Victor M.
Fernández-Alvarez
a,
Sarah K. Y.
Ho
b,
George J. P.
Britovsek
b and
Feliu
Maseras
*ac
aInstitute of Chemical Research of Catalonia, The Barcelona Institute for Science and Technology, Avgda. Països Catalans, 16, Tarragona 43007, Catalonia, Spain. E-mail: fmaseras@iciq.es; Fax: +34 977 920231; Tel: +34 977 920202
bDepartment of Chemistry, Imperial College London, Exhibition Road, South Kensington, London SW7 2AY, UK
cDepartament de Química, Universitat Autònoma de Barcelona, 08193 Bellaterra, Catalonia, Spain
First published on 4th May 2018
The photocatalyzed insertion of dioxygen into the Pt(II)–methyl bond in terpyridine platinum complexes has been shown to proceed efficiently, but its mechanism remains a challenge. In particular, there are serious counter-intuitive differences in the reactivity of structurally similar complexes. M06 calculations in solvent with a valence double-ζ basis set supplemented by polarization and diffusion shells (benchmarked against ωB97x-D calculations with a larger basis set) are able to provide a satisfactory mechanistic answer. The proposed mechanism starts with the absorption of a photon by the metal complex, which then evolves into a triplet state that reacts with the triplet dioxygen fragment. A variety of possible reaction paths have been identified, some leading to the methylperoxo product and others reverting to the reactants, and the validity of some of these paths has been confirmed by additional experiments. The balance between the barriers towards productive and unproductive paths reproduces the diverging experimental behavior of similar complexes and provides a general mechanistic picture for these processes.
Recently, Britovsek et al.13–15 and Goldberg et al.16–18 reported that platinum methyl complexes can insert dioxygen to give the methylperoxo product. One of the remarkable aspects of the reaction reported by Britovsek was the strong dependence of the insertion on the substituents in the 6 and 6′ positions of the terpyridine ligand (Fig. 1). Whereas the diamino-substituted complex readily inserted dioxygen, replacing one or both groups with hydrogen made the system unreactive towards insertion. However, if the substituents are methyl groups, O2 insertion does take place under reaction conditions. This reactivity is rather counter-intuitive since, unlike in the diamino-substituted case in which hydrogen bond interactions might facilitate oxygen coordination, the methyl groups represent a hindering factor towards the approach of dioxygen. It is for this reason challenging to find a reaction mechanism able to explain why substituents that constitute fundamentally different interactions can both insert dioxygen while the rather innocuous dihydrogen-substituted complex remains unreactive.
Fig. 1 General reaction scheme of the O2 insertion. Reaction occurs for systems 1 and 4. Reaction does not occur for systems 2 and 3. |
To address this challenge we carried out a computational study of the reaction mechanism at the DFT level. Calculations are well established as a valuable tool to determine mechanisms in homogeneous catalysis, often in conjunction with experiments.19–22 Although DFT calculations have been used to describe the photoexcitation process,23,24 they have been seldom applied to light-driven reactions, with only few examples available.25–32 In this work we report a detailed study on the reaction of the 6,6′-diaminoterpyridine Pt(II)methyl complex (1), which is later extended to three other complexes with diverse experimental behaviors.
The electronic distribution of the ground state singlet A is the expected one for a square-planar d8 Pt(II) complex 1, with four d orbitals doubly occupied in the metal. The TD-DFT calculations at the M06 level on this complex showed the existence of a strong band with a maximum absorption at 404 nm (oscillator strength f = 0.1343) and an even stronger one at 364 nm (f = 0.2789), in good agreement with the experimental spectrum of this complex (Fig. S5 in the ESI†).14 Both bands correspond to metal-to-ligand charge transfer (MLCT) transitions, specifically from a metal d orbital to a π* orbital in the disubstituted terpyridine ligand, as shown in Fig. 2. No charge transfer absorption bands were found between dioxygen and the complex when placed together forming an adduct, ruling out the existence of a donor–acceptor complex.25
Fig. 2 Leading orbitals involved in the key absorption bands in complex A. HOMO−3 (left) and LUMO (right). |
The evolution from this transient excited singlet to the long-lived triplet state 3A involves a significant electronic rearrangement. 3A contains most of the spin density on the Pt center and a very small amount on the π system of the terpyridine, in line with previous observations regarding fast interconversions between MCLT and d–d excited states in terpyridine platinum complexes.23,37 Examination of the orbitals of this intermediate points to a high-spin Pt(II) system. The analysis of the electronic structure is complicated by the fact that one of the semi-occupied orbitals is σ-antibonding with respect to the ligands, which results in spin delocalization towards both terpyridine and methyl.
