Main

The improvement of our current energy technologies in terms of sustainability for future human development represents one of the global challenges of the twenty-first century1,2. With the aim of replacing fossil fuels and decreasing the anthropogenic emissions of CO2, worldwide efforts to utilize renewable energy are currently under way. Both the Paris and more recent Glasgow Agreements aim to avoid global warming and mitigate climatic changes via net-zero CO2 emissions. The necessary paradigm shift of our present energy system can be achieved through the transformation of renewable wind, solar or hydro power to better storable chemical fuels3,4. In this respect, hydrogen is considered to be a promising chemical energy carrier5 that could be produced from renewable resources, for example, via electrochemical water splitting6,7. Indeed, the global hydrogen market is expected to reach around US$700 billion by 20508,9. However, its chemical and physical properties, for example, its low volumetric energy content at ambient conditions (0.0023 kWh l−1), its flammability and its explosive nature in oxygen-containing mixtures make its handling, storage and transportation cumbersome, energy-intensive and expensive, especially for long-term/long-distance applications10,11,12,13. Both the development of a chemical hydrogen economy and the exploration/discovery of alternative hydrogen energy carriers thus continue to attract a substantial amount of attention14,15,16,17,18,19,20,21,22,23,24,25,26.

An ideal chemical hydrogen source should fulfil the following requirements: it should have high gravimetric and volumetric energy density, low toxicity, fit the present infrastructure, be practical in terms of handling and transportation, and the hydrogen should be cost- and energy-efficient to store as well as release. However, none of the current chemical hydrogen carriers fulfil all these prerequisites. For example (Fig. 1a), methanol22 (MeOH) and ammonia27 are available on a bulk scale and possess a high hydrogen content, but they are classified as toxic and flammable according to the globally harmonized system of classification and labelling of chemicals (GHS[EC]) of the United Nations (Fig. 1d)28. Liquid organic hydrogen carriers (LOHCs)29 based on arenes have medium hydrogen densities (5.8–7.3 wt%) and can be easily handled, but are less available and have toxicity problems (Fig. 1d). Formic acid (FA)17 is directly available from CO2, has favourable thermodynamic data (Fig. 1c) and can be easily dehydrogenated even under ambient conditions, but it has a low hydrogen content (<5 wt%) and is corrosive (Fig. 1d). This means that there is still strong interest in new practical hydrogen vectors that avoid the abovementioned problems. Methyl formate (MF) drew our particular attention. As shown in Fig. 1d, MF has a hydrogen storage capacity (8.4 wt%) between those of MeOH (12.1 wt%) and FA (4.4 wt%) and comparable to other LOHCs. More specifically, the volumetric energy density of MF is equivalent to pressurized hydrogen at 1,200 bar. Notably, its dehydrogenation is thermodynamically favoured compared to other hydrogen carriers, except FA (Fig. 1c). Furthermore, MF is classified as non-toxic, non-irritating and non-corrosive (Fig. 1d), and it can easily be transported, refuelled and handled. At present, MF is produced industrially from MeOH carbonylation, and the global capacity of MF was >6 million metric tonnes per annum in 201630,31. In recent years, many research groups, including industry, have demonstrated the possibility of accessing MF through the hydrogenation of CO2 in the presence of MeOH, which can also be generated utilizing CO2 (refs. 32,33,34,35,36,37,38,39,40,41). Overall, this will allow for sustainable MF synthesis.

Fig. 1: Selected hydrogen energy systems.
figure 1

a, Energy carriers based on CO2 (the values in tonnes represent annual output from industry and nature emission). b, Using MF as a chemical hydrogen energy carrier (red indicates this work). c, Selected catalytic dehydrogenation reactions (red indicates this work). d, Properties of different hydrogen energy carriers. aThe data are based on the actual dehydrogenation chemical equations in c .

