Metal-support interactions in the design of heterogeneous catalysts for redox processes

General principles

Specific properties of supported metal NPs are affected by many factors such as their size, shape, oxidation state and interaction with the support, etc. [4], [5], [6]. Particularly, the metal particle size is crucial for the performance of a catalyst in structure-sensitive reactions. Many authors believe that HDC should be attributed to this type [7], [8], [9], [10]. Therefore, the preparation of size-selected metal nanoparticles as well as the increase in their resistance to sintering by tuning MSI is of great interest.

Another significant problem in the development of HDC catalysts is the oxidation of the active component under the influence of chlorinated reactants, intermediates or products that inevitably include HCl [11], [12]. The right choice of support gives instrument to smoothen such action; thus, zirconia is able to capture chloride ions, so Pd/ZrO₂ catalyst is much more stable to chlorination than SiO₂/AlPO₄ and MgO-supported ones, providing the highest chlorobenzene conversion [13]. As for MgO support, it dissolves under reaction conditions due to the effect of the HCl, which leads to the destruction of the catalyst. Even more effective in Cl^- trapping is sulfated zirconia, ensuring in addition the smaller size and the uniform distribution of metallic NPs on the outer surface of Pt–Pd catalyst and its high stability in HDC reaction [14].

Very profitable is the ability of a support to provide the hydrogen spillover as a contribution to the synergistic action with the active metal [15]. It resulted in the increase of the reaction rate of chlorobenzene HDC, when Pd was deposited on the oxide supports (Al₂O₃ or SiO₂), in comparison with bulk Pd catalyst or Pd NPs deposited on carbon supports (activated carbon, graphite or graphitic nanofibers).

In a significant number of researches, the role of the MSI is associated with a change in the electronic state of the active metal. For example, some authors consider the presence of Pd in two oxidation states as the necessary condition of high activity of Pd catalysts for CCl₄ HDC [16], [17]. The presence of Pd in several oxidation states provides advantage for HDC reaction according to the mechanism proposed by Gómez-Sainero et al. [17]. In this mechanism R–Cl activation proceeds on Pdⁿ⁺ to form [Pdⁿ⁺–Cl] covalent bond, Pd⁰ coordinates electrophilic R⁺ ion by a donor–acceptor bond and activates chemisorbed H₂ that then reacts with [Pdⁿ⁺–Cl] to form HCl and with R⁺ to form RH. The electronic state of metal on the surface of catalyst can be tuned using appropriate preparation technique. Thus, the strong MSI helps to obtain the electron-deficient form of palladium (Pdⁿ⁺) [18].

The Pdⁿ⁺/Pd⁰ and Rhⁿ⁺/Rh⁰ ratios also play a significant role in the catalytic performance of Pd and Rh NPs in HDC of 4-chlorophenol [18], [19]. A crossed effect of particle size and Rhⁿ⁺/Rh⁰ ratio was observed [19]. As the size of Rh nanoparticles decreased the activity increased reaching the maximum at 2.8 nm, with the lower size values leading to a significant decrease in the activity. The increase in the relative amount of Rh⁰ increases the catalytic activity. For Pd nanoparticles the authors of work [18] observed the positive effect of the increase in Pdⁿ⁺/Pd⁰ ratio on the catalytic activity calculated per surface area. In the highly active NPs Pdⁿ⁺/Pd⁰ ratio was close to the optimum reported in the literature [17]. It has been suggested that the optimal Pdⁿ⁺/Pd⁰ ratio provides many low-coordinated palladium atoms adsorbing the reagents and products.

The significant influence of the calcination temperature on the selectivity of Pd/Al₂O₃ catalyst in 4-chlorophenol HDC is connected by the authors of the work [20] with the increase in Pdⁿ⁺/Pd⁰ ratio (from 1.15 to 2.17 when increasing calcination temperature from 200 to 400°C) due to the changes in Pd particle size. Electronic effects are manifested as participation of coordinately-unsaturated atoms of surface defects in the catalytic action; thus, HDC of chlorophenols at low temperatures mainly occurs on the defects on the Pd surface on which the reagent is adsorbed by the formation of the σ-complex [21].

MSI in Ni/Al₂O₃ systems

Chemical MSI, e.g. by incorporation of metal ions into the support structure can cause the changes in the electronic state of the supported metal, as it was observed for Pd on pillared clay [22]. The resulting catalyst demonstrates high activity in aqueous-phase HDC of 4-chlorophenol under mild conditions using formic acid as the reducing agent. Often MSI leads to the formation of mixed oxides or spinel, in the volume or on the surface of a catalyst, thus providing the presence of partially oxidized metal. Classical example of such process is the MSI in Ni/Al₂O₃ catalysts [23] where spinel formation is widely used to tune Ni particle size and reduction ability, and therefore catalytic properties [24]. The changes in the reduction temperature can be registered by temperature-programmed reduction (TPR) method. When nickel ions are incorporated into alumina support or decorated by aluminum ions, the reduction peak is expected to shift to higher temperatures due to the formation of surface nickel aluminate-like species [25].

Both nickel oxide and aluminate comprise a cubic close-packed lattice of oxide ions. The structures differ only in the way of filling of cation vacancies. At the first stage, Al³⁺ ions replace Ni²⁺ in octahedral positions without noticeable distortion of the structure, but with the simultaneous creation of cation vacancies, to form the structure called substituted nickel oxide. When the degree of Ni substitution with aluminum increases, some of the aluminum ions begin to occupy tetrahedral positions, as occurs in spinel, to form a disordered spinel-oxide intermediate. The feature of such material is a higher disordering degree than in conventional nickel aluminate produced by the interaction of nickel and aluminum oxides. Such structures can form in the locations where NiO contacts with Al₂O₃.

Nickel strongly bounded with γ-Al₂O₃ support is preferentially formed due to incorporation of nickel ions into the tetrahedral vacancies [26]. Because of low concentration of tetrahedral vacancies in κ-Al₂O₃, nickel weakly interacts with this support. The nature of nickel-support bounding determines the catalytic properties in liquid phase selective hydrogenation of isoprene. The Ni/γ-Al₂O₃ catalysts with strong metal-support interaction have enough hydrogenation sites, resistant to carbon deposition, and they provide higher isoprene conversion at higher stability. In contrast, Ni/κ-Al₂O₃ with the same nickel loading undergoes significant coking due to the presence of hydrogenolytic sites.

The MSI can also have a negative or indirect impact on the catalytic activity. For example, supporting of Ni on porous Al₂O₃ by atomic layer deposition led to the NiAl₂O₄ spinel formation [27]. The resulting catalyst demonstrated the low initial methane conversion in dry reforming. But then, under the influence of CO and H₂ that are formed as the reaction products, the spinel was reduced to form highly dispersed and stable nickel NPs, providing the increase in the catalytic activity.

The preparation method and the nature of nickel precursor drastically affect the type of nickel species in alumina-supported catalysts, and influence on the phase state of alumina in the final catalyst. The adsorption of Ni²⁺ by controlled adsorption method can prevent NiAl₂O₄ formation at T<600°C [24]. Spinel formation occurs only at heating in He at 700°C, which is much higher than typically used in dry impregnation.

As for the nature of a metal precursor, it was found that the Ni/Al₂O₃ catalyst prepared from NiCl₂ contains NiO weakly bonded with the support and strongly bonded nickel aluminate species [28]. Relatively large nickel particles are formed during reduction of such precursor, and the final catalyst provides the lower initial conversion of 1,1,2-trichloroethane in HDC reaction than the catalyst prepared from Ni(NO₃)₂. On the other hand, long-term stability of the catalysts prepared from NiCl₂ and NiSO₄ were higher because of the large size of Ni NPs and their sulfated surface which prevent deactivation caused by sintering of the active component.

