Hydrogenation Studies of Iridium Pyridine Diimine Complexes with O- and S-Donor Ligands (Hydroxido, Methoxido and Thiolato) ^#

Abstract

For square-planar late transition metal pyridine, diimine (Rh, Ir) complexes with hydro-xido, methoxido, and thiolato ligands. We could previously establish sizable metal-O- and S π-bonding interactions. Herein, we report the hydrogenation studies of iridium hydroxido and methoxido complexes, which quantitatively lead to the trihydride compound and water/methanol. The iridium trihydride displays a highly fluctional structure with scrambling hydrogen atoms, which can be described as a dihydrogen hydride system based on NMR and DFT investigations. This contrasts the iridium sulfur compounds, which are not reacting with dihydrogen. According to DFT and LNO-CCSD(T) calculations, hydrogenation of the methoxido complex proceeds by a two-step mechanism, i.e., an oxidative addition step of H₂ to an Ir(III) dihydride intermediate with consecutive reductive O-H elimination of methanol. Based on PNO-CCSD(T) calculations, the reactivity difference between the O- and S-donors can be traced to the stronger H-O bonds in the water/methanol products compared to the S-H bonds in the sulphur congeners, which serves as a driving force for hydrogenation.

1. Introduction

In a seminal study, Goldberg et al. investigated the hydrogenation mechanism of late transition metal alkoxide and hydroxide ligands in square-planar PCP pincer compounds, which results in the formation of the corresponding alcohols and water, respectively, and metal hydride complexes [1]. The presence of lone pairs on the O-donors, which represent potential extra-basic sites, led these researchers to question whether the reaction paths might differ from the hydrogenolysis of alkyl or phenyl ligands. Three different pathways were considered, which are depicted in Scheme 1.

The electronic and steric properties of the O-R ligand and reaction conditions, such as solvent polarity or the presence of adventitious water, were found to influence the reaction paths. Among the three pathways considered in Scheme 1, path B is the most relevant, displaying a four-centered transition state that resembles a σ-bond metathesis pathway for metal-carbon bonds [2]. However, for alkoxides, the reaction pathway differs as the lone-pair electrons of the oxygen atom can act as a basic site, leading to the intramolecular deprotonation of a dihydrogen ligand that becomes more acidic through metal coordination. This type of reaction is termed internal electrophilic substitution (IES).

DFT calculations demonstrated that pathway A, involving H₂ oxidative addition followed by O-H reductive elimination, had significantly higher barriers and was not considered competitive. These results were subsequently confirmed by Karen Goldberg et al. and other groups through experimental and theoretical studies on square-planar group 10 transition metal systems, including Ni, Pd, and Pt [3,4]. For this study described herein, the hydrogenolysis of group 9 Rh methoxido complexes is also relevant, although the latter systems were six-coordinate octahedral [3,4]. For octahedral systems, there are also examples for the hydrogenation of rhodium and iridium sulfur bonds, which lead either to the formation of metal hydrides with [5,6] or without elimination of the thiol [7].

Our group investigates square-planar complexes with pyridine diimine NNN-donors, for which we and others have demonstrated their ability to behave as non-innocent ligands and reasonable π-acceptors [8,9,10,11,12,13,14,15]. We recently analyzed the push-pull π-interactions between sulphur and oxygen π-donor and the PDI π-acceptors in detail [11,16]. This leads to a particular high thermal stability of the rhodium and iridium OMe and SMe units, which are resistant to β-hydride elimination even at elevated temperatures [11,16,17]. In the course of this investigation, we also studied the hydrogenolysis of the iridium oxygen and iridium sulphur bonds, which is reported herein.

2. Materials and Methods

2.1. Materials

The syntheses of the iridium alkoxido and thiolato complexes, along with a related PDI iridium methyl compound utilized in this study, were previously reported [17]. The hydrogenation conditions are reported in the Supplementary Material.

2.2. Methods

(a)

NMR spectroscopy

Variable temperature T₁ ¹H NMR relaxation times were measured using standard inversion recovery experiments over the temperature range of 176 to 296 K at 300 MHz. The T_1min relaxation time represents the minimum value in the T₁ versus temperature plot (see Figure S2 in the Supplementary Material ).

(b)

X-ray crystallography

Single X-ray crystal measurements were conducted using a BRUKER AXS SMART APEX single crystal diffractometer, equipped with graphite monochromatic MoKα radiation (λ = 0.71073 Å) at 100 K. The single crystals were mounted in high-viscosity polybutene oil on a lithographic cryo-loop attached to the goniometer head. Data collection and analysis were performed using the software packages SAINT and SADABS. The structure was solved and refined using the Shelx [18] and Olex2 [19] program packages. All atoms, except for the hydrogen atoms, were refined anisotropically. Two independent complexes and THF molecules are found in the unit cell. The unreacted starting material complex partially cocrystallizes on one of the two sites and could be refined with an occupation factor of 0.44.

