Ruthenium hydrides encapsulated in sol–gel glasses exhibit new ultrafast vibrational dynamics

The dynamics of molecules in the first solvation shell surrounding a catalyst are key to enabling reactants to reach a transition state and form products.^1–4 These time-dependent solvent–solute interactions stabilize or destabilize the transition state, thereby directly influencing the rate coefficient.^5–9 It is for precisely this reason that catalytic rate constants are almost always solvent specific.^10–13 The random, thermally driven motions of solvent molecules effectively allow the catalyst (and substrate) to sample the energetic landscape that spans the reaction coordinate. In biomacromolecules, for example, key structural motions for the reactivity of enzymes have been tied to the dynamics of the surrounding solvation layer.^14–19 Ultrafast dynamics in the solvation shell have also been implicated in mechanistic steps for organometallic catalysts.^20,21

In addition to changing the identity of the solvent itself, the reactivity of catalysts can also be controlled by encapsulation in small volumes of solvent inside solid support matrices, such as zeolites or sol–gel glasses.^22–26 Alumina sol–gels have been shown to increase the catalytic activity of homogeneous catalysts entrapped inside of the sol–gel pores.^27–29 In a recent study, it was shown that encapsulation of Ru₃(CO)₁₂ in alumina sol–gels increased its catalytic activity for hydrogenation reactions.²⁸ It was proposed that the improved activity originated from the formation of catalytically active metal hydrides during the sol–gel aging process;²⁹ metal hydrides have a long history in chemical catalysis.^30–32 In a subsequent study, our group demonstrated definitively that the ruthenium hydride {[HRu₃(CO)₁₁]⁻} was indeed formed within the alumina sol–gel glass with the concomitant loss of a CO ligand;³³ however, we did not explore the role that solvent dynamics in nanoscopic pores might play in the catalytic activity.

To address this question in this study, infrared pump–probe spectroscopy and two-dimensional infrared (2D-IR) spectroscopy were performed on Ru₃(CO)₁₂ and [NEt₄][HRu₃(CO)₁₁] in tetrahydrofuran (THF) and [HRu₃(CO)₁₁]⁻ encapsulated in an alumina sol–gel. We explore how the chemical conversion to a metal hydride and the environmental encapsulation in an alumina sol–gel affect vibrational dynamics reported by the CO ligands.

Ru₃(CO)₁₂, aluminum isopropoxide [Al(O-i-Pr)₃], and tetrahydrofuran (THF) used for alumina sol–gel studies were purchased from Millipore-Sigma and used as received. Sodium borohydride (NaBH₄) used in the air- and moisture-sensitive synthesis of [NEt₄][HRu₃(CO)₁₁] was purchased from Millipore-Sigma and dried in vacuo prior to use. The solvents used in the synthesis of [NEt₄][HRu₃(CO)₁₁] (THF, pentane, dichloromethane, and Et₂O) were dried through activated alumina on a Pure Process Technology solvent purification system.

Ru₃(CO)₁₂ alumina sol–gels were prepared according to a previous procedure.²⁸ Approximately 0.02 mmol of Ru₃(CO)₁₂ was dissolved in 1 ml of THF and the solution was stirred. Separately, ∼6 mmol of Al(O-i-Pr)₃ was dissolved in 4 ml of THF, which was then decanted into the stirring Ru₃(CO)₁₂ solution. The mixture was stirred for 20 min. For the FTIR, pump–probe, and 2D-IR measurements, the Ru₃(CO)₁₂/Al(O-i-Pr)₃ mixture was sandwiched between two calcium fluoride (CaF₂) windows with a 50 μm spacer that had two small strips cut out of the spacer (one on top and one at the bottom) to allow for the introduction of water. The sandwiched sol–gel sample was then placed in a homemade solvent cell built in such a way that water surrounded the sample edges. This allowed water to diffuse into the Ru₃(CO)₁₂/Al(O-i-Pr)₃ mixture and catalyze the alumina sol–gel formation.

