Hydrogenation of carbon dioxide to methanol using a homogeneous ruthenium–Triphos catalyst: from mechanistic investigations to multiphase catalysis

The hydrogenation of CO2 to methanol using a recyclable molecular organometallic catalyst in the absence of an alcohol additive is demonstrated for the first time.


General
Safety advice: High-pressure experiments with compressed gases represent a significant safety risk and must be conducted only following appropriate safety procedures and in conjunction with the use of suitable equipment.
For complex synthesis and catalytic experiments, moisture and oxygen were excluded by working in a glove box or by using Schlenk techniques. Argon 4.8 (Messer, Germany) was used as inert gas.
Glassware was dried under vacuum with a heat gun, evacuated and refilled with argon at least three times. All solvents were purified by distillation prior to use. Tetrahydrofuran, toluene and pentane were degassed by bubbling argon with a frit, dried by passing over activated alumina in steel columns and stored over molecular sieves. Acetonitrile, dioxane, and 2-methyltetrahydrofuran were degassed by distillation under argon and dried over molecular sieves. Water was degassed by bubbling argon with a frit. All reagents were commercially supplied by Aldrich, Fluka, Alfa Aesar and Heraeus Precious Metals and used as received unless stated otherwise. Reaction gases hydrogen (5.0) and carbon dioxide (4.6) were supplied by Linde and PraxAir and used without further purification. The complexes [(Triphos)Ru(TMM)] 2 [1] , [Ru( 2 -OAc)Cl(Triphos)] 9 [2] , [(Triphos)Ru(H) 2 CO] 12 [3] and [Ru 2 (-H) 2 (Triphos) 2 ] 6 [4] were synthesised according to literature procedures.
The 31 P{ 1 H}-NMR spectrum at room temperature showed two singlets at 43. 5

General procedure for CO 2 hydrogenation experiments
All high pressure batch experiments were conducted in stainless steel autoclaves (inner volume = 13 mL) equipped with a glass inlet and a magnetic stir bar. Prior to use, the autoclave was evacuated and repeatedly purged with argon. Under an argon atmosphere, catalyst 2 (0.025 mmol), together with HNTf 2 (1 eq.) was weighed into a Schlenk tube and dissolved in THF (2.08 mL). Alternatively, a THF solution of catalyst 4 (0.025 mmol) was prepared by heating [Ru( 2 -OAc)Cl(Triphos)] 9 (0.025 mmol, 20.5 mg) together with AgNTf 2 (1.1 eq., 0.0275 mmol, 10.7 mg) in 1 mL of THF for 3 h at 60 °C. The solution was filtrated over silica and the silica washed with 1 mL of THF. Alternatively, isolated catalyst 4 (0.025 mmol) could be used in THF (2.08 mL) and gave the same results.
In either case the solution was transferred via cannula to a stainless steel autoclave under argon atmosphere. The autoclave was pressurised with carbon dioxide to 20 bar and then hydrogen was added up to a total pressure of 80 bar. The reaction mixture was stirred and heated at 140 °C in an oil bath giving a total pressure of about 120 bar. After 24 h, the autoclave was cooled to 0 °C in an ice bath and then carefully vented. The resulting clear solution was analysed by 1 H-NMR (D1 = 10 s) in d 6 -DMSO with internal standard mesitylene and the results confirmed by gas chromatography using heptane as internal standard. Figure 11 shows a representative 1 H-NMR spectrum and Figure 12 a representative GC chromatogram. Figure 11: 1 H-NMR (300 MHz, d 6 -dmso) spectrum of the reaction solution from a CO 2 hydrogenation reaction using above described standard protocol (internal standard mesitylene).

Figure 12:
Representative gas chromatogram of the reaction solution from a CO 2 hydrogenation reaction using above described standard protocol (internal standard heptane).
The solution was analysed via NMR-spectroscopy (Figures 17-19). The same formate species 3a as observed in the CO 2 hydrogenation reaction was detected. The transmission IR-spectrum recorded in
HCO 2 H (0.9 L, 0.025 mmol, 1 eq.) was added via micro-syringe and the solution became orange. of the formate-ligand to methanol. In the corresponding 31 P{ 1 H}-NMR spectra the formation of 6 was observed in about 44 % of the total intensity besides some other, yet unknown species (Figure 24). In the hydride region of the 1 H-NMR-spectrum the signal corresponding to 6 (bs, -8.8 ppm) was observed. Besides that, two small signals at -6.7 ppm (bs) and -9.4 ppm (bs) corresponding to yet unknown species were observed.

In-situ NMR study of the CO 2 hydrogenation reaction using complex 2 with HNTf 2
HNTf 2 (3.5 mg, 0.0125 mmol) and complex 2 (9.8 mg, 0.0125 mmol, 1 eq.) were dissolved in d 8 -THF (0.5 mL) at room temperature giving a deep red coloured solution. 0.3 mL of this solution were transferred to a high-pressure NMR tube (volume = 0.93 mL). Of this solution 1 H-NMR and 31 P{ 1 H}-NMR spectra were recorded at 25 °C. After pressurising with 20 bar of carbon dioxide and 60 bar of hydrogen again spectra were recorded at room temperature. The NMR-tube was heated at 80 °C in the NMR-machine and 1 H-NMR and 31 P{ 1 H}-NMR spectra were recorded directly and again after 1, 2 and 4 hours at 80 °C. The time shift between recording the 1 H-NMR spectra and the 31 P{ 1 H}-NMR spectra was 30 minutes due to shimming and time for measuring the 1 H-NMR. After cooling to room temperature again a 1 H-NMR was recorded.
The recorded 31 P{ 1 H}-NMR spectra are shown in Figure 25. The spectrum recorded at 0 bar and 25 °C showed two singlets at 48.6 and 58.8 ppm and no signal due to the initial complex 2. We may speculate that the signal at 48.6 ppm is due to the cationic complex [Ru(Triphos)(methylallyl)] + resulting from protonation of the TMM ligand in complex 2, [5] whereas the signal at 58. 8  indicates that 3a is a resting state.
In Figure 27 the hydride areas of the recorded 1 H-NMR spectra are depicted. In the spectra after 1 and 2 hours a very small signal at -6.7 ppm was observed. However, this signal disappeared again after two further hours, making it very unlikely to play a role in the catalytic transformation of CO 2 and H 2 to methanol. The hydride signal at -8.8 ppm forming after 60 minutes is due to [Ru 2 (-H) 2 (Triphos) 2 ]