2.2.1. Theoretical Methods

DFT Calculations

For the geometry optimizations of both ground and transition states, DFT calculations were performed using the PBE functional with dispersion corrections via Grimme’s D3 method with Becke-Johnson damping (D3BJ) [20] with the Turbomole program package ver. 7.7 and 7.8 [21]. The Def2-TZVP basis sets were used for all atoms, and an ECP-60-MWB was employed for iridium. For the PW6B95 hybrid functional, seminumeric exchange was used, indicated by the $senex keyword in Turbomole. Solvation effects were accounted for using the COSMO formalism with a dielectric constant of ε = 7.6 for THF. Geometries were fully optimized without imposing any constraints on geometry or symmetry.

To confirm minima, we ensured the absence of imaginary frequencies in the analytic second derivative calculations. For transition states, only one imaginary frequency was observed. Transition state optimizations utilized Kästner’s DL-FIND optimizer implemented in TCL-Chemshell 3.7, starting from geometries identified through linear transit searches [22]. IRC calculations were conducted to confirm that these transition states connect the starting materials and products using the DRC program provided by Turbomole.

Local Coupled Cluster Calculations

For closed-shell systems, local natural orbital (LNO) coupled-cluster calculations at the LNO-CCSD(T) level were performed using the freely available MRCC (2022) program package (https://www.mrcc.hu/ accessed on 6 August 2024) with default thresholds (lcorthr = normal) [23]. Geometries were optimized at the PBE-D3BJ/def2-TZVP level. The def2-TZVPP basis sets, along with complementary def2-QZVPP/C auxiliary correlation basis sets and def2-TZVP pseudopotentials, were employed. The solvation correction was derived from the energy differences between two single-point calculations at the PBE-D3BJ(COSMO(ε = 7.6)/def2-TZVP) level and the other at the PBE-D3BJ/def2-TZVP (gas phase) level. Back corrections for LNO-CCSD(T) energies to free enthalpies (ΔG298) were conducted using thermochemical data from DFT frequency calculations at the PBE-D3BJ/def2-TZVP level, with a scaling factor of 1.011 taken from Truhlar’s database (ver. 5.0) [24]. Typical values for T1 and D1 diagnostics were approximately T1 ≈ 0.015 and D1 ≈ 0.15, indicating single-reference cases.

For the evaluation of bond dissociation enthalpies, local coupled cluster calculations for open-shell (S = 1/2) systems were performed using Molpro version 2022.3, applying PNO-U(R)-CCSD(T1) theory with the setting domopt = tight [25,26]. A complete basis set (3->4) extrapolation was carried out using def2-TZVPP and def2-QZVPP basis sets. The SO-SCI SCF optimization scheme was used to converge to the ground state of the Hartree-Fock reference wave function. Enthalpy corrections to thermochemical data were provided by the “freeh” program of the Turbomole package, using analytical second derivatives obtained at the U/PBE-D3BJ/def2-TZVP level. A scaling factor of 1.011 was used from Truhlar’s database (ver. 5.0). A value of 5/2 RT was used for the hydrogen atom. The default def2-TZVP pseudopotential was employed for iridium (ECP-60MWB), and the corresponding RIJK auxiliary density fitting basis was used for all atoms. Typical values for T1 and D1 diagnostics were around T1 ≈ 0.015 and D1 ≈ 0.15, signaling single reference cases, which was further substantiated by negligible spin contamination (<S**2> ≈ 0.75) of the Hartree-Fock reference wave functions of the S = 1/2 radicals.

These computations were performed using versions 7.7.1 and 7.8 of the parallelized Turbomole program package on our local machines: a 32-core and a 96-core system equipped with 512 GB and 3 TB of RAM, respectively. Additionally, computations utilized the two 40-core nodes (each with 1 TB RAM) of the “Hummel” computing cluster at the University of Hamburg’s computing center (RRZ).

3. Hydrogenation Reactions

We previously studied the facile C-H activation in a PDI iridium methyl system, which reacted with benzene at ambient temperature to yield the corresponding phenyl complex and methane [27]. For this transformation, we proposed a mechanism with an oxidative addition and a consecutive reductive elimination step. The low barrier for oxidative addition in benzene was rationalized by the facile accessibility of a highly reactive C_2v-d⁸ ML₄ fragment [28] generated by out-of-plane bending of the methyl group. This reactivity was contrasted by the corresponding hydroxido and methoxido rhodium and iridium complexes, which were stable even when heated to reflux in benzene over several days [16]. This contrasts the CH-activation in benzene by square-planar (PNP)Rh hydroxide and phenoxide complexes [29,30] and (acac)₂iridium(III) hydroxido and methoxido complexes at higher temperatures [31,32].

