Homogeneous Catalysis for Sustainable Hydrogen ... - ACS Publications

Homogeneous Catalysis for Sustainable Hydrogen...

0 downloads 204 Views 10MB Size

Review pubs.acs.org/CR

Homogeneous Catalysis for Sustainable Hydrogen Storage in Formic Acid and Alcohols Katerina Sordakis,† Conghui Tang,‡ Lydia K. Vogt,‡ Henrik Junge,‡ Paul J. Dyson,† Matthias Beller,*,‡ and Gábor Laurenczy*,† Institute of Chemical Sciences and Engineering, École Polytechnique Fédérale de Lausanne (EPFL), Avenue Forel 2, CH-1015 Lausanne, Switzerland ‡ Leibniz-Institut für Katalyse an der Universität Rostock, Albert-Einstein-Straße 29a, D-18059 Rostock, Germany †

ABSTRACT: Hydrogen gas is a storable form of chemical energy that could complement intermittent renewable energy conversion. One of the main disadvantages of hydrogen gas arises from its low density, and therefore, eﬃcient handling and storage methods are key factors that need to be addressed to realize a hydrogen-based economy. Storage systems based on liquids, in particular, formic acid and alcohols, are highly attractive hydrogen carriers as they can be made from CO2 or other renewable materials, they can be used in stationary power storage units such as hydrogen ﬁlling stations, and they can be used directly as transportation fuels. However, to bring about a paradigm change in our energy infrastructure, eﬃcient catalytic processes that release the hydrogen from these molecules, as well as catalysts that regenerate these molecules from CO2 and hydrogen, are required. In this review, we describe the considerable progress that has been made in homogeneous catalysis for these critical reactions, namely, the hydrogenation of CO2 to formic acid and methanol and the reverse dehydrogenation reactions. The dehydrogenation of higher alcohols available from renewable feedstocks is also described. Key structural features of the catalysts are analyzed, as is the role of additives, which are required in many systems. Particular attention is paid to advances in sustainable catalytic processes, especially to additive-free processes and catalysts based on Earth-abundant metal ions. Mechanistic information is also presented, and it is hoped that this review not only provides an account of the state of the art in the ﬁeld but also oﬀers insights into how superior catalytic systems can be obtained in the future.

CONTENTS 1. Introduction 2. Formic Acid as a Hydrogen Storage Medium 2.1. The Carbon Dioxide−Formic Acid Couple 2.2. Carbon Dioxide Hydrogenation to Formic Acid 2.2.1. Noble-Metal-Based Catalysts 2.2.2. Non-Noble-Metal-Based Catalysts 2.2.3. Catalysts Active under Base-Free Conditions 2.3. Formic Acid Dehydrogenation 2.3.1. Noble-Metal-Based Catalysts 2.3.2. Non-Noble-Metal-Based Catalysts 2.4. Reversible H2 Storage 3. Methanol as a Hydrogen Storage Medium 3.1. The Carbon Dioxide−Methanol Couple 3.2. Carbon Dioxide Hydrogenation to Methanol 3.2.1. Noble-Metal-Based Catalysts 3.2.2. Non-Noble-Metal-Based Catalysts 3.3. Methanol Dehydrogenation 3.3.1. Noble-Metal-Based Catalysts 3.3.2. Non-Noble-Metal-Based Catalysts 4. Dehydrogenation of (Bio)Alcohols 4.1. Dehydrogenation of C2 Alcohols 4.2. Dehydrogenation of C3 Alcohols © 2017 American Chemical Society

4.3. Dehydrogenation of C4 Alcohols 4.4. Dehydrogenation of C5 and C6 Alcohols 4.5. Non-Noble-Metal-Based Catalysts 5. Conclusions Author Information Corresponding Authors ORCID Author Contributions Notes Biographies Acknowledgments Abbreviations References

372 373 373 374 374 381 385 386 386 396 400 403 403 404 404 407 407 407 410 411 411 416

419 419 420 420 421 421 421 421 421 421 422 422 423

1. INTRODUCTION The rapid increases in global energy demand caused by an explosion of the world’s population, increasing economic development, and higher standards of living, in addition to eﬀorts to replace traditional crude oil supplies linked to Special Issue: Sustainable Chemistry Received: March 30, 2017 Published: October 6, 2017 372

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

value and poses practical and safety concerns during handling and storage. Liquefaction, on the other hand, is a complex, energy-intensive (30−40% of energy content), and expensive multistage cooling procedure that requires storage tanks to be well insulated to prevent boil-oﬀ as a result of thermal conduction and radiation from the environment.21 To overcome the limitations of these traditional hydrogen storage methods, both physisorption in porous materials and innovative chemical methods have been proposed.22,25 Physisorption is a reversible process in which H2 is stored molecularly and not in its dissociated form through physical adsorption onto the surface of a porous solid material, where H2 uptake is governed by the speciﬁc surface area, pore structure, and pore size of the adsorbent. Several carbon materials (graphene, fullerenes, nanotubes, etc.),26 organic polymers with intrinsic microporosity,27 metal−organic frameworks,28 zeolites,29 and clathrate hydrates30 have been considered as adsorbents. The advantages of physisorption include fast kinetics, complete reversibility, and excellent cyclability.31 However, because of the weak van der Waals interactions between the hydrogen gas and the investigated solids (ΔHa = 4−10 kJ mol−1), satisfactory storage capacities are generally attained at low temperatures and/or high pressures. Chemical hydrogen storage methods refer to covalently bound hydrogen in either solid or liquid form, where hydrogen release is triggered by the thermal or catalytic decomposition of the carrier. To date, a number of solid carriers, such as simple metal hydrides, metal hydride alloys,32 metal borohydrides,33 alanates,34 imides/amides,35 and ammonia borane,36 have been proposed. Potential liquid (organic) hydrogen carriers (LOHCs) include cyclohexanes and heterocycles,37 ammonia,38 hydrazine and amine boranes,36,39 2-methylthiophene,40 formic acid,41−43 and alcohols.44 The present work focuses on homogeneous catalysts reported in the last two categories, from the ﬁrst examples through the most recent advancements as of the beginning of 2017. Following an overview of reversible CO2 hydrogenation to formic acid, its derivatives, and methanol, hydrogen release from alcohols incorporating two to six carbon atoms is discussed. (Because alcohols are not readily obtained by direct CO2 hydrogenation, their production is not summarized herein.)