Fig. 3 shows the optimized structure for both the ground state singlet A and the triplet state 3A resulting form the excitation. There is steric repulsion between the methyl ligand and the ortho substituents of the terpyridine ligand, in this case amine groups. This steric repulsion forces the methyl ligand slightly out of the plane even in the ground state A (N–Pt–Me angle, 167.2°). This distortion is significantly increased in the triplet state (N–Pt–Me angle, 117.7°), which also lengthens the Pt–Me bond from 2.08 to 2.12 Å. The increased distortion is associated with the electronic changes when moving from singlet to triplet. The singlet state contains a closed shell d8 platinum center, with an empty d orbital in the molecular plane, and thus a square planar arrangement is electronically favored. The triplet state 3A has a significant amount of spin density, 1.31, located in the platinum center, which therefore is no longer favoring the square planar geometry. The fact that the methyl ligand is entirely outside the molecular plane in 3A leaves a vacant site on the Pt centre for dioxygen to coordinate. 3A has an energy of 40.8 kcal mol−1 (1.77 eV) above A, and is thus only accessible through photoexcitation.
Fig. 3 Optimized geometries of intermediates A (left) and 3A (right). Selected distances in Å and angles in degrees. |
The rest of the manuscript will deal with the evolution of this triplet intermediate 3A. The different possibilities will be discussed in detail in separate sections, but the overall picture can be seen in Fig. 4. Complex 3A reacts with dioxygen to produce intermediate C, which can in turn evolve into product G through two different paths, or revert to the reactants through intermediate 3C.
Fig. 4 Full reaction scheme for complex 1 (R1R2NH2, blue) and key points for complex 2 (R1H, R2NH2, green) and 4 (R1R2CH3, orange). Energies relative to C in kcal mol−1. |
Complex 3A, resulting from the photoexcitation, can react with dioxygen in two major ways, as shown in Fig. 5. In both of them, the overall spin of the sum of the interacting fragments is zero, as the metallic fragment can be seen as bringing two α electrons and the dioxygen bringing two β electrons. The first reaction path leads to intermediate C through intermediate B. In this path, one of the unpaired electrons on dioxygen couples with one unpaired electron at the metal center, resulting in open-shell singlet (OSS) intermediate B. In B one unpaired electron remains in the terminal oxygen and another one in the molecular orbital. This intermediate is 23.9 kcal mol−1 above the ground state reactants. Coordination of O2 to 3A is barrierless in the potential energy surface, but we could estimate an entropic barrier of ca. 8 kcal mol−1 for this step (TS(3A–B), see the ESI† for details). B has an O–O distance of 1.30 Å, which indicates a superoxo ligand.42 The formal oxidation state at the platinum centre is thus Pt(III). A related Pd(III) superoxo dimethyl complex with a dipyridine ligand was proposed by Mirica and could be trapped by the addition of a radical scavenger and then observed by EPR.43B evolves into closed-shell intermediate C in which dioxygen displaces the methyl group to the apical position. C is 17.6 kcal mol−1 above the reactants, and has lower energy than free singlet oxygen (22.7 kcal mol−1), and thus is an energetically favored intermediate. In C dioxygen is coordinated as a side-on peroxyde ligand with an O–O distance of 1.42 Å which is between the superoxo and peroxo distances (1.33 and 1.49 Å, respectively).42 The metal centre in C is formally in a Pt(IV) oxidation state. The formation of this peroxo complex C from Pt(II) complex Avia intermediate B resembles very closely the computational work on the binding of O2 to Pd(0) complexes reported by Landis and Stahl.12,44
Fig. 5 Free energy profile for the reaction between 3A and O2. Energies relative to reactants in kcal mol−1. |
An alternative reaction path starts with the direct interaction of dioxygen with the methyl ligand and leads to intermediate D. This path involves a single electron transfer (SET) in which the Pt–Me bond undergoes homolytic dissociation. We were able to locate the transition state for this single electron transfer process, TS(3A–D),45 which is 53.7 kcal mol−1 above the reactants. The resulting fragments are a tricoordinated Pt(II) complex with a vacant site and methylperoxo radical (D, 19.1 kcal mol−1). Both fragments could combine without a significant barrier to yield the product directly. In any case, this alternative pathway is not feasible because it is not competitive. The direct attack on the metal discussed above is favored, as the barrier is ca. 5 kcal mol−1 lower. As a result, C is formed rather than D and the reaction proceeds from there. For the sake of simplicity, C will be considered as the origin of energies for the species discussed in what follows, as all thermal pathways described below involve this intermediate.
