Elucidating the Oxygen-Activation Mechanism in Nonheme MnII-, FeII-, or CoII-Containing MOFs Mimicking FeII/2-Oxoglutarate-Dependent Complexes Authors: Ziyue Huang, Yingqi Li, Xiaotian Zhang, Xi Chen, Jiawei Xu, Haiyan Wei Published: 2026 ABSTRACT A Fe/Zn metal–organic framework (MOF) that mimics the FeII/2-oxoglutarate-dependent complex (FeII/2-OGDC) was reported previously, but its oxygen-activation mechanism remains unclear. In the present research, density functional theory (DFT) calculations were used to study this Fe-based system and its Mn- and Co-substituted analogues. We found that all three systems follow the general oxygen-activation chemistry of FeII/2-OGDC but show clear metal-dependent differences. The Fe and Co systems generate high-spin TMIV–oxo complexes, whereas the Mn system forms high-spin MnIII-oxyl complexes. The energy profiles further show that the formation of the key oxidizing intermediate is more favorable in the Fe and Mn systems than in the Co system. Overall, this work explains how changing the metal center affects oxygen activation in MOF-based enzyme mimics and highlights Mn as a more promising substitute for Fe than Co. Introduction Triplet oxygen (3O2) is the most accessible oxidant in nature. Although thermochemically favored, the reaction kinetics between 3O2 and most organic substrates are strongly hindered by spin-forbidden constraints. − Most natural solutions highlight the importance of transition metals in biological oxygen activation, where iron is the most common O2-binding site. Typical examples include, but are not limited to, the cytochrome P450 (CYP450) family, − FeII/2-oxoglutarate-dependent complex (FeII/2-OGDC), − Rieske family, , etc. Comprehensive reviews have summarized the general catalytic cycle, structural features, and reactivity patterns of FeII/2-OGDCs. Recent computational studies on nonheme FeII/2-OGDC have provided important mechanistic insights into dioxygen activation, generally supporting the involvement of FeIII-superoxo/peroxysuccinate-type intermediates and a high-valent FeIV–oxo oxidant. These studies also suggested that the detailed reactivity can be influenced by the coordination mode of 2-OG, substrate conformation, and structural dynamics. − Together with earlier theoretical work on related nonheme iron systems, these studies have established an important mechanistic framework for understanding spin-state changes, O–O bond activation, and the formation of high-valent iron–oxo species. Much effort has been devoted to extending the catalytic ability of Fe-containing enzymes, especially by mimicking their structure and reactivity in material systems. The CYP450 family uses a planar Fe–N4 single-atom site as the O2-binding position. Due to the stable, rigid, and graphene-like coordination structure of the Fe-porphyrin cofactor, the Fe–N4 single-atom site is well developed and characterized in single-atom catalysts (SAC) embedded in carbon materials, showing similar C–H hydroxylation catalytic capability in the presence of oxygen. − However, in other Fe-containing kinases, instead of a stable cofactor structure, Fe (or an iron–sulfur cluster) is coordinated with the solvent and amino acids, forming a flexible coordination structure, awhich makes it difficult to design and reconstruct similar single-atom sites in material systems. − It is worth noting that a recent study by Hou et al. reported the first successful example of building FeII/2-OGDC-like sites in metal–organic framework (MOF) systems, which realized the reactivity of oxygen activation, C–H hydroxylation toward cyclohexane, and selective ethanol oxidation. In biological systems, FeII/2-OGDC utilizes oxygen and 2-oxoglutarate (2-OG; also known as α-ketoglutarate, α-KG) to catalyze redox reactions toward a broad range of substrates, typically through dioxygen activation to generate a high-valent FeIV–oxo oxidant, showing monooxygenase or dehydrogenase reactivity. − In the resting state, the FeII/2-OGDC family uses Fe2+ as a catalytic center, which is usually coordinated with solvent water and three neighboring histidine (HIS) or aspartate (ASP) residues. The general catalytic mechanism is shown in Figure a, where a high-valent FeIV–oxo complex is responsible for C–H activation via the hydrogen atom transfer (HAT) process, followed by hydroxyl rebound to give a hydroxylation product (oxygenase reactivity) or another HAT to give a desaturation product (dehydrogenase reactivity). The MOF systems designed by Hou et al. mimicking a FeII/2-OGDC single-atom site have a slightly different coordination environment with three coordinated 1,2,3-triazole ligands embedded in the frameworks (Figure b), where pyruvate (prv) and 3,3-dimethyl-2-oxobutyrate (moba) were used to replace and take over the function of the 2-OG ligand. However, the detailed oxygen-activation mechanism of this MOF system remains unclear, and it is unknown whether analogous Mn- and Co-substituted systems can support similar FeII/2-OGDC reactivity. (a) Active site structure and oxygen-activation mechanism of the FeII/2-OGDC family showing monooxygenase and dehydrogenase reactivities. (b) Structure of FeZn4 L 4(btdd)3 MOFs (H2btdd = bis­(1H-1,2,3-triazolo [4,5-b],[4′,5′-i])­dibenzo­[1,4]­dioxin; L = prv or moba) mimicking the FeII/2-OGDC family. Biocatalysis is well-known to have high efficiency and selectivity; however, compared with the material system, it still has some practical drawbacks before massive production and industrial application, such as sensitivity to the environment, high cost, limited substrates, etc. Therefore, a theoretical understanding of the rarely reported material systems mimicking nonheme-type Fe-containing enzymes will be of great importance for future designs. Clarifying these issues is important for establishing metal-dependent design principles for MOF-based mimics of nonheme oxygenases. In this work, our theoretical results reveal how triplet oxygen is utilized by the FeZn4(prv)4(btdd)3 MOF system mimicking the FeII/2-OGDC structure and reactivity. Methodology To simulate the MOF structure, a cluster model of FeZn4(prv)4(btdd)3 was built. All geometric optimizations and frequency