A prolific catalyst for dehydrogenation of neat formic acid

Formic acid is a promising energy carrier for on-demand hydrogen generation. Because the reverse reaction is also feasible, formic acid is a form of stored hydrogen. Here we present a robust, reusable iridium catalyst that enables hydrogen gas release from neat formic acid. This catalysis works under mild conditions in the presence of air, is highly selective and affords millions of turnovers. While many catalysts exist for both formic acid dehydrogenation and carbon dioxide reduction, solutions to date on hydrogen gas release rely on volatile components that reduce the weight content of stored hydrogen and/or introduce fuel cell poisons. These are avoided here. The catalyst utilizes an interesting chemical mechanism, which is described on the basis of kinetic and synthetic experiments.

In the drybox, a formic acid stock solution that was 0.005 mol% (50 ppm) in iridium precatalyst 1 and 5 mol% in sodium formate was prepared by dissolving 1.8 mg (2.6 μmol) of 1 and 180 mg (2.64 mmol) of sodium formate in 2.0 mL of formic acid. 0.5 mL of this stock solution was transferred into a 5 mL reaction flask possessing a large bore plug valve and a side arm. Water is added according to the amounts shown in the table above. Additional formic acid is added accordingly to make a 0.55 mL solution, which was used to measure the rates of formic acid decomposition using Procedure 2 (see Methods). A log/log plot of the rate of formic acid decomposition versus the water concentration yields a slope of 0.11 (5). Thus, water does not inhibit the rate of dehydrogenation. and NaO2CH (182.6 mg, 2.7 mmol) was added, in air, 0.5 mL of formic acid. The Schlenk test tube was connected to a three-way valve which was connected to a balloon. The reaction flask was heated to 90 °C. The gaseous products were collected in the balloon until almost all formic acid was decomposed and their composition analyzed by gas chromatography using a thermal conductivity detector. Three runs were performed. A 1:1 mole ratio of H2 and CO2 was detected along with small amounts of oxygen and nitrogen (figure A above, carbon monoxide was not detected; thus, the carbon monoxide level is < 0.099 v/v %). Figure B was obtained using a similar procedure with glove box and Schlenk techniques to exclude air.
Supplementary Figure 4 | A. IR spectrum of commercial CO2. B. IR spectrum of gaseous products from neat formic acid. A 50 mL Schlenk tube was charged with iridium precatalyst 1 (6.1 mg, 8.9 μmol) and sodium formate (184.0 mg, 2.7 mmol), which were dissolved, in air, in 1.0 mL formic acid. The Schlenk test tube was connected to a three-way valve, which was connected to a 2 L gas burette. The gas burette was fitted with a vent line connected to a three-way valve, which allows for direct sampling of the gaseous products for IR analysis. The reaction flask was heated to 90 °C. Gaseous products were collected until 1.1 L of gas was produced (ca. 5 h). An aliquot for IR analysis was transferred using a syringe into a Harrick temperature controlled gas cell (10 cm path length, 17 mL cell volume), which had been evacuated under reduced pressure. IR data (1 cm -1 resolution) was then obtained on a Bruker Vertex 80v FT-IR instrument equipped with a DTGS detector. Unlike a spectrum of commercial gaseous CO2 (A), CO can be observed in the IR spectrum of the gaseous reaction products (B). Using a calibration curve of different CO concentrations in air matrix, we estimated the CO concentration to be ca. 500 ppm. We also performed dehydrogenation of 5 mL of neat formic using the same catalyst loading over a period of 60 h. In this case, the CO concentration was estimated to be ca. 2000 ppm.

