**About the Editors**

**Antonio Guerrero Ruiz** is Full Professor at the Department of Inorganic and Technical Chemistry of the UNED in Madrid, where he heads the Laboratory of Chemistry at Surfaces. His research activities are concentrated in the development of new heterogeneous catalyst materials. For this aim, the preparation and characterization of metallic or bimetallic nanoparticles, functionalized carbon nanotubes, modified graphene composites, etc., has been accomplished. These materials are applied as catalysts or adsorbents for different technical processes.

**Inmaculada Rodr´ıguez-Ramos** is Research Professor at the Instituto de Catalisis y ´ Petroleoqu´ımica (ICP-CSIC) in Madrid, where she heads the Group for Molecular Design of Heterogeneous Catalysts. Her expertise is Heterogeneous Catalysis, with a particular specialization in carbon materials, their functionalization chemistry, and applications in catalysis. Her research interests involve the design and preparation of (nano)materials for their application in gas/liquid catalytic reactions related to the sustainable production of primary chemicals and energy products.

## *Editorial* **Application of New Nanoparticle Structures as Catalysts**

#### **Antonio Guerrero Ruiz 1,\* and Inmaculada Rodríguez-Ramos 2,\***


Received: 3 August 2020; Accepted: 3 August 2020; Published: 27 August 2020

Nanocatalysts, more precisely solids nanomaterials with catalytic properties to be used as heterogeneous catalysts, are an extended and very diverse group of nanostructured materials representing, at present, an active area of research with application in many catalyzed processes. Therefore, this research area not only can lead to significant advances for their potential technological applications, but also must engage a large variety of new materials. For the characterization of the studied nanocatalysts, many different techniques and experimental methods should be used to reveal both the structural and superficial properties of these materials. The scope of the Special Issue "Application of New Nanoparticle Structures as Catalysts" was to provide a non-systematic overview with current research studies in the field of developing nanocatalyts. Practically all the nanomaterials compositions presented as examples in this issue are different from each other, both in their chemical composition of the structured nanomaterials, or in relation to the catalyzed reactions where they are applied. In this way, in this Special Issue, nine selected original research paper and one comprehensive review are collected. More than 40 scientists from universities and research institutions contributed their research studies and expertise for the success of this Special Issue.

The scientific contributions are summarized in the next paragraphs.

An analysis of the recent studies concerning transition metal nitrides applied as heterogeneous catalysts is presented in the review carried out by Dr. Dongil [1]. These materials have a clear interest because they can substitute noble metals in different catalyzed processes. With these nanomaterials, numerous possibilities are opened for new structures for metal nitrides; since chemical ingredients can be combined as mono-, binary and even ternary mixtures or the addition of promoters can be accomplished during the preparation procedures. The description of the most employed synthetic methods is revisited and the application of some transition metal nitrides in different catalyzed reactions, hydrotreatments, oxidations and ammonia synthesis/decomposition is reported.

Two of the contributions of this Special Issue are related to Metal Organic Framework (MOF) nanostructures. The development of these MOF nanomaterials is, at present, one of the major subjects in fundamental research due to, among others, their potential applications as catalytic materials or as selective adsorbents. In the contribution by Zamaro et al. [2], the preparation of nanocatalysts derived from the MOF named UiO-66, when used as support for three transition elements (Cu, Co, and Fe), is described. These materials are evaluated in two CO oxidations: oxidation with air and selective oxidation in a hydrogen-rich stream. The main aim of this research is to find reaction conditions where these new nanostructures can be, for these processes, an alternative to the commercial catalysts based on expensive noble metals.

In the second contribution regarding MOFs by Ordonez et al. [3], a very important aspect when nanomaterials are proposed for a real application is emphasized. Thus, the solid material should be submitted to physic-mechanical treatments, such as grinding or pelleting, in order to transform the original powder into granules, which can be used at industrial scale. Three commercial materials ([Cu3(C9H3O6)2], [C9H3FeO6] and [C8H5AlO5]) were studied as methane adsorbents in a fixed bed reactor. It was concluded that all these materials suffer structural and textural modifications when subjected to pressure, and consequently their adsorption capacities are largely reduced.

Two other research papers in this issue are related with mixed transition metal oxides. In the contribution by Illán-Gómez et al. [4], the authors synthesized, characterized and tested a series of perovskites (BaFe1−xCuxO3 with x = 0, 0.1, 0.3 and 0.4). The target reaction for these nanocatalysts is the soot oxidation, as method for avoiding the atmospheric contamination by exhaust gases of car engines. These perovskites catalyze both the NO2 oxidation of Diesel soot and, but to a lesser extent, the soot oxidation by O2 of Gasoline engines. The catalytic activities of these perovskites seem to be related to the amount of oxygen evolved during temperature programmed desorption experiments, which decreases when increasing the copper content.

In the case of the contribution by Cauqui et al. [5], a mixed oxide (ZrO2 with different loadings of Ce, Ca and Y) is used as a support of Ni nanoparticles. In this study, the synthesized nanomaterials are extensively characterized by complementary methods and techniques, and evaluated as catalysts for the aqueous-phase reforming of methanol. Focusing on the effect of the redox properties of ceria and the basicity properties induced by Ca or Y, it is revealed that the availability of Ni-metallic at the surfaces and the presence of weak basic sites, particularly derived from Ca incorporation, is the key parameter for improving the catalytic performance.

Y. Wang et al. reported on the use of the surface plasmon resonance effect, in this case in a metal nanocomposite, AuPt/N–TiO2, used as a photocatalyst [6]. While the Au nanoparticles were used to obtain energy from visible-light, Pt nanoparticles work as a cocatalyst, trapping the energetic electrons from the semiconductor support. With this material, the selective oxidation of benzyl alcohol under visible-light irradiation can be performed with a markedly enhanced selectivity and yield. An extensive series of irradiation experiments shed light on relevant information concerning the different steps of the photocatalytic mechanism with this material.

Fernández-Morales et al. reported a comparative study of diverse materials, in general solid catalysts with acidic surface properties, when applied to the reaction of isobutene dimerization to C8 olefins [7]. The exposed surface catalytic sites were conveniently characterized in order to interpret catalytic performances. In general, catalytic materials with a higher amount of Brønsted acid sites display improved catalytic performance, but for achieving an optimum selectivity towards C8 compounds, a combination of the nature of acidic sites and structural characteristics of the catalytic materials is required.

In the contribution by Ramirez-Barria et al. [8], an extended series of graphenic materials (doped or not with nitrogen adatoms, with different textural properties, etc.) were prepared and applied as electrocatalysts for the demanding oxygen reduction reaction. The material with nitrogen doping and with smaller grain sizes was demonstrated to be the most efficient electrocatalyst. Moreover, all nitrogen-doped graphenic materials show high tolerance to methanol poisoning and good stability.

The approach of Faroldi et al. [9] for the study of a new heterogeneous catalysts for the dehydrogenation reaction of formic acid, generating high-purity hydrogen, is also of great interest. Instead of noble metal catalysts, they prepared and characterized Ni-based catalysts supported on silica, which were doped with calcium in order to facilitate the adsorption-decomposition of the reactant. From the results of the catalytic performance (100% conversion with a 92% of selectivity to hydrogen) at a moderate reaction temperature, 160 ◦C, it can be concluded that these materials were very promising for this application. In fact, these results for catalytic behavior are comparable to those reported for noble metals.

Another example of a complex catalytic multicomponent material with an interesting potential application is reported by Ivars-Barcelo et al. [10]. In this case, the composite materials are based on noble metal particles (Pd or bi-metallic Ag/Pd) supported over an iron oxide (Fe3O4) with a magnetite structure. The catalytic application of these materials is the direct methane partial oxidation into value-added chemicals as formaldehyde. The presented preliminary catalytic results confirmed the potential of magnetite-supported (Ag)Pd catalysts for CH4 partial oxidation into formaldehyde, with incipient methane conversion starting at 200 ◦C, but with very high selectivity above 95%. The prepared nanocomposite

materials were investigated by different physicochemical techniques, with the purpose of relating the structural and superficial properties of these nanocatalysts with their detected catalytic performances.

In conclusion, the papers collected in this Special Issue can be described as an impressionistic painting with brushstrokes of different aspects of new developments of catalytic materials. All of them include complementary features involved in the design of special nanocatalysts: preparation-treatments, intensive characterization and evaluation as catalysts in various reactions of applied interest. Although the present Special Issue can cover neither all the research of new structures used as nanocatalysts nor a complete list of application in catalyzed processes, the editors are confident that its contributions to fundamental research will offer new perspectives for the readers.

**Author Contributions:** All the guest editors wrote and reviewed this Editorial Letter. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded partially by the SpanishMinister of Science through the projects CTQ2017-89443- C3-1-R and CTQ2017-89443-C3-3-R.

**Acknowledgments:** We are grateful to all the authors who contributed to this Special Issue. We also acknowledge the referees for reviewing the manuscripts. And especially we have to recognize the immense work developed by Miss Tina Tian, without her support the publication of this Special Issue was impossible.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Recent Progress on Transition Metal Nitrides Nanoparticles as Heterogeneous Catalysts**

#### **A.B. Dongil 1,2**


Received: 15 July 2019; Accepted: 23 July 2019; Published: 2 August 2019

**Abstract:** This short review aims at providing an overview of the most recent literature regarding transition metal nitrides (TMN) applied in heterogeneous catalysis. These materials have received renewed attention in the last decade due to its potential to substitute noble metals mainly in biomass and energy transformations, the decomposition of ammonia being one of the most studied reactions. The reactions considered in this review are limited to thermal catalysis. However the potential of these materials spreads to other key applications as photo- and electrocatalysis in hydrogen and oxygen evolution reactions. Mono, binary and exceptionally ternary metal nitrides have been synthetized and evaluated as catalysts and, in some cases, promoters are added to the structure in an attempt to improve their catalytic performance. The objective of the latest research is finding new synthesis methods that allow to obtain smaller metal nanoparticles and increase the surface area to improve their activity, selectivity and stability under reaction conditions. After a brief introduction and description of the most employed synthetic methods, the review has been divided in the application of transition metal nitrides in the following reactions: hydrotreatment, oxidation and ammonia synthesis and decomposition.

**Keywords:** heterogeneous catalysis; transition metal nitrides

#### **1. Introduction**

In the last decade strong interest has emerged in the field of nitrogen-doped catalysts, especially since new carbon nanostructures have been successfully synthetized. In those structures nitrogen can be easily inserted as heteroatoms due to its similarity to carbon, and as a doping agent it offers new possibilities in the field of catalysis [1]. Nitrogen confers basic sites for optimal reactant adsorption and an excellent electron density that may improve the catalytic performance due to the better dispersion of the active phase and the changes in the electronic properties [2].

Transition metal nitrides have been traditionally employed as catalysts in reactions such as hydrodesulfurization or ammonia synthesis [3]. Nowadays, researchers face new challenges in catalytic transformations to find active and selective catalysts in the fields of energy or biomass including reactions such as hydrodeoxygenation, water-gas shift and CO and CO2 hydrogenation among others [4].

Noble metals have been widely studied on those reactions. However economical catalysts based on abundant materials are more suitable for industrial applications. Hence, the need of searching new catalyst able to replace noble metals and the positive effect of nitrogen doping in several transformations, have brought up renewable attention to transition metal nitrides. These materials can donate electrons from the nitrogen atom and possesses high chemical, mechanical and thermal stability. Also importantly for thermal catalytic applications, when optimal synthesis conditions are employed, nitrides can reach relatively high surface areas of at least 90 m2/g [5].

Transition metal nitrides are compounds in which nitrogen is incorporated into the interstitial sites of the metal structure. Since the size of the nitrogen atom is small (0.065 nm), it fulfills the Hägg rule (i.e., the ratio of the radii of non-metal to metal is less than 0.59, this allowing the formation of simple typical structures as compiled in Figure 1 [6].

As it can be expected the reactivity of the resulting transition metal nitride is different from that of the parent metal. In general, two main effects can be considered when nitrides are used in catalysis [7,8]:


#### **2. Synthesis and Structural Properties**

Metal nitrides can be prepared among other methods by thermal treatment of the metal precursor obtained from template methods [9], ball milling [10,11] or temperature programmed reaction/reduction procedures [12].

Template methods are based on the use of a sacrificial template, normally MgO, to which a mixture containing the metal precursor is added followed by thermal treatment at high temperatures and extensive wash to remove the MgO template. In this way Metal-N entities are well dispersed in a carbon based matrix. In the ball milling method, the solid metal precursor either commercial or lab-made are submitted to milling under pressure and connected to a supply of the nitridation agent, N2 or NH3.

So far, the most employed method to synthetize transition metal nitrides is a temperature programmed reaction/reduction procedure. The method consists in a reductive nitriding treatment under a gas flow (i.e., NH3 or a mixture of N2:H2) of the corresponding metal oxide precursor to produce the nitride [12]. The characteristics of the final nitride depend on the precursor and preparation conditions such as heating rate and final temperature. It was reported that higher areas are obtained when heating rates below 1 ◦C/min and/or high gas space velocity (150,000 h−1) are used, as these conditions reduce sintering due to water release [13,14]. Other nitrogen sources such as urea or *m*-phenylenediamine have been employed in order to improve the efficiency of the nitridation [15,16].

The use of nitrogen-rich transition metal nitrides in catalysis is very promising. However, the synthesis of TMN with high nitrogen to metal ratio is energetically demanding. This has prompted researchers to find aternative methodologies and new structures such as polymorphs of Zr3N4, Hf3N4, Ta2N3, and noble metal dinitrides OsN2, IrN2, and PtN2 have been explored [17].

**Figure 1.** Value of the nitrogen chemical potential (μN) where the nitrogen covered surface becomes more stable than the clean metal surface; Reproduced from [18], with permission from American Chemical Society, 2018.

Sautet et al. [18] recently studied the nitridation of transition metal surfaces and tried to offer some insight into the synthesis and stability of transition metal nitrides under working conditions. The authors studied fifteen transition metals and using theoretical and experimental techniques evaluated the extent of nitridation (surface vs. bulk) depending on the metal and the shape of nanoparticles. By using the nitrogen chemical potential at which metal covered by nitrogen is more stable than bare metal, the authors were able to stablish a nitridation trend among the studied serie. As depicted in Figure 1, Mo, W, Fe and Re are more easily nitrided, becomign more difficult when moving to the right hand side of the periodic table, i.e., less oxophilic metals, which agrees well with previous experimental results [19,20].

The most studied transition metal nitride in catalysis is molybdenum. In general cubic γ-MoNx (0.5 ≤ x < 1) is obtained by NH3 treatment of the oxide precursor, MoO3, while mixtures of H2 + N2 lead to the formation of tetragonal β-MoNx (x ≤ 0.5) or γ-MoNx. Hexagonal δ-MoNx (x ≥ 1) is obtained from MoS2 and NH3. Similarly, the content of nitrogen on iron nitrides influences their structure, so upon increasing the nitrogen content, the lattice structure changes from fcc γ'-Fe4N to hcp ε-FexN (2 < x ≤ 3) and to orthorhombic ζ-Fe2N. In Figure 2, the most common structures of transition metal nitrides are shown.

**Figure 2.** (**a**) bcc: TiN, ZrN, HfN, VN, CrN; (**b**) hcp: Mo2N, W2N; (**c**) fcc: MoN, TaN. Blue points represent transition metal atoms and brown points nitrogen atoms. Adapted from [6], with permission from Wiley, 2013.

Body centered cubic (bcc) Hexagonal close packed (hcp) Face centered cubic (fcc)

#### **3. Transition Metal Nitrides as Catalyst**

#### *3.1. Hydrotreatment Reactions*

Transition metal nitrides are considered excellent candidates to replace noble metals in hydrogen-treatment reactions since they show similar or even better performance than noble metals. It has been reported that Mo nitrides can easily chemisorb hydrogen due to the contraction of the d-band and the changes in the electron density that result as a consequence of the interstitial incorporation of N in the Mo metal lattice. Moreover, in the Mo2N-based catalysts hydrogenation reaction occurs over nitrogen vacancies, so Mo/N ratio also influences the catalytic performance [21].

Both CO and CO2 hydrogenations are very interesting reactions since they overcome key environmental challenges while providing energy or valuable chemicals. On the one hand CO hydrogenation can be employed to purify H2 gas streams before feeding to fuel cells to avoid poisoning. On the other hand, CO2 hydrogenation has been appointed as a solution to reduce the amount of CO2 evolved to the atmosphere from the industrial activity and convert it to fuels and chemicals [22,23].

However CO and CO2 hydrogenation face several challenges. CO2 activation is difficult due to the inert nature of the molecule and the cleavage of the C−O bonds in CO2 demands high activation energy. Also, methanation of CO and CO2 which provides an efficient alternative to conventional natural gas, is highly exotermic. The produced reaction heat favours metal sintering, which decreases the catalyst activity. Catalyst deactivation can also take place when carbon deposits on the active phase.

Zaman et al. [24,25] have studied the influence of adding alkali promoters to the Mo2N systems in CO hydrogenation. The synthesis in both cases was performed by a simple temperature programmed treatment of the molybdenum and alkali precursors under ammonia flow. This leads to a material containing a mix of different phases: Mo, Mo2N, Mo oxide and alkali-Mo oxide phase.

The promotion with alkalis favoured the conversion to oxygenates, i.e., methanol, ethanol and propanol. Cesium was found to be better promoter to oxygenates at 5% wt. (28% selectivity to oxygenates vs. 11% with Li promoted and 6.5% with unpromoted). The lower selectivity achieved with Li compared to Cs was attributed to the formation of Li2MoO4 phases during nitridation.

On the other hand, bare Mo2N leads to the preferential conversion of CO to hydrocarbons, which can be ascribed to (i) CO dissociative hydrogenation and (ii) water-gas shift reaction, as shown in Figure 3. However, the presence of alkali hinders CO dissociation, which benefits the molecular insertion of CO into—CHx intermediate and promotes the coupling of alcohols.

**Figure 3.** Plausible CO hydrogenation reaction pathways on Mo2N and K-Mo2N catalysts. Reproduced from [25] with permission from Elsevier, 2018.

Regarding the effect of alkali loading, the authors performed a thorough study with potassium as promoter using several weight percentages: 0.45, 1.3, 3 and 6.2% [24]. The best selectivity to oxygenates, 44%, was obtained over promoted K-Mo2N with a K/Mo surface ratio of 0.06 which also corresponded to the best K distribution among the samples. The XRD showed that both γ-Mo2N cubic and monoclinic K2MoO4 were formed and this latter phase seems to increase upon K addition further than 3% wt, this being detrimental for K distribution and hence for oxygenates conversion.

Methanation, is the catalytic hydrogenation of carbon oxides (CO and CO2) to obtain synthetic natural gas. Ru, Rh and Ni catalysts have proven to offer good catalytic performance in terms of activity and selectivity to methane which can be further improved by using bimetallic systems and optimizing catalyst synthesis method [26].

Methanation and hydrodesulfurization of dibenzothiophene was studied over molybdenum nitride by Zhao et al. [27]. The authors were able to synthetize a rich nitrogen molybdenum nitride using high pressure, 3.5 GPa, through a solid-state ion exchange reaction. This new nitride, 3R−MoN2 holds a rhombohedral R3m structure, isotypic with MoS2. However it offered catalytic activities three times higher than MoS2 for the hydrodesulfurization of dibenzothiophene and over twice as high in the sour methanation of syngas at 723 K.

The binary nitride, Ni2Mo3N, was studied by Leybo et al. [28] in the methanation of CO2. Yet, modest selectivites to CH4, ca. 20%, were obtained and the active phase suffered sintering upon reaction conditions, decreasing the stability of the catalyst.

Besides molybdenum other metal nitrides have been tested in methanation reaction. For example co-methanation of CO and CO2 have been evaluated by Li et al. [29] over cobalt nitrides supported on alumina. The authors studied the effect of metal loading on Co4N/γ-Al2O3 and Co/γ-Al2O3 catalysts. According to the characterization, the cobalt nitride favoured stronger interactions with the support, this improving the dispersion of the nanoparticles and also their resistance to coking and metal sintering after 250 h which, as previously mentioned, is critical in such a exotermic reaction. Moreover, it was confirmed that the nitrogen atoms improved the adsorption of reactants due to their basicity, leading to better catalytic performance compared to the Co metal supported catalysts. The better results were also explained by the uniform metal dispersion and superior metal-support interaction.