In a glovebox, Ru₃(CO)₁₂ (106 mg, 0.166 mmol), NaBH₄ (33 mg, 0.87 mmol), and 17 ml of THF were added to a 20 ml scintillation vial along with a small stir bar. The reaction vial was then sealed with a Teflon screw cap and stirred for 40 min at room temperature. [Et₄N]Br (42 mg, 0.20 mmol) was then added to the solution, and the reaction mixture was stirred for another 1 h. The solution was filtered through a pipet plug and concentrated to 2 ml in vacuum. 15 ml of pentane was added to the solution to form a precipitate. The resulting suspension was filtered and the precipitate was washed with 2 ml of pentane. The precipitate was then dissolved in CH₂Cl₂, and crystallization of [NEt₄][HRu₃(C)₁₁] was achieved by vapor diffusion of Et₂O into CH₂Cl₂, yielding [NEt₄][HRu₃(C)₁₁] as reddish-brown crystals. The final yield was 95 mg (0.13 mmol, 77%). For characterization details, see Ref. 33.

All FTIR spectra were collected using a Nicolet 6700 FTIR spectrometer (Thermo Scientific). The resolution was 1 cm⁻¹ at 16 scans. The spectrum of Ru₃(CO)₁₂ in THF was collected with THF as the background. The spectra of Ru₃(CO)₁₂ entrapped in the alumina sol–gel were collected with air as the background. In cases where “ultradry THF” was used for spectroscopy, THF was taken from a Pure Process Technology solvent purification system and stored over 4 Å sieves in a glovebox. Solvent dryness was measured through a Na/Ph₂CO ketyl radical titration, with a maximum threshold of ∼10 ppm H₂O.

The 2D-IR setup has been described previously.^8,9 Briefly, a regeneratively amplified Ti:sapphire laser (Spectra-Physics, 800 nm, 40 fs pulse duration FWHM, 30 nm bandwidth FWHM, 600 mW) pumped an optical parametric amplifier (OPA) (Spectra-Physics) with a repetition rate of 1 kHz. The near-IR signal and idler beams generated by the OPA’s β-barium borate (BBO, 3 mm thick) crystal were difference frequency mixed in a silver gallium sulfide crystal (AgGaS₂, 0.5 mm thick), generating 3 µJ mid-IR pulses. The final output pulses had a spectrum that was centered at 2040 cm⁻¹ (THF sample) or 2020 cm⁻¹ (sol–gel sample). The bandwidth was ∼200 cm⁻¹ (FWHM). The mid-IR pulse train was split into three separate beams and focused onto the sample in the BOXCARS geometry.¹⁰ The generated vibrational echo was overlapped with a local oscillator and sent to a spectrometer with a spectral resolution of 4 cm⁻¹ and detected with a liquid nitrogen cooled 64-pixel mercury cadmium telluride (MCT) linear array detector (Infrared Associates, Inc.). The entire mid-IR beam path was purged with dry air (−100 °F dew point).

The FTIR spectrum of Ru₃(CO)₁₂ in THF shows three strong peaks at 2005, 2030, and 2060 cm⁻¹ (Fig. 1). These peaks have been assigned as the E′ (radial), A₂′ (axial), and E′ (axial) carbonyl modes on the ruthenium molecule.^33–36 Sol–gel formation results in the complex losing one carbon monoxide ligand. Drastic changes to the spectral changes occur with new peaks at 1951, 1961, 1987, and 2014 cm⁻¹. The synthesized version of the hydride, [NEt₄][HRu₃(CO)₁₁], has an FTIR spectrum that is identical to that of the sol–gel (Fig. 1), confirming the formation of a hydride within the nanoscopic alumina pores.

