Single‐Atom Catalysts on Covalent Triazine Frameworks: at the Crossroad between Homogeneous and Heterogeneous Catalysis

Abstract

Heterogeneous single‐site and single‐atom catalysts potentially enable combining the high catalytic activity and selectivity of molecular catalysts with the easy continuous operation and recycling of solid catalysts. In recent years, covalent triazine frameworks (CTFs) found increasing attention as support materials for particulate and isolated metal species. Bearing a high fraction of nitrogen sites, they allow coordinating molecular metal species and stabilizing particulate metal species, respectively. Dependent on synthesis method and pretreatment of CTFs, materials resembling well‐defined highly crosslinked polymers or materials comparable to structurally ill‐defined nitrogen‐containing carbons result. Accordingly, CTFs serve as model systems elucidating the interaction of single‐site, single‐atom and particulate metal species with such supports. Factors influencing the transition between molecular and particulate systems are discussed to allow deriving tailored catalyst systems.

This Review highlights single‐atom catalysis on covalent triazine frameworks (CTFs), thus focusing on the available insights on factors influencing nuclearity and coordination environment of metal species on CTFs and the effect on catalytic performance.

1. Introduction

The landscape of catalysis can be spanned between the use of full catalysts in heterogeneous catalysis and molecularly defined complexes dissolved in liquid media in homogeneous catalysis. The former includes early examples of catalysts still in use today, such as Raney nickel or platinum–rhodium nets in the Ostwald process; the latter covers developments such as the Wilkinson catalyst for selective hydrogenation in the 1960s. ^[1] In addition, biocatalysis can be mentioned as part of this landscape, but will not be the subject of this contribution (Figure 1).

Figure 1.

The “landscape of catalysis” in the transition between heterogeneous and homogeneous catalysis, between full catalysts on the one hand and the use of defined metal complexes in solution on the other. Increasing the metal dispersion of supported particulate species results in a change of electronic structure and transition to immobilized complexes. The metal species at uttermost dispersion (single atoms) are the focus of this publication. Adapted from ref. [11].

In heterogeneous catalysis, nanoparticles became the focus of research with the advent of nanotechnology. It was recognized early on that only surface‐accessible centers participate in catalytic reactions. Accordingly, a larger surface‐to‐volume ratio enables more efficient use of the metal. ^[2] From an understanding of dedicated reaction centers on metallic surfaces coined by Taylor, Boudart developed the notion of structural sensitivity for reactions on metallic surfaces. ^[3] This describes the dependence of catalytic activity on the distribution of crystal planes on metal surfaces, which changes with size for particulate systems. This gave rise to the interpretation, still established today, of a distribution of locally unseparated reaction centers for supported, particulate as well as solid catalysts, which greatly complicates the analytical description of “the” active site. ^[4] For recent work on understanding the influence of different crystal planes in particulate catalysts, reference is made to work by Nørskov et al. ^[5]

With an increasing understanding of surface processes at catalytically active interfaces, which found an early peak with work by Ertl (Nobel Prize 2007), the development of heterogeneous catalysis today can also be seen as an approximation to homogeneous catalysis, which has prospered in parallel.[ ⁵ , ⁶ ] Nanoparticles can act similarly to homogeneous catalysts when stabilized in solution or functionalized with organic ligands. ^[2] In addition, often a delicate equilibrium exists between molecular and particulate species in solution, e.g. in Pd‐catalyzed Heck coupling without phosphine ligands. ^[7]

For supported catalysts, the increase of metal dispersion is an important field of research and has been postulated as a bridge to molecular catalysts. ^[2a] Increasing metal dispersion is accompanied by a drastic change in the electronic structure, especially for particles <1 nm.[ ^6b , ⁸ ] For even higher dispersion atomically dispersed reaction centers ultimately remain. If these centers are homogeneous in nature, the term single‐site heterogeneous catalyst (SSHC) defined by Thomas et al. is applicable. ^[9] In contrast to the above distribution of reaction centers, identical reactivity without mutual interference (“no spectroscopic cross‐talk”) is present. However, there is no need for them to be metallic or mononuclear. Active centers in molecular cluster compounds, highly crystalline zeolites as well as metal complexes can exhibit this property and allow a precise description of the reactivity of all active centers. A central, industrially relevant example is the TS‐1 catalyst with defined titanium centers in a silicate framework.[ ¹⁰ , ¹¹ ]

For metal complexes in homogeneous catalysis, a comprehensive understanding of the nature of the active species up to the detection of reaction mechanisms is possible.[ ^1b , ¹² ] The reactivity can be specifically influenced by steric and electronic factors of the ligand framework, thus forming the counterpart to the concept of structure sensitivity in particulate heterogeneous catalysis. ^[13] Immobilization of such complexes was the focus of early research efforts. ^[14] Different approaches were used, such as covalent binding to the support material as well as adsorption, or entrapment in porous supports. ^[15] Important contributions in this field were made by the development of surface organometallic chemistry (SOMC), which explicitly focuses on the reaction of organometallic complexes (or clusters) with defined silicate surfaces and subsequent modification under inert conditions.[ ^14b , ¹⁶ ] In general, the influence of the pore system regarding general accessibility of the active site and pore transport limitation must be taken into account for these materials, which often leads to significantly lower activity compared to homogeneous complexes. In addition, insufficient stability under reaction conditions is a decisive disadvantage, which has so far hindered widespread industrial use. ^[15b] Accordingly, there are only a few examples of immobilized catalysts used in chemical industry to date, including immobilized metallocene complexes of titanium and zirconium on silica,[ ¹² , ¹⁷ ] well‐defined Philips catalysts based on chromium ^[18] for polymerization reactions, and the AceticaTM process of UOP and Chiyoda using an immobilized rhodium catalyst for carbonylation of methanol. ^[19]

With the advent of covalent organic frameworks (COFs), and especially covalent triazine frameworks (CTFs), novel classes of supports evolve that go beyond oxide materials. They offer defined coordination environment and new types of metal–support interactions for stabilizing single atoms. ^[20] The presence of (mobile) atomic species in heterogeneous catalysts was postulated early on as a component of nanoparticle agglomeration according to the Ostwald principle. ^[21] For metal‐catalyzed transformations, however, continuous metal surfaces with adjacent adsorption centers were assumed to drive catalytic activity.[ ^3a , ^3c , ^16b ] In the next section, we will introduce effects that influence the nuclearity of supported metal species and the consequence for catalytic activity. This is followed by an overview of the influence of synthesis conditions on the properties of CTFs. Finally, the stabilization of metal species of different nuclearity on CTFs will be discussed.