These results favour a monometallic pathway, in contrast to previous suggestions that photoexcitation and oxygen quenching take place through a bimetallic mechanism involving two complex molecules in the form of a dimer.13 Our calculations show that once 3A is formed, coordination of dioxygen takes place without the need for a second complex. We confirmed this conclusion through calculations on the bimolecular mechanism reported in the ESI.†
In the second mechanism, the side-on peroxo ligand rearranges to an end-on configuration in which the terminal oxygen is closer to the methyl group (E, 3.8 kcal mol−1), from which the methyl ligand migrates directly to the terminal oxygen (TS(E–G), 26.6 kcal mol−1) to yield the product. The third path involves dissociation of one “arm” of terpyridine from the metal and connects C directly to G through transition state TS(C–G) (31.3 kcal mol−1). Data in Fig. 6 indicate conclusively that for system 1, the first path through TS(C–F) is clearly favored, with a barrier of 2.8 kcal mol−1 below that of the closer alternative through TS(E–G). The other two paths are nevertheless included here because they become relevant for other complexes that will be discussed below.
Fig. 7 Free energy profile for the reversion from C to the reactants. Energies relative to C in kcal mol−1. |
An alternative path for reversion to the reactants goes through intermediate H. This intermediate is an adduct between singlet oxygen and the reactant A, with an energy of 15.7 kcal mol−1, which is reached through transition state TS(C–H), with an energy of 18.3 kcal mol−1. This step is mostly an intramolecular transfer of an electron pair from the dianionic peroxide ligand to the Pt(IV) center in C. Relaxation from H to the reactants requires the transition from singlet oxygen to triplet oxygen. We estimated the rate for this process from experimental data on radiative and non-radiative decay of singlet oxygen relaxation. Specifically, we used the experimental data of the singlet oxygen relaxation constant in acetonitrile (kAcCN = 1.5 × 104 s−1) to determine an energy barrier for this transformation of restoring ground state oxygen.47 Direct application of the Eyring equation led to an “activation energy” of 11.8 kcal mol−1 above H, which is included in the figure. The competition between the two processes of reversion to the reactants clearly favors the first one, through intermediate 3C with a barrier 4.2 kcal mol−1 lower.
The two lowest energy paths are highly competitive. The difference of 0.5 kcal mol−1 means that 31% of the photoexcited systems should evolve into products, with the other 69% returning to the reactants. This results in a net observation of product formation, in agreement with experiments. Even if two-thirds of the photoexcitations are non-productive, this would represent only a minor reduction in the reaction rate, as the deactivated reactants can be photoexcited again.
As with 1, we began by studying the photophysical behavior of these complexes through TD-DFT calculations. The structures for complexes 3 and 4 before and after the excitation are shown in Fig. 8. The structures corresponding to compound 2 are not shown because of their similarity to those of compound 1. The values for the Pt–N, Pt–C and N–Pt–C parameters are 2.028 and 2.074 Å, and 173.4° for 2-A, and 2.019 and 2.095 Å, and 126.5° for 2-3A. The qualitative nature of the photophysical process is similar: the main structural modification associated with the conversion from the ground state A to the triplet state 3A is the displacement of the methyl group out of the ligand plane. This displacement is well characterized by the N–Pt–C angles shown in Fig. 3 and 8, which are provided in Table 1. It is remarkable that the two compounds which are active, 1 and 4, have the largest distortions in the triplet state. This hints at a possible correlation of this distortion with the reactivity towards dioxygen, but other differences need to be considered as well.
Fig. 8 Optimized geometries of 3-A (top left), 3-3A (top right), 4-A (bottom left) and 4-3A (bottom right). Selected distances in Å and angles in degrees. |
System | N–Pt–C | System | N–Pt–C |
---|---|---|---|
1-A | 167.2 | 1-3A | 117.7 |
2-A | 173.4 | 2-3A | 126.5 |
3-A | 178.2 | 3-3A | 134.3 |
4-A | 165.9 | 4-3A | 117.7 |
The vertical spectrum of 2 showed a single MLCT band with a strong maximum absorption at 357 nm (f = 0.2245). 3 showed only a weak π–π* band at 358 nm (f = 0.0227) and a very strong MLCT band at 316 nm (f = 0.3996). The spectrum of 4 presented a somewhat weak MLCT band at 347 nm (f = 0.0506) and a stronger one at 330 nm (f = 0.3604). Let us recall for comparison that 1 absorbed at 404 nm (f = 0.1343) and 364 nm (f = 0.2789). These results are mostly in agreement with experimentally reported data for these species.13,14 The overall picture is that the maximum MLCT absorption for all these complexes is near the frontier between visible light and ultraviolet, which is often considered to be around 380 nm. There are however significant quantitative differences between them. The highest wavelength efficient absorptions are at 404 nm, 357 nm, 316 nm and 347 nm, respectively, which is in qualitative agreement with the experimental absorption spectra reported in the ESI (Fig. S5†). Calculations and experiments thus predict that complex 3 absorbs at a significantly lower wavelength. Complex 3 should not form the long-lived triplet intermediate under sunlight conditions, which correlates with its lack of reactivity under these reaction conditions. The reason why complex 3 does not absorb can be found in the structure of the ground state for this compound shown in Fig. 8. The complex is much closer to the ideal square planar geometry, with a N–Pt–C angle of 178.2°, because of the small steric repulsions between the methyl ligand and the ortho substituents at the terpyridine ligands, in this case hydrogen atoms. This planarity has the important consequence of stabilizing the non-bonding occupied d orbitals in the d8 platinum center. These orbitals are the origin of the photoexcitation (see Fig. 2), and their stabilization necessarily increases the photoexcitation energy.