calculations were performed under the PBE0-D3­(BJ)/def2-SVP level of theory , using the Gaussian16 (revision C.01) program. PBE0-D3­(BJ) was selected because dispersion-corrected hybrid functionals are widely used in mechanistic studies of transition-metal systems and can provide a reasonable description of localized metal-centered electronic structures. We note that the absolute spin-state gaps and barriers may depend on the exchange-correlation functional. Therefore, our discussion focuses on qualitative mechanistic trends supported by structural, spin density, UNO/LOBA, and other analyses. The thermal correction to the Gibbs free energy at 298 K (G corr) was obtained at the same level. The IEF-PCM implicit solvent model was adopted to describe the aqueous environment. − The IEF-PCM continuum model cannot explicitly describe pore-confined solvent organization, solvent configurational entropy, or strong local solvent–MOF interactions such as hydrogen bonding. Nevertheless, because all spin states and metal analogues were treated using the same cluster model and solvation protocol, the calculated free energies should be adequate for comparing the qualitative trends across different spin states and metal centers, although the absolute barriers may be affected by explicit solvent effects. To obtain the precise electronic energy (E ele), single-point calculations were performed based on the optimized structure under the PBE0-D3­(BJ)/def2-TZVP level of theory. Potential energy surface (PES) scans were performed to locate the TS structures using the optimized reactant complex (RC) by scanning the bond coordinates, and then the structures corresponding to the maximum energy points in the scans were optimized using the Berny algorithm. Optimized TS structures were subjected to intrinsic reaction coordinate (IRC) calculations to confirm that the reactants and products could be connected through the TS while moving in the forward and backward pathways. The minimum energy crossing point (MECP) was optimized using the branching plane updating algorithm (BPUPD) with the XMECP code, with the energy and gradient obtained under the same level of theory as the optimization of the minimum point. Multiwfn (development version 3.8) was used for wave function analysis, and the Visual Molecular Dynamics (VMD, version 1.9.4) program was used for visualization. A limited gCP , BSSE analysis for representative competing Mn barriers is provided in Table S6, showing that BSSE has a negligible effect on the relative barrier trends relevant to the mechanistic conclusions. To model the local reaction environment of the MOF active site, a cluster model was constructed by truncating the framework around the metal center of the parent MOF structure. The model retains the complete first coordination sphere of the metal and the chemically relevant nearby framework fragments surrounding the reactive site, including the prv ligand and atoms directly involved in oxygen activation. Truncation was performed at positions remote from the reactive center, and the resulting dangling bonds were saturated with hydrogen atoms. Because the present MOF is an open-framework system, oxygen activation chemistry is expected to be governed primarily by the local coordination environment of the metal center. Although such a cluster model cannot fully describe the long-range framework effects or the full flexibility of the extended MOF lattice, it is appropriate for comparing the Fe, Mn, and Co analogues at the mechanistic level, where the key bond-breaking and -forming events occur locally at the metal site. Results and Discussion Oxygen Activation by FeZn4(prv)4(btdd)3 Iron–Dioxygen Complex All possible spin-state combinations of FeII and 3O2 are listed in Figure S1, which may lead to triplet, quintet, and septet states, noted in Figure as 3 IM1, 5 IM1, and 7 IM1, respectively. Computational results showed that 7 IM1 has a long Fe–O distance of 3.31 Å (Figure c), which shows that oxygen is bound to the active site by noncovalent interactions. This weak interaction is also supported by the IGMH analysis, as shown in Figure S4. The spin density difference (SDD) population numbers showed that FeII and 3O2 have parallel populations of 3.86 and 2.00 in 7 IM1, where the electron exchange interaction enhanced its stability to the lowest spin state. Therefore, 7 IM1 was chosen as the zero-energy level for further calculations. Nevertheless, the spin-parallel states of FeII and 3O2 are not reactive states for further single-electron transfer (SET) because all unpaired electrons are parallel to each other, as pointed out by previous calculations on the FeII/2-OGDC family. − Calculated structure, electron configuration, and singly occupied orbitals of possible spin states of iron–oxygen complexes formed in the active site of FeZn4(prv)4(btdd)3. (a) 3 IM1 (S = 1, 2.3 kcal·mol–1), (b) 5 IM1 (S = 2, 12.9 kcal·mol–1), and (c) 7 IM1 (S = 3, 0.0 kcal·mol–1). For other possible spin-states, 3 IM1 and 5 IM1 are 2.3 and 12.9 kcal·mol–1 higher than 7 IM1. Both 3 IM1 and 5 IM1 adopt an end-on coordination configuration (Figure b,c), where the Fe–O distance is reduced to 2.60 and 2.15 Å in 3 IM1 and 5 IM1, respectively, indicating the formation of a σFe–O bond upon adsorption. Compared with noncovalent adsorption (1.19 Å), the O–O distance does not change much during covalent adsorption (1.19 Å in 3 IM1 and 1.20 Å in 5 IM1). This indicates that SET has not been completed in any spin state of IM1, as the O–O bond retains its feature in the free state, giving the formal oxidation state of the FeII–O2 complex. The spin-state 3 IM1 is thermochemically more favored compared to that of 5 IM1. As can be seen from the SDD population numbers (Figure b), the antiferromagnetic coupling caused by opposite SDD populations on the FeII center (+3.85 e) and oxygen (−0.96 and −1.01 e) in 3 IM1 contributes to its stability. Possible alternative oxygen-activation pathways were also considered. In addition to the explicitly calculated spin states and coupling alternatives, several formally possible routes were excluded on chemical grounds. A side-on η2-O2 binding mode would require two