2. Stabilized Single Atoms in Catalysis

Understanding the role of metal species on supports has been of great importance since early on, with the further development of analytical methods for direct detection of isolated metal species being the main focus. ^[14c] Central developments in the past 20 years, including aberration correction in electron microscopy as well as wide establishment of and better access to analytics using synchrotron radiation, have made it possible to obtain direct analytical evidence for processes on the atomic level. ^[22]

Although the use of isolated metal centers after immobilizing metal complexes on supports is not new, these developments have led to an extension of the concept of single‐atom catalysis to the whole of heterogeneous catalysis. Linked to this is the question of what drives catalytic activity in classical particulate systems. Work by the group of Flytzani‐Stephanopoulos illustrated that in the water‐gas shift reaction over Au/CeO₂ and Pt/CeO₂, isolated ionic species rather than nanoparticles contribute significantly to activity. ^[23] Accordingly, it was shown that agglomeration to gold nanoparticles led to activity loss. ^[24] In 2011, Qiao et al. demonstrated one of the clearest examples to date of isolated metal centers with catalytic activity using Pt atoms (0.17 wt %) on iron oxide. ^[25] This work led to a significant increase in work on single‐atom heterogeneous catalysts (SAHCs) and their characterization. Given the abundance of examples, reference is made to review articles by Flytzani‐Stephanopoulos et al., ^[26] Yang et al., ^[27] Liu et al., ^[28] and Kaiser et al., ^[29] and only a few key examples and synthetic strategies are discussed below.

In the description of isolated metal species on solid supports, the large number of publications has led to a rather inflationary use of the term “single‐atom catalysis”. In the transition from homogeneous to heterogeneous catalysis, the description of catalysts with identical environments around the active site (SSHC) was given by Thomas et al. ^[4b] However, the extent to which all catalyst systems declared as single‐site indeed possess identical active site environments is unclear. In low‐order supports, this is difficult to demonstrate.[ ¹⁸ , ³⁰ ] Furthermore, a differentiation from catalysis with single atoms (ions) in the gas phase is necessary, which is why terms such as heterogeneous single‐metal‐site catalysts (HSMSCs) are sometimes used. ^[31] Single‐atom catalysis thus emphasizes the activity of isolated supported metal centers, which do not necessarily have a “single‐site” character. Furthermore, organic ligands are always present for immobilized complexes or in well‐defined SOMC, whereas for SAHCs, ligand configuration is initially undetermined.[ ^16b , ²⁶ ]

For synthesis of SAHCs, liquid‐phase impregnation, ^[32] precipitation,[ ²³ , ²⁵ , ³³ ] high‐temperature syntheses, ^[34] atomic layer deposition, ^[35] or subsequent release from nanoparticulate species are employed. ^[23] In addition, a combination of coordination and pyrolysis is frequently used for organic carrier materials and macrocycles, despite limited control over the resulting active centers. ^[36] In oxide supports, defects in the cation lattice or dislocations in the crystal lattice have been identified as coordination centers for isolated metal species. ^[37] This efficient stabilization has also been exploited to capture and immobilize mobile species formed by fragmentation of larger nanoparticles at temperatures above 600 °C.[ ^37c , ³⁸ ] A special form of single‐atom isolation is the doping of inert metal particles (Cu, Ni, Ag) with active metal species (Pd, Pt). ^[39] These so‐called single‐atom alloys ensure high activity of isolated centers of the active component in inert metal particles by exploiting synergistic effects. However, many of these approaches only allow low metal contents, since higher loadings and stable bonding configurations in carrier materials result in the formation of particles of the mainly used Pt‐group metals.[ ²⁵ , ^32a , ^32c , ^37b ]

Apart from the widely used oxidic supports, carbon‐based materials such as activated carbon and graphene have also been used and high metal loadings of more than 10 wt % could be achieved by impregnation in organic solvents. ^[40] Here, organometallic complexes are used and partial ligand exchange can be assumed, which illustrates the difficulty in distinguishing SAHCs systems from immobilized complexes and SOMCs.[ ^16b , ²⁰ , ⁴¹ ] After the initial euphoria over the activity of isolated metal centers, later studies showed the drastic influence of catalyst preparation on catalytic activity.[ ²³ , ⁴² ]

CO oxidation over isolated Pt centers highlights the distinct influence of the support. Compared to Pt₁/FeO _x of Tao Zhang's group with significantly higher activity of isolated centers, ^[25] opposite results were reported for Pt₁/Al₂O₃. Moses‐DeBusk et al. reported that nanoparticles possess 3‐fold higher activity relative to the total amount of metal (Figure 2). ^[43] Mechanistically, this is related to participation of the support in the catalytic cycle. Migration of oxygen to the metal center according to a Mars–van Krevelen mechanism has been postulated, while for Al₂O₃ a mechanism without support contribution has been discussed,[ ²⁵ , ^32a , ⁴³ ] an effect also described for other reactions such as hydrogenation or water‐gas shift reaction. ^[44]

Figure 2.