In the course of this investigation, we studied the hydrogenation of the methyl complex 1 for comparison. This reaction proceeds instantaneously under ambient conditions and yields the diamagnetic compound 2 (Scheme 2). To avoid ambiguity, unless explicitly specified and without reference to the bonding situation, we will henceforth refer to the Ir(H₃) moiety as iridium trihydride.

The observed doublet and triplet for the pyridine meta and para protons as well as a singlet for the ketimine methyl group in the 300 MHz ¹H NMR spectrum in THF-d₈ suggested a (time-averaged) C_2v-symmetrical structure of 2 in solution, which is also in agreement with the residual ¹H and ¹³C NMR data (details see SI). The hydride signal appeared as a sharp singlet at δ = −9.5 ppm at RT with the correct integration for three protons. Upon cooling, this resonance significantly broadened and displayed a half-line width of 18 Hz at 176 K (300 MHz, details see SI). However, the resonance did not split up into two signals as might be expected for a classical octahedral trihydrido complex with two chemically equivalent homotopic protons in the apical positions and one in the octahedral plane. Details of the binding mode in the IrH₃ unit and further NMR spectroscopic data will be discussed below, together with the results for the related complex with phenyl rather than methyl substituents at the ketimine carbon atom.

It deserves a special mention that methane is also formed in the reaction with a stochiometric amount of H₂. However, rather than the expected monohydrido complex, we obtained a black highly insoluble precipitate, which thwarted its further characterization/identification. The analogous reaction for a PDI rhodium methyl analogue was reported by Budzelaar et al. and yields a paramagnetic complex [28]. We could previously show for the related rhodium methyl complex that an end-on rhodium dinitrogen compound is formed when the reaction is carried out in the presence of N₂ [33]. Next, we tested the hydrogenation of the iridium complexes with the heteroatom O,S-donor ligands, i.e., (PDI)Ir-XR (X = O,S, R = H,Me).

3.1. Comparison of the Sulphido, Hydroxido, and Methoxido Systems

3.1.1. O-Donors

The green hydroxido and methoxido iridium complexes 3 and 4 cleanly react with dihydrogen within 12 h at 60 °C to the green-turquois trihydride species (Scheme 3). The quantitative formation of water and methanol, respectively, was established by their ¹H NMR signals in the reaction mixture as well as in the NMR spectrum of the vacuum-transferred volatiles. We could not detect resonances for (free) formaldehyde in the reaction of the methoxido complex, which provides strong arguments against the involvement of a ß-hydride elimination step as part of the hydrogenation process. It deserves a special mention that we were not able to isolate (or detect) the monohydrido (PDI)Ir-H complex. This is likely due to the strong thermodynamic preference for the formation of the trihydride and will be discussed below.

3.1.2. S-Donors

In contrast to the reaction of the hydroxido and methoxido complexes with dihydrogen, the analogous compounds with hydrogensulfido and methyl thiolato ligands 5, 6 did not undergo a reaction under the same conditions (Scheme 3). Even when the temperature was raised and the reaction time extended to days, no conversion was observed. As will be detailed below, we attribute this difference to a thermodynamically unfavorable situation for the complexes with S-donors. Before we do so, we will take a quick glimpse at the structure of the iridium trihydride product 7.

3.2. Characterization of the Trihydride Product

The ¹H NMR spectrum of the product 7 displays a broad singlet for the hydride re-sonances at RT at δ = −8.53 ppm (ω₁₂ = 3.6 Hz), which integrates for 3 protons. Together with the triplet and doublet for the para and meta pyridine protons and the observation of three ¹³C NMR signals for the pyridine carbon atoms and one for the ketimine carbon atom, this suggests time-averaged C_2v-symmetry of the trihyride complex 7 at RT. The time-averaged C_2v-symmetry is maintained in the 600 MHz spectrum in toluene-d₈ even upon cooling to −90 °C, which led to further broadening of the singlet for the hydrides. This matches the observations for the related trihydride 2 (vide supra), which carries methyl rather than phenyl substituents at the ketimine carbon atom (Figure 1). We will therefore discuss only the NMR results of the latter complex 2 in the forthcoming.

Inversion recovery measurements of the T₁ relaxation times in the range of 175–300 K for the hydride resonances revealed T_1min = 27 ms at 196 K and 300 MHz for complex 2. (see SI). This value lies in the well-established typical T_1min range of 10–50 ms for dihydrogen complexes, while T_1min relaxation times >> 100 ms are established for classical metal hydrides [34]. Based on the short T_1min and the time-averaged C_2v-symmetrical structure, the fast exchange process shown in Scheme 4 is proposed, which scrambles the hydrides over all three sites.

