Highly Efficient Additive-Free Dehydrogenation of Neat Formic Acid

Abstract

Formic acid (FA) is a promising hydrogen carrier which can play an instrumental role in the overall implementation of a hydrogen economy. In this regard, it is important to generate H₂ gas from neat FA without any solvent/additive, for which existing systems are scarce. Here we report the remarkable catalytic activity of a ruthenium 9H-acridine pincer complex for this process. The catalyst is unusually stable and robust in FA even at high temperatures and can catalyse neat FA dehydrogenation for over a month, with a total turnover number of 1,701,150, while also generating high H₂/CO₂ gas pressures (tested up to 100 bars). Mechanistic investigations and DFT studies are conducted to fully understand the molecular mechanism to the process. Overall, the high activity, stability, selectivity, simplicity and versatility of the system to generate a CO-free H₂/CO₂ gas stream and high pressure from neat FA makes it promising for large-scale implementation.

Introduction

Our excessive dependence on fossil fuels has resulted in the rapid increase of atmospheric CO₂,¹ contributing towards global warming, climate change, ocean acidification and other undesirable environmental effects.^2–3 In light of this, there has been increasing demand to transition from fossil fuels to renewable energy sources for a sustainable future.^4,5 Hydrogen, produced by electro-chemical water splitting, is considered a promising carrier of the intermittent renewable energy.^6,7 However, while having the highest gravimetric energy density, H₂ suffers from an inherently low volumetric energy density. Physical methods of hydrogen storage involve high costs, low capacity and safety risks. In this regard, hydrogen carriers are being intensively investigated.^8,9

Formic acid (FA) is considered as a promising hydrogen carrier because of its volumetric and gravimetric H₂ capacities, amounting to 53 g/L and 4.4 wt%, respectively, corresponding to a high energy density of 1.77 kWh/L.^9–12 FA can be synthesized by sustainable routes such as hydrogenation of CO₂,^13,14 or partial oxidation of wet biomass.^15,16 Dehydrogenation of FA to generate H₂ and CO₂ is thermodynamically favourable in the gas phase (ΔG° = -6.9 Kcal/mol). The ccientific literature on FA dehydrogenation catalysed by metal complexes is rich, with initial studies having been reported in 1960s,¹⁷ and since then many studies have been reported with transition metal catalysts, some exhibiting turnover numbers over a million.^18–22 Many of these studies employ a mixture of FA and base (such as triethylamine or alkali metal hydroxide)^23–31 for the dehydrogenation to avoid low pH of reaction solution which presumably could deactivate the catalyst. Many other studies employ catalytic amounts of additives as well as a solvent media (typically organic solvents such as THF, dioxane, or propylene carbonate).^29,31–34 Tetracoordinate iridium and ruthenium complexes have shown promising activities in catalysing FA dehydro-genation in aqueous solutions ^26,35–40 with the highest activity being observed for 1-8 M of FA. It is to be noted that many heterogeneous catalysts have also been reported for FA dehydrogenation and while they have the advantage of catalyst recyclability, they generally show lower efficiency and selectivity.^41,42

In transitioning towards a functional H₂ economy with FA as a potential H₂ carrier for storage and transport, it is desirable to avoid the use of solvents and other volatile additives whenever possible. The use of additives and solvents may compromise both the gravimetric and volumetric H₂ densities, as well as the energy density of the whole system. Furthermore, the presence of volatile additives (e.g. amine) and solvent vapours in the generated gas mixture could also damage the fuel cell; hence necessitating their removal via a hydrogen post-purification step which further increases the complexity and cost of the system. Compared to the plethora of FA dehydrogenation systems with additives and solvents, dehydrogenation of neat FA without any additive is extremely scarce.²¹ In 2016, Williams and co-workers reported iridium catalysed dehydrogenation of neat FA, using HCOONa (5 mol%, 8.6 wt%) as an additive.⁴³ Although a very high turnover number was obtained, no significant activity was observed without the basic additive. Furthermore, formation of CO in relatively high concentration (~500 ppm) was observed, whose abatement required the addition of water (10 wt%) or high HCOONa amount (up to 45 wt%), bringing down the overall H₂ storage potential. In 2019, Fischmeister and co-workers reported examples of neat FA dehydrogenation without any additive catalysed by an Ir complex.⁴⁴ However, the activity of the catalyst in neat FA was explored during only 10 min of operation time, which is far shorter than is required for practical implementation. Similar Ir complexes were also explored by Himeda and co-workers, and found to dehydrogenate neat FA (99%) (TOF< 800 h^-1) for unspecified times, although for efficient gas evolution it was necessary to use 4-8 M aqueous FA solutions.⁴⁵ Gelman and co-workers have reported a system for neat FA dehydro-genation without any additive, which again use an iridium catalyst.⁴⁶ Also, the CO contamination of the generated gas stream, whose avoidance is crucial for application in a fuel cell, was not investigated for neat FA. Moreover, dehydrogenation of neat FA in a closed system to generate high H₂/CO₂ pressure (>10 bar), which is of importance for hydrogen storage and delivery, was not reported in any of these systems.