Usually, the MSI is enhanced when the metal loading decreases. To suppress such interaction, metal NPs can be formed and stabilized before deposition on support. In this way, laser electrodispersion (LED) method [29] developed in Ioffe Physical-Technical Institute of the Russian Academy of Sciences (St. Petersburg, Russia) is very promising. In this method the catalysts are prepared by the splashing of metal droplets from the surface of metal target irradiated by a laser torch, followed by their cascade fission and uniform deposition of amorphous spherical nanoparticles with a strictly fixed size on the outer surface of a support. The size of final NPs is determined only by the electron work function of the metal (e.g. 2 nm for Pd NP). The metal loadings are extremely low (less than 0.01 wt.%). Cu, Pd, Ni, Au, Pt-containing catalysts produced by LED method were supported on carbon [highly oriented pyrolytic graphite (HOPG) and Sibunit] or oxide (silica, surface oxidized silicon wafer SiO₂/Si and alumina) supports. The wide investigation of catalytic properties of such systems [30], [31], [32] demonstrates, that the structure of active sites and, consequently, catalytic activity depends on the nature of metal and support, and therefore on the type of the MSI.

The feature of LED catalysts is their stability to metal particles sintering even at significant degree of surface coverage because of the presence of very thin oxide layer on the surface of each separate metal particle. As a result, the small size and uniform particle size distribution do not change with increasing the metal loading, in contrast to the catalysts prepared by common methods. An additional peculiarity of LED catalysts on dielectric supports is the charged state of individual metal particles, formed due to thermally activated tunnel charge transfer between neighboring non-contacting metal NP. This phenomenon is of particular importance in the case of the formation of nanoparticle ensembles on the surface of non-conductive supports. It explains the extreme dependence of the catalytic activity of LED systems on the degree of the support surface coverage with the deposited metal NPs. Therefore, such systems are suitable for studying the effect of charging of supported metal NPs on the catalytic activity.

Early, the effect of the average distance between metal NPs on the catalytic activity was observed for the nanocomposites in which NPs were stabilized in a polypara-silylene polymer matrix [33], [34], [35]. The maximum of catalytic activity in the isomerization of chlorinated olefins was attained for metal-polymer nanocomposites where the average distance between neighboring metal NPs was several nanometers. This distance is small enough to permit tunnel electron transfer between metal nanoparticles. It was also shown that the steep rise in the conductivity of these structures occurs at the same metal content at which maximal activity in isomerization was observed.

Dielectric (thermally oxidized silicon, comprising thick layer of surface SiO₂) or semiconductor (naturally oxidized Si with very thin SiO₂ layer on the surface) supports were used to reveal the effect of electron transfer between Cu, Ni, or Pd nanoparticles on the catalytic activity in several reactions (dichlorobutene isomerization, CCl₄ addition to 1-nonene, CCl₄ interaction with decane) [36], [37]. The strong dependence of the catalytic activity on the distance between metal particles was observed. Thus, the curve of catalytic activity vs. metal loading for the Cu films deposited on the dielectric support (SiO₂/Si) in dichlorobutene isomerization passes through a maximum since the number of charged particles is also maximal in densely packed one-layer films. For Cu NPs deposited on a conductive support (naturally oxidized Si) the appearance of such curve is different: the catalytic activity reduces with the increase in the surface density of Cu NPs. The possibility of metal-support electron transfer shifts the charge balance in the system of supported NPs. Very interesting results were observed for Ni-containing systems which are characterized by two types of interactions between metal and support: the electron transfer and the chemical interaction. So, catalytic activity of Ni NPs supported on the SiO₂/Si dielectric support in dichlorobutene isomerization greatly reduces with increasing particle density. But on the conductive support the surface density of Ni NPs does not affect the activity of Ni films due to possibility of the metal-support electron transfer.

The MSI of chemical nature has a significant effect on the catalytic properties even when the preliminary reduced metal NPs are deposited on a support. This effect is manifested in Ni/Al₂O₃ systems prepared by the traditional methods. For LED systems, the strength of the MSI and its results depend on the nature of the supported metal. When Pd or Au was supported on alumina by LED, no indications of their oxidation were found by XPS [31], [32]. In contrast, XPS data confirm that the strong MSI does take place between alumina and Ni particles deposited by LED, despite reduced state of Ni NPs before deposition [31]. However, this type of MSI does not adversely affect catalytic activity. Indeed, in the whole studied temperature range from 100 to 350°C the chlorobenzene conversion on LED Ni/Al₂O₃ was higher than on LED Ni/C with the same Ni loading, even though in the latter nickel is contained only in the metallic state (Fig. 1).

Fig. 1:

Comparison of the CB conversion vs. the reaction temperature for alumina- and C-supported Ni catalysts prepared by LED. Adapted from [31].

Similar results were obtained in [38]. The analysis of Ni2p XPS spectra of the Ni/Al₂O₃ (LED) catalysts comprising 0.005 and 0.007 wt.% Ni reveals the presence of NiAl₂O₄, whereas the catalyst with similar Ni loading on the carbon support (Sibunit) contains only Ni⁰. The signal of metallic nickel appears in Ni/Al₂O₃ systems only at as high nickel loading as 0.03 wt.%, because at this loading the monolayer of Ni NPs is already formed on the surface of alumina granules. Thus, the NPs of the second layer have no direct contact with the alumina surface and no possibility for the MSI. The distinctive feature of Ni/Al₂O₃ (LED) samples with relatively low Ni loading is the high initial chlorobenzene conversion at each reaction temperature, and its decrease during time-on-stream in vapor-phase chlorobenzene HDC.

The strong MSI is characteristic for another type of catalyst, in which preliminary prepared Ni⁰ NPs were deposited on alumina from its colloidal dispersion (CD) in hexane stabilized with hexadecylamine. According to XPS, in the 0.1 wt.% Ni/Al₂O₃ (CD) only 5% Ni is in the metallic state [38]. The rest of the nickel is oxidized and presented in the form of spinel or hydroxide. This catalyst shows activity in chlorobenzene HDC only at high temperatures (250–350°C), when the reduction of relatively weakly bound Ni species with hydrogen presented in the reaction atmosphere can occur. The difference between LED and CD catalysts of a similar composition is most likely associated with the different size of initial Ni particles: 1.5–2 and 4–5 nm for LED and CD systems, respectively. Therefore, the active sites of different nature are formed in these catalysts after reduction of Ni²⁺ under reaction conditions.

Attempts were made to reduce the degree of the MSI in Ni/Al₂O₃ systems by modification of a support by a thin layer of Au or by Si, Mo, W-heteropolycompounds [39], [40], [41] prior to Ni deposition.

XPS data demonstrate that the gold on the alumina surface in low metal-loaded LED catalysts is presented predominantly in the metallic state, although a very small contribution from the charged state, Au^δ+, cannot be ruled out [39]. Sequential deposition of gold and nickel layers on alumina by the LED method does not affect the electronic state of nickel. The LED systems in which Ni NPs were deposited on Al₂O₃ or on Au/Al₂O₃ comprise oxidized form of nickel with the same binding energy (856.23 eV) corresponding most likely to nickel aluminate. The similar electronic state of Ni in gold-modified and non-modified systems results in almost complete coincidence of the conversion vs. temperature curves obtained for CO oxidation (Fig. 2).