Fisher-Tropsh is a well-known transformation to convert syngas, CO + H2, into liquid hydrocarbons and that was very relevant in catalytic research in the 70–80s due to the oil crisis. Now, it has revised attention since it can also use syngas from biomass to produce fuels and chemicals. The most studied systems are those based on iron as they offer optimal results under a variety of conditions that allow to tune the selectivity, and are economically interesting. However, under reactions conditions the water produced can oxidize the catalyst with its subsequent deactivation [30].

Bao et al. [31] studied the confination effect of FeN cubic nanoparticles inside carbon nanotubes (FexN-in) (see Figure 4). Firstly, FeN supported on CNT resulted 5–7 times more active than FeN supported on silica and metallic Fe on CNT. This seems to be related to the better stability of the nitrides under CO hydrogenation conditions compared to metallic and carbide iron which are oxidized by water, resulting in catalyst deactivation. Also, the catalyst where FeN nanoparticles were selectively loaded inside the CNT was more active than the catalyst with FeN nanoparticles mainly dipersed on the external walls, FexN-out, (1.4 times). The authors explained the better activity of the confined nanoparticles by the lower particle size and the formation of more FeCxN1x entities on FexN-in than on FexN-out during reaction, leading to stronger retention of nitrogen atoms in the lattice.

**Figure 4.** Reproduced from [32], with permission from Royal Society of Chemistry, 2011.

In contrast to the activity and stability enhancement of iron nitride compared to its parent metal, cobalt nitride seems to be a poison for FT synthesis. In literature it was reported that by adding a nitrogen source such as acetonitrile or ammonia to the FT gas feed, cobalt nitride phases are formed which result in catalyst deactivation by deposition on the most active metallic cobalt sites (steps and edges) [32].

So far, theoretical results have predicted that CO adsorption and dissociation over γ-Mo2N(111) has a similar activation barrier to that of MoS2, so similar activity for syngas conversion can be expected [33].

CO2 dry reforming of methane (DRM) has received much attention, as it transforms two greenhouse gases (CH4 and CO2) into syngas. The most studied catalysts have been noble metals, Ru, Rh and non-noble metals such as Ni [34]. Some reports appeared using transition metal carbides due to their low cost and similar structure compared to noble metals. However at atmospheric pressure, carbides can easily be deactivated due to oxidation by CO2 or H2O [35]. Hence, alternatively nitrides have been tested as potential catalysts for DRM.

Gu et al. [36] studied Mo2N, Ni3Mo3N and Co3Mo3N above 550 ◦C and atmospheric pressures and found that a synergic effect is observed in the bimetallic nitrides that improve the activity and resistance to oxidation and coke deposition on the DRM compared to the monometallic nitride Mo2N. The most active and stable catalyst was Co3Mo3N which, among other factors, was ascribed to the synergistic effect between the Mo and interstitial metal Co.

Owing to environmental concerns and the depletion of fossil resources, in the last decade researchers have focus on the study of biomass derived compounds to obtain fuels and chemicals. Since the starting lignocellulosic biomass owns a high oxygen concentration (>50%), most of the transformations require selective removal of oxygen. More specifically, one of the most studied reactions is hydrodeoxygenation (HDO). Under HDO condititions, the reactants can be also converted through the decarboxylation/decarbonylation (DCO) path, promoting the C−C cleaveage which is undesirable for fuels and chemicals.

For example, Monnier et al. [37] studied the conversion of oleic acid and canola oil with nitrides of Mo, W, and V supported on γ-Al2O3. The Mo2N catalyst exhibited superior activity for oleic acid conversion compared to the other nitride catalysts, and also favored the HDO route vs. DCO. The HDO path produces preferentially n-C18H38 (diesel fuel cetane enhancers). Also, Mo2N/γ-Al2O3 was stable in continuous hydrotreatment of canola oil at 400 ◦C under 83 bar hydrogen, reaching a constant yield of 50% middle distillates.

Murzin et al. [38] studied Ni and Mo2N-MoO2 on the HDO of more complex reactants, *Chlorella* algal oil extracted with supercritical hexane and stearic acid, at 300 ◦C under 30 bar in the presence of hydrogen. The catalysts were selective to fatty acids, indicating deactivation of decarbonylation sites. The catalyst Mo2N-MoO2, despite being less active than the Ni based catalysts, was more stable and it showed no deactivation after a 360 min test. These results open new possibilities that should be explored regarding mixtures of nitrides and oxides. Nitrides are known to be deactivated in the presence of water due to oxidation, but these oxides might tolerate better the presence of impurities, enhancing their stability.

Zhang et al. [39] prepared cobalt nitride supported on a nitrogen doped carbon CoNx@NC using cellulose and ammonia as the carbon and nitrogen source respectively at different synthesis temperatures from 500 to 800 ◦C. The catalysts were tested in the HDO of eugenol at 2 MPa H2 and 200 ◦C. According to the reaction results, the reaction follows different paths when using nitrides or metallic cobalt. While the nitrides favour the cleaveage of the C-aryl−OCH3 bond to form 4-propylphenol, metallic cobalt promotes the hydrogenation of the alkene moiety. The best catalyst was CoNx@NC-650 which displayed the largest surface area and dispersion of the nitrides nanoparticles. The catalyst was also successfully tested to promote the HDO of phenolic compounds.

Supported CoNx on carbon nanotubes on the hydrogenation of nitrobenzene and hydrogenated coupling of nitrobenzene with benzaldehyde was studied by Zhang et al. [40] as schematized in Figure 5. Some catalytic tests verified that catalytic activity was mainly due to the CoNx entities and the authors also suggested that the activity was mainly due to the cobalt chelate complexes bonded to nitrogen atoms of the graphene lattice.

**Figure 5.** CoNC/CNT active sites on nitro compounds hydrogenation and hydrogenated coupling of nitrobenzene with benzaldehyde. Reproduced from [41], with permission from Royal Society of Chemistry, 2016.

Lodeng et al. [41] compared the activity of molybdenum nitride, carbide, and phosphide supported on TiO2 on the HDO of phenol at 25 bar and in a temperature range between 350 and 450 ◦C. All the catalysts were highly active to benzene and only minor amounts of aromatic ring hydrogenation were obtained. Molybdenum nitride displayed lower activity compared to its carbide and phosphide counterparts, but its selectivity to cyclohexene was higher than that of phosphide and similar to carbide.

Hydrogenation of −COH moieties constitutes one of the most interesting and studied reactions in fine chemistry since it allows obtaining a high number of compounds that are used for example in pharmaceuticals and/or fragances. Reactants such as cinnamaldehyde or crotonaldehyde have been widely studied in an attempt to heterogeneized the catalytic system. The catalysts must be selective to the unsaturated alcohols without reducing the C=C bonds. With that aim heterogeneous catalysts based on noble metals mainly Ru, Pd and Pt have been widely investigated with good results in terms of activity and selectivity [42] and the significant role of nitrogen improving the selectivity to the desired products have also been reported [43]. To date, these specific transformations has been tested with bimetallic systems of metal nitrides and noble metals.

Fu et al. [44] have used a previously functionalized support to obtain small nitrides nanoparticles supported on the mesoporous silica SBA-15. The synthesis of Mo2N over SBA-15 started by functionalizing the support with a monoamine that is located homogeneously into the pores of the support as shown ion Figure 6. This amine is then used as anchoring point for the molybdenum precursors which preferentially adsorbs on the moieties. Then, the procedure follows the previously explained temperature reduction procedure in NH3. Finally the noble metal is impregnated and reduced with NaBH4 forming bimetallic phases with Mo2N. In this way, Mo2N and Pt nanoparticles with a size of about 8.0 and 5-6 nm respectively were obtained with metal loadings of Mo (22% wt.) and Pt (3% wt.).

Then, Pt/Mo2N/SBA-15 with different Pt loadings (1–3% wt.) was tested in the chemoselective hydrogenation of cinnamaldehyde. Both activity and selectivity to the cinnamyl alcohol was higher over Pt/Mo2N-based catalysts than over monometallic Pt/SBA-15, that the authors ascribed to the synergy between Pt and Mo2N nanoparticles and the more efficient use of the Pt surface on the bimetallic sample.

In the work of Thomson et al. [45], some more insight was given regarding the reaction mechanism and nature of active phases. To do so, the authors used a thorough experiment by hydrogenating a high surface area γ-Mo2N to obtain a partially hydrogenated entity γ-Mo2N-Hx. Then, by combining H2-TPD experiments and DFT simulations three different hydrogen species were identified: surface nitrogen bound (κ1-NHsurf), surface Mo bound (κ1-MoHsurf) and subsurface Mo-bound (μ6-Mo6Hsub).

**Figure 6.** Synthesis of Pt-Mo2N-SAB-15. Reproduced from [45], with permission from American Chemical Society, 2016.

The reactivity of these species was assessed by testing them in the hydrogenation of crotonaldehyde. Accordingly, the authors proposed that reaction starts by a heterolytic dissociation of H2 to form surface NH (κ1-NHsurf) and MoH (κ1-MoHsurf) as schematized in Figure 7. Then, since subsurface interstitial H site (μ6-MoHsub) is more energetically favored than surface κ1-MoHsurf, hydrogen migrates into the lattice. Moreover, based on the catalytic results the authors proposed that surface and subsurface species MoH (κ1-MoHsurf/μ6-MoHsub) are more selective to the hydrogenation of the C=O bond and the surface κ1-NHsurf sites hydrogenate preferentially the C=C bond.

**Figure 7.** Reproduced from [45], with permission from American Chemical Society, 2016.

The selective hydrogenation of acetylene to ethylene is an important transformation since ethylene is the monomer to produce polyethylene polymers and it has a strategic relevance in refineries, being critical its high purity production.

The catalytic behaviour of β- and γ-Mo2N in the partial hydrogenation of acetylene was evaluated by Lizana et al. [46] that studied the influence of synthesis parameters on textural properties of the nitrides and its effect on the catalytic performance. The results showed that selectivity of both β- and γ-Mo2N was higher than over Pd-based catalysts. Also, β-Mo2N which displays higher surface Mo/N ratio compared to γ-Mo2N, offered lower selectivity to partial hydrogenation and a two-fold higher specific acetylene hydrogenation rate.

Altarawneh et al. [47], used computational methods to study mechanism of the selective hydrogenation C2H2 over γ-Mo2N to C2H4 rather than complete hydrogenation to the corresponding alkane.

Reactions take place through H2 adsorption followed by dissociation. The authors obtained the modes of H2 adsorption as shown in Figure 8: 3-fold hollow fcc (H1) and 4-fold hollow fcc (H3) sites over the (111) and (100) terminations of γ-Mo2N, respectively.

**Figure 8.** H2 adsorption sites on Mo2N (111) and (100) faces. Reproduced from [48], with permission from American Chemical Society, 2016.

In agreement with experimental results, this work seems to confirm that dissociation of H2 occurs over nitrogen vacancies. It is also proposed that the lower stability of the partial hydrogenated molecule, C2H4 leads the selectivity.

Another interesting reaction within fine chemistry is the hydrogenation of nitroaromatic compounds, since aromatic haloamines are important intermediates in the manufacturing of drugs, pesticides, and pigments among others. The reaction has been successfully performed over Au [48], Ir [49] and Pd [50] catalysts supported over a variety of materials.

Keane et al. [51] demonstrated experimentally that Mo2N improved the performance of Au in the selective hydrogenation of *p*-chloronitrobenzene (p-CNB) to *p*-chloroaniline (p-CAN) reaching 100% selectivity to p-CAN, a four-fold higher hydrogenation rate compared to Au/Al2O3 and showed stability upon several cycles.

The system Pd/Mo2N was an effective catalyst for the hydrogenation of p-nitrophenol (PNP) to p-aminophenol (PAP) [52]. In this study, the authors synthetized Pd/Mo2N nanoparticles of 2–3 nm size over SBA-15. The high dispersion improved the interaction between Pd and the nitride so that 1 wt% Pd–Mo2N/SBA-15 showed better catalytic performance than 1 wt% Pd/SBA-15 and 20 wt% Pd/SBA-15. In Figure 9 the proposed reaction mechanism over the Pd–Mo2N/SBA-15 is shown.

**Figure 9.** Catalytic conversion mechanism of PNP into PAP over the Pd–Mo2N/SBA-15 hybrids in the presence of NaBH4. Reproduced from [53], with permission from American Chemical Society, 2018.

Wu et al. [53] reported a green solvent-free synthesis method for CoNx entities supported on doped mesoporous carbon materials (CoNx-OMC) with surface areas in the range 678–1250 m2/g, high N content (4.3–10.8 wt%) and rich in CoNx sites as verified by XPS. The optimized CoNx-OMC, thermally treated at 800 ◦C) catalyst showed an interesting catalytic performance on the hydrogenation of several nitro compounds, i.e., 100% conversion, almost 100% selectivity and stable upon recycling, under mild conditions (5 bar H2 pressure, 110 ◦C). According to the catalytic and characterization results, a synergy effect is reached between CoNx sites and the nitrogen doped support. On the one hand, the CoNx entity provides specific sites for the adsorption and activation of nitro groups. On the other hand, the nitrogen heteroatoms of the support act as anchoring sites, increasing the dispersion of the active components and facilitating mass transportation. Also, it was confirmed that cobalt nitride was responsible of the activity and not the metallic Co nanoparticles.

Density functional theory calculations were performed by Altarawneh et al. [54] who studied the hydrogenation of p-CNB to p-CAN over the model γ-Mo2N(111) surface. The results showed that adsorption of p-CNB is thermodynamically favoured over Mo-hollow face-centered cubic (fcc) and N-hollow hexagonal close-packed (hcp) sites with adsorption energies of −32.1 and −38.5 kcal/mol, respectively. Also, the results are in agreement with previous experimental reports that described a high selectivity to p-CAN at low temperatures, the direct path being preferential versus the condensation route [55].

The theoretical results suggest that activated hydrogen, H\*, is transferred from both fcc and hcp hollow sites to the NO/−NH groups, the hydrogenation of chloronitrosobenzene being the rate-limiting step with an energetic barrier of 55.8 kcal/mol. Also, the high energy barrier for direct fission of the C−Cl bond excludes the formation of aniline.

Another industrial transformation of acetylene is the hydrochlorination to obtain vinyl chloride monomer (VCM), the basic unit of polyvinyl chloride (PVC). Industrially this reaction is performed using HgCl2 as catalyst, however due to environmental and health concerns, alternative systems have been evaluated [56]. Among them, noble metal, i.e., Au, Pd, Pt and Ru over activated carbon, have been the most studied, gold being the best in terms of activity and selectivity.

Since it has been previously found that nitrogen doping can promote adsorption for HCl [57], which is the VCM synthesis reaction rate-controlling step, a potential catalytic system is that consisting of metal nitrides.

Dai et al. [58] studied V, Mo, and W nitrides (10 wt% metal loading) supported on activated carbon. Their experiments showed that VN/AC offered very low activity in the reaction. Mo2N was initially the most active, but deactivated with time to give lower conversion values than those reached with W2N/AC. The selectivity to C2H3Cl was similar for all the tested nitrides and reached near 100%. To further explain these results, TPD of the reactants, C2H2 and HCl, was performed and showed that W2N/AC dislayed a stronger and similar interaction with both reactants, compared to the other catalysts. However, Mo2N/AC which deactivates during reaction, showed an easy desorption of HCl and difficult desorption of C2H2, this producing significant amounts of coke which can be responsible of the observed deactivation.

The authors also studied binary Mo and Ti nitrides with different Mo/Ti ratios supported on activated carbon for acetylene hydrochlorination [59]. All the binary nitrides displayed better catalytic performance compared to the mononitrides and a Mo/Ti ratio of 3 was optimal among the studied systems offering 89% conversion and selectivity over 98.5%. Apparently a synergy effect among Mo and Ti occurs so that adsorption of HCl is enhanced while adsorption of acetylene is reduced.

Other chloro compounds have been evaluated with TMN. For example Keane et al. [60] studied the gas phase hydrodechlorination of 1,3-dichlorobenzene (1,3-DCB) using molybdenum and tungsten carbide (Mo2C, W2C) and nitride (Mo2N). fcc-Mo2N showed better activity (by a factor of 20) compared to pure hcp- and fcc-carbides, that displayed similar activity.

#### *3.2. Oxidation*

The selective oxidation of carbon monoxide with low concentration of O2 (CO-PROX) is used to purify hydrogen rich streams obtained from hydrocarbon reforming. So far, the most studied and active system are the CuO–CeO2 and Pt based catalyst [61].

With the aim of finding more economical active and selective catalysts, the catalytic performance of several transition metal nitrides have been assessed. For example, Yang et al. [62] studied the effect of Co loading (1 to 10 wt%) on Co4N supported on γ-Al2O3. The cobalt nitrides displayed similar activities compared Pt-group metals in the temperature range 200–220 ◦C. The sample 3 wt% Co/γ-Al2O3 offered the best activity and selectivity in PROX reaction which could be related to the higher concentration nitrogen vacancies in the near-surface that enhance the adsorption of reactants.

Selective oxidation of alcohols plays an important role in many industrial transformations such as energy conversion and storage or the production of fine chemicals. The most studied and active catalysts are supported Au Pt, Pd, Ag and Ru and other non-noble-metal catalysts such as Co and Cu [63].

In order to test more economically viable alternatives Deng et al. [64] evaluated the catalytic activity of iron nitride, FeN4, supported on graphene in the oxidation of benzene using H2O2 as oxidant. The characterization suggested that atomically dispersed FeN4 were obtained and these entities were also stable after reaction. The catalyst reached 23.4% conversion and 18.7% yield of phenol at room temperature, however no comparison with other catalyst is given.

Later, Yuan et al. [65] studied several metal nitrides (MNx/C-T, M = Fe, Co, Cu, Cr, and Ni) synthetized at different pyrolysis temperatures, T, in the oxidation of unsaturated alcohols. The most active and selective catalyst to the corresponding aldehydes was the iron nitride. Among them, the catalysts prepared by thermal treatment at 900 ◦C displayed the better catalytic performance in the selective oxidation of HMF to DFF with almost complete conversion and selectivity exceeding 97%. According to the characterization performed the authors suggested that when thermal treatment was performed at lower temperatures, c.a. 600 ◦C, a higher concentration of nitrogen doped carbon was formed, in detriment of iron nitride; and that nitrogen doped carbon offered lower activity. In contrast, the formation of FeN4 was favored at higher synthesis temperatures, resulting in materials that offered

better activity. Moreover, the activity of the recycled catalysts could be restored by thermal treatment under NH3/N2.

With the aim of gaining more insight into the active sites of the Fe−N−C catalysts. Zhang et al. [9] studied atomically dispersed Fe−N−C catalyst synthetized using nano-MgO as a template. The catalysts were tested in the selective oxidation of the C−H bond of a wide range of aromatics, heterocyclic, and aliphatic alkanes at room temperature. The catalysts showed high activity and selectivity of up to 99%, as well as great reusability. The atomical dispersion of FeNx (x = 4 − 6) was verified using sub-Ångström-resolution HAADF-STEM along with XPS, XAS, ESR, and Mössbauer spectroscopy.

The authors also reported that the concentration of each FeNx species depends on the pyrolysis temperature. Among the studied samples, the most active was Fe−N−C-700 which is comprised of high-spin FeN6 (28.3%), low-spin FeN6 (53.8%), and medium-spin FeN5 (17.9%) species, as shown in Figure 10. This latter is over one order of magnitude more active than the other two species. Upon increasing the pyrolysis temperature, the concentration of FeN5 decreases to less than 10%, leading to lower activity.

**Figure 10.** Reproduced from [9], with permission from American Chemical Society, 2017.

#### *3.3. Ammonia Synthesis and Decomposition*

The production of ammonia through the Haber–Bosch Process was a technological breakthough of the last century since it is a relatively easy way of obtaining a synthetic fertiliser. The process uses promoted iron catalyst to produce ammonia from N2 and H2 at temperatures of ca 400 ◦C and high pressure of around 100–200 atmospheres. The process of ammonia synthesis is so relevant that it currently consumes near 1–2% of the world's energy demand [66].

This data reveals that more efforts can be made to optimize the process as small changes on the reaction conditions could have a huge impact on global energy consumption. Thus, despite being a well-known process, several attempts have been performed to obtain more active and stable catalysts to work under milder conditions [67].

In the last decades binary and ternary transition metal nitrides of the type Mo, Co and Fe have been tested in both ammonia decomposition and ammonia synthesis. Simulations and experimental tests have shown that cobalt molybdenum nitride, Co3Mo3N can be the most active catalyst for ammonia synthesis, superior to the industrial iron catalyst and promoted ruthenium catalyst [68,69].