Full size image

Taking all these facts into account, MF seems to be a promising hydrogen carrier, although, surprisingly, and to the best of our knowledge, it has not been proposed or described for this purpose yet. In addition, no precedent for MF dehydrogenation has been reported before (Fig. 1b).

Results

Catalytic MF dehydrogenation

Hydrogen generation from MF is possible in the presence of a number of ruthenium pincer catalysts, as found in previous works on FA and MeOH dehydrogenation17,21,22,42,43,44,45. Among the tested catalysts (Table 1; for gas evolution curves see Supplementary Table 2), complexes C1–C5, including commercial ones, showed high productivity, with a hydrogen turnover number (TON(H2)) of up to 21,500, and high activity, with a hydrogen turnover frequency (TOF(H2)max) of up to 8,300 h−1. It is noteworthy that the CO content in the produced gas was <10 ppm when using C1–C5, which indicates a highly selective conversion of MF to hydrogen and CO2. C2 was chosen as the catalyst for further investigation due to its good performance and commercial availability. The reaction parameters, such as the amount of water, solvent, type and amount of base, and reaction temperature were studied in more detail (Supplementary Tables 3–7). First, the amount of water was varied (Supplementary Table 3), finding an optimal water range of 2–4 equiv. to MF. Notably, an insufficient amount of water led to more CO generation. Second, with the exception of acetonitrile, the use of co-solvents, such as tetrahydrofuran (THF), dioxane, N,N-dimethyloctylamine (DMOA) and dimethylformamide (DMF), provided similar hydrogen volumes (Supplementary Table 4). Regarding the base, KOH, NaOH, CsOH, K2HPO4 and K3PO4 are all appropriate for MF dehydrogenation; among these, KOH and K3PO4 performed best (Supplementary Table 5). Varying the amount of KOH in the range 0–60 mmol revealed an optimal range of 10–40 mmol, which also showed the necessity of the base (Supplementary Table 6). Finally, the reaction temperature was varied from 23 °C to 110 °C (Supplementary Table 7). A temperature between 90 °C and 100 °C led to the best gas evolution, with a CO content below 10 ppm. Notably, MF dehydrogenation is possible even at room temperature. In summary, using MF as a hydrogen carrier allows for hydrogen release under a variety of conditions, which makes it attractive for a number of applications.

Table 1 Ru-catalysed MF dehydrogenation
Full size table

Next, we compared the gas evolution rate using MF with established hydrogen energy carriers. MeOH and FA were dehydrogenated under identical mild conditions (Fig. 2). Remarkably, the initial gas evolution rate using MF is five times faster than that with FA and 20 times faster than with MeOH. These unexpected results prompted us to conduct a more in-depth study of the mechanism of MF dehydrogenation.

Fig. 2: Comparison of MeOH, FA and MF dehydrogenation.
figure 2

Plot of gas evolution (top) and details of the different hydrogen energy carriers and reaction conditions (bottom).