Fig. 2:

Carbon monoxide steady-state conversion in the presence of Ni/Al₂O₃ (LED) and Ni–Au/Al₂O₃ (LED) catalysts. The curves were obtained in the heating (solid lines) and cooling (dashed lines) modes in micro-catalytic pulse reactor at stoichiometric CO/O₂ molar ratio (adapted from [39]).

In contrast, deposition of heteropolycompounds (HPC) of the Keggin type (K₄SiW₁₂O₄₀ and K₄SiMo₆W₆O₄₀) on alumina strongly affects the MSI in Ni/Al₂O₃ system [40], [41]. The changes in the nature of supported nickel species due to such modification are clearly seen from TPR profiles of catalytic precursors.

Thus, the comparison of TPR profiles for non-reduced precursors of Ni/Al₂O₃ (Fig. 3, curve a) and Ni/HPC/Al₂O₃ (Fig. 3, curve b) reveals the difference in Ni reduction process [40]. First peak of hydrogen consumption at 470°C in the TPR profile of Ni/Al₂O₃ corresponds to the reduction of pure NiO, while the broad peak with the maximum at 470°C can be explained by the reduction of highly dispersed NiO interacting strongly with alumina.

Fig. 3:

TPR profiles for non-reduced precursors of Ni/Al₂O₃ (a) and Ni/HPC/Al₂O₃ (b).

Two broad peaks are present in the TPR profile of modified Ni/HPC/Al₂O₃ (Fig. 3, curve b). The first low intensity peak with the maximum at 280°C most likely corresponds to the reduction of small NiO particles, while the intense peak at 360°C is associated with the reduction of larger NiO particles. The shift of the adsorption maximum into the high temperature range relative to non-modified Ni/Al₂O₃ indicates possibly stronger MSI in the Ni/HPC/Al₂O₃ system.

The results of this work show that modifying of Ni/Al₂O₃ with HPC improves the catalyst activity in HDC of chlorobenzene [40] and phenylacetylene hydrogenation [41]. It was found that modification of Al₂O₃ by HPC affects the catalytic activity in several ways: provides the formation of new centers of dissociative hydrogen adsorption, promotes the uniform distribution of metal NPs over the surface, changes the Ni interaction with alumina support, and promotes the spillover of activated atomic hydrogen from the metal active centers onto a HPC or MeO₃ oxide (Me=Mo or W) formed by decomposition of HPC during the thermal treatment of the catalysts at preparation stage.

Modification of Al₂O₃ by HPC also results in the change of the selectivity in phenylacetylene hydrogenation, depending on the nature of HPC [41]. The main difference between the two catalysts modified by different HPC appears at 250°C when an appreciable amount (12%) of alkylated and condensed side products are produced in the presence of Ni/W-HPC/Al₂O₃, while in the presence of Ni/W,Mo-HPC/Al₂O₃ the formation of the side products is substantially lower (up to 5% of the alkylated derivatives of ethylbenzene, with no condensed derivatives). High hydrogenating ability of Ni/W-HPC/Al₂O₃, in contrast to Ni/W,Mo-HPC/Al₂O₃, results from the formation of new centers of hydrogen dissociative adsorption, and by the appearance of a surface layer of atomic hydrogen that can spill from metallic acid centers to a W-HPC.

MSI in zirconia-supported catalysts

Even though the effect of chemical interaction (as a type of MSI) on catalytic activity is most often manifested in Ni/Al₂O₃, the catalytic properties of metal NPs deposited on the other types of oxides sometimes can be also explained in terms of the chemical interaction. Thus, strong bonding of deposited metals with the zirconia support can provide effective active centers formation, as in the case of the catalysts, produced by the deposition of Pd(OH)₂ on pristine or alumina-modified ZrO₂. High activity of such systems in multi-phase HDC of polychlorinated benzenes [42], [43] can be connected with the formation of binary palladium-zirconium double oxide at the stage of the catalyst precursor annealing. Indeed, the presence of this compound was confirmed by TPR study. Along with the peak of PdO reduction at approx. 80°C, TPR profiles (Fig. 4b) of such samples contain an additional peak at approximately 300°C, which can be attributed to the reduction of Pd_xZr_yO_z, because the reduction of zirconia proceeds at the temperatures much higher than 300°C. The formation of Pd_xZr_yO_z correlates with the catalytic properties: the highest activity in HDC of 1,3,5-trichlorobenzene (1,3,5-TCB) demonstrated the catalyst prepared from PdO/(1 mol.% Al₂O₃+ZrO₂) precursor (Fig. 4a), for which the maximal H₂ consumption during TPR was found [42]. A somewhat lower catalytic activity of Pd/ZrO₂ can be explained by low thermal stability of pure zirconia in comparison with alumina-modified ZrO₂.

Fig. 4:

Kinetic curves of 1,3,5-TCB conversion in the presence of Pd-containing catalysts on various supports (a) and TPR profiles of corresponding non-reduced catalysts precursors (b).

The concentration of Pd_xZr_yO_z in Pd/ZrO₂ depends on the preparation technique. In [44] such system was prepared without template or using biotemplates (cellulose or wood sawdust). In the first case crystal ZrO₂ was used as a support and palladium was deposited on its surface from hydroxide. The adsorption of Pd(OH)₂ on the acidic centers on the surface of crystalline ZrO₂ and the subsequent annealing at 600°C ensure the strong bonding between palladium oxide and the support, which is evident by the appearance of the peak at about 300°C in TPR profile. In contrast, in templated synthesis the crystallization of ZrO₂ and PdO phases proceeds during thermal removal of the template. In this process mainly individual ZrO₂ and PdO oxides are formed, so the contact between these components is not so strong, as in the previous case. Therefore, only little amount of Pd_xZr_yO_z is found in biomorphic samples by TPR. Vapor-phase HDC of chlorobenzene on these catalysts is accompanied with hydrogenation of benzene to form cyclohexane, and the selectivity to this product depends on the degree of PdO-ZrO₂ interaction.

The improvement of thermal stability of zirconia upon modification with yttrium and gallium oxides influences on the catalytic activity of palladium catalysts in multiphase hydrodechlorination of 1,3,5-TCB [43].

General principles

It is generally believed that carbon-supported metals are subjected to the MSI in significantly lower extent in comparison with metal oxides-supported. Moreover, supporting on the carbon materials prevents oxidation of transition metals, even at small loadings, which is usual when they are supported on chemically active surface of oxides, as discussed above for Ni-alumina systems. Covering of surface with carbon is a promising way both to preserve mechanical hardness of an oxide support and to prevent strong MSI. Thus, glucose pyrolysis in inert atmosphere after supporting on the surface of SiO₂ provides tunable carbon coating formation [45]. By simple increase the weight ratio of glucose to SiO₂ from 0 to 1.5, the size of supported Co particles can be decreased from 13 to 5 nm, with the cooresponding improvement in the CO conversion in Fisher-Tropsh (FT) synthesis, and decrease of C₅₊ selectivity.

Carbon materials are very attractive for the use as supports due to their relatively high chemical stability, electroconductive properties facilitating electron transfer, and availability of a large variety of materials with tunable structure, texture and morphology. Many of new types of carbon materials have been developed in last years; the majority of them can serve as promising catalyst supports.