NMR-spectroscopic analysis of the reaction mixture using complex 12 together with HNTf 2
The catalytic hydrogenation of CO 2 to methanol using the carbonyl complex [(Triphos)Ru(H) 2

17.
After performing a CO 2 hydrogenation reaction using this mixture (25 mol complex 12, 1 equivalent In Figure 36 a magnification of the hydride region of the recorded 1 H-NMR spectra is depicted. No hydride species was observed in any of the 1 H-NMR spectra.

Computational Details
All calculations in this work were carried out with the Gaussian09 program series (Revision C.01 and D.01). 7 Gas phase calculations: The geometries of all structures were optimised with no constraints or restraints using the M06-L density functional 8 and the def2-SVP 9 basis set with the associated ECP 10 for ruthenium. The automatic density fitting approximation was activated. 11 The structures were characterised by frequency calculations to be local minima (i = 0) or transition states (i = 1). The obtained energies are listed in Table S2 below. For most of the optimised transition states IRC calculations were performed to ensure the transition state to connect the independently localised local minima. Thermochemical corrections were computed for standard state conditions. Additionally, single point energies were obtained on the M06-L def2-TZVP level of theory. The thermochemical corrections from the lowerlevel geometry optimisations were added to the electronic energies of the higher-level single-point calculations to arrive at corrected values for the Gibbs free energies. These are listed in Table S2 and are used for the discussion in the main text and in the figures in the main text. Figure 3 of the main manuscript is repeated here (Figure 38) to show three additional high energy pathways leading from I to VIII.

Solvent phase calculations:
While the study was ongoing Gaussian09 in Revision D.01 became available to us, which contains the MN12-L density functional. MN12-L has been tested to describe transition metal thermochemistry very accurately 12 . Accordingly, solvent phase calculations were calculated using MN12-L. The geometries of selected gas phase structures were optimised with no constraints or restraints using the MN12-L density functional and the def2-TZVP basis set 9 with the associated ECP 10 for ruthenium.
The automatic density fitting approximation was activated. 11 Solvent effects (THF) were considered implicitly by applying the IEF-PCM as well as the CPCM 13 formalism. The structures were characterised by frequency calculations to be local minima (i = 0) or transition states (i = 1).
Thermochemical corrections were computed for a temperature of 413.15 K. A pressure of 302 atm was specified to account for entropy corrections in the condensed phase as was described elsewhere. 14 The obtained energies are listed in Table S3 below.

Influence of the coordination geometry (facial vs. meridional)
To obtain an impression of how large the influence of the coordination geometry (fac or mer) is on one of the key steps of the catalytic cycle presented in the main manuscript we constructed a model ligand, being similar to Triphos chemically, however, with no interconnected carbon backbone to allow for the setup of calculations in the facial and the meridional coordination mode. When three of such P(Ph) 2 Table S2. Calculated energies of the optimised structures (M06-L/def2-SVP) and single point energies (M06-L/def2TZVP) in Hartree and relative Gibbs free energies G rel (kcal/mol) belonging to  in the main text and to Figure 38 and 39 in the ESI. Figure 3 main text + Figure  38

Comment on the endergonicity of the net reaction CO 2 + 3 H 2 -> H 3 COH + H 2 O
The M06-L/def2-TZVP//M06-L/def2-SVP values for the above mentioned net reaction generates a reaction Gibbs free energy of 14.5 kcal/mol, i.e. an endergonic reaction, which is counter intuitive with regard to the fact that the reaction should have a negative Gibbs free reaction energy. However, it should be noted that these computed values are values for gas phase reactions which for the general understanding and implementation of a transition metal complex catalysed reaction mechanism in most cases serve sufficiently well. While the small molecules CO 2 , H 2 , H 3 COH and H 2 O can easily be reoptimised in the solvent (see Table S3) this is by no means a fast approach for the reoptimisation of all of the molecules of the rather complex reaction pathway scenario outlined in Figures 3 to 5 of the main text and Figure 39 in the ESI. We have therefore restricted the general discussion to the gas phase energy profiles and refer the reader to the fact that a recomputation in the solvent phase most likely would generate energy profiles which are shifted relative to the ones being presented here, while the relative energy differences between the various intermediates and transition states will most likely not change markedly. However, for completeness we have recalculated the net reaction with the MN12-L density functional using the IEF-PCM and the CPCM continuum model and the IEF-PCM additionally with a radii model recently developed by the Truhlar group. 15 The results demonstrate that the reaction is exergonic in accordance with experimental observation and with standard state thermodynamics (see Table S3).
For comparison the gas phase energy values are also included in Table S3. Data were integrated with SAINT 16 and corrected for absorption by multi-scan methods. 17 The structure was solved by direct methods (SHELXS-97) and refined by full matrix least squares procedures based on F 2 as implemented in SHELXL-97. 18 One of the trifluoromethylsulfonate groups in the anion was disordered over