spiraling environmental concerns, have triggered a worldwide debate about energy sustainability.1,2 The widespread use of fossil-fuel resources within the global energy infrastructure constitutes the primary source of anthropogenic emissions of carbon dioxide, an infamous greenhouse gas with implications for global climate change.3 Crude oil still accounts for the largest share of the total primary energy supply (31%) and over 90% of transportation fuels, followed by coal (29%) and natural gas (21%).4 Consequently, global CO2 emissions reached 35.7 Gt in 2014,5 translating into an atmospheric CO2 concentration of 408.84 ppm6 or a 56% increase since the 1990s. Strict emission regulations and the fear of energy supply shortages are therefore creating a need for renewable energy sources and fuels, which should be technically feasible, environmentally acceptable, and in the long term, economically competitive.7 Wind, solar, and hydroelectric energy technologies have been expanding throughout the past 50 years despite the fact that they suﬀer from time and region dependency, resulting in an intermittent energy supply. Hydrogen gas, on the other hand, constitutes a storable form of energy that can be utilized on demand and could therefore provide viable options for complementing these intermittent energy “gaps”.8 Moreover, “green” hydrogen can potentially minimize environmental impacts,9−11 depending on the characteristics and carbon footprints of the various steps constituting its life cycle, from raw material extraction and transformation, through plant construction and operation, to ﬁnal product distribution and utilization.12 Use of hydrogen in fuel cells, for instance, results in only water as a byproduct.11 A further prerequisite would be a shift in global hydrogen production from the current fossilfuel-based methods (mainly natural gas)13 to alternatives relying on either water (presently approximately 4% of the total H2 supply)14 or biomass15 as feedstocks. One of the main disadvantages of hydrogen arises from its low density. At 120 MJ kg−1, molecular H2 has the highest energy content of common fuels by weight; however, at 0.0108 MJ L−1, it also has a quite low energy density by volume. Therefore, eﬃcient handling and storage methods are key factors that need to be addressed for the successful introduction of H2 into our energy infrastructure. A storage system should be inherently safe, have satisfactory volumetric and gravimetric H2 capacities, have reasonable refueling times, have an adequate lifespan, be energetically eﬃcient, and be available at reasonable costs. For on-board H2 storage systems, the containment of a suﬃcient quantity of gas to satisfy the driving range requirements of the vehicle is a vital feature. In addition, weight and volume limitations are important; excess weight must be minimized, and excessive volumes should also be avoided.16 The U.S. Department of Energy has set the gravimetric and volumetric targets for on-board vehicle H2 storage to 5.5 wt % and 40 g L−1, respectively, by 2020.17 To date, none of the investigated H2 storage methods have been able to unite all of the aforementioned features. An overview of the various advantages and disadvantages of each technique can be found in several reviews.18−23 Conventionally, hydrogen is stored physically without the formation of strong chemical bonds, as a compressed gas (at 100−700 bar) or in its liqueﬁed form (at −253 °C).24 High pressures increase the volumetric energy density, reducing the tank volume but not its weight. In turn, advanced materials such as carbon ﬁbers with metal or polymer liners can reduce the weight of the vessel but increase the overall material cost. Furthermore, compression of H2 to 700 bar requires approximately 15% of its lower heating

2. FORMIC ACID AS A HYDROGEN STORAGE MEDIUM 2.1. The Carbon Dioxide−Formic Acid Couple

Formic acid (HCOOH, FA) is the simplest carboxylic acid (pKa = 3.75) and is a colorless liquid with a pungent odor at normal temperatures and pressures.45 Industrially, formic acid is most commonly derived from the hydrolysis of methyl formate obtained by methanol carbonylation or as a coproduct in acetic acid production, with a global capacity of 720 000 tons in 2013.46 Its liquid nature and low toxicity45 render its transportation, refueling, and handling straightforward. Depending on its concentration, materials such as stainless steel, high-density polyethylene, and polypropylene can be used for its safe containment.46 The possibility of storing hydrogen in the form of formic acid through the reduction of carbon dioxide was ﬁrst suggested by Williams et al. in the late 1970s.47 Compared to pressurized hydrogen at 200 bar with a volumetric density of 16 g L−1, formic acid has a higher hydrogen content of 53 g L−1 or 4.4 wt %. Even though this value is below the U.S. Department of Energy target of 5.5 wt 373

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 1. Reversible hydrogen storage cycles based on the FA/CO2 (left) and the formate/bicarbonate (right) couples.47,50

Table 1. Thermodynamic Data for Reversible H2 Storage Based on FA/CO2 Derivatives and Their Equilibria in Water54,57,58 ΔH° (kJ mol−1) HCOOH(l) → H2(g) + CO2(g) +31.2 HCOOH(l) → H2O(l) + CO(g) +28.7 CO2(g) + H2(g) → HCOOH(l) −31.2 CO2(g) + H2(g) + NH3(aq) → HCO2−(aq) + NH4+(aq) −84.3 CO2(aq) + H2(aq) + NH3(aq) → HCO2−(aq) + NH4+(aq) −59.8 MHCO3(aq) + H2(aq) → MHCO2(aq) + H2O(l) −20.5 CO2(aq) + H2O(aq) ↔ H2CO3(aq) ↔ HCO3−(aq) + H+(aq) ↔ CO32−(aq) + 2H+(aq) HCOOH(aq) ↔ HCOO−(aq) + H+(aq)

ΔS° (J mol−1 K−1)

ΔG° (kJ mol−1)

eq

+215 +138 −215 −250 −81 −66.4

−32.9 −12.4 +32.9 −9.5 −35.4 −0.72

(1) (2) (3) (4) (5) (6) (7) (8)

species are formed in equilibrium with CO2, whereas under basic conditions, FA is present in solution in the form of its conjugated base, formate (Table 1, eqs 7 and 8). Therefore, an equivalent hydrogen storage couple employing formate/ bicarbonate salts as initially proposed by Zaidman et al. (Figure 1, right)50 also represents a viable approach. The dehydrogenation of the formate carrier constitutes the on-demand hydrogen production step, whereas the hydrogenation reaction of the resulting bicarbonate salt can be employed for hydrogen ﬁxation (Table 1, eq 6). Conveniently, the handling of aqueous metal formate and bicarbonate solutions is uncomplicated because of their benign and noncorrosive nature. The H2 storage capacity of such a system is dictated by the solubilities of the carriers in the aqueous solution under the speciﬁc reaction conditions.