Thus, graphene and many graphene-based nanomaterials (nanofibers, nanotubes, nano-onions, nanoflakes etc.) have very large specific surface area providing enough area for highly dispersed metallic particles [46], [47], [48], [49], [50], [51], [52], [53]. However, the presence of various functional groups on the surface of carbonaceous supports, the peculiarities of their structure, and even coking can significantly influence the catalytic properties of supported metal NPs through the mechanisms which are usually attributed to the MSI. Unusual type of catalysts is formed by metal-carbon composites produced by the intended coverage of metal NPs with carbon layers. In such systems the MSI can affect the catalytic properties.

MSI in carbon-supported LED catalysts on the base of metal NPs

Weak MSI is expected for carbon-supported catalysts, produced by LED method, which was described above by the example of oxide supports, because metal nanoparticles are formed before the contact with a support. In the published works, mainly two carbon materials were used for deposition of metal NPs by LED method: HOPG having mostly flat surface, and the granules of Sibunit, the material composed of carbon black and pyrolytic graphite [54], [55], which is known as a promising support for the catalysts, providing acceptable mechanical strength and attrition resistance. Metal NPs, deposited by LED, are located on the outer surface of a support, so Sibunit-supported LED catalysts have core-shell structure, making possible recording and comparison of their XPS spectra at very low metal loading. Carbon-supported LED catalysts demonstrate extremely high TOF values in several metal-catalyzed reactions in comparison to their counterparts with traditionally higher loadings prepared by common techniques (impregnation, deposition, precipitation etc.) (Fig. 5).

Fig. 5:

The comparison of the Pd/Sibunit and Ni/Sibunit catalysts prepared by LED with Pd/ND one prepared by wet impregnation on the base of TOF values in vapor-phase chlorobenzene HDC (150°C).

At the first glance, indeed the MSI in LED catalysts seems negligible: no signs of Pd oxidation were observed in XPS spectra of 0.0002 wt.% Pd/Sibunit, as demonstrated by Pd3d_5/2 peak with the binding energy of 335.6–336.1 eV [31], [32]. However, for non-precious metals e.g. Ni the MSI does take place even at contact with carbon supports, although the degree of interaction in this case is weaker. Thus, Ni on HOPG and Sibunit supports is presented both as Ni⁰ (Ni2p_3/2 binding energy from 852.8 to 853.2 eV) and Ni²⁺ (binding energy from 856.0 to 856.3 eV). An increase of Ni loading leads to a transition from a predominantly oxidized to the metallic state of nickel. In contrast to their counterparts prepared by common methods and usually highly loaded with metals, for LED catalysts such transition is observed at much lower loadings, about 0.01 wt.% [56]. The XPS spectra of Ni/HOPG comprises the Ni2p_3/2 line with the binding energy of 856.0 eV, accompanied by intense shake-up satellite, attributed to Ni²⁺ in hydroxide, Ni salts etc., but not NiO [57], [58]. The second narrow line in Ni2p spectra is observed at binding energy of 853.1 eV which is slightly shifted to higher energy in comparison with the position characteristic for bulk Ni⁰ (852.7 eV [59]). This shift can be caused both by small size of Ni particles and by the MSI. The fraction of oxidized Ni significantly decreases in the XPS spectrum of Ni/HOPG when Ni loading increases from 0.1 to 0.5 monolayers.

Not only metallic state of Pd, but also the high stability to oxidation under the influence of HDC reaction medium, comprising hydrogen chloride as the reaction product, was found for LED 0.0004 wt.% Pd/Sibunit catalyst in [32]. No shift in the position of the Pd 3d_5/2 line (336.1 eV) in the XPS spectrum was observed after prolonged catalytic test in vapor-phase HDC of chlorobenzene in flow-type fixed-bed reactor. Such high stability which is also typical for other LED-prepared Pd catalysts can be explained in terms of the high Pd concentration in the surface layer of catalyst granules, and amorphous state of metal NPs, also confirmed by TEM results.

The MSI can strongly depend on the crystallinity of metal particles. In [60] scanning tunnel microscopy (STM) method was used to compare the properties of LED-produced Ni/HOPG systems, comprising predominantly amorphous Ni, and the catalyst of similar composition, produced by supporting of crystalline Ni nanoparticles stabilized as reverse micelles in colloid dispersion (CD) in hexane. On the base of current-voltage curve it was demonstrated, that LED system comprises not only conducting areas of Ni⁰ but also non-conducting regions indicating the presence of NiO, whereas only the latter were found in Ni/HOPG (CD) catalyst with the similar Ni particle size. Moreover, such difference influences the process of nickel oxide reduction by H₂, investigated in the special cell of STM designed for treatment of samples with gases. In CD catalyst the reduction starts at 400°C as for common Ni catalysts and only after prolonged contact with H₂. In contrast, NiO reduction in LED Ni/HOPG systems proceeds, albeit slowly, already at room temperature [61] because of the hindered MSI, due to the arrangement of nickel oxide mainly on the surface of metallic nickel particles. However, the rate of amorphous NiO reduction increases much at higher temperatures. The long-term preservation of non-conductive areas on the surface of the systems during processing with H₂ indicates that the reduction of NiO/HOPG does not start on the surface but in the bulk of NiO particles. This result corresponds to the reduction mechanism proposed in [62]. Good reducibility of NiO in Ni/C LED catalysts is very convenient for HDC catalysis, improving their stability in reaction medium comprising HCl, where the major deactivation reason is the oxidation of the active component.

Nickel catalysts deposited by LED on carbon supports (e.g. Sibunit) show very promising catalytic performance in vapor-phase HDC of chlorobenzene (CB). In addition to low cost of nickel, their advantage is the increase of activity with decreasing Ni loading. Indeed, CB steady-state conversion at 200°C increases from 4 to 52% when Ni content in Ni/Sibunit decreases ten times, from 0.01 to 0.001 wt.%. This conversion is only slightly lower than the values characteristic for similar Pd catalysts [63]. These features provide possibility to develop cheap and efficient catalysts for disposal of such toxic xenobiotics as chlorinated hydrocarbons.

However, Pd/C LED catalysts are also very attractive for the industrial use because of very high TOF values at very low Pd loading, and high stability of Pd and carbon support to the chlorination and oxidation under the influence of HDC reaction medium [31], [32]. The uniform size of Pd particles (about 2 nm), their amorphous state, and their location on the outer surface of the support granule or flat support are the main features of Pd/C LED catalysts providing their unique efficiency [64], [65], [66]. Thus, TOF value in the vapor-phase CB reductive conversion on 0.0004 wt.% Pd/Sibunit LED catalyst at 150°C was found to be 2.3×10⁵ h⁻¹ (Fig. 5), that is three orders of magnitude higher than that on 5% Pd/ND (ND – nanodiamond), prepared by impregnation technique (5.7×10² h⁻¹). As it will be demonstrated below, ND-supported catalysts often show better performance in HDC than active carbon supported counterparts. Palladium-containing LED catalysts on the Sibunit support are effective not only in vapor-phase, but also in liquid-phase HDC. The TOF value of 1,4-dichlorobenzene processing was 1.3×10⁶ h⁻¹ on 0.0004 wt.% Pd/Sibunit with the granule size of 2.5–3.0 mm prepared by LED. For similar catalysts prepared by wet impregnation TOF value of 0.6×10⁶ h⁻¹ was found for 0.001 wt.% Pd/Sibunit with the granule size of 0.6–1.2 mm, and only 370 h⁻¹ for 5 wt.% Pd/Sibunit with the granule size of 2.5–3.0 mm [67].