There are several challenges in the dehydrogenation of neat FA not otherwise encountered in systems with solvent and additives. Firstly, the low solubility of many complexes in FA limits the number of possible catalyst candidates for neat FA dehydrogenation. Additionally, the high acidity of neat FA renders many catalysts ineffective in dehydrogenation through irreversible formation of unwanted catalytically inactive species. Besides, our calculations in this study indicate that dehydrogenation of neat FA is energetically endergonic by +2.9 kcal/mol as compared to dehydrogenation in non-polar organic solvents such as benzene (ΔG = -4.4 kcal/mol; vide infra DFT discussion), which might also contribute to the difficulty of dehydrogenating neat FA, especially at low temperatures.

Herein, we report the remarkable activity of a ruthenium 9H-acridine complex towards the additive-free dehydrogenation of neat FA. The catalyst is stable in FA for prolonged periods, is thermally robust, and can selectively catalyse the hydrogenation of neat FA with TON over 1.7 million and with TOF as high as 3067 h^-1. Furthermore, this catalytic system is active even in a closed system, generating high pressures of H₂/CO₂ from neat FA and can even catalyse the dehydrogenation of commercially available 85% FA, which is the lowest and most inexpensive grade of FA. Different aspects of this catalytic system were explored, keeping in my mind its potential use in a scale-up facility as described below.

Results

Catalytic activity of the complex

Dehydrogenation of FA begins as soon as complex 1^47,48 and FA are mixed, even at room temperature (Figure 1A). Complex 1 displays good solubility in FA, forming an orange coloured homogeneous solution. When 10 µmol (5.8 mg) of 1 was mixed with 0.5 mL (13.2 mmol) of FA in a J Young NMR tube, H₂ and CO₂ gases (in 1:1 ratio; no CO detected; CO detection limit: 15 ppm; supplementary note 4) were detected in the headspace after 5 mins by gas chromatography (GC). The dehydrogenation rate increases with increasing temperature (Figure 1C). At a catalyst loading of 10 µmol and at 95 °C, the dehydrogenation of 1 g (21.7 mmol) of neat FA was complete in 50 mins with a total evolution of 690 mL of H₂/CO₂ gas (~1:1), corresponding to 66% conversion of FA (Figure 1E). The CO level was below the detection limit of the GC (15 ppm) (Figure 1B). The remaining FA condensed at the headspace of the reaction vessel, out of reach of the catalyst. The reaction profile showed an initial 10 min period of relatively low catalytic activity, during which the system was likely heating up to the desired temperature (95 °C). Subsequently, between 15 and 45 min from the beginning of the reaction, the dehydrogenation rate increased, with the TOF reaching 2186 h^-1 (for TON and TOF calculation, see supplementary methods). During this period, an average steady gas flow of ~18 mL/min was achieved. In the final 5 min of the reaction, the TOF was limited by the availability of FA. Upon completion of the reaction, the orange solid catalyst was left behind in the reactor, and upon addition of neat FA the reaction immediately resumed. This process was repeated 15 times and at the end of the 15^th cycle, a total of 14.8 L of gas were collected, from the dehydro-genation of 12.3 mL FA with a TON of 30598 (overall FA conversion = 94%) (supplementary note 5, supplementary table 2). The CO content in the generated gas stream was found below 20 ppm in all the cycles. To check the catalytic longevity of the system, we carried out an extended experiment over two months at a lower catalyst loading of 0.20 µmol. At this catalyst loading, a slightly higher TOF was observed (3067 h^-1). Remarkably, the catalytic activity of the system was retained for more than 50 days of intermittent reaction time over a period of two months (supplementary figure 45). After this time, a turnover number (TON) of 1,701,150 was observed (Figure 1F; supplementary table 3). During this time, the catalytic activity was not significantly hampered by the presence of 1 in neat FA for a prolonged period of time at high temperatures, or by the cooling down the reaction solution followed by subsequent heating, or even by letting the reactor dry up for extended periods of time. Thus, the catalyst displays the robustness required for its potential application in a large scale on-demand H₂ production system. We also performed a scaled up long time reaction with 40 µmol (23 mg) of catalyst loading in 5 mL FA at 95 °C. Fresh FA was continually added via a syringe pump (2-5 mL/h, see methods) to replenish the consumed FA in the reactor. A steady gas flow rate of ~95 mL/min was observed for more than 70 h (Figure 1G). Afterwards, the rate decreased slowly. The catalytic activity was explored for 19 days of continuous operation. At the end of 19 days, a total of 1.2 L of FA was dehydrogenated (1.5 kg), generating >1560 L of H₂/CO₂ gas, with a catalytic turnover over 802,000. After this time, the catalyst was still active in the dehydrogenation and a gas generation rate of 35 mL/min was observed (37% of the initial catalytic activity). Notably, the CO level of the generated gas stream was below 20 ppm throughout the reaction.

