports. Lower propene conversion and poorer PO selectivity have been found compared to the gold-based systems [12,19,20,[66][67][68][69][70][71][72][73][74][75][76] (Table 2). Specifically, Ag/TS-1 catalysts have been studied by different authors [67][68][69][70]. Contrary to the gold case, oxidised silver ions are the main active sites improving the selectivity to PO by the presence of framework titanium species. Optimum preparation conditions to achieve both high propelyne conversion and high selectivity to PO were investigated by Wang et al. [68]. They found that the most efficient catalyst was synthesised by a deposition-precipitation method with a resulting silver loading of 2 wt% in a support with Si/Ti ratio of 64. These authors also reported that the use of K 2 CO 3 as a precipitator improved the stability and selectivity whilst Cs 2 CO 3 enhanced the catalyst stability [68]. Oxidised palladium (II) species in small clusters are also favourable for the formation of PO compared to fully reduced Pd species and large clusters selective to hydrogenation to propane. Metal alloys have also been studied as active sites for this tandem process. In Pd-Pt/TS-1 catalysts, the addition of 0.01-0.02 wt% Pt increases the fraction of Pd(II) species achieving maximum PO yield and selectivity [12]. Jin et al. [77] prepared a compact zeolite membrane of TS-1 on the surface of Au-Pd/TiO 2 -SiO 2 catalyst pellets. This material showed much higher propylene conversion and PO yield than those of the conventional physically mixed catalyst with the same composition.

The effect of the reaction temperature
Despite the inherent difficulties of comparing catalytic data obtained at different reaction conditions, general trends can be drawn related to the effect of the reaction temperature in the overall conversation and selectivity of the tandem system. As expected, the conversion of propylene increases as the reaction temperature increases mirroring the increase in the rate of the reaction. Contrarily, the selectivity towards propylene oxide decreases as the temperature increases. It is worth mentioning that 100% selectivity to PO can be achieved at relatively high temperatures (<400 K) depending on the catalytic system (Fig. 3), however, the hydrogenation to propane and its complete oxidation to CO 2 and water are favoured at high temperatures dramatically decreasing the selectivity to PO [63]. Additionally, the hydrogen efficiency decreases as the temperature increases as the parallel decomposition of in-situ generated hydrogen peroxide to water increases. As mentioned before, this parameter is normally used to assess the economic feasibility of the system, and therefore, it should be a priority when designing catalysts for the direct epoxidation of propylene with hydrogen and oxygen mixtures.

Hydroxylation of benzene to phenol
Phenol is an important intermediate in the chemical industry for the synthesis of drugs, functional polymers and dyes, amongst others. Currently, phenol is produced at large scale from benzene using the so-called Cumene process. This three-step highly energy intensive process leads to low phenol yields with a high production of by-products, mainly acetone and methylstyrene. Different options to this process have been proposed such as the use of nitrous oxide as an oxidant [78]. Alternatively, the direct hydroxylation of the benzene aromatic ring with molecular oxygen and hydrogen has also attracted great interest as a green route for this transformation (see Table 3).
Remias et al. [79] reported the mechanistic insights of the hydroxylation of benzene and cyclohexane using in-situ generated hydrogen peroxide from oxygen and hydrogen with palladium Table 3 Performance of the different catalytic systems in the hydroxylation of benzene to phenol using in-situ generated H2O2. nanoparticles supported on vanadium and iron oxides. The study indicates that the rate limiting step of the overall system is the formation of H 2 O 2 . An electron transfer mechanism with a radical cation intermediate was proposed for the benzene hydroxylation by vanadium or iron oxide which is not affected by the presence of metallic palladium.