Single‐atom reactivity and selectivity: (a) Qiao et al. found Pt single atoms on FeO _x to be highly active in CO oxidation. Adapted with permission from ref. [25]. Copyright 2011 Springer Nature. (b) In contrast, Moses‐DeBusk et al. found that on Al₂O₃ Pt clusters and nanoparticles are more active than single atoms. Adapted with permission from ref. [43]. Copyright 2013 American Chemical Society. Illustrating that highest dispersion does not always yield highest activity. The same is true for the case of Rh on TiO₂ where nanoparticles (d) show higher conversion than single atoms (c) in the water‐gas shift reaction (e, closed symbols). However, single atoms do not catalyze the methanation reaction and therefore have superior selectivity to CO (e, open symbols). Adapted with permission from ref. [44b]. Copyright 2016 Wiley‐VCH.

Testing of SAHCs in new reactions has always been linked to the question of comparing activity to particulate systems.[ ²⁸ , ⁴⁵ ] Progress in analytical methods, analogous to the detection of isolated species, allows better understanding of subnanometer clusters with very different activities in selected model reactions.[ ^32a , ^32c , ⁴⁶ ] Guan et al. showed that in water‐gas shift reactions catalyzed by Rh/TiO₂, selectivity strongly depends on particularity (Figure 2). ^[44b] Generalizing reactivity trends as a function of reaction and metal considered is challenging due to the variance of the systems. In any discussion of the isolated character of metal centers and their catalytic activity, the principle discussed by Schlögl et al. should be kept in mind that high (initial) catalytic activity is always in proportion to stabilization of reactive states, otherwise rapid deactivation follows. ^[47]

Therefore, the examples illustrate that it is the combination of metal and coordination environment in a carrier material that determines the reactivity.[ ^32b , ⁴⁸ ] This leads to an interpretation of the carrier material as a macroligand, which finally builds the bridge to the principles of homogeneous catalysis and stresses the decisive importance of metal centers’ binding environments.[ ²⁰ , ²⁸ , ^37f , ^48a , ⁴⁹ ]

Through computational chemistry calculations, Fako et al. demonstrated that carbon‐based supports are promising for stabilizing isolated metal species while maintaining catalytic activity. ^[50] While unfunctionalized carbon can only atomically stabilize very low metal contents, heteroatoms enable increased stabilization of isolated metal centers.[ ^36c , ⁵¹ ] For carbon‐based supports, major progress has been reached understanding the reduction potential and thus stability ^[52] along with describing surface chemistry ^[53] and targeted functionalization. ^[54] In addition, Mitchell and Pérez‐Ramírez point out that, while oxides were the main focus of research efforts in early years, tailored carbon‐based supports have received increased attention recently. ^[55] Nitrogen as heteroatom has played a central role in many contributions. In addition, new classes of materials have been established, particularly with the development of nanoporous polymers.

In the following, we focus on covalent triazine‐based networks (CTFs) as a typical class of these carbon–nitrogen materials. References to similar materials containing N‐moieties are made to exemplify general concepts.

3. Synthesis and Properties of Covalent Triazine‐based Networks (CTFs)

CTFs are a representative of the group of nanoporous polymers introduced by Kuhn et al. in 2008. ^[56] The triazine structural motif is formed by trimerization of nitrile monomers in presence of a Lewis acid catalyst at temperatures between 400 and 700 °C. ^[57]

In contrast to the widespread solvothermal approach for synthesis of other nanoporous polymers, an ionothermal approach using zinc chloride, which is liquid at reaction temperature, is often used for CTF synthesis.[ ⁵⁶ , ⁵⁸ ] ZnCl₂ acts as solvent and porogen in addition to the Lewis acid property. If reaction temperature and monomer‐to‐ZnCl₂ ratio are adjusted appropriately, crystalline materials can be prepared, following the example of the modular COFs according to Côté et al. with reversibility of trimerization under reactive conditions. ^[59] This was done in the first examples based on 1,4‐dicyanobenzene prepared at 400 °C and an equimolar ratio of monomer and ZnCl₂ (“CTF‐1”).[ ⁵⁶ , ⁵⁷ ]

Further examples of ordered materials based on 2,6‐naphthalene‐dicarbonitrile and 1,3,5‐tricyanobenzene were reported in subsequent years. ^[60] Above 400 °C, additional reaction pathways gain importance, leading to irreversible bond linkages with cleavage of smaller molecules. ^[57] Here, in addition to the existing micropores, new mesopores form in a reorganization process leading to hierarchical porosity.

Mechanistically, the formation of C−C bonds after retro‐trimerization and cleavage of HCN has been postulated. Other reactions involving the cleavage of molecules such as nitrogen and HCN have been proposed and linked to decreasing N‐content of CTFs compared to their theoretical values. ^[61] Due to this carbonization, the idealized polymer structure is at best a rough approximation.[ ⁵⁷ , ⁶¹ , ⁶² ] Scheme 1 shows the reaction equation commonly used in literature for synthesis at 400/600 °C, taking into account reversibility at 400 °C and irreversible reorganization at higher temperatures, starting from 1,4‐dicyanobenzene. A variety of other monomers and mixtures have been reported for CTF synthesis, which allows targeting composition and pore structure. ^[63] Materials obtained via the ionothermal route are black, exhibit moderate to good conductivity, and possess high thermal as well as chemical stability.[ ⁵⁷ , ⁶⁴ ] Specific surface areas exceeding 3000 m² g⁻¹ can be achieved. With commonly used monomers and temperature sequences, specific surface areas are between 1000 and 2500 m² g⁻¹.[ ⁶⁴ , ⁶⁵ ] Pore size distribution can be influenced by the choice of monomers, resulting in materials of hierarchical porosity. ^[57] Similarly, the N‐content can be varied, achieving up to 30 wt %. ^[66]

Scheme 1.

Synthesis routes for the preparation of CTFs including base‐catalyzed polycondensation, ionothermal synthesis utilizing ZnCl₂ or phosphorous pentoxide. Adapted from ref. [11].

