Sakshi
Satyanand
a,
Sanjana
Srinivas
a,
Dionisios G.
Vlachos
*abc and
Stavros
Caratzoulas
bc
aDepartment of Chemical and Biomolecular Engineering, University of Delaware, 150 Academy St., Newark, DE 19716, USA. E-mail: vlachos@udel.edu
bDelaware Energy Institute, University of Delaware, 221 Academy St., Newark, DE 19716, USA
cCenter for Plastics Innovation, University of Delaware, 221 Academy St., Newark, DE 19716, USA
First published on 5th December 2024
With the increasing interest in developing catalytic materials based on atomically dispersed transition metals on heterogeneous supports, it is necessary to have an atomic-level understanding of the factors that impact their structural and electronic properties and, ultimately, their reactivity. In this contribution, we address and elucidate with electronic structure calculations open questions related to the ethane dehydrogenation mechanism on silica-supported mononuclear Fe(II) and Fe(III) sites. Contrary to prior hypotheses, we determine that the σ-metathesis on Fe(II) sites is an unlikely dehydrogenation mechanism. On tricoordinate and tetracoordinate Fe(II)@SiO2, the reaction proceeds via heterolytic C–H bond activation and β-hydride elimination facilitated by spin-crossing. Atomically dispersed Fe(III) grafted on SiO2 exhibits a more complex behavior as it seems to be undergoing autoreduction and we propose a new redox ethane dehydrogenation mechanism which, remarkably, is energetically competitive with the heterolytic C–H activation mechanism previously identified for other transition metals.
Amorphous silica-supported mononuclear first row 3d-metals M@SiO2 (M = Fe(II),5 Co(II)6–8 and Zn(II)9) have shown promising catalytic activity for ethane dehydrogenation, with Co(II) demonstrating high selectivity. The mechanism on Co(II)@SiO2 has been elucidated by Srinivas et al.,10 who showed that high-spin (quartet) Co d7 catalyzes the C–H activation heterolytically and that the ensuing β-H elimination is facilitated by spin-crossing to the doublet state. By microkinetics simulations, Srinivas et al.10 further showed that the β-H elimination is rate-controlling.
Isolated Fe species grafted on silica or incorporated in the framework of silicates also exhibit interesting catalytic properties. Hu et al.5 synthesized isolated 3-coordinate Fe(II)@SiO2 and demonstrated that it was selective and stable for propane dehydrogenation. The mechanism in this case has not been elucidated, although it has been suggested that the reaction could proceed either via the heterolytic pathway described above or via σ-bond metathesis between the Fe-hydride and an alkane molecule, to form H2 and an Fe-alkyl as in the case of late transition metal complexes.11,12 Isolated Fe(III) species have been incorporated in the matrix of SBA-15 by Nozaki et al.13 and by Cheng et al.14 Site stability of Fe(III) on amorphous silica is challenging due to mobile Fe centers that form FeOx agglomerates, which can be mitigated by deploying bulky ligands as precursors.13,15 Grosso-Giordano et al.15 have further hypothesized that the crystalline spatial arrangement of hydroxyls in siliceous zeolites provides a more favorable environment for enhanced Fe(III) uptake compared to amorphous silica in grafting experiments.15 Interestingly, Lobo and co-workers have suggested a redox mechanism for the dehydrogenation of propane over isomorphously substituted H–[Fe(III)]-ZSM5 (ref. 16 and 17) but to the best of our knowledge the mechanism has not been confirmed. Notably, d4 to d8 metal centers can cleave C–H bonds through various pathways.18
Tuning the coordination environment of surface-bound, well-defined transition metals is critical to the development of new catalysts for C–H activation on heterogeneous supports with well-defined catalyst species. In the case of amorphous silica supports, the challenge is compounded by the absence of structural uniformity.19,20 Characterization methods, such as EXAFS and XANES, provide space-averaged information rather than the local structural features of the grafted site that catalyze the reaction. Not all spectroscopically relevant sites, or the most abundant sites, are necessarily catalytically active. For instance, in the Phillips catalyst (Cr@SiO2), only 10% of the sites are active.21 Co@SiO2 presents a similar case, where small but detectable amounts of Co(III) are present in the as-synthesized Co@SiO2 catalyst.22 This necessitates consideration of other possible sites to explore possible local geometries in mechanistic investigations that may or may not have been picked up in experiments. To make progress, we first need to eliminate ambiguities related to the reaction mechanisms and how they are influenced by the structural properties of the active site.