The effect of metal and redox promoters
Group VIII metals have been generally used for the direct synthesis of phenol with hydrogen and oxygen mixtures. Negligible or relatively small phenol yields are obtained when only metals are used in the absence of redox promoters [80]. As an example, Miyake et al. [81] demonstrated the direct formation of phenol over Pd, Rh, Ir and Pt supported on silica, however, with an extremely small selectivity compared to water production. Conversely, it was observed that if the catalyst is modified with vanadium oxide, the phenol formation is significantly improved more than fifteen times. Consequently, a series of redox promoters (metal oxides) have been considered to increase phenol yield such as vanadium, wolframium, molybdenum, niobium, zinc, bismuth, manganese, tin, cerium, zirconium, chromium, lanthanum and yttrium oxides. Fig. 4 shows the effect of different main metal oxides in the rate of formation of phenol, directly related to hydrogen yield. Vanadium oxide was generally found as the most effective promoter, Y 2 O 3 and La 2 O 3 being also very efficient to promote benzene hydroxylation (Table 3). Interestingly, a similar enabling effect is observed with a Pd/Cu composite catalyst supported on silica, compared to the inactive Pd-only system, revealing the essential role that the Cu I/II redox couple played in the reaction [80]. Specific to this system, Kitano et al. [82,83] highlighted the importance of the catalysis preparation method showing that if Pd was loaded by means of ionexchange instead of impregnation, the performance of the Pd-Cu composite was nearly doubled.
The role of the promoter is believed to be related to the inhibition of the parallel decomposition of hydrogen peroxide reaction on the surface of the metals. In this way, oxygen adsorbed on monomeric vanadium species is transferred to the metallic sites, increasing the relative amount of oxygen on the surface of the metal which prevents the over-hydrogenation of the in-situ generated H 2 O 2 . As a result, a higher concentration of H 2 O 2 is present in the system, increasing not only the rate of phenol formation but also the hydrogen yield. Moreover, the metal oxide promoters might also have an effect on the dispersion of the precious metals over the support, leading to smaller particle sizes with higher reactivity towards hydrogen peroxide synthesis. Accordingly, Ehrich et al. [84] show the effect of the metal to promoter ratio (Fig. 5) on the final benzene conversion and phenol yield. In their case, the best reactivity is achieved with a low metal loading (0.5 wt% Pt) and a V/Pt ratio of 6. Therefore, optimum catalysts require both a high dispersion of the transition metal oxide and small metal particles.

The effect of the support
Despite silica being the most common support for this tandem process, the literature contains scattered studies using other supports. For example, Tatsumi et al. [85] studied a palladium supported on TS-1 catalyst and demonstrated the enhanced properties of TS-1 for selective hydroxylation of aromatics using in-situ generated hydrogen peroxide, showing an increase of both benzene conversion and phenol selectivity [86]. Resins like Amberlyst or Nafion/silica composites have also been explored as supports to carry out the direct benzene hydroxylation [87,88]. Optimum results are obtained over a strong acidic Nafion/silica composite carrier with 13% Nafion containing 0.5 wt% Pd and 0.5 wt% Pt due to a combination of strong acidic properties, high dispersion of the Nafion resin on the silica matrix and its hydrophobic/hydrophilic properties. Despite this, low phenol selectivity values of ca 41% are reported, being hydroquinone, the main byproduct.

Membrane reactors
Membrane reactors have been significantly used in this tandem process by a series of authors [86,[89][90][91][92][93][94][95][96][97][98][99]. Generally, a Pd-based membrane is used to separate feed streams on both sides of the membrane. On one side, hydrogen is fed into the system while a mixture of benzene and oxygen is introduced in the other side of the membrane. Hydrogen permeates through the metallic membrane following a three-step mechanism: (i) dissociative adsorption of hydrogen into atoms, (ii) diffusion of the hydrogen atoms through the octahedral interstitial sites of the face-centreed cubic palladium lattice and (iii) associative desorption of molecular hydrogen, however, in these kind of tandem systems, permeated hydrogen atoms are believed to react with oxygen before recombination to produce different active oxydant species such as HOO • , HO • and H 2 O 2 in-situ used for the hydroxylation of benzene [86,90]. While HOO • and HO • can be expected to be short lived and localised on the surface of the Pd membrane, H 2 O 2 can diffuse from the membrane to the catalyst [86,90]. Despite the obvious advantages of these membrane reactors in terms of safety, avoiding direct contact of oxygen and hydrogen, phenol yield and hydrogen efficiency remain as challenges (Table 4). Generally, a large amount of by-products (CO 2 and H 2 O) from side reactions are produced whereas the formation of phenol is relatively very low. Over-oxidation of benzene into CO 2 can be caused by the highly oxidative species formed on the Pd membrane. On the other hand, the hydrogen utilisation efficiency decreases due to the excess of hydrogen (respect to oxygen) in the membrane surface, being the production rate of water 500-1000 times higher than that of phenol. Further optimisation of the process and the reaction conditions have been carried out in order to enhance both yields. Thus, Sato et al. [92] demonstrated the importance of the reaction atmosphere: while oxygen-rich conditions can cause the complete oxidation to CO 2 , high hydrogen flux induce hydrogenation activity and consequently, water formation. In general, a balance of the H 2 /O 2 ratio needs to be achieved to maximise the phenol yield obtained as shown in Fig. 6.
Additionally, the reaction temperature plays a vital role in the membrane systems beyond its effect on the reaction due to the thermal membrane stability. Palladium embrittlement takes place below 573 K, initially producing pitch holes in the membrane before its complete rupture due to ␣ → ␤ phase transition [93]. To avoid this, Pd-alloy membranes are normally used [100]. Specifically, Pd-Cu membranes where used in a double membrane reactor, reaching a phenol selectivity of 67% [97]. A complement engineering approach is the use of membrane microsystems to control the localised H 2 /O 2 ratio [94,95].
However, to compete with the commercial well-established Cumene process for large-scale production of phenol, an increase of the overall benzene conversion while keeping high selectivity to phenol, the reduction of the cost of the metallic membrane and its long-term performance, remain important issues to deal with in the near future.

