Mechanistic and thermodynamic insights into binding and activation of small molecules on metallozeolites – relevance for adsorption and catalysis

Abstract

Metallozeolites exchanged with 3d transition metal ions (TMI) are versatile catalytic materials due to their well-defined framework structures, redox flexibility, and remarkable adsorption and catalytic properties. These features make them invaluable for both fundamental and applied research, underpinning numerous catalytic technologies. The binding and activation of small reactant molecules is governed by the complex mechanistic interplay of involved intrazeolite reactions, whose course is influenced by the flexible valence, spin, and coordination states of the encaged metal ions and the metal–oxo entities. Despite significant advances, the nature of active sites, confinement effects, and the complex activation mechanisms of reactant molecules, which act as both innocent and non-innocent ligands, remain subjects of ongoing debate. This has driven extensive research into the thermodynamic constraints and molecular-level insights into activation processes with orbital and spin resolution. This review critically examines the thermodynamic and molecular aspects of intrazeolite speciation of transition-metal ions and metal–oxo active sites, their structural dynamics, and reactivity toward catalytically relevant small molecules, including NH₃, H₂O, CO, N₂, O₂, NO, N₂O. Particular emphasis is placed on ligand coordination, redox activation, and the role of electronic and spin states in dictating the catalytic behaviour of metallozeolites. The discussion integrates insights from site-selective spectroscopies and computational methods to elucidate the structural, thermodynamic, and molecular aspects of metal–ligand interactions and activation pathways, with an emphasis on the role of spin states in binding and reactivity. We hope that this review can serve as a relevant and valuable reference for researchers working with zeolite catalysts, providing new insights and inspiration.

---

Filip Zasada

Filip Zasada, born 1981, received his PhD in Chemistry from Jagiellonian University in 2009, where he currently serves as Associate Professor in Inorganic Chemistry Department. He completed two research internships in the Heterogeneous Catalysis Department at the University of Lille, where he combined DFT modelling and first principles thermodynamics for applications in redox catalysis. His current research interests focus on the molecular pathways of reactions involving small molecules in contact with redox catalytic centers on oxide surfaces and within zeolite frameworks.

Piotr Pietrzyk

Piotr Pietrzyk has been a Professor at the Faculty of Chemistry, Jagiellonian University (JU) since 2019. He holds an MSc (2001) and a PhD in chemistry from JU (2005). He completed research stays at the University of Lille, France, and the University of Turin, Italy. He was awarded by the Prime Minister of Poland, the Ministry of Science and Higher Education, and the Foundation for Polish Science. Since 2024, he has been elected Dean of the Faculty of Chemistry at Jagiellonian University. His research focuses on characterizing catalysts and investigating reaction mechanisms in heterogeneous catalysis (activation of small molecules, SCR, AOP) through spectroscopic and quantum chemical studies. He has co-authored over 80 scientific publications.

Mariusz Radoń

Mariusz Radoń, born 1982, received PhD (2011) and habilitation (2019) in Chemistry and is currently an Associate Professor at the Faculty of Chemistry, Jagiellonian University, Kraków. He has completed internships in KU Leuven, HPC Centre Cyfronet AGH (Kraków), and visited Max-Planck-Institut für Kohlenforschung (Mülheim). His research interests are focused on computational (bio)inorganic chemistry using quantum chemistry methods. He received the 2024 Award of the Quantum Bio-Inorganic Chemistry Society for his contributions to the accurate prediction of energetics in transition metal complexes.

Zbigniew Sojka

Zbigniew Sojka is a full professor at the Faculty of Chemistry of the Jagiellonian University (Poland), a corresponding member of the Polish Academy of Science, and Head of the Catalysis and Solid State Chemistry Group. He earned his PhD in Chemistry from Jagiellonian University in 1984 and subsequently completed postdoctoral fellowships at Pierre and Marie Curie University (France), and then at Lehigh University (USA). His research is focused on experimental and computational catalysis, materials chemistry, molecular modelling, spectroscopy, and electron microscopy. He published over 230 papers and holds more than 25 patents. Professor Zbigniew Sojka received several awards from the Minister of Science and Higher Education of Poland, the Polish Chemical Society, and a Jagiellonian Laurel for achievements in science.

---

1. Introduction – context, theoretical and experimental background

Shortly after the development of synthetic zeolite materials, they were ion-exchanged with various transition-metal ions (TMI)^1,2 to acquire new, and often unique functionalities and reactivities, providing a solid foundation for their widespread applications in adsorption and redox catalysis.^3–6 In this context, the numerous zeolite frameworks⁷ have provided a large variety of molecular architectures with the degree of perfection required for sensible molecular-level studies of the fundamental events in chemisorption and catalytic chemistry.⁸ Due to their unique structural features, metallozeolite catalysts provide an opportunity to investigate the transition from simple (molecular ligation via lone pair donation with spin conservation) to complex (dissociative redox processes with a change in spin state) adsorption events, as well as from single-site to multiple-site chemistry. They also made it possible to explore the intermediate reaction steps both stoichiometrically and structurally, using spectroscopic methods in combination with theoretical modelling at various levels of sophistication.^1,2,8–11

As a result, a large variety of promising research areas has emerged over the past few decades, including specific issues related to the substitution of TMI for the framework Al³⁺ or Si⁴⁺ ions, extra-framework accommodation of TMI, as well as capture, redox activation, and catalytic transformations of reactant molecules into desired products.^6,8,12–14 The coupled redox and acid–base chemistries, as well as space-confined effects, significantly impact the latter processes. In this context, the unique properties of the encaged bare transition-metal ions and the metal–oxo active centres result from their flexible valence, spin, and coordination states, which, in conjunction with the space confinement effects, lead to specific catalytic performance not encountered in the counterpart homogeneous catalysis.^12,13 Among the catalytic processes involving small-molecule gas-phase reactants, NH₃-SCR,^14,15 HC-SCR,^16,17 CH₄ oxidation,^18,19 and N₂O decomposition^20,21 are the most notable.

The zeolite framework type, the Si/Al ratio, and the distribution of Al atoms within the cavities and channels (constrained by the Löwenstein rule²²), essentially govern the intrazeolite speciation of TMI. It is manifested in the simultaneous presence of exchanged cations in various valence and coordination states. The mutually related intrazeolite hydrolysis, olation, and oxolation processes lead to the formation of oligomeric metal–oxo entities, particularly for the high valent cations. Besides the framework and extra-framework oxygen ligands (oxo-centers), reactant molecules (such as NO, NO₂, O₂, H₂O, and NH₃) can also be captured in the coordination sphere of encaged transition-metal ions in various modes of hapticity and stoichiometries. Therefore, the successful delineation of TMI speciation in metallozeolite catalysts requires the application of a broad suite of spectroscopic techniques. Nevertheless, due to common speciation effects, single spectroscopic techniques rarely provide a thorough insight into the complex structure of the intrazeolite TMI species in their particular framework locus, including valence and oxidation states. Thus, a combination of dedicated techniques is typically required for the successful resolution of the geometric, electronic, and magnetic structures of the active sites.^17–20

The chemical states of transition-metal ions and their agglomeration can be monitored with several conventional and commonly available spectroscopic techniques such as UV-vis/UV-vis-NIR,^23,29,30 EPR,^1,3,17 MAS-NMR,²⁴ IR,^25,26 Raman (RS) and resonance Raman (rR),²⁷ and luminescence (PL),²⁸ applied directly or with using suitable probe molecules.^31,32 More advanced methods include correlation 2D IR,³³ pulse EPR (HYSCORE,³⁴ ESEEM,³⁵ ENDOR³⁶), and high-field EPR,³⁷ as well as a toolbox of powerful synchrotron X-ray absorption techniques (XAFS, XANES, EXAFS³⁸), which have been of growing importance, recently.³⁹ In addition to the wide-ranging applications of these methods, various techniques have been developed to study zeolites in operando/in situ modes, providing valuable insights into intrazeolite coordination chemistry and elucidating the catalytic reaction mechanisms based on the establishment of functional structure–property–performance relationships. Operando characterization techniques enable monitoring of the adsorption and desorption of reactants, as well as the identification of surface intermediates and their dynamics. Additionally, they aid in understanding how and why zeolite catalysts behave differently under working conditions in real-life applications in comparison to static experiments.^18–20 However, the actual capacity of operando techniques to deliver definitive mechanistic insights is often limited by a compromise between the optimal circumstances for recording the high-quality spectra and the reaction conditions. The imposed constraints are particularly due to the temperature dependence of peak positions, line broadening effects, loss of the signal-to-noise ratio, and accumulation of spectator species. As a result, static and often low-temperature spectroscopic techniques still play a crucial role in providing high-quality reference spectra and reliable spectroscopic fingerprints, which can be used not only to guide the interpretation of operando results but also as crucial experimental constraints for ab initio molecular modelling.

Experimental studies on metallozeolites and their adsorption properties, as well as their catalytic chemistry, are consistently corroborated by theoretical modelling for the advanced interpretation of their intrinsic catalytic performance at both molecular (DFT, post-HF calculations) and phenomenological (ab initio thermodynamics) levels. Various quantum-chemical methods have been productively applied to characterise catalytically-relevant TMI systems, starting from commonly used and well-grounded density functional theory, progressing to more involved post-Hartree–Fock (post HF) methods, such as complete active space (CAS)⁴⁰ or coupled cluster (CC)^41,42 for proper treatment of correlated electron systems. Due to its overall good compromise between accuracy and computational cost, DFT remains the method of choice in computational catalysis. However, one should be aware of the limitations in currently available density functionals, which are approximated in their accuracy.^43–45 Examples of particularly challenging problems in computational catalysis, where DFT methods may not be sufficiently precise, include the prediction of chemically accurate adsorption energies and barrier heights in extended systems,^46–48 binding of noninnocent ligands to TMI,^49–51 and the spin-state splittings in mononuclear^52,53 and polynuclear complexes.^54,55 In such cases, a critical evaluation of the DFT results may be necessary, including cross-checking with the results of post-HF wave function calculations and comparing them with experimental spectroscopic or thermodynamic quantities, wherever available. When properly used, quantum-chemical calculations have been of crucial importance in developing a comprehensive molecular description of the coordination chemistry of TMI in zeolites, explaining their structure, dynamics, reactivity, and spectroscopic features, which are not directly accessible by experimental techniques. Supplemented by molecular simulation methods including molecular dynamics (MD) at various sophistication levels,^56–59 first-principles thermodynamics (FPT),^60,61 advanced methods of electronic structure analysis, including natural orbitals for chemical valence combined with extended-transition-state method (ETS-NOCV),^62–64 crystal orbital Hamiltonian population COHP,^65,66 and machine learning (ML) approaches,^67,68 they define a robust panoply of modern computational chemistry of zeolites.

In this review, we focus on the interfacial coordination chemistry of 3d transition-metal cations encaged in zeolites and its relevance for adsorption and catalysis, with the following topics that are covered: (1) accommodation and spectroscopic characterization of intrazeolite transition metal (TMI) centres (bare single, dual, and polynuclear metal–oxo), (2) survey of their structure and spectroscopic fingerprints, (3) first principles thermodynamic account of TMI speciation and stability at various p_O2 and p_H2O conditions, (4) coordination of extra-framework non-redox (H₂O, NH₃, N₂, CO) and redox (NO, O₂, N₂O) molecules and their activation with the specific accent on the role of electronic and spin state of the active sites and (5) thermodynamic and molecular orbital interpretation of the binding and activation processes. The emphasis is placed on the utility and diagnostic capacity of selected molecular spectroscopies, particularly CW-EPR and related pulsed variants, UV-vis, Raman/resonance Raman, and IR techniques commonly applied in such investigations, and on ab initio thermodynamic and molecular modelling, supplemented by spin-resolved analysis of the nature of metal–ligand bonds using natural orbitals for chemical valence (NOCV).

The examples discussed herein demonstrate the capacity of these spectroscopies to probe the interaction of the reactant molecules with intrazeolite first-row 3d transition metal ions, and unravel various ways of their activation, particularly when the spin of TMI plays a crucial role. In selected cases of special relevance, this review is expanded to include non-transition metal ions (such as Zn) dispersed in zeolites. However, the framework cations obtained by isomorphous substitution, which form corner-sharing MO₄ tetrahedra fully integrated with the zeolite structure through covalent framework M–O–Si bonds, are only scarcely considered. Lastly, we do not aim to provide a comprehensive review herein, but rather present illustrative examples that demonstrate the value of advanced concepts in establishing the rational basis of redox catalysis on metallozeolites, which complements the previous notable reviews devoted essentially to copper and iron.^{5,14,29,69–73}

2. Localisation, structure, and speciation of cationic metallic active sites

The unique structural and topological features of zeolites, such as 3D and 2D channel networks, accessible void space, space confinement, the presence of ion-exchange sites, and Brønsted acidity, make them versatile catalysts of enormous importance for both fundamental investigations and wide-ranging practical applications, particularly when functionalized with transition metal ions.^13,74–76

The introduction of transition-metal ions to zeolite matrices gives rise to four broad types of accommodation scenarios: (1) substitution of alien cations for the framework Al or Si ions, (2) formation of charge balancing ionic pairs with the framework O–Al⁻–O centres, (3) encapsulation of metal species and entities within the channels and cages, (4) support of nanoparticles on the external zeolite surface. Depending on the Si/Al and M/Al ratio, zeolite topology, Al distribution, and the preparation method used, a variety of confined species of various site topologies, structures, and nuclearity, such as single and dual bare cations, and a diversity of polynuclear metal–oxo species, can appear (Fig. 1).

Fig. 1 Selected preparation methods applied for synthesizing the metallozeolites, including one-pot synthesis (illustrated with TEM/EDX imaging of Cu in SSZ-13),^33,77 chemical vapour deposition (with EPR spectrum of Zn⁺ cations dispersed in ZSM-5),⁷⁸ dealumination and impregnation (with UV-vis of VO₂⁺ species incorporated in Si-BEA),⁷⁹ aprotic and ionic exchange (with H₂-TPR profiles of Ni/BEA),⁸⁰ dry impregnation (with Mössbauer spectra of iron in ZSM-5),⁸¹ and solid state exchange³³ (with TEM image of segregated oxide nanocrystals). The figure was drawn based on data published in the quoted references. The microscopic images were adapted from ref. 33 with permission from Elsevier, Copyrights 2023.

2.1. Isolated TMI centres

Hydrated metallozeolites are usually catalytically inert, but elevated temperature (prevailing during catalyst pretreatment and reaction conditions) drives off the labile ligands and generates coordinatively unsaturated active 3d metal (and/or metal-hydroxo and -oxo) centres. The resultant charge-balancing extra-framework cations hosted in the zeolites exhibit diverse low coordination states depending on their electronic configuration, charge, and spin (Fig. 2a). The topology of the sitting sites, size of the rings (Fig. 2b₁), Si/Al ratio, and Al spatial distribution (Fig. 2b₂ and b₃) play a decisive role in the accommodation mode of these cations.⁸² This issue is illustrated by various coordination patterns of the bare 3dⁿ cations in the MFI matrix, as well as for selected single and dimeric Cu and Fe cations hosted in the CHA and FER frameworks, which are derived from DFT structure optimization (Fig. 2b–d).⁸³ Generally, ligation to more basic Al–O–Si oxygen atoms is preferred over Si–O–Si.⁵ Furthermore, the Al pairs (i.e., Al atoms arranged in the Al–O–(Si–O)₂–Al sequences) are capable of stabilising M²⁺ cations, whereas the single Al atoms (present in the Al–O–(Si–O)_n sequence, n > 2) can charge-balance only the monovalent ions or hydroxylated divalent cations, M²⁺–OH⁻, and any other oxo species of an effective charge q = 1.⁸⁴ As a result, zeolite hosts with abundant Al–O–(Si–O)₂–Al sequences are needed to obtain catalysts with an enhanced content of the divalent redox sites. Accommodation of TMI cations leads to significant deformation of the rings where the alien cations are sited, and the associated lowering of the local site symmetry. In this way, the flexible zeolite framework acting as a polydentate ligand tends to maximize the coordination number (M_nc, n = 1, 2, 3 …) of TMI, yet at the expense of forming appreciably strained bonds.

Fig. 2 Exemplary (DFT optimized) coordination patterns of selected isolated TMI hosted in MFI frameworks with the illustrative spin density distribution for Co²⁺ cation in the high (S = 3/2) and low (S = 1/2) spin states (a). Coordination environments of single Cu²⁺ cations showing the effect of site topology (b₁), and Al distribution (b₂) and (b₃) on the TMI relative stability (based on data published in the ref. 82). Dimer Cu–Cu sites in CHA (c) and Fe–Fe sites in FER (d).

The resultant structural deformations are well reflected in the IR spectra of the T–O–T asymmetric skeletal vibrations (T = Si or Al).⁸⁵ The extent of this perturbation and the ensuing shift of the perturbed asymmetric T–O–T vibration varies with the valence state, size, particular exchangeable position the hosted ion occupies in the zeolite framework, and the presence of co-ligands as well.⁸⁶ The corresponding IR bands appear between two strong bands of antisymmetric and symmetric T–O–T vibrations, and are shifted from 1020–1100 cm⁻¹ to about 800–1000 cm⁻¹ upon accommodation of the alien cations. This effect can be exemplified by the Cu²⁺, Ni²⁺, and Mn²⁺ hosted in the FER matrix (Fig. 3a), and Cu²⁺, Cu²⁺–OH⁻ and Cu⁺ in ZSM-5 (Fig. 3b) and SSZ-13 (Fig. 3c).^86–88 In the case of small charge cations (Cu⁺, Cu²⁺–OH⁻ or Zn⁺), these deformations and the associated T–O–T frequency shifts are less pronounced (Table 1). The perturbed T–O–T vibrations can be used, for instance, to differentiate between the single Cu²⁺ species located in 6MR (900 cm⁻¹) and the Cu²⁺–OH⁻ centres in 8MR (∼950 cm⁻¹) rings present in the CHA zeolite.⁸⁸

Fig. 3 IR spectra of the perturbed T–O–T antisymmetric mode for dehydrated ferrierite substituted with Co (a₁), Ni (a₂), and Mn (a₃),⁸⁷ for ZSM-5 zeolite loaded with Cu²⁺ (b₁) and Cu²⁺/Cu⁺ (b₂),⁸⁶ and for Cu-SSZ-13 with the Si/Al ratio of 6 (c₁) and 30 (c₂).⁸⁸ The figure was drawn based on data published in the quoted references.

Table 1 Antisymmetric T–O–T vibrations (cm⁻¹) for selected cations accommodated in the FER, CHA, and MFI zeolites

Zeolite framework	FER		MFI		CHA		Ref.
Cation	α-site	β-site	α-site	β-site	σ-site	τ-site
Fe^2+	935–940	913–915		927			89 and 90
Co^2+	942–945	918–920	970	931–935	895	924, 948(τ^3Si)	23, 87 and 89
Mn^2+	953	928					89
Ni^2+	940	918		930			89
Zn^2+	935	915–917					84
Cu^+				977			86 and 89
Cu^2+				937	900–892		86 and 89
Cu^2+–OH^−						950	88

---

Direct insight into the ligation of TMI in zeolites can be derived using classic continuous wave CW-EPR^91–94 or advanced pulse and high-field techniques (HYSCORE, ESEEM, ENDOR).^95,96 For the series of Ti³⁺ (3d¹), V³⁺ (3d²), Cr³⁺ (3d³), Cr²⁺ (3d⁴), Mn²⁺, Fe³⁺ (3d⁵), Fe²⁺ (3d⁶), Co²⁺ (3d⁷), Ni²⁺ (3d⁸) Cu²⁺, Ni⁺ (3d⁹) and Zn⁺ (3d¹⁰4s¹) ions of the first row, which can be hosted in zeolites, all tend to assume the high-spin configuration, noting the position of the zeolite oxygen (O_z) ligands in the spectrochemical series (AlO) < (O_z) < (SiO) ∼ (H₂O) proposed by elswhere.⁹⁷ Among them, those with S = 1/2 are amenable to straightforward investigations using the commonly available X-band CW-EPR technique.

The hyperfine coupling A- and the g-tensor values obtained from the EPR spectra are strongly affected by the local environment of the encaged TMI. Apart from the 3dⁿ configuration, the g- and A-tensor anisotropies are sensitive to the local symmetry and the covalency of the M–O_z bonds with the zeolite framework oxygens. Simple perturbation theory treatment within the crystal field approximation allows for a straightforward (yet coarse) molecular interpretation of the g-tensor (g_ij) values of the S = 1/2 TMI hosted in zeolites, based on eqn (1):⁹⁸

For further discussion, upon a rudimentary account for the M–O_z bond covalency, gauged succinctly by the α coefficient (it accounts for the extent of delocalisation of the unpaired electron from the metal core into the ligands (Fig. 4d), for more details see ref. 99), this equation may be reframed in a simplified handy form as; g = g_e ± mλα²/(E_SOMO − E_n), where m = 2, 6, 8 (depending on the 3d orbital involved in SOMO – singly-occupied molecular orbital), g_e = 2.0023, λ is the one-electron spin–orbit coupling constant, and L_i, L_j are the orbital angular momentum operators that couple SOMO ≈ α|3d〉, with the |n〉 states of E_n energy. The requisite energy differences (E_SOMO − E_n) can be derived from the d–d transitions observed in the complementary UV-vis spectra, as illustrated in Fig. 4a and b. Their relation with the corresponding EPR spectrum is shown in Fig. 4c.

Fig. 4 UV-Vis-NIR spectrum of dehydrated Cu-SSZ-13 (a), together with the 3d energy levels (b) and the corresponding simulated EPR spectrum (c) (panels a, b, and c inspired by ref. 100). Structural model showing influence of a shift of spin density between Cu- and ligand-based states on the magnitude of hyperfine coupling gauged by α coefficient (d). The road map showing the origin of the off-axis features in the EPR spectrum of Cu²⁺ hosted in the CHA matrix (e), based on ref. 83.

The ligand field symmetry around the intrazeolite TMI in the hosting site imposes distinct constraints on the g-values. Axial signals with g_‖ = g_zz ≠ g_⊥ = g_x = g_y are expected for the local D_4h, C_4v, D_3h, C_3v, D_2d symmetries, whereas for lower symmetries (D_2h, C_2v, C_2h, C_s) orthorhombic (or monoclinic) signals with g_zz ≠ g_xx ≠ g_yy are anticipated. As a result, for a 3d¹ (Ti³⁺) cation with the d_xy SOMO in an exemplary tetragonal C_4v environment, g_⊥ < g_‖ < g_e is expected, and indeed experimentally observed, whereas for a 3d⁹ cation (Ni⁺), g_‖ > g_⊥ > g_e for d_x²−y² SOMO (Fig. 5a and b) and g_⊥ > g_‖ ∼ g_e for d_z² SOMO are typical. The hyperfine (hf) coupling constants for the unhybridized d_x²−y² SOMO can be expressed as A_zz = A_s + P[−4/7α² + Δg_zz] and A_yy = A_s + P[−4/7α² + Δg_yy], A_xx = A_s + P[−4/7α² + Δg_xx], where the α coefficient accounts again for the M–O_z bond covalency, P = g_eg_Nμ_Bμ_N/〈r³〉, neglecting smaller terms.¹⁰¹ Thus, the value of the α coefficient can be assessed from the analysis of the A-tensor, using the calculated atomic values of the P parameters, which are tabulated for a given TMI elsewhere.⁹⁹

Fig. 5 EPR spectra of Ti/MFI (a) Ni/BEA (b). EPR spectra of the dehydrated Cu-mordenite samples of various Si/Al ratios (Si/Al = 5, 15, 46) and Cu exchange levels (2Cu/Al = 38, 24, 70) (c). Room temperature EPR spectra of Cu⋯Cu pairs in CuCe/Y zeolite oxidized at 450 °C and evacuated at 600 °C (d). Blumberg–Peisach-type (P–B) plot for the EPR active Cu²⁺ species stabilized within different zeolite environments (e), (colours indicate various bare Cu²⁺ species hosted in different zeolite matrices. In the case of CHA, yellow and blue colours correspond to various Cu²⁺ and Cu²⁺–OH⁻ species). The corresponding spin density contours are provided for exemplary Cu²⁺/6MR and Cu²⁺–OH⁻/8MR structures. EPR spectrum of Cu-SSZ-13 zeolite activated at 250 °C in vacuo (f), resolved into the contributing signals. The A_‖ and g_‖ values for the species identified in (f), labelled as (f_1–4), are shown in the P–B plot as CHA1–4. The figure was drawn based on data from ref. 80, 83, 91, 92, 104 and 105 respectively.

More ionic bonds of the encaged TMI with the framework O_z (small spin density delocalisation) increase the hyperfine A_Metal values on the metal core and decrease the superhyperfine splitting on the ligands with magnetic nuclei (A_Ligand), such as ¹⁷O (I = 5/2) labelled O_z framework oxygen atoms, and vice versa (see Fig. 4d). In the latter case, the superhyperfine patter due to ¹⁷O_z can be used to ascertain the coordination environment of intrazeolite TMI (see below, Fig. 6).

Fig. 6 Experimental (black) and simulated (red) Q-band ¹⁷O ENDOR spectra, recorded at different magnetic field settings, of ¹⁷O isotopically enriched and fully dehydrated Cu-CHA (a₁), together with the schematic representation of the isotopic enrichment of the zeolite framework (a₂). Comparison of experimental (grey) and computed (red) maximum hyperfine coupling values (A_max) for ¹⁷O and ²⁷Al nuclei for the three different Al distributions: 2Al-3NN, 2Al-3NN, and 1Al (b) (the black dashed lines represent the experimental range of A_max for ¹⁷O nuclei). Adapted from ref. 95 under the terms of the CC-BY 4.0 license, copyright 2021, published by Springer Nature.

A comprehensive listing of the equations that can be used for molecular interpretation of the g and A tensors of TMI in ligand fields of various symmetries can be found elsewhere.¹⁰² It should be noted, however, that except for the regular features along the z and x, y directions, shown in Fig. 4c, additional off-axis lines may appear in the powder EPR spectra when

2A_i² − hvA_i/m_I < (g_i²A_i² − g_j²A_j²)/(g_i² − g_j²) < 2A_j² − hvA_j/m_I (2)

where v is the microwave frequency and m_I is the nuclear magnetic spin number (Fig. 4e).¹⁰³ These extra features, which correspond to additional extrema in the angular dependence of the magnetic resonant field (∂B/∂θ = 0 and ∂B/∂φ = 0), may exhibit significant intensity in the copper EPR spectra and should not be confused with the speciation of copper. Broadening of the EPR signals caused by local strain (structural inhomogeneity of the siting environment) and the presence of abundant magnetic ²⁷Al (I = 5/2) atoms for zeolites of low Si/Al ratio, along with the high TMI loading, lead to loss of the hyperfine structure A and g-anisotropy, which can be illustrated by the EPR spectra of copper hosted in mordenite of various Si/Al (5, 15, 46) ratios, and copper exchange (2Cu/Al × 100% = 38, 24, 70) levels (Fig. 5c), where the perpendicular hyperfine structure of copper remains unresolved, and the parallel hyperfine structure becomes barely visible.⁹¹ Furthermore, at higher loadings, the decreased distance between the copper cations may result in Cu²⁺ pairs formation, which is manifested in the EPR spectrum by the appearance of additional fine structure features on both sides of the perpendicular component of the single-ion signal of Cu²⁺, separated by D/g_⊥μ_B.¹⁰⁴ They can be used for assessment of the distance between the two adjacent copper ions (D/cm⁻¹ ≅ 0.433g²/r³ for r in Å) when the direct exchange interactions (∼J) are small. The presence of a half-field signal due to the forbidden ΔM_s = 2 transition and the appearance of 2I + 1 = 7 lines of the hyperfine structure with half-splitting are other important characteristics of the Cu dimer formation (Fig. 5d).

The M–O_z covalency depends primarily on the presence of Al in the coordination environment of the encaged TMI. The framework oxygen atoms in the Si–O_z–Al entities are more negatively charged than those in the Si–O_z–Si motives. These factors make g and A values structure sensitive, which can be illustrated concisely in the form of Blumberg–Peisach-type diagrams¹⁰⁶ for Cu²⁺ in various zeolite frameworks and different Si/Al ratios.^92,107 Although the application of such a diagram to zeolites, due to the absence of suitable model compounds, requires some caution, it demonstrates its usefulness for the preliminary sorting of the intrazeolite copper species based on their EPR fingerprints.

Overall, copper ions in the more positively charged environments are found in the lower right part (smaller A_‖ and bigger g_‖), whereas Cu²⁺ species in the more negative coordination environments are located in the upper left end (bigger A_‖ and smaller g_‖) of the trend line in the Blumberg–Peisach-type diagram (Fig. 5e). This is well exemplified by the different positions in the plot of two types of the Cu-MFI zeolites, with two Al atoms in a MR6 ring generating a higher charge (MFI2) and with only one Al atom in the MR6 ring with the resultant lower charge (MFI1).⁹² For the particular case of Cu-CHA, the EPR parameters for the MR6 ring (σ-site) fall in the middle of the trend line. The scattering of the positions in the plot reflects the sensitivity to the local symmetry of copper ligation (A_‖(C_2v) > A_‖(C₁)), yet the position of the Al atoms within the 6MR appears to be of less importance. The presence of a hydroxy OH⁻ co-ligand in the coordination sphere of Cu²⁺ located in the MR8 (τ-sites, see CHA5 in Fig. 5e) considerably lowers the value of the hyperfine structure, providing a straightforward means for the distinction between the bare Cu²⁺ and hydroxylated Cu²⁺–OH⁻ species.

As a result, the structural sensitivity of the g and A values can not only reveal the speciation of the TMI hosted in zeolites, but also provide a rational basis for unambiguous site-specific assignment, particularly using the benchmarking parameters calculated by DFT or post-HF methods. This is well illustrated for the most extensively investigated Cu²⁺ hosted in CHA, MFI, and BEA zeolites, as shown in Table 2.

Table 2 Experimental and calculated EPR spin-Hamiltonian parameters for selected TMI in zeolites

TMI–zeolite	Notes^a	g tensor	A tensor/MHz	Ref.
a Calculated when the quantum chemical method is provided; otherwise, the g and A tensors are the experimental values.b Cluster models with different arrangements of Al sites.c Only one value (gzz, Azz) reported.d gx,y (g⊥) and Ax,y (A⊥) are not reported (for zeolites typical values of g⊥ ∼ 2.07 and A⊥ ∼ 30–40 MHz are structurally less informative than the g‖, A‖ values).
Ti^3+(3d^1)/TS-1	Experimental	1.922; 1.939; 1.9897		108
		1.981; 1.960; 1.91
Ti^+/(3d^1)/ZSM-5	Experimental	1.94; 1.97		109
VO^2+(3d^1)/ZSM-5	Experimental	1.9843; 1.9843; 1.931	214; 214; 542	37
VO^2+(3d^1)/BEA	Experimental	1.987; 1.983; 1.927	234; 214; 558	110
	B3LYP calculated, 6MR	1.987; 1.984; 1.952	156; 139; 440
VO2(3d^1)/SiBEA	Experimental	1.803; 1.942; 1.968	676; 892; 876	110
	B3LYP calculated, 6MR	1.853; 1.992; 1.998	838; 1124; 1043
Cr^5+(3d^1)/Na-Y	Experimental, supercage site II	1.98; 1.98; 1.92		111
	Small cage site I or II	1.98; 1.98; 1.92
Cr^5+(3d^1)/Na-MOR	Experimental	1.99; 1.98; 1.93		111
Fe^3+(3d^5)/ZSM-5 (Fe^3+OH/ZSM-5)	Experimental	geff = 4.3		112
		geff = 6.3; 5.7
Fe^3+(3d^5)/BEA (Fe^3+OH/BEA)	Experimental	geff = 4.3		112
		geff = 6.1
Co^2+(3d^7)/MFI	Experimental, S = 3/2	geff = 5.1		113
Co^2+(3d^7)/AlPO		geff = 5.80–5.44; 2.0		119 and 120
Ni^+/(3d^9)/ZSM-5	Experimental	2.49; 2.072; 2.072		64
Ni^+(d^9)/ZSM-5	B3LYP calculated, 6MR	2.337; 2.112; 2.159		114
Cu^2+(3d^9)/ZSM-5	Experimental	2.323^d	471^d	115
Cu^2+(3d^9)/ZSM-5	Experimental	2.276^d	545^d	115
	Calculated, 6MR (α site)	2.230^d	479^d
	6MR (β site)	2.192^d	497^d
	6MR (γ site)	2.206^d	531^d
Cu^2+(3d^9)/CHA	Experimental	2.352; 2.075; 2.075	470; 35; 35	116
	Experimental	2.347^d	493^d	115
Cu^2+(3d^9)/CHA	Experimental	2.325; 2.072; 2.072	490; 35; 35	116
	B2PLYP calculated, 6MR	2.282^c	455^c
Cu^2+OH(3d^9)/CHA	Experimental	2.290; 2.072; 2.072	410; 45; 45	116
	B2PLYP calculated, 8MR^b	2.275, 2.135, 2029	388^c
Cu^2+OH^− (3d^9)/CHA	Calculated, 8MR^b, B3LYP	2.194^d	203^d	115
	6MR^b, B3LYP	2.200^d	401^d
	6MR^b, B3LYP	2.242^d	492^d
	D6R, B3LYP	2.243^d	368^d
Zn^+/(3d^104s^1)ZSM-5	Experimental	1.9951; 1.9984; 2.0015	1520, 1520, 1535	78
	B3LYP calculated for:
	T7 site	1.992, 1.9963, 2.0023	1321, 1325, 1347
	T8 site	1.9946, 1.9975, 2.0023	1458, 1460, 1480
	T10 site	1.9921, 1.9965, 2.0023	1331, 1335, 1357

---

As an example, the diagnostic A_‖ region of the EPR spectra of CuSSZ-13 shows pronounced speciation of copper (Fig. 5f)⁸³, which has been resolved into two tetracoordinate Cu²⁺ cations of the local C₁ and C_2v symmetry in the σ-rings (marked as CHA-3 and CHA-4 in Fig. 5e, respectively), and two types of the Cu²⁺OH⁻ species hosted in the τ-sites of the CHA matrix (CHA-5 and CHA-6 in (Fig. 5e).

The local structure of single-site TMI cations hosted in zeolitic structures, and quantification of the covalency of the Cu–O_z bonds with an accuracy unattainable for other site-sensitive methods, can be ascertained using advanced EPR techniques combined with the ¹⁷O isotopic labelling of the zeolite framework.⁹⁵ Of particular note here is the study of the CuSSZ-13 zeolite using ¹⁷O, ²⁷Al ENDOR spectroscopy (Fig. 6a₁), supported by DFT modelling.^95,117 The requisite 17-O labelling of the zeolite framework can be achieved quite readily using isotopic exchange with H₂¹⁷O (Fig. 6a₂).

In contrast to the metal A values, the ¹⁷O superhyperfine couplings depend significantly on the Al distribution in the 6MR ring, allowing for unambiguous discrimination between various Al tetrahedra locations. Their presence and number can be directly probed by ²⁷Al HYSCORE or ENDOR spectroscopies. In particular, the σ-rings characterized by 2 Al atoms separated by two Si (2Al-3NN) or one Si atom (2Al-2NN), as well as the rings with one Al atom only, exhibit characteristic patterns of the ¹⁷O and ²⁷Al superhyperfine coupling constants (Fig. 6b), providing the basis for the Al-resolved identification of hosting sites. The exquisite sensitivity of such couplings enables, therefore, the univocal identification of the Cu²⁺ coordination environments. This approach has also been used to examine the location of VO²⁺ (ref. 95) and Zn⁺ in ZSM-5,¹¹⁷ demonstrating that it is possible to ascertain the specific location of paramagnetic TMI active centres in zeolite matrices with the level of detail needed for the rational establishment of structure–reactivity relationships. It is worth mentioning here that anomalous X-ray powder diffraction at the K-edge, which probes the long-range order of aluminium atoms within the zeolite frameworks, can also be used to determine isolated Al atoms and pairs (i.e., the Al(–O–Si–O–)_xAl sequences). Using this approach in combination with ²⁷Al and ²⁹Si MAS-NMR, the spatial distribution of Al atoms in ZSM-5 has recently been resolved.¹¹⁸

The TMI species with S > 1/2 (with more than one unpaired electron) exhibit zero-field splitting of the ground state gauged by the axial D and rhombic E parameters. The high-spin TMI can be divided into Kramers ions of half-integer spin numbers (S = 3/2, Co²⁺ and S = 5/2, Fe³⁺) and non-Kramers ions with integer spins (S = 1, Ni²⁺ and S = 2, Fe²⁺, Mn³⁺). The zero-field splitting can make the intrazeolite non-Kramers TMI “EPR-silent” using conventional X-band (∼0.3 cm⁻¹) or Q-band (∼1.2 cm⁻¹) EPR, or detectable, particularly for the Kramers ions, when the ZFS leads to a well-separated M_s = ±1/2 state as the lowest one (Fig. 7a and b₁).^119,120 Such systems can be treated simply as exhibiting effective S′ = 1/2 states, but due to closely lying energy states that promote fast relaxation, registration of the EPR spectra usually requires liquid helium temperatures.^120,121 The zero-field splitting depends on the anisotropy of the local ligand field and is mediated by spin–orbit coupling (SOC). The magnitude of the latter increases as one passes from early to late 3d TMI across the periodic table. As a result, ZFS may vary from small values (of the order of 10⁻² cm⁻¹) for Mn(II) in symmetric environments to large values of the order of even 10² cm⁻¹ for Co(II).¹¹⁹ Notable examples can be provided by liquid helium X/Q-band EPR studies of CoZSM-5 (Fig. 7b₂)¹²¹ or CoAPO zeotypes with the Kramers M_s = ±1/2 ground state, which exhibit effective g-values of g_⊥,eff ≈ 5.80–5.44 and g_‖,eff ≈ 2.00, and the zero-field splitting of 7 cm⁻¹. Another example of a Kramers high-spin system is provided by FeZSM-5 zeolites.¹²² The EPR signal at g_‖,eff ≈ 2 with small ZFS has been associated with the highly symmetric α sites. In contrast, the signal at g_⊥,eff = g_⊥(S + 1/2) ≈ 6 and g_‖,eff ≈ 2 to β sites of local axial symmetry (D ≫ hν, E ≈ 0), whereas the g_eff ≈ 4.3 signal to γ sites with rhombic ZFS (D ≫ hν, and D/E ≈ 1/3) (Fig. 7c₁ and c₂). It is more likely, however, that the trivalent ions are actually encaged in the ZSM-5 host in the form of hydroxo or oxo species, which can be more readily charge balanced by 2 Al and 1 Al atoms, respectively, since the configurations of 3 Al atoms per ring/cage are available in zeolites of low Si/Al ratio, such as X faujasite, only. A broad signal appearing at g_eff ≈ 2.3 corresponds, in turn, to small Fe_xO_y clusters. The extension of the low-field signal to g_eff ≈ 9.6 can mainly be attributed to the rhombicity D/E parameter, which characterizes the Fe³⁺ cations in highly anisotropic (low-symmetry) environments. A signal at g ∼ 3.3 may also appear for Fe³⁺ in the sites of trigonal symmetry, but some spectral features can be smeared out due to the strong dependence of their resonance magnetic fields on the spatial orientation of the paramagnets, or distribution of the ZFS values caused by local strain.

Fig. 7 Splitting of the energy levels in the magnetic field for exemplary S = 3/2 and S = 1 states of TMI (a), zero-field splitting (ZFS) for the ⁴A₂ state of Co²⁺ (d⁷) with D > 0 (b₁), with the corresponding Q-band EPR spectrum of Co/AlPO (b₂).^119,120 X-band spectra of Fe/ZSM-5 (c₁ – low field, c₂ – high field component),¹²² Mn/BEA of different loading (d),¹²³ and Q-band spectrum of Mn²⁺ in A zeolite matrix (e).¹²⁴ The figure was drawn based on data published in the quoted references.

In the case of Mn²⁺ ions exchanged into zeolite hosts, the EPR spectra show gradual smearing of the 6-line hyperfine structure due to I = 5/2 with an increasing Mn content due to dipolar broadening effects caused by the adjacent Mn²⁺ ions (Fig. 7d).¹²³ The observed spectrum results from transitions within the M_s = ±1/2 doublet produced by the combined action of the ligand field and the spin–orbit coupling.