In this contribution, we focus on ethane-to-ethylene dehydrogenation on various structural models of the active site of silica-supported mononuclear Fe(II) and Fe(III). With the aid of electronic structure calculations, we elucidate open questions related to the dehydrogenation mechanism. We consider tricoordinate and tetracoordinate Fe(II) sites and determine that σ-metathesis is an unlikely dehydrogenation mechanism on Fe(II)@SiO2, regardless of the coordination geometry. On Fe(II)@SiO2, the kinetically favored pathway involves heterolytic C–H activation by a siloxide pair Fe–OSi–, followed by β-hydride elimination facilitated by spin-crossing, akin to Co(II)@SiO2. However, the energetics and dynamics of the spin crossover phenomena during the rate-limiting β-H elimination of supported Fe(II) and Co(II) are different, in contrast to the corresponding alkyl complexes in organometallic chemistry.23,24 We show that Fe(III)@SiO2 exhibits a more complex behavior and present a new redox dehydrogenation mechanism which, energetically, is on a par with the heterolytic mechanism on Co(II)/SiO2.
![]() | ||
Fig. 1 Structural models of the Fe(II)@SiO2 active sites. (a) 4-Coordinate geometry. (b) 3-Coordinate geometry (purple: Fe; red: O; white: H; teal: Si.). |
On both the tricoordinate and tetracoordinate Fe d6 structural models, we tested the dehydrogenation mechanism shown in Fig. 2. The corresponding potential energy profiles and transition states are shown in Fig. 3(a) and (b), respectively. Following the ethane coordination to Fe (2), the C–H bond is activated by the metal center and a –SiO− ligand that abstracts the hydrogen, forming the metal ethyl intermediate 3 and a –SiOH. The ground electronic state remains a quintet, regardless of the Fe′s coordination geometry. However, the 3-coordinate site presents a slightly lower intrinsic C–H activation barrier than the 4-coordinate model by about 0.14 eV (1.68 eV on the 3-coordinate vs. 1.82 eV on the 4-coordinate). The C–H activation is indeed heterolytic as analysis of the Mulliken spin densities ruled out a proton-coupled electron transfer (PCET) mechanism (ESI† Table S1). In a PCET mechanism, the abstracted hydrogen atom would still have ended up on the oxygen atom as a proton, but its electron would have been transferred to the metal center. The Mulliken analysis showed 4 unpaired electrons on Fe d6, indicating no reduction, and zero spin density on the abstracted H and on the carbon atom of 3. The spins on the oxygen ligands remained quenched. The Fe–C bond distance in 53 is 2.0 Å in the 3-coordinate site and 2.06 Å in the 4-coordinate, both in good agreement with the distance of 1.96 Å reported in Hu et al.'s5 EXAFS studies of the as-prepared and H2 pretreated Fe(II)@SiO2 catalyst at 400 °C.