The synthesis of ammonia catalyzed by molybdenum nitride seems to be a structure-sensitive reaction so that the bigger the particle size, the higher the intrinsic activity of the catalyst [70]. On the other hand, the effect of the nanoparticle morphology does not seem to be so clear. Sun et al. studied the synthesis of ammonia over plate-like γ-Mo2N and nanorod β-Mo2N and γ-Mo2N [71] and did not find significant differences among them. Also, the synthesis conditions limit the comparison since residual sulphur from the molybdenum precursor, Mo2S, can poison the catalyst.

In order to increase the dispersion of the nitrides and improve the stability, Ding et al. [72] synthetized molybdenum nitride supported on HZSM-5, using MoOx as precursor and NH3 as nitriding agent at 973 K. In this way molybdenum oxide exchanges with hydroxyl groups on the zeolite surface and obtained a Mo to N ratio close to 2. The authors proved that this supported molybdenum nitrides are more stable against oxidation by comparing with unsupported catalysts.

The resulting catalyst, MoNx/ZSM-5, displayed excellent activity in the ammonia synthesis under ambient pressure. A kinetic approach showed that the apparent activation energy for ammonia synthesis on MoNx/ZSM-5 is around 20% lower than that on bulk Mo2N (9.8 kcal/mol vs. 12.4 kcal/mol).

Interestingly, the effect of reaction pressure was different on bulk Mo2N and MoNx/ZSM-5, i.e., higher reaction pressures are more favorable on the MoNx/ZSM-5 catalyst than on γ-Mo2N. The authors suggested that this difference could be due to the interaction of nitrogen with the zeolite framework, acting as an equivalent pressure which may enhance the pressure effect on the ammonia synthesis. Another significant finding of this work is that upon increasing the Si/Al ratio of the zeolite, the catalytic activity of MoNx increases.

Recently, Catlow et al. [73] used DFT calculations to evaluate the associative and dissociative mechanisms of ammonia synthesis over Co3Mo3N. The results showed that in the associative mechanism, Eley-Rideal/Mars van Krevelen, hydrogen reacts with nitrogen adsorbed and activated on the surface to generate ammonia.

Promotion of cobalt molybdenum nitrides with other elements have been also studied. According to Paweł Adamski [74] chromium and potassium were able to generate a well-developed porous structure, increasing the activity of the catalysts in ammonia synthesis by 50% compared to the non-promoted catalyst.

Ammonia decomposition has become a strategic research topic since it allows to obtain CO free hydrogen to feed fuel cells. Ammonia can also be employed directly as fuel in vehicles since it has a high energy density (8.9 kW h/kg), is easily liquified at room temperature and low pressure, i.e., less than 10 bar, and its narrow combustion range allows safe operation.

Many cataytic monometallic and bimetallic systems have been tested based on Fe, Ni, Co, Ru, Ir, Pt or Rh, supported on several materials. Among them, ruthenium catalysts supported on different carbon materials such as active carbon, carbon nanotubes showed the highest activity.

In all cases, researchers observed a low activity for ammonia decomposition reaction at temperatures below 400 ◦C, since recombinative desorption rate of the adsorbed N atoms from active metals is slow at those conditions [75]. Also, hydrogen molecule seems to cover active sites over Ni- and Ru catalysts, this hindering the ammonia decomposition reaction. Hence, despite being well studied systems, the temperatures required and the use of noble metals should be avoided to make the process technical and economically feasible.

Eguchi et al. [76], studied the effect of a second transition metal on Mo nitride based catalysts, prepared by temperature-programmed reaction under NH3 flow of the oxides precursors: MoO3, CoMoO4, NiMoO4, and FeMoO4. Incorporation of the second metal into the Mo nitride resulted in a significant decrease in the surface area (3.1–8.8 vs. 80 of Mo2N). However, the area of Mo2N was reduced to 23 m2/g after reaction, indicating that the material was not stable under the reaction conditions. Despite the lower surface area, the addition of a second metal was beneficial for ammonia decomposition and the activity followed the trend Co3Mo3N > Ni3Mo3N > Fe3Mo3N > Mo2N. Several transition metal nitrides have been studied for ammonia decomposition, being the binary system metal-cobalt molybdenum nitride the most interesting among them [77–81].

Based on the NH3-TPSR results, the authors suggested that the addition of Co and Fe favoured the desorption of hydrogen. However, over Mo2N de desorption of nitrogen was slower since nitrogen atoms tend to interact stronger with the nitride. According to the results the authors concluded that the order in which each metal nitride system dissociates metal nitride–N bond was Co3Mo3N ≈ Ni3Mo3N > Fe3Mo3N > Mo2N. Moreover, the presence of Co, Ni, and Fe improves the stability against poisoning by H2 on the active sites and the best results are obtained when doped catalysts are employed.

In order to improve the surface area of nitrides, alternative synthesis paths have been explored. For example Podila et al. [77] studied the use of citric acid as chelating agent to prepare bulk Co3Mo3N, with surface areas in the range 93–129 m2/g. The use of citric acid (CA) as chelating agent afforded better nitride dispersion which resulted in better catalytic performance. An optimal concentration of citric acid was related to a higher surface area, lower particle size and increased proportion of

Mo2N and Co3Mo3N phases, so that when the CA/Mo ratio was changed from 1 to 3 in the synthesis, the conversion increased from 75% to 97% at 550 ◦C.

Similarly Zaman et al. synthetized nickel [78] and cobalt molybdenum nitrides [79] using citric acid and compared its catalytic performance on ammonia decomposition with that obtained employing γ-Mo2N. Under these synthesis conditions, the surface area of the binary nitrides was increased but still below 20 m2/g. According to the results both catalyst, Ni2Mo3N and Co3Mo3N offered over 97% conversion, while the use of pure γ-Mo2N resulted in 50–70% conversion under the same experimental conditions.

Zhao et al. [80] used supported binary CoMo nitrides over several porous materials with different physico-chemical characteristics: CNTs, Al2O3, activated carbon and 5A Zolite. The experimental procedure was a simple impregnation of the precursors followed by a temperature-programmed reaction in N2–H2. The activity followed the order: CoMoNx/CNTs > CoMoNx/Zeolite 5 > CoMoNx/AC > CoMoNx/Al2O3. However, despite the clear better performance of CNT, the supports differ in many features such as surface chemical composition and morphology, making the comparison and related conclusions quite difficult.

The effect of synthesis conditions, nitridation temperature and iron loading was also evaluated in iron nitrides supported on carbon nanotubes [81]. A higher synthesis temperature of ca. 500 ◦C and Fe loading of 10% prepared under NH3 flow resulted in well-dispersed Fe2N nanoparticles which exists along with Fe2O3 entities and offered the best ammonia conversion among the catalysts tested. In contrast the synthesis under a N2/H2 flow resulted in the formation of both Fe2N and Fe4N, this latter species being detrimental for ammonia decomposition.

#### **4. Conclusions**

This short review of the most recent literature has shown that transition metal nitrides possess a wide spectrum of catalytic applications with interesting catalytic performance. The review has focused on thermal heterogeneous catalysis which requires high surface areas and good dispersion of the active phase. Despite the significant advances done up to date, there is plenty of room for research on transition metal nitrides to optimize the synthesis conditions in order to obtain higher surface areas and better nanoparticles dispersion, ideally reducing the synthesis temperature.

Also, further work to improve the stability will be required in order to obtain potential industrial catalysts. One of the main reason for deactivation is the oxidation of the nitrides, which is likely to occur in reactions that generate water such as carbon dioxide hydrogenation, or when water is already in the reactants mixtures as it happens with biomass transformations. In this sense, it seems that subsurface hydrogen can delay deactivation and more insight into this reaction mechanism would allow to propose regeneration mechanism.

Again, the use of a high surface area support for transition metal nitrides nanoparticles can improve the dispersion of the active phase and potentially improve their stability upon reaction conditions against sintering and oxidation.

The improved catalytic performance that has been reached with more complex systems that incorporate a second or third metal, should be complemented with deeper understanding of the actual active phase and the chemical structure of the nitrides. Similarly, the use of promoters like alkalis and its effect of the structure need to be further studied since these materials have also demonstrated a significant potential for future catalytic applications. However, there is still no clear correlation mainly due to the complexity of these systems and difficulties to perform in situ investigations.

**Funding:** A.B.Dongil acknowledges financial support from Fundación General CSIC (Programa ComFuturo).

**Acknowledgments:** A.B. Dongil acknowledges financial support from Fundación General CSIC (Programa ComFuturo).

**Conflicts of Interest:** The author declares no conflicts of interest.

#### **References**


© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Metal–Organic Framework-Based Sustainable Nanocatalysts for CO Oxidation**

#### **Luis A. Lozano, Betina M. C. Faroldi, María A. Ulla and Juan M. Zamaro \***

Instituto de Investigaciones en Catálisis y Petroquímica, INCAPE (FIQ, UNL, CONICET), Santiago del Estero 2829 (3000), Santa Fe, Argentina; llozano@fiq.unl.edu.ar (L.A.L.); bfaroldi@gmail.com (B.M.C.F.); mulla@fiq.unl.edu.ar (M.A.U.)

**\*** Correspondence: zamaro@fiq.unl.edu.ar

Received: 17 December 2019; Accepted: 14 January 2020; Published: 17 January 2020

**Abstract:** The development of new catalytic nanomaterials following sustainability criteria both in their composition and in their synthesis process is a topic of great current interest. The purpose of this work was to investigate the preparation of nanocatalysts derived from the zirconium metal–organic framework UiO-66 obtained under friendly conditions and supporting dispersed species of non-noble transition elements such as Cu, Co, and Fe, incorporated through a simple incipient wetness impregnation technique. The physicochemical properties of the synthesized solids were studied through several characterization techniques and then they were investigated in reactions of relevance for environmental pollution control, such as the oxidation of carbon monoxide in air and in hydrogen-rich streams (COProx). By controlling the atmospheres and pretreatment temperatures, it was possible to obtain active catalysts for the reactions under study, consisting of Cu-based UiO-66-, bimetallic CuCo–UiO-66-, and CuFe–UiO-6-derived materials. These solids represent new alternatives of nanostructured catalysts based on highly dispersed non-noble active metals.

**Keywords:** UiO-66; copper; iron; cobalt; nanocatalyst; CO oxidation; COProx

#### **1. Introduction**

The use of metal–organic frameworks (MOFs) as host matrices to disperse metal nanoparticles is a topic of great interest in the ongoing research for new nanocatalysts. MOFs have the advantage of presenting a variety of transition metals and a wide range of organic ligands in their composition, which makes them attractive for applications in catalysis [1,2]. There are several alternatives to obtain catalytic functionality in the structure of MOFs. One is based on the coordination around metal centers or structural defects that are not committed to the material framework, which can act as active sites [3]. Another possibility is to take advantage of the ligand chemistry since terephthalates or tricarboxylates to which acidic or basic functional groups can be added are usually used [4]. Moreover, the large internal volume available in these materials can be used to host active species [5]. In addition, in recent years, the use of MOFs as templates, which according to their construction units can generate nanostructured metal/metal oxide systems [6], became a consolidated strategy. For example, Sun et al. [7] reported a number of catalysts obtained with this concept, such as Co3O4 materials from Co-based MOFs or α-Fe2O3 and Fe3O4 nanomaterials from Fe-MIL-88B. CuO and CuO–CeO2 nanoparticle-based catalysts derived from the MOF HKUST-1 with high catalytic activity were also reported [8].

On the other hand, the catalytic oxidation of carbon monoxide in gaseous streams is a reaction of great environmental relevance. Carbon monoxide is one of the main pollutants of indoor and industrial environments because, due to its high affinity for hemoglobin, it is extremely toxic to living beings [9] and, therefore, numerous studies are currently being conducted for their catalytic removal [10]. Moreover, the elimination of this gas in concentrated hydrogen streams (COProx) has significance in the field of renewable energy, as it is one of the most accepted alternatives to carry out the final purification of H2 to be used in fuel cells [11]. For the CO oxidation, numerous catalytic formulations were tested to date, and those composed of supported oxides represent one of the most promising alternatives [12]. For example, catalysts based on supported particles of copper oxides, cobalt oxides, and iron oxides showed good performance [13–15]. These types of solids avoid the use of noble metals traditionally used in this reaction such as Pt, Pd, or Au [12], some of which were also supported in MOFs [16,17]. These elements have a limited abundance and involve much higher costs; thus, efforts are being made for the development of catalysts based on non-noble metals. Currently, it is considered that the use of precious metals is not sustainable compared to earth-abundant metals which are available in orders-of-magnitude higher quantities [18]. Of special interest are the metals of the first transition period, such as Cu, Co, and Fe, since they are not only less expensive but also less toxic compared to those of the second and third period [19]. Since MOFs have high specific surface areas, they represent a new support alternative for the efficient and low-cost obtainment of catalysts based on dispersed non-noble metal species. In this scenario, UiO-66 is an attractive structure for this purpose since it is a microporous zirconium terephthalate forming a three-dimensional arrangement with high specific surface area and good thermal, mechanical, and chemical stability [20]. In addition, it requires low-cost precursors for its synthesis and can be obtained under fairly sustainable conditions. Very recently, active catalysts based on atomically dispersed ionic Cu species on UiO-66 and hybrid nanostructures of CuO nanocrystals encapsulated in UiO-66 crystals were reported [21,22]. In addition, with the concept of a matrix, CuO/CeO2 active catalysts derived from Ce–UiO-66 were obtained, as well as CuCe/ZrO2 catalysts derived from metal-impregnated UiO-66 [23,24].

In this context, the present work proposes to systematically analyze the use of UiO-66 synthesized through a sustainable protocol [25] as a dispersion matrix of Cu, Co, and Fe species to obtain new nanoparticle structures with potential use in catalysis. In addition, the incorporation of these metallic species is proposed by simple procedures traditionally used to prepare supported catalysts, such as the incipient wetness impregnation with precursors and subsequent thermal decomposition. The physicochemical properties of the obtained nanomaterials were tested in the gas-phase oxidation of carbon monoxide in air streams and in the CO preferential oxidation in hydrogen-rich streams (COProx).

#### **2. Materials and Methods**

#### *2.1. Synthesis of UiO-66*

Benzenedicarboxylic acid (BDC, Aldrich, 98.0% purity, St. Louis, MO, USA), ZrCl4 (Zr, Aldrich, 98.0% purity, Darmstadt, Germany), and acetone (Cicarelli, 99.0% purity, San Lorenzo, Argentina) were used without further purification, and UiO-66 synthesis was performed employing a sustainable protocol reported elsewhere [25]. Briefly, the procedure consisted of mixing the two solid reactants together with the solvent in the molar proportions BDC:ZrCl4:solvent = 1:1:1622. After obtaining the homogeneous mixture, it was placed under solvothermal treatment at 80 ◦C for 24 h. At the end of the treatments, the solids were recovered by centrifugation (10,000 rpm, 10 min), washed twice with ethanol, and finally dried at 80 ◦C overnight.

#### *2.2. Incorporation of Metallic Species*

Cu(NO3)2·3H2O (Aldrich, 98.0–103% purity, St. Louis, MO, USA), Co(NO3)2·6H2O (Alfa Aesar, 98.5% purity, Tewksbury, MA, USA), and Fe(NO3)3·9H2O (Aldrich, ≥99.99%, purity, St. Louis, MO, USA) were employed as metal precursors. For their incorporation to MOF, incipient wetness impregnation was used as described in S1 (Supplementary Materials). The nomenclature employed was the following: First the incorporated metal, then its load in wt.% with respect to the catalyst total mass, and finally the support which was MOF (M) or its degradation product (Zr), i.e., Cu10/M or Cu16Fe7/Zr.

#### *2.3. Catalyst Characterization*

The crystalline structure of the materials was analyzed trough X-ray diffraction (XRD) with a Shimadzu XD-D1 instrument (Shimadzu Corp., Kyoto, Japan, CuK<sup>α</sup> radiation, <sup>λ</sup> <sup>=</sup> 1.5418 Å, 2 ◦C·min<sup>−</sup>1, 30 mV, 40 mA, 2θ = 5◦ to 65◦). In order to analyze the thermal evolution of the MOF and precursor, thermogravimetric analysis (TGA) and single differential thermal analysis (SDTA) were conducted with a Mettler Toledo STARe (version 4.1, Bristol, UK) TGA/SDTA 851e module from 25 to 800 ◦C at 10 ◦C·min−<sup>1</sup> in air or nitrogen flow (50 mL·min−1, standard temperature and pressure (STP)). Laser Raman spectroscopy (LRS) was performed using a LabRam spectrometer (Horiba-Jobin-Yvon, Stanmore, UK) coupled to an Olympus confocal microscope (Olympus Corp., Shinjuku, Tokyo, Japan) equipped with a charge-coupled device (CCD) detector cooled to about 200 K. The excitation wavelength was 532 nm (Spectra Physics argon-ion laser), and the laser power was set at 30 mW. Transmission electron microscopy (TEM) images of the synthesized UiO-66 crystals and metal-based nanocatalysts were acquired using a JEOL 2100 Plus microscope (JEOL Ltd., Tokyo, Japan) equipped with an energy dispersive X-ray (EDX) detector (JEOL Ltd.) and a scanning transmission electron microscope (STEM) (JEOL Ltd.). The samples were milled and suspended in ethanol by ultrasonic treatment, and a drop of the fine suspension was placed on a carbon-coated nickel grid to be loaded into the microscope.

#### *2.4. Catalytic Evaluations*

The samples were evaluated in a glass tubular reactor connected to a continuous flow system equipped with mass flow controllers (Brooks 4800, Brooks Instrument, Hatfield, PA, USA) and heated with a furnace controlled with a proportional–integral–derivative (PID) system. Before each evaluation, the reactor was heated at different temperatures (200–650 ◦C) in He or Air flow (30 mL·min<sup>−</sup>1) according to the pretreatment required by the sample before its test and maintaining such a temperature for 60 or 120 min. Then, the catalytic tests were performed with a mixture of molar composition 1% CO, 2% O2 (employing synthetic air) in He balance, maintaining a total flow of 30 mL·min−<sup>1</sup> with an initial mass of the solid of 70 mg. Some tests were also performed by adding 50% of H2 in the reaction stream (COProx), at the same total flow. The catalytic measurements were taken after stabilizing the reactor at different temperatures for 8 min. The CO conversions were determined with an on-line Shimadzu GC-2014 chromatograph (Shimadzu Corp., Kyoto, Japan) equipped with a thermal conductivity detector (TCD) and a 5-Å molecular sieve packed bed column. CO conversions (XCO were calculated as follows:

$$\text{XCO} \left( \% \right) = \left( \text{[CO]} \right)^0 - \left[ \text{CO} \right] \left[ \text{[CO]} \right]^0 \times 100 \text{ \textdegree$$

where [CO]<sup>0</sup> and [CO] are inlet and outlet gas concentrations (ppm), respectively. Moreover, for COProx, the selectivity of O2 to CO2 (S) was calculated as follows [12]:

$$\text{CS} \left( \% \right) = \lambda \text{CO} \left( \% \right) / \lambda \times \left( \left( \left[ \text{O}\_2 \right]^0 - \left[ \text{O}\_2 \right] \right) \left[ \text{O}\_2 \right]^0 \right) \text{.}$$

where [O2] <sup>0</sup> and [O2] are inlet and outlet oxygen concentrations (ppm), respectively. In our case, the value of the factor λ (excess oxygen in the reaction) was four.