Supplementary Figure 5 | A. IR spectrum of the headspace of a reaction flask containing neat formic acid heated to 90 °C for 2 h.
It is known that neat formic acid decomposes in the presence of concentrated acid or at high temperatures to form H2O and CO 3,4,5 . We therefore hypothesized that much of the CO produced in our reaction conditions may be forming because of thermal, uncatalyzed decomposition of neat formic acid. To test for this situation, we dissolved 180 mg of sodium formate in 1 mL of formic acid and heated this solution in a sealed 100 mL pear flask for 2 h at 90 °C. The solution was then allowed to cool to room temperature and a sample of the headspace gases was analyzed by IR. Indeed, even after only 2 h of heating, copious amounts of CO (which was estimated to >5000 ppm) had formed as seen in the IR spectrum above.
Supplementary Figure 5 | B. Comparison of IR spectra of gaseous products from 90 v/v% FA/H2O and ca. 10 ppm CO in air matrix. Because it is known that neat formic acid thermally decomposes to produce H2O and CO, and that the presence of water decreases the rate of this decomposition 3,4,5 , we reasoned that we might be able to suppress CO production by using 10 vol% H2O/FA instead of neat formic acid as the dehydrogenation medium. Thus, a 50 mL Schlenk tube was charged with iridium precatalyst 1 (30.2 mg, 44.0 μmol) and sodium formate (181.0 mg, 2.7 mmol), which were dissolved, in air, in 1.0 mL formic acid. The mixture was allowed to sit overnight. The light orange homogeneous solution was then heated to 90 °C for 1 h forming 840 mL of gas. Gratifyingly, the IR spectrum of the gaseous products shows much lower levels of CO.
Comparing the CO signal from the gaseous products to an independently-prepared ca. 10 ppm CO solution in air, we estimate that the CO produced is less than 10 ppm. Figure 6 | Comparison of IR spectra of gaseous products from supersaturated sodium formate solution of neat FA heated at 70 °C for 6 h and ca. 10 ppm CO in air matrix. We reasoned that we might be able to decrease the CO production using neat formic acid by using solutions that are highly concentrated in sodium formate. Thus, a solution of 30 mg (44 μmol) of the iridium precatalyst and 900 mg of the sodium formate in 1 mL formic acid was prepared in the drybox and allowed to sit overnight. The mixture was then taken out of the drybox and exposed to air. The supersaturated, very viscous mixture was heated to 70 °C without stirring for 6 h, producing 550 mL of gaseous products. IR analysis of the gaseous products shows levels of CO below 10 ppm. In the drybox, a formic acid solution that is 5 mol% in sodium formate was prepared by dissolving sodium formate (901 mg, 13.2 mmol) in formic acid (10.0 mL, 265 mmol). To this solution was added the iridium precatalyst (18.2 mg, 26.2 μmol) to make a stock solution of the catalyst (2.52 mM based on monoiridium). The solution is allowed to sit overnight before use. The initial rates of formic acid decomposition were obtained using Procedure 2 (see Methods). An Eyring plot was constructed utilizing the measured turnover frequencies. From the Eyring plot, ∆H ‡ = +29.0(4) kcal mol -1 and ∆S ‡ = +16.0(10) eu.  Figure 16 except that formic acid concentration was varied instead of iridium concentration. A log/log plot of rate of formic acid decomposition versus formic acid concentration shows a reaction order of -0.94 (9). This indicates a reaction order of -1.0 with respect to formic acid.

Supplementary Tables
Supplementary

Supplementary Table 2 | Evaluation of the effect of different bases and different Lewis acids on formic acid dehydrogenation.
Formic acid stock solutions that were 0.005 mol% (50 ppm) in [Ir atom] and 5 mol% in base were prepared in the drybox by adding measured amounts of the iridium catalyst (1.8 mg, 2.6 μmol) and base (see below for exact amounts) into a vial. Formic acid (2.0 mL) was then added to dissolve the base and the catalyst. These experiments were aimed to determine the effect of different bases on the rate of dehydrogenation. In addition, these experiments allowed evaluation of the effect on dehydrogenation by the Lewis acids Li + , Na + , K + , and Ca 2+ . The initial rates of formic acid decomposition (average of two runs) were obtained using Procedure 2 (see Methods). According to the table, base is required for dehydrogenation to occur efficiently. However, the types of bases and Lewis acids do not have an effect on the rate of dehydrogenation. One outlier, the data for calcium carbonate, gives a much lower decomposition rate because copious amounts of calcium formate precipitates.