![]() | ||
Fig. 3 Electronic energy profiles for the mechanism shown in Fig. 2 on Fe(II)@SiO2 sites. (a) 3-Coordinate Fe(II) geometry. (b) 4-Coordinate Fe(II) geometry. The profiles correspond to the two lowest-lying spin states, the quintet and the triplet. Relevant intermediates and transition states (in bold) are indicated in the energy profile with spin states in superscripts. Quintet to triplet spin-crossing takes place prior to the β-H elimination transition state. In (b), note the agostic intermediate (and associated transition state 3TSag) preceding the β-H elimination transition state (3TS2) in the triplet spin state. Intermediate 1 is the reference state, with energy equal to the sum of energies of the bare active site and C2H6(g). Intermediates 2–4 are surface bound species. C2H4(g) + H2(g) are the gaseous products of the reaction pathway, and the energy level is obtained by subtracting the DFT energy of C2H4(g) + H2(g) from C2H6(g). Full mechanism including the hydrogen recombination step is presented in Fig. S2† of the accompanying ESI.† |
TSag in Fig. 3 refers to the agostic transition state in the tetra-coordinate site preceding the β-hydride elimination step, and it involves the H2 atom (atom labelling in Fig. 4) coordinated to both the C2H5 moiety and the Fe (see Fig. S1 in the ESI†). Unlike the tetracoordinate site, which is sterically hindered, the tricoordinate site did not present any agostic transition state and intermediate 3 in Fig. 3(a) is already an agostic product. We arrive at this conclusion due to geometric similarities between intermediates 3ag and 3 and evidence from literature.23 However, the agostic intermediate is mechanistically inconsequential as it precedes the spin-crossing event.
![]() | ||
Fig. 4 Transition state geometries obtained in this work. Heterolytic cleavage mechanism: (a) 5TS1, (b) 3TS2 on the tricoordinate site; (c) 5TS1, (d) 3TS2 on the tetracoordinate site; (e) 5TS3 of the metathesis mechanism on the tricoordinate site; redox mechanism: (f) 6TS1 and (g) 4TS2 on the tetracoordinate site. Superscript denotes the spin state, and subscript denotes the reaction number in the corresponding reaction mechanism (heterolytic cleavage: Fig. 2, σ-bond metathesis: Fig. 5 and redox: Fig. 6(b)). Only the active site is shown for clarity (purple: Fe; red: O; white: H; grey: C.). |
As can be seen from the electronic energy profiles in Fig. 3 during the ensuing β-H elimination from the Fe-ethyl intermediate (53), the Fe d6 atom undergoes a spin flip into the triplet state, which lies lower in energy than the quintet. Spin pairing is induced by spin–orbit coupling and empties a metal d-orbital to accommodate the formation of the metal-hydride. It is typical in high-spin complexes and has been reported for the Fe(C2H5)+ cation,46 diketiminate-supported Fe(C2H5) complexes23 and several other systems.47 Because of the spin-crossing, the repulsive three-electron two-orbital interaction in the high-spin state becomes an attractive two-electron two-orbital interaction in the low-spin state, which lowers the β-H elimination barrier (see Fig. 5). Regardless of the coordination geometry of the active site Fe (3-coordinate or 4-coordinate), the triplet transition state assumes square planar geometry, typical of low-spin complexes. As can been seen from Fig. 2, past the β-H elimination transition state, the complex crosses over back to the quintet state, which is the ground electronic state of the metal-hydride (see Fig. 4b and d for β-H elimination transition state on tricoordinate and tetracoordinate site, respectively). Similar spin-crossing events have been reported by Srinivas et al.10 for atomically dispersed Co d7 on silica. Unlike Co, however, spin–orbit coupling in Fe is not strong and this has a couple of implications. First, the low-spin β-H elimination transition state is not significantly lower than the high-spin one compared to the corresponding states in Co(II)@SiO2, explaining the lower activity for small alkane dehydrogenation observed in experiments.5 It must be noted, however, that the quintet-triplet gap is noticeably smaller for the 3-coordinate than the 4-coordinate geometry, presumably because the former affords the system greater flexibility. Second, the dynamics of the β-H elimination is non-adiabatic with a tunneling constant of 0.86, lower than that obtained for Co(II)@SiO2 corresponding to an adiabatic transition.10
At high temperatures, entropic contributions are dominant.48 The entropies of formation of the transition state is negative for both C–H activation steps, in alignment with the dissociative nature of these steps. Free energy barriers are higher than electronic energies (see Fig. S3 in the ESI†). The rate determining step is the β hydride elimination step in the heterolytic cleavage mechanism, the same as that for Co(II)@SiO2. High-spin states result in longer bond lengths due to repulsion within the occupied orbitals, making the transition states more flexible and contributing more to the entropic term. In contrast, low-spin states have tightly bound transition states with shorter bonds, leading to lesser entropic corrections. This leads to the flipping of the spin state order at reaction temperatures in the free energy diagrams, further suggesting the minimal benefit from spin crossover during β-hydride elimination. The hydrogen recombination step has a lower free energy change because the elementary step is an association reaction with partial H-H bond formation.