%, various mobile applications based on FA are under evaluation; nevertheless, FA is probably better suited to stationary power storage units. An ideal, carbon-neutral H2 storage/release couple involving FA and CO2 is illustrated in Figure 1 (left): Readily available and abundant CO2 is combined with green H2 gas to yield formic acid in a 100% atom-eﬃcient reaction, and FA is subjected to selective dehydrogenation to generate on-demand hydrogen.48,49 The decomposition of formic acid can proceed by two pathways: decarboxylation toward CO2 and H2, which is relevant for H2 release, and decarbonylation to CO and H2O, which should be inhibited (Table 1, eqs 1 and 2). The yield of each reaction pathway is aﬀected by several parameters such as the temperature, the substrate concentration, the presence of additional acid impurities, and the vessel surface,51 but it can be controlled using a suitable catalyst and reaction conditions. An overview of heterogeneous catalysts that are active in FA dehydrogenation is beyond the scope of the present work but can be found in recent reviews.52,53 The hydrogen storage step consists of the two-electron reduction of carbon dioxide resulting in the formation of formic acid. In the gas phase, this reaction, despite being exothermic, is largely endergonic because of unfavorable entropy conditions (Table 1, eq 3). However, if the product is trapped by a base (or solvent) in the form of formate salts, the reaction is thermodynamically feasible (Table 1, eq 4).54 The presence of bases is not a prerequisite for the catalytic hydrogenation of CO2 to formic acid; however, it does beneﬁt the equilibrium concentration and can also assist in heterolytic hydrogen activation.55 The entropy diﬀerence between the starting materials and the liquid product is also inﬂuenced by the solvation of the substrates or the raising of the entropy of the produced FA through disruption of the strong hydrogen-bond network (Table 1, eq 5). The latter can be achieved through interactions with strongly polarized solvent molecules, thus explaining the suitability of polar aprotic solvents in CO2 hydrogenation to FA.56 Solvents can also inﬂuence the catalyst through stabilization of the intermediate stages of the catalytic cycle and/or through coordination to the metal center. So far, several organic media, ionic liquids, supercritical carbon dioxide, and water have been tested as solvents for the conversion of carbon dioxide. In water, a set of pH-, temperature-, and pressure-dependent

2.2. Carbon Dioxide Hydrogenation to Formic Acid

Homogeneous carbon dioxide hydrogenation to FA has been more extensively studied than the reverse FA dehydrogenation reaction. The main reasons for this diﬀerence are related to the attractiveness of converting CO2 into useful chemicals,59−64 its potential utilization as an energy vector,47 and its use in various other applications ranging from leather processing to animal feed preservation and silage additives.65 The reaction has been studied under a broad range of conditions, in a variety of solvents including water, with both precious and nonprecious metal catalysts, in the presence of organic and inorganic bases, and in the absence of basic promoters. The reaction has also been performed under biphasic conditions, which facilitate product separation and catalyst recycling and integrated with a CO2 capture step. The majority of catalytic systems produce adducts, salts, or derivatives of FA because of the low equilibrium concentration of pure FA obtained from CO2 and H2 as a result of thermodynamic constraints. Over the years, these reactions have been discussed in a number of reviews, which have highlighted both the tremendous advances in the ﬁeld and the scientiﬁc and technological challenges that remain to be overcome.54,55,66−72 2.2.1. Noble-Metal-Based Catalysts. In the 1970s, the ﬁrst reports on homogeneous carbon dioxide hydrogenation to formate salts were published by Inoue and co-workers, who tested a series of transition-metal complexes (Pd, Fe, Co, Ni, Ru, Rh, and Ir) under relatively mild conditions in the presence 374

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

of alcohols and amines as cocatalysts.73,74 The highest activity [turnover number (TON) = 87, turnover frequency (TOF) ≈ 4.5 h−1] was exhibited by [RuH2(PPh3)4] in benzene/water containing triethylamine (NEt3) under 50 bar H2/CO2 (1:1) at ambient temperature.74 During the 1990s, Leitner and co-workers investigated the reduction of CO2 to FA and achieved record activities. For example, [Rh(COD)Cl]2/DPPB [COD = cyclo-1,5-octadiene, DPPB = 1,1-bis(diphenylphosphino)butane] in dimethyl sulfoxide (DMSO)/NEt3 at 25 °C under 40 bar H2/CO2 (1:1) results in a TON of 1150 (in 22 h).75 This was later improved to 3439 (in 12 h) with the Wilkinson-type catalyst [RhCl(mTPPTS)3] (mTPPTS = meta-trisulfonated triphenylphosphine, sodium salt) in H2O/dimethylamine.76 TOFs of 375 h−1 with [Rh(COD)(μ-H)4]/DPPB56 and 1335 h−1 in the presence of [Rh(hfacac)(dcpb)] [hfacac = hexaﬂuoroacetylacetonate, dcpb = (cyclohexyl)2P(CH2)4P(cyclohexyl)2],77 in both cases in DMSO/NEt3, were subsequently obtained. In the same solvent, screening of a series of mono- and bidentate phosphines in combination with [Rh(COD)(μ-Cl)2] yielded FA concentrations of up to 2.58 M,78 speciﬁcally, with 1,1bis(diphenylphosphino)methane (DPPM). Complexes formed in the presence of chelating ligands were activated prior to catalysis to overcome an induction period. From the study, it was concluded that monophosphines should be of medium basicity and have cone angles of less than 180° to eﬃciently catalyze FA formation. Furthermore, the ideal backbone length in chelating ligands R2P(CH2)nPR2 corresponded to n = 3 or 4 with either aryl or alkyl (R) substituents. In 2012, the authors presented a system for continuous supercritical CO2 (scCO2) hydrogenation to FA in a biphasic system, namely, the mobile reactant phase (scCO2 + H2) and the stationary phase consisting of a nonvolatile, basic ionic liquid (IL) containing the immobilized catalyst.79 The FA product was partially extracted into the mobile phase, allowing it to be recovered by decompression. Following catalyst and IL optimization under batch conditions, [Ru(COD)(methallyl) 2 ]/PBu 4 TPPMS (PBu4TPPMS = monosulfonated triphenylphosphine, in which the sodium cation has been exchanged for tetrabutylphosphonium) in 1-(N,N-diethylaminoethyl)-2,3-dimethylimidazolium bis(triﬂuoromethylsulfonyl)imide was heated to 50 °C, pressurized with 200 bar CO2, and subjected to a ﬂow of supercritical CO2/H2. With this arrangement, FA was formed with a TON of 358 after 211 h, albeit with signiﬁcant catalyst deactivation. With the same catalyst in a mixture of 1-ethyl-3methylimidazolium bis(triﬂuoromethylsulfonyl)imide and a diethylamine-functionalized polystyrene resin, a stable catalytic activity and a TON of 485 were reached within 200 h. Very recently, Leitner and co-workers developed another biphasic system for integrated CO2 hydrogenation and facile product separation employing cis-[Ru(DPPM)2Cl2] as the catalyst precursor,80 demonstrating that CO2 captured with commercial scrubbing amines can be directly converted into FA−amine adducts. Two types of biphasic systems were evaluated; (1) IL/ amine + H2O and (2) organic solvent/amine + H2O. In both cases, the catalyst was immobilized in a hydrophobic solvent, and the FA adduct was extracted into the aqueous phase. With the IL [OMIM][NTf2] [OMIM = 1-octyl-3-methylimidazolium, NTf2 = bis(triﬂuoromethylsulfonyl)amide], good TOFs and acid-to-amine ratios (AARs) were obtained with amines such as HNMe2, HNiPr2, and NEt3, although catalyst deactivation was observed following recycling. Organicsolvent-based systems were also studied, with methyl isobutyl

