The production of hydrogen gas on demand is an enabling technology for the widespread deployment of hydrogen fuel cell vehicles. One approach to providing H2 on demand is to release it catalytically from a liquid organic hydrogen carrier (LOHC), provided that the economics of such a system can overcome the costs of pressurizing and delivering the gas. Gas compression contributes 49% to 83% of the total refueling cost for light-duty and heavy-duty vehicles, respectively, in US retail cases. 1 Thus, the ability to produce pressurized H2 on demand reduces the cost of H2 in vehicle refueling. Yet, to our view, most catalyst development work on LOHC dehydrogenation has been under ambient pressure conditions.
Formic acid (FA, available from biomass fermentation or CO2 electrolysis), is a low cost, sustainable hydrogen carrier with desirable volumetric density (1.22 g/mL) and H2 content (4.4 wt %). 2,3 Its dehydrogenation is significantly entropically driven, with rH o = +7.4 kcal/mol and rS o = +51 cal/mol•K, so entropic energy released upon dehydrogenation serves as a type of spring, capable of delivering compressed hydrogen without the cost of compression. Self-pressurization of FA or alcohol dehydrogenation creates a unique environment for catalysis, where H2, CO2, and/or CO can govern catalyst initiation (e.g. insitu catalyst synthesis), speciation, and decomposition. While carbonylation is a known poisoning pathway in many cases, 4,5 we find that it can be essential to catalyst activation in others: for example, CO can play a healing role, extending the life of systems that would be deactivated without it. Despite these key advantages, we know of no broad studies of how closed-reactor conditions impact dehydrogenation catalysis; 5 the healing role of CO has been missed; and there are not generalizations for when this behavior might be expected or what the role of pressurization might have in directing it.
Hydrogen release from FA has been studied extensively in homogeneous and heterogeneous systems based on precious (Ir, 6,7,8 Ru, 8,9,10 Pd, 11,12 and Au 13,14 ) and non-precious (Fe 15,16 and Mn 17,18 ) metal catalysts, but we see only a few systems that are known to produce pressurized products while maintaining catalytic reactivity and selectivity. [19][20][21][22][23][24] Pioneering work by Fellay et al. described one of the first examples of high-pressure dehydrogenation of aqueous formic acid using a ruthenium catalyst. 23,24 Since then, several groups have reported similar findings, [19][20][21][22] however, industrially relevant turnover frequencies (TOF) and turnover numbers (TON) have not been achieved using neat formic acid. Most recently, Milstein recently reported a ruthenium PNP pincer catalyst for the 1. Ambient pressure conditions: All catalysts in this study were stored in the glovebox under nitrogen, and glassware used (round bottom flasks, stir bars, water condensers, etc.) was oven dried prior to use. In the glovebox, catalyst (7.95 mmol) and sodium formate (1.20 g) were weighed out and added to a round bottom flask equipped with a magnetic stir bar. Formic acid (3.00 mL, 79.5 mmol) was measured out and added to the same flask via a syringe. The flask was connected to a water condenser with a Tygon tubing. The other end of the tubing was submerged in an inverted graduated cylinder (eudiometer). Oil bath temperature was set to 110 °C. Evolved gas volume was recorded by water eudiometry.
In the glovebox under nitrogen, catalyst (7.95 mmol) and sodium formate (1.20 g) were weighted out and added to an 8-dram vial equipped with a magnetic stir bar. Formic acid (3.00 mL, 79.5 mmol) was measured out and added to the same vial via a syringe. The solution was transferred and sealed in a 125 mL Parr apparatus. The internal temperature of the reactor was set at 110 °C (± 5 °C) and monitored closely via a thermocouple to minimize temperature difference between the oil bath and the reaction temperature. Evolved gas pressure was monitored via the reactor's pressure gauge. 3. H2/CO Gas Pre-Treatment: Similar procedure to 3.2.2., except the reaction was charged with either H2 or CO to the desired pressure at room temperature and let stir at 110 °C. Evolved gas pressure was monitored via the reactor's pressure gauge.
Catalysts for FA Dehydrogenation at Ambient and Self-pressurizing Conditions. We surveyed a wide range of complexes that generally fit into four classes: (1) Noyori-type tridentate complexes 1-8, (2) bidentate chelates complexes 9-12 and their analog 13, (3) cyclopentadienyl piano stools complexes 14-17, and (4) metal precursors for the ligated complexes 18-21 (Figure 1). Each was examined in FA dehydrogenation both under ambient pressure and self-pressurizing conditions to determine catalyst activity and efficiency (Table 1). While every complex is different under these conditions, some generalizations of each class can be identified.