For complex 7, after numerous attempts, we were able to obtain suitable crystals for X-ray crystal structure determination from the reaction of the hydroxido complex 3 with H₂. There are two independent complex molecules in the unitary cell, of which one consists of partially unreacted residual hydroxido starting material. The hydroxido group could be refined with an occupation factor of 0.44. The molecular structure of 7 is shown in Figure 2, and details of the data collection and refinement and selected bonds and angles are summarized in the ESI.

While the structure of the planar PDI iridium core unit could be clearly established, the three missing hydrogen atoms of the IrH₃ unit could not be located/refined. Unfortunately, our attempts to obtain reliable geometric parameters for the hydrogen atoms from neutron diffraction data were thwarted by the lack of suitable (sizeable) single crystals.

The average distance of the rather long ketimine unit and the short bond of the C_py-ridine-C_imine-bond of 1.348 and 1.429 Å point toward a doubly reduced, i.e., PDI²⁻ ligand. This is clearly reflected in the low value of 0.061 Å for the established Wieghardt’s Δ_geo parameter, which allows to analyze the non-innocence of PDI ligands [12]. Since it is well-documented that DFT calculations deliver good results for geometries and thermochemistry for hydride and dihydrogen complexes, we sought to obtain the missing bonding data of the IrH₃ unit from theory.

The most relevant geometric parameters optimized at the PBE-D3BJ/def2-TZVP level (Ir: def2-ECP) are shown in Figure 3, in which the aryl groups were omitted for clarity. There is good agreement between the experimental and DFT calculated PDI ligand bond distances, which is nicely reflected in the excellent match of the Δ_geo parameters (0.061 Å (X-ray) vs. 0.05 Å (DFT)).

This moiety is planar and perpendicular to the PDI Ir plane, displaying a hydrido and a significantly stretched dihydrogen ligand [34] at the brink to two separate hydrides, as evidenced by the long H-H distance of 1.343 Å. The distance of the central hydrogen atom to the apical hydride amounts to 1.754 Å, which is above the generally accepted limit of 1.6 Å between two hydride ligands [34]. The slightly shorter Ir-H distance of 1.594 Å to the hydrido ligand compared to 1.617 and 1.640 Å for the hydrogen atoms of the dihydrogen unit is in full agreement with the description as an Ir(H₂)H unit. Overall, the (PDI)Ir(H₂)(H) core displays C_s-symmetry. We also performed DFT geometry optimizations starting in C₂-symmetry, which makes the two H_apical-H_equatorial distances (r₁ and r₂) equidistant. Upon release of the symmetry constraints and further geometry optimization at the PBE3-BJ/def-TZVP DFT level, these H-H distances remain essentially unchanged (1.59 and 1.58 Å). The optimized structure corresponds to the transition state connecting the two degenerate dihydrogen hydride structures. The barrier for this process is tiny and amounts to only 1 kcal/mol (Figure 4 left) and 0.5 kcal/mol at the LNO-CCSD(T)/def2_TZVPP level. Since the energetic difference is so small, we probed different DFT functionals and basis sets accompanied with reoptimizations of the geometries. The results are compiled in

Table 1. Energetic differences and bonding parameters of the IrH₃ moiety in the dihydrogen hyrido and trihydrido structure for different density functionals, basis sets, and different types of dispersion corrections.

Method/ Basis Set	ΔΕrel = ΔE(Ir(H2)(H) − ΔE(Ir(H)3 [kcal/mol]	Ir(H2)(H) Distances in [Å]			Ir(H)3 Distances in [Å]
		Ir-H1 Ir-H2 Ir-H3	H1-H2, r1	H2-H3, r2	Ir-H1 Ir-H2 Ir-H3	H1-H2, r1	H2-H3, r2
PBE-D3BJ/ def2-TZVP	−1.0	1.594 1.640 1.617	1.754	1.343	1.598 1.638 1.602	1.579	1.584
r2scan-3c/ def2-mTZVPP	−0.3	1.591 1.625 1.605	1.769	1.521	1.592 1.629 1.596	1.637	1.664
BP86-D3BJ/ def2-TZVP	−0.2	1.596 1.636 1.614	1.751	1.411	1.600 1.637 1.603	1.597	1.619
PBE0-D3BJ/ def2-TZVP	−1.3	1.574 1.669 1.653	1.822	1.009	1.584 1.625 1.586	1.539	1.573
PBE0-D4/ def2-QZVPP		1.573 1.667 1.651	1.799	1.005
B3LYP-D4/ def2-TZVP	-0.3	1.590 1.628 1.605	1.752	1.474	1.592 1.630 1.597	1.630	1.655
WB97X-V/ def2-TZVP	n/a	1.592 1.626 1.591	1.671	1.685
WB97X-D4/ def2-TZVP		1.591 1.626 1.590	1.663	1.596
PW6B95/ def2-TZVP	−1.8	1.575 1.688 1.677	1.861	0.952	1.585 1.625 1.590	1.609	1.587

.