#### **3. Results and Discussion**

#### *3.1. Dispersion of Copper Species in UiO-66*

#### 3.1.1. Study of the Precursor Decomposition

The obtained MOF corresponded to a crystalline and pure phase of UiO-66 since all its diffraction signals were observed (Figure S1, Supplementary Materials); the most important ones were identified at 2θ = 7.38◦, 8.52◦, and 25.75◦, corresponding to the (111), (200), and (600) planes, respectively [20]. All these signals matched those indexed for this MOF from its crystallographic data (CCDC 733458) and no impurities were detected, such as benzenedicarboxylic acid. Afterward, the compatibility

between the decomposition temperatures of the metal precursors and the MOF was analyzed by TGA, both in an inert and in air atmosphere. During the impregnation process, the transition metals were incorporated as nitrate salts onto the UiO-66 surface. Afterward, the nitrate ion decomposition was required to obtain the metal oxides. In this context, it is necessary to analyze the atmosphere and temperature condition in which this decomposition was done, maintaining the MOF structure. Figure 1a, profile 3, presents the TGA of UiO-66 in inert gas, and Figure 1b, profile 3, depicts its corresponding derivative TGA (dTGA). Typical evolutions were observed, with an initial mass loss up to 130 ◦C due to physisorbed water and/or residual solvent of synthesis. Then, another small mass loss from 180 ◦C due to the dehydroxylation of the inorganic cluster of the MOF from Zr6O4(OH)4 to Zr6O6. Finally, the thermal decomposition of the ligands was observed, which was similar to the one corresponding to this MOF synthesized under conventional conditions [26], with a maximum decomposition rate (Tmax) at 555 ◦C (Figure 1(b3) and Table 1). Instead, under air atmosphere, the TGA (Figure 1(c3)) and dTGA profiles (Figure 1(d3)) showed a somewhat lower stability with a Tmax of 520 ◦C (Table 1). This evolution was more exothermic compared to the previous one as could be seen in the SDTA (Figure S2, Supplementary Materials). Meanwhile, the copper precursor showed a complete decomposition at 290 ◦C or 265 ◦C in inert gas (Figure 1(a4)) or air (Figure 1(b4)), respectively. The degradation temperature window of the MOF and the precursor suggested the possibility of obtaining dispersed copper species by applying heat treatments.

**Figure 1.** Thermogravimetric analysis (TGA) of UiO-66 (M), Cu/M, and copper precursor: (**a**) N2 atmosphere; (**c**) air atmosphere. Corresponding derivative TGA (dTGA) curves: (**b**) N2 atmosphere; (**d**) air atmosphere.

**Table 1.** Maximum decomposition temperatures (Tmax) of the fresh metal–organic framework (MOF) and of the MOF impregnated with copper (Cu/M) and that of the respective metal precursors (obtained from the derivative thermogravimetric analysis (dTGA) data).


<sup>1</sup> Decomposition temperature (◦C) of the maximum in the N2 dTGA profile. <sup>2</sup> Decomposition temperature (◦C) of the maximum in the air dTGA profile.

The MOF impregnated with 5 wt.% copper (Cu5/M) showed a marked decrease in the structural stability of the framework (Figure 1a,b), which was magnified with a higher copper amount (Cu10/M). The Tmax was reduced from 555 ◦C to 490 ◦C and 460 ◦C for fresh UiO-66, Cu5/M, and Cu10/M, respectively (Table 1). The same effect was observed in air but with a greater destabilization of the framework. This phenomenon is attributed to the copper species formed after the precursor decomposition, which catalyzed the oxidation of the organic part of the MOF, accelerating its structural collapse. However, a small temperature gap persisted in which it would be possible to decompose the said precursor before the MOF collapsed. The SDTA of Cu/M in both atmospheres (Figure S2, Supplementary Materials) showed only an endothermic process due to the evacuation of host molecules that ended at 160 ◦C, and an exothermic one near 300 ◦C due to MOF collapse. From the previous results, pretreatments of Cu10/M in inert and combinations with air were carried out at different temperatures prior to carrying out its catalytic test in order to get insight into the catalytic performances and structural stabilities after the different pretreatments.

#### 3.1.2. Copper-Based UiO-66 Catalysts

In Figure 2a it can be observed that UiO-66 by itself presented no activity in the oxidation of CO. On the other hand, Cu10/M treated for 1 h at 200 ◦C in He (1) presented activity, reaching a maximum conversion of 60% at the said temperature, while a treatment at 225 ◦C (2) yielded an improvement, reaching conversions of 75%. Meanwhile, a pretreatment at 300 ◦C impaired activity, as observed in solid (3). The diffractograms of these evaluated samples (Figure 2b) indicate that samples (1) and (2) maintained the MOF structure. Nevertheless, a degradation of the MOF took place in sample (3). The absence of definite signals of copper species in all the diffractograms should be highlighted since it indicates their high dispersion. Accordingly, in solids (1) and (2), the catalytic activity was due to copper species highly dispersed in the MOF structure, while, in sample (3), those species were dispersed in an amorphous solid.

**Figure 2.** Cu/UiO-66 solids (Cu/M): (**a**) catalytic evaluation in the CO oxidation; (**b**) X-ray diffraction (XRD) patterns after the reaction.

The change in the pretreatment atmosphere was analyzed in terms of the activity of Cu10/M by combining the pretreatment in inert atmosphere followed by a brief exposure to air for 0.5 h (sample (4)). The as-treated solid was also active (Figure 2a) even though the treatment promoted the total degradation of the MOF (Figure 2b), favoring its evolution to a crystalline phase of tetragonal zirconia (t-Zr), (JCPDS 17-923). The decomposition of the hydrated nitrate salts in the presence of air can generate various oxidizing agents such as HNO3 and NO2, which, added to the treatment in oxygen, can give rise to a hyperoxidizing atmosphere and accelerate the transformation of the MOF into zirconia. This evolution is in agreement with what was reported for the decomposition of UiO-66 in air [27]. A high dispersion of cupric oxide was observed in these solids, characterized by weak signals at 2θ = 35.5◦ (masked by a signal of t-Zr) and 38.5◦, corresponding to the (11−1) and (111) planes of a CuO monoclinic phase (JCPDS 48-1548), respectively.

The TEM image of the synthesized MOF shows nanometric crystals with sizes ranging from 30–100 nm which formed globular aggregates (Figure 3a), making it possible to distinguish the facets of the individual crystals with a polyhedral morphology (Figure 3b) similar to that reported for this MOF but obtained under conventional conditions [20]. Meanwhile, when the said crystals were functionalized with copper following the sequence of impregnation and heat treatment in He to obtain Cu/UiO-66, the porous structure of the MOF was maintained (Figure 3c). The high-resolution (HR) TEM image showed the characteristic aspect of a porous material but no particles could be distinguished inside the MOF porosity (Figure 3d). This highlights the small size of these copper species dispersed in the MOF, consistent with the XRD results.

In brief, a pretreatment of Cu10/M in He for 1 h at 225 ◦C allowed obtaining an active catalyst in the CO oxidation, based on copper species with high dispersion in the MOF structure which were preserved after the tests in reaction. This nanocatalyst represents a new alternative not only for this reaction but also for other reactions demanding a high dispersion of active copper phases and led at relatively low temperatures (<225 ◦C). These reactions could be, for example, the reduction of C–C multiple bonds and carbonyl, the hydroxylation of benzene, the reduction of aromatic nitrocompounds or NOx [28], or the synthesis of methanol from CO and H2 [29].

**Figure 3.** TEM images: (**a**) as-synthesized UiO-66 crystals; (**b**) close view of UiO-66; (**c**) Cu/UiO-66 catalyst with He treatment at 225 ◦C; (**d**) close view of activated (He treated) Cu/UiO-66.

#### *3.2. Derived Cu*/*UiO-66 Catalysts*

#### 3.2.1. Monitoring of the Thermal Transformation of Cu/MOF

Given the structural changes observed after the thermal pretreatments, the transformation of Cu/M was analyzed through temperature-programmed X-ray diffraction (T-XRD) both in an inert atmosphere and in air. In the first case, from 250 ◦C, the MOF underwent a reduction in crystallinity (Figure 4a), totally losing itself at 325 ◦C (red curve). Then, the solid persisted as an amorphous material in which signals at 43.3◦ and 50.1◦ emerged (the latter masked with a t-Zr signal), which corresponded to the (111) and (200) planes of a cubic Cu<sup>0</sup> phase, respectively (JCPDS 4-836). At higher temperatures, these species increased the crystallinity, while the support evolved into a tetragonal zirconia (t-Zr) of low crystallinity. This is consistent with what was discussed above. On the other hand, heat treatments in air showed that the structure of the MOF was more unstable (Figure 4b) losing the crystallinity at 275 ◦C (red curve). In addition, it quickly transformed into a t-Zr system with a high dispersion of copper oxide species. The stable formation of a t-Zr phase from this MOF was attributed to the initial transformation of the small zirconia nuclei from inorganic nodes, which have a low surface energy and facilitate the evolution toward a tetragonal phase instead of a monoclinic (m-Zr) [27]. From about 400 ◦C a small contribution of m-Zr was detected, characterized by strong signals at 2θ = 27.9◦, 31.2◦, 34.1◦, 40.7◦, and 49.1◦, corresponding to the (−111), (111), (200), (−112), and (220) planes, respectively (JCPDS 37-1484). An increase in the proportion of the monoclinic phase in copper-doped zirconia was attributed to copper inclusion in the ZrO2 network, which increased the size of the crystallites, causing a growth in the free surface energy and, thus, promoting the evolution toward m-ZrO2 [30]. This could be, in our case, due to a migration of part of the copper to the zirconia phase in formation during the heat treatment in air.

**Figure 4.** Analyses of X-ray diffraction at programmed temperature (T-XRD) with the Cu/M sample: (**a**) in nitrogen; (**b**) in air.

By T-XRD, it was shown that thermal pretreatments of the Cu/M solid in an inert atmosphere caused a delayed degradation of the MOF toward an amorphous solid in which Cu0 species evolved. Meanwhile, the MOF degradation was accelerated in air with a fast growth of t-Zr phase with a small contribution of m-Zr and with highly dispersed CuO species. Given the potential of solids derived from Cu/M, their catalytic behavior was analyzed.

#### 3.2.2. Catalytic Behavior of Derived Cu/UiO-66 Catalysts

Cu10/M was pretreated in situ in He flow at 225 ◦C for 1 h, and its catalytic curve showed an inflection in the profile starting at 250 ◦C (Figure 5a) due to a fall in the CO conversion. This was caused by a smaller availability of oxygen for the reaction (as observed in the insert in Figure 5a), which was consumed in the MOF degradation. Hence, 250 ◦C is the maximum temperature at which Cu10/M maintained its structure under reaction conditions. From 375 ◦C the oxygen was recovered, and the catalyst was taken up to 400 ◦C for 1 h in reaction, maintaining conversions of 100%; later, the catalyst was cooled and evaluated again. In this case, Cu23/Zr (1), a marked activation was observed (Figure 5a) due to the evolution of the solid to the system of copper oxide dispersed in a developing t-Zr phase (Figure 5b). Since the zirconia mass in the derived solid was around 37.4 wt.% with respect to the initial mass of the MOF, the proportion of copper in these systems was 23 wt.%. It is noticeable that, with this high load, the copper species were highly dispersed in the t-Zr support. Given the good catalytic performance of these solids and taking into account the studies of the transformation of a Cu/M solid into Cu/Zr, a pretreatment of Cu10/M in air was performed at 400 ◦C for 2 h. In this case, Cu23/Zr (2), a remarkable shift of the catalytic curves was observed, reaching total conversion at 225 ◦C (Figure 5a) without extra oxygen consumption due to the presence of a stabilized phase of CuO/ZrO2 (Figure 5b) with a contribution of a m-Zr phase. Compared with classical CuO/ZrO2 catalysts obtained via other techniques such as sol–gel [31] or urea combustion [32], the use of the MOF as a template allowed minimizing the generation of bulk CuO of low interaction with zirconia, which would generate a lower catalytic activity. In contrast to the MOF-derived zirconia (Zr in Figure 5b), the Cu23/Zr (2) solid showed a contribution of the m-Zr phase. When Cu10/M was pretreated in air at 500 ◦C, Cu23/Zr (3), a slight catalytic improvement was observed with respect to the former case, reaching total conversion at 175 ◦C.

**Figure 5.** Cu/UiO-66-derived solids: (**a**) catalytic evaluation; (**b**) XRD patterns after reaction.

However, the contribution of m-Zr in this sample was more evident, which could be related to its better catalytic behavior. In this sense, it was demonstrated that the adsorption capacity of CO in m-Zr supports was higher than in t-Zr, which can be explained by a higher Lewis acidity (Zr4+), as well as a higher Lewis basicity (O2<sup>−</sup>), on the surface of the m-Zr solid [33]. Finally, a pretreatment at 650 ◦C in air, Cu23/Zr (4), did not improve the conversion (Figure 5a), even though it favored the development of the monoclinic phase, probably due to a sintering of the CuO species (Figure 5b).

#### *3.3. Derived Cobalt and Iron-Based UiO-66 Catalysts*

Other non-noble metals of interest that have activity in the CO oxidation reaction are cobalt and iron [13], added to the fact that the latter is a very low-cost metal with high abundance. The decomposition temperature of cobalt and iron nitrate precursors in air was far from the limit of MOF stability (Table 2, Figure S3, Supplementary Materials) while the incorporation of 10 wt.% Co or Fe in the MOF (Co10/M, Fe10/M) decreased the framework stability due to the formed oxides, although the shift was lower than that of Cu10/M. The order of structural stability was as follows: Co/M > Fe/M > Cu/M. The SDTA profiles in air (Figure S3, Supplementary Materials) were very similar to that of Cu10/M, with an endothermic peak due to the evaporation of host molecules and an exothermic one due to structural collapse. Taking into account the similar structure stability of Cu10/M under either He or air atmosphere at temperatures lower than 275 ◦C, Co10/M and Fe10/M solids were pretreated at 250 ◦C in air, and their catalytic behavior was analyzed.

**Table 2.** Maximum decomposition temperatures (Tmax) of the MOF impregnated with cobalt (Co/UiO-66) and iron (Fe/UiO-66), and that of the respective metal precursors (obtained from the dTGA data in air).


<sup>1</sup> Decomposition temperature (◦C) of the maximum in the air dTGA profile.

For the Co10/M solid (Figure S4a, Supplementary Materials), from 200 ◦C onward, conversion increased proportionally with temperature, and, when it was over 325 ◦C, both a conversion fall and an abrupt consumption of oxygen were produced due to the MOF degradation. The activity evolved until total conversion but at a higher temperature than the Cu10/M solid, previously analyzed. Subsequently, this solid was taken to 400 ◦C and was kept 1 h under reaction. When evaluated again, an improvement in activity was observed, Co23/Zr (1). The catalyst consisted of a t-Zr phase evolving with a high dispersion of cobalt species due to the absence of characteristic signals of their oxides (Figure S4b, Supplementary Materials). Since it was observed that an improvement in ZrO2 crystallinity favored the activity, a pretreatment of the Co10/M sample was performed in air but at 400 ◦C for 2 h. This effectively improved the catalytic performance (Figure S4a, Supplementary Materials) due to the formation of a stabilized phase of Co3O4 in a t-Zr of high crystallinity (Figure S4b, Supplementary Materials). This is in agreement with what was reported regarding the formation of this cubic spinel (JPDS 43-1003) on conventional ZrO2 supports [15].

On the other hand, Fe10/M showed less activity (Figure S4a, Supplementary Materials), even over 300 ◦C when the MOF decomposed. After 1 h in reaction at 400 ◦C, the system was evaluated again, and an improvement was observed even though total conversion was not reached in the temperature range analyzed. This solid consisted of an incipiently formed t-Zr with a high dispersion of iron species (Figure S4b, Supplementary Materials). In this case, a remarkable catalytic improvement was also observed when pretreating at 400 ◦C in air. This solid consisted of a highly crystalline t-Zr with a small contribution of m-Zr which dispersed a rhombohedral hematite phase (α-Fe2O3). The previous confinement of the iron precursor in the pores of UiO-66 facilitated, after degradation, the generation

of small Fe2O3 crystals that were quite active in CO oxidation, as already observed for iron oxide crystals [34].

The MOF degradation under reaction conditions of CO oxidation started at 325, 300, and 250 ◦C for Co10/M, Fe10/M, and Cu10/M respectively. Although the thermal stability of the latter was slightly lower, its CO conversion at 250 ◦C was significantly higher (70%) than that of the other two samples (17% and 5%). The best catalytic performance corresponded to the Cu-based sample after the degradation of the MOF structure in air at 400 ◦C for 2 h (Cu23/Zr), which reached 100% CO conversion at 225 ◦C. At that temperature, the conversion for the Co23/Zr solid was 45%, while, for the Fe23/Zr solid, it was only 8%.

*3.4. Bimetallic CuCo*/*UiO-66- and CuFe*/*UiO-66-Derived Nanocatalysts*

#### 3.4.1. CO Oxidation

It was reported that mixed cobalt and copper oxides [35], as well as copper–iron mixed oxides synthesized by low-temperature co-precipitation methods [33,36,37], have synergistic effects on the oxidation of CO; therefore, bimetallic systems were prepared incorporating these metals into UiO-66. The solids obtained by successive impregnation were analyzed, firstly by incorporating the copper precursor followed by their thermal treatment (500 ◦C, air, 2 h); subsequently, a cobalt or iron precursor was added, followed by a final calcination step in air (400 ◦C, 2 h) to obtain Cu16Co7/Zr and Cu16Fe7/Zr systems. The Cu16/Zr, Cu16Co7/Zr, and Cu16Fe7/Zr samples exhibited well-developed t-Zr phases with a small contribution of m-Zr (promoted by the presence of copper as discussed above), adding to a CuO phase with high dispersion (Figure 6). Additionally, a Co3O4 phase was observed in the solid containing cobalt, while, in the iron-containing bimetallic solid, no iron oxide phases were detected. This accounts for the high dispersion of the FeO phases in this solid. The catalytic assays showed that, among these catalysts, an improvement in the activity of Cu16Fe7/Zr was found (Figure 6a). This was due to both the initial presence of a very small proportion of m-Zr phase before the incorporation of iron and the subsequent development of a Fe–Cu synergy among these species due to their intimate contact, favored by the high dispersion achieved by these oxides in the solid, as shown by their XRD patterns.

**Figure 6.** CuCo/Zr and CuFe/Zr nanocatalysts: (**a**) catalytic behavior; (**b**) XRD patterns after reaction.

The catalytically evaluated nanomaterials were analyzed by laser Raman spectroscopy (LRS). These spectra are shown in Figure 7, and the respective spectra of the used monometallic samples are included for comparison. The vibrational signals observed in all of these spectra are consistent with

the crystalline phases identified by XRD. The vibrations of the monoclinic ZrO2 (m-Zr) and tetragonal ZrO2 (t-Zr) phases were present in the Cu16/Zr spectrum (Figure 7), proving the existence of both zirconia phases. The characteristic narrow vibration signals of m-Zr were at 179, 192, 335, 347, 385, 476, 614, and 635 cm−1, with that at 476 cm−<sup>1</sup> being the strongest one [15], while the typical broad signals of t-Zr were at 145, 275, 310, 460, and 650 cm−<sup>1</sup> [27]. In the case of the sample with higher copper content, Cu23/Zr, the monoclinic phase was clearly identified due to its narrow vibration signals (Figure 7). The vibrations of CuO at 280, 335, and 615 cm−<sup>1</sup> [8] overlapped with those of the zirconia, thus hindering their identification. The spectrum of Co23/Zr pointed out the existence of a t-ZrO2 phase and a Co3O4 spinel (485, 523, and 687 cm<sup>−</sup>1) [15], in clear agreement with the XRD results. The absence of m-ZrO2 was evident, inferring that the UiO-66 degradation in the presence of cobalt hampered the formation of this monoclinic phase. The same outcome was obtained when iron was the impregnated metal. The Fe23/Zr spectrum (Figure 7) revealed the presence of just t-ZrO2 and α-Fe2O3 (226, 246, 293, 411, and 610 cm−<sup>1</sup> [36]), and no signals of m-ZrO2 were identified.

**Figure 7.** Laser Raman spectroscopy (LRS) spectra of the mono and bimetallic solids after the CO oxidation reaction.

In the spectra of the bimetallic catalysts, it could be observed that the zirconia signals were mainly associated with t-Zr (Figure 7). In the Cu16Fe7/Zr sample, a high dispersion of iron and copper oxides was achieved given the absence of defined signals of these phases, in line with what was observed by XRD and also confirming the absence of agglomerates after the reaction. This shows that the addition of Fe to the Cu/Zr system generated highly dispersed and stable iron species, since they were kept in that situation in the solid after reaction (Figure 7). This is in contrast with the higher sintering reached in the monometallic Fe/Zr system after reaction. On the other hand, in the spectrum of sample Cu16Co7/Zr, the signals of a developed Co3O4 spinel were dominant (Figure 7), in agreement with XRD observations, showing again that the cobalt species were segregated forming big crystals at the catalyst surface.