Table 1 shows the results of FA dehydrogenation conducted under both ambient and self-pressurizing conditions. All twenty-one complexes react with FA at a higher rate when pressurized than they do at ambient pressure, each without detectable reversibility (see Figure S4-S30 for time course data). The overall improvement in conversion efficiency varied between a minimum of +13% (entries 15 and 16) and a maximum of +72% (entry 14) as conditions changed from open to closed vessels. For example, mildly active complex 20 at ambient pressure promoted complete conversion when in a closed system (entry 20). Perhaps most startling, complexes 1-6, 8, 10, 13, and 21 exhibit little reactivity at ambient pressure but are dramatically more reactive under self-pressurizing conditions. An exception was observed in complex 7 26,27,28 which has competitive reaction conversion at both ambient and pressurized conditions (Figure S15).
Complexes 1-8 in the well-studied Noyori-type tridentate family, generally featuring M(PNL) (L = PPh2, P( i Pr)2, S(CH3)2) structures, tend to have lower reactivity than other catalysts at ambient pressure, giving conversions between 2% and 10%; but they are the highly impacted by pressurization relative to the other classes, reaching conversions from 58% to 84% under selfpressurizing conditions. Complex 7 is a notable exception to both of these generalizations, possibly owing to the semi-lability of its phosphine oxide and the lower hydricity of its active form; whereas in an ester hydrogenation reaction, complex 7 is one order of magnitude less efficient than complex 4. 26 Often, complexes in this class require pre-activation via hydrodechlorination with KOH or KO t Bu to generate their active hydride forms. 9, 26,31,32,33 Nevertheless, under self-pressurizing conditions, there was an increase in conversion from 21% (entry 2a) to 75% (entry 6a) without such pre-activation. For example, self-pressurization enables complex 4 (entry 4b) to achieve 79% conversion, comparable to its activated dihydride derivative (entry 5, 79%). Complex 6 can be initiated under pressurizing conditions without any base to convert 84% FA, while at ambient pressure only 3.4% FA is converted, even if the catalyst is activated with KO t Bu. This dramatic enhancement of reactivity upon pressurization suggests that one of the reaction products, like H2 or CO, is necessary to enable or maintain catalytic activity. Hydrogenation is known to activate aminecontaining Noyori-type complexes such as 1-6 in the presence of base, 9,31,32,33 which is a possible explanation. Further, we observe that thermal decarbonylation of FA is possible at our operating temperature (vide infra). We expect that the trace CO generated through this pathway is oxidized rapidly by the catalyst, but that its continued supply installs or maintains a CO ligand on the catalyst.
Bidentate chelate complexes 9-CO, 11, 6,29,34,35,36 and 12 are the most reactive precursors that we encountered at ambient pressure. 37 Complex 11 exhibits the highest rate of the entire library, where in 79.5 mmol of FA were fully converted within 1.2 hours. We found these complexes to be pressure-tolerant, but with little enhancement in reactivity because of their high baseline efficiency at ambient pressure. We believe the unique reactivity of these complexes to be a function of a novel selfassembly pathway: these convert to two-metal pseudo-pincer structures in the presence of a CO (isolated and characterized from reaction mixtures), exemplified by cases of our (pyridyl)phosphine ligand bound to ruthenium and iridium (Figure 2). 6,37 These pathways are available at ambient pressure either from FA or an alcohol. 6,29,34,35,36 The resulting bimetallic complexes have high activity and stability at ambient or elevated pressure. The active complexes have structural homology with some prolific Noyori-type and Milstein-type pincer complexes, where one arm of the tridentate ligand is replaced by the second metal. 38,39 Carbene-ligated compound 10 in this class lacks the reactivity of 9-CO, 11, or 12. While the reactivity of carbeneligated systems 9-CO and 10 should be different than their phosphine-ligated congeners 11 and 12, it is surprising that 10 does not react analogously to 9-CO under pressurized conditions, especially whereas 9-CO is prepared from its cyclooctadiene-ligated precursor 9 at ambient pressure (vide infra). Crabtree's catalyst 13 also exhibits low reactivity compared to its bidentate analog 11. We infer that tethering the pyridine and phosphine groups is important for proper catalyst self-assembly.