Full size image

Mechanistic investigations of MF dehydrogenation

First, the rate of MF dehydrogenation was studied systematically at temperatures between 60 and 90 °C. A linear Arrhenius plot was generated, resulting in an estimated activation energy, Ea, of 65 kJ mol−1 (Fig. 3a). To understand the initial activation step for MF, H2O was replaced by D2O, and a secondary kinetic isotope effect (KIE) of kH/kD = 1.40 was observed (Fig. 3b, entry 2). Notably, a similar KIE (kH/kD = 1.59) was found when using DCOOCH3 instead of HCOOCH3 (Fig. 3b, entry 3). A slightly higher KIE (kH/kD = 1.80) was obtained with the combination of DCOOCH3 and D2O (Fig. 3b, entry 4). These KIE measurements indicate that the formyl C–H group in MF can be more easily activated compared to the C–H bonds (KIE > 2) in other established hydrogen carriers (MeOH, FA and other LOHCs)16,17,21,24,46,47,48,49. To understand this special reactivity, the details of MF dehydrogenation on a molecular level were derived from density functional theory (DFT) calculations. As shown in Fig. 3c and Supplementary Figs. 4–7, starting from the active [KRu–OH] complex with nucleophilic attack of MF, intermediate [KRu–O(H)CHOOCH3] is generated through transition state TS-MF-OHK. After that, hydrogen evolution might occur via two routes: MF direct dehydrogenation (Fig. 3c, red) or MF hydrolysis followed by dehydrogenation (Fig. 3c, blue). In the direct dehydrogenation case, H2 release proceeds through TS-CHK with a free energy barrier of 80 kJ mol−1 to form the intermediate [KRu–OCOOCH3], followed by CO2 release through TS-CO2, K to generate the intermediate [KRu–OCH3]. The alternative MF hydrolysis route resulting in [KRu–OOCH] and CH3OH formation via transition state TS-COK has a free energy barrier of 46 kJ mol−1. The following CO2 and H2 release via decomposition of [KRu–OOCH] is determined by TS-H2/H2OK with a free energy barrier of 124 kJ mol−1, which also leads to the same intermediate [KRu–OCH3]. Based on these calculations, we propose that H2 and CO2 are generated by a direct MF dehydrogenation pathway via the [KRu–OCOOCH3] intermediate rather than [KRu–OOCH] due to the substantially lower effective energy barrier than that of the MF hydrolysis route (80 versus 124 kJ mol−1). Although the formation of [KRu–OOCH] via the MF hydrolysis route is more favourable than the formation of [KRu–OCOOCH3] via the MF direct dehydrogenation route, both thermodynamically and kinetically, the release of H2 is the driving force for [KRu–OCOOCH3] formation. Moreover, the decomposition of [KRu–OCOOCH3] is kinetically more favoured than that of [KRu–OOCH] (54 versus 124 kJ mol−1). By using the Eyring and Arrhenius equations, the activation energies were calculated based on enthalpy barriers (Supplementary Fig. 7). As shown in Supplementary Table 10, the calculated Ea via the direct dehydrogenation mechanism (Ea = 61 kJ mol−1) matches the experimentally obtained value of 65 kJ mol−1, which is much lower than that via MF hydrolysis route (Ea = 121 kJ mol−1). Furthermore, the dehydrogenation of [KRu–OCH3] (free energy barrier of 153 kJ mol−1) is more difficult compared to hydrogen release from [KRu–OCOOCH3]. Detailed DFT calculations for MF dehydrogenation in the absence of a base showed the same trend that the direct MF dehydrogenation pathway is more favourable than the MF hydrolysis route (95 versus 129 kJ mol−1), but the effective energy barrier is higher than that under basic conditions by 15 kJ mol−1 (Supplementary Figs. 8–11). This is also qualitatively consistent with the experimental results (Supplementary Table 6 entries 1, 3 and 9) that the volume of released H2 ((V(H2)) for MF as educt in the presence of a base is higher than without a base (3,645 versus 1,820 and 1,980 ml), as well as for HCOOH (359 ml) or MeOH (311 ml) as educts in the presence of a base (Fig. 2 and Supplementary Table 8).

Fig. 3: Mechanistic investigations of MF dehydrogenation.
figure 3

a, Arrhenius plot. b, KIE data. For the curve slopes, see Supplementary Table 9. c, Molecular details of the DFT calculations. E, PPh2. Red colour route: MF direct dehydrogenation. Blue colour route: MF hydrolysis followed by dehydrogenation. d, Solid-state structure of the key intermediate [H–Ru–OCOOCH3] without carbon-bound hydrogen atoms. The displacement ellipsoids correspond to 30% probability. e, Dehydrogenation comparison using the intermediate crystal [H–Ru–OCOOCH3] and its corresponding catalyst precursor C1.