In fact, in powder spectra only the central line due to the M_s = −1/2 ↔ +1/2 transitions of the least angular dependence on the magnetic field is strong, whereas the other allowed inter-doublet transitions (such as M_s = −±3/2 ↔ ±1/2 or −±3/2 ↔ ±5/2, marked with the grey stick diagrams in Fig. 7e) are smeared to large extent, and barely observed as mere wings flanking the central hyperfine sexted. A relevant example is provided by Mn²⁺ exchanged in A zeolites.¹²⁴

2.2. Terminal and bridging metal oxo centres

Besides the bare state (Co²⁺, Ni²⁺/Ni⁺, Cu²⁺/Cu⁺), extraframework cations, particularly those of z ≥ 3 (V⁵⁺, Ti⁴⁺ or Fe³⁺) hosted in zeolites of moderate Si/Al ratio, can be stabilised in the form of hydroxo and terminal or bridging oxo TMI centres of various nuclearity. Their appearance results from hydrolysis when the local pH during the catalysts preparation is too high ([M(H₂O)_n]^z+_(cage) + hH₂O_(cage) → [M(OH)_h(H₂O)_n−h]^(z−h)+_(cage) + hH₃O⁺/Z), launching the intrazeolite olation/oxolation reactions that lead to the formation of M=O or O=M=O oxo-species (for instance, [V(H₂O)₆]⁵⁺ → [V(OH)₄(H₂O)₂]⁺ → [VO₂(H₂O)₄]⁺ + 4H₃O⁺ (ref. 125)). The relevant thermodynamic background of these processes is provided in Chapter 3. Notable examples of the intrazeolite terminal oxo-systems of high valent metals include mono-oxo ([V=O]²⁺/BEA, [W=O]³⁺/ZSM-5¹²⁶ or [Fe=O]²⁺/MOR,²⁶ [Fe=O]²⁺/CHA, [Fe=O]²⁺/BEA,¹⁹ and di-oxo species ([MoO₂]²⁺/ZSM-5-¹²⁷ and VO₂/BEA zeolites¹¹⁰). Alternatively, metal–oxo species can also be formed during the redox-type interactions of bare (single and adjacent) TMI cations with O₂ or N₂O, and this chemistry is described in detail in Chapters 4.3.1 and 4.3.3.

Generally, the metal–oxo bonding of TMI can be described in terms of σ- and π-type overlap between the 2p orbitals of the O_t moiety with the corresponding 3d orbitals of the M^z+ core. A tetragonal environment (for clarity, with enhanced C_4v symmetry) with the M–O_t bond placed along the z-axis can be used as a basic, semiquantitative model for the TMI situated in the 6MR rings. In such case, the d_z² orbital is involved in the σ-overlap with the 2p_z(O) in the M–O_t bond, the d_zx, d_zy orbitals participate in the π-overlap, whereas the d_x²−y² orbital is primarily engaged in the σ-bonding to the framework O_z ligands (assuming the x- and y-axes intersecting O_z ligand positions), and in a considerably weaker δ-type interaction with the terminal oxygen, as well. The d_xy orbital remains essentially nonbonding. The resultant general scheme of the energy levels is shown in Fig. 8 (note, it may be altered by changes in the local site symmetry and oxidation state of the cations). All bonding oxo orbitals are occupied by the 6 electrons provided by the terminal O_t moiety. In contrast, the nonbonding d_xy and the antibonding , δ* and σ* orbitals are occupied by the electrons derived from the metal dⁿ centres. Thus, for the d⁰–d¹ TMI (exemplified by [VO₂]⁺ and [VO]²⁺ hosted in BEA²⁴ VO₂⁺/MCM-22¹²⁸ or [VO]²⁺ in ZSM-5³⁷), the M–O_t bonds acquire a distinct triple bond character (1σ and 2π), and are the strongest among the 3d series. For [VO]⁺/ZSM-5,¹²⁹ the bond order decreases to 2.5, as in the high-spin d² configuration, the antibonding becomes occupied.

Fig. 8 Qualitative energy level diagrams for 3d [MO]ⁿ⁺ cores hosted in a 6M ring containing two [AlO₄]⁻ sites representing the evolution of the metal–oxygen bond nature with the increasing number of d electrons.

Distinction between the diamagnetic [VO₂]⁺ (3d⁰) and paramagnetic VO₂ and VO²⁺ (3d¹) cations, accommodated, e.g., in the BEA zeolite, can be performed by MAS-NMR (Fig. 9a) in combination with photoluminescence techniques (Fig. 9b₁ and b₂),¹³⁰ and EPR spectroscopy (Fig. 9c1–c3),^24,110 respectively. The resolved fine structure in the PL spectra of [VO₂]⁺/BEA reveals directly not only the V=O bonds, but also speciation of the vanadium(V)-oxo species (Fig. 9b2).

Fig. 9 Exemplary NMR (a) and photoluminescence spectra of vanadium(V)-oxo species in BEA zeolite (b₁) and (b₂), (β, σ, γ indicate three vanadium species of various structures), based on data published in the ref. 130 redrawn and modified for clarity. Simulated and experimental EPR spectra of vanadium in the dealuminated BEA zeolite (c₁) and the component signals of VO²⁺ (c₂) and VO₂ (c₃) species, together with spin density contours calculated for the VO₂ and VO²⁺ species (c₄), adapted from ref. 110 with permission from American Chemical Society, Copyright 2007. Experimental (black) and simulated (red) EPR spectra of the VO²⁺/ZSM-5 sample registered at (d₁) X-band, (d₂) Q-band, and (d₃) W-band, adapted from ref. 37 with permission from Elsevier Inc., Copyright 2020.

The superimposed EPR spectra of the paramagnetic VO²⁺ and VO₂ species observed in the dealuminated BEA zeolite are shown in Fig. 9c1 and the component signals in Fig. 9c2 and c3, whereas the associated structures with spin density contours are shown in Fig. 9c4.²⁴ Different SOMO nature of both species (d_x²−y² for VO²⁺ and 3d_z²/4s hybrid for VO₂) is well reflected in the dramatically different EPR parameters: g_xx = 1.987, g_yy = 1.983, g_zz = 1.927, |A_xx| = 8.4, |A_yy| = 7.7, and |A_zz| = 20.7 mT for VO²⁺, and VO₂: g_xx = 1.803, g_yy = 1.942, g_zz = 1.968, |A_xx| = 26.8, |A_yy| = 32.8, and |A_zz| = 31.8 mT. The isotropic hyperfine constant is extremely sensitive to the number of oxo-ligands, and therefore, can be used as a simple diagnostic for resolving the oxo vs. dioxo dichotomy of the vanadium(IV) species. Application of multifrequency EPR illustrates, in turn, a possibility of separation of the spectral A_‖(A_zz) and A_x,y(A_⊥) features of the VO²⁺ entities exchanged in the ZSM-5 zeolites, and may be helpful to reveal their possible speciation (Fig. 9d1–d3).³⁷

With an increasing electron count, the antibonding orbitals become progressively populated (Fig. 8), decreasing the M–O bond order and weakening its strength. In the particular case of 3d⁴ cations, the bond order drops to 2. The resultant double bond in the M=O_t unit can be represented by a quintet [Fe=O]²⁺ ferryl species (d_xy¹d_xz¹d_yz¹(d_z²−y²)¹ σ², S = 2, Fig. 8) stabilized in various zeolite matrices,^90,131,132 which is featured by the presence of a low-lying ferric-oxyl septet (d_xy¹d_xz¹d_yz¹(d_z²−y²)¹(d_z²)¹, σ¹, S = 3),²⁰ or by isoelectronic [MnO]⁺ units hosted in ZSM-5 and SSZ-39 (AEI) modelled by DFT.¹³³ Yet, both [Fe=O]²⁺ and [MnO]⁺ species hosted in the MFI and AEI topologies exhibit different sitting locus, the structure and stability of which depends considerably on the distribution of Al atoms within the rings. The most stable [Fe=O]²⁺ species in ZSM-5 exhibit a square pyramidal (coordination number CN = 5) structure located in δ-sites of the 6MR(T11/T11) rings with the Al atoms in the para-position, whereas [MnO]⁺ species are most stable in the 10MR (T12) rings, assuming a trigonal structure of CN = 3,¹³⁴ characteristic for intrazeolite monovalent species such as [CuO]⁺ or [ZnO]⁺.

The [Fe=O]²⁺ entity is known as α-oxygen and exhibits unique reactivity toward C–H bond cleavage, which is addressed in Section 4.3.3. Particularly, for the BEA topology, the spin Hamiltonian parameters (S = 2 and the sign and magnitude of D = 13 ± 1 cm⁻¹) confirm the [⁵Fe(IV)=O]²⁺ formulation of this species, definitely.¹³⁵

In the case of the late intrazeolite transition metal cations (Co, Ni, Cu), the high-valent states (z ≥ 3) become more hardly accessible. With an increasing electron number, the corresponding M–O_t bonds gradually lose their oxo character. For the transition [Fe–O]⁺ (d_xy¹d_xz¹d_yz¹(d_z²−y²)¹(d_z²)¹ σ², S = 5/2) and [Co–O]⁺ (d_xy²d_xz¹d_yz¹(d_z²−y²)¹(d_z²)¹ σ², S = 2) species, the bond order (bo) drops to 1.5, decreasing further upon passing to Ni²⁺–O (bo = 1) and Cu²⁺–O (bo = 1/2). Since the π(3d_zx,yz–2p_x,y) overlap is significantly decreased, such entities acquire a distinct metal–oxyl (M–O˙) character with the high-spin metal core magnetically coupled to the oxyl-moiety.¹³⁶ Elongation of the M–O_t bond, e.g., in the transition state while interacting with reactants, weakens the 3d_π–2p_π overlap, favouring the highly reactive oxyl state.¹³⁷ Finally, in the case of the closed 3d¹⁰ shell the 3d–2p bond order ceases entirely, and, e.g., a [Zn^II–O]⁺ bonding is realized essentially via the 4s–2p overlap,¹³⁴ exhibiting a distinct zinc–oxyl nature ((3d)¹⁰(σ)²(O2p_x)²(O2p_y)¹(σ*)⁰), with the nearly degenerate non-bonding 2p_x (HOMO) and 2p_y orbitals (SOMO), see Fig. 8.

Summarizing, vanadium, chromium, or molybdenum terminal oxo-species are relatively rigid and disposed to protonation¹³⁸ or an electro-protic activity (M^z+=O + HR → M^(z−1)+–OH + R), which is mainly observed in the photocatalytic reactions, where the reactive excited oxyl state, M^z+=O → M^(z−1)+–O˙⁻, becomes accessible by photon (hν) absorption^139–141 or at elevated temperatures (typically 400–500 °C) for thermal catalytic processes.^142,143 In the case of Fe, Co, Ni, and Cu, the terminal oxygen functionalities are more chemically labile (unstable), and featured by the co-existence (or close lying and thermally available) electromeric M^z+–O²⁻ ↔ M^(z−1)+–O˙⁻ forms.¹³⁷ They may appear as important intermediates, particularly involved in low-temperature catalytic redox reactions (150–350 °C), entailing hydrogen or oxygen atom transfer processes, exemplified by CH₄ hydroxylation^135,144 or appear during N₂O decomposition that requires somewhat higher temperatures,¹⁴⁵ see Chapter 4.3.3 for more details.

A brief survey of typical spectroscopic Raman, IR, PL, and UV-vis techniques used for the characterization of terminal Cr-, Mn-, Fe-, Co-, Ni-, Cu–oxo, and Zn–oxyl species encaged in various zeolitic matrices is shown in Fig. 10a–g.

Fig. 10 Survey of spectroscopic techniques used for probing the intrazeolite metal–oxo species. Raman spectra for the Cr/ZSM-5 (Si/Al = 15) catalyst (a).¹⁴⁶ UV-Vis spectra of iron in CHA and BEA (b),¹⁴⁷ FTIR spectra of Ni-, Mn-, and Co-ferrierite samples showing perturbation of the region of T–O–T vibrations by the corresponding oxo-species (α-oxygen), illustrated here by distant binuclear centers (c),¹⁴⁸ UV-vis-NIR vibronic absorption spectra of the isotopically labelled Zn^II–O˙⁻ complex formed in the MFI-type zeolite framework (d).¹³⁷ In situ UV resonance Raman spectra of the high-temperature-treated Fe/ZSM-35 samples with different Si/Fe ratios (e).¹⁴⁹ In situ steady-state Raman spectra of Cu/CHA-I (top) and Cu/CHA-M (bottom) after activation with ¹⁶O₂ and ¹⁸O₂ and the corresponding structures along with the simulated Raman spectra (f).¹⁵⁰ In situ UV-vis spectra of Cu/MOR during activation in O₂ at 200 °C and the corresponding difference spectra (blue) obtained by subtracting the initial spectrum (g).¹⁵¹ The figure was based on data from the quoted references.

Apart from providing the spectroscopic fingerprints for identification and monitoring of the investigated species during catalytic reactions, these examples also illustrate the structure sensitivity of the particular signals associated with the mono-oxo and di-oxo states (Fig. 10a–f). It includes the effect related with the Al content in the rings (Fig. 10a), local symmetry distortion of the 6MR rings in the CHA vs. BEA topologies (Fig. 10b), perturbation of T–O–T vibrations by various mono-oxo Mn, Fe, Ni species (Fig. 10c), and the unique structure features of the Zn–oxo entities (Fig. 10d). In the case of bridging-oxo species (vide infra) the effect of the Si/Fe (Fig. 10e), and influence of the Al atoms distribution within the 6MR rings of the CHA framework on the Raman features are shown in Fig. 10f. Formation of the trinuclear coper-oxo clusters, [μ-CuO]₃²⁺ accommodated in mordenite can be monitored, in turn, by UV-vis spectroscopy (Fig. 10g). However, it is important to critically evaluate many of the structural assignments of such species reported in the literature.

As already mentioned above, a nucleophilic character (negative partial charge) of the TMI–OH ligands triggers interazeolite olation and oxolation processes, which lead to the appearance of various di- and tri- and even tetranuclear species containing the bridging oxygen moieties (see Chapter 3). The structures of homo-metallic bridging μ-oxo, and bis(μ-oxo) species of various nuclearity for selected cations (Fe, Ni, Co, Cu, Zn) are shown in Fig. 11a–c. Examples of heterometallic M₁–(μ-O)–M₂ analogues (predicted by DFT) include [Cu(μ-O)Zn]²⁺ in MFI and FAU zeolites¹⁵² or [Cu(μ-O)Ag]²⁺ entities hosted in ZSM-5.¹⁵³ The nearest environment (sitting topology) exhibits a noticeable effect on their stability, as illustrated by the μ-oxo [Fe(μ-O)Fe]²⁺ species, and even more markedly by the bis(μ-oxo) [Fe(μ-O)₂Fe]²⁺ species accommodated in the α, β, γ, δ positions of the ZSM-5 zeolite (Fig. 11d1 and d2).¹⁵⁴

Fig. 11 Structure of selected oligomeric oxo-species (derived from DFT optimisation). Mono-oxo species (a), dual oxygen species (b), and oligomeric M_xO_y species with x ≥ 3 (c).^{18,149,155–159} Relative stability of the μ-oxo [Fe(μ-O)Fe]⁴⁺ (d₁), and the bis(μ-oxo) [Fe(μ-O)₂Fe]²⁺ (d₂) entities accommodated at the α, β, γ, δ positions of ZSM-5 zeolite.¹⁵⁴ The figure was based on data from the quoted references.

It should be emphasized that univocal spectroscopic identification of the particular oxo-entities in the presence of speciation, which is a typical case, is a challenging endeavour. Overlapping of signals, insufficient structure sensitivity, and poor resolution are typical impediments. These can be exemplified by Fe/ZSM-5 zeolites, which may exhibit UV-vis bands virtually identical to those of isolated and dimeric oxo iron sites when the antiferromagnetic coupling is weak.¹⁶⁰ Therefore, the application of site-selective spectroscopies such as variable-temperature and variable-field magnetic circular dichroism (VTVH-MCD), resonance Raman, EXAFS, or advanced multifrequency EPR techniques, especially when combined with isotopic labelling, provides the most reliable information on metallozeolites regarding the actual nature of the active sites.⁵

Apart from the electronic structure, magnetic interactions may also exert an appreciable influence on the catalytic activity of the oxo-species. Therefore, their spin state and the structural and orbital factors that determine the magnetic exchange coupling between the TMI should also be examined for a thorough understanding of the structure–reactivity relationship. This issue has been recently addressed in the case of bridging [Cu(μ-O)Cu]²⁺ cores hosted in CHA and MFI zeolites.¹⁶¹ The dissimilar spin states, antiferromagnetic in MFI and ferromagnetic in CHA, revealed by variable-temperature/variable-field magnetic circular dichroism spectroscopy, exhibit different reactivities in CH₄ hydroxylation. Coupling between both Cu²⁺ (S = 1/2) cations of the intrazeolite [Cu(μ-O)Cu]²⁺ species leads to an antiferromagnetic (S = 0) state in Cu-MFI (Fig. 12a) and ferromagnetic (S = 1) ground state in Cu-CHA (Fig. 12b). Surprisingly, it does not dependent on the Cu–O–Cu angle (as it usually is observed in homogeneous complexes), but is governed by the mutual arrangement of the bidentate (π-type ligation) moieties linked to the two copper cations (AlO–Cu–OAl). Therefore, the zeolite framework topology imposes unique spin patterns in the [Cu(μ-O)Cu]²⁺ cores, resulting in a distinctly different magnetic structure and reactivity through the rigid, stereo-specific bidentate ligation effect.¹⁶¹

Fig. 12 Structural features of the Cu–O–Cu entities that lead to an antiferromagnetic coupling in MFI (a) and a ferromagnetic coupling in CHA (b), based on data from ref. 161. Broken symmetry picture (UKS B3LYP/SOMF TZV/CP(PPP) calculations) of the magnetic structure of the [Fe(μ-O)₂Fe]²⁺ species hosted in ZSM-5 zeolite matrix (c).

Molecular orbital interpretation of the magnetic structure of the exchange-coupled oxygen-bridging cores can be analysed with the DFT broken symmetry (BS) calculations, taking as an example the [Fe(μ-O)₂Fe]²⁺ species hosted in ZSM-5 zeolite.¹⁶² The relevant magnetic orbitals are visualized in Fig. 12c. The α/β pairs of magnetic orbitals shown in Fig. 12c constitute symmetric magnetic exchange pathways between the 3d Fe states via the in-plane σ–π overlap (orbitals 1, 2, 4), and the out-of-plane π–π overlap (orbitals 3 and 5) mediated by the 2p oxo orbitals. Such pathways indicate that the antiferromagnetic coupling is realized by the superexchange interactions via both μ-oxo bridges.¹⁶³ The spin coupling parameter J can be calculated within the broken-symmetry approach using, for example, the E_HS − E_BS = −(2S₁S₂ + S₂)J equation, which has been demonstrated to provide more accurate coupling constants for DFT calculations.¹⁶⁴ The J values for several dinuclear oxo and hydroxo species hosted in the 6MR and 8MR rings of SSZ-13 and in ZSM-5 zeolites can be found elsewhere, along with the impact of the magnetic interactions on N₂O decomposition.^164,165 These points are discussed in Chapter 4.3.3.

3. Thermodynamic factors influencing the speciation of metal centres in zeolites

The chemistry of TMI in zeolites is complex because of speciation into isolated bare cations (of different valence state Mⁿ⁺), isolated oxygen-containing species (metal–oxo and metal–oxyl complexes), dinuclear oxo-cations, larger oligomeric metal–oxo species, and segregated M_xO_y oxide particles that often coexist within the same zeolite catalyst. They may undergo distinct transformations evolving significantly under different conditions, ranging from catalyst pretreatment and activation to actual reaction conditions.^71,166

The most important chemical processes during activation of metallozeolites include hydration, hydroxylation, olation, oxolation, auto-reduction (in vacuo) or oxidation in the presence of O₂.^33,167,168 The interrelation of these processes for zeolite TMI centers (Mⁿ⁺) associated with single (z₁) and double (z₂) aluminum sites (isolated and dual exchange centers, respectively), which bear a formal negative charge of −1 and −2, is schematically illustrated in Fig. 13, and the corresponding reactions are summarized in Table 3. The possibility of further transformations of dual metallic centers into polynuclear metal–oxo species is discussed below (Fig. 16). First-principles thermodynamic modelling is often used to provide a clear background for relating the feasibility of these processes to their energetics and various T,p conditions.

Fig. 13 Typical transformations of metal-containing zeolite centers controlled by thermodynamic conditions (temperature and O₂/H₂O pressures) associated with single (z₁) and double (z₂) exchange sites (tetrahedral AlO₄⁻ units). Atom color coding: O – red; Si – yellow; Al – pink; H – light grey; TMI (Mⁿ⁺) – pale blue.

Table 3 The most important chemical processes taking place during the synthesis and activation of metallozeolites. The formal negative charge of the zeolite framework associated with tetrahedral (AlO₄)⁻ units is denoted z

Autoreduction of oxo-species
2[M^n+–O^2−](z) ⇌ 2M^(n−2)+(z) + O2
Direct oxidation to metal–oxo species
M^n+(z) + 1/2O2(g) ⇌ [M–O]^n+(z)
Intrazeolite hydrolysis
Associative hydration	M^n+(z) + mH2O(g) ⇌ (z)M^n+–(H2O)m
Dissociative hydration (hydroxylation)	M^n+(z) + H2O(g) ⇌ M(OH)^(n−1)+(z) + H^+
Agglomeration into oligomeric metal–oxo entities
Olation	M(OH)^(n−1)+(z) + M^n+(z) → (M–OH–M)^(2n−1)+(z2)
Oxolation	(M–OH–M)^(2n−1)+(z2) → (M–O–M)^(2n−2)+(z) + H^+(z)
Oxidative oxolation	3[(z)M^2+–OH^−)] + 1/2O2 → (z2)[M3(μ-O)3]^2+ + H2O + H^+(z)
Merging of metal–oxo species	2[MO]^n+(z) → [M(μ-O)2M]^2n+(z2)

---

3.1. First-principles thermodynamics of metallozeolites

One of the most useful computational techniques applied for the modelling of metallozeolite catalysts beyond the basic “0 K/UHV model”¹⁶⁹ is the first principles thermodynamics (FPT). It integrates density functional theory (DFT) with statistical physics to describe the stability, possible transformations, and reactivity of the investigated system at the atomic level, as a function of the temperature and pressure of reactants.^170,171 After appropriate adaptation, FPT is now routinely applied in modelling of the zeolite catalytic materials and their adsorption properties.^15,165,172 Typically, (i) a parent zeolite network distinct by the amount and distribution of Al atoms, where formal negative charge generated by z AlOH₄ units is counterballanced by z protons (zeolite (zH)), (ii) a bulk metal (M_(bulk) – serving as the source of the metallic centers), and (iii) relevant gaseous reagents that may interact with the zeolite centers (e.g., H₂O_(gas) and O_2(gas) in case of catalyst pretreatment) are used as convenient reservoirs of the involved species. Such an assembly is shown in Fig. 16a₁, where gas phase, bulk metal, and zeolite parts (b₁, b₂, and b₃, respectively) are used to build a specific model of metal-exchanged zeolite interacting with exemplary O_2(g) and H₂O_(g) molecules. In this case, general equilibrium reaction (eqn (3)) of the zeolitic system (M_xH_nO_m) formation may be formulated as:¹⁵⁴with the reaction energy defined by eqn (4).where E_{zeolite(MxHnOm)} is the total energy of the metal containing zeolite with active sites covered by water/oxygen, E_zeolite(zH) is the energy of the initial H-form of the zeolite with z Al sites and z protons, whereas E_M, E_H2O and E_O2 correspond to the computed total energies of bulk metal M, gaseous water, and gaseous dioxygen, respectively. The ΔE values are obtained from periodic DFT calculations, which assure fairly good accuracy in combination with good computational efficiency. Geometries and energetics computed using hybrid functionals (e.g., B3LYP, B3PW91,¹⁷³ and PBE0¹⁷⁴), complemented by inclusion of the semiempirical treatment of the dispersion interactions,^175,176 generally align well with experimental results, and high-level post-HF calculations, as well.^177,178 However, benchmark calculations comparing different levels of the DFT methods are highly valuable, allowing for sensible selection of the required computational scheme.^{172,179–181}

The change in the free energy ΔG due to intrazeolite chemical processes is given by:

where the chemical potential differences of gaseous reactants (herein Δμ_O2 and Δμ_H2O) are obtained applying standard statistical thermodynamics.¹⁸² Using gas-phase molecules as a reference state for interaction with the active sites, the changes in chemical potential correspond to the external equilibria with reactants located outside the zeolite channels (Fig. 14a₁ with the reference states taken from (Fig. 14b₁–b₃)). For modelling the internal equilibria, the intrazeolite molecules (molecule_(IC)) are used as a reference state (Fig. 14a₂ with the reference states taken from Fig. 14b₂–b₄). The entropy correction for the in-cage molecule can be calculated using molecular dynamics methods^183–185 or estimated by scaling, for instance, to the NH₃-TPD peaks for catalysts with well-defined active sites.¹⁷² It is noteworthy to recognize that the Al configuration may not be constant under experimental conditions.¹⁸⁶ As a practical solution, several stoichiometric or configurational models of the zeolite catalyst can be evaluated comparatively.

Fig. 14 A molecular model of zeolite in equilibrium with free-standing molecules (a₁), and molecules trapped inside zeolite cages (a₂) together with molecular and bulk-phase reservoirs used in FTP modelling to assemble the zeolite catalyst in equilibrium with the reactants (b₁)–(b₄). The 3D (c₁) and 2D (c₂) representations of the thermodynamic diagrams representing the stability of iron species in the Fe/SSZ-13 zeolite as a function of (T,p_O2) and (T), respectively. Bottom envelope of 3D FPT diagram representing stability regions (T,p_O2) of dual copper sites in the Cu/CHA zeolite (d), based on ref. 181.

Thermodynamic phase diagrams, constructed using first-principles data, provide insight into the various aspects of metal–zeolite chemistry within the widespread operational limits of pressure, temperature, and composition. They are typically presented in the form of 3D (see Fig. 14c₁) and 2D (Fig. 14c₂) plots, where the Gibbs energy is plotted as a function of temperature and the pressures (or chemical potentials) of the reactants. Due to the constraints imposed by the 2D/3D representation, thermodynamic properties are often presented assuming variable temperature and the partial pressure of one gas reactant, with the other variables fixed or constrained by the reaction stoichiometry (e.g., NO/NH₃ = 1 in SCR reaction).

As an example, the stabilities of different iron species present in the Fe/SSZ-13 zeolite are compared in Fig. 14c₂ as a function of temperature, with the oxygen and water partial pressures fixed at p_O2 = 0.01 atm, and p_H2O = 0.005 atm. At low temperatures, the most stable form is the hydrated Fe²⁺(H₂O)(z₂) center, which releases a water molecule at around 250 °C. The hydroxylated forms, Fe²⁺(OH)(z) + H⁺(z), are slightly less stable than the hydrated Fe cations across the entire temperature range (thus, both forms may coexist). In contrast, the terminal iron-oxo species, [FeO]²⁺(z₂), of the highest free energy, are not expected to appear in the whole temperature range.

To present the results more explicitly, often only the lower envelope of the FPT diagrams (defining the most stable species) is shown (Fig. 14d). Such a 2D diagram, for instance, reveals the thermodynamic conditions of copper auto-reduction in the Cu/SSZ-13 zeolite.¹⁸¹ For low water vapor pressure (p_H2O = 10⁻⁶), three stable dual-, and one mono-copper sites are expected depending on the (T,p) conditions. At lower temperatures (T < ∼500 K), regardless of the oxygen pressure, the Cu–(OH)₂–Cu(z₂) dimer dominates, whereas at higher temperatures it transforms into an oxygen-bridged entities, Cu–O–Cu(z) + H(z), with desorption of a water molecule, or into a Cu–(OH)–Cu(z₂) dimer upon losing the OH group. These transformations are controlled by the oxygen pressure with a threshold value of ∼0.01 atm. Under the most reducing conditions (highest T and lowest p_O2, orange area), copper bridges are expected to decompose, leading to the formation of two adjacent highly reactive monomeric copper(I) centres (Cu(z)) (Chapters 4.3.1 and 4.3.3).

FPT modeling is also a convenient tool that enables comparison of the stability and evolution of intrazeolite active sites such as (z)FeOH, (z₂)FeOH, (z)FeO, (z₂)[FeO]²⁺, (z)Fe-OH₂ in the six- or eight-membered rings (6MR or 8MR) of Fe/SSZ-13.¹⁸⁷ The corresponding phase diagram (Fig. 15a₁) implies that H₂O prefers to adsorb on a Fe⁺ cation in a wide range of temperatures. However, the hydrated sites are much less stable than the hydroxylated (z)FeOH and (z)FeO species, which dominate at low and high temperatures, respectively. The most energetically favourable sites for iron accommodation are 6MR rings (Fe in 8MR are less stable by 0.5 eV). Still, upon ligation of water, hydroxyl, and oxygen species, the energy gap between Fe located in the 8MR and 6MR sites becomes negligible. The K-edge of XANES spectra (see the simulated spectra in Fig. 15a₂) is diagnostic for the presence of those ligands in the coordination sphere of the Fe cations.¹⁸⁷

Fig. 15 2D thermodynamic diagram representing stability of iron active centers formed within the six- or eight-membered rings (6MR or 8MR) of Fe/SSZ-13 (a₁), together with the associated XANES spectra simulated for the most important species (a₂).¹⁸⁷ The expanded diagram (b) comparing the stability of Fe species on isolated (z₁) and dual (z₂) exchange centers positioned in 8MR and 6MR rings of CHA zeolite with different Al distribution (Al_1–3 vs. Al_1–4).¹⁶⁵ Calculated (HSE06-TSvdw) copper speciation phase diagrams for 1Al (c₁) and 2Al (c₂) exchange sites of Cu/SSZ-13.¹⁸⁸ Phase diagrams for iron speciation in Fe/SSZ-13 zeolite as a function of relative O₂ pressure and temperature with p(H₂O) set at 0.02 atm, constructed for sites stabilized within the z₂-8MR (d₁) and z₂-6MR (d₂) rings.¹⁶⁵ The figure was based on data from the quoted references.

A more advanced analysis of the stability of different iron forms in the isolated (z) and dual (z₂) exchange centers in the chabazite-type zeolite is ilustrated in Fig. 15b, where ring multiplicity (8MR vs. 6MR), internal aluminum distribution (Al_1–3 vs. Al_1–4), formation of bridged species and the possibility of iron existing in higher oxidation states are taken into account as well.¹⁶⁵ This diagram can be used to track the evolution of Fe sites under various pretreatment conditions, showing that autoreduction temperatures and water desorption conditions are highly dependent on the proximity of aluminum atoms and the nuclearity of iron.

In the case of Cu/SSZ-13 zeolite, the akin phase diagrams representing evolution of CuOH(z) and Cu(Z₂) as a function of temperature and O₂ pressure at fixed H₂O pressure,¹⁸⁸ are shown in Fig. 15c₁ and c₂, respectively. Both Cu sites are hydrated under ambient conditions (where solvated Cu²⁺–H₂O and [CuOH]⁺ dominate), and become dehydrated at elevated temperatures in an O₂ environment. However, only CuOH(z) species undergo autoreduction to monovalent copper (Cu(z)) in the inert environment (Fig. 15c₁). The preference of populating the Cu(z₂) sites over CuOH(z) sites is independent of the extent of hydration.¹⁵

A comparison of the Fe ions' stability in the 8MR (Fig. 15d₁) and 6MR (Fig. 15d₂) rings of the Fe/SSZ-13 zeolite as a function of T, p_O2, and p_H2O¹⁶⁵ revealed a dramatic impact of the ring multiplicity on the distribution of dimeric iron centers. At room temperature, the most stable species are the iron-dimers ligated by water (without their detachment from the framework). They are initially in the most stable trivalent state, and as the temperature rises, all sites transform into partially dehydrated entities at T = 100 °C and 450 °C for 8MR and 6MR, respectively. An additional difference of even greater significance for the catalytic activity arises from the fact that the dimers in the 8MR ring transform into (z)Fe–(μ-O)₂–Fe(z) bridging species, while the 6MR dimers are stabilized as (z)Fe–(OH)₂–Fe(z). The latter sites contain divalent iron and may be treated as “autoreduced” (similarly to the monocopper sites in Cu-/SSZ-13^61,188). The autor-eduction temperatures vary widely from site to site, and the high-temperature activated zeolites are expected to contain a mixture of Fe²⁺ and Fe³⁺. The validity of such prediction has been proven for Fe/ZSM-5 and Fe/Y zeolites heated in helium (or in vacuum) using the X-ray absorption and Mössbauer spectroscopies.^189,190

Fig. 16 Possible transformations of a cubic Fe₄O₆ cluster (a₁), together with the FPT diagram representing the stability of the most important iron forms in the 3D (a₂) and 2D (a₃) representation, based on data from ref. 154. The relative stabilities of multi-nuclear copper species hosted in the ZSM-5 zeolite, as a function of Δμ_O (b₁) and (T,p_O2) (b₂), based on data from ref. 191. The stability correlations of the Fe²⁺, [FeO]²⁺, and [FeO]²⁺ species accommodated in the ferrierite α, β, and γ sites (c), adapted from ref. 195 with permission from Wiley, Copyright 2025. The CuOH/Cu fraction plotted as a function of the Si : Al and Cu : Al, revealing regions representing a composition space where Cu²⁺ species dominate (below white line), and a space where gradual increase in the [ZCuOH]⁺ fraction is expected (above the white line), white circles denotes experimental results (d), adapted from ref. 188 with permission from American Chemical Society, Copyright 2016. Distribution of Cu²⁺, Cu⁺, [Cu₂OH]²⁺ species (coded with blue, red, and green, respectively), presented as pie charts for Cu/SSZ-13 zeolite hypothetical samples (e) based on data from ref. 61. Samples (e₁)–(e₅) differ in the distribution of the monomeric and dimeric exchangeable 1Al and 2Al centres, in Al pair arrangements, and in defect concentrations (d) and are studied under different conditions (top and bottom row).

Going beyond the charged mono- and dimeric centres requires a sensible postulation of proper chemical processes interconnecting the dispersed and oligomeric metal species in the presence of H₂O, O₂, and the zeolitic protons as well. An exemplary diagram representing possible transformations of a bulky Fe₄O₆ cluster hosted in the Fe/ZSM-5 zeolite is shown in Fig. 16a₁.¹⁵⁴ Its decomposition into binuclear species (via protonation at Brønsted acid sites) proceeds either through hydrolysis, leading to isolated oxygenated or hydroxylated mononuclear species, or through dehydration, forming oxo-bridged [Fe(μ-O)Fe]²⁺ species. Thermodynamic diagrams confirm that the intrazeolite [Fe(μ-O)Fe]²⁺ and [Fe(μ-O)₂Fe]²⁺ binuclear complexes represent the most stable species under typical reaction conditions (Fig. 16a2–a3). They preferentially occupy the γ site (8MR), but other iron varieties of comparable stability can also lodge at the α, δ, and β sites (6MR). Oxygen pressure controls the thermodynamically preferred mixing between the [Fe(μ-O)Fe]²⁺ and [Fe(μ-O)₂Fe]²⁺ binuclear clusters. The formation of Fe²⁺ ions and binuclear complexes occurs only at high temperatures, aligning well with experimental observations.¹⁵⁴

Comprehension of thermodynamic conditions is crucial when analysing multinuclear copper species in ZSM-5 zeolites, as the stability of the extra-framework complexes involved is strongly influenced by temperature and the composition of the gas phase (Fig. 16b1 and b2).¹⁹¹ The binuclear Cu sites correspond to the lowest-energy structures, yet under the O₂-rich conditions, typically employed for the activation of Cu/ZSM-5 catalysts, when T- and p-related effects are taken into account, the [Cu₃(μ-O)₃]²⁺ trinuclear complex turns out to be more stable than the mono- and binuclear species. The intrazeolite mononuclear oxo- and hydroxo-complexes are generally less stable than the oligonuclear species.^192–194

The thermodynamic analysis can also be used to draw more general conclusions than the relative stability of TMI within the given zeolite network structure. For example, a noteworthy analysis of the FTP data on iron species stabilized in the 6MR, 8MR sites of chabazite, and α, β and γ sites of ferrierite zeolites (Fig. 16c), allowed for the demonstration that there is a distinct linear correlation (R² ∼ 0.9) between the stabilities of the Fe²⁺, [FeO], and [FeOH]⁺ species.¹⁹⁵ This implies that the stability trends of the related Fe²⁺, [FeO], and [FeOH]⁺ species with the varying framework type and the Al distribution remain consistent, regardless of thermodynamic conditions, and are, therefore, of a more general character.

The impact of Al arrangements on the monomer and dimer configuration of Cu cations and their oxidation state, concerning the auto-reduction of Cu in SSZ-13,¹⁸⁸ is illustrated in Fig. 16d. Within the assumption that the exchanged copper ions initially populate all available dual sites as Cu²⁺/z₂, before occupying 1Al sites in the form of Cu²⁺OH/z, the CuOH/Cu fraction can be assessed as a function of the Si : Al and Cu : Al ratios.^196,197 Two distinct regions, where Cu²⁺ species associated with 2Al are expected to dominate (below the white line), and where a gradual increase in the Cu²⁺OH/z fraction is predicted (above the white line), can be distinguished. This diagram helps to rationalize the speciation of copper in real catalysts (white circles) depending on the Al/Si and Cu/Al ratios. Furthermore, the pie charts can be used for concise illustration of the relative contributions of key copper species in CHA⁶¹ (Cu⁺ – blue, Cu²⁺ – red, [Cu₂OH]²⁺ – green), depending on the zeolite composition (Si/Al, Al pair arrangements, defect concentrations) and thermodynamic (oxidizing or reducing) conditions (Fig. 16e).

4. Interfacial coordination chemistry – principal pathways of ligation

4.1. Molecular picture of ligation

Most simply, the binding of reactants to encaged metal centres can be rationalized in terms of the molecular orbital diagrams, which can be categorized into normal (NEL), mixed (MEL), and inverted (IEL) energy level schemes (Fig. 17a₁, a₂, and a₃, respectively).¹⁹⁸ In the case of open-shell cations such as Fe²⁺ (S = 2), Co²⁺ (S = 3/2) Ni²⁺ (S = 1) or Cu²⁺ (S = 1/2), typically involved in important catalytic reactions with zeolite materials,^13,18,199 for sensible interpretation the corresponding energy level diagrams should be constructed for the α and β channels separately, including the exchange splitting (see Fig. 8). In the normal scheme (Fig. 17a₁), the 3d levels of the metal centre lie above the occupied levels of the ligands. Since the filled 3s²p⁶ sub-shell shields the inner 3d orbitals, the resultant Pauli repulsion leads to ‘stretched’ M–L bonds, which secures the integrity of the 3dⁿ configuration upon ligation.¹⁹⁸ This classic Werner-type scheme is characteristic of the intrazeolite 3d cations, typically in the 2+ oxidation states interacting with the innocent σ-donor ligands such as NH₃ or H₂O (see Chapter 4.2). Intrazeolite adducts with more extensive metal–ligand covalency may lead to the mixed (MEL) pattern (Fig. 17a₂). In the most extreme situations, when 3d levels are situated below those of the ligand, which may be caused by enhanced oxidation state and/or exchange interactions (in the case of open-shell systems), an inverted scheme (IEL) shown in (Fig. 17a₃) is appropriate. Two mechanisms can be distinguished by which the Pauli shield (Fig. 17b) can be breached, and the 3dⁿ integrity of the metal core perturbed. When the cation oxidation state becomes too positive, the 3d orbitals are shifted down in energy so far that the Pauli shield can be broken (from ‘outside’) by the redox ligands. Analogically, when the 3d orbitals become sufficiently electron-rich, they may overcome the Pauli shield from ‘inside’ and transfer electron density onto ligands (such as O₂), altering the 3dⁿ configuration of the metal core.¹⁹⁸ These binding modes are discussed in detail in Chapter 4.3.