The bimetallic sample Cu16Fe7/Zr exhibited a nanoparticle system in intimate contact (Figure 8a), confirming the results of XRD and LRS discussed above, which corresponded to small domains of zirconia phases and oxides of copper and iron. The different crystalline planes of these phases in the individual crystals can be observed (Figure 8a). Figure 8b shows the analysis of the same particle in

dark-field mode and its nature of aggregated nanoparticles was also highlighted. The elementary mapping performed in STEM mode showed a homogeneous distribution of the zirconium (yellow, Figure 8c), iron (green, Figure 8d), and copper (magenta, Figure 8e) phases in the nanoparticle aggregates, confirming the high dispersion and intimate contact between these nano-oxides. The characterizations performed by XRD, LRS, and TEM demonstrated the small particle size reached by the phases of these oxides in intimate contact with each other, explaining the catalytic synergy in this material, as shown in Figure 6.

**Figure 8.** TEM images of Cu16Fe7/Zr catalyst: (**a**) selected area for the energy-dispersive X-ray (EDX) mapping in bright field; (**b**) selected area for the EDX mapping in dark field. Elementary mapping: (**c**) zirconium (yellow); (**d**) iron (green); (**e**) copper (magenta).

#### 3.4.2. CO Oxidation in Hydrogen-Rich Stream (COProx)

Given the high performance of the Cu16/Zr, Cu16Co7/Zr and Cu16Fe7/Zr solids, they were analyzed in the COProx reaction. The conversion curves obtained are presented in Figure 9a and show a volcano-type shape, similar to that found for catalysts based on these types of dispersed oxides in classical supports [14,36]. Initially, the conversion increase may be due to highly dispersed CuO or superficial Cu–O–Zr type sites on the zirconia [14,32], reaching a maximum of 47% (175 ◦C) for the Cu16/Zr sample. The fall in conversion at higher temperatures is probably due to a reduction in copper species dispersed in the hydrogen-rich atmosphere [14]. When comparing this behavior with that of the CO oxidation in air (Figure 6), a shift of the curves toward higher temperatures was observed for both the monometallic and bimetallic catalysts. This differs from that observed for dispersed cupric oxide crystals that exhibited a similar activity in both reactions [38], although this behavior was similar to that

observed for other types of copper–iron mixed oxide catalysts [36]. This change in conversion levels may be due to structural differences between the active sites present and in the reaction mechanism operating under an oxidizing or reductive atmosphere [39]. Meanwhile, the selectivity was greater than 70% up to 120 ◦C, after which it fell sharply (Figure 9b).

**Figure 9.** Catalytic evaluations in the preferential CO oxidation (COProx): (**a**) CO conversion; (**b**) selectivity toward CO2.

The Cu16Co7/Zr catalyst exhibited similar characteristics to those analyzed above, with a 53% maximum conversion at the same temperature. In this sense, Co and Cu oxides would compete for the formation of M–O–Zr clusters over the zirconia support and not for the formation of Cu–Co–O–Zr species that could increase the conversion levels [14]. However, cobalt modulated the activity, given the increase in the selectivity of this system (Figure 9b). At the same time, the Cu16Fe7/Zr solid exhibited a shift of the conversion curve to lower temperature, which was compatible with the higher activity shown by this solid in COTox, previously discussed. Moreover, its selectivity was the highest of all evaluated materials (higher than 85% up to 125 ◦C). This behavior again accounts for the synergy between the oxide phases in this nanocatalyst.

The LRS characterization of the used catalysts after the COProx evaluation is shown in Figure 10. From the analysis of the Cu16/Zr spectra before and after reaction, it can be inferred that the tetragonal and monoclinic phases were present in the support. Nevertheless, an increasing trend of the m-Zr strong signal at 460 cm−<sup>1</sup> was observed at the expense of the t-Zr phase after reaction. This same trend can be observed in the spectra of the Co-containing materials. In the latter sample, a more acute and defined signal of Co3O4 can be additionally seen, from which the increase of the said particles under reducing atmosphere can be inferred. However, for the Cu16Fe7/Zr material, after being under reducing reaction conditions, t-Zr was still the main phase. Moreover, no signals of Cu or Fe oxides were observed, which highlights the high stability of these species in the said reaction atmosphere, which is also consistent with what was observed for this catalyst after CO oxidation in oxidizing atmosphere (Figure 6).

**Figure 10.** LRS spectra of the bimetallic nanocatalysts before and after the COProx reaction.

#### **4. Conclusions**

UiO-66 crystals obtained through a sustainable protocol were used as a dispersion matrix for copper, cobalt, and iron species, allowing the preparation of new nanostructured catalysts active in the oxidation of carbon monoxide. The MOF was modified with the said non-noble metals through simple and classic methods of incipient wetness impregnation, followed by controlled thermal treatments. It was shown that, by precisely tuning the treatment atmosphere (He), temperature (225 ◦C), and time (2h), the solid Cu/UiO-66 could be obtained having 10 wt.% copper species in a very high dispersion inside UiO-66 crystals, maintaining the structure of the MOF. This solid proved to be an active catalyst for the CO oxidation in air streams, representing a novel nanocatalyst not only for this reaction but also for others that demand a high dispersion of active copper species. It was also shown that, if the thermal decomposition treatments of the impregnated metal precursors were carried out in air at temperatures higher than 400 ◦C, the capacity of the MOF to host metallic species could be used to obtain non-noble metal-based catalysts supported on nano-zirconia derived from UiO-66. These controlled treatments in an air atmosphere of Cu, Co, or Fe-impregnated UiO-66 promoted the rapid development of a solid composed of tetragonal (t-Zr) and monoclinic (m-Zr) zirconia, supporting highly dispersed transition metal oxides. The derived bimetallic Cu–Fe/ZrO2 nanocatalyst exhibited the best levels of activity and stability both in the oxidation of CO in air and in the COProx reaction, due to synergic effects of the very close contact between such oxides in the homogeneous nanomaterial.

This study shows the potential of UiO-66 as a dispersion matrix for low-cost and abundant metals such as copper, cobalt, and iron to obtain new sustainable nanocatalysts active in the CO oxidation reaction.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/10/1/165/s1: S1: Incipient wet impregnation procedure; Figure S1: Diffractogram of the synthesized UiO-66 crystals and the pattern simulated from its crystallographic archive (CCDC 733458). The diffractogram of the benzenedicarboxylic acid ligand (BDC) is also included; Figure S2: SDTA of UiO-66(M), Cu/M, and copper precursor: (a) under N2 atmosphere; (b) under air atmosphere; Figure S3: Thermal stability of Co/M, Fe/M solids, and their respective precursors in air: (a) TGA; (b) dTGA; (c) SDTA. Figure S4: Co/Zr and Fe/Zr nanocatalysts: (a) catalytic behavior; (b) XRD patterns after reaction.

**Author Contributions:** Conceptualization, M.A.U. and J.M.Z.; formal analysis, B.M.C.F., M.A.U., and J.M.Z.; funding acquisition, J.M.Z.; investigation, L.A.L., B.M.C.F., and J.M.Z.; methodology, L.A.L.; project administration, J.M.Z.; supervision, M.A.U. and J.M.Z.; writing—original draft, L.A.L., B.M.C.F., M.A.U., and J.M.Z.; writing—review and editing, M.A.U. and J.M.Z. All authors discussed the results and contributed to the manuscript. All authors read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Agencia Nacional de Promoción Científica y Tecnológica (ANPCyT) of Argentina under project PICT 2241 and by the Universidad Nacional del Litoral under project CAI+D 00071 LI.

**Acknowledgments:** We acknowledge the financial support from the Agencia Nacional de Promoción Científica y Tecnológica of Argentina (PICT 2241) and from the Universidad Nacional del Litoral (CAI+D 0071). The institutional support of CONICET is also acknowledged.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Densification-Induced Structure Changes in Basolite MOFs: E**ff**ect on Low-Pressure CH4 Adsorption**

#### **David Ursueguía, Eva Díaz and Salvador Ordóñez \***

Catalysis, Reactors and Control Research Group (CRC), Department of Chemical and Environmental Engineering, University of Oviedo, 33006-Oviedo, Spain; ursueguiadavid@uniovi.es (D.U.); diazfeva@uniovi.es (E.D.)

**\*** Correspondence: sordonez@uniovi.es

Received: 18 April 2020; Accepted: 19 May 2020; Published: 1 June 2020

**Abstract:** Metal-organic frameworks' (MOFs) adsorption potential is significantly reduced by turning the original powder into pellets or granules, a mandatory step for their use at industrial scale. Pelletization is commonly performed by mechanical compression, which often induces the amorphization or pressure-induced phase transformations. The objective of this work is the rigorous study of the impact of mechanical pressure (55.9, 111.8 and 186.3 MPa) onto three commercial materials (Basolite C300, F300 and A100). Phase transformations were determined by powder X-ray diffraction analysis, whereas morphological changes were followed by nitrogen physisorption. Methane adsorption was studied in an atmospheric fixed bed. Significant crystallinity losses were observed, even at low applied pressures (up to 69.9% for Basolite C300), whereas a structural change occurred to Basolite A100 from orthorhombic to monoclinic phases, with a high cell volume reduction (13.7%). Consequently, adsorption capacities for both methane and nitrogen were largely reduced (up to 53.6% for Basolite C300), being related to morphological changes (surface area losses). Likewise, the high concentration of metallic active centers (Basolite C300), the structural breathing (Basolite A100) and the mesopore-induced formation (Basolite F300) smooth the dramatic loss of capacity of these materials.

**Keywords:** coordination polymers; methane storage; XRD crystallinity measurements; mechanical shaping; compaction; VAM; gas separation; MOF pelletization

#### **1. Introduction**

Energy demand estimations for the next decades, mainly due to the global population and industrialization process increments, boost the development of techniques and processes able to make the most of available resources [1]. What is more, the recent COVID-19 pandemic, with millions of people confined to their homes, pointed out even more our domestic reliance on electricity. In most economies that have taken strong confinement measures in response to the coronavirus, electricity demand has declined by around 15%, and the share of variable renewables like wind and solar had become higher than normal [2]. Even when electricity from wind and solar would satisfy the majority of demand, systems need to maintain flexibility in order to be able to ramp up other sources of generation quickly when the pattern of supply shifts, such as when the sun sets. That is, electricity system operators have to constantly balance demand and supply in real time to prevent blackouts, which in recent times occurred mainly during periods of low demand [2]. In this context, natural gas power plants can quickly ramp generation up or down at short notice, providing in this way flexibility, underlining the critical role of gas in the longed-for clean energy transition.

In the natural gas industry, methane purification is a major process for upgrading the streams [3]. In these streams, methane concentration is originally elevated (>90%), so satisfactory results have been reported using fixed-bed adsorption techniques [4,5]. In these cases, the usual practice is to separate

the component that is in lower concentration by adsorption (typically CO2). Adsorbents usually used for this purpose are activated carbons and zeolites, which have good CO2 adsorption yields and their cost is relatively low [6,7]. On the other hand, these techniques present difficulties when methane is the component with the lowest concentration in the stream. Activated carbons and zeolites present low selectivity towards methane with respect to other very similar compounds in molecular size and polarity, like nitrogen [8,9]. This is the case of one of the new alternative methane sources that has begun to be studied in recent years, the recovery of methane from ventilation gases from mining exploitation (VAM). Until now, these streams, which contain typically less than 1% in methane, had been burned directly, with the need of an auxiliary fuel. VAM could be used in order to obtain energy or chemical products, as well as to prevent greenhouse gas emissions into the atmosphere [10,11]. For these operations to be profitable, it is necessary to perform a previous concentration step, whose success depends on the separation capacity of the adsorbent used [12].

Among the materials studied for this purpose, due to its amazing properties, metal-organic frameworks (MOFs) have been shown to present large adsorption and gas separation yields [13,14], these being among the most promising materials in this field. Their high specific surface area (even values up to 6255 m2/g [15]) combined with high total pore volume (1.303 cm3/g [16]) and great porosity (91.1% [17]) are responsible for the large adsorption capabilities, exceeding in the majority of cases other common materials [18]. The materials' structure is made up by an organic ligand, such as imidazole or pyrazine, which links different metal ions or clusters corresponding to each MOF type (copper, aluminium, etc.). These combinations form a cage-like structure that is repeated continuously, conferring on these materials a high degree of crystallinity [19]. Two of the main characteristics of the MOFs are the flexibility in the design, which means a huge variety of organic ligands and metallic ions that allow on-demand materials to be made, and the pore functionalization, presenting high interesting adsorptive and catalytic properties. The possibility of performing a large number of combinations has led to an astonishing number of works related to the synthesis of MOFs suitable for different applications, which include gas storage and separation [20]. For example, in the case of methane separation from other gases, Arami-Niya et al. [21] have tested the zeolitic imidazolate framework (ZIF-7) for the separation of methane from nitrogen, obtaining a selectivity of more than 10 for an equimolar mixture at 303 K. In addition, other authors such as Eyer et al. [22] have studied different materials capable of selectively adsorbing methane from air mixtures, obtaining promising results in the case of HKUST-1, with a selectivity methane/nitrogen of 2.8 and a large gravimetric methane adsorption capacity (171.36 mg/g) at 100 kPa and 196 K. Thus, MOFs have led to satisfactory results at the laboratory level in the case of low-concentrated methane separation from mixtures [23,24], with no experiences being performed at greater scales.

Therefore, most of the experimentation at lab scale and the properties' studies are done on the original powder form, since the most-used techniques for the MOFs synthesis are solvothermal methods, which generally produce powders [25]. Industrial-scale difficulties occur as a result of pressure drops associated with powder-filled beds, high diffusional problems and low density of the materials [26,27]. In order to reduce the pressure drop through the bed, there are techniques for increasing particle size and MOF densification: mechanical, hydraulic or hot pressing, extrusion, solid or emulsion templating, and the use of a polymeric binder [28,29]. In addition, there are also other techniques currently in development, such as the sol-gel monolithic synthesis [30]. Among them, mechanical compression is an inexpensive procedure and avoids the use of additional components like polymeric binders, which may change the physical properties of MOFs [31]. However, compression pelletization could also induce amorphization as well as phase transformations, which could influence also the adsorption capacities of the MOFs [32,33].

In this way, several studies deal with the effect of mechanical compression on hydrogen adsorption for MOF-5 [34,35] and MIL-101 [36] MOFs; as well as on CO2 adsorption [37]. By contrast, there are fewer works related to the influence of MOFs' densification on the methane adsorption. For example, Yuan et al. have studied the behavior of PCN-250 on the methane and nitrogen adsorption at densification pressures up to 300 MPa [38]. General pressure-effects are, added to the loss of gravimetric performance, an increase in the volumetric adsorption capacity, in addition to higher stability in a humid ambient. Typically, these adsorption studies are done at elevated gas pressures since the main objective is to increase material density and volumetric adsorption capacity for meeting gas storage challenges. In this work, the aim is the separation of methane from low-concentration streams, so adsorption studies have been conducted at low pressure (0.1 MPa).

Therefore, the aim of this work is to study, firstly, the pressure-induced changes on the morphology and structure of three of the most common (and commercially available) MOFs, Basolite C300, Basolite F300 and Basolite A100; and, secondly, on the methane and nitrogen adsorption capacity at low pressure (0.1 MPa). The study of the adsorption capacity for methane (component to be recovered in VAM) and nitrogen (majority component in VAM) establishes a benchmark for the use of these commercial materials at industrial scale for obtaining profiting lean emissions as a novel energy source.

#### **2. Materials and Methods**

Basolite C300 [Cu3(C9H3O6)2], Basolite F300 (C9H3FeO6) and Basolite A100 (C8H5AlO5) were manufactured by BASF and supplied by Aldrich (96% mass basis purity, Steinheim, Germany). All three materials were stored in a desiccator in order to avoid its contact with the ambient air. Particles were used in powder form, being the commercial size: Basolite C300 (16 μm, D50), Basolite F300 (5 μm) and Basolite A100 (32 μm, D50). Methane (CH4), nitrogen (N2) and helium (He), with a purity >99.995% mol, were supplied by Air Liquide (Madrid, Spain).

The pelletization method was performed using a hydraulic press (Graseby SPECAC 15.011, Orpington, UK) at compression pressures of 55.9, 111.8 and 186.3 MPa, for 30 s. Starting pressure was selected considering two considerations: ensuring the pelletization of the material to work at the actual conditions, and the lower operating limit of the hydraulic press used for this purpose. The resulting pellets were crushed and sieved in order to obtain powder (<50 μm) to perform all the successive analysis.

Breakthrough adsorption curves were obtained by flowing either CH4 or N2 (60%) diluted in He with a total flowrate of 50 mL/min, 298 K and 0.1 MPa of total pressure in a Micromeritics AutoChem II 2920 apparatus (Norcross, GA, USA) through a fixed bed of each sample (30 mg). The evolution of CH4, N2 and He signals were followed in a Pfeiffer vacuum Omnistar Prisma mass spectrometer (Pfeiffer Vacuum, Asslar, Germany). Adsorption gravimetric capacity was obtained from desorption experiments that were performed in the same apparatus flowing a He stream (20 mL/min and 0.1 MPa) with a temperature ramp of 5 K/min from 298 K to 463 K, recording also the outlet with the mass spectrometer.

The textural characteristics of specific surface area and pore volume were estimated by N2 physisorption at 77 K in a Micromeritics ASAP 2020 surface area and porosity analyzer (Norcross, GA, USA). Physisorption data was processed using Brunauer–Emmett–Teller (BET), Barrett–Joyner–Halenda (BJH) and t-plot approaches for determining surface area, total mesopore volume and total micropore volume, respectively. Variations in average pore size were calculated by assuming pore cylindrical geometry. Scanning electron microscopy (SEM) images were obtained by using a JEOL 6610LV scanning electron microscope (JEOL, Yvelines, France). The samples were coated with gold prior to observation.

The crystallographic structures of the materials were determined by powder X-ray diffraction (PXRD) using a Philips PW 1710 diffractometer (Koninklijke Philips, Amsterdam, The Netherlands), working with the Cu-Kαline (λ = 0.154 nm) in the 2θ range of 5–85◦ at a scanning rate of 2◦/min. Variations in the materials cell structure were verified by the Bragg law. Consequently, variations in lattice parameters of the structures were obtained through the standard equations for cubic, orthorhombic and monoclinic structures.

#### **3. Results**

#### *3.1. Materials Characterization*

Figures 1–3 show the SEM images of powder in the commercial form, as well as the sieved powder after the three pressure treatments. As can be observed, materials in the original form show well-defined particulate shapes, polyedric form in case of Basolite C300, and rounded shape in the case of Basolite F300 and Basolite A100. Size distribution seems to be wide for all of them, being the original size order: Basolite A100 > Basolite C300 > Basolite F300. Pressure increments lead to particle fragmentation, with the subsequent formation of irregular particle agglomerates. At the highest pressure (186.3 MPa) individual particles are practically indistinguishable, which become part of a large individual no-shaped bulk, especially for Basolite C300 and A100.

**Figure 1.** Scanning electron microscope (SEM) images of Basolite C300 (zoom in 50 μm) ((**A**): original, (**B**): 55.9 MPa, (**C**): 111.8 MPa, (**D**): 186.3 MPa).

**Figure 2.** *Cont*.

**Figure 2.** SEM images of Basolite F300 (zoom in 10 μm) ((**A**): original, (**B**): 55.9 MPa, (**C**): 111.8 MPa, (**D**): 186.3 MPa).

**Figure 3.** SEM images of Basolite A100 (zoom in 10 μm) ((**A**): original, (**B**): 55.9 MPa, (**C**): 111.8 MPa, (**D**): 186.3 MPa).

Figure 4 shows the adsorption-desorption isotherms determined by N2 physisorption analysis at 77 K. As shown in the figure, pristine samples exhibit a combination of type I (b) and type II isotherms, according to the International Union of Pure and Applied Chemistry IUPAC. The first zone (up to P/P0 = 0.8) resembles a type I (b) isotherm, with a steep elevation of the adsorbed quantity at very low pressure, and a subsequent maintenance. It is characteristic of microporous materials with wide micropores and possibly narrow mesopores [39]. The second area, up to a P/P0 = 1, shows a more pronounced increase of adsorbate retained, which resembles the final part of a type II isotherm. This indicates the adsorption onto macroporous or non-porous materials in multilayer disposition, which corresponds to the external phase of MOFs [40]. A combination of these two isotherms usually results in a type IV isotherm, but in this case no characteristic hysteresis is observed, and the end of the isotherms is not a plateau [41]. As the densification pressure increases, the isotherms are closer to type I (b), due to the material agglomeration and the consequent loss of external surface availability. In addition, in all the materials a marked reduction is observed in the quantity adsorbed at low P/P0 after pressure compression, indicative of a reduction in the total micropore volume, as it can be seen in the

expanded graph (Figure 4). Micropores are clogged when particles are agglomerated with each other, in agreement with SEM images (Figure 1). The results show a significant effect of pelletization pressure on the morphology of the three MOFs (Table 1). Basolite C300 exhibits the highest BET surface loss (95.4%) at the highest pressure, although even at 55.9 MPa, the BET surface decrease reaches a value of 54.2%, in addition to 69.4% for total pore volume, which rules out the appearance of mesopores in the structure. In agreement, Casco et al. [42] have observed a great structural collapse by applying mechanical pressure (1.5 tons) to this material.