Full size image

According to the DFT calculations, both [K–Ru–OCOOCH3] (in the presence of base) and [H–Ru–OCOOCH3] (base-free conditions) are thermodynamically stable intermediates in the proposed direct MF dehydrogenation pathway. Indeed, it was possible to obtain crystals of the corresponding [H–Ru–OCOOCH3] complex that were suitable for X-ray analysis (Fig. 3d). When performing the MF dehydrogenation with this defined complex, a pressure curve was observed that is similar to the one using precursor C1, supporting the existence of such active intermediates (Fig. 3e).

To detect other reaction intermediates and further verify the mechanism, a detailed NMR study of MF dehydrogenation was performed. The neutral complex [H–Ru–OH] was prepared by mixing complex C1 with tBuOK, followed by celite filtration and adding H2O (Supplementary Fig. 25). Stoichiometric amounts of 13C-labelled MF (H13CO2CH3) were then added, and the reaction mixture was analysed by NMR (Supplementary Fig. 34). In agreement with the calculations, the thermodynamically most stable intermediates [H–Ru–OO13CH] and [H–Ru–O13COOCH3] were detected as major products. In addition, 13CO2, H13CO2CH3, HCO213CH3, H13CO213CH3, HCO2CH3, 13CH3OH and CH3OH were all observed in solution as well as H2 in the gas phase after the reaction (Supplementary Figs. 30–33). The various 13C-labelled products are explained by the reversible hydrogenation and dehydrogenation reactions, which formally lead to a scrambling of 13C among MF and MeOH (ref. 50). Interestingly, complex [H–Ru–OCH3], which was assumed to be an important intermediate, was undetectable in the reaction mixture. To understand this observation, a stoichiometric experiment utilizing complex [H–Ru–OCH3] was performed at room temperature (Supplementary Fig. 35). After adding H2O, complex [H–Ru–OH] was detected as well as [H–Ru–OCH3]. Both complexes disappeared immediately upon further addition of MF, resulting in the immediate formation of complexes [H–Ru–OOCH] and [H–Ru–OCOOCH3] (Supplementary Fig. 42). All these NMR studies are in good agreement with the DFT calculations and strongly support the proposed direct MF dehydrogenation pathway.

Finally, a time-resolved analysis of the reaction revealed the consumption of a significant amount of MF (23%), even during the initial heating from room temperature to 90 °C, leading to H2, CO2, MeOH and FA/formate (Fig. 4a). During this period, the pH changed from 10.9 to 7.4 and then remained constant between 6.8 and 7.6 due to the formation of a buffer system. With ongoing reaction, MF was continuously consumed and the amounts of hydrogen, CO2 and MeOH were further increased. A maximum amount of FA/formate was detected between 1 and 5 h.

Fig. 4: Mechanistic investigations of MF dehydrogenation.
figure 4

a, Time course study of compound distribution in MF dehydrogenation. b, The proposed catalytic cycle of MF dehydrogenation.

Full size image

All the mechanistic investigations, including KIE measurements, DFT calculations, crystallization of intermediate [H–Ru–OCOOCH3], NMR studies and time-resolved analysis, agree with the catalytic cycle shown in Fig. 4b. Initially, under basic conditions, formation of the active species I takes place, which, after nucleophilic attack of MF, hydride transfer and H2 elimination, leads to key intermediate II. After release of CO2, complex III is formed. Subsequently, methoxy and formate group dehydrogenation occur and finally the active catalyst I is regenerated. In addition, I or II can be formed from III or IV by ligand exchange or via MeOH-assisted MF formation, respectively.