Fig. 17 Schematic representation of the metal–ligand energy levels (a₁)–(a₃), epitomic representation of the Pauli shield (b),¹⁹⁸ and principal types of the molecular orbital overlap sorted by symmetry (c). Five categories of the spin density distributions defining a magnetophore of the open-shell intrazeolite adducts (d), and an exemplary illustration of a zeolite nano-cage reactor with the associated signifiers (e).

Attachment of molecules to the caged transition-metal cations via σ, π, and δ type overlap (Fig. 17c) may occur in an isodesmic (with conservation of the number of bonds) or an alterdesmic (with a change in the total number of bonds) way. In the former case, new covalent bonds with the coordinated reactants are formed at the expense of the strained bonds between the metal center and the zeolite O_z ligands. The simplest behaviour of intrazeolite transition metal complexes of NEL character in the catalytic redox process consists of oxidation or reduction of the metal centre only, leaving the redox state of the ligands unaffected. Yet, the ligands control indirectly the redox properties of the metal core by modifying the energy levels of the 3d orbitals (see Chapter 4.2.4). In some cases, however, ligands can also participate in the redox process. Then, such ligands are referred to as non-innocent or redox ligands (see Chapters 4.3.1 and 4.3.2).

As a result, the four leading roles the ligated molecules can play in the catalytic reactions are as follows. (i) Tuning of the metal catalytic properties through reduction/oxidation, which strongly influences the affinity to capture reactants, and the mechanistic course of ensuing follow-up reactions. (ii) Acting as an “electron-reservoir” in the case of redox active ligand, which allows the metal to store and release electrons on the ligand moiety (anionic redox) in the elementary steps unavailable on the metal core only. (iii) The formation of reactive ligand-radicals that actively participate in the making and breaking of chemical bonds during catalysis. (iv) The modification of the substrate reactivity in cases where the substrate itself acts as a redox non-innocent ligand.²⁰⁰

The redox reactivity of the open-shell species, in particular, is governed by the spin density distribution between the metal core and the ligated molecule that both constitute a common magnetophore. The latter notion, analogous to the chromophore in optical spectroscopy, refers to the molecular fragment of a paramagnetic adduct where the bulk of the spin density is confined, thereby determining the magnetic properties of the entire system essentially. Three main magnetophore patterns are categorized as ligand-centred, mixed, and metal-centred (Fig. 17d). The locus of the highest spin density defines the centre of the preferential attack of paramagnetic reactants such as NO, O₂, or NO₂ (see Chapter 4.3.2). The zeolite aperture controls, in turn, the accessibility of these centres to the reactant molecules. In contrast, the speciation and reactivity of the active sites are significantly influenced by the framework topology and the associated confinement effects, as well as the Si/Al and M/Al ratios, and the spatial distribution of the Al atoms. Those aspects are epitomized in the form of a zeolite nanoreactor in Fig. 17e.

4.1.1. Orbital interaction analysis. Metal centres embedded within the zeolite framework exhibit complex interactions with the zeolite framework, and the conventional bonding descriptors (total energy, bond length, bond order) cannot sensibly characterize the reactant molecules. Thus, natural orbital for chemical valence (NOCV) analysis²⁰¹ becomes an insightful tool for probing orbital interactions within metal–zeolite systems, offering an advanced understanding of bonding, charge transfer, and electron distribution in these materials. NOCV analysis enables the separation and quantification of electron and spin density transfer channels, such as π-backdonation or σ-donation, between virtually delineated interacting fragments. The pairs of the natural orbitals for chemical valence (Ψ_−k, Ψ_k) allow for decomposition of the differential charge density, Δρ, into the particular NOCV contributions (Δρ_k):²⁰¹

The ν_k stands for the NOCV eigenvalues, and M for the number of basis functions. The eigenvalue expresses a fractional electron population transfer in each electron flow channel. For open-shell systems, the decomposition is carried out separately for the α and β electron densities (spin-resolved NOCV). Afterwards, the contributions of similar orbital character appearing in both spin densities can be treated jointly. The NOCV analysis, combined with the Ziegler–Rauk extended transition state energy decomposition scheme (ETS-NOCV),²⁰² allows for the partitioning of the total orbital interaction energy into particular NOCV channels (k,−k):

The integrals F^TS_ij = 〈i|∇²/2 + V_ne + V_C[ρ^TS] + V_xc[ρ^TS]|j〉 are matrix elements of the Kohn–Sham operator evaluated for the electron density at the “TS” state, which is given by ρ^TS = 1/2ρ + 1/2ρ⁰ (ρ and ρ⁰ represents the electron density of interacting and non-interacting fragments, respectively).²⁰² The NOCV method has enabled a deeper understanding of the origin of the IR red shift of the stretching C–C frequencies due to metal–ligand bonding on the Cu and Ni centres of ZSM-5 zeolite,²⁰³ and it has also provided a reliable description of various non-covalent interactions.^204,205 The ETS-NOCV analysis has also been used to unravel the binding mechanism of dinitrogen, carbonyl (Chapters 4.2.3 and 4.2.4), superoxide (Chapter 4.3.1), and nitrosyl adducts (Chapter 4.3.2) of nickel, copper, and cobalt ions in the exchangeable positions of various zeolites.

4.2. Ligation of innocent molecules

Innocent molecules such as H₂O, NH₃ are the non-redox ligands that interact with metal centres in metallozeolites primarily through sigma-donation (Fig. 17a1). The strength and nature of sigma-donation depend on several factors, among them, the most important are electronegativity and hardness/softness of the metal centres (according to the HSAB principle, where Cu²⁺, Ni²⁺, Fe³⁺, etc., exhibit different affinities towards innocent ligands), availability of coordination sites on the metal ion within the zeolite framework, and the intracage solvation effects that may modulate metal–ligand interactions.^206–208 Moreover, the ability of innocent molecules to form multiple coordination σ-bonds (poly-molecular adsorption) with metal ions is controlled by the oxidation state and the electronic structure of metal ions, and the cage size of the zeolite framework.^209,210 It is well established that the interaction of H₂O and NH₃ with metallozeolites can also lead to partial or complete detachment of the metal ion from the zeolite framework.²¹¹ This occurs through ligand-induced solvation, where strong coordination of ligands stabilizes the metal ion inside the cage. This effect plays a vital role in modulating the catalytic activity within the zeolite pores. In light of these facts, although such reagents are considered innocent, their impact on the catalytic and adsorption properties of metallozeolites is significant. Other innocent ligands, often used to probe the valence and coordination states of the intrazeolite TMI by IR, are CO and N₂ molecules. Although their adsorption energy has a pronounced dispersion component, a minute electron exchange between the ligated N₂ or CO and the metal is responsible for the observed site sensitivity (see Chapters 4.2.3 and 4.2.4).

4.2.1. Intrazeolite hydration of TMI. Water and hydroxyl groups serve as the primary ligands in metal-exchanged zeolites immediately after synthesis and catalyst activation (Chapter 3.1). Another common source of water is its autogenic production during catalytic reactions such as oxidation of hydrocarbons, NH₃ or NH₃-SCR, where water molecules are either final or intermediate products.^33,212

While water is often considered a co-reagent in catalytic processes and implicated in specific reaction steps,^152,213,214 more general studies aiming to identify general trends in its interaction with metal centres are relatively scarce. In this context, B3LYP-DFT cluster calculations of water adsorption on various divalent metal centres (Zn²⁺, Cu²⁺, Ni²⁺) stabilized at exchange positions I and II in zeolite Y (Fig. 18a) can be invoked as a suitable example.²¹⁵ The oxidation state and type of the cation have a greater impact on the adsorption than the specific exchange position. An ALMO (absolutely localized molecular orbitals) analysis,^216,217 allowed for the decomposition of the adsorption energy into none-CT (non-charge transfer), ligand to metal charge transfer (CT(X → MY)), metal to ligand charge transfer (CT(MY → X)), and CT(HO) terms. This reveals that the electronic H₂O-TMI interactions are primarily governed by the non-CT and CT(X → MY) contributions, in contrast to those previously studied for alkali metal sites.

Fig. 18 Adsorption of water on divalent metal centres (Zn²⁺, Cu²⁺, Ni²⁺) located at the exchange positions I and II in zeolite Y, together with ALMO EDA analysis for Zn²⁺ in II site of Y zeolite (a), based on data from ref. 215. Cluster models (b₁) and (b₂) and water adsorption energies on Cu²⁺, Zn²⁺, and Fe²⁺ incorporated in FAU (b₃), together with the iso-surface electron density contour for Fe²⁺–H₂O (b₃, insert), adapted from ref. 218 with permission from Royal Society of Chemistry, copyright 2022. Adsorption energy of water on various TMI centres incorporated in MOR framework in the high (HC) and low coverage (LC) limits (c), based on data from ref. 219. Optimised geometries of the copper poly-aqua complexes (d₁)–(d₅), together with the corresponding 2D thermodynamic stability diagrams constructed for the 8MR (e₁) and 6MR (e₂) exchangeable sites of CHA. The panels (a)–(d) were based on data from the respective references. Panels (e₁) and (e₂) were based on data from ref. 83.

Comparison of water adsorption on divalent cations (Cu²⁺, Zn²⁺, Fe²⁺) incorporated in FAU zeolite (see Fig. 18b₁–b₃) shows that Zn²⁺ and Cu²⁺ exhibit the strongest interactions with water, with the adsorption energies of −105.3 kJ mol⁻¹ and −81.6 kJ mol⁻¹, respectively.²¹⁸ These values, comparable to those of alkaline earth metal cations, may be attributed to their significant HSAB hardness. This is further supported by the iso-surface electron density contours (Fig. 18b₃, insert), which agree well with theoretical studies on mordenite, where the influence of high (HC) and low (LC) coverage is illustrated (Fig. 18c).²¹⁹

A more thorough approach to studying water ligation involves a set of copper multi-aqua complexes in different configurations (Fig. 18d₁–d₅)⁸³ to construct 2D thermodynamic stability diagrams for the 8MR (Fig. 18e₁) and 6MR (Fig. 18e₂) exchangeable sites of CHA zeolite. This analysis reveals the level of ligation at which a copper cation is pulled out from the zeolite rings, gradually acquiring considerable mobility within the cages. Regardless of the site, this occurs when four H₂O ligands are bound to a Cu cation. When the charge of the cation is reduced, a smaller number of solvent molecules is needed for the detachment, as illustrated by Cu⁺(H₂O)₂ species entrapped in mordenite.¹⁶¹ Then, the polyaqua-complexes exhibit sufficient mobility to migrate between the zeolite channels. For selected H₂O pressure of 10⁻⁴ (typical for the NH₃-SCR reaction, marked with a dashed line), this condition is met below 125 °C (8MR) and below 75 °C (6MR).

4.2.2. Coordination of NH₃ and interpretation of NH₃-TPD profiles. Essentially, NH₃ exhibits stronger binding to TMI than H₂O due to its higher electron-donating ability, resulting in the formation of more pronounced σ-bonds.²¹¹ In the metal–zeolite catalysis, the adsorption and desorption kinetics of NH₃, and the associated modification of the coordination state and redox properties of the intrazeolite TMI play an essential role, particularly in NH₃-SCR^33,77,93 and AMO³³ reactions. These processes are chiefly studied for Cu in the small-pore zeolites, and both mononuclear and binuclear copper species serve as the primary adsorption sites.^220,221 Much less attention has been paid to other transition metal ions.^40,222 Brønsted acid sites also contribute to NH₃ capture, leading to the formation of NH₄⁺·nNH₃ associations, which can complicate the interpretation of experimental NH₃ uptakes,²²³ whereas extra-framework Al species play a minor role.²²⁴ In addition to spectroscopic techniques (mainly IR), temperature-programmed desorption (TPD) of NH₃ is widely used for coarse characterisation of both metal Lewis (LAS) and Brønsted (BAS) acid sites in zeolite catalysts.^196,225,226 This technique provides valuable insights into the thermodynamic stability of various intrazeolite metal–ammonia complexes across different temperatures. It can serve as an essential experimental benchmark for validating the FTP modelling of NH₃ interactions with TMI present in zeolites. The combination of TPD and FTP methods can be applied for in-depth investigations into NH₃ adsorption, e.g., on various monomeric and dimeric copper centres, and BAS sites present simultaneously in Cu/SSZ-zeolite catalysts, differing in Si/Al ratios and aluminium distribution within the zeolite framework.^172,227 For each adsorption site, the energies of multiple NH₃ adsorption (see Fig. 19a for the exemplary 6MR site) can be used for the construction of the 3-dimensional FPT diagram (Fig. 19b₁), and its recasting into variation of Θ_NH3 (site coverage) as a function of temperature and partial pressure of NH₃ (Fig. 19b2). The cross-sections at p_NH3 set to the experimental value (10⁻⁴ atm) allow for a straightforward determination of the temperature windows in which NH₃ desorption peaks are expected due to ammonia release from the particular centres (Fig. 19b₃ and b₄). The overall theoretical desorption profile results from adding the contributions of all involved adsorption sites (Fig. 19c₁). They can be compared with the experimental TPD-profile obtained for Cu/SSZ-13 catalysts differing in the Si/Al ratio (Fig. 19c₂). The comparison implies that the desorption peaks resolved in TPD cannot be univocally assigned to NH₃ release from specific Cu centres or Brønsted acid sites. The experimentally observed low-, medium-, and high-temperature desorption bands arise from multiple copper species whose speciation and redox states evolve dynamically throughout the NH₃-TPD process. This behaviour is driven by copper migration, its reduction, and condensation within the zeolite framework, as revealed by complementary in situ EPR and IR investigations.¹⁷²

Fig. 19 Optimized geometries of the Cu²⁺–(NH₃)_n adducts located in the 6MR site of CHA (a). The 3-dimensional diagrams of ΔG vs. T, p_NH3 (b₁) and Θ_NH3 vs. T, p_NH3 (b₂), together with the corresponding cross-sections for p_NH3 set to 10⁻⁴ atm (b₃) and (b₄). The overall calculated TPD profile with all component bands (c₁). The experimental NH₃-TPD profile obtained for samples differing in Si/Al ratio (c₂). Adapted from ref. 227 with permission from John Wiley and Sons, Copyright 2021.

At lower temperatures, when higher coordination numbers of TMI to ammonia are expected (n > 2), detachment and mobilisation of the metal cations from the zeolite framework is typically observed (analogous to the water ligation case). This phenomenon has been extensively investigated due to its crucial role in the low-temperature NH₃-SCR process.^179,229 A combination of computational (DFT-HSE06) and experimental (XANES, XAS) studies has been applied to thoroughly examine different Cu–(NH₃)_n complexes under in situ, ex situ, and operando SCR conditions, considering variations in the catalyst composition.²²⁸ The copper speciation, mobility, and its actual siting are primarily dictated by environmental factors, rather than the initial location of the Cu cations or the type of zeolite framework. As an example, the mobility of [Cu(NH₃)₂]⁺ within the CHA framework, as revealed by ab initio meta-dynamic simulations, is shown in Fig. 20a1. The Cu ion migration through the 8-MR rings between two cages requires an activation energy ∼55 kJ mol⁻¹, confirming that the CHA framework provides sufficiently large windows for efficient intrazeolite diffusion of the [Cu(NH₃)₂]⁺ species, triggering the oxidation half cycle (OHC) of the SCR reaction.^15,179,229

Fig. 20 The migration pathway of [Cu(NH₃)₂]⁺ species presented inside the channels of CHA zeolite (a₁), together with the corresponding free energy profile plotted against Cu–Al distance (a₂), adapted from ref. 228 with permission from the American Association for the Advancement of Science, Copyright 2017. Representative configurations of [Cu(NH₃)₂]⁺ in the initial state (ξ < 0), transition state (TS, ξ = 0), and final state (ξ > 0) during the diffusion through the 8-membered ring windows of Cu-CHA without co-adsorbed molecules (b₁) and in the presence of O₂ and NO (b₂), together with the corresponding free energy profiles as a function of the reaction progress ξ (b₃) at 423 K, adapted from ref. 229 with permissions from American Chemical Society and The Authors, Copyright 2021. The initial (c₁), transition (c₂), and final state (c₃) structures of the [Cu(NH₃)₂]⁺ complex migrating from 6MR rings with a single Al to 6MR with a pair of Al, adapted from ref. 231 with permission from American Chemical Society, Copyright 2024. Postulated route of formation of mobile μ-η²,η²-peroxo diamino dicopper(II) complexes upon oxidation of the [Cu(NH₃)₂]⁺ species (c), adapted from ref. 232 under the terms of the CC BY 4.0 license, Copyright 2022, published by American Chemical Society.

The role of Cu ion mobility in Cu/SSZ-13 has also been explored using large-scale simulations, employing an interatomic potential obtained from machine learning that accurately replicates the ab initio results.^229,230 A combination of biased and unbiased simulations enabled an in-depth analysis of the [Cu(NH₃)₂]⁺ mobility in Cu-CHA catalysts, revealing that the Al pairing within the 8-MR rings facilitates local hopping, whereas higher NH₃ concentrations enhance long-range diffusion. The [Cu(NH₃)₂]⁺ migration under NH₃–SCR–NO_x conditions obtained from AIMD simulations is illustrated in Fig. 20b₁–b₃. The diffusion is hindered when additional NO/O₂ molecules are present in the zeolite cavity (Fig. 20b₂). Furthermore, a decrease in the number of BAS sites leads to an increase in the diffusion barrier of [Cu(NH₃)₂]⁺. The van der Waals interactions between [Cu(NH₃)₂]⁺ and the zeolite framework (Fig. 20c₁–c₃) are significantly influenced by the BAS distribution, inhibiting the low-temperature NH₃-SCR activity.²³¹ For zeolites of uneven Al distribution, variations in the interaction energy between [Cu(NH₃)₂]⁺ and the zeolite framework are mainly responsible for the inter-cage diffusion. The migration of [Cu(NH₃)₂]⁺ from an Al-poor cage to an Al-rich cage exhibits a lower diffusion barrier, favouring the formation of active dimer-Cu species and enhancing the SCR reaction.

The oxidation of mobile diamino copper complexes (Fig. 20d) monitored by XANES and EXAFS techniques shows consistently that the Si/Al ratio of the zeolite hosts influences the structure of the mobile μ-η²,η²-peroxo diamino dicopper(II) adducts, which are formed during the oxidation of [Cu(NH₃)₂]⁺ complexes by O₂ (OHC).²³²

4.2.3. Adsorption of N₂ and the probe molecule characteristics. A non-polar N₂ molecule with no lone pairs interacts primarily through non-specific van der Waals forces when adsorbed onto metal centres within zeolite frameworks. Despite its chemical inertness, weak residual covalent interactions can also be formed, which are notably sensitive to the chemical state of the metal center. The nature of this sensitivity to the metal oxidation and coordination environment within the zeolite matrix can be better understood through computational NOCV analysis. It enables a deeper understanding of these interactions, as it identifies the regions of electron and spin density flow between the metal center and the N₂ ligand. These fluxes are indicative of weak but significant covalent interactions, even though N₂ itself remains largely non-polar upon the ligation.

The interest in dinitrogen coordination chemistry is mainly associated with gas separation, such as N₂/O₂, CO₂/N₂, or CH₄/N₂²³³ and the challenging activation of the highly inert N₂ molecule. Catalytic nitrogen fixation, as well as air separation, begins with the coordination of N₂ to the transition metal centres. As a ligand, N₂ exhibits certain similarity to the isoelectronic CO, since both molecules are diatomic, having a strong triple bond (σ,2π) and empty antibonding 2π* orbitals. Dinitrogen usually coordinates to metal cations in an end-on (η¹) mode. In contrast to CO, which often forms polycarbonyl complexes (see Chapter 4.2.4), both experimental and theoretical results show that the dinitrogen complexes with the intrazeolite TMI ions usually contain one or two (at low temperatures) N₂ ligands only.^80,234 Typical examples include Cu⁺ and Cu²⁺ centers in ZSM-5^235,236 MOR,²³⁷ and BEA,²³⁸ as well as Ni⁺ and Ni²⁺ hosted in ZSM-5^80,239 or Fe²⁺, Cu²⁺, Zn²⁺ in CHA.²³³ Apart from the most common η¹-attachment, stable trans and cis μ-η¹:η¹ and μ-1,2 adducts of N₂ with Cu⁺ in ZSM-5 (computationally optimized) have also been reported, see Fig. 21a.²⁴⁰ Dinitrogen interacts relatively strongly with the intrazeolite TMI with energies varying from −100 to −20 kcal mol⁻¹, depending on the framework type and accommodation site (see Fig. 21b).^234,240

Fig. 21 Stable binding modes of N₂ to Ni⁺ and Cu⁺ in ZSM-5 (a). Adsorption energies of N₂ on selected metallozeolites (b). IR spectrum of ¹⁴N₂ and ¹⁵N₂ adsorption on Ni⁺/ZSM-5 zeolite (c), together with the calculated IR bands for two possible conformations of di-N₂ adducts (d). The NOCV analysis of the σ- and π-type bonds in {(η¹-N₂)₂Ni⁺}/ZSM-5 complex (e). Panels a, b, and c were based on the data from ref. 234 and 240. Panels d and e adapted from ref. 80 with permission from Elsevier, Copyright 2020.

While forming intrazeolite adducts, N₂ is coordinated by σ- and π-type bonds. The TMI–N₂ bonding depends on the overlap between the available 3d-orbital with the σ (HOMO) and 2π* (LUMO) orbitals of the dinitrogen molecule. Such an overlap pattern implies a synergy of the σ-donation and π-backdonation interactions.²³⁶ Therefore, the resultant electron density redistribution between the metal and N₂ ligand leads to a decrease in the N–N bond order. As a result, the N–N stretching frequency is lower than the corresponding frequency of a gas-phase molecule (2331 cm⁻¹), see, e.g., IR spectra of N₂ with Ni^I/ZSM-5 in Fig. 21c. Apart from orbital contributions, the N–N vibration frequencies also have an electrostatic component, Δν_tot = Δν_(σ+π) + Δν_el, which makes them particularly sensitive to the oxidation state of the metal centre.²³⁶ These features make the N₂/IR method a convenient spectroscopic probe for measuring the electron donor/acceptor properties of the intrazeolite TMI. The N–N stretching frequency of the Cu^I–¹⁴N₂ species is observed in the region of 2300–2285 cm⁻¹,²³⁶ for Ni^I–N₂ at 2254–2252 cm⁻¹.⁸⁰ In contrast, for Ni^II–N₂ it increases to 2324 cm⁻¹,²³⁹ revealing the different extent of back π-donation and the diverse oxidation states of the TMI centres as well.

The Ni⁺ ions hosted in Ni–ZSM-5, in contrast to Cu⁺/ZSM-5,²⁴¹ can coordinate two N₂ molecules simultaneously, and the ensuing {(η¹-N₂)₂Ni⁺}⁹ adducts are characterized by ν_s at 2288 cm⁻¹ and ν_as at 2270 cm⁻¹.²³⁴ The coupling between the two N₂ ligands is usually confirmed by co-adsorption of ¹⁴N₂ and ¹⁵N₂ isotopic mixtures. Because of the presence of mixed ligands (¹⁴N₂ ¹⁵N₂), the resultant structural dichotomy of possible formation of the geminal (η¹-N₂)₂Ni⁺/ZSM-5 and two coupled vicinal (η¹-N₂)Ni⁺⋯(η¹-N₂)Ni⁺ adducts can be resolved by complementary DFT modelling, see Fig. 21a.⁸⁰ The calculated spectroscopic IR signatures of both types of dinitrogen adducts, shown in Fig. 21d, allow for a straightforward distinction between the types of complexes. In particular, the coupling through bonds in the geminal adducts is stronger (25 cm⁻¹) than the coupling through space in the vicinal ones (12 cm⁻¹).

The intimate mechanism of N₂ bonding and the delicate balance between the σ-donation and the back-donation effects, which are responsible for the IR shifts, can be unravelled and quantified by NOCV analysis. The main orbital contributions to the α and β electron density flows are shown in Fig. 21e, together with the corresponding energy values. The α-polarized σ-donation from the N₂ ligand (1α) is accompanied by electron relaxation of the β-3d manifold of the Ni^I centre, consisting of depopulation of 3d_x²−z² and population of 3d_xz (1β), and both flow channels exhibit similar energy contribution to the binding. Such deep electronic relaxation is typical of the TMI that undergoes anisodesmic changes in its structure upon binding (CN = 4 for bare Ni⁺/ZSM-5 centre changes into CN = 3 in the case of dinitrogen adduct), see also Chapters 4.2.4, 4.3.1, and 4.3.2. The π channel of the back-donation is constituted by the 2α + 2β spinorbitals, and involves the out-of-plane orbitals of N₂ and 3d_yz of Ni. The resultant electron density flow is primarily responsible for the lessening of the N–N stretching frequency. The second (3α + 3β) channel of the π-symmetry is formed by the overlap between and 3d_xz orbitals. The resultant flow of electron density along this channel leads to the accumulation of the electron density within the Ni–N bond at the expense of the π and 3d states. Therefore, it can be designated as a dative coordinative contribution. Overall, the π donation/back-donation effects account for 53% of the orbital interactions, whereas the σ channel is in charge of 24% of the binding strength. This allows for unravelling the intimate nature of the Ni⁺–N₂ bonding and the observed bathochromic shift of the stretching vibrations.⁸⁰

4.2.4. Coordination of CO molecules.
Structure of carbonyls and formation of TMI–CO bond. Transition-metal cations hosted in zeolites upon contact with CO form carbonyl adducts rapidly. The CO molecule prefers to bind via a metal–carbon bond, despite carbon being less electronegative than the oxygen atom. The HOMO of CO exhibits an essentially non-bonding character, and this orbital is primarily localized at the carbon moiety. Since the π and π* orbitals are relatively close in energy to the HOMO, they can also be involved in the bonding of carbonyls.

Bonding of CO to TMI in zeolites can be described by the classical synergistic effect of σ-donor and π-acceptor interactions (Fig. 22a), commonly referred to in textbooks as the Dewar–Chatt–Duncanson model. It serves as the basis for more advanced analysis based on quantum chemical calculations.²⁴² Briefly, the σ-donor interaction relies on donating the electron density from the lone pair at the carbon into the empty metal 3d-orbitals, and in this way, a dative bond is formed. The donated electron density raises the energy levels of the 3d electrons due to an increased electron–electron repulsion. As a result, the 3d electrons are more easily accepted by the CO ligand through the π-acceptor interaction. The two lobes of the π* orbitals overlap with the lobes of a metal 3d orbitals of proper symmetry. This type of binding is highly efficient for low-valent metal centers, as the energy of the 3d electron manifold approaches the π* level. The efficiency of the π* back-bonding reflects the stability of the resulting carbonyls. Generally, the more nucleophilic the CO moiety upon bonding, the higher the stability of the adduct. Positive charges destabilize TMI–CO complexes. The stability of neutral carbonyls, which often appear in the zeolite systems, is intermediate. Apart from this short-range electron interaction, specific long-range electrostatic effects imposed by the zeolite lattice, such as polarization and dispersion forces, must be included when analysing the chemical state of intrazeolite carbonyl adducts.^243,244

Fig. 22 Schematic representation of the σ-donation and π-back-donation effects during binding of CO to a transition-metal cation (a). IR spectra of CO adsorption on reduced Cu⁺-exchanged ZSM5, BEA, MOR, and SSZ-13 zeolites. The consecutive sets of spectra correspond to increasing doses of CO, which lead to predominant formation of mono- (b₁), di- (b₂), and tricarbonyl species (b₃) adapted from ref. 257 with permission from the Royal Society of Chemistry, Copyright 2015. A schematic diagram showing the splitting of the C–O vibration IR bands due to the coupling of two and three CO molecules in CuZSM-5 zeolite (c). The ΔG vs. T and p_CO phase diagrams of the isolated Cu⁺, Cu⁺(CO), and Cu⁺(CO)₂ species hosted within SSZ-13, ZSM-5, and MOR zeolites, adapted from ref. 244 under the terms of the CC BY 4.0 license, Copyright 2023, published by Elsevier B.V.

Cu–zeolites are well-suited for studying the coordinative unsaturation of the Cu⁺ cations and binding of carbon monoxide. Exemplary IR spectra of CO sorption at increasing pressure for Cu⁺–zeolites (ZSM-5, BEA, MOR, and SSZ-13) are shown in Fig. 22b. Depending on the CO pressure and temperature, various copper carbonyl adducts are formed. At room temperature and low p_CO, Cu⁺(CO) monocarbonyls are characterized by a single band observed at 2154–2159 cm⁻¹, Fig. 22b₁. Upon introducing more doses of CO, the formation of dicarbonyls Cu⁺(CO)₂ is observed (see Fig. 22b₂), as indicated by splitting of the monocarbonyl band into two peaks corresponding to the asymmetric (ν_asym, 2150–2152 cm⁻¹) and symmetric (ν_sym, 2178–2180 cm⁻¹) vibrations of the two coupled CO oscillators. In such a case, the I_asym/I_sym ratio can be used to estimate the angle the geminal ligands form in the {TMI(CO)₂}ⁿ adducts, using the I_asym/I_sym = tg²(θ/2) equation proposed by Cotton.²⁴⁵ The presence of two bands is indicative of a C_2v local symmetry of the dicarbonyl species.

A further increase in p_CO (Fig. 22b3) leads to the transformation of dicarbonyls into Cu⁺(CO)₃ species, and the appearance of a new triplet of the IR bands with approximate positions at 2135, 2168, and 2192 cm⁻¹. The scheme of splitting of the carbonyl bands upon increasing the number of CO ligands coordinated by Cu⁺–zeolite centers is shown in Fig. 22c. This triplet is indicative of a local symmetry lower than C_3v. The presence of individual species or the joint occurrence of the mono-, di-, or tricarbonyls is linked to the accessibility of the copper centres and the size of the intrazeolite channels and pockets. The IR spectra shown in Fig. 22b1–b3 reveal that ZSM-5 and BEA zeolites are more apt for the formation of well-defined polycarbonyls Cu⁺(CO)_n (n = 1, 2, 3) than MOR or SSZ-13 zeolites, suggesting CO sensitivity to the speciation of the copper active centres.

The stability of the carbonyl species can be predicted with first-principles thermodynamic calculations (see Chapter 3.1) and compared with spectroscopic observations. The calculated phase diagrams (ΔG vs. T and p_CO) shown in Fig. 22d1–d3 delineate the stability regions of [Cu(CO)]⁺ and [Cu(CO)₂]⁺ adducts and the bare Cu⁺ centres located in various sites of the SSZ-13, ZSM-5, and MOR zeolites.²⁴⁴ In all cases, formation of the carbonyl adducts is thermodynamically favoured. The ZSM-5 topology exhibits the highest propensity among the described zeolite series for the formation of [Cu(CO)]⁺ adducts. In contrast, the SSZ-13 framework exhibits the lowest energy difference between the mono- and dicarbonyl species.

The phase diagrams can be used for elucidating the critical conditions for the decomposition of the involved adducts at either a fixed temperature or CO pressure. For instance, at room temperature, transformation between di- and monocarbonyls is irreversible only for the 8-MR sites of the SSZ-13 framework. The decomposition temperature of [Cu(CO)]⁺ to bare Cu⁺ species at 10⁻³ mbar (corresponding to a negligible amount of CO in typical catalytic conditions) in the case of SSZ-13 is much lower (162 °C) than that for the other two frameworks (275 °C and 340 °C for MOR and ZSM-5, respectively), especially for the Cu⁺ cations located in the 6-MR. This influence of the framework topology was observed experimentally during the TPD experiments for the SSZ-13, ZSM-5, and MOR samples, corroborating the FPT results.²⁴⁴

Similar behaviour was observed for Ni⁺-exchanged zeolites subjected to CO adsorption. However, since Ni⁺ centres are paramagnetic, EPR spectroscopy can also be used to elucidate the structure and properties of the resulting carbonyls.^{80,114,247,248} Yet, the primary identification of the adducts formed is provided by IR spectroscopy.^31,239,249 Typical IR spectra are shown in Fig. 23a. The spectral region of the CO vibrations can be separated into two parts, imposed by a band located at 2175 cm⁻¹, which is attributed to the CO molecule interacting with the Brønsted centres. The IR bands located at higher frequencies originate from CO interacting with the oxidized nickel centres (Ni²⁺, Ni²⁺-oxo species). In contrast, those of lower frequencies are associated with CO adsorbed on the reduced intrazeolite Ni⁺ cations.

Fig. 23 IR and EPR fingerprints for the formation of nickel(I) polycarbonyls in the reduced NiZSM-5. The IR spectra of CO adsorption (a) on thermally activated NiZSM-5 (1), sample reduced in CO and evacuated at 150 °C (2), sorption–desorption equilibrium showing conversion of mono- and dicarbonyls (3), sorption of CO excess (4), adapted from ref. 247 with permission from the American Chemical Society, Copyright 2011. The set of CW-EPR spectra obtained after variable pressure CO sorption (up to 30 Torr) (b). Changes in the intensity of the mono-, di-, tri-, and tetracarbonyl EPR spectra recorded during microwave power saturation experiments (c). Decomposition of the experimental EPR spectra into individual carbonyl signals (d). Panels b, c, and d adapted from ref. 246 with the author's permission.

Briefly, the adsorption of CO at room temperature leads to a band at 2214 cm⁻¹ due to the Ni²⁺–CO adducts in ZSM-5. At low temperatures, this band reveals a shoulder at 2198 cm⁻¹ associated with the Ni²⁺-oxo species. After reduction of NiZSM-5 with CO (400 °C), this new band disappears, indicating a reactive character of the adsorption of CO associated with the redox transformation of Ni-oxo species to Ni⁺. This is accompanied by the development of a new doublet signal characteristic of nickel(I) geminal dicarbonyl adducts, with symmetric (2136 cm⁻¹) and antisymmetric (2092 cm⁻¹) vibrations. Desorption of CO at progressively increasing temperatures (Fig. 23a, spectra (2) and (3)) results in transformation of the Ni⁺(CO)₂ dicarbonyl into the Ni⁺CO monocarbonyl adducts (see the band at 2109 cm⁻¹). On the other hand, upon increasing the number of CO doses, a set of bands at 2124, 2118, and 2112 cm⁻¹ appears between the symmetric and antisymmetric dicarbonyl bands, which are attributed to the tricarbonyl Ni⁺(CO)₃ species. The corresponding vibrational splitting scheme is similar to that of Cu⁺(CO)₃ (Fig. 22c), although the positions of the bands are different. Therefore, both Cu- and Ni-containing zeolites provide easily accessible, reduced Cu⁺ and Ni⁺ centres, capable of forming several polycarbonyl adducts, whose stoichiometry and structure depend on CO pressure and temperature. More accurate measurements of the CO frequencies, and the definite confirmation of the polycarbonyl structures (ligation stoichiometry), can be ascertained using isotopically labelled gas mixtures ¹²C¹⁶O/¹²C¹⁸O/¹³C¹⁶O/¹²C¹⁸O.^249,250

Magnetic properties of carbonyl adducts and the binding mechanism. The CO adsorption process can be followed successfully by in situ EPR measurements.^{80,114,239,248} The advantage of EPR over IR spectroscopy stems from the lack of limitations imposed on the CO pressure, as the high pressures lead to the appearance of excessive CO in the gas phase, obscuring the IR spectra of the surface species. The EPR spectra of Ni⁺/ZSM-5 (and Ni⁺/BEA) zeolites, pre-reduced with CO, show dramatic changes upon increasing the pressure of CO (Fig. 23b). CW-EPR measurements at liquid helium temperature with an increasing microwave power allowed to saturate successively the component signals of the individual carbonyl species coexisting upon CO adsorption. Then, using computer simulation, the individual EPR spectra of the adducts [Ni⁺(CO)_n]ZSM-5 with n = 1, 2, 3, 4 (mono-, di-, tri-, and tetra-carbonyl nickel(I)) can be obtained (see Fig. 23c and d). Their magnetic properties have been examined using the DFT calculations of spin-Hamiltonian parameters for model complexes [Ni⁺(CO)_nL_m] (n = 1–4, L = H₂O, OH⁻), which can be regarded as discrete, well-defined molecular analogues of the intrazeolite polycarbonyls.²⁵¹ The molecular nature of the calculated g and ¹³C hyperfine tensors, when using ¹³CO, can be interpreted in terms of the nickel coordination state and its local symmetry, providing clear guidelines for justifying their assignment.

Although the DFT calculations of the g and ¹³C tensors has been restricted to the rudimentary [Ni⁺(CO)_nL_m] models, despite missing of the relevant EPR data for such discrete systems, generic categories of the g tensor structure–relationship can be delineated, and associated with the particular magnetophore [Ni⁺(CO)_n] that is embedded in the zeolite matrices. Additionally, the calculated data can be confronted with the available EPR parameters for nickel(I) polycarbonyls that are produced in ZSM-5 zeolites or supported on silica,^252,253 as the surface functional groups (≡SiO⁻, –OH, ≡Si–O–Si(Al)≡) play the role of mere ligands and can be inserted into the spectrochemical series close to aqua and hydroxyl ligands.²⁵⁴ Additionally, it has been shown that the [Ni⁺(CO)_n] core acts as the proper magnetophore, whose spectroscopic properties are primarily controlled by the spatial arrangement of the CO ligands, being only slightly perturbed by the number and position of the H₂O, OH⁻, or ≡SiO⁻ co-ligands.

The comparison of results obtained from DFT calculations reveals that the polycarbonyl complexes can be divided into two classes with distinct ground states, characterized by distinctive EPR spectra. They are schematically illustrated in Fig. 24a₁ and a₂ and accompanied by the simulated EPR spectra using the DFT-calculated parameters and typical experimental linewidths. The class of the |3d_z²〉 ground state (e.g., T-shaped monocarbonyl, tri- and tetracarbonyls) is characterized by a rhombic EPR spectrum with the g_xx, g_yy components varying between 2.4 and 2.1 (the actual shift depends on the number of CO ligands), and the smaller g_zz value close to g_e = 2.023. In the case of the |3d_x²−y²〉 ground state, observed for the dicarbonyl complexes, the reversed g-tensor values are expected with one component being larger (2.2–2.13) and two components lying closer to the g_e value (2.07–2.01).

Fig. 24 EPR spectra of nickel(I) carbonyls in ZSM-5. Simulated X-band EPR spectra for the generic structural models of Ni⁺ carbonyl adducts with dominant 3d_z² (a₁) or 3d_x²−y² (a₂) ground state of the nickel core (depicted structures show SOMO contours, derived from DFT calculations). Experimental X-band EPR spectra of the isotopically labelled carbonyl species, showing hyperfine structures due to the ¹³CO ligands (hyperfine splitting is shown with stick diagrams) (b₁), (b₂), (b₃). For better comparison of the hyperfine splittings, the spectra are recalculated to the common magnetic field axis and supplemented with the corresponding spin density contours calculated for the identified carbonyl structures. HYSCORE spectra of the monocarbonyl species, including echo-detected EPR spectrum (c₁), where the arrows indicate the observer positions at which HYSCORE spectra were recorded, whereas the unit spheres show orientation selectivity connected with the observer positions. Spin density contour of the Ni⁺CO magnetophore (c₂), and HYSCORE spectra recorded at B₂ (c₃) and B₃ (c₄) positions in the magnetic field. Adapted from ref. 248 and 251 with permissions from the American Chemical Society, Copyrights 2008 and 2013.