Since the first publications, further approaches for preparing CTFs have been reported, mainly aiming at lower reaction temperatures and the use of other catalysts, since residues of ZnCl₂ of up to 5 wt % remained in the material after preparation.[ ^63b , ⁶⁷ ] Alternatives are ionothermal synthesis by other metal salts such as SnCl₂ or CuCl₂, ^[61] and metal‐free synthesis by phosphorus pentoxide simplifying preparation and workup. ^[68] Moreover, in the latter approach, diamides can be applied avoiding the need for cyanation reactions to prepare the dinitrile monomers.

Cooper's group reported in 2012 the use of trifluoromethane‐sulfonic acid (TfOH) in liquid phase to trimerize dinitriles. Compared to the ionothermal route, the synthesis was successful at room temperature and shorter reaction times. ^[67] These TfOH‐based materials better approximate the ideal structure due to absence of irreversible side reactions. The obtained white–yellow materials absorb light in the visible range, associated with pronounced photoluminescence. ^[69] Their comparatively low specific surface area could be significantly increased in a subsequent treatment in ZnCl₂ with the possibility of using an open vessel, since the TfOH‐based materials are no longer volatile. ^[70] Another option for synthesis of CTFs in a solvo‐ rather than ionothermal process was described in 2017 by Wang et al. using polycondensation in dimethyl sulfoxide (DMSO). ^[71] Here, dialdehyde compounds react with diamidines with the release of water and ammonia. The latter derivatives can be prepared from the respective dinitriles in a one‐step synthesis. The authors showed that similar to the use of TfOH, yellow materials with a specific surface area of 660 m² g⁻¹ result.

While in classical ionothermal synthesis thermodynamic control and ordered material structures are achieved by reversibility of bond formation, ^[56] in the amidine‐based route the number of nucleation centers is crucial to avoid kinetic control and formation of a disordered structure. Ordered CTFs were achieved by using a dialcohol as second monomer ^[72] or controlled addition of the monomer, ^[73] thus reducing the number of active chains.

While ionothermal synthesis dominates early publications, the number of contributions using solvothermal synthesis has steadily increased. Here, the goal is to use appropriate methods to produce more ordered materials exhibiting semiconductor properties due to absence of carbonization during synthesis. ^[74] In addition to the four main routes for CTF synthesis, a number of other triazine‐containing polymers have been reported, for which reference is made to relevant publications by Meier et al., Puthiaray et al., and Krishnaray et al.[ ^63a , ⁷⁵ ] The multitude of synthetic approaches for this class of materials with very diverse properties has led to investigations in various fields of application. We herein focus on CTFs as supports for catalytically active metal species. This application was already postulated by Kuhn et al. early on, based on the specific coordination sites at high N‐content enabling the preparation of CTF‐based single‐site and single‐atom catalysts.[ ⁵⁶ , ⁷⁶ ]

4. CTFs as Support Material in Catalysis

Supported metal species range from particulate systems to subnanometer clusters to immobilized metal complexes. “Metal species” is accordingly used as an umbrella term independent of the nuclearity of the supported metal, which is further differentiated into nanoparticles (particulate) and molecular or even isolated metal species. Immobilized metal complexes resembling the binding environment of a homogeneous catalyst belong to molecular metal species (Figure 3).

Figure 3.

Nomenclature of immobilized metal species in CTFs as a distinction between particular and molecular systems. Adapted from ref. [11].

4.1. Particulate Metal Species Supported on CTFs

Coordination of metal complexes prior to reduction to nanoparticles has been identified as a major advantage in using CTFs as supports, as it avoids the need for additives to stabilize nanoparticles in solution in advance. ^[64] After reduction, highly dispersed particulate systems were obtained exhibiting high activity and selectivity for a range of transformations (vide infra). However, during coordination, the formation of nanoparticles can occur as a minor component next to isolated metal species. ^[77] A positive effect by pre‐coordination on the dispersion of nanoparticles was also shown for silicates. Metal complexes immobilized via silanol groups led to smaller particle size (<1 nm) after reduction in a hydrogen stream compared to impregnation by electrostatic adsorption. ^[78]

One of the first examples of classically particulate systems based on CTFs was demonstrated by Chan‐Thaw et al. using Pd/CTF. ^[79] They showed increased stability compared with the use of activated carbon as support for glycerol oxidation, whereby the Pd nanoparticles were stabilized in solution in advance and then impregnated onto the CTF (Figure 4). The aforementioned combination of impregnation and reduction led to a high metal dispersion of Ru on CTF, as described by Artz et al. ^[65b] This in turn had a positive effect on activity in the oxidation of 5‐hydroxymethylfurfural compared to Ru/C. Other examples of particulate systems cover the hydrogenation of sugar alcohols with Ru/CTF, where the influence of nitrogen species on metal species improved selectivity and highly dispersed nanoparticles (d<2 nm) were detected. ^[80] A similar system based on Ru nanoparticles was applied in NH₃ decomposition, where increased electron density at Ru was postulated to be beneficial for the reaction rate. ^[81] Siebels et al. demonstrated that nanoparticle preparation can also be carried out by microwave‐induced decomposition of metal carbonyl complexes. They successfully used these systems in the hydrogenation of benzene as model reaction. ^[82] While all previous examples involved ionothermally prepared CTFs, Liu et al. showed that Pd nanoparticles in a solvothermally prepared CTF can act as co‐catalyst for photocatalytic production of hydrogen. ^[83] Here, the reduction after impregnation was carried out by irradiation with light (>420 nm) and variability of the metal–support interaction was illustrated by significantly different particle size distributions.

Figure 4.

Pd nanoparticles on CTF (a) show improved performance compared to Pd/activated carbon in glycerol oxidation (b). Adapted with permission from ref. [79]. Copyright 2010 American Chemical Society.

