Evaluation of hydrogen generation step: dehydrogenation of alkali formates (FD)In an ideal hydrogen storage system, the same catalyst and solvent system should be applied for both hydrogenation and dehydrogenation steps. Nevertheless, to identify most suitable conditions for a combined process, initially the hydrogenation of bicarbonate (BH) and formate dehydrogenation (FD) were investigated separately. In general, the nature of the formate/bicarbonate cation has a decisive influence on the volumetric and gravimetric energy content of the reaction system. Consequently, FD was studied utilizing Li, Na, K, Rb, and Cs formates under identical conditions (Table 1 and Supplementary Table 4). In previous works, it has been shown that the latter reaction proceeds well in aqueous DMF using [RuCl2(benzene)]2 and biphosphine ligands as in situ catalysts28.Table 1 Comparison of the dehydrogenation of alkali formates (FD) in the presence of Ru-1Following this protocol, HCOOLi provided H2 in 96% yield with a TON of 3820 at 60 °C after 6 h (Table 1, entry 1). Notably, the obtained gas contained a mixture of H2/CO2 in a ratio of 65/35, demonstrating the loss of a significant amount of carbon dioxide in this system. In contrast, simple HCOONa showed high H2 productivity (98% yield; TON = 3920) and improved gas purity (H2/CO2 = 92/8) (Table 1, entry 2). Appling HCOOK as the starting material led to decreased dehydrogenation activity (45% yield and TON = 1780), but at the same time almost all carbon dioxide remained in the reaction solution (H2/CO2 = 99.5/0.5; Table 1, entry 3). HCOORb (58% yield and TON = 2320) and HCOOCs (53% yield and TON = 2120) showed comparable reactivity with a gas proportion of H2/CO2 = 90/10 and H2/CO2 = 97/3, respectively. Based on these results and considering substrate costs, molar solubility, and the composition of produced gas, HCOOK was chosen for further investigations.Next, the impacts of different catalysts, co-solvent, and the amount of water as well as reaction temperature were studied at increased substrate concentration in more detail (Table 2, and Supplementary Tables 5–8). First, different catalysts were tested. The combinations of [RuCl2(benzene)2]/dppe (Ru-1), [RuCl2(benzene)2]/dppm (Ru-2) and RuCl3/TPPTS (Ru-3) showed low productivity with H2 yields <15% and TONs <1600. In all cases, the ratio of H2/CO2 was around 99/1 (Table 2, entries 1–3). The activity of the defined complex Ru-4 was slightly improved compared to the in situ generated system (Table 2, entry 4). Higher productivity was achieved with ruthenium pincer complexes. Using Ru-PiPrN(H)PiPr (Ru-5), Ru-MACHO (Ru-6), Ru-MACHO-BH (Ru-7) and Milstein´s Ru-catalyst (Ru-8) H2 yields up to 63%, and TONs up to 7110 were obtained (Table 2, entries 5-8). Again, the CO2 release was negligible (≤0.5%). Besides ruthenium, different Ir-pincer catalysts were tested. Ir-1 showed good results (69% H2 yield, TON 7790), but the reactivity decreased with the reaction time (Table 2, entry 9). Notably, the water-soluble Ir-Cp* complexes Ir-2 and Ir-3, which work well in formic acid dehydrogenation, showed no activity under these basic conditions (Table 2, entries 10–11). Overall, Ru-5 was selected as the catalyst for further investigations. Almost full conversion of HCOOK was obtained by increasing the reaction time from 6 h to 13 h (TON 11,260, Table 2, entry 13). Further increasing the substrate amount to 275 mmol, a TON of 17,390 and 62% H2 yield were observed (Table 2, entry 14). Additionally, increasing the reaction temperature and/or decreasing the catalyst amount (3.6 ppm) gave excellent H2 yields up to 99% and TONs up 263,200 (Table 2, entries 15–17). Noteworthy, utilizing Ru-5 for HCOONa dehydrogenation instead of HCOOK gave significantly lower H2 yield and purity (Table 2, entry 12). Among the different co-solvents tested, sulfolane, t-amylol, and triglyme allowed for significant hydrogen production rate; however, DMF performed best in FD (Supplementary Table 8). Then, the ratio of DMF and water was varied. Optimal results with respect to catalyst activity and CO2 release were obtained in DMF/water (15/10) (Table 2, entry 18 and Supplementary Table 8, entries 15–19). With regards to the activity of Ru-5 in the BH (see next chapter, Table 4), different ratios of triglyme were investigated for FD, too (Table 2, entries 18–20 and Supplementary Table 8). Using the solvent ratio of 15 mL triglyme and 10 mL water, the highest gas purity within the triglyme experiments was reached. Under the standard conditions 32% H2 yield were observed with a gas ratio of 94/6 H2/CO2 (Table 2, entry 19; for 275 mmol HCOOK: 30% H2 yield, 92/8 H2/CO2; Table 2, entry 20). Elongation of the reaction time from 6 h to 46 h increased the yield to 91% (Table 2, entry 21). Thus, FD productivity in triglyme/H2O is comparable to