All but the range separated WB97X-D4 and WB97X-V density functionals favor the dihydrogen, hydride structure, albeit by a very small margin of −0.2–−1.8 kcal/mol. The H-H bond lengths of the dihydrogen ligands vary from 1.009–1.512 Å, with the latter distance obtained with the r2scan-3c functional reaching the range observed for compressed dihydrides [35]. For the WB97X functionals, the trihydride structure represents the minimum for both dispersion correction methods employed, i.e., D4 and VV10.

These diverging results prompted us to calculate an energy hypersurface with the two H_apical-H_equatorial distances r₁ and r₂ as parameters. The contour plot for the PBE-D3BJ functional shown in Figure 4 reveals an extremely shallow energy hypersurface with the two degenerate dihydrogen hydride structures displaying the global minima, which are connected by the trihydride motif of the TS. Expectedly, this contrasts the energy hypersurface for the WB97X-V functional, which was constructed from single point calculations at the PBE-D3BJ optimized geometries.

For the W97X-V density functional, the trihydrido system represents the global minimum. Without further experimental investigations, e.g., analysis of temperature-dependent H-D coupling constants, it is not possible to conclude, however, whether this corresponds to a true trihydride motif or a structure with two compressed diyhdrides. This study is beyond the scope of this paper. Overall, the hypersurface is extremely flat and resembles previous findings for dihydride and trihydride systems reported by several groups [35], including the Heinekey [36,37] and Gusev [38] group. An excellent, extensive study on this topic was published recently by Wendt et al. [39].

Taking the NMR data and the DFT calculations into account, it is anticipated that scrambling over all three sites proceeds by the mechanism shown in Scheme 4.

For the exchange processes by rotation of the dihydrogen ligand shown in steps 2 and 5, we calculated a small barrier of 4.5 kcal/mol for the PW6B95-D3BJ DFT functional and a comparable value of 6.1 kcal/mol at the LNO-CCSD(T) level, which is in the mid-range for dihydrogen ligands (0.8–11 kcal/mol) [34,35]. The corresponding transition state is shown in Figure 5 (right).

In principle, the observed dynamics in the ¹H NMR spectra could also be explained with an alternative process involving dissociation/reassociation of H₂ according to (PDI)Ir(H₂)H ((PDI)IrH + H₂). Based on our DFT and LNO-CCSD(T) calculations (cf. Equation (1) below), which revealed that H₂ dissociation is strongly endothermic (+21 kcal/mol (and endergonic ΔG₂₉₈ = +10 kcal/mol), we anticipate that this process cannot compete with the intramolecular exchange pathway, however.

The type of exchange mechanism shown in Scheme 4 is well-established for trihydride systems. A prominent example is the cationic half sandwich rhodium complex Cp*Rh(PMe₃)(H)(H₂)⁺, which displays a dihydrogen hydride structure and a T_1min relaxation time of 23 ms at 500 MHz [40]. The analogous iridium system is a classical trihydride Cp*Ir(PMe₃)(H)₃⁺, which exhibits quantum mechanical exchange coupling between the hydrogen nuclei [41]. The H…H distances observed in the neutron diffraction structure of the iridium complex are relatively short and amount to 1.674 (14) and 1.698 (13) [42]. A related study for an octahedral iridium trihydride investigated by Gusev revealed a dihydrogen hydride structure in the solid state and a trihydrido tautomer in solution.

4. Thermodynamics and Mechanism of the Hydrogenation Reaction

4.1. Thermodynamics

After having unambiguously established the trihydrido product of the hydrogenation reaction, we analyzed the thermodynamics of this reaction by DFT and LNO-CCSD(T) calculations. At first, however, we turned our attention to the question of why the trihydride rather than the monohydride complex is formed. The thermodynamics for the equilibrium shown in Equation (1) immediately provided the answer.

(PDI)IrH + H₂ ⇆ (PDI)IrH₃

(1)

The reaction lies strongly on the side of the trihydrido product, as evidenced by the value of −10.04 kcal/mol for ΔG₂₉₈ calculated at the LNO-CCSD(T)/def2-TZVPP level (ΔE_R = −21.01 kcal/mol). Therefore, it is anticipated that the hydrogenation leads to a monohydride intermediate, which is then quickly converted to the trihydride product. It deserves a special mention at this point that only very few square-planar iridium complexes with a terminal hydrido ligand are known [43,44,45,46].