adjacent open coordination sites or substantial disruption of the prv/triazole coordination environment. An Op attack on the α-keto C2 atom would disrupt the M–O interaction, whereas an Od attack would preserve the metal–dioxygen connection and provide a direct route for coupled C–O bond formation, electron transfer, and decarboxylation. The attack at the carboxylate C1 atom is also unlikely because it does not naturally lead to C1–C2 oxidative cleavage or TM–oxo/oxyl formation. Finally, direct O–O cleavage before prv participation was not considered chemically viable because the formation of a high-valent TM–oxo/oxyl species requires electron supply from 2-oxoacid oxidative decarboxylation. Therefore, the present pathway search focuses on the chemically relevant Od attack/decarboxylation mechanism across accessible spin-state surfaces. Multispin-State Oxygen Activation For clarity, Op denotes the oxygen atom coordinated to the metal center, while Od denotes the distal oxygen involved in the attack on the C2 atom of prv. The labels C1 and C2 are shown in Figure a. Despite the noncovalent oxygen adsorption structure (7 IM1 Fe), 3 IM1 Fe is the most favored initial adsorption structure for oxygen activation. Two spin states were located for the triplet potential energy surfaces, as marked with suffixes A and B in Figure a. Pathway A indicates that the unpaired electrons on the Fe and O2 parts have opposite spins to form antiferromagnetic coupling, while pathway B indicates no antiferromagnetic coupling. This was confirmed by the SDD population numbers (Table S1). (a) Multispin-state Gibbs free energy diagram of oxygen activation by FeZn4(prv)4(btdd)3 (energy in kcal·mol–1 relative to 7 IM1 Fe). (b) Structure of the active site and spin density difference. (c) Electron transfer route of the favored reaction pathway. As revealed by the potential energy surface (Figure a), on the triplet surface, the Gibbs free energy barriers for the distal oxygen (Od) attacking the C2 atom of the prv ligand are 28.0 kcal·mol–1 (3 TS1A Fe) and 36.6 kcal·mol–1 (3 TS1B Fe) relative to 3 IM1 Fe. The antiferromagnetic coupling pathway has a lower barrier than the pathway without antiferromagnetic coupling, although the barrier is still too high to be overcome at room temperature, and the following transition states (3 TS2A/B Fe and 3 TS3) have lower barriers. Therefore, triplet states are not the desired spin states for the oxygen-activation process. This is consistent with the conclusions reached in biochemical systems. Considering that the final FeIV–oxo complex adopts a quintet spin state, spin crossing is crucial to the whole pathway, and the minimum energy crossing point (MECP) should be located before reaching TS1 Fe to lead the spin-state transition from 3 IM1 Fe to 5 TS1A/B Fe. As shown in Figure a, this MECP is 3.5 kcal·mol–1 higher than 3 IM1 Fe, which is available with only thermal activation at room temperature. On the quintet potential energy surfaces, the Gibbs free energy barriers of 5 TS1A Fe and 5 TS1B Fe are 3.5 kcal·mol–1 and 18.9 kcal·mol–1, respectively, showing that both quintet spin states are available at room temperature, and pathway A with antiferromagnetic coupling is more favored. The productive Fe pathway in the present MOF model is somewhat higher in energy than those reported for several nonheme Fe enzyme systems. However, this increase is still moderate and remains within a chemically accessible range. This is reasonable because the present system is a more robust artificial MOF mimic, whereas enzyme active sites are more finely tuned by the surrounding protein environment. As shown in Figure b, in the pathway 3 A, the rate-determining transition state is closer to its product intermediates (the one next to this transition state), 3 IM2A Fe, in terms of structural parameters. Specifically, in 3 TS1A Fe, the key reaction coordinate R(Od–C2) is 1.64 Å, which is close to 1.55 Å in its product intermediate 3 IM2A Fe, showing that the Od–C2 bond has almost been formed at the transition state, while R(Od–C2) = 3.22 Å shows no interaction in reactant intermediate 3 IM1. Besides, the Fe–Op coordination bond is strengthened during this step, as 3 TS1A Fe has a shorter R(Fe–Op) = 2.25 Å compared with 2.60 Å in 3 IM1 Fe, which is closer to 2.44 Å in 3 IM2A Fe. R(Op–Od) is the same case, with Op–Od being elongated from 1.19 Å in 3 IM1 Fe to 1.27 Å in 3 TS1A Fe, and finally, reaching 1.28 Å in 3 IM2A Fe. The progressive O–O activation is also supported by the Mayer Bond Order (MBO) analysis. The Op–Od MBO decreases from 1.13 in 3TS1AFe to 0.57 in 3TS3Fe, whereas the Fe–Op bond order increases from 0.25 to 1.03, consistent with O–O bond weakening coupled with the Fe–O bond formation (Table S4). All these structural parameters indicate that 3 TS1A Fe resembles the structure of product intermediate 3 IM2A Fe, which, according to the Hammond’s postulate, , is a “late transition state” and corresponds to a highly endergonic reaction. Nevertheless, compared with the “early transition state” 5 TS1A Fe, which resembles its reactant intermediate 5 IM1 Fe, 3 TS1A Fe gives a much higher barrier of 28.0 kcal·mol–1 than 5 TS1A Fe does (3.5 kcal·mol–1). As for 3 TS2A Fe and 5 TS2A Fe, both transition states resemble their reactant intermediates (3 IM2A Fe and 5 IM2A Fe), which explains why 3 TS2A Fe and 5 TS2A Fe have similar low Gibbs free energy barriers of 9.2 and 5.0 kcal·mol–1, respectively. We may naturally ask why 5 TS1A Fe resembles 5 IM1 Fe, and why 3 TS1A Fe does not. The electronic structure provides an answer to this question. As shown in Figure b, the less-favored spin state of 5 IM1 Fe can be described as high-spin FeII and low-spin close-shell 1O2, instead of 3O2 in 3 IM1 Fe or 7 IM1 Fe, which is the reason for its high energy level. 