The selective oxidation of methane into oxygenates
One of the great challenges in the field of selective oxidation is the direct transformation of methane (the primary component in natural gas) into oxygenates. Currently, a two-step highly energy intensive process is used where the activation of the C-H bond takes place at high temperatures to form syn-gas prior its further conversion into oxygenates. Evidently, economic and environmental advantages can be gained through the direct oxidation of methane into oxygenates in a single step with careful control of the over-oxidation of the oxygenated products into CO 2 . Pioneering studies in the field used concentrated acids, such as H 2 SO 4 , with Pt-based catalysts [101,102], however, a breakthrough from the environmental point of view is the use of green oxidants Table 4 Performance of the different catalytic systems in the hydroxylation of benzene to phenol using in-situ generated H2O2 in membrane reactors.  [103]. Following the lines of this review, further advantages can be obtained with in-situ generated hydrogen peroxide or peroxo species (Table 5). Lin et al. [104] selectively oxidised methane, ethane and butane in the presence of oxygen and carbon monoxide on a mixture of trifluoroacetic acid and water using Pd/C and copper chloride as a catalytic system. The proposed mechanism shown in Fig. 7, involves a metal catalysed water gas shift reaction and the subsequent formation of H 2 O 2 . Further experiments suggest that the principal role of metallic palladium was to generate hydrogen peroxide in-situ, being the copper chloride species responsible for the hydroxylation of the hydrocarbons into their corresponding alcohols [73,105]. A similar catalytic system was studied by Park et al. [106] who observed that in the presence of copper chloride, the metallic palladium can be partly oxidised into Pd(II) species, responsible of the selective activation of methane and the formation of CF 3 COOCH 3 and formic acid. On the other hand, metallic palladium remains key for the in-situ formation of H 2 O 2 . Heteropoly compounds such as Pd 0.08 Cs 2.5 H 0.34 PVMo 11 O 40 have also been explored as an alternative. In these cases, the formation of hydrogen peroxide is believed to be catalysed by Pd with acidic sites in the heteropoly Table 5 Performance of the different catalysts in the selective oxidation of methane into oxygenates using in-situ generated H2O2.