Furthermore, our calculations provided insights into the thermodynamics of the hydrogenation of the iridium complexes with O,S-donor ligands.

Table 2. Thermodynamics for the hydrogenation of the O,S donors according to Equation (2).

Method/Complex	(PDI)Ir-OH 3	(PDI)Ir-OMe 4	(PDI)Ir-SH 5	(PDI)Ir-SMe 6
ΔEhydrogenation DFT [kcal/mol] a	−10.48	−10.33	+8.97	+8.19
ΔEhydrogenation LNO-CCSD(T) b [kcal/mol]	−13.34	−12.10	+7.55	+6.61
ΔG298, hydrogenation LNO-CCSD(T) c [kcal/mol]	−0.78	−2.07	+16.08	+14.14
ΔG298 LNO-CCSD(T) COSMO d (ε = 7.6) [kcal/mol]	−4.19	−5.81	+15.22	+13.27

^a DFT: PW6B95-D3BJ, basis def2-TZVP, Ir: ECP-60-MWB; ^b LNO-CCSD(T), basis def2-TZVPP, Ir: ECP-60-MWB, MRCC program package (localcc = 2021, default parameters); ^c thermodynamic corrections from vibrational analysis data at the PBE-D3BJ/def2-TZVP level; ^d COSMO solvation estimated from the DFT calculated energies of the educts and products in the gas phase and in solution.

summarizes the results for the equilibrium shown in Equation (2) for the complexes with the full PDI ligand system presented above in Scheme 3. We disregarded the formation of the hydrogen-bonded water dimer 2 H₂O (H₂O)₂, which is favored by 3 kcal/mol [47] and would shift Equation (2) for complex 3 by 1.5 kcal/mol even further in the direction of the hydrogenation products. For H₂S, this additional driving force is smaller and amounts to a mere 0.85 kcal/mol [48].

(PDI)IrX + 2 H₂ ⇆ (PDI)IrH₃ + X-H

(2)

Inspection of Table 2 immediately reveals that the hydrogenation of the iridium complexes with S-donor ligands is thermodynamically unfavorable. Considering THF solvation and employing thermodynamic corrections, these reactions are uphill by 15.22 and 13.27 kcal/mol, corresponding to equilibrium constants of ca. 10⁻¹¹ and 10⁻⁹ at 298 K. This is in sharp contrast to the results for the hydrogenation of the hydroxido and methoxido complexes. The latter exhibits negative free enthalpy of reactions of ΔG_298,COSMO = −4.19 and −5.81 kcal/mol, thus signaling a favorable (downhill) situation for the iridium hydroxido and methoxido compounds. These data readily explain the different outcomes for the iri-dium complexes with either O- or S-donor ligands. We will briefly analyze the origin of this difference before we eventually turn to the mechanism(s) involved in the hydrogenation reaction.

Analysis of the Thermodynamic Data

In order to study the thermodynamic situation, we calculated the BDEs of the Ir-H, X-H, Ir-X, and H-H bonds, where X = Me, OH, Ome, and SH, SMe, with high-level PNO-UCCSD(T)/def2-T(Q)ZVPP calculations with Molpro 2022.3 (

Table 3. BDEs of Ir-X and H-X bonds in kcal/mol calculated with various theoretical methods and basis sets. For the H-X bonds, the experimental values were taken from the literature and are included in parentheses [55].

Bond/Method	PBED3BJ	PW6B95D3BJ	PNO-CCSD(T)//PBED3BJ/def2-TZVP
	def2-TZVP/def2-QZVPP		def2-TZVPP	def2-QZVPP	CBS(3-4)
	Ir: def2-ECP		Ir: def2-ECP
(PDI)Ir-H	58.58	63.29	63.20	63.33	63.44
(PDI)Ir-Me	53.07	51.24	55.64	55.42	55.36
(PDI)Ir-OH	90.09	84.80	87.91	87.91	87.99
(PDI)Ir-SH	81.41	78.22	79.08	80.21	81.20
(PDI)Ir-OMe	72.47	69.52	76.36	76.01	75.83
(PDI)Ir-SMe	75.06	72.34	74.36	75.10	75.82
H-H	99.35/99.44	102.71/103.01	103.01	103.74	104.20 (104.20)
H-CH3	102.62/102.65	104.76/104.87	103.54	104.34	104.91 (105.00)
H-OH	115.98/117.62	114.85/116.51	116.07	117.96	119.12 (118.98)
H-SH	89.77/90.07	90.60/90.99	89.55	90.85	91.72 (91.29)
H-OMe	97.57/103.71	98.87/104.71	102.93	104.32	105.18 (105.32)
H-SMe	83.76/85.05	85.05/86.81	85.18	86.31	87.08 (87.49)

Calculated with Molpro ver. 2022.3 with option domopt = tight and thermodynamic corrections from vibrational analysis data obtained at the PBE-D3BJ/def2-TZVP level (scaling factor 1.011) [56].