DMF/water (although activity is lower), and full conversion of formate in triglyme/H2O is possible.Table 2 Dehydrogenation of HCOOK: Variation of catalysts and solventTo showcase the catalytic efficiency, the dehydrogenation of HCOOK was performed on >250 g-scale (3 mol) with only 3.3 ppm catalyst Ru-5, using DMF/H2O (150/100). Full conversion was observed with a remarkable gas release of >72 L in 25 h. This corresponds to a practically relevant catalyst turnover number (TON = 290,000), frequency (TOFmax = 38,400 h−1), maximum space-time-yield (0.76 mol h−1 L−1), and rate of hydrogen releasemax (9.3 L h−1) (Table 3 and Supplementary Table 10; for reaction in triglyme/H2O see Supplementary Table 11). The purity of the produced hydrogen was 97% and no CO was detected (detection limit <10 ppm) under these reaction conditions.Table 3 Scaled up dehydrogenation of HCOOKEvaluation of the hydrogen storage step: hydrogenation of potassium bicarbonate (BH)Similar to the dehydrogenation experiments, different catalysts, solvent mixtures, and temperatures were examined for the hydrogenation step (Table 4 and Supplementary Table 12). In an ideal hydrogen storage device both H2 loading and release should be performed in the presence of the same catalyst system (catalyst, solvent). Consequently, selected well-performing catalysts in FD were investigated for their capability of BH in DMF/H2O (20/5 mL) at 90 °C (Table 4, entries 1–7). Since Ru-1, Ru-5 and Ru-7 exhibited similar activities (≤7% yield) in BH, Ru-5 was selected for further studies due to its superior performance in the dehydrogenation. Unfortunately, DMF was partially decomposed under hydrogen pressure due to catalytic reductive cleavage of the amide29. To avoid this issue, further catalytic experiments were conducted using triglyme as co-solvent. Triglyme was chosen for its anticipated stability, thermal resistance (boiling point 216 °C) and previous performance in the formate dehydrogenation vide supra. Noteworthy, applying triglyme water mixture as solvent in FD, hydrogen yields >91% were achieved (Table 2, entries 18–20; Supplementary Table 8).Table 4 Catalytic hydrogenation of KHCO3 to HCOOK in aqueous DMF and triglymeInterestingly, the concentration of this co-solvent has a significant effect on the hydrogenation reaction. While a low yield (4%; Table 4, entry 8) was observed at a triglyme/H2O ratio of 20/5 mL, increasing the water concentration led to a significant increase in the formate yield up to 74% (Table 4, entries 9–12). As expected, decreasing the reaction temperature from 90 °C to 50 °C resulted in reduced formate yields (Table 4, entries 13–14 and Fig. 2).Combination of hydrogenation and dehydrogenation steps: demonstration of H2 storage and release cyclesTo demonstrate the stability of the catalysts system, we investigated the cyclability of the (de)hydrogenation over a long period with an appropriate number of cycles. Principally, the storage capacity of a hydrogen battery relies on both, FD and BH performence. Thus, a compromise of productivities achieved in FD and BH has to be made. Taken all results from the (de)hydrogenation into account, triglyme/H2O (15/10 mL) was chosen as the appropriate solvent due to its high stability and activity in BH (up to 74%) and almost full conversion for FD (up to 91%). The individual formate dehydrogenation and bicarbonate hydrogenation steps were combined using a 100 mL autoclave in the presence of KHCO3 (135 mmol), Ru-5 (10 µmol, 74 ppm). As shown in Fig. 3, 40 consecutive cycles of hydrogen storage and release were performed over a period of 6 months. While in the first two experiments, a slight decrease of the hydrogen yield was observed, from the third run onwards, highly stable activity and hydrogen yields were obtained (3rd cycle: 1195 mL; 37th cycle: 1116 mL). In 40 hydrogen storage and release cycles, nearly 50 L of hydrogen with an average purity of 99.5% were generated utilizing ppm amounts of Ru-5. In addition, the produced gas contained <0.5% carbon dioxide and less than 10 ppm carbon monoxide. As shown above, the hydrogen yield can even be increased by the elongating the reaction time. By shortening the FD reaction time, we were able to show that the optimal molecularly defined Ru complex allowes for a stable operation of the hydrogen generation for 6 months, demonstrating its potential for practical applications.Fig. 3: H2 storage and release cycles of the KHCO3/HCOOK system using Ru-5.H2 storage and release cycles were started with BH: 135 mmol KHCO3 (13.5 g), 10 µmol Ru-5 (4.7 mg), triglyme/water (15 mL/10 mL), 60 bar H2, 90 °C, 18 h. After pressure release, FD was started: 90 °C, 12 h. The generated gas was measured with manual burettes. The gas was analyzed by GC. CO is undetectable in all cases (CO quantitation limit of the applied GC is <10 ppm).