Piano stool Cp*Ir complexes 15-17 are not very efficient in this study, although they are moderately aided by pressure. Complexes 16 and 17 have been known to have excellent reactivity in alcohol dehydrogenation, 30,39 but their activity towards FA is moderate, respectively 16% and 42% conversion under pressurizing conditions. Notably, Shvo's cyclopentadienone-ligated catalyst 14 is much more reactive than Cp*Ir systems under pressurizing condition. The Shvo system is known to rest in its dimeric form 14 in the presence of H2, 40,41,42 so we reason that the availability of CO to trap the system's oxidized monomer and prevent formation of 14 could account for its rate advantage upon pressurization, because it is known that H2 pressure will drive the system back to dimer 14. 41,42,43 While several of the ligated species in Table 1 are efficient catalysts-they were designed as such-we were surprised to find that their synthetic precursors 18-21 6,36,37 have reactivity that rivals their ligated congeners. We find, however, that unlike the ligated congeners, the unligated precursors seem to deactivate easily. Overall, one piece of traditional wisdom that seems to be preserved is that ligated complexes tend to have good stability, sometimes at the cost of reaction rate. For example, we had difficulty replicating entries 18-21 whilst other entries were very reliable. Apparently, these more naked species tend to react quickly, yet the reactivity is short-lived and difficult to replicate.
Whereas many of the complexes we screened are more productive under self-pressurizing conditions, we conclude that initially formed products, probably CO and H2, are involved in activating the precatalysts 5,37 and healing the active catalyst by preempting deactivation processes. We propose that these processes could be emulated by adding CO or H2 at the outset of the reaction. To test this, reactions involving four catalyst precursors, 5, 10, 14, and 15, were examined representing the 3 respective classes of ligated precatalysts that benefitted significantly from self-pressurizing conditions. These were alternatively pretreated with H2 or CO in FA and their catalytic activity was evaluated (Figure 3).
Neither H2 (green triangles) nor CO (orange circles) uniformly improved catalytic activity over baseline (black squares) of every catalyst tested. While complexes 15 and 10 benefit respectively from H2 and CO pretreatment, other combinations of catalyst and treatment did not significantly improve reactivity: there is not a generalization that explains why these four complexes are accelerated by pressure. By contrast, both 10 and 5 are deactivated by H2 pretreatment. In the case of the Shvo system 14, H2 pressure slows the reaction but did not affect maximum total pressure (27-28 bar), consistent with the above proposal of dimer formation.
Notably, after 10 was pretreated with CO, the activity was significantly improved, reaching 89% conversion in five hours, surpassing its carbonylated homolog 9-CO. This is an interesting contrast to the relatively low reactivity of 9 under selfpressurizing conditions (vide supra): apparently insufficient CO Please do not adjust margins Please do not adjust margins is generated by formic acid dehydrogenation to realize the full benefit of carbonylation. Also very interesting about this catalyst system is that at ambient pressure, reactivity slows after about 3 minutes, which is not observed under pressurizing conditions. We view this as evidence that CO heals the catalyst and maintains fast kinetics when it does not have the opportunity to escape the reactor. Despite numerous cases of catalyst poisoning by metal carbonylation, 4, 5 complexes 9-CO and 10 exhibit the opposite effect: in the absence of CO, 10 has low activity for FA decomposition, but when CO is introduced, either by self-generation or pretreatment, complex 10 performed ca. three times (added CO, Figure 3) to four times (self-generated CO, Table 1, entry 9b) better.
Whereas CO is essential to the activation of these catalyst systems, it must be available in the reactor, although it is not detected in the product stream of FA dehydrogenation as reported in many studies from our lab and others. 5 Although Scheme 1. Synthesis and molecular structures of 9-CO and 11-CO. a a CCDC 2142637 contains supplementary crystallographic data for 11-CO.