) [49] extrapolated to the complete basis set limit using a two-point (3→4) extrapolation scheme [50]. A previous work by Ess et al. focusing on the binding of late transition metals (Ru, Rh, Ir, Pt) to he-teroatoms deserves a special mention [51]. They employed DFT and canonical CCSD(T) calculations with a small basis set (6-31G(d,p)) due to computational and methodological limitations at that time. However, their study successfully reproduced the experimentally established correlation between M-X and H-X bond energies [52]. The BDEs of the O-H and S-H bonds in water, hydrogen sulfide, and methyl sulfide were previously calculated up to the canonical CCSD(T)/aug-ccpV6Z level [53,54].

As previously noted, complete basis sets are required to achieve chemical accuracy for the small sulphur molecules [54]. This is also reflected in our PNO-CCSD(T) data, which provides excellent results for the H-X BDEs with the def2-T/QZVPP(3-4) complete basis set extrapolation scheme. For the smaller triple and quadruple zeta basis sets, the calculated BDEs are on average 2.1%, respectively 0.8% too low.

The calculated data reflect the expected trend that Ir-XR (X = O,S; R = H,Me) bonds are stronger for R = H. In addition, it is noteworthy that the Ir-H BDE of 63.44 kcal/mol is rather low compared to experimentally established data for other iridium hydride complexes, which are typically >>70 kcal/mol [57]. We have no explanations for this difference but are currently looking into it. The iridium methyl bond dissociation energy is lower (55.36 kcal/mol) but lies in the expected range [57,58].

Turning back to the question, why do the iridium complexes with O-donor ligands undergo hydrogenation while the thiolates are reluctant to react with H₂? This becomes immediately clear upon inspection of the BDEs for the Ir-OR and Ir-SR bonds and the corresponding H-O,S bonds in methanol (MeOH) and methyl sulfide (MeSH). While the Ir-S and Ir-O bonds display essentially the same BDEs (Ir-O: 75.83 vs. Ir-S: 75.82 kcal/mol), the MeO-H bond is ca. 18 kcal/mol stronger than the S-H bond in methyl sulfide. For water vs. hydrogen sulfide, this difference of the H-OH and H-SH BDEs is even further extended to 27 kcal/mol. While the Ir-OH bond is ca. 12 kcal/mol stronger than the Ir-OMe bond, this increase is compensated by the larger O-H BDE of 119 kcal/mol in water, which is +14 kcal/mol higher compared to the one in methanol. It can therefore be concluded that the stronger O-H bonds in water and methanol drive the reaction to the product side, while the weaker S-H bond strengths in H₂S and MeS-H are not sufficient. At this point, it has to be recalled that there are examples for the hydrogenation of rhodium sulfur bonds in octahedral trispyrazolyl rhodium(III) complexes, which operate under ambient conditions [5,6]. Interestingly, the latter process can be reversed when the hydrogen atmosphere is replaced by dinitrogen. This implies a nearly thermoneutral situation for hydrogenation of the Rh-S bond, which might be explained by the weaker rhodium sulfur bond compared to the stronger bond Ir-S bond in the 5D system, with the latter being further stabilized by the partial Ir-S π-bond in our PDI complexes. After this thermodynamic analysis, we will turn to the investigation of the mechanism of the hydrogenation process.

4.2. Hydrogenation Mechanism

In a seminal study, the Goldberg group examined the hydrogenation process of square-planar, d⁸-configured palladium alkoxo PCP pincer complexes in great detail using both kinetic and theoretical modeling [1,4]. Gunnoe et al. reported later a thorough mechanistic and theoretical study on the hydrogenation of a Rh-OMe bond in an octahedral Rh(III) system [59]. Herein, we will restrict ourselves to the theoretical modeling of this reaction for the iridium methoxido system.