It was found that increasing the number n of CO ligands resulted in smaller Δg_ii shifts as the energy separation between occupied and virtual spinorbitals increases from 0.8 (n = 1) to 1.6 eV (n = 4). The other reason stems from the fact that the corresponding matrix elements of the g-tensor (see Chapter 2.2) are dominated by the spin–orbit couplings (SOC) of the metal centre and the ligands. Noting the substantial difference in the SOC constants, λ_Ni⁺ = 565 cm⁻¹ and λ_C = 30 cm⁻¹, the resulting g tensor shifts become reduced considerably in magnitude, upon passing from the mono- to the tetracarbonyl adducts.²⁵¹

Direct confirmation of the structure of the [Ni⁺(CO)_n] adducts formed within the channels of ZSM-5 zeolite can be achieved using EPR measurements with an isotopically labelled ¹³CO. Due to the nuclear spin I = 1/2 of ¹³C, the labelled carbon monoxide can be used for counting the number of the CO ligands and their possible equivalence, as the EPR spectrum acquires an informative superhyperfine structure (Fig. 24b_1–3). Additionally, molecular models of the individual polycarbonyl species, along with the corresponding spin density contours, are also shown in these figures. The ¹³CO EPR measurements eventually prove the presence of the tri- and tetracarbonyls. In the latter case, almost complete solvation of the Ni⁺ adsorption centres takes place, leaving a residual attachment of the adduct to the zeolite backbone with one Ni–O_z bond only. Notably, it has been previously observed that nickel undergoes significant migration upon formation of Ni⁺(CO)_n adducts in NiCaX zeolite.²⁵⁵

The EPR measurements are usually performed at cryogenic temperatures. Only mono- and dicarbonyls were observed in the IR spectra at room temperature. Higher polycarbonyls (n > 2) are possibly formed only at cryogenic conditions. Indeed, a comparison of the energies of the successive adsorption steps for the mono- (−42.9 kcal mol⁻¹), di- (−20.2 kcal mol⁻¹), and tricarbonyl species (−5.7 kcal mol⁻¹) with the TΔS term of CO_(g) reveals that already above 140 °C, the dicarbonyls are unstable. Therefore, for simple thermodynamic reasons, in most cases, only the monocarbonyl species can be expected at actual catalytic conditions. Thus, their structure and properties have been studied in more detail so far.

The expected hyperfine doublet is not resolved in the [Ni⁺CO]–ZSM-5 adduct. This issue has been explained elsewhere,²⁵¹ based on the composition of SOMO and its structural misalignment with the Ni–C bond direction (see Fig. 24c₂). To describe the mechanism of CO–Ni⁺ bond formation, pulse EPR measurements (HYSCORE) are of particular relevance. They allow for the successful determination of the small ¹³C hyperfine couplings.²⁴⁸ The corresponding echo-detected EPR spectrum and HYSCORE spectra are collated in Fig. 24c₁, c₃ and c₄. The interaction of ¹³CO with Ni⁺ZSM-5 under low p_CO pressures gives rise to the formation of a monocarbonyl adduct of the trigonal Ni⁺ with g_zz > g_yy ≫ g_xx > g_e, |A_yy| > |A_zz| > |A_xx|, which is characteristic of the T-shaped conformation (Fig. 24a₁). The binding of CO is accompanied by an intrinsic d–d relaxation of the nickel 3d manifold, due to the allied charge and β-spin density relocation between the 3d_xz and 3d_x²−y² orbitals, which stabilizes energetically the carbonyl adduct. The resolved spectroscopic parameters (g_xx = 2.018, g_yy = 2.380, g_zz = 2.436, A_xx = +1.0 MHz, A_yy = −3.6 MHz, A_zz = −1.6 MHz for ¹³C hyperfine interaction, and e²Qq/h = −13 MHz, η = 0.8 for ²⁷Al quadrupole interaction), reinforced by DFT calculations within the SOMF/B3LYP calculation scheme, provide a suitable background for validation of the T-shaped structure of the Ni^I–CO intrazeolite adduct.²⁴⁸

Electronic details on the CO bonding mechanism can be deduced from the frontier molecular orbital (FMO) interaction diagram, and NOCV analysis of the interacting CO and Ni⁺–zeolite model fragments (see Chapter 4.1.1). The resultant orbital overlap correlation diagram of the π(2p) and σ states of the CO ligand with the 3d states of the zeolite [Ni⁺_2cM7]_frag fragment is shown in Fig. 25a. The ligation of CO results primarily from the sizable π overlap between low-lying π(2p) CO states with the symmetry-adopted 3d_xz and 3d_yz orbitals of the nickel core. It is enhanced by a slight contribution owing to the σ-type overlap of σ(2p + 2s) with one of the lobes of the in-plane 3d_xz orbital, made possible by an unusual T-shape geometry of the CO–Ni–zeolite entity.

Fig. 25 The molecular energy diagram and electronic structure of the Ni⁺–CO adduct in ZSM-5 (a) with the corresponding Kohn–Sham frontier orbital interaction diagram, and NOCV deformation density contours (b). For the charge density (α + β) and the spin density (α − β) contours, the corresponding values indicate the interaction energies associated with the individual channels of the orbital σ and π symmetries. Adapted from ref. 248 with permission from the American Chemical Society, Copyright 2013.

The energy levels of the bonding and antibonding π(2p) CO orbitals define the low and high-energy boundaries encompassing the 3d nickel-based states of the [Ni^I_2cM7] fragment. The interaction of 3d orbitals with the corresponding and MO of the CO ligand gives rise to strong spin polarization, reflected by the pronounced shift of the α and β spin manifolds of the carbonyl adduct. The singly-occupied α-orbital (SOMO) exhibits small delocalization on the CO moiety, which accounts well for the lack of the ¹³C hyperfine splitting in the CW-EPR spectra. Yet, the residual spin density on the ligand is responsible for a very weak ¹³C hyperfine interaction, as detected with the HYSCORE technique.

The delicate balance between the donation and back-donation effects can be quantified using the ETS-NOCV analysis, which enabled the identification and visualization of specific charge and spin flow channels. The main orbital contributions to the α and β electron and spin density flows constituting those channels are shown in Fig. 25b, along with the corresponding energy values. As a result, the analysis enabled the disentanglement of the cooperative σ donation and π back-donation effects involved in the ligation of a CO molecule to Ni⁺ ions hosted in zeolites.²⁴⁸ The dominant effect is the back-donation along the π channel of a spinless charge transfer. This explains the position of the carbonyl band in the IR spectrum and its bathochromic frequency shift. The spin density redistribution within the CO moiety primarily operates via the 4α and (3α + 4β)σ channels. Due to the substantially uneven contributions of the 3α and 4β components, the ensuing unbalanced spin flow leads to a spin polarization of the carbon sp-hybrid. On the contrary, the 4α channel allows for a direct delocalization onto the carbon 2s orbital. Since the polarization effect prevails, this explains the negative sign of the isotropic constant a_iso(¹³C).²⁴⁸ These two effects virtually cancel each other, resulting in a minute net spin density flow within the σ channel. Because upon the CO attachment the unpaired electron relaxation (1β) is restricted to the 3d-based orbitals of nickel, and the main channels (one σ and two π) of the electron density flow within the Ni⁺–CO unit involve both α and β spins in a similar manner, the magnetophore of the monocarbonyl adduct is essentially constituted by the metal core. All the discussed effects provide a rational basis for an in-depth molecular interpretation of the IR and EPR parameters in terms of charge and spin flows along the σ and π channels.

Quantification of TMI with CO as an IR molecular probe. The IR spectra of adsorbed CO can be used to explore the nature and content of TMI in zeolites, which primarily relies on the formation and quantification of C-bonded carbonyls. The small kinetic size of the CO and its high affinity to ligation make this molecule well-suited as an effective probe for investigating Lewis and redox sites present inside the channels of zeolites. The vibrational spectrum of the diatomic CO molecule is relatively simple and strongly depends on the nature of the adsorption site. The advantage of CO as a probe molecule lies in its sensitivity to the speciation of TMI due to the subtle balance of electrostatic effects, σ-donation and π-back-donation as well as the static and dynamic interactions between the adsorbed molecules.^256,257 As a result, the CO-IR technique can inspect various valence states of TMI and their speciation (bare TMI, isolated TMI–oxo species, oxide nanoclusters). Additionally, it is often possible to selectively detect cations in different valence and coordination states, even when present simultaneously.²⁵⁸ Yet, although CO can serve as a probe for analysing the coordination vacancies of exchangeable cations within zeolites, in some cases, the adsorption of CO is isodesmic, i.e., formation of new TMI–CO bonds occurs at the expense of the TMI–O_z bonds with the zeolite framework, keeping the overall coordination number intact.²⁴⁸ Spectroscopic fingerprints for commonly studied intrazeolite carbonyl adducts of Co, Cu, Fe, and Ni are presented in Table 4.

Table 4 Spectoscopic fingerprints for commonly studied intrazeolite carbonyl adducts of Co, Cu, Fe, and Ni

Intrazeolite TMI–CO adduct		Frequency/cm^−1	Ref.
Co/MFI	Co^2+–CO	2204	32
	(CoO)n–CO	2194
Ni/ZSM-5	Ni^2+–CO	2220–2180	31, 80, 249 and 259
	Ni^+–CO	2109
Cu/ZSM5	Cu^+–CO	2157	260
	Cu^0–CO	2124
	Cu^2+–CO	2206
Fe/MFI	Fe^2+–CO	2196	261 and 262
	Fe^2+–(CO)2	2188

---

There are certain limitations to using CO as a probe, primarily when reactive CO adsorption occurs, resulting in chemical modification of the TMI centre by the adsorbed species. The typical products of reactive CO adsorption include carbonates, bicarbonates, carboxylates, and even formates. In some instances, CO can also be oxidised to carbon dioxide.²⁶³ The undesired reactive CO adsorption can often be mitigated by using low-temperature experiments. These issues should be considered when aiming for a more accurate determination of the oxidation states of TMI ions in zeolites. A more general account of the application of IR spectroscopy in zeolite research can be found in previous works.^256,257

After establishing the experimental conditions for generating the monocarbonyl adducts (see the ΔG(p,T) diagrams), IR absorption coefficients can be obtained through careful CO titration under controlled conditions (CO pressure and temperature). The most convenient situation occurs when the specific probe molecule reacts selectively with only one type of adsorption center, and the stoichiometry of this process is well understood. It can be inferred, for instance, from the FPT phase diagrams (Fig. 22d₁–d₃), which also reveal differences in the formation of monocrbonyls between the zeolite frameworks and TMI localisation within the particular channels.²⁴³ In most cases, it is reported that the value of the absorption coefficient for the respective carbonyls is independent of the zeolite structure, despite the observed various CO stretching frequencies.

In the case of cobalt cations accommodated in the pentasil zeolites (e.g., MFI, FER), the CO band at 2204 cm⁻¹ is attributed to Co²⁺–CO adducts formed with the exchanged Co²⁺ cations, whereas the bands at 2194 cm⁻¹ and 2179 cm⁻¹ are associated with Co²⁺ in oxide clusters and segregated CoO, respectively.³² The plot of the intensity of the characteristic monocarbonyl IR band as a function of the amount of CO sorbed is used to determine the absorption coefficient. For the 2204 cm⁻¹ band, the absorption coefficient is equal to 0.123 ± 0.002 cm² μmol⁻¹, and its value is independent of the zeolite type and the adsorption temperature.³² For the 2194 cm⁻¹ band, a value of 0.129 ± 0.004 cm² μmol⁻¹ has been reported, based on the intensity (height) of the corresponding IR signal. In addition, the absorption coefficient of the CO–OH band at 2175 cm⁻¹ due to adsorption of CO on Brønsted centres is equal to 0.120 ± 0.003 cm² μmol⁻¹ (determined for FER zeolite). All these coefficients can be used to evaluate the complete balance of cationic sites in pentasil zeolites from a single IR adsorption measurement.

Coordination of CO to Ni²⁺ cations leads to the formation of Ni²⁺–CO monocarbonyl adducts with the IR bands located in two spectral regions: 2220–2180 cm⁻¹ and 2170–2130 cm⁻¹.^80,249,259 The low-energy region is characteristic of carbonyls formed on nanosized NiO and mixed oxides. The high-energy band is characteristic of isolated Ni²⁺ cations present in the exchangeable positions. The absorption coefficient for this band in NiZSM-5 zeolite (2212 cm⁻¹) equals 0.143 ± 0.003 cm² μmol⁻¹.³¹ The same value was obtained for NiY, despite slight differences in the CO stretching frequencies of the Ni²⁺–CO monocarbonyls (2212 and 2220 cm⁻¹ for NiZSM-5 and NiY, respectively). During the reactive adsorption of CO on nickel–oxo compounds, CO acts simultaneously as a reducing agent and as a probe for the generated Ni⁺ cations.²⁴⁷ The characteristic band of the Ni⁺–CO adduct is located at 2109 cm⁻¹, and the absorption coefficient is equal to 1.104 ± 0.018 cm² μmol⁻¹.³¹ A strong increase in the absorption coefficient of Ni⁺–CO compared to Ni²⁺–CO originates from more efficient transfer of the 3d electron density from the Ni⁺ cation onto the π* antibonding orbital of CO (π-back-donation effect) due to the better energy matching of the overlapping orbitals.²⁴⁹

The interaction of CO with copper sites in thermally treated ZSM-5 zeolite results in the formation of a strong band at 2157 cm⁻¹ due to Cu⁺–CO monocarbonyls at low CO doses. In contrast, the band at 2130 cm⁻¹ produced when CO was sorbed in higher amounts was attributed to Cu⁺ ions present in an oxide form.²⁶⁰ Treatment of CuZSM-5 zeolite with H₂ results in the appearance of a band at 2124 cm⁻¹, attributed to the monocarbonyls of metallic copper (Cu⁰–CO). When oxidised with O₂, some of the Cu⁺ cations in exchange positions are transformed into Cu²⁺ species, which is recognised by the development of a band at 2206 cm⁻¹ due to the Cu²⁺–CO monocarbonyls. By measuring the intensity (height) of the 2157 cm⁻¹ band corresponding to the sorbed CO, the slope of the resulting linear correlation defines the extinction coefficient, yielding a value of 1.30 cm² μmol⁻¹. The extinction coefficient of CO interacting with Cu²⁺ species in zeolite CuZSM-5 was obtained after treatment with oxygen. The intensity of a relatively weak Cu²⁺–CO band at 2206 cm⁻¹ can be used to calculate the absorption coefficient, which equals 0.11 cm² μmol⁻¹. Note that this value is similar to other divalent cations hosted in ZSM-5 (see above). The experimental and calculation details can be found elsewhere.²⁶⁰

Iron speciation in zeolites is also extensively studied using IR spectroscopy with a CO probe. For instance, the corresponding carbonyl bands appear at 2200–2194 cm⁻¹ for the zeolites BEA,²⁶⁴ FER,²⁶⁵ and MFI.²⁶¹ In general, iron species in zeolites prepared from a ferrous solution were found to be mainly in the Fe²⁺ form. Fe²⁺–CO monocarbonyls (with the band at 2195 cm⁻¹) convert easily into Fe²⁺(CO)₂ dicarbonyls (2188 cm⁻¹) at low temperatures, but the reaction progress depends on the exchangeable sites within the zeolite. Other identified Fe²⁺ ions in FER produced exclusively monocarbonyls (band at 2189 cm⁻¹), and in comparison to the previous hosting site, they were easily oxidised to Fe³⁺ ions in the presence of oxygen (or air). Sorption of CO on FeZSM-5 zeolite resulted in monocarbonyl bands at 2188 and 2200 cm⁻¹, which were attributed to the Fe²⁺(CO) based on the results of oxidation/reduction reactions with O₂ and H₂.²⁶² The calculated extinction coefficient based on the integral absorption is equal to 2.52 ± 0.07 cm μmol⁻¹, while the value obtained from the band height measurement drops to 0.11 ± 0.01 cm² μmol⁻¹.

4.3. Non-innocent molecules – ligation pathways

Non-innocent (ambivalent) ligands such as O₂/O₂⁻/O₂²⁻, NO/NO⁺/NO⁻, NO₂/NO₂⁻ play an essential role in a large variety of catalytic reactions involving metallozeolite catalysts, including HC (hydrocarbons) oxidation and HC-SCR,^{17,156,157,266} NH₃-SCR,^{156,188,267,268} deNO_x^269,270 or oxidation of CH₄^271–275 and NO^172,276,277 in particular. Their ambiguous electronic and magnetic structures, as well as their electronic flexibility upon binding to intrazeolite TMI, pose challenges for the clear assignment of the metal and ligand oxidation states, and therefore, for establishing their distinct roles in the redox mechanisms of the investigated reaction. In contrast to the innocent H₂O or NH₃ ligands (which are characterised by a well-defined charge that is usually preserved upon ligation, see Chapter 4.2), the non-innocent ligands can change their redox states due to breaching of the Pauli shield, and participate in electron transfer to or from the TMI centres. As a result, the valence and coordination states of non-innocent ligands can change to facilitate the mechanistic redox steps during a catalytic reaction. The non-innocent ligands can play an auxiliary role, meaning that they do not interact directly with the reactants, but alter the redox state of the metal centre (redox functionality tuning), or participate explicitly in the reaction by making and breaking covalent bonds with the captured reactant molecules.¹⁷² For instance, the appearance of an unusual 3d⁹ state of cobalt is made possible upon the coordination of redox-active NO ligands to Co²⁺ hosted in ZSM-5, which triggers the subsequent capture of olefin molecules by the Co⁰ centres for driving the HC-SCR reaction efficiently with the involvement of the Co²⁺/Co⁰ redox couple.¹⁷ An intricate network of NO oxidation¹⁷² and NO reduction with ammonia over Cu/CHA and Cu/MFI catalysts^199,278,279 or with ethylene over NiSZM-5¹⁵⁶ can also be invoked here as illustrations.

To circumvent the ambiguity of oxidation state assignments and dichotomy of the electrons division, the Enemark–Feltham notation {XYM}^d+π can be used, where XY indicates a diatomic ligand, M – metal, and d + π is the sum of the electrons on d orbitals of M and π orbitals of XY.²⁸⁰ Two alternative coordination situations between a transition metal cation, M^p/p±1/z, and an attached non-innocent ligand, L^q/q±1, involve resonance structures, XY^qM^p/z ⇌ XY^q−δM^p+δ/z (redox counterparts²⁸¹), with the delocalized electrons in a single or double energy minimum. In this notation, p and q represent the charges of the TMI and XY ligand, respectively, whereas δ corresponds to the partial charge shifts. This ambivalence is particularly amplified in the case where the 3d levels of the intrazeolite transition metal cations, such as Co, Ni, or Cu, are intermingled with the 2π* levels of the non-innocent XY ligands (Fig. 26a). The classic ligation dichotomy involves here spin pairing and electron transfer categories.

Fig. 26 Alignment of the 2π* levels of the non-innocent NO and O₂ molecules to the 3d levels of Co, Ni, and Cu cations hosted in the exemplary ZSM-5 zeolite (a). Isodesmic routes of the intrazeolite coordination via spin addition, spin pairing, and spin crossing coupled with pairing mechanisms (b), calculated at the ZORA/TPZ UKS level.

The alignments of the energy levels shown in Fig. 26a favour electron transfer from Cu⁺ and Ni⁺ to O₂ and spin-pairing for NO. In the case of Cu²⁺ and Co²⁺, since the β-3d manifold is located below the 2π* level, coordination of the donor co-ligands (e.g., NH₃) is needed to trigger the MLET (see Chapter 4.3.1). Ligation of open-shell non-innocent XY molecules leads to changes in the overall spin state of the {XYM}^d+π unit. Exemplary scenarios collated in Fig. 26b illustrate computational isodesmic pathways of spin addition, spin pairing, and spin crossing coupled with spin pairing upon NO molecule coordination to selected transition metal ions of different spin multiplicity hosted in the MFI matrix (see Chapter 4.3.2. for more details).

4.3.1. Binding and activation of O₂. The interaction of O₂ molecules with intrazeolite transition metal ions yields a variety of mono- and bidentate M–O₂ and bridging M–O₂–M adducts (Fig. 27a), which are capable of the O–O bond scission or direct attack on an organic substrate, exhibiting an intricate combination of interfacial coordination and redox chemistries. It involves the partial reduction of the ligated O₂ via one- (1e⁻) or two-electron (2e⁻) transfers from the TMI active sites, producing superoxide and peroxide 1 : 1 and 1 : 2 species. The deep four-electron dissociative reduction of O₂ on single sites occurs at electron-rich TMI. However, attainment of elevated oxidation states (z ≥ 4) is hindered by high energy costs, particularly for those TMI that are positioned after Mn in the periodic table. Actually, it takes place in a more complex manner, favourably on dimeric centres (1 : 2), and can be promoted by coupling with proton relocation (electroprotic route), leading to more stable hydroperoxo M–OOH and hydroxyl M–OH intermediates. Various combinations of these processes, involving the removal of the inert triplet ground state of molecular oxygen (³Σ_g⁻) and an apposite stabilisation of the resultant species by coordination to the metal centres, make the activation thermodynamically more favourable and kinetically accessible. Isotopic labelling is widely explored in the elucidation of the O₂ binding and activation processes, being particularly useful for identifying activated oxygen species and studying the activation mechanisms, using ¹⁸O (IR, RS) and ¹⁷O (EPR) labelled dioxygen.

Fig. 27 Typical binding modes of oxygen to mono- and dinuclear metal centres hosted in zeolites (a). The influence of Al distribution on the stability of the cis μ-η¹:η¹-O₂ di-copper adducts presented in TON zeolite (b).²⁸⁵ The effect of the zeolite framework on the structure of η²-O₂–Co in various sites of the distinct topology of MFI (c), and on the structure of η¹-O₂–Cu, η²-O₂–Cu, and trans μ-η¹:η¹-O₂ di-copper adducts in MOR and CHA zeolites (d).²⁸⁷ The d-O–O vs. 1/(ν_O–O)^2/3 2D correlation diagram for classification of oxygen adducts with the intrazeolite transition metal ions (e). The calculated positions of the reference O₂, O₂⁻, H₂O₂, and bare η¹-O₂–Cu, η²-O₂–Cu and cis, trans μ-η¹:η¹-O₂, μ-η²:η²-O₂ species (terminated with –OH) are marked in large circles, whereas the positions of the corresponding adducts of Cu, Co, Fe, hosted in CHA, TON, MFI, MOR zeolites are marked with small circles.^286,288 Panels a, b, d, and e based on data from the quoted references. Panel c adapted from ref. 286 with permission from the American Chemical Society, Copyright 2020.

The variety of ways in which single and two adjacent TMI cations (M∪M) can bind an O₂ molecule includes a considerable diversity of the resultant structural motifs.^282–285 End-on η¹-O₂ superoxide (Pauling structure), the side-on η²-O₂ superoxide or peroxide (Griffith structure), the cis μ-η¹:η¹ (Yeager structure) and trans μ-η¹:η¹, side-on μ-η²:η², μ-η²:η¹, as well as the bis(μ-oxo) or η²⁻ozonide molecular adducts with Fe, Co, Ni, Cu or Zn are the most common in the case of metallozeolites. Their energetic, structural, and spectroscopic characteristics depend on Al location within the rings (Fig. 27b), accommodation locus in the given zeolite framework (Fig. 27c and d), and the framework type (Fig. 27d).^285–287 The bridging peroxo species, upon further reductive dissociation, can be converted into highly reactive terminal (such as Fe=O) or bridging (e.g., Cu–O–Cu), mono-oxo species, accomplishing the complete 4e⁻ reduction and splitting of O₂, discussed below.

Dioxygen attachment and activation on single TMI centres. The reductive monodentate (η¹) and bidentate (η²) attachment of O₂ to single intrazeolite TMI centres can occur through the σ, π and δ-overlap (see Fig. 17c) between the metal donor 3d (or 4s for Zn) and the 2π* acceptor orbitals of dioxygen leading to the formation of paramagnetic superoxide (1e⁻ transfer) or diamagnetic peroxide (2e⁻ transfer) 1 : 1 species. However, due to the simultaneous operation of bonding and back-bonding interactions, fractional changes in the charge and spin transfers between the metal and ligand are actually observed. Reasonably, the adducts with the O–O bond length of 1.4–1.5 Å and ν_O–O 750–930 cm⁻¹ are designated as peroxides, whereas those with 1.2–1.3 Å bond length and ν_O–O lying in the range of 1050–1200 cm⁻¹ are classified as superoxides.²⁸⁹ Their dinuclear nature can be deduced from the Δ¹⁸O shifts (∼50–60 cm⁻¹) in IR spectra, and from the character of ¹⁷O hf pattern in the EPR spectra of the paramagnetic superoxide complexes (11 equidistant lines are diagnostic for η² and up to 36 lines for η¹ attachment).²⁹⁰ The correlation between the d_O–O bond length and the ν_O–O vibration frequencies based on empirical Badger relation (d_O–O = C/(ν_O–O)^2/3 + b), established previously for homogeneous complexes,²⁸⁹ shows that the vibration frequencies smoothly span the whole range from the superoxide-like to the peroxide-like species (Fig. 27e). As a result, assignment of the discrete (integer) oxidation states of oxygen species has a formal character only. Thus, the covalent communication between the metal centre and the O₂ ligand leads to the formation of intrazeolite adducts, which may be regarded as valence hybrids of the limiting superoxide and peroxide extremes.²⁸⁹ In addition to assigning the formal oxidation states to metal and dioxygen moieties, the mode of binding and the orbital overlap responsible for bond formation, as well as the MLET pathway, are of key importance. Especially, recognising the impact of the O₂ activation type on the catalytic reactivity with reactant molecules.

The intrazeolite superoxide adducts can then be categorised into the σ, π and δ-type paramagnetic species, exemplified by the [η¹-O₂⁻–Zn^II]⁺/MFI,²⁹¹ [η¹-O₂⁻–Co]²⁺Y²⁹² and [η²-O₂⁻–Ni]⁺ZSM-5⁶⁴ cage adducts identified univocally by EPR spectroscopy using ¹⁷O-labelling.

The Zn⁺/ZSM-5 (Si/Al = 11.9) species react with O₂ (5 mbar, 300 K) with the formation of [O₂˙⁻Zn]⁺/MFI adducts, as demonstrated by the EPR spectrum with g_zz = 2.039, g_yy = 2.0108, g_xx = 2.0033 (Fig. 28a).^78,291 Based on the internuclear bond distance of d_O–O = 1.33 Å and ν_O–O = 115 cm⁻¹ (see Fig. 27e) this adduct has been assigned to top-on [¹η-O₂˙⁻–Zn]⁺/ZSM-5, produced by transfer of the electron density (0.7–0.8) from 4s¹ to the in-plane orbital, via the σ-donation channel that is favourable for the top on ligation (Fig. 28a₁). The HOMO is mainly responsible for the O₂ binding (−150 kJ mol⁻¹), whereas the out-of-plane SOMO is localized essentially on the oxygen moiety, interacting weakly with the Zn ion.

Fig. 28 X-band CW-EPR spectrum of the [η^1-16O₂⁻–Zn]⁺/ZSM-5 adduct (a) with the corresponding MO energy diagram (a₁) (adapted from ref. 291 with permission from the American Chemical Society, Copyright 2013) and of [η^2-17O₂⁻–Zn]⁺/ZSM-5 (b) with the corresponding NOCV analysis of the spin-polarized MLET along the β-channel and back donation along the α-channel (b₁). EPR spectrum (c),²⁹³ and UV-vis spectrum (c₁) of the [η²-O₃⁻–Ni]⁺/ZSM-5 adduct (adapted from ref. 294 with permission from the American Chemical Society, Copyright 2021). The EPR spectrum of the cage [η¹-O₂⁻–Co(NH₃)_n]²⁺/ZSM-5 adduct (d) together with the associated MO diagram of the energy levels (d₁).¹¹³ Panels b, b₁, d, and d₁ are inspired by the quoted references.

An alternative η²-attachment of dioxygen, distinct by a higher g_zz = 2.061 value (and similar g_yy = 2.010, g_xx = 2.003), is disclosed by the ¹⁷O hyperfine structure with the resolved 11 equidistant lines that are separated by A_xx = 7.68 mT (Fig. 28b), observed in the EPR spectrum when ¹⁷O₂ is contacted with Zn^I/ZSM-5 (Si/Al = 140).²⁹⁴ Such ¹⁷O hf pattern confirms that both oxygen atoms are magnetically equivalent, which is tantamount to the side-on ligation of dioxygen with ΔE_ads = −110 kJ mol⁻¹. The internuclear bond distance of d_O–O = 1.24 Å is indicative of the superoxide nature of the bound O₂. The NOCV analysis reveals that formation of the O₂˙⁻ species results from the dominant flow of electron density from the 4s¹ to 2π* orbital in the β-channel, which is accompanied by a small back-donation of the α-spin density towards the zinc core (Fig. 28b₁). The recognition of both the top-on [η¹-O₂˙⁻–Zn]⁺/MFI (Si/Al = 11.9) and the side-on [η²-O₂˙⁻–Zn]⁺/MFI (Si/Al = 140) adducts illustrates a subtle role played by the intermolecular electrostatic interactions in supporting the particular binding mode of dioxygen. It is also plausible that the η² superoxide adduct is actually produced via the η¹ intermediate, since the nil overlap between the redox 4s¹ to 2π* orbitals in the side-on mode impedes charge flow. In the case of [Zn–O˙⁻]/MFI species where the spin density is mainly localized on the ligand (characterized by g_‖ = 2.37 and g_⊥ =1.98 and the UV/vis band at 12 020 cm⁻¹), a room temperature interaction with O₂ (p_O2 > 0.1 Torr) leads to the reversible formation of a zinc–ozonide adduct, [Zn–O˙⁻]⁺/MFI + O₂ ↔ [η²- O₃˙⁻Zn]⁺/MFI, produced by the attack of dioxygen at the O˙⁻ moiety (see Chapter 4.1 and Fig. 17d). It is characterized by g_xx = 2.0019, g_yy = 2.0106, g_zz = 2.0124, A_xx = −8.37, A_yy = 0.14, A_zz = 0.36 (Fig. 28c), and the n_b → π*(O–O–O) band centred at ∼450 nm, which is featured by a well resolved vibronic progression due to the stretching O–O–O oscillations with the 860 and 900 cm⁻¹ splitting for η^2-16O₃˙⁻Zn^II and η^2-18O₃˙⁻Zn^II species, respectively (see Fig. 28c1).²⁹⁴ The binding of O₂ occurs via an [η¹- O₃˙⁻Zn]⁺ intermediate,²⁹³ and a rather small overall energy of this process (96 kJ mol⁻¹), together with low activation barriers (the highest is equal 68 kJ mol⁻¹), accounts well for the experimentally observed complete reversibility of this process. The a₁-bonding orbital (labelled in the local C_2v symmetry of the zinc–ozonide magnetophore) with the largest overlap between the Zn and O₃ counterparts, contributes mainly to the Zn–O₃ bond. In contrast, the b₁-SOMO is almost entirely localised on the bidentate ozonide moiety. The remaining orbital interactions (a₂, b₁, and b₂ MO) are less important for the strength of the Zn–O₃ interaction due to the small overlap. The distinct polarisation of two O–O bonds in the O₃ ligand lead to a unique pattern where both side oxygens that are directly bound to Zn²⁺ become more negatively charged (−0.415|e|). In contrast, the central oxygen acquires surprisingly a partially positive charge (0.202|e|), in comparison to the free ozonide anion (−0.167|e|). The zeolite framework plays a vital role in maintaining the effective charge of the zinc cation at approximately +2, stabilising the highly polarised η²-O₃⁻–Zn²⁺ bond.

In the case of di-oxo vanadium ↑VO₂/BEA, a room temperature contact with dioxygen (2 Torr) leads to the formation of a bent (α = 16°) top-on [η¹-O₂⁻ VO₂⁺]/BEA complex (g_z = 2.0236, g_y = 2.113, g_x = 2.0054, A_z = 0.98, A_y = 0.68, A_x = 0.54 mT) via the MLET, d_z² → 2π*, pathway. The π*-type SOMO is essentially confined to the η¹-O₂⁻ moiety.²⁹⁵

The π-type adducts are exemplified by superoxide species attached to cage Co²⁺(NH₃)_n complexes. The intrazeolite Co²⁺ (S = 3/2) cations cannot directly activate O₂ into the superoxide state due to the inverse energy levels (IEL) of the donor 3d and acceptor 2π* orbitals (see Fig. 26a). However, when boosted by the coordination of stronger σ-donor NH₃ co-ligands (which can be allied with an “orthosteric effect”) the resultant [Co(NH₃)_n]²⁺ adducts hosted in FAU(Y)²⁹² and ZSM-5 zeolites¹¹³ upon exposure to O₂ form the corresponding [η¹-O₂⁻Co(NH₃)_n]²⁺ heteroleptic superoxide complexes identified readily by EPR spectroscopy (Fig. 28d). The reductive binding of O₂ is accompanied by a spin crossing in the 3d manifold of the cobalt core (↑↑O₂ (S = 1) + ↑↑↑Co²⁺ (S = 3/2) → ↑O₂⁻–Co³⁺↑↓ (S = 1/2)), giving rise to the observed doublet state of the superoxide adduct. In an alternative spin pairing process (↑↑O₂ (S = 1) + ↑↑↑Co²⁺ (S = 3/2) → ↑O₂⁻↑↓Co³⁺↑ (S = 1), the adduct should assume a triplet state. Yet, the g-tensor anisotropy (g₁ = 2.084, g₂ = 2.01, g₃ = 2.00) is diagnostic for the formation of a superoxide species with S = 1/2. The superhyperfine structure (^CoA₁ = 1.78 mT, ^CoA₂ = 1.2 mT and ^CoA₃ = 1.25 mT)²⁹² can be interpreted in terms of the β-electron density transfer to formed by the out of plane π-type overlap, and the spin polarization of the close-lying in-plane of the mixed π/σ character, as discussed elsewhere.²⁹⁶ Such in-plane d/π interactions impart a bent configuration of the Co–O–O moiety and local C_s symmetry of the ensuing magnetophore. A qualitative molecular energy diagram shown in Fig. 28d1 accounts for the reductive coordination of O₂ by [Co(NH₃)_n]²⁺/ZSM-5 via the MLET mechanism associated with spin crossing.

The superoxide η²-species can also be produced in the δ-type MO overlap. This relatively rare binding mode of O₂ results from an isodesmic ²Ni⁺_4c/ZSM-5 + ³O₂ → [η²-O₂⁻Ni_2c]⁺ZSM-5 process, where the initial coordination number of the nickel cation (Ni⁺_4c) equalled 4, is preserved. The corresponding EPR spectrum obtained for ¹⁷O-enriched dioxygen is shown in Fig. 29a. The 11-line pattern proves the side-on attachment of the dioxygen ligand. The diagnostic features of the heteroleptic [η²-O₂⁻Ni_4c]⁺ZSM-5 adduct (g_e < g_xx(3) = 2.0635 < g_yy(2) = 2.0884 < g_zz(1) = 2.1675, A_yy(2) (5.67 mT) ≫ A_xx(3) (1.0 mT) ∼ A_zz(1) (1.3 mT)) with the highest hf splitting along the g_yy(2) mid value direction are distinctly different from those observed for the π-type O₂⁻ adducts (g_e ≈ g_xx(3) < g_yy(2) < g_zz(1), and A_zz(1) > A_xx(3) ∼ A_yy(2) with the maximum of hf splitting along the highest g_zz(1) value). The [η²-O₂⁻Ni]⁺ZSM-5 complex is best described as a mixed metalloradical with two supporting oxygen donor framework O_z ligands and nearly even spin density repartition between the superoxide moiety and the nickel centre (ρ_Ni = 0.36, ρ_O(1) = 0.32, ρ_O(2) = 0.30), as shown in Fig. 29a₁. The associated changes in the charge distribution upon O₂ attachment (Δq_Ni = 0.32, Δq_O = −0.09, Δq_O = −0.11) confirm the MLET mechanism of dioxygen activation with back donation playing a crucial role (see below). Notably, the observed triangle spin density and charge distribution resemble those observed for η²-O₃˙⁻Zn²⁺ species, where the middle Ni cation and the middle oxygen are both positively charged.

Fig. 29 X-band EPR spectrum of [η²-O₂⁻Ni_4c]⁺/ZSM-5 adducts (a), and the structure of the η²-O₂Ni moiety with the spin density distribution (a₁) adapted from ref. 64 with permission from the American Chemical Society, Copyright 2011. NIR spectra of η^2-16O₂ and η^2-18O₂ adducts of nickel (b), and cobalt (c) hosted in MFI zeolite, and the corresponding structures of these species (b₁) and (c₁), adapted from ref. 286 and 297, with permissions from the American Chemical Society (Copyrights 2020 and 2019, respectively). Thermodynamic stability diagram of η¹-O₂ and η²-O₂ adducts of Cu²⁺ and Cu⁺ hosted in CHA (d), based on data from ref. 83.

The SOMO orbital is constituted by a δ overlap between the out-of-plane MO of dioxygen and the 3d_x²−y² orbital of nickel, whereas LUMO results from the π-type overlap between d_zy and in-plane orbitals (see Fig. 30a), following the superoxide nature of the [η²-O₂⁻Ni]⁺ZSM-5 species. The overall binding energy of dioxygen, equal to −16 kcal mol⁻¹, results from the molecular (−141 kcal mol⁻¹) and electrostatic (−87 kJ mol⁻¹) contributions that are attenuated by a sizable steric repulsion (+212 kJ mol⁻¹).

Fig. 30 Molecular orbital correlation diagram of O₂ binding to Ni⁺/ZSM-5 (a). The NOCV contours that describe the bonding between the dioxygen and the intrazeolite Ni⁺ cation (b₁), along with the corresponding diagram showing the charge and spin flow channels and the associated energies (b₂), together with the orbital channels of the overall charge (Δq) and spin (ΔS) flows (b₃). The red and blue contours correspond to loss and gain of the electron and spin density, respectively. Adapted from ref. 64 with permission from the American Chemical Society, Copyright 2011.

The intrazeolite [η²-O₂⁻Ni_4c]⁺/ZSM-5 species exhibit significantly different electronic and magnetic structure than the akin dioxygen adducts with nickel in homogeneous complexes,^298,299 where the spin pairing mechanism prevails. The SOMO is then constituted mainly by the 3d_z² orbital, giving rise to g_zz ≪ g_xx, g_yy, and a small ¹⁷O hf splitting (which usually remains unresolved). Thus, dissimilar magnetic structures of both types of adducts can be accounted for by the different mechanisms of dioxygen binding, which are based on coupled electron and spin transfer in the case of [η²-O₂⁻Ni_4c]⁺ZSM-5, and a spin pairing model for the homogeneous complexes. The ETS-NOCV analysis of the [η²-O₂⁻Ni_4c]⁺ZSM-5 electronic structure reveals three distinct orbital channels of the congruent and incongruent charge and spin density flows, associated with δ, π, and σ overlap, which contribute jointly to the attachment energy and the reductive activation of the dioxygen ligand (Fig. 30b1 and b2).

The unequal charge and spin density transfers along the π-channel result from the π-donation from 3d_yz to (1β) and the π-backdonation (1α) processes, with highly uneven energetic contributions −83.2 and −5.0 kcal mol⁻¹, respectively. The δ-channel (2β) of a congruent but opposite spin and charge transfer contributes to −30.5 kcal mol⁻¹ of the O₂–Ni interaction energy, whereas the spinless charge back-donation, , through the σ channel (2α + 3β) is associated with a relatively small energy gain of −14.2 kcal mol⁻¹. As a result, the β-spin polarised charge δ-donation, reinforced by the partly counterbalancing β-charge donation and α-polarised back-donation along the π-channel, together with a small charge-only σ-back-donation (Fig. 30b₂, and b₃) are mainly responsible for the formation of nickel-bound superoxide O₂⁻ radical of the (π)²(δ)²(3d_z²)²(3d_x²−y²)²(3d_zy²)²(δ*)¹(π*)⁰ configuration, and its S = 1/2 ground state. This illustrates the inherent complexity of what appears to be a simple process.⁶⁴

The vibronic progression with the O–O stretching frequency of 1010 cm⁻¹ and the isotopic shift, Δ¹⁸O = 30 cm⁻¹, associated with the δ → δ* excitation seen in the NIR spectra (Fig. 29b) provides another argument that the δ channel interaction contributes to the formation of the superoxide species.²⁹⁷ The bonding δ orbital is mainly localised on the Ni²⁺ centre, and the antibonding δ* orbital on the O₂⁻ moiety (Fig. 29b₁), hence the δ → δ* transition lengthens the O–O bond from 1.308 to 1.405 Å, contributing additionally to dioxygen activation via MLCT. Furthermore, the ionicity of the δ bond has been linked to the abnormal electrophilic reactivity of the unoccupied β-2π* orbital of the superoxide moiety.²⁹⁷

In contrast to the superoxide species produced on Ni⁺/ZSM-5, the side-on attachment of O₂ to Co⁺ hosted in ZSM-5 leads, via 2e⁻ transfer, to the formation of a peroxo [η²-O₂²⁻Co]⁺/ZSM-5 adduct (Fig. 29c).²⁸⁶ Actually, this species is indirectly produced from a dicarbonyl precursor [Co(CO)₂]⁺ + 2Co²⁺ + O₂ → [Co(O)₂]⁺ + 2[Co(CO)]²⁺. The positioning of the Al atoms at the interface between the straight and sinusoidal channels of the MFI framework is essential for the successful implementation of this process. In the [η²-O₂²⁻Co_4c]⁺/ZSM-5 adduct with S = 1, the Co³⁺ core is chelated by the peroxo ligand, and supported by two weak donor anionic lattice oxygen atoms, forming a square planar structure (similar to that observed for the superoxide adduct with the intrazeolite Ni²⁺) with the ground state electronic configuration of (π)²(δ)²(d_z²)²(d_x²−y²)²(d_zy)¹(δ*)¹(π*)⁰. This structure of the [η²-O₂²⁻Co_4c]⁺/ZSM-5 adduct is ascertained by a distinct NIR absorption band around 9840 cm⁻¹, assigned to the δ → δ* transition in the [Co(O₂)]⁺ moiety (Fig. 29c). The interval between the first and second vibrational fine structure features equal to 560 cm⁻¹ for the ¹⁶O₂, and the isotopic shift of Δ¹⁸O ≈ 20 cm⁻¹ are typical for the symmetric vibrational frequency of the peroxo side-on species (O₂²⁻).