Figure 1. Different aspects of the catalytic activity of 1 in the dehydrogenation of neat FA.

(A) FA decomposition cata-lysed by 1. (B) Gas chromatogram of the generated gas mixture during the dehydro-genation of neat FA at 95 °C; (C) Effect of temperature on the catalytic activity (from 65 to 95 °C; reaction conditions: FA (1g, 21.7 mmol), 1 (10 µmol), T (as specified). (D) Reaction setup. (E) Gas evolution with time at a reaction temperature of 95 °C (F) Changes in TON with reaction time (at a catalyst loading of 0.2 µmol) (G) Plot of flowrate vs time in a continuous scaled up experiment; reaction conditions: FA (5 mL; 0.13 mol), 1 (40 µmol), 95 °C, fresh FA injected at a rate of 2-5 mL/h via syringe pump to replenish the reacting FA, flowrate measurement error: ±5 mL/min. (H) Pressure generation with time during the dehydrogenation of neat FA in a closed 30 mL autoclave; reaction conditions: FA ( 5 mL, 0.13 mol), 1 (40 µmol, 23 mg), 103 °C. The reported gas evolution volumes (except for Figure 1G), TON and TOFs are averaged over two runs. Maximum error is ±10%.

To further understand the practicality of this system, the catalytic behaviour of complex 1 towards the dehydrogenation of FA was also explored in a closed system. Many FA dehydrogenation catalysts lose their catalytic activity in a closed system since the pressure built up due to H₂/CO₂ evolution diminishes the catalytic activity. This is an undesirable situation as high pressure H₂ generation is of much importance in the practical implementation of FA as a hydrogen carrier because it allows for H₂ storage in convenient times and places where immediate consumption is not necessary. Furthermore, it is advantageous to feed the input H₂ gas stream to a PEM fuel cell at a positive pressure (to increase the energy output), with the U.S. DOE recom-mended limit being 5-12 bars.⁴⁹ To our delight, complex 1 was able to cata-lyse neat FA decomposition in a closed system. When 4 mL of FA with 40 µmol of 1 was heated at a bath temperature of 103 °C inside a 30 mL autoclave, constant pressure generation inside the vessel was observed. The pressure reached 100 bars after 52 minutes, at which point the reaction was stopped (Figure 1H). During the reaction, pressure generation was almost linear with time, indicating that the catalytic activity of 1 does not decrease within this H₂/CO₂ pressure range. Also, the catalytic species present after the reaction were analysed by NMR spectroscopy and the presence of 1 as the major species (>90%, supplementary figure 38) was observed. Thus, it is demonstrated that complex 1 can generate high H₂/CO₂ pressure from neat FA. While our current laboratory facilities do not allow us to determine the maximum H₂/CO₂ pressure achievable with catalyst 1 using neat FA dehydrogenation in a close system, it can be surmised to be much higher than 100 bars, based on the rate of pressure increase in the 0-100 bar region.