carbinol (MIBC) and methyldiethanolamine (Aminosol CST 115) found to provide a good compromise between catalytic activity and stability. Under 90 bar CO2/H2 (1:2) at 70 °C, an initial TOF of 41 000 h−1 and a total TON of more than 18 000 over 10 runs were reported (AARaver = 1, TOFaver = 8109 h−1). Despite an average P leaching of approximately 1% per run (Ru leaching was only 0.21%), reproducible pressure−time proﬁles were obtained. Further studies on coupled adduct formation and extraction were conducted in the presence of monoethanolamine. The highest AAR of 0.94 was obtained when the amine was presaturated with either 5 or 15 bar of CO2 and then pressurized to 85 or 75 bar with H2. At a constant CO2 pressure of 5 bar, the AAR increased linearly with H2 pressure between 25 and 85 bar. The overall sequence, namely, CO2 hydrogenation and product isolation, was repeated 11 times, reaching a total TON of 150 000, albeit resulting in almost complete catalyst deactivation. In conclusion, the feasibility of converting feedstocks obtained from a CO2 scrubber was illustrated, even though further improvements are necessary to ensure steady and continuous operation. The use of scCO2 as a reaction medium for its own hydrogenation to FA in the presence of an amine was introduced by the group of Noyori.81−84 The authors expected good catalyst activities, reasoning that the high solubility of H2 in scCO2 could help overcome mass-transfer limitations, which typically limit gas−liquid reactions. Indeed, using [RuH2(PMe3)4] and [RuCl2(PMe3)4], a maximum TOF of 1400 h−1 and TON of 7200 were obtained, in the presence of triethylamine and traces of water under 205 bar pressure at 50 °C.81 Addition of a promoter such as methanol or DMSO instead of water led to higher turnover frequencies of over 4000 h−1 in the presence of [RuH2(PMe3)4], likely due to enhanced catalyst solubility in the supercritical medium and/or stabilization of catalytic intermediates through hydrogen bonding.82 The rate of the hydrogenation reaction was found to be dependent on the number of phases present during catalysis. Following an extensive screening of additives, bases, and phosphine ligands,83 an unprecedented TOF of 95 000 h−1 (at 50 °C) was later reported by Jessop and co-workers upon replacing methanol with pentaﬂuorophenol in the presence of [RuCl(OAc)(PMe3)4] (OAc = CH3COO−) and NEt3 under 190 bar H2/CO2 pressure.84 The authors speculated that even higher rates should be expected with triﬂic acid and 1,8diazabicyclo[5.4.0]undec-7-ene (DBU); however, no concrete values were reported because of instrument limitations. In addition, equally high eﬃciencies were obtained with catalysts generated in situ from [RuCl2(C6H6)]2 and an appropriate phosphine ligand [PMe 3 , PPhMe 2 , DPPM, 1,1-bis(diphenylphosphino)ethane (DPPE), and cis- or transPh2PCHCHPPh2) in a MeOH/tripropylamine (NiPr3) solvent mixture.83 In the latter study, a general correlation between the properties of the ligands and the resulting catalytic activities was not possible. Nevertheless, the authors concluded that, in the case of monodentate phosphines, steric demands rather that electronic factors dictate catalyst performance. On the other hand, only weakly basic diphosphines with small bite angles resulted in active systems, whereas for more basic diphosphines, large bite angles led to more eﬃcient catalysts. The [RuCl(OAc)(PMe 3) 4 ] catalyst also promoted CO 2 conversion into an FA−amine adduct in neat NEt3 with a TOF of 500 h−1 obtained for the ﬁrst 30 min under 100 bar H2/CO2 (2:3) at 50 °C.85 375

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

CO2, the ﬁnal FA concentration was approximately halved. Moreover, the reaction was inhibited in heptane, benzene, and THF, whereas a 2-fold improvement, namely, to 2.5 M FA, was obtained in methanol. In addition to organic solvents, water has been extensively used in homogeneous CO2 hydrogenation at various pH values. Because of the involved acid−base equilibria (Table 1, eqs 7 and 8), a number of research groups have considered bicarbonate as a CO2-like hydrogen carrier. In these cases, despite bicarbonate being the substrate and reports on catalytic intermediates incorporating bicarbonate ligands,92,93 the formation of CO2 under the reaction conditions and its participation in catalysis (i.e., its direct reduction) cannot be unequivocally excluded. The ﬁrst studies were published by the groups of Joó, Laurenczy, and Gonsalvi, who studied the aqueous hydrogenation of bicarbonate to formate94−103 as an alternative reaction to the amine-promoted CO2 reduction investigated by the groups of Leitner, Noyori, and Jessop. Several ruthenium and rhodium catalysts incorporating watersoluble mTPPMS or PTA (PTA = 1,3,7-triaza-7-phosphaadamantane) ligands were evaluated under various conditions (pH, temperature, substrate concentration/type, H2/CO2 pressure). During the ﬁrst investigations, the initial reaction rate, which was strongly dependent on the pH of the solution, peaked at a value of 807 h−1 at 80 °C with [RuCl2(PTA)4] at pH ≈ 5.9 (i.e., 10% HCO3−/90% CO2).96 Subsequently, [RuCl2(mTPPMS)2]2/mTPPMS was found to hydrogenate a 0.3 M sodium bicarbonate solution with a TOF as high as 9600 h−1 under 60 bar H2 and 35 bar CO2 at 80 °C.97 The same catalyst was also used in reversible H2 storage in a formate/ bicarbonate couple, in the absence of additional CO2 gas.103 A rhodium(I) mTPPMS catalyst, namely, [RhCl(mTPPMS)3]/ mTPPMS, was reported to catalyze the formation of free FA in addition to formate when pressurized with 4H2/CO2 (Ptot = 100 bar) in a 0.1 M CaCO3 solution.99 Furthermore, this catalyst aﬀorded FA solutions of up to 0.13 M under 100 bar H2/CO2 (1:1) in the presence of 0.5 M NaOOCH.102 In 2003, Behr et al. introduced a biphasic process for homogeneous FA synthesis from CO2 coupled to catalyst recycling104,105 that was later developed by several research groups.79,80,106−108 The reaction occurs in basic, aqueous solution with TONs of up to 4980 (50 bar H2/CO2 at 60 °C) with the in situ generated RuCl3·xH2O/DPPBTS {DPPBTS = 1,3-bis[di(3-sulfonatophenyl)phosphino]propane, sodium salt} catalyst, and the FA product was subsequently extracted with N,N-dibutylformamide. Analysis of the raﬃnate revealed that between 6% and 9% of the catalyst was lost during this procedure and, therefore, an additional extraction of the organic phase with water was implemented prior to FA isolation by distillation. In 2004, the group of Himeda extensively investigated CO2 hydrogenation to formate with numerous noble-metal-based catalysts featuring proton-responsive ligands (mainly hydroxyfunctionalized).109−112 In these cases, an acid−base equilibrium of the substituents of the ligands was exploited to tune the catalytic activity and/or facilitate recycling through adjustment of their water solubility. Through changes in the pH of the reaction solution, the electronic properties and polarity of the catalyst can be controlled. Under basic conditions, the watersoluble oxyanion form of the catalyst, which was highly active in bicarbonate hydrogenation, was generated by deprotonation, whereas in acidic solution, the protonated hydroxyl form catalyzed FA dehydrogenation. It was initially reported that