Refno
Catalyst  Performance of the different catalytic systems in the Fenton process using in-situ generated H2O2.
Chlorophenol degradation = 100% Mineralisation degree = 70% for DCP The consecutive reduction-oxidation protocol was the most efficient. Fig. 7. The mechanism for the selective oxidation of lower alkanes with in-situ generated H2O2. Reprinted with permission from [73].
compound while the hydroxylation takes place in the heteropoly compound itself [107].
Additionally, H 2 /O 2 mixtures are more active than O 2 /CO/H 2 O systems, as in the former case, the water gas shift reaction step is not required. Interestingly, in this case, a chlorine-free system can be used in combination with a heterogeneous Pd catalyst, where the presence of copper (Cu(CH 3 COO) 2 ) remains essential for the activation of methane and metal palladium for the in-situ generated H 2 O 2 [108]. Alternatively, the same research group used a series of vanadium compounds such as NH 4 VO 3 , H 4 PV 1 Mo 11 O 40 or V 2 O 5 to substitute the addition of copper [109]. Although copperbased catalysts are more active than vanadium ones, no difference in product yield was observed. In any case, the absence of copper or vanadium compounds leads to negligible amounts of oxidation products which means that hydrogen peroxide in the presence of Pd/C cannot drive the methane oxidation under these conditions. Similarly, the heavy metal copper can be replaced by p-tetrachlorobenzoquinone (TCQ) used in the industrial synthesis of H 2 O 2 with similar roles in the catalytic system [110].
In addition to palladium-based systems, RhCl 3 along with several equivalents of Cl − and I − ions dissolved in an acidic solvent (mixture of perfluorobutyric acid and water) has also been studied for the direct oxidation of methane with CO/O 2 mixtures [111]. Selectivity towards methanol or acetic acid can be tuned by a simple change of solvent system. A curious aspect of the previous metallic Pd and the homogeneous Rh-based systems is that, apart from their ability to activate both dioxygen and the alkane, both require the presence of a co-reductant (carbon monoxide), showing a remarkable resemblance with the monooxygenases enzymatic system.
As an alternative to the addition of homogeneous co-catalysts, Hutchings et al. [112] have recently demonstrated the direct oxidation of methane to methanol over AuPd/TiO 2 catalysts using either pre-formed hydrogen peroxide or in-situ generated from H 2 and O 2 mixtures. Interestingly, improved methanol selectivity is obtained in the latter case compared to the former with a similar productivity. Bimetallic AuPd/TiO 2 showed a remarkably higher productivity to methanol than a physical mixture of Au/TiO 2 and Pd/TiO 2 with the same metals loadings, showing once again the superior performance of AuPd alloy nanoparticles with unique catalytic properties. In this case, the primary product is methyl hydroperoxide which is believed to be formed by the reaction of On a related study, Bravo-Suárez et al. [113] have studied the selective oxidation of propane to acetone and 2-propanol with a H 2 and O 2 mixture on Au/TS-1 catalyst. According to the spectroscopic results, H 2 O 2 seems to be formed on the Au sites while propane is partially oxidised on Ti centres, demonstrating the formation of hydroperoxo species as true intermediates in the tandem reaction.
Despite the demonstration of the feasibility of the direct oxidation of methane into oxygenated products by in-situ generated H 2 O 2 , the productivities reported are still low and far away from values required for industrial application. Additionally, control of over-oxidation of the oxygenated products into worthless CO 2 remains a challenge for the future strategies to increase yields. Methane is the least reactive of the alkanes, thermodynamically more stable than its oxygenated counterparts (e.g. methanol) which tend to over-oxidise under reaction conditions. Further development of the bimetallic Au-Pd systems seems to have the potential to achieve high selectivity and high methane conversion at low temperatures. Additionally, engineering alternatives such as the use of membranes for a more controllable formation of hydrogen peroxide by H 2 and O 2 dosing might aid towards the medium term implementation of this system at large scale.