The coordination of O₂ to Cu⁺ (d¹⁰) cations hosted in zeolites of different topologies (AFI, CHA, TON, MOR, MFI) leads to the formation of stable magnetic, η¹, and non-magnetic, η², adducts with the side-on attachment being energetically more favourable by ∼0.5 eV.²⁸⁵ The η² adsorption energies are around −1.3 ± 0.1 eV for CHA(8MR), TON(10MR), MFI(10MR), AFI(12MR), dropping to −1.03 eV for MOR(12MR), indicating that the size of the zeolite cavities plays a secondary role in controlling the strength of the Cu–O₂ binding. The O–O stretching frequencies varying about 1172–1198 cm⁻¹ are characteristic of superoxide species (Cu^I/z + O₂ → [η²-O₂⁻–Cu^II]⁺/z). The thermodynamic stability diagram of η¹-O₂ and η²-O₂ adducts of the Cu²⁺ and Cu⁺ cations hosted in the 6MR and 8MR sites of CHA at various temperatures and p_O2 pressures (Fig. 29d) shows that O₂ is only weakly bound to Cu²⁺ in both 6MR and 8MR positions, desorbing already below 0 °C under the O₂ pressure of 5% (horizontal orange line). This is consistent with a weak activation of dioxygen (d_O–O of 1.235 Å is close to the free O₂ value of 1.21 Å), primarily due to the formation of inverted energy levels (IEL, see Chapter 4.1 and Fig. 17). The stability of the superoxide [η²-O₂⁻–Cu]⁺ species hosted in the 6MR and 8MR extends to ∼200 °C, following the sizable charge transfer stabilization of the ligated superoxide moiety. Interestingly, attachment of the ancillary NH₃ ligands of stronger donor properties makes the mononuclear [O₂–Cu(NH₃)₂]⁺ adducts that are solvated in the zeolite cavities unstable (at the DFT calculation level of BEEF-vdW), regardless of the attachment mode.³⁰⁰ Yet, the [η¹-O₂–Cu(NH₃)₃]⁺ species, proposed to be formed upon NH₃-induced dissociation of the peroxo diamino dicopper(II) complexes, are apparently stable.³⁰¹

Dioxygen attachment to dinuclear metal and metal–oxo centres. The binding of O₂ by two adjacent TMI cations typically leads to the formation of end-on cis μ-η¹:η¹, trans μ-η¹:η¹, the side-on μ-η²:η², and the bis(μ-oxo)²⁸² species of predominantly peroxo and oxo character, mentioned above (Fig. 27). Notable examples are provided by the bridging oxygen species, produced upon O₂ attachment to dimeric Cu⁺∪Cu⁺ centres in MOR,^285,302 CHA,^285,303 BEA,³⁰⁴ MFI^285,305,306 and Fe²⁺∪Fe²⁺ centres in BEA,³⁰⁷ MOR or FER.³⁰⁸ Other less explored species include trimeric Cu₃O₃ centres in MOR³⁰² or Co²⁺∪Ca in FAU(X) zeolites.³⁰⁹ The metal–metal distance and the presence of ancillary ligands (such as common NH₃³¹⁰) play a notable role in controlling the nature of the produced species, as well as the pathways of eventual splitting of the bound O₂ moiety.

When the auto-reduced Cu/ZSM-5 is contacted with O₂ at room temperature, side-on peroxo dicopper species ([Cu(μ-η²:η²-O₂²⁻–Cu]²⁺) are formed.³¹¹ The resonance Raman ν(Cu⁺,Cu⁺) band at 269 cm⁻¹, the isotopically sensitive ν(O–O) vibrations at 736 cm⁻¹ (for ¹⁶O) and 695 cm⁻¹ (¹⁸O), respectively, and the UV-vis peak around 29 000 cm⁻¹ (assigned to ) are their typical diagnostic features (Fig. 31a1 and a2). The [Cu(μ-η²:η²-O₂²⁻–Cu]²⁺ adduct after heating at T > 150 °C transforms into mono-μ-oxo dicopper [Cu(μ-O²⁻)–Cu]²⁺ characterized by the UV-vis band at 22 700 cm⁻¹ and the rR lines at 237 cm⁻¹ (ν_δ, Δ¹⁸O = 3 cm⁻¹), 456 cm⁻¹ (ν_s, Δ¹⁸O = 8 cm⁻¹) and 870 cm⁻¹(ν_as, Δ¹⁸O = 40 cm⁻¹), see Fig. 31a₂ and a₃. Although the splitting of the O–O bond is an uphill process, a relatively small barrier <0.4 eV (for the SSZ-13 zeolite) makes the subsequent transformation of the μ-η²:η² peroxo adduct into the bis(μ)oxo descendant quite feasible. Indeed, this process becomes thermodynamically favourable above 200 °C at log(p_O2/p₀) < 1 (Fig. 31a4). The bis-μ-oxo species are the precursors for mono(μ-oxo)dicopper(II) species,²⁸⁴ which exhibit high activity in C–H bond activation, and are recognised as the active sites for CH₄ or benzene hydroxylation.^304,311,312

Fig. 31 Resonance Raman (a₁) and (a₃) and UV-vis-NIR (a₂) spectra of Cu²⁺–(O–O)Cu²⁺ and Cu²⁺–(O)–Cu²⁺ species hosted in CHA, together with corresponding FPT diagram (a₄).^83,311 Influence of copper distance on the nature of oxygen adducts, together with the associated key molecular orbitals responsible for O–O bonding (b₁) and (b₂).³⁰⁶ Evolution of the dinuclear diamino adducts of O₂ with the corresponding energy profile (c).³¹⁰ Energy profile for dioxygen splitting on Fe/FER catalysts (d), and dependence of the activation energy of O₂ dissociation on the zeolite topology (d₁). Panels d and d₁ are adapted from ref. 308 with permission from the American Chemical Society, Copyright 2021. Panels a–c are based on data form quoted references.

Molecular modelling studies show that the distribution of Al atoms within the cavities/channels and the resultant Cu–Cu distance control to a large extent the nature of the bridging oxygen species. For 2NNAl pairs configuration (d_Cu–Cu = 2.224 Å) a cis μ-η¹:η¹-O₂ binding mode is preferred, in the case of 3NNAl pairs a μ-η²:η²-O₂ (d_Cu–Cu = 3.397 Å) and a trans μ-η¹:η¹O₂ (d_Cu–Cu = 3.887 Å) adducts are formed, whereas at larger distances between the Al atoms (4NN) only a trans μ-η¹:η¹O₂ (d_Cu–Cu = 4.249 Å) can be produced.²⁸² The O–O bond activation gauged by lengthening of the interatomic distance, varies significantly between the adducts.³⁰⁶ In the 2NN configuration, the end-on trans μ-η¹:η¹O₂ complex has the O–O bond of 1.376 Å, whereas for trans μ-η¹:η¹O₂ adducts in 3NN and 4NN environment this bond is shorter (∼1.31 Å). The largest O–O bond length (1.436 Å) is observed for the side-on μ-η²:η²-O₂ peroxo species (Fig. 31b1). The corresponding MO energy diagram of these species, shown in Fig. 31b2, reveals a lucid molecular orbital rationale for different extents of dioxygen activation. The antibonding character of the overlap in the HOMO and HOMO−1 orbitals of the O–O moiety for the side-on 3NN adduct is responsible for lengthening of the O–O bond in comparison to 2NN top-on species with only one antibonding HOMO in the frontier orbitals region.²⁸² In the case of the 4NN end-on adduct, HOMO exhibits in-phase overlap, whereas HOMO−1 exhibits out-of-phase overlap of the oxygen orbitals. The cancellation of both effects accounts for the shorter O–O bond in the trans μ-η¹:η¹O₂ 4NN in comparison to cis μ-η¹:η¹O₂ 2NN species.

The μ-dioxygen adducts are also formed upon reaction of O₂ with [(NH₃)₂Cu]⁺ complexes, where the parent framework O_z ligands are entirely replaced by the NH₃ molecules, which are produced in the reduction half-cycle (RHC) of the SCR reaction.^310,313,314 DFT calculations show that the most probable products include trans μ-η¹:η¹-peroxo and μ-η²:η²-peroxo dicopper(II) adducts with a bidentate attachment of O₂, without complete dissociation of the O–O bond, or a bis-μ-oxo dicopper(III) adduct with the split O–O bond (Fig. 31c). Despite that, the corresponding energy profiles are quite sensitive to the calculation scheme, the μ-η²:η²-peroxo dicopper(II) species with the O–O bond retained is favoured.³¹⁰ It remains in agreement with the experimental UV-vis (a band at 13 850 cm⁻¹) and EXAFS (Cu–Cu distance of ∼3.4 Å) investigations, which show that the side-on μ-η²:η² peroxo dicopper(II) complex predominates after O₂ interaction with a pair of the cage diamino copper(I) complexes hosted in CHA zeolites (Si/Al = 15, Cu/Al = 0.5) at 200 °C.³⁰¹

The zeolite FER, MOR, and *BEA frameworks containing iron that exhibit high concentrations of Al atoms (Si/Al < ∼10) with a large part of the Al pairs of apposite spatial organisation are capable of dioxygen capture and subsequent O–O bond splitting. For this purpose, the two cationic sites that form the distant binuclear Fe²⁺ centres in six rings must face each other in parallel and axial configuration, and the Fe⋯Fe distance should be within a narrow range, from ∼7 to ∼8 Å.³⁰⁸ However, the experimental confirmation of the involved face-to-face configuration of Fe–O species has not been definitive so far. Such active centres are capable of splitting dioxygen to produce, upon dissociation, a pair of ferryl species. The reaction is driven by the cooperative four-electron 2Fe²⁺ → 2Fe⁴⁺ redox cycle, and its primary molecular steps, along with the corresponding energetic profiles, are shown (Fig. 31d1), taking the Fe/FER system as an example. The reaction pathway involves a top-on capture of O₂, followed by subsequent transformation of the η¹-O₂ into more stable η²-O₂ species. The trans μ-η¹:η¹ O₂ transition state, located at −0.4 kcal mol⁻¹ to separate reactants, is achieved by returning to the monodentate η¹-O₂ binding. The dissociation of the dioxygen moiety is energetically slightly favourable by −2.4 kcal mol⁻¹ for the most stable side-on adduct. The activation energy of the O–O bond scission depends on the Fe⋯Fe distance for the iron cations accommodated in the β sites with the Al pairs placed in the para-position, preferably in FER, MOR, and BEA zeolites, with a distinct exception of LTA (where both rings are not parallel), see Fig. 31d2 and a suitable ref. 308.

4.3.2. Coordination and activation of NO. Although nitric oxide (NO) is thermodynamically prone to both decomposition and disproportionation, these processes do not proceed significantly without a catalyst. This is due to the considerable kinetic stability of the NO molecule, which arises primarily from a spin barrier. The latter, together with constraints imposed by orbital symmetry, present the most critical molecular obstructions that prevent efficient activation of NO along a straightforward route. Activation through the coordination of the NO molecule to the intrazeolite TMI centres imparts unique catalytic properties to the metal core and the nitrosyl ligand.

The properties and distinct chemical reactivity of the coordinated NO are often explained in terms of electron density transfer (ET), resulting in the withdrawal of a single electron from the 2π* orbital to produce a nitrosonium (electrophilic) NO^δ+ species, or the addition of an electron to the 2π* orbital, forming a nitroside NO^δ− species of nucleophilic character.^14,315 Yet, using the Enemark–Feltham notation of the TMI nitrosyl adducts, ^2S+1{TMI–NO}ⁿ (see Chapter 4.3), the direct attribution of a TMI–NO to nitrosonium (NO^δ+) or nitroside (NO^δ−) category can be circumvented. The other pathway to NO activation involves an oxygen transfer (OT), resulting in the oxidation of NO to NO₂. The occurrence of the elementary ET and OT events depends critically on the nature of the active site. The electron transfer is characteristic of the bare cations,^17,247 while oxygen transfer occurs in the case of the various mono or polynuclear oxo-cations.³¹⁶

The epitomic NO⁺ and NO⁻ species, which are expected to bind in linear and bent η¹-N mode, respectively. But typically observed adducts with the partial charge accumulated on the NO ligand (NO^δ−), exemplified by ({²CuNO}¹¹, ²{ZnNO}¹¹), are slightly bent with elongated and polarized N–O bonds. The NO^δ+ containing adducts, such as (²{NiNO}⁹, {³CoNO}⁸), are also bent with the bonds shortened and polarized.^317,318 The NO₂⁻/NO₃⁻ oxidation products, which play an essential role in some catalytic processes (e.g., SCR of NO, oxidation of NH₃, selective oxidation reactions), are produced along mechanistically more intricate pathways involving the interaction of the activated (bound) NO with O₂ or via involvement of elusive HONO species produced upon the interaction of NO with the intrazeolite TMI–OH centres.¹⁷²

The appearance of nitrosonium and nitroside, or other NO_x species can be deduced from the corresponding IR spectra, featured by characteristic position (blue- or red-shifted compared to the bare NO band) and the number of N–O bands.³¹⁹ For zeolitic systems, such simple discrimination of the nitrosonium vs. nitroside reactivity of NO^δ with δ close to +1 or −1, has been rarely observed, and situations with polarisation of the TMI–NO bond rather than formation of ionic adspecies via electron transfer are more common.^{247,317,320,321}

Role of spin and metal electron configuration. The behaviour of the spin density of NO upon binding to TMI can be followed with EPR measurements of in situ NO adsorption. This procedure results in characteristic powder EPR spectra (Fig. 32a–e) attributed to the paramagnetic nitrosyl adducts. Depending on the investigated system (the number of unpaired 3d-electrons of the TMI core), the following processes can be distinguished: (1) spin pairing, (2) spin crossing, and (3) spin addition. The simple spin pairing was observed for VO₂ (3d¹) species (being indirectly deduced from disappearance of the EPR signal of VO₂),²⁹⁵ high-spin ⁴Co²⁺ (3d⁷) centres (leading to the formation of {²Co(NO)₂}⁹ dinitrosyls upon pairing of two odd 3d electrons with two NO molecules),¹⁷ and high-spin ³Ni²⁺ (3d⁸) cations (formation of the mononitrosyls {²NiNO}⁹ adducts).²⁴⁷ The reaction of ⁵Fe²⁺ (3d⁶) with NO in ZSM-5 zeolite represents an example of spin pairing combined with spin crossing.^322–324 In contrast, the adsorption of NO onto reduced Cu⁺ (3d¹⁰) represents a simple spin addition process to the closed shell Cu⁺ cation.³²⁰ In all five cases, the formation of nitrosyl complex can be rationalized in terms of the spin processes occurring between the spin state of 3d TMI and that of NO (²Π_1/2) radical ligand:

Fig. 32 Reactivity of TMI in zeolites toward NO monitored with the CW-EPR spectroscopy. NO adsorption on VO₂/SiBEA (a), Co²⁺/ZSM-5 (b), Ni²⁺/ZSM-5 (c), reduced Fe²⁺ZSM-5 (d), reduced Cu⁺/ZSM-5 (e) metallozeolites. The arrows schematically indicate the number of unpaired electrons involved in the TMI active centre and the resulting nitrosyl adducts. The dotted grey EPR spectra correspond to iron and copper in their parent oxidation states before autoreduction. Figure was based on data from ref. 247, 295, 316 and 322.

(1) ²VO₂/SiBEA + NO_(g) (²Π_1/2) → ¹{ON-VO₂}²/SiBEA;

(2) ⁴Co²⁺/ZSM-5(BEA) + 2NO_(g) (²Π_1/2) → ²{Co(NO)₂}⁹/ZSM-5(BEA);

(3) ³Ni²⁺/ZSM-5(BEA) + NO_(g) (²Π_1/2) → ²{NiNO}⁹/ZSM-5(BEA);

(4) ⁵Fe²⁺/ZSM-5 + NO_(g) (²Π_1/2) → ³{FeNO}⁷/ZSM-5;

(5) ¹Cu⁺/ZSM-5 + NO_(g) (²Π_1/2) → ²{CuNO}¹¹/ZSM-5.

In the case of the ⁴Co²⁺ adducts with NO, analysis of the observed cobalt hyperfine splitting indicates that approximately 80% of the spin density (ρ^3d ≈ 80%) is localized on the metal centre. This suggests that the magnetic structure of the dinitrosyl intermediates can be represented as NO(↑↓)–↑Co²⁺–(↓↑)NO, where the unpaired electron resides primarily on the cobalt ion. This adduct, however, undergoes interesting transformation upon thermal treatment, resulting in the reduction of the metal centre to zero-valent cobalt. The latter state was unambiguously confirmed by consecutive adsorption of CO and propene, resulting in EPR spectra characteristic of 3d⁹ electron configuration of the cobalt cation.¹⁷

Similarly, the g-tensor values of Ni²⁺NO adduct in ZSM-5 zeolite are characteristic of a metal-centred paramagnet with the 3d⁹ electron configuration.³ This signal has been assigned to the adducts of the type [Ni–NO]²⁺/ZSM-5 (S = 1/2, 3d⁹). A similar signal was reported earlier for surface adducts in the Ni–silica systems.³²⁵ The [Ni–NO]²⁺ notation does not reflect the actual redistribution of partial charges that occurs upon NO coordination to Ni²⁺, and the corresponding shift of the ν_N=O IR band (1890–1898 cm⁻¹, Fig. 33a) concerning the 1874 cm⁻¹ value of the physisorbed NO. The actual charge redistribution is closer to the Ni^(2−δ)+–NO^δ+ formulation. The relative position of the donor level regarding the semi-occupied Ni 3d_z² (σ donation channel) and the position of the occupied 3d_xz orbital in comparison to the empty level (π back donation channel), dictate the extent of the actual ligand to metal charge transfer, where the coordinated NO molecule is partially oxidized to NO^δ+. This is consistent with the experimentally observed blue shift in the NO stretching frequency in the IR spectrum.²⁴⁷

Fig. 33 Spectroscopic evidence of the formation of nitrosyl upon NO interaction with Ni²⁺/ZSM-5. The IR spectra of NO adsorption at various temperatures (a), spin density contour of the [NiNO]²⁺ adduct in ZSM-5 (b), CW-EPR spectra (77 K) of NO adsorption on Ni and isotopically enriched ⁶¹Ni (I = 3/2) hosted in ZSM-5 zeolite (c), electron spin echo detected EPR signal (10 K), the arrows indicate observer positions (B₁, B₂, B₃) at which HYSCORE signals can be measured (d₁). HYSCORE spectrum due to the ¹⁴N hyperfine interaction recorded at B₂ position with the matched pulse sequence (d₂). The figure was based on data from ref. 156 and 247.

The EPR results also support such a molecular picture. The g-tensor parameters extracted from the CW-EPR spectrum of {²NiNO}⁹ adducts in ZSM-5 zeolite (Fig. 33c) are characteristic of a 3d⁹ electron configuration, which is supported by the presence of ⁶¹Ni (I = 3/2) hyperfine splitting when an isotopically-enriched sample was used. At the same time, hyperfine splitting due to ¹⁴N (I = 1) is not resolved due to an effective quenching of the spin density at the nitrogen nucleus resulting from the spin pairing ligation. The residual hyperfine structure can be unraveled using HYSCORE spectroscopy, a powerful method for characterizing paramagnetic species in porous materials.^{95,116,117,156} Electron spin echo (Fig. 33d₁) and the HYSCORE spectrum for one selected orientation, corresponding to the maximum amplitude of the echo, is shown in Fig. 33d2 for the {²NiNO}⁹/ZSM-5 adduct. The characteristic correlation ridges in the so-called strong coupling regime correspond to the following ¹⁴N hyperfine parameters A_xx = 14.0 MHz, A_yy = 9.0 MHz, A_zz = 6.0 MHz, implying that approximately 4% of the total spin density is localised on nitrogen.¹⁵⁶

The above observations indicate that nitrosyl adducts are formed via a spin-pairing process, which in the case of Ni²⁺/ZM-5 involves a one-electron Ni²⁺(3d⁸)/Ni⁺(3d⁹) couple.²⁴⁷ For the Co²⁺/ZSM-5 (or BEA) system, adsorption of NO leads to the spontaneous formation of dinitrosyls, engaging a two-electron Co²⁺(3d⁷)/Co⁰(3d⁹) couple.³¹⁶

The IR and EPR evidence for the formation of paramagnetic ²{CuNO}¹¹ is shown in Fig. 34a and b. The bonding between the closed-shell Cu⁺ and NO radical in the channels of ZSM-5 zeolite has been rationalized in terms of the spin addition mechanism and the s/d hybridization of the empty 4s and two occupied 3d_σ orbitals (3d_z² and 3d_xz) of Cu, which interact with the 2π* MO of NO.^{320,321,326,327} The empty 3d_z² + 4s acceptor state can overlap more efficiently with the in-plane of NO. At the same time, the interaction of the NO lone pair is weaker as the latter is situated relatively low in the energy within the 3d manifold (see the MO interaction diagram in Fig. 34c). Such an interaction provides a pathway for the σ-type bond, producing the singly occupied molecular orbital (SOMO). The interaction of the occupied 3d_yz orbital with the remaining out-of-plane empty gives rise to the lowest-unoccupied molecular orbital (LUMO). The local C_s symmetry allows for admixing of the Cu 4s state to the 3d manifold. The interaction with the 2π* states of NO results in a charge and spin donation , accompanied by a back-donation. SOMO and LUMO are antibonding between Cu⁺ and NO, exhibiting a mixed 2π*-3d character with a prevailing NO contribution. The redistribution of electronic charge within the Cu–NO unit leads to a partial oxidation of the copper centre and a concurrent reduction of the NO ligand, as evidenced by the red-shift of the N=O stretching vibration to 1805 cm⁻¹. In addition, such an orbital picture explains all characteristic features of the EPR spectrum of {²CuNO}¹¹ adducts (Fig. 34b), i.e., monoclinic symmetry, large hyperfine coupling constants (^63,65Cu, I = 3/2) due to involvement of Cu 4s in SOMO, and the ¹⁴N (I = 1) hyperfine splitting that is typical of ligand-centred nitrosyls radical.³²⁰

Fig. 34 Spectroscopic evidence of the nitrosyl adduct formation upon NO interaction with Cu⁺/ZSM-5. IR spectra of NO adsorption with the highlighted mononitrosyl band at 1805 cm⁻¹ (a), CW-EPR spectra of the parent Cu²⁺/ZSM-5 zeolite and the adsorption complex, produced after thermal autoreduction of Cu²⁺/ZSM-5 and subsequent NO adsorption (b), Kohn–Sham correlation diagram of the {²CuNO}¹¹/ZSM-5 adduct (c), and the corresponding spin density contours of the mononitosyl species (d). Panels a and d adapted from ref. 345 with permission from Elsevier, Copyright 2007. Panel b adapted from ref. 172 under the terms of the CC BY 4.0 license, Copyright 2025, published by American Chemical Society. Panel c adapted from ref. 320 with permission from he American Chemical Society, Copyright 2003.

Beyond the previously discussed interaction, the primary factor driving the bending of the NO ligand is the enhanced orbital overlap between the orbital of NO and the 4s orbital of the metal centre. This interaction becomes symmetry-allowed only in a bent geometry, making it inaccessible in the linear coordination mode. The spin density contour shows principal localization on the NO ligand with a sizable Cu contribution, and it is practically confined to the CuNO moiety (Fig. 34d). Thus, the copper nitrosyl in ZSM-5 is a ligand-centred {²CuNO}¹¹ radical for spin classification purposes.

Charge flow during the TMI–NO bond formation and bonding mechanisms. The electronic density flows accompanying the TMI–NO bond formation can be elucidated and quantified using the NOCV method (see Chapter 4.1.1). Since NO is an open-shell ligand, the NOCV analysis should be performed in a spin-resolved (SR) fashion, i.e., separately for the α (↑) and β (↓) spin densities. Moreover, the unpaired electrons on the NO and TMI–zeolite non-interacting fragments should be distributed and coupled consistently with the TMI–NO bond; this is necessary to avoid spurious intra-fragment charge flows that would blur the interpretation of the NOCV results.³²⁸

The SR-NOCV method has been applied to a series of increasingly complicated NO interactions with zeolitic TMI, starting from the simplest case of closed-shell Cu⁺ cations,^329,330 through open-shell 3d⁸ Ni²⁺ (ref. 156) and 3d⁹ Cu²⁺,³³⁰ up to the most complex 3d⁷ Co²⁺ active centres with NH₃ co-ligands^40,331 hosted in ZSM-5 zeolite. The SR-NOCV analysis of the NO binding to the Cu⁺/Cu²⁺ sites^328–330 revealed the following most relevant electron-flow channels: (1) π*-backdonation, (2) donation of an unpaired electron from NO π*, (3) σ-donation from the nitrogen lone pair, (4) covalent Cu–NO contribution. They are illustrated in Fig. 35, using the SR-NOCV results for the representative {CuNO}¹¹, {CuNO}¹⁰, and ¹{Co(NH₃)₃(NO)}⁸ species.

Fig. 35 Dominant SR-NOCV components of the differential density during the NO binding to intrazeolite Cu⁺, Cu²⁺ cations and to ¹{Co²⁺(NH₃)₃} complexes in ZSM-5, along with heuristic assignments of the electron flow channels and the corresponding NOCV eigenvalues. Contour values ±0.001 a.u. Adapted from ref. 40 and 330 with permissions from the Royal Chemical Society (Copyright 2016) and from the Canadian Science Publishing (Copyright 2013), respectively.

The channel (1), which represents backdonation from the doubly occupied Cu 3d_⊥ to the empty MO of NO, increases the population of the 2π* (antibonding) orbital, leading to a nitroside activation of the ligand (NO^δ−). The channel (2), which represents the unpaired electron displacement from the initially singly occupied to the Cu–zeolite fragment, acts oppositely. The relative importance of these two channels is sensitive to the oxidation state of Cu and the embedding of zeolite. The 2π*-backdonation channel (1) is most efficient for the zeolitic Cu⁺ site, consistently with a pronounced red-shift of the NO stretching vibration for the intrazeolite {CuNO}¹¹ systems. The π*-backdonation is considerably diminished in the zeolitic Cu²⁺ site (cf. Fig. 34). In this case, this channel features strong polarization on NO and involves the electron density transfer from Cu 3d_⊥ and nitric oxide O 2p_⊥ to the bonding region, rather than straightforward π*-backdonation.³³⁰ Moreover, in the Cu²⁺NO system, the effect of π*-backdonation is effectively cancelled by the opposite effect of the unpaired electron donation from the 2π* SOMO along the channel (2). The net outcome of this competition is the deactivation of NO, which is also reflected by the blue shift of the ν_N–O stretching frequency.^329,330 These results explain why the intrazeolite Cu²⁺ sites are inefficient in activating NO, although it binds this molecule more strongly than the intrazolite Cu⁺ centers.^330,332

The remaining two electron density flow channels, shown in Fig. 35, are of minor importance for the NO activation and are relatively insensitive to the metal oxidation state and embedding.³²⁸ The σ-donation channel (3) appears in all studied models, but hardly influences the strength of the NO bond, as the electron density is donated mainly from the nitrogen lone pair. The covalent Cu–NO channel (4) can be heuristically interpreted as a donation from the NO 2π_‖ and Cu 3d_‖ orbitals to the bonding region, accompanied by simultaneous intra-NO density redistribution. It may also be understood as the formation of a three-electron bond , followed by some mixing in NO. As this channel only exists in the β spin manifold, it contributes to the partial spin transfer from NO to Cu. Spin transfer is also connected with the channel (2), which only exists in the α spin manifold, and channel (1), which exists in both spin manifolds, but its α- and β-spin components are not balanced (compare eigenvalues of 0.37 and 0.25 for the{CuNO}¹¹ adduct in Fig. 35).

In summary, the discussed SR-NOCV results nicely corroborate the spin addition and backdonation mechanisms discussed above for the Cu–NO system, and the interpretation of the EPR experimental data. By showing competition between the two most relevant electron flow channels: (1) π*-backdonation and (2) donation of the 2π* unpaired electron, they rationalize the propensities of different Cu sites to activate NO. Such an insight can hardly be obtained from more popular descriptors, such as atomic charges and spin populations, alone. In particular, the spin density on NO is not a direct indicator of its effective redox state, a conclusion that has also been pointed out in a broader context for other TMI–nitrosyl complexes.⁵⁰ The fractional charge of the ligated NO is a net result of the opposing electronic flows corresponding to donation and backdonation processes.

DFT studies, supplemented with SR-NOCV analyses, have also been used to elucidate the mechanism of NO bonding to zeolitic Co²⁺ sites in the presence of ammonia co-ligands.^40,331 The experimental studies of mononitrosyl species in Co²⁺-exchanged MOR and FER zeolites demonstrate that the NO stretching frequency undergoes a blue shift, which changes to a considerable red shift upon NH₃ co-adsorption.³³¹ Following the experimental conditions, DFT calculations were conducted on simplified cluster models of the {CoNO}⁸ species containing none, two, three, or five NH₃ co-ligands, which are embedded in a simple one-tetrahedron T1.^40,331 The construction of the more realistic zeolitic models is described elsewhere.^333,334

The DFT modelling of the (NH₃)_n{CoNO}⁸ adducts fully confirmed the electron-donating effect of NH₃ lone pairs, leading to enhanced Co 3d → 2π*-backdonation, and explaining the red-shift of the NO stretch in the (NH₃)_n{CoNO}⁸ adducts. Interestingly, this electron flow channel turned out to be finely spin-state controlled, i.e., highly more effective in the singlet than in the triplet spin state.³³¹ The relative energies of the two spin states are challenging to predict using the available approximate DFT methods, which is a typical problem for 3d-electron TMI complexes in various coordination environments.⁵³ Yet, the CCSD(T) and CASPT2 results show that the singlet and triplet spin states of the intrazeolite {CoNO}⁸ species are close in energy, allowing them to coexist in equilibrium under experimental conditions.⁴⁰ The calculations also reveal that the most substantial red shift of the NO stretching occurs for the complex with three NH₃ co-ligands, rather than for the complex with five ligands. While being counterintuitive at first sight (ammonia is a stronger electron donor than zeolitic oxygen groups), this result can be rationalized by invoking the trans effect of the NH₃ ligand, present in the pentammine complexes. It leads to elongation of the Co–NO bond, and hereby to the reduction of the π*-backdonation effect in [Co(NH₃)₅(NO)]²⁺ as compared with the [Co(NH₃)₃(NO)]²⁺ zeolite adduct.³²⁸ A similar pattern of complex competition between the binding of strong NH₃ co-ligands and preserving the covalent linkages to the zeolite framework, which is an electron-rich, but weaker ligand, has been observed for more realistic zeolite models.³³⁴ It is not necessarily full ammonia saturation of the Co²⁺ site that leads to the highest extent of the NO activation.

According to the SR-NOCV analysis,⁴⁰ the Co–NO bond formation in the spin singlet {Co(NH₃)₃(NO)}⁸ zeolitic species can be best characterized as a combination of (1) spin-pairing between Co²⁺(3d_‖, ↑) and NO (, ↓) with concomitant (2) π*-backdonation from Co 3d_⊥ to NO (Fig. 35). The spin pairing channel is consistent with the formation of a covalent σ Co–NO bond (having a significant biradical character – see below). The π*-backdonation channel, which activates the N–O bond, is found in both spin manifolds with the identical eigenvalues, and hence, it does not transmit the spin polarization.

The electronic structure of the discussed ¹[Co(NH₃)₃(NO)]²⁺ zeolite model (T1) has been described as unpolarized (closed-shell) singlet,⁴⁰ caused by the application of a pure density functional (BP86). In contrast, hybrid functionals can easily lead to a spin-polarized BS-DFT description, as seen for the intrazeolite {CuNO}¹⁰ species discussed earlier. In general, both the existence of a BS solution³³⁵ and the distribution of spin density^50,51 in the nitrosyl complexes are critically dependent on the choice of the exchange–correlation functional. The BS spin density, although helpful in visualizing the pattern of spin coupling accompanying TMI–NO bond formation, is not expected to accurately reproduce the physical spin density. It merely provides a means of simulating nondynamic correlation effects in the metal–ligand bond, within the limitations of the single-configurational DFT formalism.^50,336

A more rigorous description of these effects can be obtained using the multiconfigurational wave function formalism, in which the principal electronic configuration (PEC) is mixed with other configurations, obtained by promoting electron pairs from the bonding to antibonding orbitals. In this way, the antibonding orbital (unoccupied in the PEC) gains fractional occupations, whereas the bonding orbital (doubly occupied in the PEC) reduces its effective occupation below two. Thus, the TMI–NO bond acquires a partial biradical character. In a sense, it is “stretched” or “partially dissociated” already in the equilibrium geometry due to non-optimal overlap between the metal and ligand orbitals.^337–339 This effect is most pronounced in first-row TMI due to the small size of their 3d valence orbitals. The ligand cannot approach close enough to achieve the optimum overlap with the 3d orbitals due to Pauli repulsion with the doubly occupied 3s3p outer-core orbitals, which are of approximately the same size as the 3d ones (see Pauli shield in (Fig. 17b)).³³⁸ The NO ligand has doubly occupied 2π orbitals in a very similar space region as 2π* ones, giving rise to powerful Pauli repulsion.³³⁹ These nondynamic correlation effects exacerbate problems associated with the non-innocent nature of NO, as they influence both the strength of the TMI–NO bond and the effective oxidation state of the NO ligand.

The discussed nondynamic correlation effects can be accounted for by using the CAS method with an active space containing the TMI 3d and NO 2π* orbitals as well as other relevant orbitals, e.g., the additional shell of d orbitals and NO π orbitals, which are needed to achieve a balanced approximation of the electronic wave function in nitrosyl species.^40,50,337 The CAS calculations with 13–14 active orbitals have been performed for the intrazeolite ^1,3{CoNO}⁸ adducts with ammonia as co-ligands,⁴⁰ and for the ²{NiNO}⁹ complex in ZSM-5 zeolite.¹⁵⁶ Based on these results, Fig. 36a shows the pairs of bonding and antibonding orbitals that describe the σ component of the TMI–NO bond.

Fig. 36 CASSCF natural orbitals with bonding and antibonding character concerning the TMI–NO σ bond, and their fractional occupation numbers for zeolite cluster models hosting (NH₃)₃{Co(NO)}⁸ and {NiNO}⁸ adducts (a) based on results from. Contour values ±0.04 a.u. Breakdown of the contributions from the TMI–NO resonance structures resulting from VB analysis of the CAS wave functions for the (NH₃)₃{Co(NO)}⁸ (b₁) and {NiNO}⁹ (b₂) adducts into electronic configurations for the individual VB oxidation states. Figure was based on data from ref. 40 and 156 with panel b₁ adapted from ref. 328 under the terms of the CC BY license 4.0, Copyright 2021, published by MDPI.

The contributions of TMI and NO fragment orbitals to the bonding and antibonding orbitals are comparable, demonstrating a highly covalent character of the σ-bonds, which is consistent with the spin pairing ligation mechanism.¹⁵⁶ The orbitals’ fractional occupation numbers strongly deviate from the ideal integer values (i.e., 2 for the bonding orbital, 0 for the antibonding orbital), indicating considerably reduced bond order, and hence a pronounced biradical character of these bonds. The effective bond order, b = (n_b − n_a)/2, amounts to 0.765 for the intrazeolite {CoNO}⁸ complex and to 0.485 only for the {NiNO}⁸/ZSM-5 complex. Thus, the σ-bond in the equilibrium geometry of Ni–NO exhibits a biradical character exceeding 50% (the biradical character is calculated as (1 − b) × 100%).

The spin density of the {NiNO}⁸/ZSM-5 complex primarily originates from the SOMO, which is predominantly (95%) composed of the Ni 3d_xy orbital interacting with lone pairs of the zeolite oxygen atoms. Due to the spin coupling within the Ni–NO bond, the total spin population on Ni increases to +1.22, which is compensated by the negative spin population on NO (−0.17 on N, −0.10 on O) and residual positive spin population on the zeolite oxygens, according to Mulliken spin populations of the CAS wave function.¹⁵⁶ Only a minor part of the spin density is retained on the NO ligand (in agreement with the spin-pairing mechanism of ligation), which is consistent with the interpretation of the HYSCORE experimental results discussed above.

Further insight into the multiconfigurational electronic structure of TMI–NO bonds can be gained by analyzing the valence bond (VB) structure of the CAS wavefunction.^40,50 In this approach (VB-CAS), the active orbitals undergo a localization procedure, after which each active orbital can be assigned to either the TMI- or NO-fragment. The CAS wave function can be reinterpreted as a superposition of the VB resonance structures with a defined number of electrons assigned to the TMI fragment and the NO ligand (i.e., definite oxidation states of these fragments). The VB-CAS method thus allows circumvention of the limitations of the Feltham–Enemark notation (see Chapter 4.3) by quantifying contributions from the resonance structures belonging to different oxidation states. The technique, initially developed for bioinorganic architectures,⁵⁰ has been subsequently applied to describe nitrosyl complexes in diverse coordination environments,^51,340,341 including intrazeolite TMI–NO adducts.^40,156 The VB-CAS method should be clearly distinguished from alternative approaches to assigning oxidation states in nitrosyl complexes, e.g., based on the IUPAC rules (ionic approximation) and/or geometric arguments^342,343 and other types of orbital analyses.³⁴⁴

The contributions of the VB-CAS resonance structures for {Co(NH₃)₃NO}⁸ are illustrated in Fig. 36b₁,⁴⁰ and for the {NiNO}⁹ complex in Fig. 36b₂.¹⁵⁶ The most important individual electronic configurations for a given resonance form are also shown. For both analysed nitrosyl adducts, the CAS wave function is clearly dominated by the resonance form with neutral NO radical, which can be formulated as TMI²⁺–NO˙ (where TMI = Co or Ni). In the most important configurations of this type, the local spin on NO is antiferromagnetically coupled with the TMI spin. For example, in the {NiNO}⁹ adduct, 87.3% of the CAS wave function is provided by the Ni²⁺(S_Ni = 1)NO˙(S_NO = 1/2) resonance structure, with the critical single configuration constituting 78.5% of the CAS wave function.¹⁵⁶ The large contribution of the TMI²⁺–NO˙ resonance form, particularly in the {NiNO}⁹ complex, is thus consistent with the above-discussed considerable biradical character of the TMI–NO bonds, and corroborates lucidly the spin-pairing ligation mechanism. The shares of the essential ionic resonance structures, such as M²⁺–NO⁻ or M⁺–NO⁺, are notably different in the two discussed complexes, indicating partial nucleophilic character of NO^δ− in the case of ¹{CoNO}⁸ and partial electrophilic character of NO^δ+ in ²{NiNO}⁹. This is consistent with the positions of the NO stretching bands, which are either red-shifted or blue-shifted, in these two cases. Moreover, in the series of {CoNO}⁸ adducts,⁴⁰ the relative shares of the Co³⁺–NO⁻ or Co⁺–NO⁺ resonance forms align well with changes of the N–O force constant and the stretching frequency, showing that predictions of the VB-CAS method generally correlate with other descriptors of the NO valence state. Concluding, the VB-CAS shares of the resonance structures appear to be more robust descriptors than the SR-NOCV eigenvalues when comparing complexes with different spin states or geometries.