4.2. Mononuclear Metal Species Supported on CTFs

The structural motif of a (bi)pyridinic ligand can be realized in CTFs by appropriate choice of monomer, such as 5,5′‐bipyridine‐dicarbonitrile or 2,6‐pyridine‐dicarbonitrile.[ ⁷⁶ , ⁸⁴ ] Functionalization is then achieved by impregnation, in which a solubilized precursor complex coordinates to the support surface by ligand exchange. ^[84] This concept was used by Palkovits et al. realizing a comparable chemical environment in CTFs as in the molecular Periana catalyst (Figure 5). The immobilized Pt complex enabled oxidation of methane in concentrated H₂SO₄ at 210 °C.[ ⁷⁶ , ⁸⁵ ]

Figure 5.

Literature known systems with (bi)pyridine structural motif for coordination of Pt complexes for partial oxidation of methane under harsh reaction conditions. Adapted from ref. [11].

A similar system also based on immobilized Pt complexes was used in ORR in methanol by Kamiya et al. The reactivity of the isolated metal centers was demonstrated by the absence of methanol reduction, which usually competes in catalyst systems in which Pt is present in particulate form. ^[86] The same group succeeded in a later publication in using immobilized Ru complexes (RuCl₃/CTF) to perform the electrochemical oxidation of alcohols selectively over the oxidation of water with a similar reference to high metal dispersion. ^[77a] More recently Hu et al. demonstrated that Co single‐atoms can be stabilized on a bipyridine‐derived CTF. This system was then shown to be active in the photoreduction of CO₂ in the presence of a dye and a hole scavenger. ^[87]

In general, the immobilization of metal complexes in solid supports is intended to preserve the reactivity of homogeneous metal complexes, while improving the separability of the catalyst.

Following this model, Tahir et al. immobilized an Ir methoxy‐1,5‐cyclooctadiene dimer [Ir(OMe)(cod)]₂ in CTFs and demonstrated catalytic activity in the boronation of C−H bonds in aromatics, a reaction usually performed with homogeneous catalysts. ^[88]

For the catalyzed synthesis of β‐lactones from epoxides, where poor separability of the homogeneous complexes has previously hindered further commercialization, a bifunctional system with a Lewis acidic aluminum salt and a Co complex as the metal component was reported. ^[89] The latter was immobilized as the counterion of the aluminum salt coordinated to the CTF. The bifunctional character of the corresponding homogeneous complex was thus successfully preserved. Nevertheless, recycling experiments showed that the ionically bonded Co complex was susceptible to leaching, indicating the need for direct coordination to the N‐species of the CTF for high stability. Further examples of molecular catalysts were published in catalytic transfer hydrogenation with immobilized Ir ^[90] and Ru complexes. ^[91] After the metal‐free oxidation of alcohols by activation of oxygen on CTFs in a prior study, Abednatanzi et al. were able to improve the system by functionalization with (IrCp*Cl₂)₂. ^[92] Away from Pt group metals, the activity of Ni complexes in the oligomerization of ethene was demonstrated, although the activity was an order of magnitude lower than in a comparable homogeneous catalyst. ^[93]

CTFs have also been considered for the catalytic dehydrogenation of formic acid. Using an immobilized (IrCp*Cl₂)₂ species, Bavykina et al. showed in 2015 that at 80 °C and 2 wt % metal loading, the use of an additional base could be omitted and an initial TOF of 21 300 h⁻¹ was achieved. ^[94] Gunasekar et al. used a similar experimental setup again with (IrCp*Cl₂)₂ as metal precursor to functionalize the CTF. ^[95] Here, largely independent of the metal loading between 1.4 and 4.7 wt % Ir, a significantly lower initial TOF of 2 820 h⁻¹ at 80 °C was measured. By increasing the maximum synthesis temperature of the CTF from 400 to 500 °C as well as increasing the metal loading to 11.3 wt %, the activity could be increased to 7 930 h⁻¹. Our group recently immobilized Ir(COD)(acac) (1 wt %) on CTFs synthesized via the amidine route. TOFs of 24 400 h⁻¹ with CO contents below 40 ppm could be realized. Pretreatment conditions significantly influenced activity as well as selectivity (vide infra). ^[96] For the monomer 5,5′‐bipyridine‐dicarbonitrile, Hug et al. demonstrated that although the triazine motif was no longer detectable in CTFs above 500 °C synthesis temperature in the infrared spectrum, there was always high metal uptake with dispersion as a molecularly coordinated species. ^[84] This is in agreement with results of Soorholtz et al. on K₂PtCl₄/CTF.[ ⁷⁶ , ⁸⁵ ] Defined coordination centers inspired by homogeneous ligands are thus not mandatory for the stabilization of immobilized metal species, although the description of structure–activity relationships as well as the coordination environment is significantly complicated in their absence.

From the publications presented, it is evident that binding environments similar to those of homogeneous ligands exist in CTFs for immobilization of complexes, while very high metal dispersions with particle sizes of <2 nm can be achieved for nanoparticles. These systems also exhibit increased stability against leaching.

For metal particles, modulation of electronic properties by coordination in CTFs has also been postulated.[ ⁸⁰ , ⁸¹ ] From a statistical point of view, the influence of nitrogen centers of a support on the electronic properties of an immobilized metal species should increase with decreasing particle size. Thus, with a steady increase in dispersion of metal particles, the influence of the carrier material increases and it must be increasingly understood as a macroligand.[ ^49a , ⁵⁰ ]

This implies the transition to immobilized complexes that exhibit coordinative bonds to the support and further ligands. Thus, the interaction of metal species and CTF acquires a strong importance to understand the reactivity of highly dispersed metal species and to study their stabilization against agglomeration. The following section exemplifies studies on CTFs and related material classes with comprehensive characterization of the metal species as well as derived bonding models.