As discussed in the first section, the charged state of individual metal particles on semiconductor oxide supports creates the possibility of reagents activation due to electronic effects. It would seem that the charged state of separated metal NPs is impossible on the surface of conductive carbon supports due to the inability of electron tunneling between individual NPs. However, it was found that similar mechanism can take place, if metal particles were covered by oxide layer prior to deposition on a conductive support. In this case the electric field generated by charged particles will be located predominantly in the gaps between the closely located particles, and its strength will depend on the thickness of the oxide layer and the difference in the work function of metal and conducting support. The fewer particles are located on the surface, the less likely is the tunneling of the electron between them [56]. Therefore, this effect will be more pronounced for the low metal-loaded systems. The molecules of reagents, when adsorbed in the gap between metal particles, will be additionally activated by the strong electric field. The presence of a charge on active metal particles will play a particularly positive role in reactions in which the key stage is the transfer of an electron between an active center and reagent. Therefore, in contrast to dielectric oxide supports, the catalytic activity of Pd LED catalysts on a conductive carbon support, e.g. Sibunit, in HDC reactions increases with decreasing metal content (see Fig. 5).

The authors of work [31] on the base of TEM investigation supposed that LED catalysts can contain, in addition to size-selected metal NPs (about 1–1.5 nm for Ni, about 2 nm for Pd), the fraction of very fine dispersed active metal, possibly in the form of single atoms or very small clusters. Heterogeneous catalysts on the base of supported metals often comprise different types of particles, including NPs, clusters and single atoms. The latter are harder to detect by microscopic methods, in comparison with NPs. However, in some cases it is the small fraction of individual metal atoms on the surface of the support that can provide the main catalytic effect. Such particles can demonstrate unique properties, including those reflecting the MSI. Some of their features will be discussed in the next part.

The role of functional groups on the carbon surface in the MSI

Recently single atom catalysts (SACs) become to attract a significant attention as promising model systems to elucidate the role of metal interaction with carbon supports [68]. This type of catalysts is very efficient in several important catalytic reactions. Thus, the stabilization of Pt clusters and single atoms produced by atomic layers deposition on graphene nanosheets [69] confirmed by high-angle annular dark field mode STEM, provides catalytic system for electro-catalytic methanol oxidation that is very stable to CO poisoning.

It is evident that the catalyst comprising supported metal in the form of single-atom will provide the best efficiency of metal use. However, very high surface energy will cause fast sintering of metal atoms, if they are not strongly anchored on the support surface. Such stabilization can be achieved by forming a bond with the support, through surface functional groups already existing on the surface or introduced intentionally.

Single atom of a metal can be anchored to graphene lattice by the direct bonding or with an intermediate bridge. Functionalization of carbon materials with oxygen and nitrogen-containing groups opens the way for stabilization of very small NPs, clusters and even single atoms of metal. Zhang et al. demonstrated [70] that single Co atom can be fixed on the N-modified surface of graphene by coordination with four pyridinic nitrogen atoms within graphitic layers to form Co–N=species. Such a unique structure exhibited excellent activity, chemoselectivity and stability for the synthesis of aromatic azo compounds through hydrogenative coupling of nitroarenes. For Co/CNT catalyst containing larger Co NPs it was found that the surface geometry of CNT is much important factor for its stability in FT synthesis than the presence and the nature of surface oxygen-containing functional groups [71].

Single Fe sites confined in a graphene matrix showed an excellent catalytic performance for the four-electron reduction of dioxygen to water [72] and oxidation of benzene [73]. The presence of topological defects causes local disruption of the smoothness of graphene layers providing the sites for metal atom or cluster adsorption. Qiu et al. [74] anchored single-atom nickel dopants to three-dimensional nanoporous graphene by physical adsorption onto the hollow centers of the graphene lattice, as it was found by STEM. Such systems are promising catalysts for electrochemical hydrogen production in acidic solution.

It is commonly believed that weaker interaction of metal with carbon supports cannot provide high metal dispersion because of sintering of small metal particles; but sufficient reducibility [75] of metals on carbon supports makes possible to smooth reduction conditions thus preventing agglomeration proceeding at high temperatures.

If the cause of the MSI is the bonding of metal to the surface through functional groups, it can be tuned by changing the nature of surface functional groups by grafting less-active moieties. In the case of Co-containing catalysts for FT synthesis the surface of SiO₂ can be effectively modified by organosilanes [76], [77]. Instead of suppression of non-desirable surface groups, the tuning of functional coverage composition can be used to change the strength of the MSI. It was demonstrated in [78] that the presence of oxygen surface groups significantly changes the adsorption strength of aromatics and double-bonded compounds on the surface of commercial high surface area graphite, since the removal of electrons from the π-electron system of the basal planes weaken adsorption bonds.

Modification of novel carbon materials (CNT, CNF etc.) is often performed via introducing reactive oxygen- [79] or nitrogen-containing surface functional groups [80]. Oxidation of carbon is an easy way to impart hydrophilic properties to initially hydrophobic surface changing the process of metal supporting, which is often performed via water solutions. Therefore, it will impact on the extent and nature of the MSI, changing the activity, selectivity and stability of final catalysts [81], [82].

Oxidation of CNT with nitric acid provides significant amount of oxygen-containing furctional groups on the carbon surface, which facilitates the production of smaller Co NPs; the increase in the number of oxygen-containing groups from 6.8 to 9.2 at.% transforms predominantly crystalline Co particles to amorphous ones [83]. All these changes influence the activity, stability and selectivity in the FT synthesis. There is the optimal concentration of defects on the CNT surface generated by oxidation observed as a maximum of specific surface area, pore size, defectiveness and oxygen content. It can be achieved at middle oxidation duration and provides maximum stability of Co particles to sintering in the FT reaction conditions [71]. The stability of metal particles is more dependent on the surface geometry of a support rather than on the total amount of functional groups. Residual oxygen groups, left on the surface after calcination and reduction, may also play a role in preventing Co sintering during catalytic reaction but this role is less important than the surface geometry of CNT support.

Modification of MSI in Pd-carbon system by addition of the second metal

In several cases interaction between active metal and carbon support can be modified by the introduction of the second metal-modifier. The catalytic properties of bimetallic catalysts are often much better in comparison with single metal counterparts, and synergistic effects are complex in nature. In some cases, modification occurs due to the creation of a layer of metal oxide on the carbon surface. Thus, in [84] palladium-enriched bimetallic PdFe/C catalysts were more effective in multiphase HDC of polychlorinated organic substrates (1,4-dichlorobenzene, hexachlorobenzene and 2,4,8-tetrachlorodibenzofurane) than pure Pd/C one (Fig. 6). Using magnetic methods, it was demonstrated, that metallic NPs in such catalysts consist of Pd-enriched PdFe alloy, but Fe₂O₃ is also presented on the surface. Based on the catalytic performance, magnetic measurements, and TPR data, four aspects of the influence of iron addition were considered:

Fig. 6:

Time of 30% substrate conversion (HCB=hexachlorobenzene, TCDBF=2,4,8-trichlorodibenzofurane) on Pd/C and Pd-Fe/C with different metal loadings and Fe:Pd ratio under multiphase conditions (50°C, Aliquat 336, isooctane, 5% KOH, H₂). The diagram is plotted using the data from [84].