Industrially, the majority of FA is produced via the hydrolysis of methyl formate (produced from CO and methanol), a process that requires large excess of water.⁵⁰ As a result, most of the commercially available FA contains various amounts of water, based on the purity level of the acid. In transitioning towards FA as an H₂ carrier, it is thus important to devise catalytic systems that can catalyse the dehydrogenation of neat FA, while withstanding some amount of water. To this extent, three different purity grades of commercially available FA, i.e. 98-100%, 97%, and 85%, were screened for dehydrogenation in our system. At 95 °C, similar dehydrogenation activity was observed for all three grades of FA (Figure 2). This displays the ability of complex 1 to withstand the presence of water and other possible contamination that might be present even in the lowest grades of commercially available FA. Notably, although in many systems using iridium com-plexes, the presence of water is known to facilitate the dehydrogenation rate of FA, no such ef-fect is observed in our system (supplementary note 1).

Figure 2. Dehydrogenation of different grades of FA.

Gas evolution vs time profile is shown during the de-hydrogenation of 98-100%, 97% and 85% formic acid. Reaction conditions: FA (1 g), [Ru] 0.01 mmol, 95 °C. Average of two / four (85%) runs. Maximum error in values ±10%.

Mechanistic investigation

The reactivity of complex 1 with HCOOH was explored further to understand the molecular mechanism of the catalytic process (Figure 3, Supplementary note 3). Upon addition of one equiv. of HCOOH to complex 1 in C₆D₆, immediate formation of two new Ru-hydride species were observed (Supplementary figure 28). Based on the ¹H, ³¹P{¹H} and ¹³C{¹H} NMR spectra, the predominant species was identified as the formic acid adduct of 1, intermediate 1a-mer (Figure 3a). The two protons of the FA ligand were observed at 8.78 and 10.86 ppm in the ¹H NMR spectrum, whereas the characteristic ³¹P{¹H} signal was observed at 78.5 ppm, 3 ppm downfield from the parent complex. The ³¹P{¹H} chemical shift of the minor Ru species appeared upfield at 62.7 ppm. A formate peak corresponding to this species was observed at 8.17 ppm in the ¹H NMR spectrum. Based on the NMR evidence, this minor species was identified as the formato complex (1b-mer), in which, a proton shifted from the FA ligand to the N atom of the ligand. Interestingly, these two new peaks vanished after 15 min at room temperature to give back the parent complex 1, accompanied by the generation of H₂ and CO₂. This is likely to proceed via the formate complex 1c, although it was not observed due to rapid decarboxylation.

Figure 3. Mechanistic studies.

a) reactivity of com-plex 1 with formic acid, CO₂ and H₂ at RT. b) reac-tivity of complex 1 with acetic acid at RT. c) Crystal structure of the acetate complex 1c’. Atoms are drawn with a probability level of 50%. Colour code: ruthenium (magenta), carbon (light grey), oxygen (red), phosphorus (yellow) and nitrogen (blue). Isopropyl groups are in wireframe and hydrogen atoms are not displayed for clarity. Selected bond lengths (Å) and angles (degree) Ru(1)-P(1) 2.2868(7), Ru(1)-P(2) 2.2918(6), Ru(1)-O(1) 2.1952(16), Ru(1)-O(2) 2.1863(15), Ru(1)-N(1) 2.1583(18), Ru(1)-C(30) 1.862(2), C(30)-Ru(1)-N(1) 179.71(8), C(30)-Ru(1)-O(2) 93.64(7), N(1)-Ru(1)-O(2) 86.64(6), C(30)-Ru(1)-O(1) 94.43(8), N(1)-Ru(1)-O(1) 85.78(6), O(2)-Ru(1)-O(1) 60.23(6), C(30)-Ru(1)-P(1) 92.18(7), N(1)-Ru(1)-P(1) 87.59(5), O(2)-Ru(1)-P(1) 157.56(4), O(1)-Ru(1)-P(1) 97.73(5), C(30)-Ru(1)-P(2) 92.54(7), N(1)-Ru(1)-P(2) 87.35(5), O(2)-Ru(1)-P(2) 95.64(5), O(1)-Ru(1)-P(2) 155.23(4), P(1)-Ru(1)-P(2) 105.73(3), O(2)-C(28)-O(1) 119.9(2).