4.3. Influence of Nitrogen Functionalities on Supported Metal Species

Compared to the industrially widely used oxides, significantly less has been reported on the interaction of metal species with N‐centers. A number of examples have been reported for CTFs, but the interaction of metal species with the nitrogen species of the support has only occasionally been deeply investigated. In the following, if not explicitly mentioned, examples after impregnation with a dissolved metal complex are discussed instead of the immobilization of nanoparticles after stabilization as sol in solution. A major contribution to the understanding of the interaction of metal species in CTFs was provided by the work of Soorholtz et al. in which immobilized Pt species in ionothermally prepared CTFs were characterized by a number of analytical methods (Figure 6). ^[20] The interaction with nitrogen in the material was demonstrated by XPS, where the formation of a shoulder at higher binding energies indicates successful coordination. This is in agreement with results of Gascon's group.[ ⁶¹ , ⁹⁴ ] Deviations of the found coordination environments from the planar structure in the homogeneous Periana catalyst as a model complex were attributed to irregular pore geometry and thus, a distribution of coordination centers.

Figure 6.

Employment of TEM (a), EXAFS (b) and XPS (c) enabled the elucidation of Pt single‐atom coordination environment on a CTF (c, inset). Coordination is equivalent to a Pt 2,2′‐bipyrimidine complex. Adapted with permission from ref. [20]. Copyright 2016 American Chemical Society.

The influence of nitrogen species in CTFs on the immobilization of metal precursors (Rh or Ir) as isolated species was also confirmed by Gunasekar et al. ^[62b] However, support materials hold a limited capacity (e.g. coordination centers) to stabilize isolated metal centers as derived by Lykhach et al. for Pt/CeO₂. ^[97] This limit was also exceeded for CTFs by Gunasekar et al. when the loading was increased beyond 4.7 wt %, demonstrating a distinct influence of N‐centers during impregnation. The high porosity and number of free N‐centers is associated with the successful formation of isolated species. However, at high loading and reduction of pore volume, metal particles form. It was postulated that nitrogen centers in the material serve as nucleation centers. Similar observations were made by Hug et al. for impregnation with Pt and Pd precursors. ^[84]

Nanoparticulate catalysts were characterized in early publications by their high stability. Here, the N‐centers in Pd/CTF led to higher stability against deactivation of the metal centers compared with activated carbons (Pd/C). The metal precursor was reduced in advance in solution with NaBH₄ and stabilized as a sol with polyvinyl alcohol (PVA). ^[79] Functionalizing the support by impregnation with a dissolved metal precursor analogous to Soorholtz et al., ^[20] resulted in significant differences in reactivity. ^[98] Although small particles were shown to be present after reduction and the lower reactivity could be correlated with a significantly higher fraction of ionic (potentially isolated) Pd²⁺ species in XPS analysis even after reduction. Thus, coordination of metal complexes by nitrogen has a strong influence on the behavior during reduction (vide infra). For the interaction after reduction, also a dependence on the metal species exists. Beine et al. and Fei Chang et al. demonstrated an increase in electron density for Ru/CTF in contrast to the cationic species in Pd/CTF.[ ⁸⁰ , ⁹⁹ ] For the nanoparticles formed during reduction, a shift of −0.6 eV was determined by XPS and attributed to basicity of the support. However, the nature of Ru species after reduction is not fully understood, as another work on Ru/CTF demonstrated a high proportion of cationic Ru species after reduction. ^[77b] This is consistent with reports by Pilaski et al. on Rh/CTF. ^[100] Although the activity in the hydroformylation of 1‐octene was significantly lower than with homogeneous systems, the presence of highly dispersed Rh species after reduction in H₂ was confirmed by X‐ray absorption spectroscopy. No signals for Rh−Cl were observed in the analysis of the near‐edge fine structure compared with the RhCl₃ precursor complex used. A certain fraction of nanoparticles was detected in electron micrographs, but these were ruled out as the main species. These results again suggest incomplete reduction upon pre‐coordination of the metal precursors and predominant formation of isolated cationic metal species.

As discussed in Section 3, CTFs prepared by ionothermal synthesis and by polycondensation result in a range of N‐rich systems with different properties. For this reason, both g‐C₃N₄, which has been intensively studied in recent years, and N‐doped carbon materials emerge as suitable comparative systems for studying metal–nitrogen interactions. The former has received considerable attention for stabilizing Pd species through the work of Pérez‐Ramírez's group. ^[101] The central concept of the material system is a cationic Pd species present in the material after impregnation. ^[102] Atomic dispersion was ensured by low metal contents (<2 wt %), as well as ultrasonic and microwave treatment. By XPS, Pd²⁺ and Pd⁴⁺ metal species were observed, although a formal distribution of charge between metal and support was critically put in perspective at various points. Away from formal oxidation states, the Pd²⁺/Pd⁴⁺ ratio depends strongly on the choice of monomer and hence the bonding environment.[ ¹⁰² , ¹⁰³ ] The cationic nature of immobilized metal species could furthermore be generalized for a range of metals (Pt, Ag, Ir) supported on g‐C₃N₄. ^[104] A model for electron transfer from isolated metal centers (Pd) to a N‐doped support was proposed by Arrigo et al. for Pd immobilized in N‐doped carbon nanotubes (Pd/N‐CNT). ^[105] Combining NEXAFS measurements at the nitrogen K‐edge and XPS measurements (Pd_3d), the contribution of a π‐backbonding of filled d‐states of the metal was deduced to outweigh the electron‐shifting σ‐bonding of the free electron pair at the nitrogen. The transferability of this concept to N‐doped carbons upon interaction with pyridinic nitrogen centers has been demonstrated for Pd and Pt by Büchele et al., ^[106] Bulushev et al., ^[107] He et al., ^[108] and Melke et al., ^[109] among others. Unlike graphitized carbon nitrides, however, N‐doped carbons encompass a wide range of materials for which generalization of influence is complicated by the distribution of N‐species as a function of substrate, preparation method, and the presence of other heteroatoms. ^[110] For example, for a carbonized disordered material, the formation of nanoparticles was clearly observed next to isolated Pd centers. ^[104] This is consistent with publications by Bulushev et al. on Pd/CNT, Pt/CNT, and Ru/CNT, in which the isolated species were always present next to particles (1–3 nm) after impregnation and reduction.[ ¹⁰⁷ , ¹¹¹ ]