1. Stabilization of small metal particles and change in the nature of active centers;

2. Pd-enriched PdFe alloy formation, confirmed for 6% Pd₁Fe₁/C by the measurement of the Curie temperature. It can cause the variation in the geometry of active sites and/or in the electronic state of metals, observed in the changes of Pd3d XPS spectra;

3. Mixed sites formation. The presence of Fe₂O₃ in the surface layer of bimetallic catalysts under study, even after reduction by H₂ at 500°C, along with the deteriorated reducibility of Fe³⁺ in oxide-supported catalysts known from the literature [85], give reasons to describe bimetallic catalysts as Pd/Fe_xO_y/C systems rather than PdFe/C one. As a result, spillover of hydrogen and changes in the heat adsorption of both chlorinated substrate and hydrogen become possible;

4. Prevention of Pd poisoning: much stronger oxidation of Fe_xO_y to FeCl₃ than Pd⁰ to Pd²⁺ was found by analysis of Fe 2p and Pd3d lines in XPS spectra of active bimetallic catalyst after prolonged usage in the catalytic test in HDC reaction. Therefore, poisoning effect of HCl on Pd or bimetallic active sites is reduced by the elimination of chlorine as FeCl₃. In deactivated catalyst not only Fe, but also a significant fraction of Pd⁰ is also oxidized to Pd²⁺. In addition, it was demonstrated by TPR that close contact between metals in PdFe/C catalysts leads to the significant changes in reducibility of Fe, which are visible by the strong shift of the second peak in TPR profile towards lower temperatures.

The other base metals (Cu, Ni) also can serve as effective modifiers of Pd/C catalysts for HDC of polychlorinated benzenes [43] and act as it was described above.

It is seen from Table 1, that the efficiency of carbon-supported bimetallic catalysts depends on the nature of the substance being transformed. Thus, HDC of 1,3,5-trichlorobenzene proceeds well on PdNi/C; PdCu/C and PdFe/C with the similar metal loading are not so active as PdNi/C, but much better than Pd/C. In contrast, HCB transformation proceeds better on PdFe/C, whereas PdNi is the least active among bimetallic catalysts. Considering that twice as much HCl is evolved during HDC of hexachlorobenzene in comparison with trichlorobenzene, it seems that Fe is the most efficient in preventing Pd chlorination. Copper addition to Pd was the least efficient. In the literature, the effect of encapsulation of palladium particles with the addition of copper was noted, providing a decrease in the activity of the catalyst [86].

MSI in nanodiamond-supported catalysts

Detonation nanodiamonds are the unique type of carbon support that is of special interest in catalysis. Nanodiamond (ND) material [87], [88], [89] consists of nano-sized particles with diamond core covered with highly disordered carbon shell on which great variety of functional groups are fixed. Thus, FTIR-ATR study demonstrated the presence of O–H groups (major absorption band at 3341 cm⁻¹), cyclic ether or cyclic ester (1064 cm⁻¹, stretching vibrations of C−O−C moiety), C=O in carboxylic (COOH), lactone (COO), or cyclic ketone (−CO−) structures (the band at 1840–1850 cm⁻¹) [89], as well as aliphatic CH₂ and CH₃ groups (their symmetric and asymmetric stretching vibrations are observed as intense peaks at 2946 and 2870 cm⁻¹).

In addition, the surface of ND usually comprises nitrogen as an impurity in the structure of diamond and in the composition of surface functional groups. Total nitrogen content can be as large as 2 wt.% or 2×10⁴ ppm [90]. In addition to functional groups, N-atoms can be connected directly with carbon surface of ND forming P1 paramagnetic centers (carbon atoms substituted by nitrogen). The concentration of P1 centers in detonation ND was estimated in [91] on the base of CW EPR spectrometry. It was found to be (8±4)×10¹⁶ centers/g or 2±1 ppm N, i.e. only a small fraction of nitrogen form such paramagnetic centers.

Several methods providing effective tuning of the ND surface properties are described in the literature [92]. Therefore, there is a possibility to tune the type of metal anchoring on the surface of ND [93]. It can be performed by metal coordination on the functional groups of specific type (stronger MSI) as it was demonstrated for Ni on the base of FTIR and XRD data [93], or directly with carbon shell (milder MSI), changing by this way the catalytic efficiency of ND-supported metals. Surface bonding with nitrogen- and oxygen-containing functional groups is a significant reason of the unique properties of ND-supported active metals (Ni, Pd, Au etc.).

ND-supported metals were tested in several reactions, such as methanol decomposition [93], [94], hydrolysis of ammonia borane [95], electrocatalysis [96], nitrobenzene reduction [97], reforming/dehydration of alcohols [98], hydroamination [99], hydrogenation of double and triple carbon-carbon bonds [100], etc. This support is very promising for metal-containing catalysts for structure-sensitive reactions such as CO oxidation [89], [101] and HDC of chlorobenzenes [66]. As noted above, the latter reaction is sensitive not only to the particle size of the active metal but also to the electronic state of metal [17].

Excellent catalytic properties of Pd/ND catalysts in HDC of chlorobenzenes were demonstrated in [66], [102], [103]. Due to electronic and structure factors low-loaded ND-supported Pd catalysts can be more efficient than their analogs with higher metal content. Thus, XRD and TEM data demonstrate that the average size of Pd NPs decreases from 8–10 to 2 nm with decreasing Pd loading from 5 to 0.5 wt.%; turnover number value of Pd/ND in structure-sensitive multiphase HDC of 1,3,5-TCB simultaneously increases from 200 to 1510 (mole 1,3,5-TCB)/(mole Pd) [103]. The authors of this work consider the electronic interaction of palladium NPs with a support as a reason for such dependence. Like for Pd catalysts on oxide supports, small Pd particles on the very defective ND surface can be stabilized in the electron-deficient state, providing the optimal for HDC Pdⁿ⁺/Pd⁰ ratio in the catalyst.

The results of the IR spectroscopy of CO adsorption on the surface of Pd/ND catalysts, comprising from 0.5 to 2 wt.% of Pd, confirm the presence of palladium in both the metallic and partially oxidized state. Indeed, in addition to absorption bands at 2077 and 1960 cm^-1 corresponding to linearly and bridged adsorbed CO molecules on Pd⁰ atoms and crystals, IR spectra also contain absorption bands at 2172 and 2120 cm⁻¹ related to CO adsorption on Pd²⁺ and Pd⁺, respectively. When Pd loading grows to 5 wt.%, single adsorption band at 2335 cm⁻¹ attributed to weakly disturbed CO₂ becomes dominant in the spectrum. It was speculated in [102], that such CO adsorption spectrum may be due to the oxidation of CO to CO₂ after dissociation on defects of Pd crystals activated by the presence of metal impurities typical for ND. Therefore, the Pdⁿ⁺/Pd⁰ ratio, which is essential for HDC reaction, decreases with the Pd loading decrease in Pd/ND catalysts, providing the changes in catalytic properties. Additional proofs of this fact were found by the XPS analysis.

Vivid contrast was observed in the selectivity of ND and active carbon-supported Pd catalysts, including commercial 5% Pd/C (Fluka). On the latter catalyst the HDC of 1,3,5-TCB proceeds through well-known consecutive scheme, where at 30% 1,3,5-TCB conversion no benzene was found in products (only chlorobenzene and dichlorobenzene in 2:1 ratio). In contrast, benzene was the main product (84%) on 5% Pd/ND at similar conversion. Moreover, in the presence of catalyst with significantly lower Pd loading (0.5% Pd/ND) benzene content in the reaction mixture at 30% TCB conversion was even higher, 88%. This difference in selectivity can be related to the improved adsorption affinity of ND and Pd/ND to chlorinated benzenes in comparison with activated carbon and Pd catalysts, prepared using this support.