The acetate analogue of complex 1c is kinetically stable and was synthesized by addition of acetic acid to complex 1 (Figure 3b, supplementary figures 18-21). The ¹H NMR spectrum of 1c’ did not show any characteristic hydride peaks as expected. Single crystals of 1c’ were grown and analysed by X-ray diffraction. Interestingly, the 9H-acridine ligand was found to coordinate the Ru centre in a facial manner, and the acetate in a bidentate fashion (Figure 3c). Though not isolable, the formate complex 1c can be successfully generated and characterized in situ by treating complex 1 with 4 bar CO₂ in a J Young NMR tube (supplementary figures 11-16). The formate proton of the formato ligand is observed at 8.07 ppm in the ¹H NMR spectrum as a triplet due to coupling with the ligand P atoms (J = 4.4 Hz). Similarly, the ¹³C{¹H} peak of the κ²-formato ligand is observed at 178.8 ppm. Interestingly, complex 1c, in the presence of unreacted 1, can form a proposed dimeric complex 1d with the formato ligand bridging the two Ru centres (based on NMR evidence). Upon removal of the CO₂ atmosphere, complex 1c slowly and spontaneously reverts back to complex 1 through decarboxylation, whereas in case of the dimeric complex 1d, a slightly elevated temperature of 60 °C was required for CO₂ loss and regeneration of 1. Notably, complexes [1a+1b] can also be accessed from 1 under a CO₂/H₂ atmosphere (supplementary figures 5-10).

Based on these observations, a catalytic cycle is proposed as depicted in Figure 4, in which catalysis proceeds via an isomer of 1 with fac-PNP coordination. In the first step, FA coordinates to the vacant site of complex 1 to form the formic acid adduct 1a-fac. Subsequently, H₂ liberation from complex 1a-fac generates complex 1c with a bidentate formate ligand. CO₂ elimination from 1c via β hydride elimination regenerates complex 1, thus completing the catalytic cycle.

Figure 4. Mechanistic cycle.

A Plausible scheme for the de-hydrogenation of neat formic acid.

Computational Studies

DFT studies were carried out to investigate the energy profile of such a FA dehydrogenation pathway and probe the high catalytic activity of 1 (Figure 5; Supplementary Discussion 2). To account for both the neat catalytic conditions and the mechanistic studies, computations were performed in the gas phase as well as with the SMD (solvent model based on density) variation of the IEFPCM (integral equation formalism polarizable continuum model) of Truhlar and co-workers in both FA and benzene.⁵¹ The fac isomer of 1 was found to be 7.2 kcal/mol higher in energy than the mer isomer in FA (10.9 kcal/mol in benzene). The energy difference suggests that 1-mer can access 1-fac at ambient temperature. FA coordination to 1-fac forms 1a-fac, which is energetically more stable than 1-fac by 4.7 kcal/mol (7.0 kcal/mol in benzene). The stability is provided by a dihydrogen bond interaction between the Ru-H^δ- and H^δ+-O of the coordinated formic acid ligand. A short distance of 1.449 Å was found between H^δ-… H^δ+ in the DFT optimized structure of 1a-fac, signifying the strong interaction (Figure 6).⁵² Subsequent H₂ elimination from this complex has low activation barrier (4.1 kcal/mol in both benzene and FA), which results in the formation of the κ²-formato complex 1c. Complex 1c is thermodynamically quite stable due to the chelating formate ligand. β-hydride elimination from 1c has an activation barrier of 18.6 kcal/mol in benzene and 23.4 kcal/mol in FA, both of which are accessible at room temperature. This also indicates that it is more difficult to dehydrogenate neat FA compared to a solution of FA in benzene. Besides, the dehydrogenation of neat FA was found to be endergonic by 2.9 kcal/mol (akin to aqueous FA dehydrogenation), compared to its exergonic nature in the gas phase (ΔG^Ꝋ= -6.9 kcal/mol) and benzene (ΔG^Ꝋ = -4.4 kcal/mol), likely due to the dissolution of the gases in neat FA. The calculated kinetic barriers of the H₂ elimination (4.1 kcal/mol in FA) and β-hydride elimination (23.4 kcal/mol in FA) steps suggest that the latter is the rate determining step in the reaction. Noteworthy, Eyring analysis of the rate of dehydrogenation of neat FA at different temperatures (Figure 1C) yielded an activation barrier of 23.1 kcal/mol (supplementary note 2), which is in close agreement with the calculated value.