One way that encapsulation could affect catalyst reactivity is by perturbing the coupling to the solvent bath, which then impacts the rate of vibrational energy relaxation (VER).^37–41 We performed IR pump–probe measurements to compare the relaxation dynamics of Ru₃(CO)₁₂ in THF, [HRu₃(CO)₁₁]⁻ in an alumina sol–gel glass, and [NEt₄][HRu₃(CO)₁₁] in THF. The congested nature of the FTIR spectra in Fig. 1 led to convoluted pump–probe spectra for all three samples. To avoid artifacts, the blue edges of the highest energy carbonyl modes (∼2065 or ∼2019 cm⁻¹) in the pump–probe spectra were analyzed.

Selected population decays for all three samples were fit to biexponential decay functions as shown in Fig. 2 (fit parameters tabulated in Table S1). The fast time constants are on the order of 1 ps in all cases, and we attribute this to fast intramolecular redistribution (IVR). This was confirmed by directly measuring IVR occurring on the same time scale by analyzing the 2D-IR spectra as shown in the supplementary material (Fig. S1). The most salient difference between the dynamics in Fig. 2 is that [HRu₃(CO)₁₁]⁻ in the sol–gel (T₁ = 30 ± 8 ps) and in THF (T₁ = 37 ± 2 ps) exhibits drastically shorter lifetimes than Ru₃(CO)₁₂ (T₁ = 144 ± 15 ps). This difference suggests that the hydride (encapsulated or synthesized) couples to its environment differently than Ru₃(CO)₁₂. This difference cannot be attributed to small amounts of water that are formed by alumina hydrolysis during sol–gel synthesis since both Ru₃(CO)₁₂ and synthesized [HRu₃(CO)₁₁]⁻ were measured in ultradry THF and still reflect this difference. We have previously shown that the formation of the hydride places additional electron density on the Ru nuclei. The Ru nuclei then participate in π-backbonding and place additional electron density on the carbonyl ligands, which shift their vibrational frequencies.³³ A shift in vibrational frequency can allow for better energy-match between the solvent and the solute vibrational modes, which would expedite VER. For example, THF has a weak mode at 1980 cm⁻¹ that would be in registry with the red-shifted hydride spectra and may provide this pathway.

On the other hand, the comparison of VER between the two hydrides in this study also reveals smaller differences: The carbonyl mode encapsulated by the sol–gel relaxes 7 ps faster than the same mode in ultradry THF. Although the majority of the solvent in the pores has been exchanged with THF, there is certainly some residual water in this sample that impacts VER for the CO modes. We showed that adding water to a Ru₃(CO)₁₂ in THF solution, for example, can lower the lifetime by about 75%, which is consistent with the decrease between the two hydrides (Fig. S2 and Table S1). Modified THF bath modes inside the sol–gel pores, newly formed low-lying energy-accepting modes among the alumina sol–gel matrix, or finite amounts of water in the pores could all provide alternate mechanisms of VER that expedite vibrational relaxation. The similarity in T₁ values for the synthesized hydride in dry THF and sol–gel encapsulated hydride provides indirect evidence that the pores contain predominantly THF. Nonetheless, residual water cannot explain the acceleration in relaxation from Ru₃(CO)₁₂ to [HRu₃(CO)₁₁]⁻. We find that even adding copious amounts of water (1:5 H₂O:THF) to the Ru₃(CO)₁₂ solution is only able to lower the CO lifetime to 110 ps (Table S1).

The solvent dynamics of the three compounds were further studied via 2D-IR spectroscopy. The representative 2D-IR spectra are shown in Fig. 3 for Ru₃(CO)₁₂ in ultradry THF, alumina sol–gel, and [NEt₄][HRu₃(CO)₁₁] in ultradry THF at short (0.5 ps) and long (10 ps) waiting times (T_w). The x axes of the 2D-IR spectra are the pump frequencies (ω₁) and the y axes are the probe frequencies (ω₃). The red peaks arise from the ground state bleach of the ν = 1 → 0 transition, while the blue peaks arise from the ν = 1 → 2 excited state absorption. All spectra show strong on-diagonal peaks corresponding to the peaks in the FTIR spectra in Fig. 1 as well as off-diagonal peaks that grow as T_w is increased. The off-diagonal peaks indicate the exchange of excitation between modes that are coupled.