In general, by using 1 mol aqueous MF as a hydrogen carrier, up to four moles of hydrogen and two moles of CO2 can be generated (Fig. 1c, equation in red). To prove this, and to demonstrate the stability of our catalyst system, we performed several consecutive runs of MF dehydrogenation and gas release by applying 84 mmol MF/168 mmol H2O (Supplementary Table 19). As the dehydrogenation reaction reaches an equilibrium state in a closed system, the formed gases were released after reaching a pressure plateau. The volumes and gas constitution were measured after each run. The catalyst system was stable for more than 25 consecutive runs, resulting in a gas release of >4.3 l that contained 2.5 l (103 mmol) H2, corresponding to a catalyst TON of >25,000. In an additional long-term experiment, >9.4 l of gas, overall, was released, with 60% H2 yield (241 mmol) based on MF (100 mmol, 6.1 ml; Supplementary Table 20). The detected ratio of H2 to CO2 was 1.7:1, which is near the expected ratio (H2/CO2 = 2:1) for MF aqueous reforming as in Fig. 1c. Apart from a continuous supply of hydrogen at low pressure, MF can be used advantageously for the rapid generation of high-pressure hydrogen, which is of specific interest for combustion or electric engines51. Using 310 mmol MF in the presence of only 6.5 ppm of catalyst C5 resulted in a remarkable pressure (70 bar (2 h) and 128 bar (10 h)) (Supplementary Table 21). This corresponds to a catalyst TON(H2) of >107,000 and TOF(H2)max of >44,000 h−1. Notably, CO was undetectable in this reaction (Supplementary Fig. 45) and the produced hydrogen could be used directly52. Finally, directly using a mixture of the educts MF and H2O, solvent-free dehydrogenation was performed in a closed autoclave (Supplementary Table 22). Remarkably, a pressure of more than 75 bar was obtained due to gas evolution in the presence of KOH and 25-ppm Ru-catalyst with TON(H2) > 16,871.

In conclusion, we propose MF as a hydrogen storage material that allows for a carbon-neutral hydrogen energy cycle. Due to its physical and chemical properties, MF complements currently discussed chemical energy carriers. It is available as a multi-million-tonne-scale annual output, has good hydrogen density and is classified as non-toxic, non-irritating and non-corrosive (Fig. 1d). Furthermore, MF can be easily transported, refuelled and handled. In the presence of an appropriate catalyst system, the gas evolution from aqueous MF reforming proceeds five times faster than with FA and 20 times faster than with aqueous MeOH under identical mild conditions. The developed optimal Ru-pincer complex for MF dehydrogenation is highly selective (CO undetectable) and highly active, with TOF(H2)max > 44,000 h−1 and TON(H2) > 100,000. Moreover, solvent-free MF dehydrogenation also demonstrates its application potential. Hydrogen generation from MF proceeding by this mechanism is supported by KIE measurements, DFT calculations and X-ray crystal structure and NMR studies.

Methods

Materials and characterization methods

All catalytic experiments were carried out under an Ar or N2 gas atmosphere with exclusion of air. All liquid reagents were degassed or distilled before use and stored under Ar. All liquid reagents were protected by Ar after distillation or degassed with three freeze–thaw cycles using liquid N2. Chemicals were purchased from Aldrich, TCI, Alfa, Fisher Chemical, Abcr, Deutero, Eurisotop and Cambridge Isotope Laboratories. Catalysts C1–C4, C6 and C7 were purchased from Stem Chemicals. Catalyst C5 was synthesized according to the procedure reported in ref. 42. Catalyst C8 was synthesized according to the procedure reported in ref. 53. Air- and moisture-sensitive syntheses were performed under an Ar atmosphere in heating-gun vacuum-dried glassware. The liquid and solid products were characterized by 1H NMR, 13C NMR and 31P NMR spectroscopy. The NMR spectra were recorded on Bruker Avance 300 (300 MHz) or 400 (400 MHz) NMR spectrometers. Quantitative 13C NMR measurements were performed with a Bruker AV 400-MHz spectrometer, and the analysis time for each sample was no less than 1.5 h. MestReNova (version 14.0.1-23559) was used for interpreting and processing the NMR spectra. Gas chromatography (GC) analysis was performed on an Agilent Technologies 7890A GC system (HP Plot Q/FID, hydrocarbons, Carboxen/TCD, permanent gases; Ar carrier gas), with a CO quantification limit of 78 ppm, and on an Agilent Technologies 7890A GC system (HP Plot Q/FID, hydrocarbons, Carboxen/TCD, permanent gases; He carrier gas), with a CO quantification limit of 10 ppm. pH values were measured on a laboratory digital pH meter (Mettler Toledo AG, SevenEasy pH 8603) at room temperature (24 °C).