Unique ability of ND to the Pd NPs stabilization and electronic modification makes it very promising support for the catalysts efficient in processing of polychlorinated aromatics, which are reckoned among the persistent organic pollutants of the environment. The comparison of catalytic properties of Pd/ND and 5% Pd/C (Fluka) catalysts in HDC of chlorobenzene, 1,3,5-TCB, hexachlorobenzene (HCB) and trichlorodibenzofurane (2,4,8-TCDBF, non-toxic congener of dioxin series) [102] presented in Table 2 clearly demonstrates the advantages of 5% Pd/ND in comparison with commercially available analog produced by Fluka in HDC of all studied molecules. Especially significant difference in activity was found for HDC of chlorobenzene, 1,3,5-TCB and 2,4,8-TCDBF: the time of near complete CB and 2,4,8-TCDBF transformation (t₉₀) decreases two times, for 1,3,5-TCB – more than 16 times. Nearly equal properties of these two catalysts were observed in HCB processing. Even at lower Pd loading (1 wt.%) Pd/ND still is much more active in CB and 1,3,5-TCB transformation, than 5 wt.% Pd/C (Fluka), and only slightly worse in HDC of hexachlorobenzene.

Not so prominent, but still visible advantage of 5% Pd/ND in comparison with 5% Pd/C (Fluka) was observed in vapor-phase HDC of CB in continuous-flow system: the CB conversion at 50°C was 68 and 83% correspondingly, whereas at higher temperatures both catalysts provided nearly 100% conversion of CB. The TPR results demonstrate the easier reduction of Ni and Pd from oxides when supported on ND in comparison with active carbon-supported counterparts [66], [102].

ND surface not only provides several oxidation states of Pd; as the structured support, it facilitates also the spillover of hydrogen activated on Pd⁰ centers to the locations of [Pdⁿ⁺–Cl] and R⁺ moieties. The presence of functional groups on the surface of ND may contribute to the activation of chlorinated substrate on the electron-reach areas of the ND surface. High crystallinity is typical even for small Pd particles supported on ND [103]. According to the widespread opinion [104], [105] the crystalline state promotes the strong MSI and provides significant difference between palladium catalysts prepared using ND and active carbon.

Even more bright confirmation of the influence of the method of metal NPs anchoring to ND surface was found for Au/ND systems [89]. If Au was supported by deposition-precipitation (DP), providing direct contact of gold nanoparticles precursor with ND surface, final catalysts were much more active in CO oxidation than their counterparts prepared by deposition of Au particles stabilized in colloid dispersion using sodium citrate as reducing agent and stabilizer (Turkevich method) [106]. Due to direct attachment of Pd to the surface of ND and smaller size of Au particles (less than 10 nm vs. 13 nm in comparative sample) CO conversion was much higher on 2 wt.% Au/ND (DP) [65% at 400°C vs. 30% for 2 wt.% Au/ND (Turkevich)]. The decrease of Au content results in the growth of CO conversion at 250–300°C [39% on 0.05 wt.% Au/ND (DP) vs. 13% on 2 wt.% Au/ND (DP) at 300°C], because of its better surface distribution and interaction with surface functional groups.

Detailed investigation of the ways of NiO anchoring on the surface of detonation ND was performed in [107]. Two-step reduction of NiO/ND during TPR with the maxima at 260 and 390°C was explained by the presence of two types of NiO, strongly and weakly bounded with ND surface (Fig. 7). According to authors’ opinion, weakly bonded Ni species are attached to the ND surface by van der Waals forces, but strongly bonded ones are chemically anchored through the functional groups on the ND surface. To clean the surface from functional groups as much as possible, ND was annealed in Ar at 900°C. The changes in ND activity and selectivity in methanol decomposition after similar treatment were early noticed in [108]. Indeed, in the TPR profile of NiO supported on this sample the peak of weakly bonded Ni becomes the predominant one. In contrast, calcination of ND in air at 320°C provides an increase in the number of oxygen-containing functional groups on the ND surface. Significant increase in the intensity of the TPR peak corresponding to the reduction of NiO strongly bonded with the ND surface was observed in the TPR profile of NiO supported on such-treated ND. The changes in the composition of functional groups adsorbed on ND particles during these two types of treatment were additionally confirmed by the variation in zeta-potential, which changes from −4.3 mV for ND to −17.5 mV after treatment in air at 320°C for 3 h, and to 23.8 mV after Ar treatment at 900°C for 2 h. Moreover, the attachment of Ni to ND surface through oxygen-containing functional groups finds its proofs in disappearance of the low-intensity absorption bands at 1192 and 1257 cm⁻¹ in FTIR spectrum of ND after Ni supporting.

Fig. 7:

TPR profiles of ND and catalyst precursors, prepared by nickel supporting on pristine ND (NiO/ND), after calcination at 320°C (NiO/ND_320) and after annealing in Ar at 900°C (NiO/ND_Ar) [107].

The difference in types of Ni bonding with ND and the features of the structure of the catalysts, in which only one bonding type is predominant, were investigated by the in situ EXAFS combined with TPR. The comparison of wavelet transform (WT) of EXAFS signal for the catalysts under study and for model substances (Ni, NiO, Ni(CH₃COO)₂, and NiCO₃) was used to reveal the presence of Ni–O–C bonds typical for Ni species strongly anchored to the ND surface. The Ni salts were chosen to simulate the coordination of the metal precursor by carboxyl groups. In this study the tested samples were heated to a desired temperature (150, 300 or 900°C) in H₂ in the EXAFS cell and then equilibrated at the chosen temperature during several hours; in these conditions the reduction of each form of NiO is achieved at lower temperatures than at continuous temperature ramping during TPR data acquisition. In Ni/ND reduced at 150°C Ni²⁺ is mainly present, as it is evident from the intense peak at R=1.56 Å (R is the distance from the central atom to first coordination sphere) demonstrating Ni–O as the predominant path of X-ray scattering. These data confirm strong bonding of Ni with ND through the oxygen-containing surface groups. The reduction of nickel species strongly bonded with the ND surface proceeds above 150°C. Indeed, the reduction at 300°C provides an increase in the intensity of the peak at R=2.3 Å, k≈6.4 (k is the photoelectron wavenumber), characteristic for Ni-Ni scattering path, because of NiO reduction to Ni⁰. Only the treatment at 900°C provides the complete NiO to Ni⁰ reduction, as WT for such reduced Ni/ND sample is identical to that of metallic Ni and comprises two peaks at R=2.03 and 4.15 Å, characterizing Ni–Ni scattering paths in the first and second coordination spheres. Nickel carbide formation proceeds during reduction at 300°C as it is visible from a new peak appearing at R=2.1 Å (k≈3.8) and associated with the Ni–C scattering path. The possible mechanism of Ni₃C formation includes thermolysis of Ni complexes with oxygenated surface groups and subsequent carbidization of Ni⁰ formed in this process by CO released during thermolysis, as it was described in [109] for thermal decomposition of nickel acetylacetonate in oleylamine under inert atmosphere.

A removal of weakly bonded Ni species provides an increase of selectivity in phenylacetylene hydrogenation to form styrene, whereas the use of sample containing predominantly weakly bonded Ni leads to over-hydrogenation to produce ethylbenzene. These results open the way of tuning the catalytic properties by changing the degree of the MSI.

Strong anchoring of metal precursor through the surface functional groups needs to be mentioned among the reasons of high stability of ND supported metals to the sintering, which was underlined in several publications [110], [111].

In some articles it was noted that the presence of metal impurities characteristic for detonation ND can contribute to unusual catalytic properties. Thus in [66], [103] Fe and other metal minor components were found in ND by XPS. Considering significant influence of Fe on the activity of Pd/C catalysts, described above, it can be expected that metal impurities play an important role in the catalytic performance of ND-supported catalysts.