Figure 5. DFT-calculated energy profile for FA dehydrogenation catalysed by 1.

a) elemental steps of the dehydrogenation b) energy profile as calculated by DFT studies. Energy values correspond to Gibbs free energies in kcal/mol with respect to 1-mer+FA at 298 K at atmospheric pressure.

Figure 6. DFT-optimized structure of 1a-fac.

The close H^δ-…H^δ+ bond dis-tance is shown. This interaction likely provides stability to the molecular structure.

Consistent with the computation, a large kinetic isotope effect (KIE) was observed for the dehydrogenation of DCOOH [k_H/k_D = 2.12(3); Table 1]. On the other hand, the dehydrogenation rate of HCOOD was similar to that of HCOOH under similar reaction conditions (k_H/k_D = 1.10(6)), in line with the low activation energy for H₂ elimination, as compared to β-hydride elimination. The small KIE observed for HCOOD also verifies that during the catalytic cycle, H₂ elimination precedes CO₂ evolution.

Table 1. Kinetic isotope effect data.

Reactant	KIE
HCOOH	1
HCOOD	1.10(6)
DCOOH	2.12(3)
DCOOD	2.31 (3)

Reaction conditions: Reactant (13 mmol), 1 (5 µmol), 95 °C

Catalyst Evolution

To identify any potential deactivation pathways during the dehydrogenation, the orange solid remaining in the reactor after complete FA dehydrogenation by complex 1 (after TON 7260) was dissolved in C₆D₆ and analysed by NMR spectroscopy. The ³¹P{¹H} and ¹H NMR spectra showed the presence of complex 1 as the major species (80%). Interestingly, a different species, not observed in the mechanistic studies, was also present in the solution with a signature ³¹P{¹H} NMR peak at 89.7 ppm and a hydride peak at -6.65 ppm (triplet, J_PH = 23.5 Hz) in the ¹H NMR spectrum (supplementary figure 36). The downfield shift of the hydride peak indicates the presence of an opposite ligand with strong trans effect. When one equiv of CO was introduced to this NMR tube, the existing complex 1 converted to this species, indicating the structure of biscarbonyl 9H-acridine complex for 1e (Figure 7). The ¹H and ³¹P{¹H} chemical shifts are in accordance with that of previously reported in the literature.⁵³ Complex 1e likely forms via coordination of the trace amount of CO generated (below the detection limit of GC) under the reaction condition. Interestingly, 1e was also found to be active in neat FA dehydrogenation at 95 °C, although with reduced activity (Figure 7). Upon completion of the reaction, 1e was found to be intact, with no observable CO ligand dissociation. Thus, an alternate mechanistic pathway via a single available coordination site is also plausible (supplementary figure 49), similar to mechanisms postulated in previous reports.^31,33,54 Our DFT calculations suggest an overall activation barrier of 32.9 kcal/mol for intramolecular single site decarboxylation via β-hydride elimination for the 1e system (in FA). The higher activation barrier, when compared to the bis-coordination mechanism of 1, would explain the slower catalytic activity of 1e as compared to 1. Notably, the aromatic acridine ligand based hydrido-chloride complex 1-Cl was also active towards neat FA dehydrogenation, presumably via a similar mechanism to that of 1e, but with lower TOF (Figure 7).

Figure 7. Comparison of the catalytic activity of 1, 1e and 1-Cl towards neat FA dehydrogenation.

Reaction conditions: FA (1 g, 21.7 mmol), catalyst (10 µmol), 95 °C.