In the case of complex 2, our calculations show that the energy difference between TS(C–F) and CP(C–3C) increases by 2.5 kcal mol−1 compared to 1. The reversion to the reactants is now favored by 3.0 kcal mol−1, which corresponds to an efficiency of 1:99. 99% of the photoexcitations are thus unproductive. The reaction would certainly take place given sufficient time, but with a much lower reaction rate. The result is thus compatible with the experimental observation that this system does not react. Inspection of Fig. 4 indicates that the difference between systems 1 and 2 is in the energies of the CP(C–3C) crossing-point, as those of the TS(C–F) set are very similar. The difference is likely related to the hydrogen bonds formed by the NH2 substituents. The arrangement of the two oxygens in both C and TS(C–F) is rather symmetrical, close to the Me–Pt–O plane, which favors the formation of two hydrogen bonds, one with each NH2 group. In contrast, in 2-CP(C–3C), the dioxygen is significantly bent towards one side, an optimal arrangement for a single hydrogen bond, as shown in the left-hand side of Fig. 9. CP(C–3C) is thus favored in system 2 with only one NH2 group. As CP(C–3C) leads to reversion to the reactants, this system is significantly less reactive.
The case of complex 4 presents some differences. In this case, the evolution of products through TS(C–F) (29.1 kcal mol−1) is clearly disfavored with respect to reversion to the reactants through CP(C–3C) (25.1 kcal mol−1). This is likely associated with the presence of the methyl substituents, which bring steric pressure without any option for hydrogen bonds. The weaker coordinating labile behaviour of the 6-methyl substituted pyridine ligand donors compared to a non-substituted pyridine donor due to steric reasons has been commented on previously,48 and this is well documented, and was also encountered during synthesis of the palladium(II) analogue of complex 4. Preparation of this complex from ligand L, [PdClMe(cod)] and AgSbF6 in acetone resulted initially in a bis(acetone) complex, [PdCl(L)(acetone)2]+, with non-coordinating 6-methyl pyridine moieties (see the ESI†). Conversion to the palladium analogue of complex 4, where the 6-methylpyridine moieties are coordinated, was achieved only after prolonged heating in acetone or in acetonitrile. The subsequent reaction of this palladium methyl analogue with oxygen resulted in the formation of a rather unstable methylperoxo complex, probably due to the lack of any stabilising hydrogen bonding interactions. The key result here is that the increased steric repulsion brings into the picture TS(C–G), which proceeds through partial decoordination of the terpyridine ligand. In this case, TS(C–G), with a free energy of 23.5 kcal mol−1, is the most favored channel out of C, thus explaining the efficiency of this system. The structure for TS(C–G) in system 4 is shown in the right-hand side of Fig. 9, where the release of steric pressure resulting from the cleavage of one Pt–N bond (distance of 3.117 Å) is apparent.
Our calculations thus reproduce the experimental behavior. Moreover, the four systems studied happen to share the same general picture presented in Fig. 4, but present significant differences due to the quantitative changes associated with different processes. System 3 does not react because it is unable to absorb in the visible or near UV light range. System 2 does not react because the deactivation pathway through CP(C–3C) is favored in the presence of a single hydrogen bond with substituents in the ligand. System 1 does react because a pair of hydrogen bonds is sufficient to make the forward reaction through transition state TS(C–F) efficient. System 4 does react despite the lack of hydrogen bonds because of the steric release associated with partial decoordination of one arm of the ligand through transition state TS(C–G).
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c8sc01161c |
This journal is © The Royal Society of Chemistry 2018 |