The 2D-IR line shapes change with T_w for all three systems. At early T_ws, the spectra are elongated along the ω₃ = ω₁ diagonal. At later T_ws, the peaks become increasingly round. Qualitatively, the vibrational modes in Ru₃(CO)₁₂ appear to have experienced more spectral diffusion over the first 10 ps than either of the hydride species. Spectral diffusion refers to the process by which vibrational frequencies are modified by structural changes and movements of the solute and nearby solvent molecules. The dynamics were quantified as a function of T_w via the center line slope (CLS) method.⁴² A plot of the CLS values as a function of waiting time (T_w) is shown in Fig. 4. CLS values for the Ru₃CO₁₂ and [NEt₄][HRu₃(CO)₁₁] samples in ultradry THF were obtained out to 70 ps. The shorter vibrational lifetime and heterogeneous nature of the sol–gel sample lead to increased scatter and worse signal to noise, even with the use of scattering cancellation techniques,⁴³ thereby limiting the T_w range to under 20 ps.

For clarity, we focus our analysis here on the 2060 and 2014 cm⁻¹ modes for Ru₃(CO)₁₂ and the hydride species, respectively, since these avoid overlapping with other spectral features. Each CLS decay was fit to an exponential decay plus a constant offset. The time constants for the fast component of the biexponential decays (τ₁) are the same within errors (Table I) and are all near 3 ps. Hence, the frequency fluctuations experienced by the CO ligands due to their surroundings for the Ru₃CO₁₂ and hydride compounds occur on similar time scales. The alumina sol–gel matrix dynamics cannot be the origin of these fluctuations, given that the THF samples lack sol–gel. These motions most likely stem from the movements of surrounding THF molecules, which are the major component of the ruthenium complex solvent shells. The shorter time scale dynamics of the surrounding solvent shell do not appear to be modified by confinement in the alumina sol–gel. The similarities in shorter time scale dynamics among the three samples can be rationalized by referring to previous studies conducted by Yamada and co-workers that examined the effects of nanoconfinement on solvent reorientational dynamics in nanoporous silicate sol–gel glasses.⁴⁴ In these nanoconfined spaces, the fastest observable solvent motions occurred on time scales similar to bulk solvent motions. These motions were attributed to solvent molecules far enough away from the walls of the pore that they were essentially bulk-like. Conversely, slower solvent dynamics arose from solvent molecules interacting with the surfaces of the silica nanopore. Previous work from our group reported a similar separation of dynamics in silica sol–gel pores into bulk-like and surface-like categories.^45,46 The similarity in τ₁ among the three samples here indicates that the carbonyl groups are sensing solvent molecules that behave like bulk THF rather than those found near the walls of the nanopores.

In the current study, the slower motions of the surroundings, which include the solvent and possibly the alumina backbone, result in variable offsets in the CLS decays in Fig. 4. The magnitude of the CLS offset can be used to identify differences in unresolvably slow inhomogeneous dynamics sensed by the carbonyl ligands among the Ru₃CO₁₂ and the free and encapsulated hydrides. For all T_ws beyond 6 ps, the CLS offset is the largest for [Ru₃CO₁₁H]⁻ in the sol–gel, only slightly smaller for [Ru₃CO₁₁H]⁻ in THF, and nearly zero for Ru₃CO₁₂ in THF. The sol–gel offset exceeding that of the hydride implies that the solvent molecules near the walls of an alumina nanopore and motions of the alumina walls contribute toward slower dynamics. However, the offset of the hydride dissolved in THF differs by less than 10% from the sol–gel, implying that bulk solvent motions are the primary source of the slower dynamics.