5 TS1A Fe resembles 5 IM1 Fe in structure but not in electronic structure. As shown in Figure b, although the SDD population shows almost no net spin density difference on both oxygen atoms in 5 TS1A Fe, the isosurface still presents a notable distribution of both spin-up (shown in white) and spin-down (shown in blue) densities, showing the shape of a pair of orthogonal π* orbitals and indicating the electronic state of open-shell 1O2. The pathway from 3 IM1 Fe to 5 TS1A Fe via MECP can therefore be concluded as intersystem crossing from the 3O2 state to the open-shell 1O2 state with one unpaired electron flipped. The structural variation on the 5 A potential energy surface, including 3 IM1 Fe and MECP, is accompanied by electronic structure relaxation, resulting in a slow energy increase. However, in the 3 A potential energy surface, although 3 TS1A Fe highly resembles the structure of 3 IM2A Fe, the electronic structure relaxation is notably delayed compared with the geometry. The SDD population number of Fe atoms merely changes from 3.84 e in 3 IM1 Fe to 3.70 e in 3 TS1A Fe, which is far from the targeted 2.99 e in 3 IM2A Fe, which is a medium-spin FeIII. Distorted geometry, but without a fully relaxed electronic structure, increases the energy significantly, leading to a high Gibbs free energy barrier of 3 TS1A Fe. Another important discovery in 3 TS1A Fe and 5 TS1A Fe is revealed by the electronic structure. Generally, the process of Od-attacking the C2 atom can be divided into two parts: (1) addition of the C2 atom to form a C–O bond and (2) electron transfer from FeII to the π* of O2. This feature can be classified as bond formation coupled electron transfer (BFCET), and it is the same for both 3 TS1A Fe and 5 TS1A Fe. Both localized molecular orbitals and LOBA analysis (Table S3) show the feature of FeIII in 3 IM2A Fe and 5 IM2A Fe. The difference between 3 TS1A Fe and 5 TS1A Fe is that, in 3 TS1A Fe, the bond formation is already complete because its structure resembles that of 3 IM2A Fe, while the electron has not yet been transferred. This should be further noted as an asynchronous BFCET, where the electron transfer is triggered by bond formation. As shown in Figure b, the SDD isosurface of 3 TS1A Fe has a notable distribution on the σ-framework of the prv ligand, especially including the C1–C2 σ orbital. This indicates that in 3 TS1A Fe, the C1–C2 σ bond has a radical feature, although it is reserved in 3 IM2A Fe. Bond formation in 3 TS1A Fe is therefore a radical attack process, in which the C1–C2 σ bond acts as an electron transfer bridge. This can be described as shown in Figure S2a using the arrow-pushing scheme, where the red arrows show the radical attack and the blue arrows show the electron transfer route from FeII to the prv ligand. However, the opposite is true for 5 TS1A Fe, where electron transfer has already been completed, resulting in a high-spin FeIII in the transition state, while bond formation has yet to occur, as indicated by the long R(Od–C2) distance of 2.54 Å. The SDD distribution is not observed on the prv ligand, and electron transfer only occurs between FeII and O2. The C–O bond formation should be characterized as nucleophilic addition, as indicated by the arrow-pushing scheme in Figure S2b, discarding the possibility of a radical attack. Thus, 5 TS1A Fe should also be noted as an asynchronous BFCET, but nucleophilic addition is triggered by electron transfer, which activates 3O2 to be in the superoxide form with higher nucleophilicity. After generating IM2A Fe, the 3 A and 5 A pathways share a similar electron transfer route in the following reactions. The Fe atom adopts a medium spin state in the 3 A pathway and a high-spin state in the 5 A pathway, with an energy gap of 8.2∼16.4 kcal·mol–1 between the spin states, indicating that the two spin states are well separated and both pathways are spin-conservative. This also holds true for 3 B and 5 B. The Fe atom adopts a low-spin state in the 3 B pathway and a medium-spin state in the 5 B pathway, as shown in Table S1. FeIII in IM2A Fe is first reduced to FeII in IM3A Fe, as verified by LOBA analysis (Table S3). The electronic state of TS2A Fe should be described as radical decarboxylation, as the SDD distribution is observed on the C1–C2 σ orbital, regardless of the different spin states (Figure b). Electron transfer is triggered by the superoxide radical, which is reduced to peroxide by receiving one electron from the C1–C2 σ orbital and leads to the heterolysis of the C1–C2 σ bond. This makes the carboxyl group require one more electron to fulfill the structure of CO2, which is provided by the heterolysis of the FeIII–O11 bond, and FeIII is reduced to FeII at the same time. The electron transfer route is shown in Figure c. If we consider FeIII and superoxide motifs as a whole, this transition state can be considered a generalized reductive elimination model that adopts a radical mechanism. The generation of 5 IM4 Fe via 5 TS3 Fe has a Gibbs free energy barrier of 10.3 kcal·mol–1, which can be easily overcome at room temperature. The SDD isosurface shows no radical feature in 5 TS3 Fe, which confirms that the transition state is a 2-electron transfer, converting FeII to FeIV along with the cleavage of the peroxide bond. The final reactive oxygen species (ROS) IM4 Fe, also known as Cpd I Fe in biochemical references, has a quintet ground state, which is 13.6 kcal·mol–1 lower than the triplet state. Here, we also added the “Fe” suffix to distinguish it from the Mn- and Co-containing analogues. This result is identical to that observed for nonheme Fe-containing enzymes. The SDD isosurface of 5 IM4 Fe presents an antibonding feature with a node plane between the Fe and O atoms, as well as σ-symmetry, which corresponds to the combination of two Fe–O π* orbitals (π xz * and π yz *). The Fe atom has a 3.17 SDD population number, which is close to 3 unpaired α electrons, one of which can be attributed to the π* orbitals, while the other two unpaired electrons are located on the 3d orbitals of FeIV, holding a high-spin d4-configuration. This electronic state is similar to that reported for the nonheme FeIV–oxo complex. The FeIV–oxo bond length is a key factor to reflect the reactivity of Cpd I Fe. Reported experimental 3FeIV–oxo bond lengths lie within 1.60∼1.67 Å, while nonheme 5FeIV–oxo has a more flexible coordinating environment and may extend this range to 1.58∼1.68 Å. As observed by in situ DRIFTS analysis of FeZn4(prv)4(btdd)3, the absorption of 5FeIV–oxo bond stretching