there has not been a full explanation of this, we believe that formation of CO occurs thermally, 44,45 possibly catalyzed by traces of metals in the reactor vessel, and that CO is oxidized rapidly by the catalyst in our conditions. We tested these ideas with two experiments: (1) when the reaction was run with precursor 11 under pressurizing conditions and utilizing rapid heating, the reactor reaching over 129 o C at times, we detected CO concentration up to 0.63%, concurrent with fast H2 generation (107 L/h). No CO (< 10 ppm) is observed under analogous conditions below 100 o C. This suggests that thermal decarbonylation of FA can produce significant CO concentration if not controlled; 44,45 (2) In an aqueous methanol photodehydrogenation experiment (Table S9), 6% of CO was generated in the absence of catalyst 10, whereas none can be detected when 10 is present. Complex 10 was chosen for this experiment for its relatively slow reactivity in FA dehydrogenation, thus to allow longer life and easier observation of C1 intermediates. We infer from this observation that, when CO is produced by a non-catalytic reaction, the presence of an appropriate metal complex will reform the CO efficiently: CO is available, but not detectable. We suspect that this is a general feature of homogeneous catalysts for formic acid dehydrogenation that has not previously been described.
Whereas CO is vital to the initiation and speciation of some catalysts, we attempted to prepare species by independent synthesis that could be responsible for the observations. Upon treating 10 with 1 atm CO, we found that 10-CO was not stable to isolation. Treatment of 11 with CO results in a broad diversity of structures, which we have previously reported. 29 While these systems failed, the clean carbonylated species 9 readily yielded 9-CO upon carbonylation. 29 We measured the kinetics of dehydrogenation with 9 and 9-CO at ambient pressure to test the hypothesis that CO plays a role in precatalyst activation. At ambient pressure, complex 9-CO dehydrogenates FA faster than precursor 9 (Figure 4): both show saturation catalysis through a 4-hour experiment, with 9-CO at 24.6% conversion (16.5(1)x10 -2 TOF) relative to 9 at 11.7% conversion (8.3(1)x10 -2 TOF). These data indicate an important role for CO in the reactivity of catalyst 9. Further investigation of CO pressure revealed the expected inhibitory role at higher loading (Figure 5 and S31). At low concentration of CO either from FA decomposition or treatment with 2 bar of CO, 9 initiates at a faster rate than in the absence of CO. By contrast, under 8 bar of CO, we observe slower conversion of the catalytic reaction following rapid initiation as shown in Figure 5. As expected, 9-CO performed substantially similar to 9 when 9 is treated with 2 bar CO, but like 9, 9-CO exhibits inhibited rate when 8 bar CO is applied.
While seeking to understand the activation pathway of our most active precursor 11, a stable species 11-CO was isolated from a FA dehydrogenation reaction at ambient pressure (Scheme 1). Complex 11-CO was characterized by 1 H, 13 C, 19 F, and 31 P NMR spectroscopy and its molecular structure was established by single-crystal X-ray diffraction. Formation of carbonyl complex 11-CO under these conditions is a remarkable development, since FA dehydrogenation catalyzed by 11 is known to produce no free CO gas (< 10 ppm) and returns noncarbonylated catalytic species when operated at 90 o C. 6 This teaches us that at sufficient temperature and pressure, the previously characterized resting species from the 11-catalyzed dehydrogenation of FA can be further converted into a carbonylated system 11-CO. Again, we see that while FA decarbonylation happens during catalysis, CO is reformed rapidly to products and remains undetectable in the product stream.
To the best of our knowledge, precursor 11 continues to demonstrate comparable or superior activity to all known homogeneous systems for dehydrogenation of neat FA (Figure S4). It also provides excellent stability, longevity (TON > 2 million) and selectivity 6 (H2:CO2 1:1, CO < 10 ppm). We thus scaled this system to generate a pressurized product stream (> 103 bar) while demonstrating longevity and exceptional kinetics. To acquire high resolution data, we used a 600 mL stirred pressure vessel equipped with an internal temperature probe and a pressure transducer (see Supporting Information). We report volumetric flow rate (standardized to 1 atm at 0 o C) in units of liters per hour (L/hr corrected to ambient conditions) for all H2 evolution rates.
A 20-cycle pressure experiment was accomplished using 99.6 mg (145 µmol) of complex 11 and 20 g (294 mmol) of sodium formate co-catalyst in 55 mL of FA (Figure 6). During each cycle, ca. 50 mL of FA was added (1 L, 17.9 mol over 20 cycles), the reactor was sealed and heated to 120 °C, then the pressure was allowed to build to 117 bar (approximately 25 L of H2). Once the desired pressure was achieved, the reaction was quenched by rapid cooling in a dry-ice bath and then depressurized to repeat the cycle. The reaction rate in the form of evolved H2 per hour is plotted in Figure 6. This experiment illustrates that once pre-catalyst 11 is initiated, there is no detectable deactivation of the catalyst through 200,000 turnovers as evidenced by the consistently high peak reaction rates, varying only due to concentration differences between individual experiments. Cycle 4 demonstrated that when allowed to run near dryness, the peak reaction rate exceeded 160 L/hr, corresponding to a TOF of nearly 50,000 hr -1 .