The CLS decays were used to quantify the homogeneous and inhomogeneous contributions to the FTIR linewidth via the frequency–frequency correlation function (FFCF).⁴⁷ The following correlation function was assumed for the fitting:

where T₂ is the dephasing time associated with the homogeneous linewidth; Δ₁ and τ₁ are the inhomogeneous amplitude and time constants, respectively; and Δ₀ is the contribution to the linewidth from long time scale inhomogeneous dynamics. T₂ is related to the homogeneous linewidth by $Γ=πT2−1$⁠; the time constants obtained from the CLS decays are used without adjustment in the $Ct$⁠, while the amplitudes, Δ_x, are floated to account for inhomogeneity in the linear line shape. The results of the FFCF fitting analysis are shown in Table I.

At the outset, one might expect that the dynamics of the entrapped hydride species would be distinct from those of the hydride in THF and that perhaps Ru₃(CO)₁₂ and [Ru₃CO₁₁H]⁻ in THF would be relatively similar. Indeed, spectral diffusion is usually found to be more reflective of the dynamics in the solvation shell rather than the molecular structure to which the oscillator is attached.^21,48,49 There are also examples in the literature in which encapsulation in sol–gel glasses leads to modification of vibrational dynamics on specific time scales, presumably due to changes in the surrounding solvent or excluded volume effects (guest species partially solvated by the pore wall).^50–52 However, the FFCF analysis of [Ru₃CO₁₁H]⁻ dissolved in THF and entrapped in a sol–gel shows that this is not always the case.

The FTIR peak widths are relatively insensitive among the three samples, decreasing from 12 to 11 to 10 cm⁻¹ for Ru₃CO₁₂, [Ru₃CO₁₁H]⁻ in THF, and [Ru₃CO₁₁H]⁻ in the sol–gel, respectively. However, the FFCF analysis reveals that the dynamic contributions to these linewidths do change, as anticipated from the 2D peak shapes in Fig. 3. For Ru₃(CO)₁₂ in THF, the homogeneous linewidth $Γ$ contributes about 25% of the total FTIR linewidth, while in the sol–gel and synthesized hydride, they contribute about 20%. Therefore, the relative homogeneous contributions to the FTIR linewidth decrease upon conversion to the hydride, while the relative inhomogeneous contributions to the FTIR linewidth (summation of Δ₁ and Δ₀) increase. This suggests the presence of slower overall solvent dynamics for both hydrides compared to Ru₃CO₁₂ in bulk THF. If the slowed dynamics were primarily due to confined solvent in the sol–gel pores, the [Ru₃CO₁₁H]⁻ dissolved in THF would not experience a relative increase in the inhomogeneous linewidth similar to that of the sol–gel. For both hydrides, the measurable spectral diffusion (Δ₁) shrinks, while the static inhomogeneous offset (Δ₀) increases. This offset in the FFCF represents the dynamics that are relatively slow on the time scale of our measurements and appear as static inhomogeneity.

Overall, the dynamic differences between the two hydrides are subtle; they are after all the same molecule in nearly the same solvent. In analyzing these subtleties, we note that complexes near a pore wall would be in closer proximity to slower moving solvent molecules interacting with the pore wall and possibly slow movements of the alumina itself; this would potentially lead to an increase in Δ₀. Concomitantly, some of the CO oscillators on these molecules would be less influenced by the bulk-like motions of solvent molecules near the middle of the pore wall leading to a decrease in Δ₁. In fact, [Ru₃CO₁₁H]⁻ in the sol–gel is likely to be found near the surfaces of a nanopore and the FFCF parameters for its CO vibrations are consistent with these changes.