Reactivity of the TMI–NO adducts with NO. The reactivity of nitrosyl adducts is closely related to how the NO molecule is coordinated, and to its electron configuration and spin state. For instance, the chemoselective capture of NO and the formation of mononitrosyl intermediates are pivotal prerequisites for its direct decomposition to N₂ and O₂ or its reduction into N₂ with the aid of hydrocarbons.^172,247,316 Since the formation of the N–N bond requires the proximity of two NO molecules, there are two possible pathways along which the adsorption of the second NO molecule can occur. These pathways have been distinguished based on the NO attachment locus (Fig. 37a) and can be termed inner-sphere and outer-sphere attacks. This mechanistic dichotomy determines the regioselective addition of NO to the mononitrosyl intermediate. The inner-sphere addition process leads to the formation of geminal dinitrosyl ^2S+1{TMI–(NO)₂}^d+2 adducts, whereas the outer-sphere pathway results in a dimer bound to the metal centre ^2S+1{TMI(N₂O₂)}^d+2.

Fig. 37 The inner- and outer-sphere attack of the second NO molecule on intrazeolite nitrosyl complex (a). Comparison of DFT calculated IR dinitrosyl bands of Cu⁺(NO)₂ species in the attracto and repulso conformations (b), based on data from ref. 345. DFT optimised pathways of the second NO molecule attachment to Cr⁺–NO (c₁) and Zn²⁺–NO (c₂) mononitrosyls, along with the respective spin density repartition that governs the inner- (Cr) or outer-sphere (Zn) direction of the NO attack. DFT-calculated free-energy diagrams at various temperatures for NO decomposition over a mononuclear Cu⁺ (d₁) and vicinal Cu⁺∪Cu⁺ dimer (d₂) centres hosted in ZSM-5, adapted from ref. 350 with permission from Springer Nature, Copyright 2021.

The formation of dinitrosyl complexes in metallozeolite catalysts is confirmed, mostly by IR spectroscopy.³¹⁹ The disappearance of the band attributed to the N–O vibration of mononitrosyls and the simultaneous formation of two bands of the collective symmetric and antisymmetric vibrations of the TMI(NO)₂ species, provides direct evidence for the formation of dinitrosyl complexes at the expense of the parent mononitrosyl adducts.^{319,345–347} Similarly, formation of N₂O₂ dimers, which are diagnostic of the outer-sphere mechanism, has also been observed by IR in the metallozeolites in the spectral range of 1420–1320 cm⁻¹ corresponding to the N–N vibrations.³¹⁹

Dinitrosyl complexes have often been postulated as intermediates of the NO disproportionation into N₂O and NO₂, catalysed by transition metal complexes in homogeneous^348,349 and zeolitic³⁵⁰ systems. The two dinitrosyl conformations, namely attracto and repulso, are shown in Fig. 37b. According to the classification proposed by Richter and Legzdins,³⁵¹ the 3d^5–10 transition metals prefer an attracto conformation. In this geometry, the bending of both NO groups results in a decrease in the oxygen–oxygen distance (“inward” bending). An “outward” bending, in turn, increases the distance between the oxygen atoms of both NO ligands, with a simultaneous reduction of the distance between the nitrogen atoms (repulso conformation). Such structures have been described in detail for copper(I) centres in zeolites,^346,347 and they differ in the respective symmetric and antisymmetric IR bands.³⁴⁵

The interaction pathway of the second NO molecule with a given mononitrosyl complex can be related to the spin repartition within this species. For the first-row paramagnetic TMI–NO/ZSM-5 adducts,³¹⁷ the spin density is confined almost entirely to the TMI–NO fragment, spreading on the zeolite backbone to a small extent only. Thus, the TMI–NO unit constitutes the magnetophore of the entire TMI–NO/ZSM-5 system (see Chapter 4.1).³²¹ Two examples of the spin density repartition in the ³{CrNO}⁶ and ²{ZnNO}¹¹ magnetophores hosted in the ZSM-5 framework are shown in Fig. 37c₁ and c₂. They represent two extreme cases of metal-centered and ligand-centered metallo-radicals, respectively (see Fig. 18e).

The effect of the spin density distribution in TMI–NO on the attachment of the second NO molecule is illustrated by the reaction pathways shown in Fig. 37c₁ and c₂. For the ³{CrNO}⁶ magnetophore, the ligation trajectory along which the interaction energy is minimized results in the formation of a dinitrosyl complex ²{Cr(NO)₂}⁷ via an inner-sphere attack. A different scenario occurs for the ²{ZnNO}¹¹ magnetophore. Here, the binding of NO leads to the bond formation directly between the nitrogen atoms, as the incoming NO molecule is repelled from the first coordination sphere of the zinc centre. Therefore, a ³{ZnN₂O₂}¹² complex is formed along the outer-sphere mechanism. Thus, the outer- vs. inner-sphere dichotomy of NO coordination to the paramagnetic mononitrosyl intermediates is essentially controlled by the spin density distribution within the corresponding magnetophores. The inner-sphere attack, which occurs directly at the metal core, is expected for the metal-centered metallo-radicals. In contrast, for the ligand-centered metallo-radicals, the outer-sphere mechanism is preferred. The bent structure of the NO ligand with the associated accumulation of spin density on the nitrogen atom is particularly beneficial for directing the NO attack at the nitrosyl moiety.

The possible mechanisms of NO activation differ depending on the nuclearity of the active centres, as demonstrated for monomeric³⁴⁵ or dimeric copper species³⁵⁰ in zeolites. In the case of mononuclear copper centres, the spin density localisation on the ligand in the Cu–NO magnetophore suggests that the attack of the second NO on the copper nitrosyl intermediate is preferred via an outer-sphere pathway. It occurs with nearly zero activation energy and leads to the direct formation of the N–N bond via the N₂O₂ intermediate. The latter decomposes into N₂O, leaving behind highly reactive intrazeolite CuO/z species (its electronic structure is briefly discussed in Chapter 2.2). It can be brought to the initial state by interacting with the N₂O byproduct (CuO/z + N₂O → O₂ + N₂ + Cu/z) or by reduction with auxiliary hydrocarbons (HC-SCR). However, the energy profiles of the reaction pathway (Fig. 37d₁) favour the dinitrosyl formation (inner-sphere) rather than the direct NO coupling (outer-sphere).³⁴⁵ Two possible conformers of the dinitrosyl intermediates can be identified using IR spectroscopic signatures (the frequency difference between the symmetric (ν_sym) and antisymmetric (ν_asym) components, and the I_sym/I_asym intensity ratio).³⁴⁵ Seemingly, only the relative intensities of the stretching vibrations are markedly altered upon changing the conformation, and the IR results confirm an attracto conformation of the copper dinitrosyls in ZSM-5 (see Fig. 37b). Although the attracto conformation brings both oxygen atoms closer to each other, the development of an O–O bond (for initiating direct NO decomposition into O₂ and N₂) is significantly impeded by the concomitant accumulation of the negative charge on these atoms when they approach. In the alternative repulso conformation, both nitrogen atoms are getting closer to each other. It is beneficial for the incipient N–N bond, and along this mechanistic variant, the NO decomposition takes place. As a result, for direct NO decomposition into N₂ and O₂, the attracto dinitrosyls are inert, and the repulso conformation favours the reaction initiated by the inner-sphere N–N bond formation. It requires, therefore, an initial transformation of the dinitrosyl intermediates from the attracto into the repulso conformation.^345,346 An energy diagram for this transformation is presented in Fig. 37d₁. As a result, the mechanistic cycle of the NO direct decomposition on the mononuclear centres of Cu accommodated in zeolite catalysts involves the following steps:

¹Cu⁺ + NO_(g) → ²{CuNO}¹¹ → cis-¹{CuN₂O₂}¹² → trans-¹{CuN₂O₂}¹² → ³Cu²⁺O⁻ + N₂O_(g).

It can be accomplished by decomposition of the N₂O byproduct on the ensuing copper–oxo centres (see Chapter 4.3.3).^350,352 Such a mechanism of direct NO decomposition through the dimeric N₂O₂ intermediates is substantiated for CuZSM-5 zeolite by spectroscopic evidence.³¹⁹

Another scenario of NO decomposition occurs over copper dimers and two vicinal single copper centers in ZSM-5 zeolite. The fraction of copper dimers and vicinal Cu centres increases with a decreasing Si/Al ratio and with a rise of the copper site density.^350,353,354 Such centres can interconvert from the vicinal Cu⁺∪Cu⁺ into the dimer (Cu–O–Cu)²⁺ species upon interaction with NO, and the corresponding energy profile is shown in Fig. 37d₂. The NO decomposition mechanism can be summarized in the following steps:

(1) capture of NO, Cu⁺∪Cu⁺ + 2NO → (ON)Cu⁺∪Cu⁺(NO), the ensuing mononitrosyls are observed in IR spectra at 1813 cm⁻¹,

(2) N–N bond formation with release of N₂O byproduct, (ON)[Cu⁺∪Cu⁺](NO) → (Cu–O–Cu)²⁺ + N₂O, formation of (Cu–O–Cu)²⁺ is confirmed by the band at 1903 cm⁻¹ observed upon subsequent NO adsorption (Cu–O–Cu)²⁺ + 2NO → (ON)[(Cu–O–Cu)²⁺](NO),

(3) formation of dinitrogen, Cu⁺∪Cu⁺ + N₂O → (Cu–O–Cu)²⁺ + N₂.

The latter step is discussed in more detail in Chapter 4.3.3.

Reactivity of TMI–NO with O₂ – examples for Cu–zeolites. Recent studies have explored the interaction of NO and O₂ with Cu centres of differing nuclearity, in both reduced and oxidized states, focusing on elucidation of the reactivity patterns and thermodynamic stability of intermediates under conditions that closely resemble those of actual SCO and SCR processes. A typical reaction involving nitrosyls as intermediates is the reduction of NO to dinitrogen in the presence of O₂, using various reductants, such as hydrocarbons (HC-SCR) or ammonia (NH₃-SCR). Since early studies, the redox character of these reactions has been recognized, and Cu-containing zeolites, such as ZSM-5, CHA, have been identified as the most active catalysts.^355–360 Although there are several mechanistic proposals of the SCR reaction course,^{33,277,361,362} their common point is the coexistence of the mutually coupled reduction half-cycle (RHC) and oxidation half-cycle (OHC).^{214,228,300,363,364} The OHC is associated with the selective catalytic oxidation (SCO) of NO to NO₂ and to various charged (nitrates/nitrites) and protonated (HONO) NO_x species.^{33,355–357,364} Spectroscopic IR data support the theoretical and thermodynamic predictions of the formation of Cu²⁺–NO (1950–1880 cm⁻¹) and Cu⁺–NO (1810–1800 cm⁻¹) intermediate species. The bands at 1620, 1595, and 1575 cm⁻¹, in turn, correspond to the symmetric stretching modes of bidentate NO₃⁻, whereas the bands at 1500 and 1310 cm⁻¹ come from the antisymmetric and symmetric modes of monodentate NO₃⁻ species,^214,365,366 overlapping with the antisymmetric and symmetric stretchings of the N-bonded nitrites (NO₂⁻).³¹⁹ They are typically observed during the NO oxidation and SCR reactions in zeolites by operando IR.^278,367,368 The reactivity of NO and O₂ toward NO₂, NO₃⁻/NO₂⁻ formation involves various copper species, such as bare Cu²⁺/Cu⁺ and Cu²⁺–OH⁻, and dimeric [Cu–O–Cu]²⁺, which can act as potential active centres for such reactions.³⁶⁹

A valuable thermodynamic insight into NO oxidation over CuSSZ-13 catalyst can be obtained from FPT modelling. The relative stability of the exemplary copper–NO_x⁻ species, formed in CHA zeolite after the reaction between NO and O₂, at various T and p_NO conditions is shown in Fig. 38, for the Cu²⁺, Cu⁺, and oxo [Cu–O–Cu]²⁺ centers. In general, the oxidized isolated Cu²⁺ centres are unable to capture dioxygen, and the thermodynamic diagrams are dominated by copper(II)–nitrosyls and diverse NO_x adducts (x = 1–3) of a radical nature. Under the conditions similar to that of NO SCO, the stability regions of Cu²⁺NO_x, bare Cu²⁺ and Cu²⁺NO species are shown in Fig. 38a. In the case of reduced Cu⁺ (Fig. 38b), which can be generated, i.a., via reduction of Cu²⁺–OH⁻ with NO producing HONO,^95,370 reaction with NO/O₂ and charge transfer lead to NO₃⁻ dominant anionic species. The complexes formed by the independent capture of NO and O₂ are unstable relative to nitrate and nitrite species. Under the experimental SCR/SCO conditions, nitrates decay at approximately 400 °C. For the dimeric active centres, the dinitrate Cu²⁺–(NO₃⁻)₂–Cu²⁺ adspecies dominate the experimental range of SCR/SCO. The simultaneous adsorption of water and NO is thermodynamically disfavoured, due to the lateral repulsion between H₂O and NO ligands. Notably, under the NO–SCO/SCR conditions, several metastable intermediates with comparable thermodynamic profiles can coexist and mutually interconvert. This dynamic equilibrium enhances the functional versatility of bridging copper centres, enabling them to participate in the stoichiometric transformation of nitrates,³³ promoting fast SCR through efficient NO₂ generation, for instance.^278,357

Fig. 38 Thermodynamic ΔG(p,T) diagrams for Cu²⁺ (a), Cu⁺ (b) cations, and dual copper centres Cu²⁺–O²⁻–Cu²⁺ (c) in CHA zeolite as a function of temperature (T) and partial pressure of NO (p_NO). Thermodynamic profiles for the NO oxidation reaction (NO₂ formation) on the isolated copper centres, according to the cationic metal-redox pathway (d). The calculated NO₂ production profiles based on the thermodynamic modelling for the isolated copper active centres (e₁) and copper dimers (e₂). Adapted from ref. 172 under the terms of the CC BY 4.0 license, Copyright 2025, published by the American Chemical Society.

Recently, thermodynamic profiles have been constructed based on DFT/FPT calculations to unravel the course of NO–SCO elementary steps and underly driving forces under different experimental conditions.¹⁷² Analysis of such profiles for the typical isolated Cu centres in SSZ-13 reveals the presence of three mechanistic routes of NO oxidation: (1) copper-redox pathway, (2) ligand-redox pathway, and (3) via HONO intermediates. In the case of dimeric Cu species, the NO SCO reaction can proceed (1) via direct NO₂ release (thermal decomposition of anionic NO_x⁻ adspecies) and (2) via formation of NO₂ due to the comproportionation of nitrates with the encaged NO. The corresponding concentration profiles of the involved adspecies and intermediates are shown in Fig. 38d and e.The predicted evolution of the NO₂ concentration with reaction temperature, along with the advanced mechanistic tenets, agrees well with the experimental results using isotopically labeled N¹⁸O, N¹⁶O, and ¹⁸O₂.¹⁷²

4.3.3. Dissociative activation of N₂O. The catalytic chemistry of nitrous oxide is driven by technologies for abating anthropogenic N₂O emissions, and the use of N₂O as a unique oxidant that can provide atomic oxygen upon N₂O dissociation, brought about by its interaction with TMI centres.^18,20 Among the numerous efforts to understand the molecular routes of N₂O activation, those focused on metallozeolites with mono and dimeric iron^20,164 and copper as active centres³⁵³ are particularly widespread and have led to notably detailed insights. Other notable examples of 3d TMI include Co,³⁷¹ Ni,³⁷² and, less frequently, Mn.³⁷³
Intrazeolite coordination of N₂O. The N₂O molecule is a relatively weak ligand acting as both a σ donor and a π acceptor.³⁷⁴ It is attached most frequently in the top-on monodentate (η¹-O and η¹-N) way to single TMI cations (Fe, Co, Ni, Cu),³⁷³ and in the cis μ-1,3-O,N, μ-1,1-O fashion on bridging M–O–M species.^353,375 Rather unique trans μ-1,3-O,N adducts attached to adjacent close-distant Fe cations are proposed in Fe/FER, based on DFT modelling.⁹⁰ In the case of (N₂O)₂TMI/Z adducts, η¹-O,O, η¹-O,N and η¹-N,N binding forms, as well as the bidentate η²-O,N attachment of N₂O to single TMI have been reported,³⁷⁶ see (Fig. 39a). Since the bis-N₂O adducts are observed only at high p_N2O pressures and low temperatures, they are not relevant for catalytic considerations.

Fig. 39 Survey of the coordination modes of N₂O to single and dual TMI sites (a), based on data from ref. 20, 353, 371 and 376, with the middle structure adapted from ref. 90 with permission from Elsevier, Copyright 2010. Variable pressure (p_N2O = 5–250 Pa) IR spectra of N₂O adsorbed at room temperature on Fe, Na-CHA (with Fe predominantly located in 6MR) (b₁), and Fe-CHA (with Fe predominantly located in 8MR) zeolites (b₂), and variable temperature adsorption of N₂O on the Fe/SSZ-13 catalyst followed by IR (b₃) adapted from ref. 377 under the terms of the CC BY 4.0 license, Copyright 2024, published by the American Chemical Society. Calculated adsorption energies for η¹-O and η¹-N binding of N₂O to various TMI hosted in the ZSM-5 zeolite (c) together with the corresponding experimental TPD-N₂O profiles (d).³⁷³

The variable pressure IR spectra of N₂O adsorption recorded at room temperature, shown in Fig. 39b1, indicate that for the Fe/6MR species hosted in the CHA matrix, the IR peak of the η¹-N mode dominates, whereas in the case of Fe/8MR the η¹-O (2229 cm⁻¹) and η¹-N (2276 cm⁻¹) adsorption modes exhibit IR signals of comparable intensities (Fig. 39b2).³⁷⁷ This illustrates directly the influence of the zeolite ring size on the N₂O ligation by iron. The 2248 cm⁻¹ band corresponds to Al⋯O–NN species. In the case of Fe/FER zeolites, the η¹-N binding is generally stronger than the η¹-O one, and significantly depends on the α or β locus, and the particular arrangement of Al atoms in the corresponding rings.⁹⁰ The temperature dependence of N₂O adsorption on Fe/SSZ-13 zeolite indicates that the beneficial η¹-O active mode persists till 530 K, whereas the spectator η¹-N one is completely depleted at this temperature (Fig. 39b3). The survey of the DFT-calculated N₂O adsorption energies for selected 3d TMI hosted in ZSM-5 catalysts, and the related N₂O-TPD profiles, are shown in Fig. 39c and d, respectively. The thermal stability of the N₂O-TMI adducts does not exceed 150 °C (520 K), and the position of the desorption peaks maxima (T_max) is linearly correlated with the calculated ΔH_ads values. The extent of metal-to-ligand charge transfer is the key factor controlling the adsorption strength.³⁷³ However, alternatively to desorption, the ligated N₂O may react with the TMI centres, dissociating into N₂ and leaving the bound oxygen moiety behind (as discussed below).

General mechanistic considerations of N₂O activation. A characteristic feature of triatomic molecules like N₂O is that bending due to the ν₂ vibration (589 cm⁻¹ in the gas phase) or upon interaction with the active sites (TMI) leads to its intrinsic preactivation. This is revealed by a significant buildup of negative charge in the oxygen moiety (q_O = −0.2|e|), and the merging of charge at both nitrogen atoms to the same value of q_N = 0.1|e|. An elongation of the NN–O bond (by ∼0.3 Å) and a dramatic decrease in the NN–O bond order accompany such changes, making the bent N₂O molecule prone to oxygen release (Fig. 40a1). All these features play an important role in the N₂O activation process (Fig. 40a₂). Furthermore, since the ground state of N₂O(¹Σ⁺) is singlet, whereas the ground state of O₂(³Σ_g⁻) is triplet, the decomposition of N₂O into N₂ and O₂ follows a nonadiabatic reaction coordinate with changing spin multiplicity.

Fig. 40 Variation of the charge redistribution and the N–O bond order for the N₂O molecule upon bending (a₁), and overlay of the potential energy surfaces of N₂O (grey iso-energy lines) and N₂O⁻ (colour coded areas) with the schematic reaction pathway leading from linear N₂O molecule, through bent transition state at Frank–Condon (FC) locus into shallow minimum on the N₂O⁻ PES, prompting an eventual dissociation of the N₂O⁻ transient (a₂). Adapted from ref. 379 under the terms of the CC BY 4.0 license, Copyright 2024, published by the American Chemical Society. Evolution of the EPR spectra recorded during the reaction of N₂O with VO₂ centres entrapped in BEA zeolite (b₁), together with simulation of the EPR signal due to [VO₂⁺]–O˙⁻ species (b₂) (adapted from ref. 295 with permission from the Royal Society of Chemistry, Copyright 2016) together with the corresponding plot of gradual transformation of [VO₂]–N₂O into [VO₂⁺]–O˙⁻ via an electron transfer mechanism (c).

The two generic mechanisms of N₂O activation are based on electron transfer (ET) and oxygen atom transfer (OT) scenarios.³¹⁷ However, in the case of metallozeolite catalysts, any of those occur in the exclusive form. Most N₂O dissociation reactions on metallozeolites are initiated by an electron density transfer from the TMI core to the N₂O ligand (cationic redox pathway). The extent of the charge transfer, which assists in weakening the N–O bond and promoting transfer of the O-moiety to the TMI core, depends on its nature, location, and the type of zeolite framework. Alternatively, the O-species can also be accommodated on the μ-O²⁻ fragment of the bridging M–O–M centres, producing a peroxo (O₂²⁻(¹Σ_g⁺)) intermediate (defining anionic redox pathway).

Catalytic decomposition of nitrous oxide based on electron transfer (ET) can be formulated in terms of the classic three-step mechanism:

(1) N₂O(¹Σ⁺) + e⁻ → [N₂O⁻]* → O⁻(²P) + N₂(¹Σ_g⁺),

(2) O⁻(²P) + O⁻(²P) → O₂(³Σ_g⁻) + 2e⁻,

(3) O⁻(²P) + N₂O(¹Σ⁺) → O₂(³Σ_g⁻) + N₂(¹Σ_g⁺) + e⁻,

where the requisite electron provided by the donor 3dⁿ orbitals of the M^p+/z active site is accommodated on the antibonding 3π* LUMO of N₂O. This mechanism involves the formation of a highly unstable N₂O⁻ transient, which imparts the necessary geometric and electronic rearrangements of the N₂O molecule, contributing to the inherent barriers of this process. To initiate the reaction, the intrazeolite TMI (i) provides orbitals of appropriate energy and symmetry for N₂O capture and subsequent accommodation of the resulting O⁻ intermediate; (ii) acts as a redox centre (allowing for electron shuttling); and (iii) as spin catalyst (facilitating nonadiabatic conversion of singlet N₂O(¹Σ⁺) reactant into triplet O₂(³Σ_g⁻) product). The fact that the neutral N₂O is linear (C_∞v), whereas the N₂O⁻ transient is bent (C_s), implies a need for significant internal reorganisation energy of the N₂O molecule to make ET feasible, according to the Franck–Condon principle (see the FC locus in Fig. 40a2).³⁷⁸ Notably, upon bending, the adiabatic electron affinity of N₂O increases to 0.22 eV, favouring electron acceptance.³⁷⁹ Since the potential energy surface of N₂O⁻ is relatively flat, this transient species immediately dissociates into N₂ and O⁻ moieties, because the FC locus is situated higher in energy than the N₂O⁻ dissociation barrier (Fig. 40a2). As a result, the NN–O bond breaking, which ET brings about, implies considerable bending of the N₂O molecule, which is a distinctive mechanistic feature of the dissociation process.

These considerations allow for the formulation of N₂O decomposition along the cationic redox mechanism for single M^p+/z centres in the following way:

(1) N₂O(¹Σ⁺) + M^p+/z → [OM]^p+/z + N₂(¹Σ_g⁺),

(2) N₂O(¹Σ⁺) + [OM]^p+/z → O₂(³Σ_g⁻) + N₂(¹Σ_g⁺) + M^p+/z,

(3) 2[OM]^p+/z → O₂(³Σ_g⁻) + 2M^p+/z.

Since the specific charge transfer between the metal centres M and the ligated O-moieties depends on their actual nature, in these general equations only the overall charge “p+” of the involved species is explicitly indicated. For instance, the cationic redox mechanism may involve a single electron transfer producing radical oxyl species O˙⁻/M, as in the case of ²VO₂/BEA, or consecutive two-electron transfer forming oxo-species (α-O), exemplified by ⁵Fe^II/BEA.²⁰

The attack of the second N₂O molecule can be directed at the terminal oxygen or the metal centre, depending on the chemical characteristics of the accommodation sites imparted by their electronic structure and location in the zeolite matrix.³⁸⁰ In the case of dimeric M–O–M species, the cation M centres or the bridging μ-O²⁻ centres may serve as a possible locus for lodging the O-moieties, forming a peroxo μ-O₂²⁻ species with a diagnostic band at near 867 cm⁻¹, typically observed by Raman spectroscopy.³⁸¹

Using the N₂O decomposition on Fe/ZSM-5³⁸⁰ as a straight example, the resultant mechanisms can be epitomised as:

N₂O(¹Σ⁺) + [M–μ-O²⁻–M]^p+/z → [O–M–O²⁻–M]^p+/z + N₂(¹Σ_g⁺),

[O–M–O²⁻–M]^p+/z → [M––(μ-O²⁻)₂–M]^p+/z,

N₂O(¹Σ⁺) + [M––(μ-O²⁻)₂–M]^p+/z → [O–M––(μ-O²⁻)₂–M]^p+/z + N₂(¹Σ_g⁺),

[O–M–(μ-O²⁻)₂–M]^p+/z → [(O₂²⁻)M–μ-O²⁻–M]^p+/z → [M–μ-O²⁻–M]^p+/z + O₂(³Σ_g⁻).

Actually, by involving both the metal and the bridging oxygen, it exhibits a mixed redox (cationic and anionic) character. The less common anionic redox alternative (attack on μ-O²⁻) assumes the following sequence:

N₂O(¹Σ⁺) + [M–(μ-O²⁻)–M]^p+/z → [M–(μ-O₂²⁻)–M]^p+/z + N₂(¹Σ_g⁺),

N₂O(¹Σ⁺) + [M–(μ-O₂²⁻)–M]^p+/z → [M–(μ-O²⁻)–M]^p+/z + O₂(³Σ_g⁻) + N₂(¹Σ_g⁺),

suggested, e.g., for the Cu–O–Cu in ZSM-5,³⁵⁰ with possible O₂ evolution via the following step:

2[M–(μ-O₂²⁻)–M]^p+/z → 2[M–(μ-O²⁻)–M]^p+/z + O₂(³Σ_g⁻).

Herein, the formal oxidation state of M remains essentially intact, whereas for the bridging O²⁻ moiety it changes from −2 to −1 due to the formation of μ-O₂²⁻ (peroxo) species. Additionally, depending on the N₂O coordination energy to the particular intrazeolite TMI and the reaction conditions, adsorption quasi-equilibria such as N₂O + M^p+/z ↔ (N₂O–M)^p+/z, N₂O + [OM]^p+/z ↔ [OM–N₂O]^p+/z and O₂ + M^p+/z ↔ (O₂–M)^p+/z have also been taken into account in the microkinetic analysis.^{20,371,382,383}

Exemplary N₂O activation pathways on well-defined centres. Investigations documenting N₂O activation on metallozeolites induced by a distinct metal-to-ligand electron transfer (MLET) are hardly available in the literature. Nevertheless, a notable example is provided by the VO₂(d¹) species encaged in the BEA zeolites interacting with N₂O.²⁹⁵ The progress of the reaction ²V⁴⁺(=O)₂ + N₂O(¹Σ⁺) → ²[(O=)₂V⁵⁺–O˙⁻] + N₂(¹Σ_g⁺) at room temperature followed by EPR is shown in Fig. 40b1, whereas the EPR spectrum of the final ²[(O=)₂V⁵⁺–O˙⁻] product and its simulation are shown in Fig. 40b2. The obtained kinetic profiles of the reactant decay and the product formation (Fig. 40c) are entirely consistent with the MLET-induced dissociation of N₂O. A concerted capture of the resulting radical O˙⁻ moiety by the [(O=)₂V⁵⁺]⁺ core produces the ²[(O=)₂V⁵⁺–O˙⁻] species. The EPR parameters (g_xx = 2.0202, g_yy = 2.0173, g_zz = 2.0284 > g_e and A_zz = 1.65 mT, A_yy = 1.58 A_zz = 1.49 mT) are characteristic of the ligand-centred O˙⁻ radical (manifested by the g-tensor features) attached electrostatically to the (O=)₂V⁵⁺ core (demonstrated by small superhyperfine values due to the coupling with ⁵¹V, I = 7/2). This signal disappears (after 30 minutes) upon co-adsorption of CO or propene at ambient temperature, due to the highly reactive character of the O˙⁻ radicals.

The mechanism of N₂O activation on cationic centres has been meticulously elaborated by DFT modelling, corroborated to some extent by spectroscopic studies, for the most widely investigated single Fe^II/z and [FeO]²⁺/z sites, and dimeric Cu⁺∪Cu⁺, Fe²⁺∪Fe²⁺, Fe–O–Fe/z and Cu–O–Cu/z centres, accommodated in ZSM-5, BEA, CHA or FER zeolites.^{20,164,353,380,384} The Co, Mn or Ni zeolites are less frequently examined.^385,386

In particular, high activity of dimeric metal centres (Cu⁺∪Cu⁺) has been reported for Cu/ZSM-5 zeolites.³⁵³ Among the bridging μ-1,1-O and μ-1,3-O,N, and terminal η¹-O binding modes of the N₂O molecule, only the μ-1,1-O attachment leads to the small activation energy (2 kcal mol⁻¹) that is compatible with the experimental value of 2.5 kcal mol⁻¹. The corresponding energy profile for the singlet and triplet PES indicates that the minimum energy crossing point (MECP) appears after the bent transition state (∠(ONN) = 143°) located on the singlet PES (Fig. 41a₁). In the MECP locus, the O–N₂ distance is elongated to ∼1.55 Å, and the O–N–N angle decreased to 122°. Moreover, the associated charge and spin flows from both copper sites are distinctly uneven due to the slightly different local environments of the Cu ions (Fig. 41a₂ and a₃). The most significant changes in spin and charge transfer occur around the MCEP, but not at the TS point, which is a characteristic feature of the nonadiabatic character of N₂O dissociation, primarily driven by metal-to-ligand electron transfer.

Fig. 41 N₂O decomposition over the bridging Cu⁺∪Cu⁺/ZSM-5 (a₁)–(a₃) and the single Fe²⁺/*BEA (b₁)–(b₃) centres. Singlet and triplet potential energy surface for N₂O dissociation on the Cu⁺∪Cu⁺ centres (a₁), and the changes in the charge (a₂) and spin (a₃) densities along the O–N₂ reaction coordinate, adapted from ref. 353 with permission from the American Chemical Society, Copyright 2014. Operando UV-vis-NIR spectra of the Fe²⁺/*BEA catalyst recorded during N₂O adsorption under variable p_N2O pressures (0–0.05 atm) at 308 K (b₁). Gradual decomposition of the coordinated N₂O into [FeO]²⁺ species and N₂ at 373 K in the time interval of 15 min (b₂), together with the corresponding energy profile accompanied by the associated molecular orbital structures of the key intermediates (b₃), adapted from ref. 20 with permission from Springer Nature, Copyright 2021.

In the case of intrazeolite iron, multiple spectroscopic studies (Mössbauer, VTVH-MCD, UV-vis, IR) corroborated by DFT modelling are available for the definite description of Fe²⁺/z accommodated in the 6-member rings (β-6MR) present in BEA, MFI, and FER, as high-spin square-planar ferrous species (Fe²⁺(d⁶), with S = 2 and D > 0) that are particularly active in N₂O decomposition.⁵ Such centres also appear in the 6MR sites of CHA, where the C_s-symmetric Fe²⁺ cations are less distorted from the square planar arrangement in comparison to C₂-symmetric Fe²⁺ centres in BEA.¹⁴⁷ They are collectively called α-Fe²⁺ species, and their structural deviations across different zeolites are reflected, among others, in the position of the diagnostic 3d_z² → 3d_x²−y² band appearing at 15 900 cm⁻¹ in Fe/BEA, 15 200 cm⁻¹ in Fe/ZSM-5, 16 100 cm⁻¹ in Fe/FER, and 13 000 cm⁻¹ in Fe/CHA.^135,147

The remarkable reactivity of the α-Fe²⁺ centres to N₂O activation is realised by transferring formally two electrons from metal to ligand for complete N–O bond cleavage to produce [Fe=O]²⁺ species (α-O) with concomitant N₂ release. At moderate temperatures (∼150–250 °C), the reaction occurs quickly with an apparent activation energy of 9–13 kcal mol⁻¹ for Fe/BEA,²⁰ Fe/ZSM-5, and Fe/FER catalysts.⁵ This process can be followed quantitatively in the UV-vis-NIR operando set-up by observing the gradual transformation of the band at 15 900 cm⁻¹ into 11 500 cm⁻¹ with the isosbestic point at 13 500 cm⁻¹, which corresponds to the α-Fe²⁺ + N₂O → N₂O–α-Fe²⁺ process (Fig. 41b1). Subsequent transformation of the band at 11 500 cm⁻¹ into the bands located at 16 900 cm⁻¹ and 28 000 cm⁻¹ (Fig. 41b2) is diagnostic of the N₂O dissociation (N₂O–Fe²⁺ → [α-Fe=O]²⁺ + N₂). The appearance of an isosbestic point at 14 000 cm⁻¹ confirms this transformation univocally.²⁰

The prime factors governing efficient N₂O reduction include the binding of N₂O to the active α-Fe²⁺ site in the η¹-O fashion. This promotes the cleavage of the N–O bond, accompanied by the requisite charge redistribution and the formation of a metal–oxo bond. It generates the driving force for the eventual O-atom transfer, without restraining the subsequent evolution of O₂ upon capture of the second N₂O molecule. A more detailed insight into the reaction coordinate for the oxygen release step on α-Fe/BEA (Fig. 41b3) reveals that this process occurs in two sequential MLET events, with a notable internal rearrangement of electrons within the 3d manifold of the iron core. Upon η¹-O attachment, an electron is transferred from the 3d_z² orbital into the 3d_yz orbital, which enables an efficient π-back donation from the Fe²⁺ core into the 3π* LUMO of the N₂O ligand.²⁰ The extent of this back-bonding is controlled, among others, by the presence or absence of an axial ligand. At the bent transition state (138°), located at 25 kcal mol⁻¹, the NN–O bond is elongated to 1.45 Å, whereas the Fe–ON₂ bond is contracted to 1.83 Å. The TS structure evolves into a highly unstable Fe³⁺(S = 5/2)↑↓N₂O˙⁻ (S = 1/2) species. Once the N–O bond is eventually cleaved, a transient Fe³⁺–O˙⁻ oxyl is formed with the LUMO composed mainly of the 2p_x orbital. A second electron transfer converts the oxyl species into the final oxo [Fe=O]²⁺ units with S = 2 (Fig. 41b3).²⁰

The complete energetic profiles of N₂O binding, activation and decomposition into elements on different active sites present in the exemplary Fe/ZSM-5 is shown in Fig. 42. They illustrate two mechanistic variants of this reaction for the iron cations, hosted in the β (Al atoms in the T4, T10 position, 4-fold coordination of Fe) and α (Al atoms in the T1, T7 positions, 5-fold coordination of Fe) positions, see Fig. 42a1 and a2, respectively.³⁸⁰

Fig. 42 Mechanistic variants of N₂O decomposition over single and bridging Fe–oxo. Energetic profiles for N₂O decomposition on Fe²⁺ cations hosted in the β (a₁) and α sites (a₂), and dimeric Fe–O–Fe centres (b) hosted in the ZSM-5 zeolite. The reaction energy diagram for the ferro (navy blue) – and antiferromagnetic (green) state of the dimeric iron active centres (c). The figure was based on data from ref. 164 and 380.

The energies of η¹-O adsorption of the N₂O ligand (−38 kJ mol⁻¹ for α and −39 kJ mol⁻¹ for β sites), the location of the bent transition states TS₁ (44 and 41 kJ mol⁻¹), and the N₂O dissociation energies with concomitant formation of the [FeO]²⁺ species (ΔE = −67 and −63 kJ mol⁻¹) are quite similar for both α and β sites Fig. 42a₁ and a₂. However, the reactivity of the resultant [FeO]²⁺ units with N₂O, which closes the catalytic cycle by releasing O₂, evolves along different pathways.

In the case of β-sites, the N₂O attack is directed at the metal (Fe) centre with the activation barrier of 167 kJ mol⁻¹ (TS₁) and d(Fe–ON₂) = 2.065 Å (Fig. 42a1). Upon dissociation of the ligated N₂O, the ensuing dioxo-iron intermediate is transformed into dioxygen, and the O₂ desorption barrier of 141 kJ mol⁻¹ for S = 2 is decreased to 114 kJ mol⁻¹ for S = 3. For α sites (Fig. 42a2), the N₂O attack is directed at the terminal oxo moiety, and the reaction is driven by the formation of an incipient O–O bond in the TS₂ state with a relatively high barrier of 243 kJ mol⁻¹. The ensuing η²-O,O intermediate (d_(O–O) = 1.345 Å) evolves next into gas phase O₂, restoring the Fe²⁺/α active site. This uphill step costs 141 kJ mol⁻¹ on the quintet (S = 2 + 0) potential energy surface, but is substantially decreased by 53 kJ mol⁻¹ (α) for the septet (S = 2 + 1 = 3) PES, where oxygen is produced in its ground triplet state O₂(³Σ_g⁻). The corresponding energy profile for S = 3 is marked in blue, indicating that only the last two steps are energetically favoured over the S = 2 pathway (Fig. 42a2).

These differences show that the local coordination environment around the Fe²⁺ cations and the PES spin multiplicity exhibit a considerable influence on the detailed mechanism and energetics of N₂O decomposition, modifying the rate constant via the transmission κ coefficient in the Eyring equation:

The Landau–Zener transition probabilities P_LZ (κ = 2P_LZ − P_LZ²) between the PES of different spin states, calculated for N₂O decomposition on both Fe^II/ZSM-5 and [FeO]⁺/ZSM-5 sites, are larger than ∼0.05.²¹ Such values indicate that without correcting the reaction rates of N₂O decomposition for changes in the spin multiplicity, the resultant inaccuracies are equivalent to the activation barrier of ∼4 kcal mol⁻¹ at 700 K. The spin state exhibits substantial influence on the reaction energetics by modifying the electronic structure of the reactants, which is particularly important for attaining the ground triplet state of the O₂ product. Owing to the changes in the spin multiplicity, the reaction pathway becomes energetically more favourable.³⁸⁰

A remarkable mechanistic variant of this reaction has been reported for conceivably adjacent Fe centres hosted in the ferrierite matrix.³⁸⁷ The unique short separation of the neighbouring face-to-face β-6MR sites (7.5 Å in FER vs. >9 Å in MFI and BEA), favoured by the suitable location of Al atoms (see Chapter 4.3.1), allows the adjacent α-Fe cations to act cooperatively for efficient N₂O activation by forming a trans μ-1,3-O,N bridge between both Fe(II) centres (Fe–NNO⋯Fe → Fe + O–Fe + N₂). This finding has been attributed to explaining the superior activity of Fe/FER (∼3 times higher) in comparison to Fe/BEA and Fe/ZSM-5 at 230 °C,⁹⁰ revealing a distinct synergy between the framework topology and Al distribution. Accordingly, the calculated apparent barrier for the O-atom transfer for trans μ-1,3-O,N bridged intermediates (3–4 kcal mol⁻¹) is smaller for the η¹-O species (9–13 kcal mol⁻¹).