Supplementary Table 3 | Initial rates of dehydrogenation in the presence and absence of air.
A formic acid stock solution that is 0.005 mol% in the iridium precatalyst and 5 mol% in sodium was prepared in air by adding 1.8 mg (2.6 μmol) of the iridium precatalyst and 180 mg (2.64 mmol) of sodium formate into a vial. In air, formic acid (2.0 mL) was then added to dissolve the base and the precatalyst. Dehydrogenation rates were obtained 1 day and 2 weeks after preparation of the stock solution using Procedure 2 (see Methods). Under otherwise identical conditions, the dehydrogenation of reaction mixtures prepared in air is ca. 3 times slower.

Supplementary Table 5 | High turnover number experiments.
In the drybox, a 5 mL high pressure reaction flask with a side arm and a large bore plug valve was charged with sodium formate (0.18 g, 2.64 mmol), which was then dissolved in 0.6 mL of formic acid. A stock solution of the iridium precatalyst (which is actually a stock solution of the iridium dimer catalyst) in formic acid was added (100 μL, 1.26 mM precatalyst stock solution, 0.13 μmol precatalyst). The reaction flask was taken out of the drybox and connected to a vent line leading to a 1 L gas burette filled with oil. The flask was heated to 90 o C for 24 h. After 24 h, 420 mL of gas was produced, corresponding to 8.6 mmol of formic acid decomposed and a turnover number of 67,615. At the end of the reaction, a white solid residue remains at the bottom of the flask. Not all of the formic acid is decomposed because some of the formic acid ends up on the neck and the side arm of the flask and does not mix with the catalyst. Nevertheless, we can recharge the reaction flask with formic acid. To do this, we disconnect the sealed reaction flask from the gas burette, clean the side arm with acetone, and take the flask back into the drybox to be refilled with formic acid. Then we repeat this procedure through 40 cycles. After 40 cycles over a period of 4 months, 13.71 L of gas were produced. This corresponds to 0.

Observation of catalytic intermediates by NMR.
Data regarding the elementary steps of the conversion of the precatalyst to the active dimer catalyst were obtained using NMR. We can observe the formation of a Pfaltz-type dimer 2 from precatalyst 1 in two different ways: 1) reaction of 1 with formic acid in a coordinating solvent such as acetonitrile or 2) reaction of 1 with H2 in various solvents. In formic acid solvent, iridium dimer 2 is converted further into the (di--formate)iridium dimer 3a.

Observation of Intermediate 2 in CD3CN
Room temperature 1 H NMR studies in CD3CN show that addition of 1 equivalent of sodium formate and 10 equivalents of formic acid to a solution of iridium precatalyst 1 leads to formation of a new species (intermediate A) with a hydride signal at -19.43 ppm ( Supplementary  Figures 7-8). Data for this species is consistent with oxidative addition of formic acid to the iridium precatalyst. 20 minutes after addition of sodium formate and formic acid, a time course experiment was collected over a period of 66 min (Supplementary Figure 8). The resulting stacked 1 H NMR spectra shows that intermediate A grows then intermediate B appears ca. 30 min after addition of sodium formate and formic acid. Intermediate B is consistent with a species where one of the cyclooctadiene double bonds is bound to iridium and the other is free. Then the final product (a Pfaltz-type dimer) appears ca. 1 hour after addition. The 1 H NMR spectrum of the final product is consistent with dimer 2 (Supplementary Figure 9; S = CD3CN). The cyclooctadiene in the iridium precatalyst is reduced to cyclooctene and free cyclooctene is seen in the 1 H NMR spectrum of the products. Although 3 does not form in CD3CN, it is reasonable to conclude that formation of the catalyst resting state (diformate 3a) proceeds in a similar fashion through the intermediacy of 2 when formic acid or tetraglyme is used as solvent. It is worth noting that, in CD3CN, iridium 1 reacts in the absence of base with formic acid to form 2, but in a much slower rate.
The formation of Pfaltz-type dimer 2 is also observed upon reaction of precatalyst 1 with hydrogen. Supplementary Figure 10 shows the Pfaltz type dimer and cyclooctene formed from this reaction in CD3CN solvent. Cyclooctene is slowly hydrogenated when the reaction mixture is allowed to sit at room temperature.