The most substantial changes are observed upon conversion from Ru₃(CO)₁₂ to [HRu₃(CO)₁₁]⁻, regardless of whether the hydride is in THF or a sol–gel pore. The fast dynamics captured by Γ get faster for the hydrides. Previous studies have shown that solvent molecules near surfaces exhibit order.^12,13 This molecular order decreases molecular inhomogeneity at the pore surface and, therefore, the inhomogeneous linewidth. Returning momentarily to the VER discussed above, the faster relaxation time of the hydrides contributes to this homogeneous dynamic difference; however, the trend is in the wrong direction and the faster VER should increase Γ rather than decreasing it. Furthermore, for the long relaxation times in Table I, this could be attributed to no more than 0.15 cm⁻¹ of the change in Γ. Clearly, these faster homogeneous dynamics of molecules in the hydride solvation shell that perturb the CO frequencies are slowed down. At the same time, the quasi-static dynamics captured by Δ₀ are 50% less for Ru₃(CO)₁₂, and a concurrent increase in the 3 ps dynamics is also observed. Thinking of the dynamics as a continuum where the dynamics that are categorized by Γ (fastest), Δ₀ (medium), and Δ₀ (slow), the overall trend is a shift of faster movements in Ru₃(CO)₁₂ to slower movements in [HRu₃(CO)₁₁]⁻.

A pertinent comparison should be made between the hydride results and those in the literature that have shown modified dynamics for charged compounds relative to structurally similar but neutral species.^53,54 Dutta and co-workers reported that charged and neutral azides exhibited different spectral diffusion dynamics with the charged variants displaying significantly slower dynamics, consistent with the charged hydrides in the current work.⁵³ Similarly, it was shown recently by Kiefer and co-workers that charged organometallic catalysts showed slower spectral diffusion in spectroelectrochemical 2D-IR measurements.⁵⁴ In all cases, the operational principle is that the solvation shell structure is modified by the charge state of the analyte, and stronger solute–solvent interactions lead to slower vibrational dynamics. This is a plausible explanation for the observed differences of the hydrides relative to the neutral parent compound. Further work to simulate the solvation shell structure around these two species may shed light on the origins of these changes.

The ruthenium hydride [HRu₃(CO)₁₁]⁻ studied in this work experiences fundamentally different dynamics than its parent compound Ru₃(CO)₁₂. The accelerated VER shows that the CO vibrations couple more effectively to the surrounding solvent environment for this species, whether dissolved in neat THF or entrapped in an alumina sol–gel matrix. At the same time, the homogeneous and inhomogeneous dynamics that are characterized by 2D-IR line shape analysis report an overall slowing of the frequency fluctuations in the hydride. The dynamic changes experienced by the hydride encapsulated in alumina were found to be small but measurable. This is interesting because the alumina encapsulated complex has been reported as a potent catalyst for alkene hydrogenation reactions, while the parent compound and even silica sol–gel encapsulated compounds are relatively inactive for the same chemistry. In this work, we explored some of the fundamental differences in solvent dynamics that might contribute to this change in reactivity. Although we cannot definitively connect dynamics to reactivity with these measurements alone, we can conclude that the most likely dynamic contributions are experienced upon transformation to the hydride, not encapsulation in the sol–gel. It is certainly true that the ruthenium hydride experiences a new palette of dynamics that perturb the CO ligands on the time scale of chemical reactions. The subtlety of the dynamic changes upon encapsulation strongly suggests that the new reactivity in the alumina matrix is not due to a dynamic change in the confined solvent pool. Instead, our work shows that it is more likely that the sol–gel environment stabilizes the reactive hydride structure and protects it from degradation.

See the supplementary material for tabulated biexponential fit parameters to vibrational relaxation decays, 2D-IR off-diagonal peak analysis for Ru₃(CO)₁₂ in ultradry THF, and FTIR spectrum of Ru₃(CO)₁₂ in THF with water added.

The authors gratefully acknowledge partial support from the National Science Foundation under Grant No. CHE-1856589. C.G.P. was supported, in part, by a Newman and Lillian Bortnick Fellowship. J.G.P. was supported, in part, by a Mistletoe Research Fellowship.

The authors have no conflicts to disclose.

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The data that support the findings of this study are available from the corresponding author upon reasonable request.