MSI in metal core – carbon shell composites

In common catalysts metal particles are located on the surface of a support. However, in search for the methods of stabilization of the finely dispersed metal state for the use in industry and medicine several composite materials were created, comprising carbon and metal NPs. They attract significant attention for the use in chemical, electro- and photocatalysis [112]. In particular, such materials can be produced by evaporation of metal drop hanged in a magnetic field in a flow of an inert gas containing light hydrocarbon [113]. It was demonstrated by HRTEM (Fig. 8), XRD and other methods that the particles formed by this technique from Ni and Fe (Ni@C and Fe@C) have a core-shell structure with predominantly zero-valent metal single-crystal core less than 10 nm in diameter, and the shell consisting of several defected graphene layers, sometimes with an admixture of amorphous carbon. Similar structure can be produced from 5% Ni 95% Pd alloy, but not from Pd, on which carbon shell is not formed. Carbon coated metal particles have fairly uniform particle size distribution, and they are very stable to sintering and oxidation during years-long storage in air. It was found, that such ‘reverse’ systems demonstrate catalytic properties in hydrogen dissociation [113], [114], hydrodechlorination of chlorobenzene [115], [116] and hydrogenation of phenylacetylene [117], [118].

Fig. 8:

High resolution TEM images of Ni@C (a) and Fe@C (b) composites.

In special experiments it was demonstrated that carbon shell is very stable during catalytic tests in fixed-bed continuous-flow chlorobenzene HDC: no signs of its decomposition or disappearance were found by TEM after three cycles of increase and decrease of the reaction temperature from 100 to 350°C for Ni@C [115]. Note that the catalytic activity of Ni@C does not decrease during such prolonged catalytic test.

It is impossible to exclude the presence of a limited number of metal particles not covered with the carbon shell, but the acid treatment (with HCl or HF) aimed at removing of uncovered particles did not lead to the disappearance of catalytic properties. On the base of literature data [119], [120], DFT calculations, and experimental results it was demonstrated that the catalytic action in the reactions above listed, associated with the catalytic activation of hydrogen, is the feature of the graphene shell, activated by the presence of defects (e.g. of Stone-Wales type), and the presence of a transition metal in the subsurface layer [114], [118]. Therefore, such systems can be considered as providing the ‘reverse’ MSI where transition metal is a support, and defected graphene plays a role of active component, that is able to form temporary chemical bond with hydrogen atoms.

The features of graphene layers interaction with different metals determine the catalytic properties. Calculation of H₂ chemisorption energy on graphene layers activated by Ni or Fe helped to understand the difference in their catalytic action in phenylacetylene hydrogenation [118]. Thus, according to DFT calculation, chemisorption of hydrogen on graphene defects activated with the presence of iron will be exothermic and will cause irreversible hydrogenation of imperfect parts of the carbon shell. Therefore, adsorbed hydrogen will be active in catalytic hydrogenation only at the temperatures above 200°C in qualitative agreement with experiment, where phenylacetylene conversion on Fe@C grows from 2 to 75% between 200 and 250°C [118]. More defected graphene areas become active at higher reaction temperature providing high styrene selectivity between 250 and 300°C. In contrast, nickel in subsurface layer provides hydrogenation of large areas around defects with smooth increasing of chemisorption energy from 0.4 to 0.9 eV/H₂. According to the calculations, only selective to styrene formation defected graphene areas are catalytically active on Ni@C at low temperature. Non-defective graphene areas, providing deep hydrogenation to form ethylbenzene, become involved in catalytic action only at high reaction temperatures, again with exact agreement with the results of the catalytic test. Indeed, styrene selectivity is about 59–70% at 50–150°C but falls to 5% in the temperature range of 150 to 200°C [118].

It is known that many types of carbon (the most known example is active carbon) can be prepared on the base of pyrolyzed biomass, and the choice of biomass source influences the structure, morphology and other properties of carbon produced. Interesting branch of this topic is the preparation of metal-carbon composites by simultaneous formation of carbonaceous component from biomass and metal component from the other sources – from metal salts, MOF structures etc. [121], [122], [123]. In our recent work [121], by pyrolysis in inert atmosphere of wood sawdust impregnated with Pd salt a catalyst was obtained, unexpectedly comprising uniformly sized palladium NPs in the reduced state, that was confirmed by TEM and XPS. Metallic state of Pd NPs is explained by the reduction in pyrolysis conditions (the reducing atmosphere results from the presence of pyrolysis gases and the temperature of 400°C, appropriate for Pd reduction from its salts). However, due to relatively low S_BET (from 2 to 350 m²/g, depending on the preliminary sawdust treatment) [124], [125] a significant portion of the palladium particles are located in the bulk of the carbon material or coated with the carbon shell, that make them somewhat similar to Me@C composites described above. Therefore Pd/C (SD) composites (where SD denotes sawdust as a precursor of carbon support) include at least two types of Pd nanoparticles, strongly (if they have carbon shell or they are immersed in the bulk of a carbon support) or weakly (bare particles located on the outer surface of material) interacting with a carbon support. Catalytic properties were tested in HDC of chlorobenzene [121], [124], [125]. It seems that carbon-coated Pd particles are very efficient in vapor-phase HDC of chlorobenzene, but only those weakly interacting with carbon surface provide liquid-phase HDC of hexachlorobenzene. Note that the 0.6 wt.% Pd/C (SD) catalyst is very efficient in processing of hexachlorobenzene: the benzene selectivity exceeds 80%, and TOF value calculated for reaction time of 5 h is 58.8 h⁻¹, that is about four times higher than the value 14.4 h⁻¹ found for 0.39 wt.% Pd/Sibunit catalyst in [126]. Among the reasons of good catalytic performance of such composites the presence of ash impurities of alkali and alkaline earth metals inherited by the carbon material from the precursor (wood sawdust) can play an important role. It was confirmed by the fact that the removal of such impurities by preliminary acid treatment of sawdust leads to a decrease in the catalytic activity [121].

Recently the catalytic efficiency of three types of Co-carbon nanocomposites, namely Co/C (SD), Co@C and Co/CNT (where CNT denotes carbon nanotubes) prepared using Co(NO₃)₂ as a metal precursor has been compared in CB HDC [122], [127]. The characterization of these materials by TPR, XPS and magnetometry demonstrated that the use of different supports and preparation techniques affected both the oxidation state of cobalt and the structure of carbon-cobalt catalysts. The metal oxidation state varied from Co⁰ in Co@C to Co²⁺ (CoO) in Co/C and Co²⁺/Co³⁺ (Co₃O₄) in Co/CNT. In Co@C Co NPs were encapsulated by the defect carbon layer, the same way as it was discussed above for Fe@C and Ni@C [115], [116]; in Co/C (SD) CoO species were predominantly immersed in the carbon matrix, as Pd NPs in Pd/C (SD) in [121], and in Co/CNT Co₃O₄ particles were found by TEM both inside CNT channels and on their surface, mostly near defects. Vapor-phase CB conversion to benzene proceeded most efficiently on the Co/C composite with the lowest Co content (1.3 wt.%). This observation allowed authors to conclude that CoO could be active in HDC. However, the possibility of reduction of a small fraction of CoO to Co in the reaction medium at elevated reaction temperatures cannot be excluded. Co/CNT comprising 14.6 wt.% of Co and Co@C with 79.1 wt.% of Co were less active, which once again confirms the key role of MSI in carbon supported catalysts.