In 1995, Lau and Chen achieved CO2 hydrogenation TONs of up to 5000 (in 8 h) in an ethanol solution containing cis[Ru(Cl2BPY)2(H2O)2][CF3SO3]2 (BPY = 2,2′-bipyridine) and triethylamine (60 bar H2/CO2 at 150 °C).86 Under these reaction conditions, FA was trapped either as a salt or as an adduct with the amine. In the absence of the base, ethyl formates were formed at signiﬁcantly lower yields, through CO2 reduction followed by esteriﬁcation. Interestingly, reducing the gas pressure to 40 bar while keeping the other parameters constant had almost no eﬀect on the yield or rate of the hydrogenation reaction. A few years later, Lau and co-workers observed that the CO2 hydrogenation barrier with [TpRuH(PPh3)(CH3CN)] [Tp = hydrotris(pyrazolyl)borate] (1) in tetrahydrofuran (THF)/NEt 3 mixtures was signiﬁcantly reduced in the presence of water.87 Under 50 bar of equimolar H2/CO2, the TON peaked at 760 within 16 h at 100 °C (THF/ NEt3/H2O = 3:0.4:1 volume ratio). Based on experimental and theoretical studies, the authors attributed the water-promoting eﬀect to more facile CO2 insertion into the RuH bond of a [TpRuH(PPh3)(HOH)] intermediate, because of CO2 stabilization through hydrogen bonding with the coordinated water molecule. Therefore, a catalytic cycle was proposed featuring a simultaneous hydride and proton transfer to CO2, yielding FA in a single elementary step. In a subsequent study, the activity of 1 was improved to 1815 turnovers after 16 h at 100 °C, when H2O was replaced by triﬂuoethanol.88 An analogous intermediate, namely, [TpRuH(PPh3)(CF3CH2OH)], enabling a stronger interaction between the acidic alcohol hydrogen and the CO2 substrate, was suggested to be responsible for the superior catalytic performance. An in situ generated catalyst based on RuCl3·xH2O/PPh3 was reported in 1996 for the synthesis of formate salts from CO2 by Zhang et al.89 Here, NEt3 was imperative for the eﬃciency of the reaction. The TON value peaked at 200 under 60 bar H2/CO2 in an ethanol/water solvent mixture, whereas inorganic basic additives such as ammonia or sodium carbonate signiﬁcantly hindered FA formation. Similarly, the presence of carbon monoxide had a detrimental eﬀect on catalysis, which was related to the formation of the catalytically inactive species [Ru(CO)3(PPh3)2]. The binuclear ruthenium phosphine complex [Ru2(μ-CO)(CO)4(μ-DPPM)2] was tested for reversible H2 ﬁxation in FA by Puddephatt and co-workers.90 Although a basic additive was not required during FA decomposition, the CO2 hydrogenation reaction proceeds only in the presence of triethylamine. A TOF of 116 h−1 and a TON of 1050 corresponding to a 1.4 M FA solution were reached in an acetone/NEt3 mixture at room temperature (Ptot = 38 bar). Under otherwise-identical reaction conditions, doubling the gas pressure enhanced the FA formation (2.8 M), aﬀording a TON of 2160 without compromising the reaction rate (TOF = 103 h−1). Wilkinson’s catalyst had already been considered for CO2 hydrogenation in the pioneering work by Inoue and coworkers,73,74 when it was subsequently reported by the group of Kolesnichenko to exhibit the best performance from among several rhodium catalysts tested.91 With 6 equiv of free triphenylphosphine per Rh center, a 0.95 M FA solution (TON = 950) was obtained under 60 bar H2/CO2 (2:1) in DMSO/NEt3 after 20 h at 25 °C. The applied gas pressure, the ratio between PPh3 and rhodium, the type of gaseous atmosphere during dissolution of the metal , and the nature of the solvent were crucial. A protective argon or H2 atmosphere led to similar FA yields, whereas under air or 376