Conclusions

In conclusion, a highly efficient and robust neat FA dehydrogenation system catalysed by the Ru-9H-acridine complex (1) is described. As opposed to most FA dehydrogenation systems, this system catalyses the dehydrogenation of neat FA without any additives/solvent, allowing the full H₂ storage capacity of FA to be realized. Complex 1 catalyses the dehydrogenation of FA even at ambient temperature, with the catalytic rate significantly increasing with temperature (tested up to 95 °C). The system was catalytically active for long times and a TON as high as 1,701,150 was obtained. In a large-scale experiment, a total of 1.5 kg of neat FA was dehydrogenated after 19 days of continuous reaction, generating >1560 L of H₂/CO₂ gas ((TON>802,000). CO content in the generated H₂/CO₂ stream was found below 20 ppm throughout the reaction. Furthermore, the catalyst can be used even in a closed system to produce high pressure (up to 100 bars in this study) of H₂/CO₂ gas storage from liquid FA and is also active with commercially available low grade and inexpensive 85% FA. Mechanistic investigation and computational studies suggest a catalytic pathway via the cis available coordination site to be the major pathway in the dehydro-genation. A catalyst evolution pathway to a biscarbonyl 9H-acridine complex (1e) is identified which also catalyses neat FA dehydrogenation. Overall, the system described here fulfils, or has the potential to fulfil, with further optimizations, many of the U.S. Department of Energy technical performance targets including usable specific-energy from H₂, usable energy density from H₂, storage system cost, min. delivery pressure from storage system and others (supplementary discussion 1). Given the high tunability of the acridine-based ligand, we believe that further research might lead to even more active acridine-based complexes for neat FA dehydrogenation. Essentially, this study lays the foundation for a practical additive-free neat FA dehydrogenation system, with the goal of its final use on an industrial level upon further improvement. At the same time, we are also exploring the possibility of similar systems based on first-row transition metal catalysts.

Methods

Standard procedure for neat formic acid dehydrogenation

Inside a N₂ filled glove box, 10 µmol of complex 1 was dissolved in 0.6 g of FA (13 mmol) in a 5 mL glass vial. The resulting orange colour solution was transferred to a Schlenk flask equipped with a side arm. The vial was further washed with 0.4 g of FA (8.7 mmol) and the solution was added to the flask. The Schlenk flask was sealed, taken out of the N₂ glove box, and connected to the condenser attached to an inverted graduated cylinder filled with silicone oil on the other side (for the quantification of evolving gas) via a triple valve (drawing in Figure 1D), while keeping the solution under N₂ atmosphere all the time. The flask was further dipped into a preheated oil bath (65 °C to 95 °C). Immediate gas evolution was observed which continued over time and the amount of gas evolved over time was manually recorded. Samples of gas from the generated gas stream were collected via syringe for GC analysis using a triple valve. Turnover numbers (TON) and turnover frequencies (TOF) were calculated as described in the supplementary methods (supplementary equation 1-3).

For the long-time reaction with 0.2 µmol catalyst loading

For the low catalyst loading experiment, 0.2 µmol catalyst was used along with 1 g (21.7 mmol) of FA initially. A solution of 0.2 µmol of 1 in 1g of FA was prepared by diluting a stock solution of 1 (20 µmol) in 1 g FA twice. A slightly modified system without a condenser was used. The gas was collected via the side arm, and the top of the flask was sealed with a rubber septum. Upon completion of the dehydrogenation of the available FA, fresh FA was added via syringe though the rubber septum. When not in active catalysis, the Schlenk flask was kept at RT inside a nitrogen glove box. Further observations regarding this experiment is in Supplementary Note 6.

For the scaled up continuous FA dehydrogenation experiment

In a two-neck round-bottom flask, 23 mg of catalyst 1 was dissolved in 5 mL (0.13 mol) FA under N₂ atmosphere. The two-neck round bottom flask was then connected to a condenser under nitrogen atmosphere. The condenser was attached to a three-way valve which in turn was connected to a flowmeter and an inverted graduated cylinder. The flask was then dipped into a preheated oil bath (95 °C). Immediate gas evolution was observed. The generated gas flow over time was monitored by passing the gas through flowmeter and recorded manually. Fresh FA was pumped in though the second neck of the round bottom flask (sealed by rubber septum) at a rate between 2-5 mL/h by a syringe pump. GC samples of the generated gas mixture were collected via syringe through the rubber septum. Total amounts of gas generated were calculated by integrating gas flowrate over time. TON was calculated from the total evolved gas amount as per the calculations previously stated. The exact addition rate of FA was determined based on the rate of its consumption which was calculated from the flow rate of the generated gas stream.