Calculation of the hydrogen volume, mole, yield, TON and TOF

The GC was calibrated with certified commercially available gas mixtures. GC samples were taken from the collected gas in the burette after every reaction to obtain the hydrogen percentage (\({\mathrm{{GC}}}_{{\rm{H}}_{2}}\)) and CO2 percentage (\({\mathrm{{GC}}}_{{{\rm{CO}}}_{2}}\)) in the total gas. The amounts of H2 and CO2 (n) in mmol were calculated according to

$${n}_{{\rm{H}}_{2}}={\frac{{V}_{{\rm{gas}}}\times {\mathrm{{GC}}}_{{\rm{H}}_{2}}}{{V}_{\rm{m},{\rm{H}}_{2},25^\circ {\rm{C}}}}}$$
(1)
$${n}_{{{\rm{CO}}}_{2}}={\frac{{V}_{{\rm{gas}}}\times {\mathrm{{GC}}}_{{{\rm{CO}}}_{2}}}{{V}_{{\rm{m,}}{{\rm{CO}}}_{2},25^{\circ }{\rm{C}}}}}$$
(2)

The H2 TON and TOF were calculated by

$${\rm{TON}}_{{\rm{H}}_{2}}={\frac{{n}_{{\rm{H}}_{2}}}{{n}_{{\rm{cat}}}}}$$
(3)
$${\rm{TOF}}_{{\rm{H}}_{2}}={\frac{{n}_{{\rm{H}}_{2}}}{{n}_{{\rm{cat}}}\times t}}$$
(4)

where Vgas is the gas volume corrected by the blank volume, and the calculation of standard gas molar volume \({V}_{\mathrm{m},{\rm{H}}_{2},{25}^\circ {\rm{C}}}\) and \({V}_{\mathrm{m},{{\rm{CO}}}_{2},{25}^\circ {\rm{C}}}\) were carried out using

$${V}_{\mathrm{m},{\rm{H}}_{2},{25}^\circ {\rm{C}}}={\frac{{RT}}{p}+b-\frac{a}{{RT}}=24.48\,{\rm{l}}\,{{\rm{mol}^{-1}}}}$$
(5)
$${V}_{\mathrm{m},{{\rm{CO}}}_{2},{25}^\circ {\rm{C}}}={\frac{{RT}}{p}+b-\frac{a}{{RT}}=24.36\,{\rm{l}}\,{{\rm{mol}^{-1}}}}$$
(6)

where R = 8.3145 m3 Pa mol−1 K−1, T = 298.15 K, p = 101,325 Pa, a(H2) = 24.7 × 10−3 Pa m6 mol−2, a(CO2) = 36.5 × 10−2 Pa m6 mol−2, b(H2) = 26.6 × 10−6 m3 mol−1 and b(CO2) = 42.7 × 10−6 m3 mol−1.

General measurement of MF dehydrogenation

All experiments were performed under an inert atmosphere (N2 or Ar) with exclusion of air. An amount of one Ru-catalyst with a defined amount of base was added in an autoclave (pressure tube) under an ice bath, followed by certain amount of solvent, MF and H2O injected by syringe. Next, the autoclave was flushed with N2 (5 bar) and the pressure was released three times. The reaction was performed at a set temperature for a certain number of hours. The autoclave was then cooled using an ice-bath, and the pressure was carefully released through a room-temperature (25 °C) condenser (the water temperature was controlled by a thermostat) to a manual burette to obtain the gas volume. The gas was analysed by GC. Additional CO2 was collected via burette by adding HCl aqueous solution to the reaction liquid phase.