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 2. Structures of iridium complexes tested in aqueous CO2 hydrogenation in basic media.109,111,113−115,117

analogue 5 (Figure 2d) revealed that catalytic activity was not signiﬁcantly aﬀected by the number of iridium centers.114 The presence, position, and amount of hydroxyl groups, on the other hand, was directly related to the rate of bicarbonate/CO2 hydrogenation as a result of a combined pendent-base eﬀect and accelerated proton transfer through a water bridge, thus enhancing the rate-determining H2 heterolysis step. Consequently, 5 exhibited an activity comparable to that of 4 under equally mild reaction conditions, namely, an initial TOF of 66 h−1 under 1 bar H2/CO2 (1:1) at 25 °C in an aqueous 2 M KHCO3 solution. Interestingly, similar values were later achieved in the presence of iridium complex 6 (Figure 2e) bearing nonfunctionalized imidazoline groups.115 At ambient temperature and pressure (H2/CO2), formate was formed at a rate of 43 h−1 in a 1 M NaHCO3 solution. Evidence was provided that deprotonation of the NH moieties was not responsible for the catalytic performance. Under comparable conditions, only Fukuzumi and co-workers’ phenylpyrazolyl iridium catalyst (see Figure 19b, below) was reported to be active, with a TOF of 6.5 h−1.116 Complex 7 (Figure 2f) incorporating pyrimidine and imidazole ligands was somewhat less active with TOFs ranging from 440 h−1 at 50 °C to 6440 h−1 at 120 °C under 10 bar equimolar H2/CO2 pressure in basic media.117 Nozaki and co-workers synthesized and tested the iridium trihydride PNP1 pincer catalyst 8 [PNP1 = 2,6-bis(di-isopropylphosphinomethyl)pyridine] that was capable of hydrogenating CO2 in a 1 M aqueous KOH solution in the presence of THF (Figure 3a).118 The rigid, tridentate binding mode of pincer ligands enhances the stability of the metal center, and the activity of the system can be tailored through steric and electronic ligand modiﬁcations, without signiﬁcantly altering

half-sandwich Ru(II), Ir(III), and Rh(III) complexes bearing the 4,7-dihydroxy-1,10-phenanthroline (DHPHEN) ligand eﬃciently catalyzed bicarbonate hydrogenation in alkaline solutions.109 In a typical experiment, a 1 M KOH solution containing the catalyst was saturated with CO2, subsequently pressurized with H2/CO2 (1:1), and thermostated. The highest activity, namely, initial a TOF of 23 000 h−1 and a TON of 21 000, was exhibited by [Cp*Ir(DHPHEN)Cl]Cl (Cp* = 1,2,3,4,5-pentamethylcyclopentadienyl) (2, Figure 2a) at 120 °C under 60 bar pressure, which translated into a ﬁnal formate concentration of 0.42 M. This complex was active even at ambient pressure with a TOF of 35 h−1 at 80 °C. The ruthenium analogue, [(C6Me6)Ru(DHPHEN)Cl]Cl, aﬀorded 3-fold-higher ﬁnal formate concentrations, albeit with a decreased TOF of 3360 h−1 at 120 °C. Interestingly, the iridium and rhodium complexes with nonsubstituted 1,10phenanthroline (PHEN) groups were practically inactive. Moreover, under the basic reaction conditions that were crucial to ensure high catalytic activity, the hydroxyl groups of the PHEN ligand were fully deprotonated. The authors concluded that the strong electron donation resulting from the resonance structure incorporating the oxyanions, in addition to the acquired water solubility attributed to the presence of these substituents, was responsible for the signiﬁcant catalyst performance. By reducing the catalyst concentration, the obtained activity was later improved to 33 000 h−1 under otherwise-identical reaction conditions.111 Recycling of [Cp*Ir(DHPHEN)Cl]Cl was realized by simple ﬁltration of the catalyst, which spontaneously precipitated at the end of the hydrogenation reaction as a result of a pH shift to less basic values (related to the conversion of bicarbonate to formate), albeit with approximately 8% loss of iridium per cycle due to leaching.110 An equivalent iridium catalyst (3) with a 4,4′dihydroxy-2,2′-bipyridine donor (4,4′-DHBP) (Figure 2b) that was later developed based on the same concept exhibited an even higher initial activity of 42 000 h−1 and TON of 190 000 under identical reaction conditions.111 The groups of Hull and Fujita prepared the binuclear iridium complex 4, incorporating a bipyrimidine ligand with four hydroxyl functionalities (THBPM = 4,4′,6,6′-tetrahydroxy-2,2′-bipyrimidine) (Figure 2c) that yielded initial TOFs of 53 800 h−1 (TON = 79 000) and 70 h−1 in the hydrogenation of 2 M KHCO3 under 50 bar H2/CO2 (1:1) at 80 °C and 1 bar H2/CO2 (1:1) at 25 °C, respectively.113 Comparison of 4 with its mononuclear

Figure 3. Structures of metal pincer catalysts active in CO2 hydrogenation in basic media.118,120,122 377

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

the coordination environment of the metal.119 The TOF value peaked at 150 000 h−1 at 200 °C under 50 bar H2/CO2 (1:1), and the TON peaked at 3 500 000 at 120 °C under 60 bar gas pressure. At that time, the aforementioned values constituted 1.5-fold and 15-fold improvements, respectively, over previous activities under similar reaction conditions and were only recently surpassed by the group of Pidko.120 The strength of the base, the presence of a very polar solvent, the pressure, and the temperature were crucial parameters inﬂuencing the activity of 8.121 Through computational and experimental studies, Hazari and co-workers demonstrated that a strongly electron-donating trans ligand (for example, a hydride) and the presence of a Hbond donor in the second coordination sphere of the metal facilitated CO2 insertion through weakening of the iridium hydride bond and stabilization of the formed formato complex 9 (Figure 3b).122 The stabilization was attributed to hydrogen bonding between the nonbound oxygen of the formate group and the NH proton of the PNP2 pincer linker [PNP2 = HN(CH2CH2(PiPr2))2]. Reasoning that CO2 insertion was reversible, the authors speculated that 9 should be active in carbon dioxide hydrogenation. As expected, 9 catalyzed the reaction with rates of over 18 000 h−1 at 185 °C under 55 bar H2/CO2 (1:1) in a 1 M aqueous KOH solution. Furthermore, a TON of 348 000, equivalent to a formate yield of 70%, was obtained after 24 h. A ruthenium PNP3 [2,6-bis(di-tert-butylphosphinomethyl)pyridine] pincer complex (10) bearing hydride, chloro, and carbon monoxide ligands (Figure 3c) was reported in 2014 by Pidko and co-workers and is the most eﬃcient catalyst to date for CO2 hydrogenation under basic conditions.120 The base aﬀects the thermodynamics of the reaction, that is, the obtained acid-to-amine ratio, but not the rate of formate formation. In contrast to other catalysts that exhibited optimal performance in the presence of NEt3, 10 aﬀorded a maximum TOF of 1 100 000 h−1 and an AAR of 1.1 in a dimethylformamide (DMF)/DBU solvent mixture at 120 °C under 40 bar pressure (3H2/1CO2), whereas higher reaction temperatures yielded improved AARs, allowing better H2 storage capacities without compromising the reaction rate. Even under mild conditions, namely, 5 bar of an equimolar gas mixture and 90 °C, an activity of 60 000 h−1 was obtained. Several neutral and cationic ruthenium complexes bearing chelating (η2-N,O) ligands (Figure 4) were synthesized and tested for CO2 hydrogenation in alkaline aqueous solution by Süss-Fink and co-workers.123 Even though low activity was observed with several catalysts, the highest TON of 400 (t = 10 h) was obtained with [(η6-p-cymene)Ru(η2-HN,H,HO)(H2O)]-