In our previous study, DFT calculations indicated that the ionization of the methyl thiolato ligand in complex 6 in tetrahydrofuran (THF), as represented by the equation (PDI)Ir-SMe + THF ⇌ (PDI)Ir-(THF)⁺ + SMe⁻, is energetically highly unfavorable. The calculated values at the DFT/PBE-D3BJ/def2-QZVPPD level with solvation (COSMO(ε = 7.6, THF)) in THF are +48.7 kcal/mol, while even in water (ε = 80.0), this value amounts to +33 kcal/mol. For the ionization of the methoxido ligand, 4 values of 45.5 (ε = 7.6) and 28.7 kcal/mol (ε = 80) were obtained at this level. Considering that our previous molecular conductivity measurements showed that a related PDI iridium complex bearing a significantly better ionizable triflato ligand is only partially ionized in THF solution [16], we anticipate that ionization of the O- and S-donors does not play a role in the hydrogenation process. Therefore, our investigations focused solely on neutral and uncharged species. The results of our DFT study with single-point LNO-CCSD(T)/def2-TZVPP calculations for the stationary points are shown in Figure 6.

We considered both a σ-bond metathesis pathway and a two-step mechanism with consecutive oxidative H₂ addition (oxadd) and C-O reductive elimination (redel) of methanol. Inspection of Figure 4 immediately reveals a substantially higher activation of 33 kcal/mol (TS13) for σ-bond metathesis vs. the overall barrier of 27 kcal/mol for the two-step oxadd/redel process. The latter displays an activation barrier of 18 kcal/mol (TS12) for the formation of the octahedral iridium dihydrido intermediate I-2, which is slightly energetically uphill by 7 kcal/mol. The reductive O-H elimination step passes over the transition state TS23 and exhibits a barrier of 20 kcal/mol. It leads to the five-coordinate square-pyramidal intermediate I-3 with a methanol-bound ligand, which is 12 kcal/mol downhill from I-2. In the consecutive steps, the methanol ligand is displaced by another H₂ molecule to give the final trihydrido product.

This mechanistic scenario resembles our proposed two-step mechanism for the C-H activation in benzene (C₆H₆) in a (PDI)Ir methyl complex, (PDI)Ir-CH₃ + C₆H₆ → (PDI)Ir(H)(C₆H₅)(CH₃) → (PDI)Ir-C₆H₅ + CH₄, which also involves an octahedral intermediate [27]. It deserves a special mention, however, that the reductive C-H elimination step to give methane involves only a small barrier of 6 kcal/mol, contrasting the value of 20 kcal/mol for O-H elimination in I-2. This difference is expected based on previous reports in the literature, however [60].

We also considered an alternative pathway involving a hydroxymethylene intermediate formed by isomerization of the methoxido complex, which is then hydrogenated in the consecutive step (Equation (3)).

$$(\mathrm{PDI})\mathrm{Ir}\text{-}{\mathrm{OCH}}_3\rightleftarrows (\mathrm{PDI})\mathrm{Ir}(\mathrm{H})({\mathrm{CH}}_2\mathrm{O})\rightleftarrows (\mathrm{PDI})\mathrm{Ir}\text{-}{\mathrm{CH}}_2\mathrm{OH}\stackrel{H_2}{\to }(\mathrm{PDI})\mathrm{Ir}\text{-}\mathrm{H}+{\mathrm{CH}}_3\mathrm{OH}$$

(3)

Such an alternative pathway was previously discussed by Milstein et al., who studied the reductive elimination in octahedral rhodium and iridium complexes. For their iridium(III) system [61,62], direct reductive O-H elimination to methanol from the methoxido complex was observed [61].

DFT calculations for a smaller model system with a modified imine group (py-CH = N(o-xlyl)) indeed confirmed a very small activation energy of only 5 kcal/mol for the C-H elimination of methanol from the H-Ir-CH₂OH intermediate (details see SI). However, as analyzed in our previous paper, this has to be traded in for the overall much larger barrier of 37 kcal/mol for the isomerization process of the hydroxymethylene complex [10]. Therefore, we rule out this pathway and anticipate that the hydrogenation of the iridium complexes with O-donor ligands proceeds as shown in Figure 6. For the hydrogenation process, we therefore propose the oxidative addition/reductive elimination pathway, which is in agreement with the suggested mechanism for the hydrogenolysis of a square-planar palladium alkoxido complex by Goldberg et al. (path A in Scheme 1) [1].

5. Conclusions

In this article, we analyzed the hydrogenation of square-planar PDI rhodium and iridium complexes with O,S-π-donors. The Ir-O,S metal ligand bonds display a partial multiple bond character with calculated large homolytic BDEs. This explains the reluctance of the thiolato complexes to undergo hydrogenation of their Ir-SR bonds. For the alkoxo complexes, on the other hand, this is overcompensated by the substantially stronger H-O bonds in the water and methanol product, and hence hydrogenation becomes possible. Furthermore, based on our extended theoretical study, we propose a two-step mechanism, i.e., (i) oxidative addition of H₂ to give an octahedral intermediate with a (ii) consecutive rate-determining O-H reductive elimination of methanol. We are currently investigating the hydrogenation of rhodium methoxido complexes and will provide a comprehensive report in the near future.