For N₂O decomposition on dimeric iron–oxo centres ([Fe–(μ-O²⁻)–Fe]²⁺/ZSM-5), the minimum energy pathway after spin-state optimisation is shown in Fig. 42. Markedly, the reactivity of such species is dramatically higher in comparison to the isolated Fe²⁺ and [FeO]⁺ active sites. It is characterised by the switch of the rate-determining step (RDS) from O₂ evolution at low temperatures (400–600 K) to N₂O dissociation at high temperatures (T > 600 K). The ferromagnetic vs. antiferromagnetic coupling within the [Fe–(μ-O²⁻)–Fe]²⁺ centres exhibits a rather minor influence on the N₂O decomposition.¹⁶⁴ All the energy differences between the key mechanistic steps are below 5.5 kcal mol⁻¹ (Fig. 42c). For molecular oxygen to desorb in its ground triplet state, the barrier for the PES crossing from the antiferromagnetic singlet to the antiferromagnetic triplet is equal to 7.9 kcal mol⁻¹, whereas for the ferromagnetic coupling it increases to 9.5 kcal mol⁻¹.

Survey of N₂O activation and decomposition on various metallozeolite catalysts. The overall reactivity trends in N₂O decomposition for the most commonly investigated Fe, Cu, Co, and Ni zeolites are summarised in Table 5. It is rather difficult to establish a definite universal order of the TMI activity variation since incoherent results have been reported. It is caused, most probably, by insufficient control over the speciation of the active sites, differences in the framework topology, and in the Si/Al and M/Al ratio (loading).

Table 5 Selected activity trends for N₂O decomposition observed for Fe-, Co- and Cu-exchanged SSZ-13, ZSM-5, BEA and MOR metallozeolite catalysts

Topology	Si/Al	M/Al	Activity order	Ref.
CHA (SSZ-13)	12	0.13–0.32	Fe ≫ Co > Cu	371
MFI (ZSM-5)	25	0.5	Cu ≈ Fe > Co	388
MFI (ZSM-5)	18	0.73–1.3	Cu ≫ Co > Fe	389
MFI (ZSM-5)	12.5	0.02–0.245	Fe > Co > Ni > Cu	372
BEA	15	0.11–0.21	Co > Fe > Cu	145
BEA	23–42	0.5–1	Cu > Co > Fe	385
MOR	9.2	0.64–0.80	Co > Fe ≫ Ni ∼ 0	386

---

The effect of TMI nature and loading on N₂O reactivity is illustrated in Fig. 43a₁ for Fe, Co, and Cu/SSZ-13 zeolites.³⁷¹ The reaction rates (Fe ≫ Co > Cu) exhibit appreciably different responses to p_N2O (see insert), with considerable deviation from the pseudo-first-order kinetics. The activation energies decrease in the order 141–148 kJ mol⁻¹ (Co) > 111–113 kJ mol⁻¹ (Cu) > 81–85 kJ mol⁻¹ (Fe), whereas the preexponential factors for Fe (12.6–12.8 ([s kP]⁻¹) and Cu (11.8–12.87 [s kP]⁻¹) are quite similar, and significantly lower than that for Co (19.0–19.9 [s kP]⁻¹). On Co/SSZ-13, the N₂O decomposition occurs on the isolated Co cations, which tend to enhance the preexponential factor at the cost of a higher activation energy. For iron and copper, this process mainly occurs on dual-oxo centres.³⁷¹ Thus, the rate constant for Co/SSZ-13 increases proportionally to the Co/Al ratio (Fig. 43a2). In the case of Fe/SSZ-13 catalysts, the rate constant increases till Fe/Al < 0.06 and then levels off, implying that only a fraction of the iron species is active in the N₂O decomposition. An exponential activity increase with the Cu/Al ratio suggests, in turn, that dimeric centres are responsible for deN₂O reactivity of Cu/SSZ-13, since statistically the number of pairs increases exponentially with copper loading. The same trend of the activity (Fe ≫ Co > Cu) is also observed for BEA zeolites (Fig. 43b).¹⁴⁵ However, since the corresponding plots in Fig. 43a2 are derived from relatively small datasets, these conjectures warrant additional validation.

Fig. 43 Influence of the transition metal type (Co, Cu, Fe) hosted in CHA zeolites (a₁) and the M/Al ratio (a₂) on the N₂O decomposition performance, together with the apparent rate order (a₁, insert). The catalytic performance of Co, Cu, and Fe accommodated in BEA zeolites (b), the influence of the framework type (c) and (d_1–2), and the Si/Al ratio for Co/ZSM-5, Cu/ZSM-5, and Fe/ZSM-5 catalysts (e) on the deN₂O activity. Figure based on data from ref. 145, 371, 388, 390 and 391.

The influence of zeolite topology on N₂O decomposition over various Co/zeolites is illustrated in Fig. 43c, varying as Co/BEA > Co/MOR > Co/ZSM-5 > Co/MCM-49 > Co/ZSM-23 > Co/ZSM-35 > Co/SSZ-13.³⁹² In all cases, the active sites for this reaction are the isolated Co²⁺ cations. There is a consistent relationship between the self-diffusion coefficient of N₂O in different topological zeolites (established by computer simulation) and the N₂O conversion (apparently measured under diffusion limitation), revealing that the pore structure of the zeolites (size and dimensionality of the channels) affects the kinetics of the N₂O decomposition by influencing the N₂O diffusion within the zeolite pores. The three-dimensional pore structure and large pore size of Co/BEA appeared particularly beneficial among the investigated zeolites, leading to the N₂O conversion X > 90% at 400 °C. The worst activity of Co/SSZ-13 (X < 10%) is attributed to the blocking of the small-sized pores due to the presence of segregated cobalt oxide entities, which are inactive. The significant influence of the framework topology on N₂O decomposition was also observed for Fe/Zeolites (Fig. 43d1). Yet the pore size effect is less pronounced in this case (Fe/ZSM-5 > Fe/BEA > FeSSZ-13).³⁹⁰ The dramatic differences in the activity for the Fe/FER ≫ Fe/BEA > Fe/ZSM-5, in turn, are explained by the presence of unique two face-to-face Fe(II) cations cooperating in the efficient activation of the N₂O molecule (discussed above). Notably, in the case of copper, the reactivity decreases as Cu/ZSM-5 > Cu/FER > Cu/MOR (Fig. 43d2),³⁹¹ implying that the effects of the zeolite framework and the nature of the encaged zeolites are of comparable relevance, and cannot be rigorously separated.

Another parameter that controls N₂O activation is the Si/Al ratio³⁸⁸ and the distribution of Al atoms in the T sites of the zeolite framework.^377,384 The dependence of TOF values for N₂O decomposition on Co/ZSM-5, Cu/ZSM-5 and Fe/ZSM-5 catalysts with various Si/Al ratios at different temperatures for M/Al = 0.5³⁸⁸ is shown in Fig. 43e. Generally, for all cations, the TOF values initially increase until Si/Al ∼ 100–150, and then tend to decline except for Cu/ZSM-5 at T > 475 °C. The drop of TOF for Si/Al > 150 is particularly pronounced in the case of the Fe/ZSM-5 catalyst.

The combined effects of structural heterogeneities in both the encapsulated TMI and the zeolite frameworks (including topology, Si/Al ratio, and Al distribution) on N₂O reactivity cannot be rigorously separated. Given the mutual interrelation of these factors, their individual control is insufficient to establish universal reactivity trends for 3d metallozeolites as a function of a selected parameter. The inaccuracies are further emphasised by different conditions of the reaction, and variances resulting from distinct catalyst preparation methods (see Fig. 1).

Conclusions and outlook

After decades of extensive research on metallozeolites and the catalytic chemistry of small molecules, remarkable developments have been made on both fundamental and applied levels. The seemingly simple activation processes occurring in metal–ligand moieties can exhibit a complex nature that can be analyzed in detail by modern computational tools, such as NOCV-ETS, with orbital resolution. For a sensible elucidation of the bonding of non-innocent molecules to transition-metal ions, multireference methods, such as CASSCF or MP2, can provide an adequate treatment of the impact of static and dynamic correlation on the electronic and magnetic structure of the resulting intrazeolite adducts, in close relation to their reactivity. Nevertheless, it is rather difficult to establish definite, universal experimental TMI structure–reactivity trends, as incoherent results have often been reported. This is primarily caused by insufficient control over the speciation of the active sites, as well as by differences in the framework topology and the Si/Al and TMI/Al ratios among the investigated catalysts of nominally the same compositions. Therefore, there is still a need for further advances in the synthesis of metallozeolite materials with atomic precision, their site-specific characterisation, and complex mechanistic studies that benefit from the synergy between modern experimental and theoretical approaches. So far, most efforts have been focused on a few TMI compounds only, being limited essentially to the Cu, Fe, or Co triad, which exhibits the best catalytic performance in many critical catalytic processes, such as SCR, SCO, or C–H bond hydroxylation. However, a deep understanding of the principles of catalytic chemistry of TMI zeolite requires broadening the scope through more comprehensive studies across the periodic table to establish general patterns of bonding, activation, and reactivity, concerning the valence, coordination, and spin states of the encaged cations. These molecular foundations can provide rational tenets for classifying reactivity patterns, optimising existing structures, and predicting new multinuclear and heteronuclear active sites with enhanced reactivity. They may contain not only TMI but also auxiliary non-redox cations (e.g., Ca or Ga) with electronegativities tuned to their purpose. The biocatalysis of small molecules can further inspire the constitution of such multinuclear centres. Additionally, the integration of novel zeolite synthesis and functionalization methods that reduce reliance on poorly controlled Al distribution and minimize unintended TMI speciation is crucial. Equally important is the development of new frameworks that promote deliberately designed supramolecular interactions with reactant molecules, intermediates, and transition states. This may facilitate essential future advancements in the application of metallozeolites in the catalytic chemistry of small molecules.

Further experimental progress can be achieved by applying, for example, phase-sensitive detection to enhance the signal-to-noise ratio and time resolution, particularly in operando catalytic studies of metallozeolites under working conditions.³⁹³ Insights into the dynamical processes occurring in the cages and channels can be provided, in turn, by pulsed-laser excitations, synchronised with time-resolved spectroscopic detection of the resulting molecular changes on the scale from picoseconds to milliseconds.³⁹⁴ Future research into speciation of active sites can also be inspired by recent advances in TEM imaging techniques, such as iDPC-STEM (integrated differential phase contrast STEM) or ptychography, which are capable of single-atom detection located in microporous channels, proving that TMI species can be directly visualised in beam-sensitive materials such as zeolites.^395,396

Finally, many of the topics covered in this review are also relevant to the sister MOF materials functionalized by TMI.³⁹⁷ Due to their unique hybrid structure, they provide even more versatile scaffolds for intentionally tuning the TMI properties, opening new areas of catalytic chemistry where progress is chiefly needed.

Conflicts of interest

There are no conflicts of interest to declare.

List of acronyms

n-MR	n-Member ring
ALMO	Absolutely localized molecular orbitals
BEA	Beta framework type code
BEEF-vdW	Bayesian error estimation functional with van der Waals
BS	Broken symmetry
CAS	Complete active space
CASPT2	Complete active space with second-order perturbation theory
CASSCF	Complete active space self-consistent field
CEC	Cation exchange capacity
CC	Coupled cluster
CCSD(T)	Coupled cluster with single, double, and perturbative triple excitations
CHA	Chabazite framework type code
COHP	Crystal orbital Hamiltonian population
CT	Charge transfer
CVD	Chemical vapour deposition
DFT	Density functional theory
EDE	Dispersion interactions
EDX	Energy-dispersive X-ray spectroscopy
ENDOR	Electron nuclear double resonance
EPR	Electron paramagnetic resonance
ESEEM	Electron spin echo envelope modulation
ET	Electron transfer
ETS-NOCV	Extended-transition-state natural orbitals for chemical valence
EXAFS	Extended X-ray absorption fine structure
FAU	Faujasite framework type code
FC	Franck–Condon
FPT	First principles thermodynamics
FTIR	Fourier transform infrared spectroscopy
HC	High coverage
HC-SCR	Hydrocarbon selective catalytic reduction
HF	Hartree–Fock
HOMO	Highest occupied molecular orbital
HSAB	Hard soft acid base
HYSCORE	HYperfine sublevel CORrElation spectroscopy
iDPC-STEM	(Integrated differential phase contrast STEM)
IEL	Inverse/inverted energy levels
IZA	International zeolite association
LC	Low coverage
LMCT	Ligand-to-metal charge transfer
LUMO	Lowest unoccupied molecular orbital
MAS-NMR	Magic angle spinning nuclear magnetic resonance
MD	Molecular dynamics
MEL	Mixed energy levels
MFI	MFI zeolite framework type code
ML	Machine learning
MLET	Metal-to-ligand electron transfer
MO	Molecular orbital
MOR	Mordenite framework type code
MP	Möeller–Plesset
CN	Coordination number of TMI e.g. Cu4c
NEL	Normal energy levels
NOCV	Natural orbitals for chemical valence
OT	Oxygen transfer
PEC	Principal electronic configuration
PES	Potential energy surface
PL	Photoluminescence
rR	Resonance Raman spectroscopy
SCO	Selective catalytic oxidation
SCR	Selective catalytic reduction
SOC	Spin–orbit coupling
SOMO	Singly occupied molecular orbital
TMI	Transition metal ion
TPD	Temperature-programmed desorption
UV-Vis	Ultraviolet-visible spectroscopy
UV-Vis-NIR	Ultraviolet-visible-near infrared spectroscopy
VB	Valence band
VB-CAS	Valence bond complete active space
VTVH-MCD	Variable-temperature, variable-field magnetic circular dichroism
XAFS	X-ray absorption fine structure
XANES	X-ray absorption near edge structure
X-band CW-EPR	X-band continuous wave electron paramagnetic resonance
ZFS	Zero-field splitting

Data availability

No primary research results, software, or code have been included, and no new data were generated or analysed as part of this review. In a few cases, parts of the compound figures were based on the data available at: https://ruj.uj.edu.pl/entities/publication/2c6d31d7-da08-466d-987d-fcaebbe5db6a, https://ruj.uj.edu.pl/entities/publication/8ae5ec29-e753-4aed-b1e5-1430e36cb24d, https://ruj.uj.edu.pl/entities/publication/d8b1269b-f698-4e80-9051-d96caf48fb52.

Acknowledgements

The publication has been supported by the Faculty of Chemistry and the Priority Research Area (SciMat) under the Strategic Programme Excellence Initiative at Jagiellonian University.