The Fenton process
Fenton systems are one of the cleanest and most efficient processes to eliminate particularly toxic compounds in both wastewaters and soil due to its ability to mineralise organic pollutants. The Fenton reaction involves the generation of a hydroxyl radicals (OH • ) via hydrogen peroxide and ferrous ions. Recently, several groups have studied the one-pot process of this system in tandem with the in-situ generation of hydrogen peroxide normally using a combination of iron and palladium-based catalysts for both reactions respectively (Table 6). Contrary to previous processes, most of these studies avoid the use of potentially hazardous O 2 /H 2 mixtures by using a series of H 2 substitutes, mainly formic acid, hydrazine and hydroxylamine [114][115][116]. Although hydrazine and hydroxylamine are, in general, more efficient as H 2 substitutes than formic acid for hydrogen peroxide formation, the controlled production of H 2 O 2 with O 2 /formic acid mixtures results in more efficiency in terms of H 2 O 2 utilisation when combined with the oxidation of organic pollutants. On the other hand, hydrazine/O 2 mixtures are highly efficient for H 2 O 2 production at pH between 7 and 9 which can be highly advantageous when neutral or slightly basic effluents are treated [114]. Alternatively, hydrogen peroxide can be produced in-situ from electro-generated H 2 and O 2 using a Pd/C catalyst, as shown in the degradation of rhodamine B in an aqueous solution [117].
From a catalytic point of view, two different types of systems can be distinguished: (i) semi-heterogeneous, when a heterogeneous Pd-based catalyst is used for the in-situ generation of H 2 O 2 and a homogeneous ferrous ion for the Fenton reaction and (ii) heterogeneous, when both Pd and Fe are supported. Although in both cases H 2 O 2 is formed on the surface of Pd 0 , in the semi-heteregeneous case, H 2 O 2 is released into the solution where it is converted into OH • radicals by the ferrous ions. On the other hand, in heterogeneous bi-metallic Pd-Fe systems, the in-situ generated H 2 O 2 diffuses from the Pd 0 surface towards the Fe 2 O 3 particle where radical formation occurs. Interestingly, Yalfani et al. show that a Pd-Fe alloy does not favour the Fenton reaction but a synergetic effect is observed by close conjunction with metallic Pd and Fe 2 O 3 particles [115]. A variety of supports have been shown to be successful for this tandem system ranging from alumina [114,115], titanate nanotubes [118] and carbon [117]. Although its importance in the overall process is not fully elucidated, there are some indications in the literature of its potential role on the H 2 O 2 diffusion in the heterogeneous case [115], apart from its obvious effect determining metal particle size as previously discussed.
Similarly, Osegueda et al. [119] prepared a catalytic membrane reactor with Pd impregnated on Fe 2 O 3 , CuAl 2 O 4 , TiO 2 or CeO 2 supports for the phenol oxidation reaction, demonstrating the viability for waste water treatment.
These studies clearly show the remarkable potential of in-situ generated hydrogen peroxide in waste treatment oxidation reactions, lowering the relevant current costs of H 2 O 2 production. Additionally, it is important to highlight that the use of bi-metallic Fe-Pd catalytic systems not only to increase the degradation performance but also to decrease the environmental impact of the process by heterogenisation of the Fenton ferrous ion and its associated synergetic effects.

Oxidative desulphurisation
The permitted sulphur content in petrol and diesel fuels is continuously decreased by increasingly restrictive environmental regulations. The current method for sulphur removal from liquid fuel is mainly based on hydrodesulphurisation (HDS) which requires high temperature and pressure, making it a very costly option, especially for deep desulphurisation, to reduce the sulphur levels below 10 ppm. Based on this problem, there are many research and development efforts on alternative methods to eliminate the sulphur compounds from fuels, such as selective adsorption, alkylation, reactive desorption, bio-desulphurisation, oxidation/extraction (oxidative desulphurisation, ODS) or adsorption, etc. The development and application of oxidative desulphurisation are considered to be one of the most desired options due to its sulphur removal capability at mild conditions. Sulphides, such as thioether, thiophene and methylthiophene, which are primary organic sulphur components present in commercial gasoline, can be efficiently oxidised over TS-1, using hydrogen peroxide as an oxidant [120,121]. New research is now demonstrating the efficient removal of such components with insitu generated hydrogen peroxide from H 2 and O 2 mixtures using metal supported on TS-1 based catalysts. Gold [122] and Pd [123] supported on TS-1 seems to be the active species responsible for the direct synthesis of H 2 O 2 whereas oxidation takes place on the Ti sites. Interestingly, a higher oxidation yield is achieved by Pd-modified titanosilicate compared to the impregnated counterparts [123]. Additionally, Song et al. [124] studied the oxidation of benzothiophene (BT) and dibenzothiophene (DBT) over Au/Ti-HMS catalysts observing almost 100% removal of sulphur with a higher stability after 5 consecutive cycles.