Basolite F300 presents high decreases in specific surface area (up to 93.3%) and micropore volume (96.3%), but lower in mesopore volume (up to 56.8%). The sharp BET decrease at 55.9 MPa shows the ease with which micropores collapse. However, the scarce total mesopores volume variation in the whole pressure range indicates the appearance of narrow mesopores in the structure (Table 1), as it is confirmed by the presence of some hysteresis (H4 type, according to IUPAC) at high P/P0 values, marked in case of 55.9 and 111.8 MPa. For this material, the appearance of two leaps in total mesopore volume value is also remarkable, one between original material and 55.9 MPa and the other between 111.8 and 186.3 MPa. This indicates that the appearance of mesopores is higher at 111.8 MPa, increasing the total mesopore volume even above of the previous applied pressure (55.9 MPa). Despite that, the total pore volume is reduced (0.15 to 0.13 cm3/g) in that pressure increment. This could be attributed to the formation from the voids of interparticular pore volume, as a result of the compaction.

**Figure 4.** *Cont*.

**Figure 4.** Adsorption (•) and desorption (-) N2 isotherms (77 K). Basolite C300 (**A**), Basolite F300 (**B**) and Basolite A100 (**C**). Original sample (Blue), 55.9 MPa (Orange), 111.8 MPa (Grey) and 186.3 MPa (Green). The graphs on the left are zoom of the low pressure zone (up to P/P0 = 0.1) on logarithmic scale.


**Table 1.** Variations of Brunauer–Emmett–Teller (BET) surface area, Barrett–Joyner–Halenda (BJH) total mesopore volume, *t*-plot total micropore volume and average pore size with applied pressure.

Finally, Basolite A100 presents also high specific surface and total pore volume losses, but with a different trend than the others. The first applied pressure (55.9 MPa) provokes the highest BET surface decrease (42.1%). However, the following pressure does not affect greatly either the BET surface or the pore volume (Table 1). In agreement, Ribeiro et al. [43], after application of 62 and 125 MPa to the material, observed null relation between applied pressure and the morphological parameters, obtaining really similar results for both pressures. Finally, at the maximum pressure, a BET surface and total pore volume decrease of 94.6% and 93.3% was reached, respectively.

Figure 5 illustrates the effect of the applied pressure on the crystallinity of both pristine and pressure-modified MOFs. The relative crystallinity is obtained by comparison of the main peak among the series of each material, assuming 100% of crystallinity for the commercial material (Table 2) [44]. In addition, peaks' displacement along the x-axis and the appearance of new ones may mean changes in the material structure (Table 3).

**Figure 5.** Powder X-ray diffraction (PXRD) patterns of the three materials at different applied pressures ((**a**): Basolite C300, (**b**): Basolite F300 and (**c**): Basolite A100). Applied pressures are ordered from top to bottom in increasing order (0, 55.9, 111.8 and 186.3 MPa).


**Table 2.** Relative crystallinity losses associated with applied pressure (referred to the original material).

**Table 3.** Cell total volume and lattice parameters for each structure depending on applied pressure.


The pristine Basolite C300 powder X-ray diffraction (PXRD) pattern shows the typical peaks reported for this material at 2θ = 6.7◦, 9.5◦, 11.65◦, 13.5◦, 19.3◦ and 26◦, in addition to three little peaks at 35.5◦, 38.7◦ and 36.43◦, which indicate some CuO and Cu2O impurities [45,46]. After pressure is applied, the intensity of the peaks decreases progressively, indicative of crystallinity loss (Table 2). As its PXRD pattern is practically coincident with HKUST-1, a face-centered cubic structure is assumed [47], consisting of 16 copper atoms, 8 at the corners, as well as 6 at the center of the cube faces. Low-angle peaks (9.5◦, 11.65◦ and 13.5◦) present (220), (222) and (400) as Miller indices [48]. The net parameter (a) is obtained from the lattice plane of (222), Table 3. As shown, the cell volume remains practically unalterable (maximum variation of 1%), due to the structure rigidity [49]. In agreement, McKellar et al. reported variations of 2.6% for densification pressures of 3.9 GPa [50]. Likewise, the non-appearance of new crystalline peaks indicates that the cubic structure is maintained [51,52]. Therefore, the pressure effect on Basolite C300 consists of crystallinity destruction, in agreement with the BET surface area and pore volume reduction with the pressure, but remaining unaltered the cubic structure of unaltered cells. In agreement, Peng et al. [53] have studied the effect of mechanical pressure (up to 5 tons) onto HKUST-1, indicating a great micropore volume loss (N2 physisorption analysis), in addition to a total collapse of the crystalline structure (PXRD analysis).

For pristine Basolite F300, a characteristic peak at 2θ = 11◦ is observed, despite the low resolution of the pattern as a consequence of the semiamorphous nature of the material and the elevated background values due to the iron fluorescence [54]. In fact, Basolite F300 is a distorted form of crystalline MIL-100(Fe) [55], and possesses a zeolite MTN topology [56]. In this case, the semiamorphous nature of the material just allows observing an increase of the amorphous matter with the pressure (Table 2). As it is a semiamorphous material, crystallinity is slightly reduced in relative terms [57], not reflected in the BET surface, which does not depend on crystallinity and is severely affected by increased pressure. Particle agglomeration causes the collapse of micropores, as it was demonstrated in Figure 4.

Finally, in the case of Basolite A100, this shows a structure practically coincident with MIL-53(Al) MOF, with characteristic peaks at 2θ = 8.8◦, 15.25◦ and 17.75◦ [58]. The original pattern obtained is close to that of the large-pore (lp) phase of MIL-53(Al), which is coincident with an orthorhombic structure [59]. For this result, three different net parameters make up the structure and all the angles are right. Diffracting planes that match the characteristic peaks are (101), (011) and (210) [60,61]. As the applied pressure progresses, the appearance of new peaks around 2θ = 20◦ indicates a movement to the narrow-pore (np) phase [62,63]. In this case, structural changes are high, due to the phase transition, reaching differences up to 13.7% for total cell volume (Table 3). In fact, according to Ghoufi et al. [64], the cell shows a monoclinic structure [65] from, approximately, 53 MPa onwards. As observed, transition to the np structure has an associated reduction of the total cell volume, as well as a decrease of the *a* parameter, in conjunction with increasing trends in the rest of the parameters, including the β angle. This increase in the β angle denotes a flattening on one of its axes [62,66], being these phase changes reversible [59]. Thus, this structure is characterized by its great flexibility. Regarding crystallinity, after the initial loss at the lowest pressure, it remains practically unchanged. The first applied pressure changes the material structure to np phase, which is known for its high resistance to external pressure and flexibility [63], thus maintaining crystallinity for successive applied pressures. The same occurs at 55.9 and 111.8 MPa in the case of BET available surface and total pore volume, which are practically maintained after an abrupt decrease despite the increase of applied pressure.

#### *3.2. Performance Analysis*

The gravimetric adsorption capacity of the samples was calculated from desorption analyses. Figure 6 plots adsorption capacity at different applied pressures as well as the relationship between adsorption capacity and BET specific surface area for each material. Basolite C300 shows a dramatic total decrease of its adsorption capacity with applied pressure, following a progressive trend as in the case of crystallinity and BET surface area. After the first pressure applied, some microporosity is still available, observing decreases of the adsorption capacity of 10.8% for nitrogen and 6.25% for methane. A further pressure increase will led to the total loss of adsorption capacity, BET surface and crystallinity. Additionally, the adsorption capacity/BET surface area ratio is practically linear at low applied pressures, showing certain dependence on BET surface. At the highest pressure, a sharp increase is observed, probably due to the increased role of active metal centres in the adsorption, once the crystalline structure was collapsed.

In the case of Basolite F300, a decreasing trend of the capacity of adsorption with the applied pressure is observed, the downward trend being more pronounced at the highest pressure (loss of 41.3% for N2 and 36.5% for CH4), Figure 6B. Adsorption capacity follows a similar trend to BJH total mesopore volume (Table 1), which could be related to its originally semi-amorphous properties, in which the adsorption capacity is not drastically reduced until a certain pressure limit. The accessibility to metal adsorption sites is maintained due to the appearance of mesopores and, thus, the intracrystalline diffusivity increases. This increase in accessibility is closely related to the smooth downward trend in adsorption capacity, showing an almost linear relationship with the specific available surface.

**Figure 6.** Adsorption capacity for pure methane (blue) and nitrogen (orange) for different applied pressures at 298 K and 0.1 MPa of total pressure (figures on the left), and its relation with BET specific surface area (figures on the right). Basolite C300 ((**A**), -), Basolite F300 ((**B**), ) and Basolite A100 ((**C**), •).

For Basolite A100, a sharp decrease is observed after the first applied pressure, coincident with the asymptotic trend of BET surface area to the last applied pressures (Figure 6). This may be due to the presence of pure CH4 and N2, which provokes the transition to the lp phase at ambient conditions, thus increasing the adsorption capacity by increasing the accessibility to metallic adsorption centers [67]. In fact, from adsorption capacity/BET surface ratio, a constant behavior is observed at the lowest pressures, and a sudden increase at the highest one, due to the drastic reduction of specific surface area after the transition to the lp phase which allows the metallic adsorption centers to have great relevance in the adsorption. Comparing this with other techniques, Finsy et al. [68] have studied the effect of making pellets of MIL-53(Al) using polyvinyl alcohol as a binder. They indicated a reduction of 32% in micropore volume with a pore accessibility reduction of 19% in the best of the cases, which hinders adsorption processes. In fact, it must be pointed out that the presence of a binder can affect the adsorption behavior of the material [69].

From Figure 6 it is observed that the relative adsorption capacity decreases are higher for N2 than for CH4, and it may be related to metal adsorption centers being available, and more selective towards CH4 than N2 [70]. Thus, after surface area and total pore volume reduction, the available active metallic centers play a more relevant role in the selective gas adsorption, especially in Basolite C300 and A100 cases. The influence of the applied pressure in the CH4/N2 selectivity (mass basis) is shown in Figure 7. The increasing slope for Basolite C300 is markedly higher than for the other materials, due to the presence of a higher percentage of metal in its structure (31.5% of copper, vs. 21.2% and 12.9% of iron and aluminum for Basolite F300 and A100, respectively). Therefore, the higher metal content in the structure, the greater the selectivity-increasing trend with applied pressure.

**Figure 7.** Adsorption CH4/N2 selectivity (mass basis) for each material at different applied pressures, at 298 K and 0.1 MPa of total pressure. Basolite C300 (-), Basolite F300 () and Basolite A100 (•).

Breakthrough adsorption curves for CH4 and N2 in a fixed bed are shown in Figure 8. In general, for all the samples, breakthrough times (hence, adsorption capacity) are higher for CH4 than for N2, being attributed to the presence of metallic active adsorption sites and the difference in polarizability of both molecules [70].

**Figure 8.** *Cont*.

**Figure 8.** Adsorption breakthrough curves for CH4 and N2 onto the three MOFs at different applied pressures (Original: blue •, 55.9 MPa: orange , 111.8 MPa: grey , 186.3 MPa: yellow -). Basolite C300 ((**a**): methane, (**b**): nitrogen), Basolite F300 ((**c**): methane, (**d**): nitrogen) and Basolite A100 ((**e**): methane, (**f**): nitrogen). Black lines are used to guide the view.

In the case of Basolite C300, there is a slight difference in the slope between the original material and the others, most obvious in N2 case. As N2 molecular size is lower than CH4 (3.65 and 3.82 Å, respectively), this molecule may penetrate in narrower pores than CH4. Likewise, a decrease in the Knudsen diffusion coefficient led to more inclined curves [71,72]. The Knudsen diffusion coefficient (DK) depends on the pore diameter (dp), since the other parameters are constant for all the experiments. Variations in the Knudsen diffusion coefficient affects directly the breakthrough curve, since it influences adsorbate mass transfer kinetics within the microporous adsorbent.

The reduction in total available specific surface, especially in the micropores zone (low P/P0), indicates that these narrower pores have been totally collapsed by compression (Figure 4). This collapse is common in MOFs when pressure is applied, due to their extraordinary initial porosity [73]. This provokes the following applied pressures to present less-inclined breakthrough curve slopes, but also having less adsorptive capacity, as evidenced by the x-axis order of their breakthrough times (Figure 8). Breakthrough times follow, approximately, the same trend as BET surface area.

In the case of Basolite F300, all the samples, except the original one, show the same slope for breakthrough curves, but in this case the difference is lower than in C300 case. The original sample presents a more inclined breakthrough curve for both adsorbates, which indicates a lower Knudsen diffusion coefficient. Applied pressure modified the pore structure, plugging the micropores, but without reducing greatly the total pore volume by the appearance of mesopores that facilitate the penetration, so the differences in accessibility are softer (Figure 4). The high resemblance between the 55.9 and 111.8 MPa curves (Figure 8) is remarkable and can be related to the no-clear total mesopore volume dependence on pressure (Table 1). The appearance of mesopores in the structure enhance the intracrystalline diffusivity [74], which may be the dominant factor in this case, since the crystallinity is not great affected by mechanical pressure. Dhakshinamoorthy et al. have studied the high relevance of the intracrystalline diffusivity in this material, applied to the case of an oxidation reaction [75].

Finally, in the case of Basolite A100, despite the change from orthorhombic to monoclinic structure and the total cell volume reduction, the presence of pure CH4 and N2 causes the return to the lp phase at ambient conditions, for which the penetration is easier, obtaining a steep curve for all cases due to the structure flexibility [76]. As is also observed, the breakthrough curves of the original material present more resistance than the others, especially for N2. Despite the return to lp phase, the agglomeration provoked a certain irreversible reduction of micropore volume, which increases the Knudsen diffusion coefficient since the average available pore size is higher (Table 1). It is remarkable that differences in CH4 breakthrough times follow almost the same trend as crystallinity, whereas in the N2 case, the trend is similar to specific surface or total pore volume.

#### **4. Conclusions**

Structural and morphological transformations of three MOFs (Basolite C300, Basolite F300 and Basolite A100), as well as CH4 and N2 uptakes variation, were studied after appliance of mechanical pressure to the materials. Basolite C300, a rigid crystalline material, experimented a dramatic and progressive loss of crystallinity, as well as surface area and pore volume, which implies lower adsorption capacity due to its characteristic pores collapse. In the case of Basolite F300, a semiamorphous material, this experienced also a high decrease of surface area and micropore collapse due to agglomeration, but keeping total pore volume due to the appearance of mesopores in the structure. This transformation implies an increase of intracrystalline diffusivity and, then, lower adsorption capacity losses. For Basolite A100, a flexible crystalline MOF, a transformation is observed from orthorhombic disposition to monoclinic structure from 55.9 MPa onwards, in addition to high permanent losses of microporosity due to agglomeration. This structure change is reversible, returning to the lp phase in presence of CH4 and N2 at ambient conditions. This fact increases the accessibility to metallic active centers and an asymptotic decrease of the adsorption capacity is observed. Additionally, the key role of metal active sites in the CH4/N2 selectivity was pointed out. In fact, an increased selectivity for the three MOFs was observed with the applied pressure, decreasing this positive effect in the order: Basolite C300 (Cu, 31.5%) > Basolite F300 (Fe, 21.2%) > Basolite A100 (Al, 12.9%). However, the total gravimetric adsorption capacity has experienced high losses for all of them. Despite that, Basolite C300 stands out above the other two. It has greater adsorption capacity and also a higher metallic content in its structure. In addition, it is able to retain 94% of its adsorption capacity when applying a pressure of 55.9 MPa, enough to increase its particle size and be able to operate in real adsorption stages.

**Author Contributions:** Conceptualization, S.O.; methodology, E.D.; formal analysis, D.U.; investigation, D.U.; data curation, E.D.; writing—original draft preparation, D.U.; writing—review and editing, E.D. and S.O.; supervision, S.O.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Research Fund for Coal and Steel of the European Union, contract 754077 METHENERGY PLUS.

**Acknowledgments:** D. Ursueguía acknowledges the Spanish Government for the FPU fellowship (FPU18/01448). The authors would like to acknowledge the technical support provided by *Servicios Científico-Técnicos* de la Universidad de Oviedo.

**Conflicts of Interest:** The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **BaFe1**−**xCuxO3 Perovskites as Active Phase for Diesel (DPF) and Gasoline Particle Filters (GPF)**

#### **Verónica Torregrosa-Rivero, Carla Moreno-Marcos, Vicente Albaladejo-Fuentes, María-Salvadora Sánchez-Adsuar and María-José Illán-Gómez \***

Carbon Materials and Environment Research Group, Department of Inorganic Chemistry, Faculty of Science, University of Alicante, Av. Alicante s/n, San Vicente del Raspeig, 03690 Alicante, Spain;

vero.torregrosa@ua.es (V.T.-R.); carlamorenomarcos1@gmail.com (C.M.-M.);

vicentealbaladejo@gmail.com (V.A.-F.); dori@ua.es (M.-S.S.-A.)

**\*** Correspondence: illan@ua.es; Tel.: +34-965-903-975

Received: 29 July 2019; Accepted: 29 October 2019; Published: 31 October 2019

**Abstract:** BaFe1−xCuxO3 perovskites (*x* = 0, 0.1, 0.3 and 0.4) have been synthetized, characterized and tested for soot oxidation in both Diesel and Gasoline Direct Injection (GDI) exhaust conditions. The catalysts have been characterized by BET, ICP-OES, SEM-EDX, XRD, XPS, H2-TPR and O2-TPD and the results indicate the incorporation of copper in the perovskite lattice which leads to: (i) the deformation of the initial hexagonal perovskite structure for the catalyst with the lowest copper content (BFC1), (ii) the modification to cubic from hexagonal structure for the high copper content catalysts (BFC3 and BFC4), (iii) the creation of a minority segregated phase, BaOx-CuOx, in the highest copper content catalyst (BFC4), (iv) the rise in the quantity of oxygen vacancies/defects for the catalysts BFC3 and BFC4, and (v) the reduction in the amount of O2 released in the course of the O2-TPD tests as the copper content increases. The BaFe1−xCuxO3 perovskites catalyze both the NO2-assisted diesel soot oxidation (500 ppm NO, 5% O2) and, to a lesser extent, the soot oxidation under fuel cuts GDI operation conditions (1% O2). BFC0 is the most active catalysts as the activity seems to be mainly related with the amount of O2 evolved during an. O2-TPD, which decreases with copper content.

**Keywords:** Iron-based perovskites; copper; NO oxidation to NO2; NO2-assisted diesel soot oxidation; soot oxidation under GDI exhaust conditions

#### **1. Introduction**

The high toxicity of particulate matter (PM) or soot, mainly produced by internal combustion engines, is well established. As in Europe the transport sector generates a 14% of PM2.5 (particulates with a size lesser than 2.5 m, the most hazardous portion), the actual European emissions legislation (Euro 6c) for new passengers vehicles meet or decreases the Particulates Numbers (PN) generated by Gasoline Direct Injection (GDI) to the level corresponding to Diesel engines [1]. GDI engines are considered more effective than diesel engines due to the substantial decrease of fuel intake and CO2 emissions [2]. Consequently, a growth in the US and European market of GDI cars is being observed. To attend the actual European emission legislation, the use of Gasoline Particulate Filter (GPF) is necessary for GDI vehicles, as the Diesel Particulate Filter (DPF) was for Diesel vehicles. In both filters, periodic regeneration is demanded to avoid soot accumulation in the channels of the filter [3–5].

In Diesel engine, as NO2 promotes soot oxidation, a catalyst able to oxidize NO to NO2 is incorporated into the DPF to carry out the NO2-assisted soot oxidation. In fact, several systems (most of them containing Platinum Group Metals, PGM) were developed and implemented in diesel cars to oxidize soot. However, recently, the EU [6] has highlighted that the use of critical raw materials (such as PGM) must be optimized.