Instead of the hydrogen recombination (6) completing the cycle, as shown in Fig. 2, it has been proposed5 that the metal-hydride 5 could react with incoming ethane as shown in Fig. 6 (intermediate 7). Through a σ-metathesis mechanism, the metal-ethyl intermediate could then form with concomitant release of H2. This is the general metathesis mechanism for a d6 center and involves discrete σ-complexes undergoing ligand exchange.49 Once the metal-ethyl intermediate has formed, β-H elimination could ensue, as described earlier. We tested this hypothesis12 and were able to optimize the related intermediates and transition states (see Fig. 4e) for the 3-coordinate active site. We were unable, however, to identify the respective intermediates and transition states for the 4-coordinate site, possibly because this geometry does not afford the active site complex sufficient flexibility to accommodate six ligands on the metal center. For the 3-coordinate geometry, the σ-bond metathesis requires significant activation energy, with an overall energy span near 3.5 eV (Fig. S4,† TS3), making this pathway very unlikely compared with the mechanism in Fig. 2. The ligand exchange step presents a high free energy barrier as shown in Fig. S5 of the ESI.†
A redox mechanism has been proposed by Lobo and co-workers17 for the dehydrogenation of propane to propylene on isomorphously substituted H–[Fe(III)]-ZSM5 adapted for ethane the mechanism is depicted in Fig. 7a. In that mechanism, the Fe d5 atom is first reduced by accepting an electron from the bound ethane molecule. Next, the unstable ethane radical (C2H6˙+) releases H2 and the resulting C2H4˙+ radical accepts the electron back from Fe, which returns to the 3+ oxidation state. Although never computationally confirmed to the best of our knowledge, it raises the intriguing possibility of a redox mechanism of C–H activation on atomically dispersed Fe(III) on amorphous SiO2. A C–H bond may in principle be activated by metal-oxygen pairs on transition metal oxides, either through a hydrogen atom transfer (HAT) mechanism or through a PCET mechanism.50,51 In the HAT mechanism, the reaction essentially proceeds through the formation of radicals: an H atom is abstracted by the oxygen of the metal-oxide pair while the alkyl radical coordinates to the metal center. In the PCET mechanism, as described above, the abstracted hydrogen atom ends up on the oxygen atom as a proton while its electron is transferred to the metal center, which is reduced as a result.
![]() | ||
Fig. 7 Dehydrogenation mechanism on Fe(III)@SiO2. (a) Redox mechanism proposed by Yun and Lobo (ref. 16). (b) Revised redox mechanism proposed in this work. In (b), Fe(III) is unstable and spontaneously undergoes reduction to the Fe(II) state while the system. |
We tested the possibility of a redox dehydrogenation mechanism on an Fe(III)@SiO2 site. We used the structural model shown in Fig. 8. In this, the Fe atom was grafted on three SiO− groups, making it formally d5. We considered different spin states for the bare active site and the one with multiplicity 6 was the ground electronic state. Remarkably, analysis of the Mulliken spin density of the bare active site showed that the Fe center had 4 unpaired α-electrons instead of 5, which means that Fe was autoreduced to the d6 state. The spin density values are listed in Table 1 and spin density plots are shown in Fig. S6.†
![]() | ||
Fig. 8 Structural model of the Fe(III)@SiO2 active site. The Fe center is coordinated to 3 silanolate ligands (purple: Fe; red: O; white: H; teal: Si.). |
Species | Fe | O1 | O2 | O3 | C1 | H1/H2 |
---|---|---|---|---|---|---|
61′ | 4.18 | 0.25 | 0.26 | 0.24 | 0 | 0 |
6TS | 4.20 | 0.24 | 0.104 | 0.18 | 0.16 | 0.02 |
43′ | 3.37 | 0.08 | 0.09 | 0 | −0.55 | 0.014 |
4TS | 3.17 | 0.13 | 0 | 0.13 | −0.22 | −0.32 (H2) |
64′ | 4.25 | 0.2 | 0 | 0.21 | 0.08 | −0.22 (H2) |