Among a library of late-transition metal complexes, we found that in every case studied the dehydrogenation of neat FA is more productive under self-pressurizing reaction conditions. This is a stunning outcome, whereas the cost of hydrogen provided for retail vehicle filling is dominated by the cost of compressing the gas yet we have few detailed and broad-based studies that show how pressure evolution impacts the efficacy of homogeneous FA dehydrogenation catalysts. We grouped catalysts for neat FA dehydrogenation into four general classes, pincers, bidentate chelates, piano stool complexes, and metal precursors. Each structural class uniquely responds to pressurized condition. The bidentate chelates excel beyond others, which we attribute to a transformation from monomers to two-metal pseudo-pincer complexes in which the second metal seems to impart special reactivity. We find an enabling role for CO and/or H2 in a number of cases, typically impacting catalyst initiation (in-situ catalyst synthesis) and defining the course of catalyst speciation. This hypothesis is supported by previous studies 5,37 , namely, the observation and isolation of the carbonylation derivative of 11, and the 3-fold increase in a healing process of 10 by CO. In addition, complex 11, which exhibits exceptional catalytic activity, stability, and selectivity, supersedes existing systems in the production of a highpressure product stream from neat FA dehydrogenation. This catalyst was used to convert over 1 L of formic acid into pressurized H2/CO2 product over the course of 30 hours, proving that the catalytic activity could be maintained at a high level for 200,000 turnovers at 117 bar without any loss of reactivity. Due to the favorable economics of producing H2 at pressure, fully automated H2 generation using a continuous feed, stirred tank reactor will be developed to evaluate the ultimate longevity of the catalyst. This technology and the discovery of a detailed mechanism and speciation of 11 will be reported in future work. The reaction mixture volume was partially reduced under vacuum and filtered. The resulting filtrate was concentrated under vacuum, layered with pentane in the glovebox to obtain turmeric yellow crystalline solid in 2 days. They were filtered, washed with pentane, and dried in vacuum.
1 H NMR (600 MHz, C6D6): δ -11.6 (s, 1H, RuH), -7.5 (dd, 2 JHH = 18 Hz, 1H, RuH), δ 1.72 (s, 3H), δ 1.86-2.35 (m, 7H), δ 2.97 (d, 2H), δ 6.93-7.12 (m, 12H), δ 7.17-7.32 (m, 4H), δ 7.67 (t, J = 7.67 Hz, 8H). time (0 -25 hours): complex 1 -lavender crosses; complex 1 w/ t BuOK -purple squares; complex 2orange asterisks; complex 2 w/ t BuOK -grey circles; complex 3 -yellow plusses; complex 4 -blue squares; complex 4 w/ t BuOK -peach squares; complex 5 -green triangles; complex 6 -cream crosses; complex 7 -red asterisks; complex 8 -red circles; complex 9-CO -blue plusses; complex 10 -orange hyphens; complex 11 -black triangles;-; complex 12 -navy diamonds; complex 13 -grey hyphens; complex 14yellow crosses; complex 15 -blue asterisks; complex 16 -lavender diamonds; complex 17 -pink squares; time (0 -5 hours): complex 1 -lavender crosses; complex 1 w/ t BuOK -purple squares; complex 2 -orange asterisks; complex 2 w/ t BuOK -grey circles; complex 3 -yellow plusses; complex 4 -blue squares; complex 4 w/ t BuOK -peach squares; complex 5 -green triangles; complex 6 -cream crosses; complex 7 -red asterisks; complex 8 -red circles; complex 9-CO -blue plusses; complex 10 -orange hyphens; complex 11 -black triangles;-; complex 12 -navy diamonds; complex 13 -grey hyphens; complex 14yellow crosses; complex 15 -blue asterisks; complex 16 -lavender diamonds; complex 17 -pink squares; complex 18 -green hyphens; complex 19 -grey triangles; complex 20 -green circles; complex 21 -yellow
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 7