The oxidation of sulphides to sulphones
An interesting application of this tandem system is the in-situ generation of hydrogen peroxide for the selective oxidation of sulphides to produce sulphoxides and sulphones feedstocks for commodity chemicals and pharmaceuticals. Karmee et al. [125] reported for the first time the use of a supercritical CO 2 /H 2 O biphasic system for the in-situ generation of H 2 O 2 from H 2 and O 2 and the chemoselective oxidation of thioanisole to methyl phenyl sulphone. Pd nanoparticles supported on SiO 2 was used for the first step while the oxidation takes place on TiO 2 . Similar thioanisole conversions (99%) and selectivity to methyl phenyl sulphone (96%) values are obtained compared to the same system using pre-generated H 2 O 2 as an oxidant [126].

Conclusions and future perspectives
Selective oxidation using in-situ generated hydrogen peroxide requires a two-step tandem process involving the generation of H 2 O 2 on noble metals (mainly Au, Pd, Pt and bi-metallic combinations), normally from molecular oxygen and hydrogen, followed by its use in the oxidation of reactants into selected products, in most cases avoiding their over-oxidation into CO 2 . Although in most cases a combination of two active sites are needed for each of the tandem steps, future catalytic development should focus on single active sites capable of undertaking both reactions in parallel to avoid diffusion limitations of the generated H 2 O 2 . Alternatively, when the need for two active sites is dictated by the reaction mechanism, it has been generally observed that although no synergetic effect is found in their combination, close proximity of both active sites (e.g. metal nanoparticles) greatly enhances the overall process yield. Accordingly, future research should promote the heterogenisation of oxidation catalysts, especially Fenton catalysts, for this specific purpose. In this task, the nature and properties of the support have demonstrated to be of paramount importance as they not only determine the particle size of nanoparticles, site dispersion and their stability versus leaching, but also they, in some cases, present the active site of oxidation. Accordingly, typical supports are based on TiO 2 and SiO 2 with the mesoporous TS-1 material being the one most frequently reported due to the presence of tetrahedral coordinated Ti sites. Oxides of non-noble metals have also been reported to exert a promoting effect improving the catalytic performance. Additionally, the modification of such supports via metal doping still needs to be fully explored to unveil its potential.
This tandem system presents a series of advantages in terms of the environmental benefits associated with the direct synthesis of hydrogen peroxide versus the anthraquinone route and the associated reduced costs to the reduced number of process steps and absence of hydrogen storage. However, these benefits can be masked by the overall lower productivity achieved when compared to those oxidation processes using pre-generated hydrogen peroxide. These low yields are normally related to the low concentration of hydrogen peroxide achieved in-situ and/or its parallel decomposition in water. In both cases, careful analysis of the process is required in order to balance the kinetics of both tandem steps. Additionally, further strategies need to be developed to enhance the overall H 2 efficiency of the system as a key parameter determining the process economic feasibility.
More importantly, the tandem system seems to efficiently promote the in-situ formation of hydroperoxo species able to activate relatively inert compounds such as methane and selectively oxidise them into their oxygenated counterparts. As an exciting example, bi-metallic AuPd catalysts have recently been shown to achieve higher yields in the oxidation of methane to methanol in the one-pot system compared to the use of pre-generated hydrogen peroxide. This development has been possible thanks to the synergetic combination of the efficiency of these catalysts towards the direct synthesis of H 2 O 2 from H 2 and O 2 with high selectivity and towards selective oxidation achieved by the catalytic development carried out in Hutching's group. Similar strategies and a deep understanding of the complex reaction mechanisms are needed to reveal the full potential of these tandem processes to efficiently promote the selective oxidation rather than the complete oxidation to CO 2 . Enhancing the stability of these catalysts during the selective oxidation is also a critical issue for future commercial applications. Even though Ti-modified silica support is known to enhance the stability of metal nanoparticles, there is little understanding of the underlying basis of this phenomenon, demanding a more systematic study with the assistance of a model system.
Finally, considerations should be given to the safety issues associated to the flammability of the H 2 /O 2 mixture in the tandem systems. Some studies have explored a series of O 2 and/or H 2 substitutes such as formic acid and hydroxylamine or the use of supercritical solvents. An even more attractive alternative is the use of membrane reactors to avoid the direct contact of oxygen and hydrogen in the gas phase while taking advantage of the atomic hydrogen diffusion mechanism to enhance the oxidant formation. Despite the potential of these systems and the improved safety, the over-oxidation control and the stability of the metallic membrane remain challenges.