The spin densities on the oxygen atoms of the ligands showed that the β-electron that transferred to the Fe atom originated from one of the SiO− ligands. In addition, the majority spin density on the silanolate ligands was delocalized over the oxygen atoms of all three of them, conserving the total spin multiplicity of 6. Thus, it appears that Fe d5 is not stable in the environment of three silanolate ligands on amorphous silica. The immediate implication, therefore, is that the dehydrogenation of ethane to ethylene would not proceed via the redox mechanism in Fig. 7a. That notwithstanding, we proceeded to investigate the full dehydrogenation mechanism, curious to explore the mechanistic implications of the delocalized spin density on the SiO− ligands and whether the mechanism would be similar to the one we presented in Fig. 2 for the Fe d6 site with two SiO− ligands. Our calculations showed that the system retained the multiplicity of 6 both upon ethane binding and at the C–H activation transition state (see Fig. 4f). However, we found spin-crossing over to the more stable quartet state upon formation of the Fe-alkyl intermediate; the H atom was accepted by a SiO− ligand with Fe coordinated to the three surface oxygens and the carbon atom of the alkyl group (this is in stark contrast with the mechanism over Fe(II)@SiO2, shown in Fig. 2, where spin-crossing from the quintet to the triplet occurred after the Fe-alkyl intermediate had formed and before the β-H elimination.). In the quartet state of the Fe-alkyl intermediate, the spin density of the Fe atom indicated about 3 unpaired electrons—it was 4 prior to the spin-crossing. In addition, in the quartet state, the spin density on the SiO− oxygen atoms were fully quenched, and the transferred H atom carried no spin density. Taken together, the calculations showed that the C–H bond broke heterolytically in the sextet state and, upon formation of the Fe-alkyl intermediate, the Fe atom was re-oxidized from the d6 to the d5 electronic configuration by donating an electron back to oxygen ligands, quenching their spins. As a result, the Fe atom ended up with an empty d-orbital which prepared the system for the β-H elimination over the quartet transition state (Fig. 4g), as can been seen in Fig. 9. The system returned to the sextet spin state upon formation of the metal-hydride intermediate and ethylene. Based on this analysis, in Fig. 7b, we present the revised redox mechanism.
![]() | ||
Fig. 9 Electronic energy profiles for the redox mechanism shown in Fig. 7(b) on initially prepared Fe(III)@am–SiO2. The profiles correspond to the two lowest-lying spin states with spin multiplicities 6 and 4. Relevant intermediates and transition states (in bold) are indicated in the energy profile with spin states in superscripts. Intermediate 1 is the reference state, with energy equal to the sum of energies of the bare active site and C2H6(g). Intermediates 2–4 are surface bound species. C2H4(g) + H2(g) are gas phase ethylene and hydrogen gas generated as part of the catalytic cycle. Full mechanism including the hydrogen recombination step is presented in Fig. S7 of the accompanying ESI.† The overall reaction is endothermic. Complete free energy profile is shown in Fig. S8 of the ESI.† |
The energy requirements are quite favorable compared with the mechanism on Fe(II)@SiO2. The highest-lying transition state in the redox mechanism is at ca. 1.6 eV and it corresponds to the C–H activation. In the heterolytic mechanism on 3-coordinate Fe(II)@SiO2, the highest-lying transition state is at ca. 2.3 eV and it is associated with the β-H elimination. The 1.6 eV barrier also compares favorably with the 1.5 eV barrier (β-H elimination) calculated by Srinivas et al.10 on Co(II)@SiO2. The rate-determining step in the redox pathway on Fe(III)@SiO2 is the first C–H activation step with a significantly lower energy barrier compared to the heterolytic cleavage pathway, as shown in Fig. S8 of the ESI.† In the redox mechanism, the spin state ordering is conserved in the free energy profile unlike the heterolytic pathway.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4cy01118j |
This journal is © The Royal Society of Chemistry 2025 |