In carbonized materials, in addition to pyridinic nitrogens, pyrrolic and graphitic nitrogen species are also present, comparable with the distribution of nitrogen species for ionothermally prepared CTFs.[ ⁶¹ , ¹¹⁰ ] Accordingly, for a N‐doped carbon analogous to Pt/CTF (heterogenic “Periana catalyst”, Figure 6), high activity in the partial oxidation of methane could be demonstrated, although a faster deactivation occurred. ^[85] Isolated metal species coordinate primarily at pyridinic nitrogen centers,[ ¹⁰⁵ , ¹⁰⁷ ] and DFT calculations emphasized that no bonds to graphitic nitrogen centers form for Pd and Pt. Instead, these act as nucleation centers for nanoparticle formation through an electron‐shift contribution. ^[112]

With this information, the observations of nanoparticle formation by Gunasekar can be rationalized for Ir/CTF at higher loading dependent on the type of nitrogen species. ^[62b] After immobilization in stable bonding configurations with pyridinic nitrogen, graphitic nitrogen centers lead to nanoparticle formation. This is consistent with results of Ning et al. who considered the influence of immobilization methods for Pt in N‐CNTs. ^[113] Impregnation is predominantly followed by an interaction with pyridinic nitrogen centers, while particulate metal species after re‐reduction with ethylene glycol in solution followed by impregnation were predominantly postulated to interact with graphitic nitrogen centers resulting in more electron‐rich Pt nanoparticles. Correspondingly, in other work, an XPS shift of the Pd main signal to smaller binding energies (more electron‐rich) was found when the proportion of graphitic nitrogen in the substrate increased compared to Pd/C. ^[114]

A review connected the chemical distribution of nitrogen species in materials, and thus the electronic interaction with metal species, to the precursor material used to prepare the N‐doped carbon. ^[110] Calculation of band structures of the carbon‐based materials allows generalizing the interaction and extends the often molecular view in the literature. Accordingly, Mao et al. showed that the charge transfer between metal and support depends not only on isolated nitrogen species, but also on the electronic structure of the entire material in which they are embedded. ^[112] The electronic interaction is determined to a large extent by the level of the Fermi energy of the lowest unoccupied state (LUMO) of the support relative to the position of the highest occupied states in the metal species ( d‐band position) and leads to a charge transfer. This consideration was derived from solid‐state physics according to a Mott–Schottky transition, and the extent of charge transfer depends on the conductivity of the components in contact. ^[115] For carbon‐based materials, Antonietti et al. introduced the concept of “noble carbons,” i.e., the dependence of the location of the energy levels of a material on the substrates used for its production. ^[52] In this context, incorporation of pyridinic and graphitic nitrogen into a graphitized material was interpreted as p‐doping and n‐doping, respectively. ^[112]

In summary, the central contribution to the interaction with metal species can be assigned to pyridinic nitrogen centers independently of the nature of the N‐containing support. As has been shown for CTFs and g‐C₃N₄, impregnation leads to coordination of metal complexes and formation of isolated metal centers, which, using e.g. Pd, are characterized by electron transfer from the metal to the support. However, when a certain loading is exceeded with complete occupancy of stable coordination centers or with an increasing proportion of graphitic nitrogen centers in carbonized materials, nucleation at such centers follows with the formation of nanoparticles.

Pérez‐Ramírez's group reported that stabilizing Pd against reduction is directly linked to the extent of interaction with pyridinic nitrogen centers. ^[104] For considering the transition from immobilized complexes in CTFs to classically particulate systems by reduction, the influence of reductive treatment on isolated metal species will be discussed below using selected systems.

4.4. Influence of Reductive Treatment on Isolated Metal Species

Stabilizing isolated metal centers is one of the key challenges of single‐atom catalysts, since thermodynamically the formation of particles (or agglomeration to small clusters) minimizing the surface energy is favorable.[ ²⁶ , ²⁸ , ^48b ] For this reason, usually only small metal loadings are used. The tendency to agglomeration is always in relation to stabilization by coordination centers of the support and to the reducing agent and temperature (generally the reduction potential) imposed.[ ^37d , ¹¹⁰ , ¹¹⁶ ] Even if the exclusive presence of isolated metal centers is accessible in activated carbons by choosing suitable synthesis conditions, ^[40a] the formation of nanoparticles is observed under reaction or reductive conditions. ^[117] Isolated metal centers can also be qualitatively detected in commercial catalysts (Pd/C and Pt/C) after reduction, since a number (albeit small) of stable coordination geometries exists for isolated metal centers in the support. ^[118] For N‐containing materials, stabilizing higher metal dispersions and avoiding agglomeration are listed as key advantages.[ ⁷⁹ , ⁹⁸ , ¹⁰⁸ , ¹¹⁹ ] For the common sequence of impregnation with metal salts and reduction to nanoparticles using a suitable reducing agent to prepare nucleated catalysts, pre‐coordination of the dissolved metal salt to pyridinic nitrogen centers with formation of isolated metal species was identified as a key aspect (vide supra).[ ⁶⁴ , ⁸⁰ , ¹¹³ ]