BF4 (R1 = R2 = R3 = H) in an aqueous 0.25 M NEt3 solution under 100 bar H2/CO2 at 100 °C. Fachinetti and co-workers demonstrated that [RuCl2(PMe3)4] can catalyze the hydrogenation of CO2 in neat triethylamine as the solvent without further promoters or cocatalysts.124 This ruthenium catalyst was previously investigated by Noyori and co-workers under slightly diﬀerent reaction conditions, namely, in scCO2/NEt3 containing traces of H2O.81,82 To overcome an induction period, Fachinetti and co-workers chose to use a biphasic medium incorporating the pure (free) amine and a separate phase, an acid−amine adduct with a ratio of 1.33. Their experiments suggested that the latter was the active species, whereas even in the absence of free amine (i.e., in a single-phase medium), a 1.78 adduct was formed at 40 °C under 120 bar equilibrating pressure (i.e., the pressure leveled oﬀ that the end of the reaction). Following distillation, the AAR was increased to 2.35, corresponding to the azeotropic concentration. The reaction temperature was kept at the aforementioned value to avoid catalyst deactivation. A maximum AAR of approximately 2 was reached under a higher equilibrating pressure of 160 bar. Beller and co-workers evaluated a series of phosphine ligands in combination with [RuCl2(benzene)]2 for their activity in the hydrogenation of sodium bicarbonate in water or water/THF mixtures in the case of ligands with low solubility in the aqueous solvent.125 Initially, the highest formate yield of 30% (TON = 1374) was obtained with a Ru/DPPM (1:4) catalyst under 50 bar H2 after 2 h at 70 °C. At lower temperatures, the reaction did not occur unless the catalyst was preactivated at 70 °C. Addition of CO2 gas (35 bar) further enhanced formate formation with a maximum TON of 2793 obtained at 100 °C. The yield of NaHCO3 hydrogenation was later improved to 96% (TON = 1108) under 80 bar H2 and otherwise-identical reaction conditions.126 Furthermore, a family of air- and moisture-stable benzoyl- and naphthoyl-substituted phosphine ligands was synthesized and tested together with ruthenium precursors for sodium bicarbonate hydrogenation.127 Moderate activity was exhibited by several of them, with dicyclohexyl 1naphthoyl phosphine (L1, Figure 5) together with [Ru(COD)-

Figure 5. Structure of naphthoyl-substituted phosphine reported by Beller and co-workers for bicarbonate hydrogenation.127

(methylallyl)2] in methanol giving the highest formate yield of 65% and a TON of 7188 under 60 bar H2 at 80 °C. Replacing the bicarbonate salt with CO2 and performing catalysis in a NEt3/MeOH solution produced methyl formate with a maximum TON of 4126 at 100 °C. Paciello and Schaub developed and optimized an industrially scalable process, related to one previously reported by Behr et al.,104,105 based on integrated CO2 hydrogenation, catalyst recycling, and formic acid isolation.106,107 Catalysis was realized in a biphasic diol/trihexylamine (NHex3) medium, resulting in the formation of a NHex3·FA salt that dissolved in the diol but was immiscible with the free amine. Upon completion of the reaction, the separate amine phase containing most of the homogeneous ruthenium catalyst [Ru(H)2(PnBu3)4] was recycled back to the hydrogenation step, and the product-rich

Figure 4. Structure of ruthenium catalysts tested for CO2 hydrogenation in basic media by Süss-Fink and co-workers.123 378

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 6. Structures of ruthenium and iridium catalysts tested for CO2 hydrogenation in basic media by Peris and co-workers.128−130

phase was extracted with NHex3 to recover catalyst traces. Finally, pure FA was separated from the diol and amine by distillation. In the absence of diol, calorimetric measurements showed that FA was not formed, because the exothermic amine protonation reaction did not supply enough enthalpy to justify an exergonic process. The concept of transfer hydrogenation of CO2 toward formate was ﬁrst proposed by the group of Peris in 2010.128,129 Ruthenium and iridium complexes incorporating strongly electron-donating N-heterocyclic carbene (NHC) ligands, expected to endow the metal with high thermal stability, were evaluated for FA formation from CO2 in the presence of either hydrogen gas or isopropanol (iPrOH) as the hydrogen donor. The initial maximum TON of 1600, obtained with iridium catalyst 11 under 60 bar H2/CO2 in a 2 M KOH solution after 64 h at 80 °C,128 was subsequently improved to 23 000 (75 h) in the presence of the bis-NHC1-containing ruthenium complex 13 (Figure 6), interestingly under lower pressure (40 bar) and less basic conditions (1 M KOH) but at a higher reaction temperature of 200 °C.129 In the case of transfer hydrogenation with iPrOH as the hydrogen donor, the activity of 13 was again superior, with a TON of 874 (50 bar CO2, 0.5 M KOH, 200 °C, 16 h)129 compared to the initial value of 150 aﬀorded by 12 (50 bar CO2, 0.5 M KOH, 110 °C, 72 h).128 Encouraged by these promising results, Peris and co-workers decided to modify an analogous iridium complex for CO2 hydrogenation in water.130 The catalyst was chosen based on its high stability toward air and moisture, as well as its excellent performance in the transfer hydrogenation of ketones.131 The introduction of sulfonato groups further enhanced the water solubility of the bis-NHC2-containing catalyst 14 (Figure 6) that aﬀorded a TON of 190 000 under optimized conditions, namely, 60 bar H2/CO2 in a 1 M KOH solution at 200 °C.130 Performing the reaction in H2O/iPrOH in the absence of H2 gas led to a decrease in the catalytic activity, to TON = 2700, under comparable conditions (0.5 M KOH, 50 bar CO2, 200 °C). Two additional alcohols, namely, cyclohexanol and 1-phenylethanol, were evaluated as transfer hydrogenation agents in the presence of 12 but led to signiﬁcantly lower formate yields.128 The group of Aresta proposed the use of glycerol, which is a side product in biodiesel production, as a H donor for CO2 reduction, in an approach that could convert two waste products simultaneously into FA.132 However, the authors reported that, even with their best catalyst precursors, [RuCl2(PPh3)3] [Ru(COD)Cl2]x/PPh3, only a few turnovers could be accomplished. In 2014, Beller and co-workers reported the transfer hydrogenation of CO2 and bicarbonate with methanol as the hydrogen donor and solvent under basic conditions.133 The ruthenium PNP4 [PNP4 = HN(CH2CH2(PPh2))2] pincer complex 15 (Figure 7) catalyzed formate formation in a