Based on the success of DPF in diesel engines, GPF is proposed as a solution for GDI engines. The operating requirements of GPF differ largely from those of the DPF, as NO2 is not present and a very low amount of O2 is available in the GDI exhaust downstream the TWC [7–9]. Thus, active catalysts to oxidize soot in poor (or even null) oxygen conditions must be developed. However, even though it is a challenging issue, the soot oxidation reaction in the severe GDI exhaust requirements (i.e., <10,000 ppm of O2) has been scarcely studied [8–10].

Among the catalysts suggested for O2-soot oxidation [8–18], one of the most interesting are mixed oxides with perovskite structure (ABO3), as their properties can be tailored by selecting the nature of the A and B ions according to the specific needs of the oxidation reaction [19]. Certainly, perovskites are an option with future potential as soot oxidation catalysts in DPF conditions [16–26], as well as for other oxidation reactions such as CO, hydrocarbons and volatile organic compounds [27–33]. In previous papers [26,34], the beneficial result of the incorporation of copper into the structure of BaMnO3 and BaTiO3 perovskites for NO2-assisted diesel soot oxidation was explored. Lately, Hernández et al. [8,9] stated that iron-based perovskites are also appealing as soot oxidation catalysts in GDI exhaust requirements.

Considering this background, and taking into account the promising performance previously featured by a BaFe1−xCuxO3 catalysts series for soot oxidation in the most severe GDI exhaust requirements (regular stoichiometric GDI operation, i.e., 0% O2) [35], the objective of this research is to further study the influence of the partial replacement of iron by copper in the properties of a BaFeO3 perovskite which will define its catalytic performance for soot oxidation. Therefore, BaFe1−xCuxO3 catalysts (*x* = 0, 0.1, 0.3 and 0.4) were synthetized, characterized and tested for soot oxidation in both diesel and "fuel cuts" GDI exhaust conditions (i.e., 1% O2).

#### **2. Materials and Methods**

#### *2.1. Catalyst Preparation*

BaFe1−xCuxO3 catalysts (*x* = 0, 0.1, 0.3, 0.4) have been obtained using a citrate sol-gel method [26]. Ba(CH3COOH)2 (Sigma-Aldrich, 99%), Fe(NO3)2·9H2O (Sigma-Aldrich, 97%) and Cu(NO3)2·3H2O (Panreac, 99%) have been employed as metal precursors. Briefly, a 1M citric acid solution, with a 1:2 molar ratio with respect to barium has been heated to 60 ◦C. The solution pH has been raised to 8.5 with ammonia solution. Subsequently, the corresponding amounts of barium, iron, and copper precursors have been incorporated, and the pH value has been readjusted to 8.5 with ammonia solution. The solution was hold at 65 ◦C during 5 h and later dried at 90 ◦C for 48 h. The dried gel has been calcined at 150 ◦C for 1 h and then, at 850 ◦C6h[26]. Table 1 includes the catalysts nomenclature.

**Table 1.** Molecular composition, specific surface area, copper content, and Goldschmidt tolerance factor (*t)* **\*** values for BaFe1−xCuxO3 catalysts.


\* In the range of experimental detection limit. \*\* Calculated as: *<sup>t</sup>* = RBa+RO <sup>√</sup> 2·(((1−x)·RFe+xRCu)+RO)

.

#### *2.2. Characterization*

To measure the metal content in the samples by Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-OES), a Perkin-Elmer equipment (Optima 4300 DV) has been used. For the analysis, copper was extracted dissolving the samples with magnetic stirring in 8M HCl solution by reflux heating.

An Autosorb-6B instrument (Quantachrome Instruments, Boynton Beach, FL, USA) was used to determine, by N2 adsorption at −196 ◦C, the Brunauer Emmet Teller (BET) surface area of the samples, which were previously degasified at 250 ◦C for 4 h.

X-ray diffraction (XRD) tests were performed between 20–80◦ 2θ angles with a step rate of 1.5◦/2 min and using CuKα (0.15418 nm) radiation in a Bruker D8-Advance device. The Rietveld analysis of XRD data was developed with the Automatic Rietveld Refinement (HIGHScore Plus from PANalytical program).

A ZEISS Merlin VP Compact Field Emission Scanning Electron Microscopy (FESEM) equipment (Quantax 400 from Bruker, Berlin, Germany) was employed to analyze the morphology of the catalysts and to determine the elemental composition of the catalysts (by Energy Dispersive X-Ray analysis, EDX).

X-Ray Photoelectron Spectroscopy (XPS) was used to obtain the surface composition. To register the XPS spectra, a K-Alpha photoelectron spectrometer by Thermo-Scientific, with an Al Kα (1486.6 eV) radiation source, was used in the following conditions: 5 <sup>×</sup> 10−<sup>10</sup> mbar pressure in the chamber and setting the C1s transition at 284.6 eV. The binding energy (BE) and kinetic energy (KE) values were then determined with the peak-fit software of the spectrophotometer, to regulate the BE and KE scales.

Reducibility of the catalysts was evaluated by Temperature Programmed Reduction with H2 (H2-TPR). The experiments were developed in a Pulse Chemisorb 2705 device from Micromeritics fitted with a Thermal Conductivity Detector (TCD to find out the outlet gas composition changes. 20 mg of the sample was heated at 10 ◦C/min from room temperature to 1000 ◦C in 5% H2/Ar atmosphere (40 mL/min, *Pt* = 1 atm). The H2 consumption amount was determined using a CuO sample supplied by Micromeritics.

#### *2.3. Activity Tests*

The catalytic activity for NO to NO2 oxidation and NO2-assisted diesel soot oxidation was established by Temperature Programmed Reaction (NOx-TPR) using of a gas mixture composed of 500 ppm NOx and 5% O2, balanced with N2 (500 mL/min gas flow). For NO oxidation experiments, 80 mg of the catalyst were mixed with SiC, in a 1:4 mass ratio, and warmed from 25 to 800 ◦C, at 10 ◦C/min, in a quartz fixed-bed reactor. The activity for diesel soot oxidation was evaluated adding 20 mg of Printex U from Degussa (employed as surrogated soot, which represents the least reactive fraction of particulate matter [8–10,16,26,34–38]), in loose contact with the catalyst. For the catalyst with the highest activity, isothermal soot oxidation reactions at 450 ◦C were also performed. The gas composition was monitored by specific Non-dispersive Infrared Ultraviolet NDIR-UV gas analyzers for NO, NO2, CO, CO2, and O2 (Rosemount Analytical Model BINOS 1001, 1004 and 100). The NO2 generation, soot conversion and CO2 selectivity percentages were calculated using Equations (1), (2), and (3), respectively:

$$\text{NO}\_2(\%) = \frac{\text{NO}\_{2,\text{out}}}{\text{NO}\_{\text{x,out}}} \times 100 \tag{1}$$

$$\text{Soot conversion}(\%) = \frac{\sum\_{0}^{t} \text{CO}\_{2} + \text{CO}}{(\text{CO}\_{2} + \text{CO})\_{\text{total}}} \times 100\tag{2}$$

$$\text{CO}\_2\text{ velocity } (\%) = \frac{\text{CO}\_{2,\text{ total}}}{(\text{CO}\_2 + \text{CO})\_{\text{total}}} \times 100 \tag{3}$$

where NO2,out and NOx,out are the NO2 and NOx concentrations determined at the reactor exit, *t* 0 CO2 + CO is the quantity of CO2 and CO evolved at a time *t*, and and (CO2 + CO)total are the CO and CO2 + CO evolved during all the experiment time.

To determine the catalysts performance for soot oxidation in GDI exhaust conditions, a gas mixture with 1% O2 in He was used as it is the typical O2 concentration at the turbine-GDI engine exit, i.e., upstream the TWC [8], but also because it simulates the fuel cut GDI operation conditions (<20% O2) [7]. These experiments were developed as Temperature Programmed Reactions (6 ◦C/min from

150 ◦C temperature till 900 ◦C, 500 mL/min gas flow) in a quartz fixed-bed reactor using 80 mg of the catalysts and 20 mg of Printex-U (1:4 soot/catalyst ratio in loose contact) mixed with SiC. A Gas Chromatograph (Hewlet Packard 8690) with two packed columns (Porapak Q and Molecular Sieve 5a) connected to a Thermal Conductivity Detector (TCD) was used for the measure of the gas composition. Previous to the soot oxidation reaction, the catalysts were preheated in the reaction mixture (1% O2 in He) at 150 ◦C during 1 hour. The soot conversion and CO2 selectivity percentages were calculated using Equations (2) and (3), respectively.

#### **3. Results and Discussion**

#### *3.1. Characterization of the Fresh Catalysts*

#### 3.1.1. Chemical, Morphological, and Structural Properties

Table 1 features the real copper content and BET surface area of the BaFe1−xCuxO3 (*x* = 0, 0.1, 0.3, 0.4) perovskites obtained by ICP-OES and N2 adsorption, respectively. All the BaFe1−xCuxO3 catalysts present a very low surface area, as it corresponds to mixed oxides with perovskite structure [19]. The data of the real copper content (very close to the nominal corresponding to the stoichiometric formula) reveal that nearly all the copper used in the synthesis appears in the catalysts. Concerning morphology, FESEM images (Figure S1 in Supplementary Information) show that catalysts are formed by highly agglomerated irregular grains with a size in the range of micrometer. The presence of a low amount of copper (BFC1 and BFC3) does not significantly change the morphology of the bare perovskite; however, for the catalyst with highest copper content (BFC4), a different type of grains is detected which could correspond to a new phase. The EDX data (see Table S1 in Supplementary Information) reveal an identical atomic percentage of Ba and Fe for BFC0, as expected according to perovskite composition (BaFeO3). However, for BFC4, in addition to the presence of Cu, larger atomic percentages of Ba and O are detected. This fact supports the existence of a new phase composed by barium, oxygen, and copper in the surface of this catalyst.

Figure 1 features the XRD profiles, showing (as expected according to the calculated t values shown in Table 1) a perovskite structure as an almost unique crystalline phase for all catalysts. Additionally, a Fe(III) and Fe(IV) mixed-oxide with triclinic structure is identified as a minority phase for BFC0 and BFC1, while, for BFC4, a BaOx-CuOx phase (with a suggested stoichiometry of BaCuO2) appears. This oxide could correspond to the different type of grains observed by FESEM for BFC4 catalyst (Figure S1d in Supplementary Information) and justifies the EDX data (Table S1 in Supplementary Information) for this catalyst.

**Figure 1.** XRD patterns for fresh BaFe1−xCuxO3 catalysts.

For BFC0 and BFC1 catalysts, the diffraction peaks are assigned to a hexagonal perovskite structure; however, for BFC3 and BFC4, the peaks are consistent with a cubic structure. These results agree

with the decrease in the *t* parameter values (Table 1), which becoming closer to 1 (corresponding to an ideal cubic structure) as the copper content increases. This structural modification (which has been verified by the Rietveld analysis presented in Figure 2 was previously noticed for other barium-based perovskites [26,34,38] and also for Sn- doped BaFeO3 perovskites [39], and supports that Cu has been introduced into the perovskite lattice. Concerning BFC1, the reduction in the intensity of the main perovskite peak (at. 31.5◦) evidences that copper has been inserted into the perovskite structure [26,34,38]. Moreover, except for the catalyst with a highest copper content (BFC4), peaks corresponding to a copper segregated phase are not clearly identified, revealing that copper species are not segregated or, if they are, they would present a size under the detection limit of XRD. Finally, for BFC4, the presence of the BaOx-CuOx phase as minority segregated phase shows a limit in the amount of copper introduced into the perovskite framework [26,34,35,38].

**Figure 2.** *Cont.*

**Figure 2.** Rietveld analysis for (minority phase not included in the analysis): (**a**) BFC0: in red the original XRD pattern, in blue the Rietveld simulation corresponding to hexagonal perovskite structure and in green the residual data corresponding to triclinic structure; (**b**) BFC1: in red the original XRD pattern, in blue the Rietveld simulation corresponding to hexagonal perovskite structure and in green the residual data corresponding to triclinic structure; (**c**) BFC3: in red the original XRD pattern, in blue the Rietveld simulation corresponding to cubic perovskite structure and in green the residual data corresponding to BaCuO2; (**d**) BFC4: in red the original XRD pattern, in blue the Rietveld simulation corresponding to cubic perovskite structure and in green the residual data corresponding to BaCuO2.

The average crystal size for the catalyst has been determined from the Full Width at Half Maximum (FWHM) of the main perovskite XRD peak (in hexagonal or cubic structure) applying the Scherrer equation [40]; data are included in Table 2. The average crystal size is smaller for the catalyst containing copper with hexagonal structure (BFC1) than for the bare perovskite (BFC0). On the contrary, for catalysts with cubic structure (BFC3 and BFC4), the average crystal size increases with the copper content. The lattice parameter for hexagonal (*a* and *c*) and cubic (*a*) perovskites have also been estimated from XRD data (Table 2). As the average crystal size, the decrease in *a* and *c* values is observed in the presence of copper for the catalyst with hexagonal structure (BFC1), confirming that copper has been inserted into the lattice. However, as the ionic radii of copper (as Cu2<sup>+</sup>, 0.73 **Å**) is larger than the Fe3<sup>+</sup> ionic radii (0.65 **Å**) or Fe4<sup>+</sup> (0.59 **Å**), an increase in the lattice parameters would be expected if this was the unique factor affecting the values. Nevertheless, it has been reported that a modification in the amount of the oxygen vacancies affects the lattice parameter [41], thus, it seems that the amount of oxygen vacancies is also being affected by copper incorporation into the BaFeO3 perovskite framework. For cubic perovskites (BFC3 and BFC4), the lattice parameter is almost constant but larger than the corresponding to a reference BaFeO3 with cubic structure (4.012 **Å**) [39], again supporting that copper has been inserted into the lattice.


**Table 2.** XRD characterization data of BaFe1−xCuxO3 catalysts.

\* Calculated using the main XRD perovskite peak.

Summarizing, from XRD data, it can be concluded that copper is inserted into the perovskite structure causing: (i) the distortion of the original hexagonal perovskite structure for the catalyst with the lowest copper content (BFC1), (ii) the modification from hexagonal to cubic structure for the catalysts with high copper content (BFC3 and BFC4), (iii) the formation of a BaOx-CuOx oxide as minority segregated phase for BFC4 catalyst, and iv) a possible increase in the amount of oxygen vacancies/defects.

#### 3.1.2. Surface Properties

XPS analysis provides data about the surface composition of the BaFe1−xCuxO3 perovskite catalysts. The XPS profiles corresponding to the Cu 2p3/<sup>2</sup> transition are presented in Figure 3. Reduced copper species, such as metallic copper or Cu2O, usually appear at a binding energy (BE) close to 933 eV, while, for Cu(II) species, the Cu 2p3/<sup>2</sup> transition appears above 933 eV [36,42–44]. In Figure 3, the BE maximum of the main XPS band appears slightly above 933 eV in the three catalysts containing copper, suggesting the presence of Cu(II) species. Moreover, Cu(I) and Cu(II) species can be distinguished by the presence of a satellite peak at 942–945 eV, due to an electron transfer from Cu 2p3/<sup>2</sup> to 3d free level in Cu(II) [45]. The existence of the satellite peak for the three copper-content catalysts, that reveals the presence of Cu(II) species [45], confirms that copper is present as Cu(II) species. Additionally, based on the use of Auger data (Cu LMM) [45], the existence of Cu(II) species has been verified as reveals the Wagner (chemical state) plot shown in Figure S2 (Supplementary information). The deconvolution of the normalized Cu 2p3/<sup>2</sup> bands reveals two contributions with maxima at around 933 eV and 935 eV, which seem to correspond to two different Cu(II) species [42–44]: (i) the band at lower BE, assignable to copper species with a weak electronic interaction with perovskite, that is, to CuO species located on the surface (CuS) and (ii) the band at higher BE, corresponding to copper species with a strong electronic interaction with perovskite (CuL), i.e., copper inserted in the lattice, near the surface. As the percentage of the area for the XPS band at 935 eV (CuL band) increases with the copper content from 26% to 33%, it seems that the presence of copper with a strong electronic interaction with perovskite is favored as copper content increases. However, a slight decrease of this value is observed for the BFC4 catalyst with respect to BFC3 (33% versus 35%), confirming that a limit for the copper insertion has been achieved. In fact, a comparison between the Cu/Cu+Fe+Ba ratio calculated by XPS and the corresponding nominal ratio (both data included in Table 3) confirms that copper has been inserted into the perovskite structure, as the XPS ratio are lower (for BFC1 and BFC3) or similar than (for BCF4) the nominal ratio [25,34–38]. It is remarkable that the smallest difference between these two values is presented by BFC4 catalyst, supporting, again, the limit in the copper insertion. Therefore, the copper

which is not introduced into the lattice has to be dispersed on the surface forming the BaOx-CuOx phase, which was detected by XRD and EDX, as copper content is higher for BFC4 (Table 1).

**Figure 3.** XPS spectra for Cu2p3/<sup>2</sup> transition.

**Table 3.** XPS characterization data of BaFe1−xCuxO3 catalysts.


Figure 4 features the XPS spectra of the Fe 2p3/<sup>2</sup> for BaFe1−xCuxO3 catalysts and the corresponding to a Fe2O3 commercial sample use as reference. The maximum of the main XPS band for the four catalysts does not appear at exactly the same (BE) value than the corresponding to the reference suggesting the presence of Fe species with a different oxidation state or with the same oxidation state but in different proportion. The deconvolution of the main band shows two significant contributions at around 709 eV and 711 eV. Even though the identification of iron oxidation states by XPS is very difficult [46], according to literature [39,46–50], the first peak corresponds to Fe(III) species, and the second one could be assigned to Fe(IV) species [48–50]. It has been established that the position of the satellite peak is the key finger to detect the oxidation state of Fe [46,48]. Thus, the shake-up peak observed at 717 eV (which corresponds to the satellite peak of Fe(III)) supports the existence of this oxidation state [39,46,48,51]. However, the presence of Fe(IV) seems not to be supported by the XPS data, as the high BE peak at approximately 711 eV is not always unequivocally assigned to this oxidation state [46,48]. Thus, more evidence from other characterization techniques is needed to assume that Fe(IV) exits. The TPR-H2 results (see below) indicate that Fe(IV) and Fe(III) oxidation states co-exist in

the BaFe1−xCuxO3 catalysts, as it is observed that the experimental H2 consumption is in between the nominal (calculated) H2 consumption expected, considering that iron as Fe(III) or Fe(IV) is reduced to Fe(II). The presence of Fe(IV) in BaFe1−xCuxO3 catalysts is additionally supported by the well-known stabilization of high oxidation state for B cation, as Fe(IV), in perovskites [19,39,48–50]. On the basis of the BaFeO3 stoichiometric formula, Fe(IV) must be the oxidation state for Fe in the perovskite, and, in the presence of copper, a rise in the Fe(IV) amount and /or the generation of additional oxygen vacancies into the perovskite structure would be expected to compensate the deficiency of positive charge due to the partial iron substitution [19]. In fact, the decrease in the lattice parameter observed by XRD for BFC1 with respect to BFC0 (Table 2) suggests an increase in the Fe(IV), which presents a lower ionic ratio that Fe(III). However, for BFC2 and BFC3, the lattice parameter (Table 2) increases revealing that the amount of Fe(IV) cannot be higher; thus, the generation of additional oxygen vacancies should be observed to balance the positive charge deficiency due to the increase of the copper content in the catalyst. This larger amount of oxygen vacancies has to cause the lattice expansion detected [41].