Observation of Intermediate 3 in Formic Acid
We studied the catalyst resting state(s) in formic acid by dissolving, in the drybox, 10.0 mg of the iridium precatalyst and 10.0 mg of sodium formate in 1.0 mL of proteo formic acid in a J-Young tube. The J-Young tube was taken out of the drybox and was connected to a three-way valve, which was connected to a nitrogen line and a 50.0 mL gas burette. The tubing and the gas burette were purged with nitrogen for ca. 15 minutes. The plug valve and the three-way valve were opened such that gas produced go directly to the gas burette. The reaction flask was heated in an oil bath to 70 o C. About 15 mL of gas was produced during heating. The solution was then allowed to cool to room temperature. The proteo formic acid solvent was evaporated under high vacuum and the resulting residue was left under high vacuum overnight. The residue was then dissolved in formic acid-d2. The 1 H NMR spectrum of the residue is shown in Supplementary  Figure 11. Examination of the hydride region reveals two resting states: (A) a minor resting state with two different hydrides at -27.24 and -28.51 ppm that integrate in a 2:1 ratio (consistent with the hydrides in the 1 H NMR of 3b, whose X-ray structure is known). We formulate this species as dimer 3a. (B) The major resting state contains three different hydrides at -19.41, -25.05, and -27.13 ppm. We formulate this species as dimer 4 (see Figure 4).

Proton-Hydride Fidelity Experiment
In the drybox, a J-Young tube was charged with the iridium precatalyst (1.8 mg, 2.64 μmol), sodium formate (1.7 mg, 25.0 μmol), formic acid-d1 (O-D; 5 μL, 0.13 mmol), and 0.6 mL of methanol-d4. A 1 H NMR time course experiment spanning over 72 minutes was performed at 70 °C. Before the time course experiment was begun, the NMR tube had been heated ca. 10 minutes at 70 °C and much H-D had formed (Supplementary Figure 12). Nevertheless, formation of predominantly H-D with a small amount of H2 is observed. This observation is consistent with a formic acid dehydrogenation mechanism that proceeds through formation of an iridium monohydride where the hydride comes from the formyl C-H bond of formic acid. This iridium monohydride is then protonated (or, technically, deuterated) by the formic acid O-D. We expect that an iridium dihydride that undergoes reductive elimination to yield H2 should enable scrambling of proton and hydride, thus we disfavor this possibility. The small amount of H2 that forms can be rationalized by the presence of small amounts of formic acid O-H bonds which protonates the iridium monohydride. Interestingly, the HD signal disappears after extended heating, which is consistent with a slow reverse reaction when the reaction is run in a sealed vessel. Such back reaction is impossible under our kinetics acquisition conditions because the H2 product is sequestered and quantified in an eudiometer.