appears at 831 cm–1. In recently reported bond length data collected by multiple spectroscopic techniques, the synthesized Cpd I Fe model with a 1.634 Å bond length gives a Raman signal at 811 cm–1, while the Cpd IFe model with a 1.65 Å bond length gives an NRVS signal at 809 cm–1. This suggests the FeIV–oxo bond length in FeZn4(prv)4(btdd)3 should be shorter, which is consistent with the 1.60 Å obtained by our DFT calculations. Oxygen Activation by TMZn4(prv)4(btdd)3 (TM = Mn or Co) Transition metals, including Mn and Cu, which have flexible oxidation states and more unpaired electrons in the 3d orbitals, may have unique capabilities toward various redox reactions and are commonly involved in nonheme types of metalloenzymes as redox centers. For oxygen activation, Mn, Fe, and Cu are popular candidates among first-row transition metals. However, Cu-catalyzed oxygen activation usually requires multinuclear sites, which are difficult to design and reconstruct in synthesized catalysts. Therefore, biomimetic catalysts mainly focus on Mn- and Fe-containing mononuclear active sites. With one less electron but a similar ionic radius, replacing the FeII center with MnII should have little influence on the MOF structure; however, the reactivity of MnZn4(prv)4(btdd)3 is unclear. Although CoII-containing enzymes are rarely reported to catalyze oxygen activation, we also included CoZn4(prv)4(btdd)3 to investigate the reactivity with one more electron in the 3d orbitals. Therefore, in the following sections, we present the DFT results of the oxygen-activation process catalyzed by Mn and Co-replaced MOFs, to compare them with the FeZn4(prv)4(btdd)3 MOFs. MnZn4(prv)4(btdd)3 Replacing FeII with MnII in the resting state gives MnZn4(prv)4(btdd)3. As shown in Figure a, the final ROS adopts a quartet ground state (4 IM4 Mn), which is 11.0 kcal·mol–1 lower in Gibbs free energy than its sextet state (6 IM4 Mn) and 39.1 kcal/mol lower than its doublet state (2 IM4 Mn). From the doublet state, the pathway was excluded from consideration due to the prohibitively high energies of both intermediates (>29 kcal·mol–1) and transition states (>40 kcal·mol–1). Considering that the antiferromagnetically coupled combination of the high-spin d configuration of MnII and 3O2 also has a quartet ground state, the oxygen activation by MnZn4(prv)4(btdd)3 has one spin-conservative pathway on the quartet potential energy surface, as presented by the black line in Figure a. (a) Multispin-state Gibbs free energy diagram of oxygen activation by MnZn4(prv)4(btdd)3 (energy in kcal·mol–1 relative to 4 IM1). (b) Structure of the active site and spin density difference. (c) Electron transfer route of the favored reaction pathway. Similar to what has been discussed in Fe-catalyzed oxygen activation, the Mn-catalyzed 4 A pathway (high-spin Mn, with antiferromagnetic coupling) has a much lower energy compared with the 4 B pathway (medium-spin Mn, without antiferromagnetic coupling). Besides, the sextet potential energy surface provides a lower pathway involving high-spin Mn without antiferromagnetic coupling. Generally, the 4 A pathway has the lowest Gibbs free energy barrier of 18.2 kcal·mol–1 at 4 TS1A Mn and is the most favored reaction pathway. Sharing the same high-spin Mn center but without antiferromagnetic coupling, the sextet pathway has a close but higher energy compared to the quartet state, and the rate-determining step is 6 TS3 Mn, with an energy barrier of 22.0 kcal·mol–1. Specifically, in pathways 4 A and 4 B, the Gibbs free energy barriers for the distal oxygen (Od) attacking the C2 atom of the prv ligand reach 18.2 kcal·mol–1 (4 TS1A Mn) and 30.4 kcal·mol–1 (4 TS1B Mn). The first step is the rate-limiting step. Structurally, the Od–C2 bond length in the quartet transition-state 4 TS1A Mn shortens from 3.15 Å (in 4 IM1 Mn) to 2.05 Å, and finally forms a covalent bond (1.54 Å) to generate the product intermediate 4 IM2A Mn. During this step, the Mn–Op coordination bond is strengthened, which is similar to the case of the Fe-catalyzed system. The difference is that compared with 5 TS1A Fe, 4 TS1A Mn has a longer Mn–Op bond (4 TS1A Mn: 2.24 Å; 5 TS1A Fe: 2.03 Å), indicating that the Mn–Op coordination bond is weaker than the Fe–Op coordination bond. The following is the C1–C2 bond cleavage: C1–C2 bond dissociation in 4 IM2A Mn (via 4 TS2A Mn to 4 IM3 Mn) has a low barrier (10 kcal·mol–1); in contrast, cleavage in 4B (via 4 TS2B Mn to 4 IM3 Mn) features a higher barrier (24.3 kcal·mol–1) and exergonicity (33.6 kcal·mol–1). As shown in Figure b, the structural parameters correspond to the elongation and cleavage of the C1–C2 and Op–Od bonds. The SDD isosurface reveals an important difference between the Mn- and Fe-catalyzed pathways. In the two displayed Fe-catalyzed pathways (Figure b), the triplet and quintet electronic states differ at the spin state of the Fe-center; therefore, the medium-spin Fe (triplet state) is highly unfavorable in terms of energy, i.e., the triplet and quintet states are separated by a notable energy gap. In such a case, intersystem crossing from the initial triplet state to the productive quintet state is necessary for oxygen activation, and also methods that may enhance intersystem crossing can be applied to Fe-catalyzed systems to help increase their reactivity. In Mn-catalyzed pathways, especially for TS1, IM2, and TS2, the quartet and sextet states are close in energy because both electronic states have the same high-spin Mn and only differ with or without antiferromagnetic coupling between the Mn center and ligand, as shown in Figure b. In most mononuclear metallic complexes, antiferromagnetic coupling only slightly decreases the energy. However, intersystem crossing between these two states is not favored due to the electron spin flip without a change in the orbital angular momentum, as pointed out by El-Sayed’s rule. Thus, intersystem crossing and MECP are not involved in Mn-catalyzed pathways, even if the potential energy surfaces are close in energy. Based on the above discussion, we only need to focus on the most favored 4 A pathway when discussing the electronic states along the oxygen-activation