**Figure 4.** XPS spectra for Fe2p3/<sup>2</sup> transition.

Figure 5 presents the XPS spectra of the O1s transition for all catalysts, where three contributions are usually observed [36,42–44]: (i) at low BE (around 528 eV), corresponding to lattice oxygen (OL) in metal oxides, (ii) at intermediate BE (between 529 and 531 eV), assigned to adsorbed oxygen species such as, O2 <sup>−</sup>2, surface CO3 <sup>−</sup>2, and/or OH<sup>−</sup> groups, and (iii) at high BE (533 eV approximately) due to oxygen in adsorbed water [52–55]. The intensity of the bands is modified in the presence of copper revealing changes in the amount of oxygen species on the catalysts surface. The values of OL/Cu+Ti+Ba ratio in Table 3 (determined from the peak area of OL, Fe2p<sup>3</sup>/2, Ba3d3/2, and Cu2p3/<sup>2</sup> transitions) is higher for BFC1 than for the bare BFC0 perovskite, which means a lower amount of surface oxygen vacancies. This fact supports that the oxidation of Fe(III) to Fe(IV) occurs in the BFC1 perovskite to compensate the positive charge deficiency due to copper incorporation. However, for BFC3 and BFC4 catalysts, the lower OL/Cu+Ti+Ba ratio with respect to the nominal value confirms the generation of additional oxygen vacancies to balance the positive charge deficiency due to partial iron substitution by copper. Additionally, these results justify the change in the values of lattice parameters observed, that is: (i) the lower lattice parameters values for BFC1 catalyst with respect to BFC0 (Table 2) are due to the decrease in the amount of oxygen vacancies (Table 3), as, for this catalysts, the oxidation of Fe(III) to Fe(IV) takes place and (ii) the larger values for BFC3 and BFC4 (Table 2) are due to the rise in the amount of oxygen vacancies with respect to BFC0 (Table 3).

**Figure 5.** XPS spectra for O 1s transition.

#### 3.1.3. Redox Properties

Reducibility and redox properties of the fresh BaFe1−xCuxO3 catalysts were analyzed by Temperature Programmed Reduction with H2 (H2-TPR), which are the H2 consumption profiles shown in Figure 6. In Figure 7, the nominal (calculated) H2 consumption (mL of H2 per gram of catalysts) expected considering that iron, as Fe(III) or Fe(IV) in the perovskite, is reduced to Fe(II), is compared with the experimental H2 consumption determined from the H2-TPR profiles (Figure 6). It is observed that the experimental H2 consumption is in between both nominal values revealing that Fe(IV) and Fe(III) oxidation states co-exist in the BaFe1−xCuxO3 catalysts. Furthermore, the experimental H2 consumption data indicates that the amount of Fe(IV) increases in the presence of copper.

In the complex H2 consumption profiles shown in Figure 6, three regions can be established [56–58]:


c) At high temperatures (T > 700 ◦C), broad TCD signals assigned to the reduction of Fe(II) to Fe(0) (causing the consequent destruction of the perovskite structure) could be found [56–58]. Nevertheless, the XRD data for catalysts after H2-TPR (not shown) reveal that the perovskite structure is still present, thus, the reduction to Fe(0) is not taking place and, consequently, H2 consumption is hardly observed at T > 700 ◦C. Therefore, the most relevant information related to the redox properties of the BaFe1−xCuxO3 catalysts is located at T < 700 ◦C.

BFC0 profile shows two peaks at temperature lower than 700 ◦C: a broad peak with maximum at ca 300 ◦C and a more defined peak with a maximum at around 670 ◦C. The H2 consumption detected at temperature lower than 300 ◦C is usually related to the presence of Fe(IV) [56–58], supporting the existence of this oxidation state. The second H2 consumption peak, with a maximum at 670 ◦C, corresponds to the reduction of Fe(III) to Fe(II) and to desorption/reduction of oxygen surface species formed on oxygen vacancies (α'-oxygen) [34].

**Figure 6.** H2-TPR profiles for BaFe1−xCuxO3 catalysts.

**Figure 7.** H2 consumption (mL/g of catalyst).

In the H2 consumption profile of BFC1 catalyst, a broad peak with two maxima, at approximately 350 ◦C and 450 ◦C, is identified. The first maximum is ascribed to the Cu(II) to Cu(0) reduction (appearing at lower temperature than the CuO used as a reference [38]) and also to the consumption due to the partial Fe(IV) and Fe(III) reduction to Fe(III) and Fe(II), respectively. The second maximum

of this broad peak at 450 ◦C seems to correspond to: i) the reduction of Fe(III) to Fe(II), taking place at lower temperature than for the BaFeO3 catalyst, due to the presence of reduced copper [26] and ii) the desorption/reduction of strongly bonded oxygen species ('-oxygen) [34].

In the H2-TPR profile of the BFC3 catalyst, a broad peak between 300 and 500 ◦C is detected with a well-defined maximum at 320 ◦C, followed by a shoulder around 380 ◦C. The first maximum corresponds to the reduction of Cu(II) to Cu(0) and it is better defined than the corresponding to BFC1 due to the higher copper content. As for BFC1, the low H2 consumption at T < 300 ◦C confirms the presence of Fe(IV). The shoulder at 380 ◦C has to be related with the reduction of Fe(III) to Fe(II) that seems to take place at lower temperature than for BFC1. This fact supports that the formation of metallic copper (which is more easily reduced than iron) promotes the reduction of Fe(III) to Fe(II), as was previously observed for the reduction of manganese species in BaMn1−xCuxO3 catalysts series [26].

Concerning BFC4, a sharp H2 consumption peak with a maximum at 315 ◦C is found, followed by a low intensity peak with a maximum at around 475 ◦C. The presence of this well- defined peak, which is ascribed to Cu(II) to Cu(0) reduction, confirms the existence of copper oxide (II) species [34]. In fact, for this catalyst, BaOx-CuOx oxide has been detected by XRD and FESEM, thus, the sharp peak corresponds to the reduction of this copper oxide. The second peak has to be due to the Fe(III) reduction to Fe(II) that, for BFC1 and BFC3 catalysts, takes place at lower temperature that for BFC0.

After the analysis of the H2-TPR profiles for three catalysts, it can be concluded that the Fe(III) reduction to Fe(II) takes place at similar temperature for BFC1 and BFC4 (450 ◦C and 475 ◦C, respectively); this happens at lower a temperature (380 ◦C) for the BFC3 catalyst, probably due to its higher content of lattice copper (see Table 3).

Concluding, H2-TPR results indicate the co-existence of Fe(III) and Fe(IV) and suggest that copper incorporation promotes the reduction of Fe(III) to Fe(II).

#### 3.1.4. O2 Release During Heat-Treatment in He (O2-TPD)

In the O2 profiles evolved by perovskite mixed oxides during a heat treatment in He (O2-TPD), three regions are usually observed [26,39,59–68]. The lower temperature region, at T < 400 ◦C corresponds to weakly chemisorbed oxygen upon surface-oxygen vacancies (denoted as oxygen). The intermediate region, between 400 ◦C and 700 ◦C, is ascribed to near-surface oxygen associated to lattice defects such as dislocations and grains frontiers (designed as α' oxygen). Therefore, the presence of α and α'-oxygen is directly linked with the presence of surface vacancies/defects of oxygen in the structure [64–66]. Finally, the oxygen evolved at temperature higher than 700 ◦C, named β oxygen, is generally related with the lattice oxygen (which comes from the reduction of B cation (Fe in this case) of the perovskite [66]) and it is related with the oxygen mobility and with the inner bulk oxygen vacancies.

Figure 8 shows the O2 profiles for the BaFe1−xCuxO3 catalysts. The O2–TPD profiles show that α and α'-oxygen are mainly evolved by most of the BaFe1−xCuxO3 catalysts [8,9]. BFC0, BFC3, and BFC4 exhibit a higher oxygen signals than BFC1 and, therefore, higher quantity of surface oxygen vacancies, agreeing with the XPS results (lower OL/Cu+Ti+Ba ratio). Regarding α'-oxygen, BFC1 presents the highest signal, which evidences the great structure distortion (as exhibited by XRD) promoted by a small Cu incorporation. The total quantity of O2, calculated from the area under the O2 profiles, diminishes as copper content grows: 424 μmol/g cat (BFC0) > 333 μmol/g cat (BFC1) > 282 μmol/g cat (BFC3) > 275 μmol/g cat (BFC4). Thus, as it has been previously published [39], the addition of a dopant seems to stabilize the oxygen bonded to Fe and leads to a decrease in the desorbed O2. For BFC4, a grown in the β oxygen has been detected, probably related to the structural modification and the presence of the BaOx-CuOx phase identified by XRD.

**Figure 8.** O2-TPD profiles for BaFe1−xCuxO3 catalysts.

The phase composition of BaFe1−xCuxO3 catalysts after the O2-TPD has been determined by XRD (Figure S3 in Supplementary Information). For BFC0 and BFC1 catalysts, the hexagonal perovskite structure is replaced by a monoclinic BaFeO2.5 phase (with ordered oxygen vacancies) after losing a fraction of the lattice oxygen. On the contrary, BFC3 and BFC4 catalysts preserve the cubic perovskite structure after O2 release. This founding agrees with the conclusions of Huang et al. [39], who pointed out the increase in the structure stability due to the presence of a dopant (copper in our case). Note that the most stable catalysts (BFC3 and BFC4) are those with ideal (cubic) perovskite structure. The higher structural stability in the presence of copper could be relevant for catalytic applications at high temperature.

#### *3.2. Catalytic Activity*

#### 3.2.1. NO2 Generation and Diesel Soot Oxidation

Figure 9 shows the NO2 generation profiles obtained in TPR conditions for BaFe1−xCuxO3 catalysts including, as reference, the thermodynamic equilibrium profile. As observed for other perovskite-based catalysts [26,34–38,64,67], the thermodynamic equilibrium limits the NO2 percentage at T > 500 ◦C. All catalysts accelerate the NO to NO2 oxidation at temperature lower than 500 ◦C, being the copper-free catalyst (BFC0) the most active. In general terms, the NO2 generation follows the same sequence than the amount of oxygen evolved during O2-TPD, except for BFC4 catalyst. Note that the two catalysts evolving large amount of and '-oxygen, that is, BFC0 and BFC1, are also the catalysts generating more NO2 at low temperature (T < 300 ◦C). This is in agreement with the relationship found by Onrubia et al. [64] between the amount of and ' oxygen evolved by the catalysts (Sr-doped LaBO3 (B = Mn or Co perovskites) and the activity for the NO to NO2 oxidation. Note that BFC4 shows a slightly higher NO2 generation capacity than BFC3, which has to be related with the presence of copper species on the surface (BaOx-CuOx) that also catalyze the NO2 production [34].

**Figure 9.** NO2 generation profiles in TPR conditions s for BaFe1−xCuxO3 catalysts.

To evaluate the activity of the catalysts for NO2-assisted diesel soot oxidation, Temperature Programmed Reactions in a NO/O2 atmosphere (see Experimental Section for details) were carried out, and the TPR-NOx soot conversion profiles (calculated based on the amount of CO and CO2 evolved) are featured in Figure 10. Relevant data, such as the ignition temperature (T5%), the temperature required to reach 50% of soot conversion (T50%), and the selectivity to CO2, are included in Table 4. It can be concluded that all the catalysts shift the soot conversion profiles to lower temperatures compared to the uncatalyzed reaction (blank corresponding to bare soot) and, consequently, the T5% and the T50% are lower. In agreement with the NO2 profiles (Figure 9), BFC0 is the most active catalyst for diesel soot oxidation as the addition of copper decreases the catalyst activity for soot conversion. Moreover, for BFC0 the T50% value is close to 500 ◦C, thus, this perovskite could be used as potential catalyst for the soot removal from diesel engine exhaust [69]. The decrease in the activity for soot oxidation after the addition of a dopant (copper in our catalysts) was also observed by Huang et al. [39] for Ag-doped LaFeO3 catalysts. These authors related the lower activity of Ag-perovskites for soot oxidation with the reduction in the amount of surface oxygen vacancies due to the anchorage of Ag nanoparticles. Furthermore, the reaction rate for methane combustion of a series of oxygen deficient SrFeO3 perovskites was related with the quantity of oxygen vacancies in the structure [58]. In fact, a relationship between soot oxidation performance and oxygen vacancies has been published [70]. Thus, in BaFe1−xCuxO3 catalysts, the decrease in the total amount of O2 evolved as copper content increases apparently leads to a decrease in the activity for both NO to NO2 and soot oxidation. Note that BFC4 does not match this trend as it shows the lowest T5% y T50% values among the catalysts containing copper (BFC1, BFC3, and BFC4). This catalyst presents the highest fraction of surface copper species, which also catalyzes the NO2/soot oxidation reaction [26,34,36], and hence, improves its catalytic performance. Therefore, the activity for NO2 generation and the amount of surface copper species seem to determine the catalytic performance. Thus, the highest NO2 generation capacity of BFC0 catalyst seems to justify its highest soot oxidation activity, while it is the largest fraction of surface copper species present in BFC4, which seems to justify its higher soot oxidation activity compared to BFC3.

**Figure 10.** TPR soot conversion profiles in NOx for BaFe1−xCuxO3 catalysts.

**Table 4.** Data for NO2-assisted diesel soot oxidation in TPR conditions BaFe1−xCuxO3 catalysts.


Additionally, the data in Table 4 reveal that, as could be expected [26,34,36,37], all the catalysts show a higher CO2 selectivity than the uncatalyzed reaction (bare soot), with BFC4 being the most selective. Thus, CO2 selectivity increases with the amount of surface copper species (Table 3) as this metal is a well-known catalyst for CO to CO2 oxidation.

Due to its high activity, the performance of the BFC0 catalyst was deeply analyzed and five consecutives TPR-NOx soot oxidation cycles were carried out using the same portion of catalyst. As the T50% values for the first (543 ◦C) and fifth cycle (561 ◦C) are still under those corresponding to the uncatalyzed reaction (612 ◦C), it can be concluded that the catalyst is not significantly deactivated. In fact, the XRD data (shown in Figure S4 in Supplementary Information) of this used catalyst (after five TPR-NOx cycles) reveal that the hexagonal perovskite structure is not significantly modified. This means that, in the presence of oxygen in the reaction atmosphere, the catalyst keeps its structure and, consequently, its activity for soot oxidation in TPR conditions.

Finally, to complete the BFC0 evaluation, its catalytic performance for NO2-assisted diesel soot oxidation in isothermal conditions was determined by carrying out two consecutive soot oxidation experiments at 450 ◦C. The soot oxidation profiles at 450 ◦C (featured in Figure S5 in Supplementary Information) show that BFC0 catalyst is able to oxide soot without a significant deactivation and showing a high CO2 selectivity (close to 80%). The soot oxidation rate was calculated at the beginning of the reaction, in order to avoid the effect of high soot consumption, as being 1.2 min−1, which is not too far from a commercial model Pt/Al2O3 catalyst (1.8 min−1) used in the same experimental conditions. Thus, it seems that BaFeO3 perovskite could be a potential catalyst for diesel soot oxidation and, consequently, it could be used as an active phase for DPF.

#### 3.2.2. Soot Oxidation in GDI Conditions

A preliminary study about the use of BaFe1−xCuxO3 perovskites to catalyze the oxidation reaction of soot under the highest demanding GDI exhaust requirements (regular stoichiometric GDI operation, i.e., 0% O2) revealed that these oxides could be used as active phase for GPF [35]. It was concluded that the copper content has an essential role on the performance of the BaFe1−xCuxO3 catalysts for soot oxidation, agreeing with previous reports focused on diesel soot removal [10,26,34,36–38] and with the report for copper-supported ceria-zirconia catalysts for soot oxidation in GDI conditions [10]. These results confirm that the higher soot conversion presented by BFC4 with respect to the catalysts with lower copper content (BFC1 and BFC3) is linked both to the largest amount of β oxygen evolved by this catalyst, and with the presence of surface copper species (as BaOx-CuOx) [10,34,36–38].

To further analyze the performance of these BaFe1−xCuxO3 perovskites in GDI exhaust conditions, a study using 1% O2 in He (which represents the named "fuel cuts" GDI exhaust conditions) has been developed [9]. Figure 12 shows the profiles of CO2 and CO evolved during temperature programmed reaction experiments, including the profiles for the uncatalyzed reaction (blank corresponding to bare soot), as reference. Note that in the presence of BaFe1−xCuxO3 catalysts, the amount of CO decreases and the amount of CO2 increases. This means that, as could be expected [26,34,36,37], and as was observed during soot conversion reaction in NOx atmosphere, the catalysts increase the selectivity to CO2 from 56% for bare soot to 93% for BFC0, 70% for BFC1, 79% for BFC3, and 87% for BFC4. As for NO2 assisted diesel soot oxidation, the BFC0 catalyst is the most active and the addition of copper seems to decrease the ability of the catalysts to improve the CO2 selectivity. Additionally, as it has been deduced from soot conversion results in NOx atmosphere (see Table 4) that BFC4 presents the highest CO2 selectivity among the copper containing catalysts due to the presence of the surface copper species.

**Figure 11.** *Cont.*

**Figure 12.** CO2 (**a**) and CO (**b**) profiles during TPR soot oxidation in 1% O2 for BaFe1−xCuxO3 catalysts.

Figure 13 shows the soot conversion profiles in 1% O2, calculated based on the amount of evolved CO and CO2 featured in Figure 12.

**Figure 13.** TPR soot conversion profiles in 1% O2 for BaFe1−xCuxO3 catalysts.

In general terms, and in agreement with previous results for BaMn1−xCuxO3 catalysts [26], the catalytic effect is less relevant than the observed for NO2-assisted diesel soot oxidation. Thus, only BFC0 and BFC1 catalyze the soot oxidation with oxygen as the conversion profiles are shifted to lower temperature with respect to bare soot for these two catalysts. As observed for NO2 generation (see previous section), and in agreement with published conclusions [39,58,70], the soot conversion with oxygen follows the same trend than the amount of oxygen evolved during O2-TPD (Figure 8) as the most active catalysts (BFC0 and BFC1) are those generating the largest amount of oxygen. In fact, the highest activity of BFC1 at low temperature, i.e., at T < 600 ◦C approximately, seems to be related with the largest amount of and ' evolved (Figure 8). Thus, it seems that a similar reaction pathway is followed by soot oxidation with oxygen and with NO2-assisted diesel soot oxidation, even though

the catalytic effect is more relevant for the latter reaction than for the former. Hence, it could be concluded that the BaFe1−xCuxO3 perovskites catalyze more effectively the NO2-soot reaction than the O2-soot reaction. Additionally, the comparison of these results with the obtained in the most demanding GDI conditions (0% O2) [35] reveals that copper has an essential role on the performance of the BaFe1−xCuxO3 catalysts for soot oxidation only in the absence of oxygen in the reaction atmosphere.

Summarizing, the activity data above discussed reveals that the BaFe1−xCuxO3 perovskites catalyze both, the NO2-assisted diesel soot oxidation (500 ppm NO, 5% O2) and, to a lesser extent, the soot oxidation in the "fuel cut" GDI exhaust conditions (1% O2). BFC0 is the most active catalyst as the activity seems to be mainly related with the amount of O2 evolved during an O2-TPD, which decreases with copper content.

#### **4. Conclusions**

The results obtained for the BaFe1−xCuxO3 (*x* = 0, 0.1, 0.3 and 0.4) catalyst series allows us to conclude that:


**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/9/11/1551/s1, Figure S1: FESEM pictures for BaFe1-xCuxO3; Table S1: EDX data (atomic percentage) for BFC0 and BFC4 catalysts; Figure S2: Wagner (chemical state) plot for BaFe1−xCuxO3 catalysts; Figure S3: DRX patterns after O2-TPD for BaFe1−xCuxO3 catalysts (dotted lines). As reference, XRD patterns of fresh catalysts (solid line) have been included; Figure S4: DRX patterns after TPR-NOx with soot for BFC0 catalyst (dotted line). As reference, XRD pattern of fresh catalyst (solid line) has been included; Figure S5: Soot conversion profiles at 450 ◦C for BFC0.

**Author Contributions:** Conceptualization, V.A.-F. and M.-J.I.-G.; methodology, V.A.-F. and M.-J.I.-G.; validation, V.A.-F., M.-S.S.-A. and M.-J.I.-G.; formal analysis, V.T.-R., C.M.-M., V.A.-F., M.-S.S.-A. and M.-J.I.-G.; investigation, V.T.-R. and C.M.-M. and V.A.-F.; resources, M.-S.S.-A. and M.-J.I.-G.; data curation, V.A.-F., M.-S.S.-A. and M.-J.I.-G.; writing—original draft preparation, V.T.-R., C.M.-M. and V.A.-F; writing—review and editing, M.-S.S.-A. and M.-J.I.-G.; visualization, C.M.-M., V.A.-F. and M.-S.S.-A.; supervision, V.A.-F. and M.-J.I.-G.; project administration, M.-J.I.-G.; funding acquisition, M.-S.S.-A. and M.-J.I.-G.

**Funding:** This research was funded by the Generalitat Valenciana (PROMETEO/2018/076 and Ph.D. Grant ACIF/2017/221 for V.Torregrosa-Rivero), Spanish Government (MINECO Project CTQ2015-64801-R), and EU (FEDER Founding).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