References

1. J. Turkevich, J. Soria and Y. Ono, J. Catal., 1972, 25, 44–54.
2. K. Klier, R. Kellerman and P. J. Hutta, J. Chem. Phys., 1974, 61, 4224–4234.
3. E. Giamello, D. Murphy, G. Magnacca, C. Morterra, Y. Shioya, T. Nomura and M. Anpo, J. Catal., 1992, 136, 510–520.
4. W. Li, Y. Chai, G. Wu and L. Li, J. Phys. Chem. Lett., 2022, 13, 11419–11429.
5. B. E. R. Snyder, M. L. Bols, R. A. Schoonheydt, B. F. Sels and E. I. Solomon, Chem. Rev., 2018, 118, 2718–2768.
6. P. Vanelderen, J. Vancauwenbergh, B. F. Sels and R. A. Schoonheydt, Coord. Chem. Rev., 2013, 257, 483–494.
7. International Zeolite Association Home Page https://iza-online.org/.
8. K. Klier, Langmuir, 1988, 4, 13–25.
9. W. Yan, S. Xi, Y. Du, M. K. Schreyer, S. X. Tan, Y. Liu and A. Borgna, ChemCatChem, 2018, 10, 3078–3085.
10. J. A. Nasir, J. Guan, T. W. Keal, Y. Lu, A. A. Sokol and C. R. A. Catlow, J. Catal., 2024, 438, 115696.
11. B. Ipek, M. J. Wulfers, H. Kim, F. Göltl, I. Hermans, J. P. Smith, K. S. Booksh, C. M. Brown and R. F. Lobo, ACS Catal., 2017, 7, 4291–4303.
12. Y. Li and J. Yu, Nat. Rev. Mater., 2021, 6, 1156–1174.
13. M. Dusselier and M. E. Davis, Chem. Rev., 2018, 118, 5265–5329.
14. J. A. Nasir, A. M. Beale and C. R. A. Catlow, Chem. Soc. Rev., 2024, 53, 11657–11691.
15. C. Paolucci, J. R. Di Iorio, W. F. Schneider and R. Gounder, Acc. Chem. Res., 2020, 53, 1881–1892.
16. Y. Xu, X. Wang, M. Qin and Q. Li, J. Environ. Chem. Eng., 2022, 10, 107270.
17. P. Pietrzyk and Z. Sojka, Chem. Commun., 2007, 1930–1932.
18. M. A. Newton, A. J. Knorpp, V. L. Sushkevich, D. Palagin and J. A. van Bokhoven, Chem. Soc. Rev., 2020, 49, 1449–1486.
19. B. E. R. Snyder, M. L. Bols, H. M. Rhoda, D. Plessers, R. A. Schoonheydt, B. F. Sels and E. I. Solomon, Science, 2021, 373, 327–331.
20. M. L. Bols, B. E. R. Snyder, H. M. Rhoda, P. Cnudde, G. Fayad, R. A. Schoonheydt, V. Van Speybroeck, E. I. Solomon and B. F. Sels, Nat. Catal., 2021, 4, 332–340.
21. A. Heyden, B. Peters, A. T. Bell and F. J. Kiel, J. Phys. Chem. B, 2005, 109, 1857–1873.
22. W. Loewenstein, Am. Mineral., 1954, 39, 92–96.
23. P. Sazama, K. Mlekodaj, J. Dedecek, V. Pashkova, E. Tabor, P. Klein, M. Urbanova, R. Karcz, S. R. Whittleton, H. M. Thomas, A. V. Fishchuk and S. Sklenak, J. Phys. Chem. C, 2019, 123, 7968–7987.
24. S. Dzwigaj and M. Che, J. Phys. Chem. B, 2005, 109, 22167–22174.
25. C. Lamberti, A. Zecchina, E. Groppo and S. Bordiga, Chem. Soc. Rev., 2010, 39, 4951–5001.
26. K. Tarach, J. Sobalska, A. Held, J. Dedecek, E. Tabor and K. Góra-Marek, J. Phys. Chem. C, 2024, 128, 3759–3769.
27. K. Sun, F. Fan, H. Xia, Z. Feng, W.-X. Li and C. Li, J. Phys. Chem. C, 2008, 112, 16036–16041.
28. S. Dźwigaj, M. Matsuoka, R. Franck, M. Anpo and M. Che, J. Phys. Chem. B, 1998, 102, 6309–6312.
29. J. Chen, W. Huang, S. Bao, W. Zhang, T. Liang, S. Zheng, L. Yi, L. Guo and X. Wu, RSC Adv., 2022, 12, 27746–27765.
30. L. Čapek, V. Kreibich, J. Dědeček, T. Grygar, B. Wichterlová, Z. Sobalík, J. A. Martens, R. Brosius and V. Tokarová, Microporous Mesoporous Mater., 2005, 80, 279–289.
31. K. Góra-Marek, A. Glanowska and J. Datka, Microporous Mesoporous Mater., 2012, 158, 162–169.
32. K. Góra-Marek, B. Gil and J. Datka, Appl. Catal., A, 2009, 353, 117–122.
33. M. Fedyna, B. Mozgawa, F. Zasada, K. Góra-Marek, J. Gryboś, W. Piskorz, C. Yin, Z. Zhao, P. Pietrzyk and Z. Sojka, Appl. Catal., B, 2023, 325, 122309.
34. S. Zamani, M. Chiesa, V. Meynen, Y. Xiao, B. Prelot, J. Zajac, F. Verpoort, P. Cool and S. Van Doorslaer, J. Phys. Chem. C, 2010, 114, 12966–12975.
35. D. Goldfarb and K. Zukerman, Chem. Phys. Lett., 1990, 171, 167–174.
36. D. Goldfarb, Phys. Chem. Chem. Phys., 2006, 8, 2325–2343.
37. V. Lagostina, E. Salvadori, M. Chiesa and E. Giamello, J. Catal., 2020, 391, 397–403.
38. S. Bordiga, E. Groppo, G. Agostini, J. A. van Bokhoven and C. Lamberti, Chem. Rev., 2013, 113, 1736–1850.
39. S. H. van Vreeswijk and B. M. Weckhuysen, Nat. Sci. Rev., 2022, 9, nwac047.
40. A. Stępniewski, M. Radoń, K. Góra-Marek and E. Broclawik, Phys. Chem. Chem. Phys., 2016, 18, 3716–3729.
41. W. Jeevapong, J. Sittiwong, M. Probst, B. Boekfa, C. Wattanakit, T. Maihom and J. Limtrakul, J. Phys. Chem. C, 2023, 127, 8473–8481.
42. J. Klimeš and D. P. Tew, J. Chem. Phys., 2019, 151, 234108.
43. A. M. Teale, T. Helgaker, A. Savin, C. Adamo, B. Aradi, A. V. Arbuznikov, P. W. Ayers, E. J. Baerends, V. Barone, P. Calaminici, E. Cancès, E. A. Carter, P. K. Chattaraj, H. Chermette, I. Ciofini, T. D. Crawford, F. De Proft, J. F. Dobson, C. Draxl, T. Frauenheim, E. Fromager, P. Fuentealba, L. Gagliardi, G. Galli, J. Gao, P. Geerlings, N. Gidopoulos, P. M. W. Gill, P. Gori-Giorgi, A. Görling, T. Gould, S. Grimme, O. Gritsenko, H. J. A. Jensen, E. R. Johnson, R. O. Jones, M. Kaupp, A. M. Köster, L. Kronik, A. I. Krylov, S. Kvaal, A. Laestadius, M. Levy, M. Lewin, S. Liu, P.-F. Loos, N. T. Maitra, F. Neese, J. P. Perdew, K. Pernal, P. Pernot, P. Piecuch, E. Rebolini, L. Reining, P. Romaniello, A. Ruzsinszky, D. R. Salahub, M. Scheffler, P. Schwerdtfeger, V. N. Staroverov, J. Sun, E. Tellgren, D. J. Tozer, S. B. Trickey, C. A. Ullrich, A. Vela, G. Vignale, T. A. Wesolowski, X. Xu and W. Yang, Phys. Chem. Chem. Phys., 2022, 24, 28700–28781.
44. A. M. Teale, P. Verma and D. G. Truhlar, Trends Chem., 2020, 2, 302–318.
45. K. D. Vogiatzis, M. V. Polynski, J. K. Kirkland, J. Townsend, A. Hashemi, C. Liu and E. A. Pidko, Chem. Rev., 2019, 119, 2453–2523.
46. S. Svelle, C. Tuma, X. Rozanska, T. Kerber and J. Sauer, J. Am. Chem. Soc., 2009, 131, 816–825.
47. F. Berger, M. Rybicki and J. Sauer, ACS Catal., 2023, 13, 2011–2024.
48. K. Sillar, A. Hofmann and J. Sauer, J. Am. Chem. Soc., 2009, 131, 4143–4150.
49. M. Radoń and K. Pierloot, J. Phys. Chem. A, 2008, 112, 11824–11832.
50. M. Radoń, E. Broclawik and K. Pierloot, J. Phys. Chem. B, 2010, 114, 1518–1528.
51. Q. M. Phung, H. N. Nam and A. Ghosh, Inorg. Chem., 2023, 62, 20496–20505.
52. S. Vancoillie, H. Zhao, M. Radoń and K. Pierloot, J. Chem. Theory Comput., 2010, 6, 576–582.
53. M. Radoń, G. Drabik, M. Hodorowicz and J. Szklarzewicz, Chem. Sci., 2024, 15, 20189.
54. R. K. Carlson, S. O. Odoh, S. J. Tereniak, C. C. Lu and L. Gagliardi, J. Chem. Theory Comput., 2015, 11, 4093–4101.
55. D. A. Pantazis, J. Chem. Theory Comput., 2019, 15, 938–948.
56. Z. A. Alaithan, N. Harrison and G. Sastre, J. Phys. Chem. C, 2021, 125, 19200–19208.
57. C. J. Heard, L. Grajciar and A. Erlebach, Nanoscale, 2024, 16, 8108–8118.
58. A. J. Porter and A. J. O’Malley, J. Phys. Chem. C, 2021, 125, 11567–11579.
59. V. Van Speybroeck, M. Bocus, P. Cnudde and L. Vanduyfhuys, ACS Catal., 2023, 13, 11455–11493.
60. G. Wang, C. Zhi, Y. Wang and Q. Wang, Comput. Theor. Chem., 2023, 1227, 114228.
61. D. J. Hutton, D. H. Lopez and F. Göltl, React. Chem. Eng., 2024, 9, 1685–1695.
62. M. P. Mitoraj, A. Michalak and T. Ziegler, J. Chem. Theory Comput., 2009, 5, 962–975.
63. P. Kozyra, M. Świętek, J. Datka and E. Brocławik, J. Comput. Chem., Jpn., 2013, 12, 30–37.
64. P. Pietrzyk, K. Podolska, T. Mazur and Z. Sojka, J. Am. Chem. Soc., 2011, 133, 19931–19943.
65. V. L. Deringer, A. L. Tchougreeff and R. Dronskowski, J. Phys. Chem. A, 2011, 115, 5461–5466.
66. C. Liu, I. Tranca, R. A. van Santen, E. J. M. Hensen and E. A. Pidko, J. Phys. Chem. C, 2017, 121, 23520–23530.
67. X. Li, H. Han, N. Evangelou, N. J. Wichrowski, P. Lu, W. Xu, S. J. Hwang, W. Zhao, C. Song, X. Guo, A. Bhan, I. G. Kevrekidis and M. Tsapatsis, Nat. Commun., 2023, 14, 3152.
68. S. Ma and Z. P. Liu, Chem. Sci., 2022, 13, 5055–5068.
69. D. Plessers, M. L. Bols, H. M. Rhoda, A. J. Heyer, E. I. Solomon, B. F. Sels and R. A. Schoonheydt, Comprehensive Inorganic Chemistry III, 3rd edn, 2023, pp. 148–164.
70. Q. Zhang, S. Gao and J. Yu, Chem. Rev., 2023, 123, 6039–6106.
71. N. Kosinov, C. Liu, E. J. M. Hensen and E. A. Pidko, Chem. Mater., 2018, 30, 3177–3198.
72. H. M. Rhoda, A. J. Heyer, B. E. R. Snyder, D. Plessers, M. L. Bols, R. A. Schoonheydt, B. F. Sels and E. I. Solomon, Chem. Rev., 2022, 122, 12207–12243.
73. Y. Wu, P. Deng, L. Liu, J. Zhang, H. Liu, X. Gao, F.-S. Xiao and L. Wang, Chem. Soc. Rev., 2025, 54, 4745–4762.
74. Y. Li, L. Li and J. Yu, Chem, 2017, 3, 928–949.
75. J. Pérez-Ramírez, C. H. Christensen, K. Egeblad, C. H. Christensen and J. C. Groen, Chem. Soc. Rev., 2008, 37, 2530–2542.
76. C. Chizallet, C. Bouchy, K. Larmier and G. Pirngruber, Chem. Rev., 2023, 123, 6107–6196.
77. C. Cheng, P. Li, X. Yang and R. T. Yang, Catal. Sci. Technol., 2016, 6, 7561–7568.
78. E. Morra, G. Berlier, E. Borfecchia, S. Bordiga, P. Beato and M. Chiesa, J. Phys. Chem. C, 2017, 121, 14238–14245.
79. K. A. Chalupka, W. K. Jozwiak, J. Rynkowski, W. Maniukiewicz, S. Casale and S. Dzwigaj, Appl. Catal., B, 2014, 146, 227–236.
80. P. Pietrzyk, K. Podolska-Serafin, K. Góra-Marek, A. Krasowska and Z. Sojka, Microporous Mesoporous Mater., 2020, 291, 109692.
81. T. Yu, Y. Su, A. Wang, B. M. Weckhuysen and W. Luo, ChemCatChem, 2021, 13, 2766–2770.
82. P. J. Smeets, J. S. Woertink, B. F. Sels, E. I. Solomon and R. A. Schoonheydt, Inorg. Chem., 2010, 49, 3573–3583.
83. B. Mozgawa, Copper active sites in chabazite–type zeolites in reactions of selective catalytic reduction of NOx with ammonia – insight into structure and reactivity using spectroscopic methods and DFT modelling, PhD thesis, Jagiellonian University, 2024.
84. J. E. Olszowka, M. Lemishka, K. Mlekodaj, P. Kubat, D. Rutkowska-Żbik, J. Dedecek and E. Tabor, J. Phys. Chem. C, 2021, 125, 9060–9073.
85. Z. Sobalik, Z. Tvaruzkova and B. Wichterlova, Microporous Mesoporous Mater., 1998, 25, 225–228.
86. E. Broclawik, J. Datka, B. Gil and P. Kozyra, Phys. Chem. Chem. Phys., 2000, 2, 401–405.
87. M. Lemishka, J. Dedecek, K. Mlekodaj, Z. Sobalik, S. Sklenak and E. Tabor, Pure Appl. Chem., 2019, 91, 1721–1732.
88. Y. Jangjou, Q. Do, Y. Gu, L.-G. Lim, H. Sun, D. Wang, A. Kumar, J. Li, L. C. Grabow and W. S. Epling, ACS Catal., 2018, 8, 1325–1337.
89. B. Wichterlova, J. Dedecek and Z. Sobalik, Single Metal Ions in Host Zeolite Matrices: Structure-Activity-Selectivity-Relationships, in Catalysis by Unique Metal Ion Structures in Solid Matrices, ed. G. Centi, B. Wichterlova and A. T. Bell, 2001, pp. 31–53.
90. S. Sklenak, P. C. Andrikopoulos, B. Boekfa, B. Jansang, J. Nováková, L. Benco, T. Bucko, J. Hafner, J. Dedecek and Z. Sobalík, J. Catal., 2010, 272, 262–274.
91. P. J. Carl and S. C. Larsen, J. Phys. Chem. B, 2000, 104, 6568–6575.
92. A. Godiksen, P. N. R. Vennestrøm, S. B. Rasmussen and S. Mossin, Top. Catal., 2017, 60, 13–29.
93. R. P. Vélez, U. Bentrup, W. Grünert and A. Brückner, Top. Catal., 2017, 60, 1641–1652.
94. Z. Sojka, Catal. Rev., 1995, 37, 461–512.
95. P. C. Bruzzese, E. Salvadori, S. Jäger, M. Hartman, B. Civalleri, A. Pöppl and M. Chiesa, Nat. Commun., 2021, 12, 4638.
96. V. Nagarajan, D. Rings, L. Moschkowitz, M. Hartmann and A. Pöppl, Chem. Lett., 2005, 34, 1614–1615.
97. E. Marceau, L. Bonneviot, S. Dzwigaj, J.-F. Lambert, C. Louis and X. Carrier, J. Catal., 2021, 396, 104–121.
98. J. A. Weil, J. R. Bolton and J. E. Wertz, Electron Paramagnetic Resonance Elementary theory and Practical Applications, Wiley, NY, 1994.
99. F. E. Mabbs and D. Collison, Electron Resonance of d Transition Metal Compounds, Elsevier, Amsterdam, 1992.
100. F. Giordanino, P. N. R. Vennestrøm, L. F. Lundegaard, F. N. Stappen, S. Mossin, P. Beato, S. Bordiga and C. Lamberti, Dalton Trans., 2013, 42, 12741–12761.
101. P. Rieger, Coord. Chem. Rev., 1994, 135/136, 203–286.
102. F. E. Mabbs and D. Collison, Electron Resonance of d Transition Metal Compounds, Elsevier, Amsterdam, 1992.
103. V. Ovchinnikov and V. N. Konstantinov, J. Magn. Reson., 1979, 32, 179.
104. J. C. Conesa and J. Soria, J. Phys. Chem., 1978, 82, 1575–1578.
105. A. Tuel, J. Diab, P. Gelin, M. Dufaux, J.-F. Dutel and Y. B. Taarit, J. Mol. Catal., 1990, 63, 95–102.
106. J. Peisach and W. E. Blumberg, Arch. Biochem. Biophys., 1974, 165, 691–708.
107. P. J. Carl and S. C. Larsen, J. Catal., 1999, 182, 208–218.
108. E. Morra, E. Giamello and M. Chiesa, Chem. – Eur. J., 2014, 20, 7381–7388.
109. A. Tuel, J. Diab, P. Gelin, M. Dufaux, J.-F. Dutel and Y. B. Taarit, J. Mol. Catal., 1990, 63, 95–102.
110. P. Pietrzyk, Z. Sojka, S. Dźwigaj and M. Che, J. Am. Chem. Soc., 2007, 129, 14174–14175.
111. B. M. Weckhuysen and R. A. Schoonheydt, Stud. Surf. Sci. Catal., 1994, 84, 965–972.
112. A. M. Volodin, G. M. Zhidomirov, K. A. Dubkov, E. J. M. Hensen and R. A. van Santen, Catal. Today, 2005, 110, 247–254.
113. E.-M. El-Malki, D. Werst, P. E. Doan and W. M. H. Sachtler, J. Phys. Chem. B, 2000, 104, 5924–5931.
114. K. Podolska-Serafin and P. Pietrzyk, J. Mol. Struct., 2019, 1180, 754–763.
115. E. Fernández, M. Moreno-González, M. Moliner, T. Blasco, M. Boronat and A. Corma, Top. Catal., 2018, 61, 810–832.
116. P. C. Bruzzese, E. Salvadori, B. Civalleri, S. Jäger, M. Hartmann, A. Pöppl and M. Chiesa, J. Am. Chem. Soc., 2022, 144, 13079–13083.
117. E. Morra, M. Signorile, E. Salvadori, S. Bordiga, E. Giamello and M. Chiesa, Angew. Chem., Int. Ed., 2019, 58, 12398–12403.
118. P. Rzepka, T. Huthwelker, J. Dědeček, E. Tabor, M. Bernauer, S. Sklenák, K. Mlekodaj and J. A. van Bokhoven, Science, 2025, 388, 423–428.
119. J. Krzystek, A. Ozarowski and J. Telser, Coord. Chem. Rev., 2006, 250, 2308–2324.
120. B. M. Weckhuysen, A. A. Verberckmoes, M. G. Uytterhoeven, F. E. Mabbs, D. Collison, E. de Boer and R. A. Schoonheydt, J. Phys. Chem. B, 2000, 104, 37–42.
121. A. Peng, M. C. Kung, M. O. Ross, B. M. Hofman and H. H. Kung, Top. Catal., 2020, 63, 1708–1716.
122. S. A. Bonke, T. Risse, A. Schnegg and A. Brückner, Nat. Rev. Methods Primers, 2021, 1, 33.
123. R. Baran, L. Valentin and S. Dźwigaj, Phys. Chem. Chem. Phys., 2016, 18, 12050.
124. D. E. De Vos, B. M. Weckhuysen and T. Bein, J. Am. Chem. Soc., 1996, 118, 9615–9622.
125. J. Livage, Materials, 2010, 3, 4175–4195.
126. J. Grand, S. N. Talapaneni, A. Vicente, C. Fernandez, E. Dib, H. A. Aleksandrov, G. N. Vayssilov, R. Retoux, P. Boullay, J. P. Gilson, V. Valtchev and S. Mintova, Nat. Mater., 2017, 16, 1010–1015.
127. D. Zhou, Y. Zhang, H. Zhu, D. Ma and X. Bao, J. Phys. Chem. C, 2007, 111, 2081–2091.
128. S. Wannakao, B. Boekfa, P. Khongpracha, M. Probst and J. Limtrakul, ChemPhysChem, 2010, 11, 3432–3438.
129. M. F. Fellah and I. Onal, Phys. Chem. Chem. Phys., 2013, 15, 13969–13997.
130. S. Dźwigaj, M. Matsuoka, M. Anpo and M. Che, J. Phys. Chem. B, 2000, 104, 6012–6020.
131. B. E. R. Snyder, L. H. Böttger, M. L. Bols, J. J. Yan, H. M. Rhod, A. B. Jacobs, M. Y. Hu, J. Zhao, E. E. Alp, B. Hedman, K. O. Hodgson, R. A. Schoonheydt, B. F. Sels and E. I. Solomon, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, 4565–4570.
132. J. B. Lim, S. H. Cha and S. B. Hong, Appl. Catal., B, 2019, 243, 750–759.
133. M. H. Mahyuddin, Y. Shiota, A. Staykov and K. Yoshizawa, Inorg. Chem., 2017, 56, 10370–10380.
134. A. Oda, T. Ohkubo, T. Yumura, H. Kobayashi and Y. Kuroda, Inorg. Chem., 2019, 58, 327–338.
135. B. E. R. Snyder, P. Vanelderen, M. L. Bols, S. D. Hallaert, L. H. Böttger, L. Ungur, K. Pierloot, R. A. Schoonheydt, B. F. Sels and E. I. Solomon, Nature, 2016, 536, 317–321.
136. S. Ye and F. Neese, Proc. Natl. Acad. Sci. U. S. A., 2011, 108, 1228–1233.
137. Y. Shimoyama and T. Kojima, Inorg. Chem., 2019, 58, 9517–9542.
138. F. Tielens and S. Dźwigaj, Catal. Today, 2010, 152, 66–69.
139. M. Matsuoka, S. Higashimoto, H. Yamashita and M. Anpo, Res. Chem. Intermed., 2000, 26, 85–92.
140. M. Anpo, T.-H. Kim and M. Matsuoka, Catal. Today, 2009, 142, 114–124.
141. S. Higashimoto, M. Matsuoka, H. Yamashita, M. Anpo, O. Kitao, H. Hidaka, M. Che and E. Giamello, J. Phys. Chem. B, 2000, 104, 10288–10292.
142. H. S. Lacheen and E. Iglesia, J. Phys. Chem. B, 2006, 110, 5462–5472.
143. M. Smoliło-Utrata, K. A. Tarach, K. Samson, M. Gackowski, E. Madej, J. Korecki, G. Mordarski, M. Śliwa, S. Jarczewski, J. Podobinski, P. Kuśtrowski, J. Datka, D. Rutkowska-Żbik and K. Góra-Marek, Int. J. Mol. Sci., 2022, 23, 5584.
144. M. H. Mahyuddin, A. Staykov, Y. Shiota and K. Yoshizawa, ACS Catal., 2016, 6, 8321–8331.
145. N. Liu, R. Zhang, B. Chen, Y. Li and Y. Li, J. Catal., 2012, 294, 99–112.
146. J. Gao, Y. Zheng, Y. Tang, J.-M. Jehng, R. Gryboś, J. Handzlik, I. E. Wachs and S. G. Podkolzin, ACS Catal., 2015, 5, 3078–3092.
147. M. L. Bols, S. D. Hallaert, B. E. R. Snyder, J. Devos, D. Plessers, H. M. Rhoda, M. Dusselier, R. A. Schoonheydt, K. Pierloot, E. I. Solomon and B. F. Sels, J. Am. Chem. Soc., 2018, 140, 12021–12032.
148. K. Mlekodaj, M. Lemishka, S. Sklenak, J. Dědeček and E. Tabor, Chem. Commun., 2021, 57, 3472–3475.
149. J. Wang, G. Li, X. Ju, H. Xia, F. Fan, J. Wang, Z. Feng and C. Li, J. Catal., 2013, 301, 77–82.
150. D. T. Bregante, L. N. Wilcox, C. Liu, C. Paolucci, R. Gounder and D. W. Flaherty, ACS Catal., 2021, 11, 11873–11884.
151. T. Ikuno, S. Grundner, A. Jentys, G. Li, E. Pidko, J. Fulton, M. Sanchez-Sanchez and J. A. Lercher, J. Phys. Chem. C, 2019, 123, 8759–8769.
152. I. Kurzydym and I. Czekaj, Molecules, 2022, 27, 300.
153. G. Wang, L. Huang, W. Chen, J. Zhou and A. Zheng, Phys. Chem. Chem. Phys., 2018, 20, 26522.
154. G. Li, E. A. Pidko, R. A. van Santen, C. Li and E. J. M. Hensen, J. Phys. Chem. C, 2013, 117, 413–426.
155. M. H. Mahyuddin and K. Yoshizawa, Catal. Sci. Technol., 2018, 8, 5875–5885.
156. P. Pietrzyk, K. Góra-Marek, T. Mazur, B. Mozgawa, M. Radoń, M. Chiesa, Z. Zhao and Z. Sojka, J. Catal., 2021, 394, 206–219.
157. P. Pietrzyk, C. Dujardin, K. Góra-Marek, P. Granger and Z. Sojka, Phys. Chem. Chem. Phys., 2012, 14, 2203–2215.
158. D. W. Fickel and R. F. Lobo, J. Phys. Chem. C, 2010, 114, 1633–1640.
159. A. R. Kulkarni, Z. J. Zhao, S. Siahrostami, J. K. Nørskov and F. Studt, ACS Catal., 2016, 6, 6531–6536.
160. G. D. Pirngruber, P. K. Roy and R. Prins, Phys. Chem. Chem. Phys., 2006, 8, 3939–3950.
161. A. J. Heyer, D. Plessers, J. Ma, B. E. R. Snyder, R. A. Schoonheydt, B. F. Sels and E. I. Solomon, J. Am. Chem. Soc., 2024, 146, 6061–6071.
162. L. Noodleman and E. J. Baerends, J. Am. Chem. Soc., 1984, 106, 2316–2327.
163. T. Soda, Y. Kitagawa, T. Onishi, Y. Takano, Y. Shigeta, H. Nagao, Y. Yoshioka and K. Yamaguchi, Chem. Phys. Lett., 2000, 319, 223–230.
164. N. Hansen, A. Heyden, A. T. Bell and F. J. Keil, J. Phys. Chem. C, 2007, 111, 2092–2101,  DOI:10.1021/jp065574q.
165. S. Li, Y. Wang, T. Wu and W. F. Schneider, ACS Catal., 2018, 8, 10119–10130.
166. G. Li and E. A. Pidko, ChemCatChem, 2019, 11, 134–156.
167. C. Liu, G. Li, E. J. M. Hensen and E. A. Pidko, ACS Catal., 2015, 5, 7024–7033.
168. G. Li, E. A. Pidko, R. A. van Santen, C. Li and E. J. M. Hensen, J. Phys. Chem. C, 2013, 117, 413–426.
169. L. Grajciar, C. J. Heard, A. A. Bondarenko, M. V. Polynski, J. Meeprasert, E. A. Pidko and P. Nachtigall, Chem. Soc. Rev., 2018, 47, 8307–8348.
170. K. Reuter and M. Scheffler, Phys. Rev. B:Condens. Matter Mater. Phys., 2002, 65, 035406.
171. K. Reuter and M. Scheffler, Phys. Rev. Lett., 2003, 90, 046103.
172. B. Mozgawa, F. Zasada, M. Fedyna, K. Góra-Marek, C. Yin, Z. Zhao, Z. Sojka and P. Pietrzyk, ACS Catal., 2025, 15(4), 2715–2734.
173. A. D. Becke, J. Chem. Phys., 1993, 98, 5648–5652.
174. C. Adamo and V. Barone, J. Chem. Phys., 1999, 110, 6158–6170.
175. S. Grimme, Wiley Interdiscip. Rev.:Comput. Mol. Sci., 2011, 1, 211–228.
176. K. E. Riley, M. Pitoňák, P. Jurečka and P. Hobza, Chem. Rev., 2010, 110, 5023–5063.
177. F. Göltl, A. Grüneis, T. Bučko and J. Hafner, J. Chem. Phys., 2012, 137, 114111.
178. T. Kerber, M. Sierka and J. Sauer, J. Comput. Chem., 2008, 29, 2088–2097.
179. L. Chen, T. V. W. Janssens and H. Grönbeck, Phys. Chem. Chem. Phys., 2019, 21, 10923–10930.
180. M. Fischer, Z. Kristallogr. – Cryst. Mater., 2015, 230, 325–336.
181. K. Stanciakova, J. N. Louwen, B. M. Weckhuysen, R. E. Bulo and F. Göltl, J. Phys. Chem. C, 2021, 125, 20261–20274.
182. T. Hirano, in MOPAC Manual, ed. J. P. Stewart, 7th edn, 1993.
183. J. A. Greathouse, M. J. Paul, G. Xu and M. D. Powell, Clays Clay Miner., 2023, 71, 54–73.
184. G. Pireddu, F. G. Pazzona, P. Demontis and M. A. Załuska-Kotur, J. Chem. Theory Comput., 2019, 15, 6931–6943.
185. G. Piccini and J. Sauer, J. Chem. Theory Comput., 2014, 10, 2479–2487.
186. M. L. Bols, J. Devos, H. M. Rhoda, D. Plessers, E. I. Solomon, R. A. Schoonheydt, B. F. Sels and M. Dusselier, J. Am. Chem. Soc., 2021, 143, 16243–16255.
187. R. Zhang, E. Anderst, K. Groden and J.-S. McEwen, Ind. Eng. Chem. Res., 2018, 57, 13396–13405.
188. C. Paolucci, A. A. Parekh, I. Khurana, J. R. Di Iorio, H. Li, J. D. Albarracin Caballero, A. J. Shih, T. Anggara, W. N. Delgass, J. T. Miller, F. H. Ribeiro, R. Gounder and W. F. Schneider, J. Am. Chem. Soc., 2016, 138, 6028–6048.
189. R. Garten, W. Delgass and M. Boudart, J. Catal., 1970, 18, 90–107.
190. J. Pérez-Ramírez, G. Mul, F. Kapteijn, J. Moulijn, A. Overweg, A. Doménech, A. Ribera and I. Arends, J. Catal., 2002, 207, 113–126.
191. G. Li, P. Vassilev, M. Sanchez-Sanchez, J. A. Lercher, E. J. M. Hensen and E. A. Pidko, J. Catal., 2016, 338, 305–312.
192. E. A. Pidko, E. J. M. Hensen and R. A. van Santen, Proc. R. Soc. London, Ser. A, 2012, 468, 2070–2086.
193. E. A. Pidko, E. J. M. Hensen, G. M. Zhidomirov and R. A. van Santen, J. Catal., 2008, 255, 139–143.
194. E. A. Pidko, R. A. van Santen and E. J. M. Hensen, Phys. Chem. Chem. Phys., 2009, 11, 2893–2902.
195. S. Wang, C. Li, C. Liu and W. Zhuang, ChemCatChem, 2025, 17, e202401416.
196. S. A. Bates, A. A. Verma, C. Paolucci, A. A. Parekh, T. Anggara, A. Yezerets, W. F. Schneider, J. T. Miller, W. N. Delgass and F. H. Ribeiro, J. Catal., 2014, 312, 87–97.
197. B. R. Goodman, K. C. Hass, W. F. Schneider and J. B. Adams, Catal. Lett., 2000, 68, 85–93.
198. R. J. Deeth, Phys. Chem. Chem. Phys., 2024, 26, 18138–18148.
199. R. Zhang, N. Liu, Z. Lei and B. Chen, Chem. Rev., 2016, 116, 3658–3721.
200. V. Lyaskovskyy and B. de Bruin, ACS Catal., 2012, 2, 270–279.
201. M. Mitoraj and A. Michalak, J. Mol. Model., 2007, 13, 347–355.
202. A. Michalak, M. Mitoraj and T. Ziegler, J. Phys. Chem. A, 2008, 112, 1933–1939.
203. P. Kozyra, E. Broclawik, M. P. Mitoraj and J. Datka, J. Phys. Chem. C, 2013, 117, 7511–7518.
204. F. Sagan and M. P. Mitoraj, Non-covalent interactions in selected transition metal complexes. in Transition Metals in Coordination Environments: Computational Chemistry and Catalysis Viewpoints, ed. E. Broclawik, T. Borowski and M. Radoń, Springer International Publishing, Cham, Switzerland, 2019, pp. 65–89.
205. M. P. Mitoraj, F. Sagan, D. W. Szczepanik, J. H. de Lange, A. L. Ptaszek, D. M. E. van Niekerk and I. Cukrowski, ChemPhysChem, 2020, 21, 494–502.
206. Metal-Ligand Interactions: From Atoms, to Clusters, to Surfaces, ed. D. R. Salahub and N. Russo, Springer, Dordrecht, 1990.
207. B. Zouchoune and A. Saiad, Inorg. Chim. Acta, 2018, 473, 204–215.
208. Y. Jing, J. Liu, Z. Ye, J. Su, Y. Liu and Z. Ke, Catal. Sci. Technol., 2021, 11, 7189–7199.
209. Y. Gu, Q. Zhu, Z. Liu, C. Fu, J. Wu, Q. Zhu, Q. Jia and J. Ma, J. Mater. Chem. A, 2022, 10, 14976–14988.
210. A. Delabie, K. Pierloot, M. H. Groothaert, B. M. Weckhuysen and R. A. Schoonheydt, Microporous Mesoporous Mater., 2000, 37, 209–222.
211. H. Petitjean, C. Chizallet and D. Berthomieu, Ind. Eng. Chem. Res., 2018, 57, 15982–15990.
212. P. Sazama, J. Moravkova, S. Sklenak, A. Vondrova, E. Tabor, G. Sadovska and R. Pilar, ACS Catal., 2020, 10, 3984–4002.
213. Y. Mao, H.-F. Wang and P. Hu, Int. J. Quantum Chem., 2015, 115, 618–630.
214. T. V. W. Janssens, H. Falsig, L. F. Lundegaard, P. N. R. Vennestrøm, S. B. Rasmussen, P. Georg Moses, F. Giordanino, E. Borfecchia, K. A. Lomachenko, C. Lamberti, S. Bordiga, A. Godiksen, S. Mossin and P. Beato, ACS Catal., 2015, 5, 2832–2845.
215. C.-Y. Sung, S. Al Hashimi, A. McCormick, M. Cococcioni and M. Tsapatsis, Microporous Mesoporous Mater., 2013, 172, 7–12.
216. R. Z. Khaliullin, A. T. Bell and M. Head-Gordon, J. Chem. Phys., 2008, 128, 184112.
217. R. Z. Khaliullin, E. A. Cobar, R. C. Lochan, A. T. Bell and M. Head-Gordon, J. Phys. Chem. A, 2007, 111, 8753–8765.
218. A. Daouli, E. P. Hessou, H. Monnier, M.-A. Dziurla, A. Hasnaoui, G. Maurin and M. Badawi, Phys. Chem. Chem. Phys., 2022, 24, 15565–15578.
219. L. Benco and D. Tunega, Phys. Chem. Miner., 2009, 36, 281–290.
220. F. Gao, D. Mei, Y. Wang, J. Szanyi and C. H. Peden, J. Am. Chem. Soc., 2017, 139, 4935–4942.
221. L. N. Wilcox, S. H. Krishna, C. B. Jones and R. Gounder, Catal. Sci. Technol., 2021, 11, 7932–7942.
222. M. Colombo, G. Koltsakis, I. Nova and E. Tronconi, Catal. Today, 2012, 188, 42–52.
223. F. Giordanino, E. Borfecchia, K. A. Lomachenko, A. Lazzarini, G. Agostini, E. Gallo, A. V. Soldatov, P. Beato, S. Bordiga and C. Lamberti, J. Phys. Chem. Lett., 2014, 5, 1552–1559.
224. I. Lezcano-Gonzalez, U. Deka, B. Arstad, A. Van Yperen-De Deyne, K. Hemelsoet, M. Waroquier, V. Van Speybroeck, B. M. Weckhuysen and A. M. Beale, Phys. Chem. Chem. Phys., 2014, 16, 1639–1650.
225. L. Chen, T. V. W. Janssens, M. Skoglundh and H. Grönbeck, Top. Catal., 2019, 62, 93–99.
226. K. Leistner, K. Xie, A. Kumar, K. Kamasamudram and L. Olsson, Catal. Lett., 2017, 147, 1882–1890.
227. B. Mozgawa, F. Zasada, M. Fedyna, K. Góra-Marek, E. Tabor, K. Mlekodaj, J. Dědeček, Z. Zhao, P. Pietrzyk and Z. Sojka, Chem. Soc. Rev., 2021, 50, 17159–17180.
228. C. Paolucci, I. Khurana, A. A. Parekh, S. Li, A. J. Shih, H. Li, J. R. Di Iorio, J. D. Albarracin-Caballero, A. Yezerets and J. T. Miller, Science, 2017, 357, 898–903.
229. R. Millan, P. Cnudde, V. Van Speybroeck and M. Boronat, JACS Au, 2021, 1, 1778–1787.
230. R. Millan, E. Bello-Jurado, M. Moliner, M. Boronat and R. Gomez-Bombarelli, ACS Cent. Sci., 2023, 9, 2044–2056.
231. Y. Fu, W. Ding, H. Lei, Y. Sun, J. Du, Y. Yu, U. Simon, P. Chen, Y. Shan, G. He and H. He, J. Am. Chem. Soc., 2024, 146, 11141–11151.
232. A. Martini, C. Negri, L. Bugarin, G. Deplano, R. K. Abasabadi, K. A. Lomachenko, T. V. W. Janssens, S. Bordiga, G. Berlier and E. Borfecchia, J. Phys. Chem. Lett., 2022, 13, 6164–6170.
233. A. Daouli, J. Rey, E. H. Lahrar, V. Valtchev, M. Badawi and R. Guillet-Nicolas, Langmuir, 2023, 39, 15962–15973.
234. V. Zdravkova, M. Mihaylov and K. Hadjiivanov, J. Phys. Chem. C, 2012, 116, 12706–12711.
235. A. Itadani, M. Tanaka, T. Mori, M. Nagao, H. Kobayashi and Y. Kuroda, J. Phys. Chem. C, 2007, 111, 12011–12023.
236. G. Spoto, S. Bordiga, G. Ricchiardi, D. Scarano, A. Zecchina and F. Geobaldo, J. Chem. Soc., Faraday Trans., 1995, 91, 3285–3290.
237. Y. Kuroda, Y. Yoshikawa, S.-I. Konno, H. Hamano, H. Maeda, R. Kumashiro and M. Nagao, J. Phys. Chem., 1995, 99, 10621–10628.
238. J. Datka, P. Kozyra, E. Kukulska-Zając, M. Szutiak and N. Kumar, Stud. Surf. Sci. Catal., 2004, 154B, 1655–1660.
239. A. I. Serykh and M. D. Amiridis, J. Phys. Chem. C, 2007, 111, 17020–17024.
240. S. Morpurgo, G. Moretti and M. Bossa, Phys. Chem. Chem. Phys., 2007, 9, 417–424.
241. V. Zdravkova, N. Drenchev, E. Ivanova, M. Mihaylov and K. Hadjiivanov, J. Phys. Chem. C, 2015, 119, 15292–15302.
242. G. Frenking, I. Fernández, N. Holzmann, S. Pan, I. Krossing and M. Zhou, JACS Au, 2021, 1, 623–645.
243. E. Mansoor, J. Van der Mynsbrugge, M. Head-Gordon and A. T. Bell, Catal. Today, 2018, 312, 51–65.
244. G. Deplano, M. Signorile, C. Atzori, D. de Salusso, E. Borfecchia, V. Crocella and S. Bordiga, Catal. Today, 2024, 427, 114403.
245. F. A. Cotton and G. Wilkinson, Advanced Inorganic Chemistry, Wiley, New York, 1980.
246. K. Podolska, Interaction of nickel ions in zeolites with small molecules: spectroscopic investigations and molecular modeling, PhD thesis, Jagiellonian University, 2011.
247. P. Pietrzyk, K. Podolska and Z. Sojka, J. Phys. Chem. C, 2011, 115, 13008–13015.
248. P. Pietrzyk, T. Mazur, K. Podolska-Serafin, M. Chiesa and Z. Sojka, J. Am. Chem. Soc., 2013, 135, 15467–15478.
249. K. Hadjiivanov, H. Knozinger and M. Mihaylov, J. Phys. Chem. B, 2002, 106, 2618–2624.
250. M. Mihaylov, O. Lagunov, E. Ivanova and K. Hadjiivanov, Top. Catal., 2011, 54, 308–317.
251. P. Pietrzyk, K. Podolska and Z. Sojka, J. Phys. Chem. A, 2008, 112, 12208–12219.
252. L. Bonnevoit, D. Olivier and M. Che, J. Mol. Catal., 1983, 21, 415.
253. K. Dyrek and M. Che, Chem. Rev., 1997, 97, 305.
254. M. Che, Stud. Surf. Sci. Catal., 1993, 75, 31–68.
255. D. Oliver, M. Richard and M. Che, Chem. Phys. Lett., 1978, 60, 77–80.
256. K. I. Hadjiivanov and G. N. Vayssilov, Adv. Catal., 2002, 47, 307–511.
257. S. Bordiga, C. Lamberti, F. Bonino, A. Travert and F. Thibault-Starzyk, Chem. Soc. Rev., 2015, 44, 7262–7341.
258. X. Deng, D. Yang, W. Li, Y. Chai, G. Wu and L. Li, Trends Chem., 2023, 5, 892–905.
259. G. Busca, V. Lorenzelli and V. Sanchez-Escribano, Chem. Mater., 1992, 4, 595.
260. K. Góra-Marek, A. E. Palomares, A. Glanowska, K. Sadowska and J. Datka, Microporous Mesoporous Mater., 2012, 162, 175–180.
261. I. Malpartida, E. Ivanova, M. Mihaylov, K. Hadjiivanov, V. Blasin-Aubé, O. Marie and M. Daturi, Catal. Today, 2010, 149, 295–303.
262. K. Góra-Marek, K. Brylewska, K. A. Tarach and M. Choi, Dalton Trans., 2015, 44, 8031–8040.
263. M. I. Shilina, T. N. Rostovshchikova, S. A. Nikolaev and O. V. Udalova, Mater. Chem. Phys., 2019, 223, 287–298.
264. S. M. Maier, A. Jentys, E. Metwalli, P. Müller-Buschbaum and J. A. Lercher, J. Phys. Chem. Lett., 2011, 2, 950–955.
265. E. Ivanova, M. Mihaylov, K. Hadjiivanov, V. Blasin-Aubé, O. Marie, A. Plesniar and M. Daturi, Appl. Catal., B, 2010, 93, 325–338.
266. S. S. Goryashenko, Y. K. Park, D. S. Kim and S.-E. Park, Res. Chem. Intermed., 1998, 24, 933–951.
267. M. Saeidi and M. Hamidzadeh, Res. Chem. Intermed., 2017, 43, 2143–2157.
268. N. B. Younes, J. Martínez Ortigosa, O. Marie, T. Blasco and M. Mhamdi, Res. Chem. Intermed., 2021, 47, 2003–2028.
269. Y. Yokomichi, T. Yamabe, T. Kakumoto, O. Okada, H. Ishikawa, Y. Nakamura, H. Kimura and I. Yasuda, Appl. Catal., B, 2000, 28, 1–12.
270. M. Anpo, M. Matsuoka, H. Mishima and H. Yamashita, Res. Chem. Intermed., 1997, 23, 197–217.
271. G. Brezicki, J. Zheng, C. Paolucci, R. Schlögl and R. J. Davis, ACS Catal., 2021, 11, 4973–4987.
272. W. Kaim and B. Schwederski, Coord. Chem. Rev., 2010, 254, 1580–1588.
273. P. Xiao, Y. Wang, Y. Lu, K. Nakamura, N. Ozawa, M. Kubo, H. Gies and T. Yokoi, J. Am. Chem. Soc., 2024, 146, 10014–10022.
274. M. J. Wulfers, S. Teketel, B. Ipek and R. F. Lobo, Chem. Commun., 2015, 51, 4447–4450.
275. A. R. Kulkarni, Z.-J. Zhao, S. Siahrostami, J. K. Nørskov and F. Studt, Catal. Sci. Technol., 2018, 8, 114–123.
276. A. A. Verma, S. A. Bates, T. Anggara, C. Paolucci, A. A. Parekh, K. Kamasamudram, A. Yezerets, J. T. Miller, W. N. Delgass, W. F. Schneider and F. H. Ribeiro, J. Catal., 2014, 312, 179–190.
277. M. Moreno-González, R. Millán, P. Concepción, T. Blasco and M. Boronat, ACS Catal., 2019, 9, 2725–2738.
278. C. Liu, G. Malta, H. Kubota, T. Toyao, Z. Maeno and K.-I. Shimizu, J. Phys. Chem. C, 2021, 125, 21975–21987.
279. X. Guo, R. Zhang, Z. Di, B. Kang, H. Shen, Y. Wei, J. Jia and L. Zheng, Appl. Catal., B, 2024, 343, 123519.
280. J. H. Enemark and R. D. Feltham, Coord. Chem. Rev., 1974, 13, 339–406.
281. W. Kaim and B. Schwederski, Coord. Chem. Rev., 2010, 254, 1580–1588.
282. C. J. Cramer, M. Włoch, P. Piecuch, C. Puzzarini and L. Gagliardi, J. Phys. Chem. A, 2006, 110, 1991–2004.
283. P. Vanelderen, R. G. Hadt, P. J. Smeets, E. I. Solomon, R. A. Schoonheydt and B. F. Sels, J. Catal., 2011, 284, 157–164.
284. J. S. Woertink, P. J. Smeets, M. H. Groothaert, M. A. Vance, B. F. Sels, R. A. Schoonheydt and E. I. Solomon, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 18908–18913.
285. L. Vilella and F. Studt, Eur. J. Inorg. Chem., 2016, 1514–1520.
286. A. Oda, Y. Mamenari, T. Ohkubo and Y. Kuroda, J. Phys. Chem. C, 2019, 123, 17842–17854.
287. I. A. Pankin, A. Martini, K. A. Lomachenko, A. V. Soldatov, S. Bordiga and E. Borfecchia, Catal. Today, 2020, 345, 125–135.
288. G. Yang, L. Zhou, X. Liu, X. Han and X. Bao, Catal. Commun., 2007, 8, 1981–1984.
289. C. J. Cramer and W. B. Tolman, Acc. Chem. Res., 2007, 40, 601–608.
290. K. Sobańska, A. Krasowska, T. Mazur, K. Podolska-Serafin, P. Pietrzyk and Z. Sojka, Top. Catal., 2015, 58, 796–810.
291. A. Oda, H. Torigoe, A. Itadani, T. Ohkubo, T. Yumura, H. Kobayashi and Y. Kuroda, J. Am. Chem. Soc., 2013, 135, 18481–18489.
292. R. F. Howe and J. H. Lunsford, J. Am. Chem. Soc., 1975, 97, 5156–5159.
293. A. Oda, T. Ohkubo, T. Yumura, H. Kobayashi and Y. Kuroda, Angew. Chem., Int. Ed., 2017, 56, 9715–9718.
294. A. Oda, J. Kumagai, K. Sawabe, T. Ohkubo, Y. Kuroda and A. Satsuma, J. Phys. Chem. C, 2021, 125, 5136–5145.
295. P. Pietrzyk and K. Góra-Marek, Phys. Chem. Chem. Phys., 2016, 18, 9490–9496.
296. Z. Sojka, E. Giamello, M. Che, A. Zecchina and K. Dyrek, J. Phys. Chem., 1988, 92, 1541–1547.
297. A. Oda, T. Nanjo, T. Ohkubo and Y. Kuroda, J. Phys. Chem. C, 2020, 124(21), 11544–11557.
298. S. Yao, E. Bill, C. Milsmann, K. Wieghardt and M. A. Driess, Angew. Chem., Int. Ed., 2008, 47, 7110–7113.
299. E. Kimura, R. Machida and M. Kodama, J. Am. Chem. Soc., 1984, 106, 5497–5505.
300. L. Chen, H. Falsig, T. V. W. Janssens and H. Grönbeck, J. Catal., 2018, 358, 179–186.
301. C. Negri, T. Selleri, E. Borfecchia, A. Martini, K. A. Lomachenko, T. V. W. Janssens, M. Cutini, S. Bordiga and G. Berlier, J. Am. Chem. Soc., 2020, 142, 15884–15896.
302. M. H. Mahyuddin, T. Tanaka, A. Staykov, Y. Shiota and K. Yoshizawa, Inorg. Chem., 2018, 57, 10146–10152.
303. H. Li, C. Paolucci, I. Khurana, L. N. Wilcox, F. Göltl, J. D. Albarracin-Caballero, A. J. Shih, F. H. Ribeiro, R. Gounder and W. F. Schneider, Chem. Sci., 2019, 10, 2373–2384.
304. K. Shimizu, R. Maruyama, T. Hatamachi and T. Kodama, J. Phys. Chem. C, 2007, 111, 6440–6446.
305. A. B. Ene, M. Bauer, T. Archipova and E. Roduner, Phys. Chem. Chem. Phys., 2010, 12, 6520–6531.
306. T. Yumura, M. Takeuchi, H. Kobayashi and Y. Kuroda, Inorg. Chem., 2009, 48, 508–517.
307. A. Kornas, E. Tabor, D. K. Wierzbicki, J. E. Olszowka, R. Pilar, J. Dedecek, M. Śliwa, H. Jirglova, S. Sklenak, D. Rutkowska-Żbik and K. Mlekodaj, Appl. Catal., B, 2023, 336, 122915.
308. E. Tabor, M. Lemishka, J. E. Olszowka, K. Mlekodaj, J. Dedecek, P. C. Andrikopoulosk and S. Sklenak, ACS Catal., 2021, 11, 2340–2355.
309. X. Yang, Z. Liu, B. Gao, Z. Chen, K. Yan, S. Wang, Y. Xia, Y. Zhang, L. Wang, X. Xu and Y. Tang, ACS Catal., 2023, 13, 15572–15580.
310. T. V. W. Janssens, E. Borfecchia, K. A. Lomachenko, H. Grönbeck and G. Berlier, ChemCatChem, 2024, 16, e202400384.
311. P. J. Smeets, R. G. Hadt, J. S. Woertink, P. Vanelderen, R. A. Schoonheydt, B. F. Sels and E. I. Solomon, J. Am. Chem. Soc., 2010, 132, 14736–14738.
312. H. M. Rhoda, A. J. Heyer, B. E. R. Snyder, D. Plessers, M. L. Bols, R. A. Schoonheydt, B. F. Sels and E. I. Solomon, Chem. Rev., 2022, 122, 12207–12243.
313. C. Liu, H. Kubota, T. Toyao, Z. Maeno and K. Shimizu, Catal. Sci. Technol., 2020, 10, 3586–3593.
314. H. Lei, D. Chen, J.-Y. Yang, A. Khetan, J. Jiang, B. Peng, U. Simon, D. Ye and P. Chen, Environ. Sci. Technol., 2023, 57, 12465–12475.
315. I. Karamanis, A. Daouli, H. Monnier, M.-A. Dziurla, G. Maurin and M. Badawi, Mol. Syst. Des. Eng., 2023, 8, 1165–1181.
316. P. Pietrzyk, C. Dujardin, K. Góra-Marek, P. Granger and Z. Sojka, Phys. Chem. Chem. Phys., 2012, 14, 2203–2215–144.
317. P. Pietrzyk, F. Zasada, W. Piskorz, A. Kotarba and Z. Sojka, Catal. Today, 2007, 119, 219–227.
318. Past and Present in DeNO_x Catalysis: From Molecular Modeling to Chemical Engineering, in Stud. Surf. Sci. Catal., ed. P. Granger and V. I. Parvulescu, 2007, vol. 171, pp. 27–65.
319. K. Hadjiivanov, Catal. Rev.:Sci. Eng., 2000, 42, 71–144.
320. P. Pietrzyk, W. Piskorz, Z. Sojka and E. Brocławik, J. Phys. Chem. B, 2003, 107, 6105–6113.
321. P. Pietrzyk and Z. Sojka, J. Phys. Chem. A, 2005, 109, 10571–10581.
322. A. M. Volodin, K. A. Dubkov and A. Lund, Chem. Phys. Lett., 2001, 333, 41–44.
323. R. Pérez Vélez, I. Ellmers, H. Huang, U. Bentrup, V. Schünemann, W. Grünert and A. Brückner, J. Catal., 2014, 316, 103–111.
324. S. Malykhin, A. Volodin and G. Zhidomirov, Appl. Magn. Reson., 2008, 33, 153–166.
325. Z. Sojka, P. Pietrzyk, G. Martra, M. Kermarec and M. Che, Catal. Today, 2006, 114, 154–161.
326. C. W. Bauschlicher Jr., S. R. Langhoff and H. J. Partridge, J. Chem. Phys., 1991, 94, 2068–2072.
327. V. Umamaheswari, M. Hartmann and A. Pöppl, J. Phys. Chem. B, 2005, 109, 1537–1546.
328. E. Brocławik, P. Kozyra, M. Mitoraj, M. Radoń and P. Rejmak, Molecules, 2021, 26, 1511.
329. P. Kozyra, M. Radoń, J. Datka and E. Broclawik, Struct. Chem., 2012, 23, 1349–1356.
330. M. Radoń, P. Kozyra, A. Stępniewski, J. Datka and E. Broclawik, Can. J. Chem., 2013, 91, 538–543.
331. K. Góra-Marek, A. Stępniewski, M. Radoń and E. Brocławik, Phys. Chem. Chem. Phys., 2014, 16, 24089–24098.
332. F. Göltl and J. Hafner, J. Chem. Phys., 2012, 136, 064503.
333. P. Rejmak, E. Broclawik, K. Góra-Marek, M. Radoń and J. Datka, J. Phys. Chem. C, 2008, 112, 17998–18010.
334. E. Broclawik, K. Góra-Marek, M. Radoń, M. Bucko and A. Stępniewski, J. Mol. Model., 2017, 23, 160.
335. M. Radoń, Inorg. Chem., 2015, 54, 5634–5645.
336. F. Neese, J. Biol. Inorg. Chem., 2006, 11, 702–711.
337. M. Radoń and E. Broclawik, Electronic properties of iron sites and their active forms in porphyrin-type architectures, in Computational methods to study the structure and dynamics of biomolecules and biomolecular processes – from bioinformatics to molecular quantum mechanics, ed. A. Liwo, Springer International Publishing, Cham, 2019, pp. 755–823.
338. M. Kaupp, J. Comput. Chem., 2007, 28, 320–325.
339. G. Monsch and P. Klüfers, Angew. Chem., Int. Ed., 2019, 58, 8566–8571.
340. N. C. Tomson, M. R. Crimmin, T. Petrenko, L. E. Rosebrugh, S. Sproules, W. C. Boyd, R. G. Bergman, S. DeBeer, F. D. Toste and K. Wieghardt, J. Am. Chem. Soc., 2011, 133, 18785–18801.
341. Q. M. Phung, H. N. Nam, V. Austen, T. Yanai and A. Ghosh, Inorg. Chem., 2025, 64, 1702–1710.
342. P. Karen, Angew. Chem., Int. Ed., 2015, 54, 4716–4726.
343. T. Ampßler, G. Monsch, J. Popp, T. Riggenmann, P. Salvador, D. Schröder and P. Klüfers, Angew. Chem., Int. Ed., 2020, 59, 12381–12386.
344. M. Gimferrer, J. Van der Mynsbrugge, A. T. Bell, P. Salvador and M. Head-Gordon, Inorg. Chem., 2020, 59, 15410–15420.
345. P. Pietrzyk, B. Gil and Z. Sojka, Catal. Today, 2007, 126, 103–111.
346. A. Pulido and P. Nachtigall, Phys. Chem. Chem. Phys., 2009, 11, 1447–1458.
347. M. Radoń and E. Bocławik, J. Phys. Chem. A, 2011, 115, 11761–11774.
348. R. Lin and P. J. Farmer, J. Am. Chem. Soc., 2001, 123, 1143–1150.
349. I. M. Wasser, S. de Vries, P. Moënne-Loccoz, I. Schröder and K. D. Karlin, Chem. Rev., 2002, 102, 1201–1234.
350. P. Xie, T. Pu, G. Aranovich, J. Guo, M. Donohue, A. Kulkarni and C. Wang, Nat. Catal., 2021, 4, 144–156.
351. G. B. Richter-Addo and P. Legzdins, Metal Nitrosyls, Oxford University Press, New York, 1992.
352. K. M. Farhan and P. K. Sajith, Ind. Eng. Chem. Res., 2025, 64, 5257–5268.
353. M.-L. Tsai, R. G. Hadt, P. Vanelderen, B. F. Sels, R. A. Schoonheydt and E. I. Solomon, J. Am. Chem. Soc., 2014, 136, 3522–3529.
354. A. Wijerathne, A. Sawyer, R. Daya and C. Paolucci, J. Am. Chem. Soc. Au, 2024, 4, 197–215.
355. A. R. Fahami, T. Günter, D. E. Doronkin, M. Casapu, D. Zengel, T. H. Vuong, M. Simon, F. Breher, A. V. Kucherov, A. Brückner and J.-D. Grunwaldt, React. Chem. Eng., 2019, 4, 1000–1018.
356. L. Chen, T. V. W. Jansen, P. N. R. Vennestrom, J. Jansson, M. Skoglundh and H. A. Grönbeck, ACS Catal., 2020, 10, 5646–5656.
357. M. Bendrich, A. Scheuer, R. E. Hayes and M. Votsmeier, Appl. Catal., B, 2018, 222, 76–87.
358. W. Hu, T. Selleri, F. Gramigni, E. Fenes, K. R. Rout, S. Liu, I. Nova, D. Chen, X. Gao and E. Tronconi, Angew. Chem., Int. Ed., 2021, 60, 7197–7204.
359. C. Liu, H. Kubota, T. Amada, K. Kon, T. Toyao, Z. Maeno, K. Ueda, J. Ohyama, A. Satsuma, T. Tanigawa, N. Tsunoji, T. Sano and K.-I. Shimizu, ChemCatChem, 2020, 12, 3050–3059.
360. W. Hu, F. Gramigni, N. D. Nasello, N. Usberti, U. Iacobone, S. Liu, I. Nova, X. Gao and E. Tronconi, ACS Catal., 2022, 12, 5263–5274.
361. R. Daya, D. Trandal, U. Menon, D. J. Deka, W. P. Partridge and S. Y. Joshi, ACS Catal., 2022, 12, 6418–6433.
362. C. Liu, H. Kubota, T. Toyao, Z. Maeno and K.-I. Shimizu, Catal. Sci. Technol., 2020, 10, 3586–3593.
363. A. Oda, H. Shionoya, Y. Hotta, T. Takewaki, K. Sawabe and A. Satsuma, ACS Catal., 2020, 10, 12333–12339.
364. C. Paolucci, A. A. Verma, S. A. Bates, V. F. Kispersky, J. T. Miller, R. Gounder, W. N. Delgass, F. H. Ribeiro and W. F. Schneider, Angew. Chem., Int. Ed., 2014, 53, 11828–11833.
365. M. P. Ruggeri, I. Nova, E. Tronconi, J. A. Pihl, T. J. Toops and W. P. Partridge, Appl. Catal., B, 2015, 166–167, 181–192.
366. H.-Y. Chen, M. Kollar, Z. Wei, F. Gao, Y. Wang, J. Szanyi and C. H. F. Peden, Catal. Today, 2019, 320, 61–71.
367. Y. Zhang, Y. Peng, K. Li, S. Liu, J. Chen, J. Li, F. Gao and C. H. F. Peden, ACS Catal., 2019, 9, 6137–6145.
368. L. Negahdar, N. E. Omori, M. G. Quesne, M. D. Frogley, F. Cacho-Nerin, W. Jones, W. T. Price, R. A. Catlow and A. M. Beale, ACS Catal., 2021, 11, 13091–13101.
369. F. Gao, E. D. Walter, M. Kollar, Y. L. Wang, J. Szanyi and C. H. F. Peden, J. Catal., 2014, 319, 1–14.
370. C. Liu, S. Yasumura, T. Toyao, Z. Maeno and K.-I. Shimizu, J. Phys. Chem. C, 2022, 126, 11594–11601.
371. F. Lin, T. Andana, Y. Wu, J. Szanyi, Y. Wanga and F. Gao, J. Catal., 2021, 401, 70–80.
372. G. He, B. Zhang, H. He, X. Chen and Y. Shan, Sci. Total Environ., 2019, 673, 266–271.
373. B. Zhang, G. He, Y. Shan and H. He, Catal. Today, 2019, 327, 177–181.
374. W. B. Tolman, Angew. Chem., Int. Ed., 2010, 49, 1018–1024.
375. V. I. Avdeev and A. F. Bedilo, Chem. Phys. Lett., 2018, 695, 222–227.
376. D. Sengupta, J. B. Adams, W. F. Schneider and K. C. Hass, Catal. Lett., 2001, 74, 193–199.
377. P. Xiao, Y. Wang, Y. Lu, K. Nakamura, N. Ozawa, M. Kubo, H. Gies and T. Yokoi, J. Am. Chem. Soc., 2024, 146, 10014–10022.
378. Z. Sojka and M. Che, J. Phys. Chem., 1996, 100, 14776–14785.
379. L. Nowakowski, C. Hudy, F. Zasada, J. Gryboś, W. Piskorz, A. Wach, Y. Kayser, J. Szlachetko and Z. Sojka, J. Am. Chem. Soc., 2024, 146, 24450–24466.
380. G. Li, E. A. Pidko, I. A. W. Filot, R. A. van Santen, C. Li and E. J. M. Hensen, J. Catal., 2013, 308, 386–397.
381. H. Xia, K. Sun, K. Sun, Z. Feng, W. X. Li and C. Li, J. Phys. Chem. C, 2008, 112, 9001–9005.
382. E. V. Kondratenko and J. Perez-Ramırez, J. Phys. Chem. B, 2006, 110, 22586–22595.
383. C. Sang, B. H. Kim and C. R. F. Lund, J. Phys. Chem. B, 2005, 109, 2295–2301.
384. Z. Sobalik, J. Novakova, J. Dedecek, N. K. Sathu, E. Tabor, P. Sazama, P. Stastny and B. Wichterlova, Microporous Mesoporous Mater., 2011, 146, 172–183.
385. Z. Piwowarska, E. Micek, M. Rutkowska and L. Chmielarz, Microporous Mesoporous Mater., 2015, 209, 54–65.
386. D. Pietrogiacomi, M. C. Campa, L. R. Carbone and M. Occhiuzzi, Appl. Catal., B, 2019, 240, 19–29.
387. K. Jíša, J. Nováková, M. Schwarze, A. Vondrová, S. Sklenák and Z. Sobalik, J. Catal., 2009, 262, 27–34.
388. B. M. Abu-Zied, W. Schwieger and A. Unger, Appl. Catal., B, 2008, 84, 277–288.
389. F. Kapteijn, G. Marban, J. Rodriguez-Mirasol and J. A. Moulijn, J. Catal., 1997, 167, 256–265.
390. A. Li, P. Wang, L. Zhang, H. Zhao, M. Jin and L. Lei, Fuel, 2024, 378, 132888.
391. P. J. Smeets, M. H. Groothaert, R. M. van Teeffelen, H. Leeman, E. J. M. Hensen and R. A. Schoonheydt, J. Catal., 2007, 245, 358–368.
392. B. Kang, R. Zhang, M. Guo, X. Guo, Z. Di, Y. Wei and J. Jia, Energy Fuels, 2023, 37, 18019–18029.
393. J. W. A. Fischer, F. Buttignol, A. Brenig, D. Klose, D. Ferri, V. Sushkevich, J. A. van Bokhoven and G. Jeschke, Catal. Today, 2024, 429, 114503.
394. A. P. Hawkins, A. E. Edmeades, C. D. M. Hutchison, M. Towrie, R. F. Howe, G. M. Greetham and P. M. Donaldson, Chem. Sci., 2024, 15, 3453–3465.
395. K. Mitsuishi, K. Nakazawa, R. Sagawa, M. Shimizu, H. Matsumoto, H. Shima and T. Takewaki, Sci. Rep., 2023, 13, 316.
396. D. Liu, L. Liu, K. Wu, J. Zhou, Q. Cheng, J. Lv, T. Cao, D. Zhang, F. Lin and Y. Han, J. Phys. Chem. C, 2021, 125, 18952–18960.
397. C. H. Hendon, A. J. Rieth, M. D. Korzyński and M. Dinca, ACS Cent. Sci., 2017, 3, 554–563.

---