process. The generation of 4 IM2A Mn via 4 TS1A Mn shares the same electron transfer route as 5 TS1A Fe, where one electron is transferred from the 3d orbital of the metallic center to the π* orbital of 3O2, and the lone pair of electrons of Od attacks the C2 atom to form 4 IM2A Mn, as depicted in Figure c. The Mn atom in 4 IM2A Mn has an SDD population number of 4.03 e, behaving as a high-spin d4 configuration of MnIII, which is consistent with the one-electron transfer mechanism. However, while the Fe-catalyzed 5 TS2A Fe leads to radical decarboxylation (Figure c), no SDD distribution is observed on the C1–C2 bond in either 4 TS2A Mn or 6 TS2 Mn, which indicates that the C1–C2 bond undergoes heterolysis during the decarboxylation process. The electron transfer route should therefore be described as shown in Figure c, where one electron is transferred from the 3d orbital of the Mn atom to the O–O π* orbital, and the Mn atom receives another pair of electrons from the carboxyl and is reduced to MnII. After the formation of the peroxide intermediate 4 IM3 Mn, the Mn-catalyzed peroxide cleavage shares a similar heterolysis mechanism with the Fe-catalyzed reaction, as shown in Figure b, and the SDD isosurface is only observed on the Mn atom in both 4 IM3 Mn and 4 TS3 Mn. However, the Mn atom in 4 IM4 Mn has an SDD population number of 4.01 e, which corresponds to the high-spin d4 configuration of MnIII, while another unpaired electron is mainly located on the O atom, with an SDD population number of −0.95 e. This demonstrates that the electronic state of 4 IM4 Mn (or 4 Cpd I Mn) should be described as a MnIII–oxyl complex instead of a MnIV–oxo complex, which is also the case for the sextet state (6 IM4 Mn). CoZn4(prv)4(btdd)3 In biochemical systems, Mn and Fe are common metallic centers used for oxygen activation, while Co is rarely reported to participate in redox reactions, although three oxidation states (+1, + 2, and +3) are biochemically available in vitamin B12-dependent enzymes to mediate radical reactions. One possible reason is that Co is not as abundant as Mn or Fe in nature; however, this phenomenon may also be related to the reactivity of Co-containing complexes. Considering that CoII has an ionic radius similar to that of Mn and Fe, we therefore replaced FeII with CoII in the resting state, which has one more electron and gives CoZn4(prv)4(btdd)3 as a model to investigate the oxygen-activation reactivity of the nonheme CoII center. The Gibbs free energy diagram for the Co-catalyzed pathway is displayed in Figure a. For the heme-like Co center in vitamin B12, the macrocyclic ligand usually generates a strong ligand field and prefers a low-spin metallic center. , Nonheme Co-containing enzymes have not yet been discovered in nature, and the preference for spin states should follow that of the Fe-containing systems to be in the high-spin resting state. This is also verified by the DFT results, which also showed that CoII in the resting state has a high-spin d7-configuration. The noncovalent adsorption structure of 3O2 will then take a quartet state (S = 3/2), as shown in Figure a, and this further explains that covalent adsorption at the doublet state (S = 1/2) will have higher energy (6.6 kcal·mol–1), in the form of a CoIII–superoxo complex after SET from CoII to 3O2. (a) Multispin-state Gibbs free energy diagram of oxygen activation by CoZn4(prv)4(btdd)3 (energy in kcal·mol–1 relative to 4 IM1). (b) Structure of the active site and spin density difference. (c) Electron transfer route of the favored reaction pathway. The reaction pathway on the quartet state has Gibbs free energy barriers of 24.0 kcal·mol–1 at 4 TS1A Co and 23.0 kcal·mol–1 at 4 TS1B Co, which are the rate-determining steps of the quartet pathway. There are two distinct reaction mechanisms: for 4A, the distance between the Od–C2 gradually shortens, followed by the formation of a new Od–C2 bond via the transition-state 4 TS1A Co to yield intermediate 4 IM2A Co; for 4B, the nucleophilic attack of Od on the C2 atom occurs concurrently with the elongation of the C1–C2 bond, triggering decarboxylation and directly forming intermediate 4IM3Co. In the doublet state, compared to 4 IM1 Co, the Co–Op and Od–C2 distances are shortened to 2.22 and 2.95 Å, respectively, suggesting an enhanced propensity for the nucleophilic attack. Our subsequent computational results further confirm this observation. Two reaction pathways emerge depending on whether antiferromagnetic coupling occurs between Co and O2: pathway A (with coupling) and pathway B (without coupling). The spin density difference population numbers are shown in Table S1. The potential energy surface clearly shows that all transition states and intermediates in 2B are higher in energy than those in 2A; additionally, the doublet pathway has a much lower barrier of 5.0 kcal·mol–1 at 2 TS1A Co. The O–O cleavage step (TS3 Co) on either the doublet or quartet state has a similar barrier of 23.6 kcal·mol–1 or 23.8 kcal·mol–1, which indicates that the O–O cleavage transition state is not influenced by the spin state of the Co center and is a nonradical bond cleavage. This is also verified by the SDD distribution in Figure b, as no radical feature is observed for 2 TS3 Co and 4 TS3 Co, which is consistent with the previously discussed Fe- or Mn-catalyzed pathways. 