For the reaction in a close system for generating H₂/CO₂ pressure

Inside a N₂ glove box, 23 mg of catalyst 1 was dissolved in 5 g (0.108 mol) of neat FA in a 20 mL vial. The resulting orange solution was transferred to a 30 mL autoclave with Teflon jacket. The autoclave was sealed, taken out of the box and then placed in a preheated oil bath (bath temp: 103 °C). Generation of pressure inside the reactor was observed with time which was recorded every five minutes. The pressure reached 100 bars at 52 minutes of reaction time at which point the reactor was taken out of oil bath and the gases were released. The autoclave was further cooled down to room temperature, taken inside a nitrogen glove box, opened and the catalyst in the remaining solution was recovered by removing the FA in vacuo. The obtained solid was analysed by ¹H and ³¹P{¹H} NMR with C₆D₆ as the deuterated solvent.

Computational studies

All geometries were optimized using Truhlar’s M06-L functional,⁵⁵ the triple-ξ def2-TZVP basis set⁵⁶ and W06 density fitting to increase computational efficiency⁵⁷ as well as Grimme’s D3(0) empirical dispersion correction.⁵⁸ Frequency calculations at this level of theory were run in order to confirm stationary points and transition states, as well as to compute thermodynamic properties. Single point energies of the optimized structures were computed using the range-separated meta-GGA hybrid functional ωB97M-V of the Head-Gordon group⁵⁹ including dispersion correction,^60,61 together with the triple-ξ def2-TZVPP basis set⁵⁶ and the corresponding auxiliary basis sets, def2/J⁵⁷ and def2-TZVPP/C, ⁶² for RIJCOSX density fitting. Gibbs free energies were computed by adding the free energy correction terms from the frequency calculations to the single point energies according to

$${\text{G}}^{\text{ωB}97\text{M}-{\text{V}}_{\text{gas}}}={\text{E}}^{\text{ωB}97\text{M}-{\text{V}}_{el\text{/}gas}}+{\text{corr}}^{M06-L_{freq/gas}}$$ (1)

or

$${\text{G}}^{\text{ωB}97\text{M}-{\text{V}}_{\text{SMD}}}={\text{E}}^{\text{ωB}97\text{M}-{\text{V}}_{el/gas}}+{\text{corr}}^{M06-L_{freq/gas}}+{\text{corr}}^{M06-L_{SMD}}$$ (2)

where corr^M06-L_freq/gas is a thermal correction term to the Gibbs free energy from the frequency calculation, and where corr^M06-L_SMD is a solvation (formic acid or benzene) correction term obtained from single point energy calculations using the integral equation formalism variant (IEFPCM) of the PCM model in the SMD variation of Truhlar and co-workers,^51,63 according to

$${\text{corr}}^{M06-L_{SMD}}={\text{E}}^{M06-L_{SMD}}{-\text{E}}^{M06-L_{gas}}$$ (3)

Optimizations and frequency calculations were done using the Gaussian 16 software suite in the C.01 revision.⁶⁴ Single point calculations were performed using Gaussian 16⁶⁴ or the ORCA Software in the 4.2.1 release.⁶⁵

Figure 8.

Data availability

Synthetic procedures, NMR spectra and characterization data for all the new compounds are available within this article and its Supporting Information. The X-ray crystallographic coordinates for the structure reported in this article have been deposited at the Cambridge Crystallographic Data Centre (CCDC), under deposition number CCDC 2011708. DFT-optimized geometries of atomistic models are provided as part of the Supplementary Data. The data can be obtained free of charge from The Cambridge Crystallographic Data Centre via http://www.ccdc.cam.ac.uk/data_request/ cif. Any further relevant data are available from the authors upon reasonable request.