General Experimental Information
All air and water sensitive procedures were carried out either in a Vacuum Atmosphere glove box under nitrogen (2-10 ppm O2 for all manipulations) or using standard Schlenk techniques under nitrogen. Dichloromethane-d2, acetonitrile-d3, methanol-d4, water-d2, formic acid-d1 (C-D), and formic acid-d1 (O-D) NMR solvents and reagents were purchased from Cambridge Isotopes Laboratories. Formic acid-d2 was purchased from SynQuest Laboratories. Dichloromethane, ethyl ether, and hexanes were purchased from VWR and dried in a J. C. Meyer solvent purification system with alumina/copper(II) oxide columns; chloro(1,5cyclooctadiene)iridium(I) dimer (Strem), sodium trifluoromethanesulfonate (Sigma-Aldrich), and formic acid (Sigma-Aldrich) were purchased and used as received; 2-((di-tbutylphosphino)methyl)pyridine was synthesized using a literature procedure. 18 NMR spectra were recorded on a Varian VNMRS 500 or VNMRS 600 spectrometer. All chemical shifts are reported in units of ppm and referenced to the residual 1 H or 13 C solvent peak and line-listed according to (s) singlet, (bs) broad singlet, (d) doublet, (t) triplet, (dd) double doublet, etc. 13 C spectra are delimited by carbon peaks, not carbon count. 31 P chemical shifts are referenced to an 85% phosphoric acid external standard. Air-sensitive NMR spectra were taken in 8" J-Young tubes (Wilmad or Norell) with Teflon valve plugs. MALDI mass spectra were obtained on an Applied Biosystems Voyager spectrometer using the evaporated drop method on a coated 96 well plate. The matrix was 2,5-dihydroxybenzoic acid. In a standard preparation, ca. 1 mg analyte and ca. 10 mg matrix was dissolved in methanol and spotted on the plate with a glass capillary. X-ray crystallography data were obtained on a Bruker APEX DUO single-crystal diffractometer equipped with an APEX2 CCD detector, Mo fine-focus and Cu micro-focus X-ray sources. Elemental analysis data were obtained on a Thermo Flash 2000 CHNS Elemental Analyzer. Gas chromatography data were obtained on a Thermo gas chromatograph (Supelco Carboxen ® -1010 plot, 30 m × 0.53 mm) equipped with a TCD detector (detection limit: 0.099 v/v %). IR data (1 cm -1 resolution) were obtained on a Bruker Vertex 80v FT-IR using a deuterated triglycine sulfate (DTGS) detector.

Visual Appearance
The reaction mixture is a translucent pale yellow solution during catalysis, with no dark precipitates forming. At the end of the reaction, a pale orange solid (the catalysts) remains.

Well Behaved Kinetics
The reaction follows well-behaved saturation catalysis kinetics through its pseudo-zero order region, which persists until solvent level drops appreciably as solvent is consumed (Supplementary Figure 1).

Mercury Drop Test
The mercury drop test was performed by placing 0.5 mL of a formic acid stock solution with 0.01 mol% in iridium precatalyst 1 and 5 mol% sodium formate into a 5 mL reaction flask possessing a large bore plug valve and a side arm. A drop of mercury was then added to this solution. The rate of formic acid decomposition (average of two runs) was then measured to be 1.24 ×10 -7 mol s -1 at 90 °C, which is just slightly slower compared to the measured rate of formic acid decomposition in the absence of mercury (1.35 ×10 -7 mol s -1 ).

Quantitative Poisoning
Phenanthroline was utilized as catalytic poison. 1.00 mL (1.26 μmol iridium) of a formic acid solution that is 1.26 mM in iridium atom and 1.26 M in sodium formate was placed in a 2 mL volumetric flask. One-half mole equivalent of the poison relative to the iridium catalyst was added as a stock solution in formic acid (100 μL of 6.3 mM solution, 0.63 μmol). The solution was diluted to 2.00 mL. Using 0.5 mL of this solution, the rate of formic acid decomposition (average of two runs) was measured to be 1.27 ×10 -8 mol s -1 at 86 °C, which is 62% of the formic acid decomposition rate in the absence of phenanthroline (which was measured as 2.04 ×10 -8 mol s -1 ).

Kinetic Isotope Studies
To obtain values for kH/kD, the initial rates of dehydrogenation using formic acid, formic acid-d2, formic acid-d1 (O-D), and formic acid-d1 (C-D) were obtained.
Stock solutions of formic acid, formic acid-d2, formic acid-d1 (C-D), and formic acid-d1 (O-D) that are 0.005 mol% (50 ppm) in iridium precatalyst 1 and 5 mol% in formate were prepared. In the drybox, a vial was charged with 1.8 mg (2.6 μmol) of the iridium precatalyst and 180 mg of either sodium formate (for the FA and FA-d1 (O-D) solutions) or sodium formate-d1 (for the FA-d2 and FA-d1 (C-D) solutions). The appropriate proteo/deutero formic acid (2.0 mL) was then added to dissolve the base and the precatalyst. The solution was allowed to sit overnight before use. The initial rates of formic acid decomposition were obtained using Procedure 2 (see Methods). See main text for the obtained kinetic isotope effects.