2 TS3 Co also acts as the rate-determining step in the doublet pathway. Comparing the barriers on all three metallic centers discussed in this work, we should consider that the Co center has the least reactivity toward oxygen activation. The reactive oxygen species IM4 Co generated by CoZn4(prv)4(btdd)3 has the high-spin-state 4 IM4 Co, which is 12.1 kcal·mol–1 lower in energy compared with the low-spin state 2 IM4 Co. As revealed by the SDD population numbers, the unpaired electron of the low-spin state is mainly localized on the oxygen atom (1.06 e), corresponding to the electronic state of the CoIII–oxyl complex. The Co–oxygen complex is less commonly reported compared with other early transition metals; however, the CoIII–oxyl electronic state has already been verified in experiments with macrocyclic ligands such as porphyrin, corrin, or their derivatives. The comparison between the heme-type iron center and the nonheme-type iron centers shows that, at the nonheme site, a flexible coordinating environment with a weaker ligand field tends to yield a high-spin state at the metal center. Based on our calculations on CoZn4(prv)4(btdd)3, the conclusion can be extended to Co-centered systems where the high-spin state is the ground state for the Co–oxygen complexes and thus gives the CoIV–oxo complex electronic state, with the SDD population number reaching 2.79 e on the Co atom. Although oxygen activation by CoZn4(prv)4(btdd)3 has a higher barrier and can be difficult to realize at room temperature, there are many studies that indicate that such a process is available with light excitation. A typical example is the photoinduced degradation of EDTACoII in the presence of peroxymonosulfate (PMS), a systematically controllable model system for the simultaneous removal of transition metals and organic ligands via advanced oxidation processes (AOPs), which is a hot topic in environmental chemistry. Although the EDTACoII/PMS/light system exhibited high efficiency, none of the reactions were observed to degrade without light excitation, providing indirect evidence that oxygen activation by the Co center is not available via thermal activation in a nonheme-type ligand environment, even if a more reactive oxidant, such as PMS, is provided. On the basis of our results for the CoZn4(prv)4(btdd)3 system, we can also predict the existence of a CoIV–oxo intermediate in the light-induced oxidative degradation of EDTACoII. Mayer bond order analysis of representative transition states (Table S4) further shows different extents of M–Op bond formation and Op–Od bond activation in the Fe, Mn, and Co systems. From 4 TS1ACo to 4 TS3 Co, the Co–Op MBO increases from 0.30 to 1.04, whereas the Op–Od MBO decreases from 1.18 to 0.43 (Table S4), indicating progressive Co–O bond formation and O–O bond activation. Together with the Fe and Mn results, these MBO changes are consistent with the metal-dependent transition-state characteristics and mechanistic differences. Structure and Reactivity of TM–Oxo/Oxyl Complexes In low-spin cases, the TM–oxo complex is not available for d6-configuration, or further, i.e., the M–O π bond cannot be formed, which is known as the “oxo-wall” phenomenon. − Considering MOL5-type TM–oxo complexes with ideal C 4v symmetry as a typical example, according to ligand field theory, orbital interactions between d orbitals and 2p orbital of the oxo ligand are displayed in Figure a. The M–O σ bond is formed between the d z2 and p z orbitals, generating a low-lying bonding orbital, usually noted as σ z 2 orbital (or simply σ z ) and a corresponding antibonding σ z 2 * orbital (or simply σ z *). If formed, two M–O π bonds are perpendicular to each other in the xz and yz planes, composed of d xz and 2p x orbitals or d yz and 2p y orbitals, respectively, which correspond to two pairs of degenerated π and π* orbitals. In addition, the d xy and d x2–y2 orbitals are predominantly nonbonding with respect to the axial oxygen. In oxo complexes, the σ z 2 , π xz , and π yz orbitals are fully occupied by the 2p electrons of the oxo ligand, and 3d electrons of the transition-metal center fill the higher energy levels, as marked in the box in Figure a. For low-spin cases, the M–O triple bond can be formed by the d0- to d2-configuration with the nonbonding d xy becoming populated. Since the d3-configuration starts to occupy the M–O π xz * and π yz * orbitals and M–O π bond is weakened. For the d6-configuration and further, the M–O π bond is no longer formed with both π xz * and π yz * orbitals doubly occupied. (a) General orbital interaction scheme in the C 4v ligand field for MOL5-type the oxo complex and (b) orbital energy level crossing in the high-spin MnIII, MnIV, FeIV, and CoIV–oxo/–oxyl complexes. The most studied and applied TM–oxo complex is the FeIV–oxo complex (d4-configuration), which has been proven to be crucial for the redox reactivity of many Fe-containing enzymes and related catalysts. Both triplet and quintet states are available for the FeIV–oxo complex. With a strong field macrocyclic ligand like porphyrin, Cpd I Fe,P450 has a triplet FeIV center with a doubly occupied d xy orbital, and π xz * and singly occupied π yz * orbitals. This has been extensively discussed in P450-related catalytic systems. However, for the nonheme FeIV–oxo complex, a high-spin quintet state is usually observed as the ground state, and the electron configuration therefore follows, as shown in Figure b. This has been demonstrated in the FeII/2-OGDC-related catalytic systems. Here, the unrestricted natural orbitals (UNOs) of 5 IM4 Fe in Figure a are sorted according to the natural orbital occupation number (NOON) and are listed in Figure S3b. It can be seen that the d xy orbital has NOON of 1.9, which is close to that of the doubly occupied. Mixed with 3d and 2p orbitals, HOMO and HOMO-1 show π symmetry and have nearly degenerated NOON of 1.0 and 1.0, corresponding to the π xz * and π yz * levels, respectively. This further demonstrates that the electronic structure of 5 IM4 Fe fits well with the bonding scheme shown in Figure b, sharing the same form of the FeIV–oxo complex as Cpd I Fe,P450. Several studies have demonstrated the stability of MnV–oxo complexes (d2-configuration), − and according to the “oxo-wall” phenomenon, the stability of the MnIV–oxo complex (d3-configuration) should be between those of MnV–oxo and FeIV–oxo complexes. However, for 4 IM4 Mn in Figure a, the resonance structure of the MnIII-oxyl complex is observed instead of the MnIV–oxo complex, as discussed in previous sections. The UNOs of 4 IM4 Mn are sorted according to the NOON in Figure S3a, and the bonding scheme is summarized in Figure b. Compared with the standard MOL5 model (or FeIV–oxo complex), the antiferromagnetic ground state is observed for 4 IM4 Mn, and π xz and π yz orbitals are asymmetrically occupied with three electrons, where, according to the Jahn–Teller effect, degenerated orbitals are further split to decrease the system energy, as indicated by the