Little is known about the stability of nitrogen‐bound isolated metal species under reductive conditions, in part because nanoparticulate systems are advantageous for most applications and thus were the target of the respective work.[ ^65b , ⁹⁸ ] While in some publications no dedicated reduction of immobilized metal species in the support was performed prior to catalytic testing,[ ¹⁰² , ¹⁰⁶ ] reduction with H₂ or NaBH₄ generally results in the formation of nanoparticles with metal species in oxidation state zero.[ ⁹⁸ , ¹⁰⁵ , ¹²⁰ ] However, Bulushev et al. demonstrated the coexistence of isolated, cationic species and nanoparticles. They assigned a significant contribution on catalytic activity in the dehydrogenation of formic acid to isolated metal species.[ ¹⁰⁷ , ¹¹¹ ] Also, in work by Chan‐Thaw et al., cationic Pd species were present after impregnation and reduction, showing low activity in glycerol oxidation. ^[98] Our group recently demonstrated the stability of Ir species on CTF under strongly reducing conditions up to 400 °C. We even observed an increase in activity and selectivity for formic acid dehydrogenation and were able to attribute that to the removal of organic ligands, adoption of a different coordination geometry and increased electron transfer from the CTF (Figure 7). ^[96] Further proof for the high stability of isolated metal species on C−N supports under hydrogenative conditions was provided by Chen et al. for Pd/g‐C₃N₄ on a 5 % H₂/He mixture at 150 °C. ^[104] In a variation of the support material, this stabilization was correlated with the Pd⁴⁺/Pd²⁺ ratio, i.e., the strength of electron transfer from the support. Furthermore, Shao et al. showed that cationic Ir species were always atomically dispersed in an N‐containing porous organic polymer (POP) after reduction with NaBH₄. ^[121]

Figure 7.

Scheme of the functionalization of CTFs with Ir(acac)(COD) and subsequent reduction in hydrogen (a). Corresponding TEM images of the sample as functionalized (b), after reduction at 400 °C (c), and after reduction at 500 °C (d). Ir single atoms are largely stable up to 400 °C and show improved performance in formic acid decomposition upon reductive treatment at that temperature. Reprinted with permission from ref. [96]. Copyright 2022 Wiley‐VCH.

5. Summary and Outlook

In this contribution, we discuss CTFs as versatile hosts for metal species ranging from single atoms to nanoparticles. Catalysts based on this material often show improved stability if compared to oxide materials or carbons without nitrogen. The same is true for chemically related materials like graphitic carbon nitride or N‐doped carbons. In general, the tendency for isolated species to agglomerate results from the relationship between the metal complex interacting with the support and the cohesive energy upon formation of crystalline particles. ^[110] The strength of the interaction with the support depends on the electronic structure of the transition metal and generally decreases from left to right in the periodic table. ^[122] For oxide materials, the strong interaction of early transition metals with oxide supports leads to very stable, isolated systems, which has been successfully exploited in the field of surface organometallic chemistry on silica (SOMC). ^[123] When Pt‐group metals are used, there is a weaker interaction of the metal centers with oxygen centers, which coupled with the higher reduction potentials generally causes agglomeration to nanoparticles.[ ^78b , ¹²⁴ ] This has been attributed to poor metal–support interaction, which has been justified by the HASB principle or the energetic location of the states involved according to the Mott–Schottky model.[ ¹¹⁵ , ¹²⁵ ] The loss of metal dispersion is thus a common observation when oxidic support materials are used with an increase in reduction temperature or metal loading.[ ²⁵ , ^32a , ^32c , ¹²⁶ ]

In contrast, pyridinic nitrogen was demonstrated to facilitate stabilization of single atoms for Pt‐group metals. Therefore, material classes containing this functionality offer great potential to explore the crossroad of homogeneous and heterogeneous catalysis, where single‐atom catalysts are located. We see the future of this field in investigating the exact coordination environment of metal species and their evolution under reaction conditions as was done for an oxide support by DeRita et al. ^[32b] Further inspiration can be taken from the field of SOMC and from the extensive studies of the Gates group on agglomeration mechanisms of single atoms stabilized on oxide supports. ^[127]

At the moment, comparable studies are missing for carbon–nitrogen materials—especially under reducing conditions. Filling this gap would pave the way to a more rational design of single‐atom catalysts with high stability against agglomeration and tailored coordination environment for maximum catalytic performance.

Conflict of interest

The authors declare no conflict of interest.

Biographical Information

Andree Iemhoff is a chemist interested in catalysis and sustainable industrial processes. He studied chemistry at RWTH Aachen University, focusing on catalysis, analytical methods and biomass‐derived platform molecules. Research stays with Prof. Louis Bouchard (UCLA) and Prof. James Clark (University of York) complemented his education. Following his PhD, he joined BASF SE as lab team leader for heterogeneous catalysis in 2021.

Biographical Information

Maurice Vennewald studied chemistry at RWTH Aachen University specializing in catalysis and synthesis. During his Master′s he conducted a research internship in the group of Prof. Enrique Iglesia (Department of Chemical Engineering) at UC Berkeley and performed his master thesis research at hte GmbH under the supervision of Stephan Schunk. Currently he is a PhD student in the group of Prof. Regina Palkovits, studying the dynamics of palladium single atoms on carbon–nitrogen support materials.

Biographical Information

Regina Palkovits is a full professor for Heterogeneous Catalysis & Chemical Technology at RWTH Aachen University. She graduated in Chemical Engineering from Technical University Dortmund and finished her Ph.D. with Prof. Ferdi Schüth at the Max‐Planck‐Institut für Kohlenforschung. Since 2010 she has been Professor at RWTH Aachen University. She received numerous awards, including the 2019 EFCATS Young Researcher Award and the 2016 DECHEMA Award. She is a Max Planck Fellow at the Max Planck Institute for Chemical Energy Conversion and as of 2020, a member of the North Rhine–Westphalian Academy of Sciences, Humanities and the Arts.

Iemhoff A., Vennewald M., Palkovits R., Angew. Chem. Int. Ed. 2023, 62, e202212015; Angew. Chem. 2023, 135, e202212015.