red connecting lines. In parallel, the π xz * and π yz * orbitals are no longer degenerated. In this case, the 2p x/y orbitals of the O atom and 3d xz/yz orbitals of Mn are not coupled and behave as localized orbitals. According to the “oxo-wall”, for low-spin cases, CoIII with a d6-configuration cannot form a CoIII,LS–oxo complex, as both the π xz * and π yz * orbitals are fully occupied. Instead, the CoIII,LS–oxyl complex is the preferred electronic state and can be generated by providing peroxides as oxidants, which has been proven by both experimental and theoretical studies in vitamin B12-related systems. The corrin ligand in vitamin B12 has limited space for the CoIII center, so that stronger metal–ligand interactions usually lead to a low-spin electron configuration. Li et al. and Akogun et al. sucessfully detected the CoIII,LS–oxyl complex by EPR. As shown in Figure b, in the doublet state, 2 IM4 Co is observed to be a CoIII,LS–oxyl complex, as the SDD isosurface is only distributed on the 2p orbital of the O atom, while the rest of the distribution on the Co center presents a higher angular momentum component and should be viewed as numerical results brought by the extended basis applied in calculations. However, the CoIII,HS–oxyl complex is unstable if formed, as the singly occupied σ z 2 * orbital leads to a weakened Co–O σ bond. Meanwhile, the CoIV,HS–oxo complex is possible as the π xz * and π yz * orbitals are both singly occupied. For 4 IM4 Co in Figure b, UNOs present a similar bonding scheme to that of the FeIV–oxo complex with one more electron populating on the d xy orbital (Figure b). This is an important difference between Co and Fe. By adjusting the coordinating environment, both the low- and high-spin states of Fe prefer the FeIV–oxo complex instead of the FeIII–oxyl complex. However, with one more electron, Co shows different preferences and so distinct reactivities on different spin states, indicating that potential efforts can be made to explore the multistate redox reactivity of the synthesized Co-containing catalysts. The difference in the bonding schemes of Mn, Fe, and Co can be understood from ligand field theory. Considering the high-spin TM–oxo complex form of the three metallic centers, the ionic radius decreases from Mn, Fe to Co, leading to stronger metal–ligand interactions, and a larger ligand field splitting energy, which makes the energy of the d xy orbital comparatively lower, yet d x 2–y 2 , π xz *, and π yz * orbitals become comparatively higher, as indicated in Figure b. Assuming that the MnIV–oxo complex exists for MnZn4(prv)4(btdd)3, the smaller d-π energy gap brought by the lower d x 2–y 2 , π xz *, and π yz * orbitals compared with FeIV enhances the ligand-to-metal electron transfer. The high redox potential of the MnIV center may further help this, which finally leads to π → d SET and gives the MnIII–oxyl complex as the ground electronic state. Moreover, since the loss of one electron on the π orbitals results in one unpaired electron, the unsymmetrically occupied π xz and π yz orbitals lead to further splitting of the π orbitals and π* orbitals, as pointed out by the Jahn–Teller effect. This explains the unique bonding scheme of the MnIII–oxyl complex displayed in Figure b. Another minor difference for the CoIII,HS–oxo complex is that the d xy orbital is lower than the π orbitals due to the larger ligand field splitting energy. For the high-spin d5-configuration, the quintet state is usually available for a smaller splitting energy. Conclusion The application of triplet oxygen as an oxidant in catalysis is challenging due to its low reactivity toward most closed-shell organic substrates. Biological solutions have been successful in constructing different transition-metal-catalyzed oxygen-activation pathways, and biomimetic materials have provided much insight into biological processes. On the basis of the recently reported structure of FeZn4(prv)4(btdd)3 MOFs, which reconstructs the active site of FeII/2-OGDC enzymes, we investigated the oxygen activation reactivity of FeZn4(prv)4(btdd)3 using DFT calculations and explored the possibility of replacing FeII with MnII or CoII. Computational results showed that FeZn4(prv)4(btdd)3 shares the same oxygen-activation mechanism as FeII/2-OGDC enzymes, applying a quintet state FeIV–oxo complex (5 IM4 Fe, or Cpd I Fe) as a reactive oxygen species. Similarly, a CoIV–oxo complex is also formed in the high-spin quartet state of 4 IM4 Co (or Cpd I Co), with one more electron occupying the d x2‑y2 orbital compared with the FeIV–oxo complex. However, for 4 IM4 Mn (or Cpd I Mn), the preferred electronic state is the MnIII-oxyl complex instead of the MnIV–oxo complex due to the Jahn–Teller effect caused by ligand-to-metal single-electron transfer. The oxygen-activation mechanism of TMZn4(prv)4(btdd)3 (TM = Mn, Fe, or Co) was also proven to be the same as that of FeII/2-OGDC enzymes. Reaction energy profiles obtained by DFT calculations also suggested that the generation of Cpd I Mn and Cpd I Fe species is thermodynamically favored at room temperature, while the generation of Cpd I Co has a higher Gibbs free energy barrier and is therefore not available. This indicates that replacing Fe with Mn may result in a similar catalytic performance, while Co may not. The present results show that the different reactivities of the Fe, Mn, and Co systems are not determined by barrier heights alone but arise from the combined effects of spin-state accessibility, metal–oxygen bonding, and O–O activation ability. Fe and Co favor high-spin TMIV–oxo species, whereas Mn stabilizes a high-spin MnIII–oxyl species, reflecting clear metal-dependent differences in the electronic structure. Meanwhile, the lower oxygen-activation barriers in the Fe and Mn systems than in the Co analogue indicate that an effective FeII/2-OGDC-like catalyst should provide both accessible reactive spin states and a favorable balance between metal–oxygen bond formation and O–O bond activation. This mechanistic picture offers useful design insights for developing MOF-based biomimetic oxidation catalysts with alternative transition metals. Supplementary Material