Figure 7. Structure of ruthenium pincer catalyst 15 active in the transfer hydrogenation of CO2 and bicarbonate with methanol as the hydrogen donor.133

NaOH/MeOH solution even in the absence of bicarbonate/ CO2, through methanol dehydrogenation,134 with a TON of 130 after 20 h at 100 °C. Nevertheless, when sodium bicarbonate was introduced into the system as a hydrogen acceptor, improved TONs were reached under otherwiseidentical reaction conditions. At an optimized NaOH/NaHCO3 ratio of 3, 3266 turnovers were obtained in a MeOH/H2O solvent after 20 h at 150 °C. Higher substrate concentrations were not beneﬁcial for catalysis because of partially hindered methanol dehydrogenation at lower pH values. Use of potassium salts instead of their sodium analogues, further enhanced formate formation with a maximum TON of 18 422 after 36 h at 150 °C. (In the absence of KHCO3, the corresponding formate salt was formed with a TON of 8108.) In this case, the ideal KOH/KHCO3 ratio was 4. CO2 was found to be a suitable hydrogen acceptor, as indicated by the increase in the TON value from 10 880 to 12 308 at 150 °C in the absence and and presence of 5 bar CO2, respectively. However, higher carbon dioxide pressures inhibited formate formation. The eﬃcient capture and subsequent in situ hydrogenation of CO2 was ﬁrst realized in 2013 by the group of He135 and later by the groups of Heldebrant,136,137 Prakash and Olah,108 and Leitner and Franciò.80 Polyethylenimine PEI600 (Mw = 600, Figure 8a) is stable and nonvolatile and can absorb CO2 with a capacity of 0.159 g of CO2/0.3 g of amine (in ethylene glycol), resulting in the formation of an ammonium alkylcarbonate in the presence of an alcohol (or ammonium carbamate in the absence of alcohol).135 Screening experiments showed that RhCl3·3H2O and PPh2Cy (10 equiv) in methanol under 80 bar CO2/H2 (1:1) at 60 °C and PEI600 aﬀorded a TON of 852 after 32 h (or a TON of 542 after 16 h). Subsequently, CO2 captured by the amine was hydrogenated (40 bar H2) in an ethylene glycol/MeOH solvent at 60 °C, with 37% conversion to formate or a TON of 260 (16 h). In a subsequent study under identical reaction conditions, the in situ hydrogenation of CO2 gave nearly quantitative yields (96−99%) when PPh2Cy was replaced by bis(2-diphenylphosphine phenyl)ether and PEI600 was replaced by either of the superbases shown in Figure 8b,c.138 The CO2 capacities of DBUOH, DBUOH/TGDE (3:5 mol ratio), DBUOH@silica/TGDE (0.24:5), and DBUOH@ silica/ethylene glycol (0.24:24) are 0.58, 0.95, 3.02, and 3.48 379

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 8. Structures of bases used for CO2 capture and in situ hydrogenation to formate by the group of He.135,138

with a TON of 10 775 and a TOF of 1096 h−1 under 61 bar H2/20 bar CO2 at 70 °C. Further details of the reactions are provided in section 2.4. More recently, catalyst 16 and its iron-based equivalent (see section 2.2.2) were reported for the simultaneous capture and conversion of CO2 to formates.108 In this approach, carbon dioxide was eﬃciently absorbed and activated by several commercial amines containing primary, secondary, and tertiary amino groups, forming carbamates and/or bicarbonates/ carbonates. Although pentaethylenehexamine (PEHA) has the highest CO2 absorption capacity of 12.1 mmol g−1 (amount captured by bubbling CO2 through the solvent for 30 min at room temperature), in situ hydrogenation with 16 aﬀorded higher yields (91−95%) in the presence of the superbases diazabicyclo[2.2.2]octane (DABCO), DBU, or tetramethylguanidine (TMG). Catalysis was performed in a 1,4-dioxane/water medium under 50 bar H2 at 50−55 °C, and the TON and TOF values peaked at 7375 (yield = 95%) and 433 h−1, respectively, with TMG and DABCO. Subsequently, to examine the recyclability of 16 in a biphasic system, 1,4-dioxane was exchanged for 2-methyltetrahydrofuran containing the catalyst, and the bicarbonate/carbonate salts of DABCO were dissolved in the aqueous layer. Under 50 bar H2 pressure at 55 °C, the organic phase was reused ﬁve times with TOFs in the range of 560−690 h−1 and with excellent yields between 94% and 97%. Overall, a TON of 7130 was reached with minimal ruthenium leaching into the aqueous phase (less than 3 ppm). The identiﬁcation of [(PNP4)RuH(OOCH)(CO)] as the catalyst resting state led the authors to conclude that the formate elimination step was most likely rate-determining. Three noble-metal-based complexes with bis-N-heterocyclic carbene ligands were synthesized and tested for both H2 storage and release by Kühn and co-workers (Figure 10a).141 Despite signiﬁcantly diﬀerent activities in FA dehydrogenation (see section 2.3.1), all three catalysts exhibited similar behaviors in bicarbonate hydrogenation (50 bar) at 100 °C, with TONs in the range of 1000−1300 after 24 h (conversion = 54−64%). Complex 18 was chosen for further investigations, presumably because of its superior activity in FA dehydrogenation compared to the iridium and ruthenium analogues, with a maximum TON of 3600 (36% conversion) reached at a catalyst concentration of 0.2 mM after a reaction time of 72 h. Taking into account the availability of a single coordination site during catalysis (because of the presence of the Cp* ring and chelating bis-NHC ligand), a computational study was performed to elucidate possible reaction mechanisms. From the calculated energy barriers, it was concluded that the reduction of coordinated bicarbonate by a neighboring hydride was more favorable than that by molecular dihydrogen. As such, a second catalyst molecule providing the aforementioned hydride was likely to be involved in the reaction mechanism, thus

molCO2/molbase, respectively (TGDE = tetraethylene glycol dimethyl ether). The advantage of using silica-supported DBUOH is related to the facile isolation of free FA by ﬁltration (to remove MeOH and ethylene glycol or TGDE and catalyst) and distillation. The recovered DBUOH@silica could be reused for CO2 hydrogenation, albeit with a 10% decrease in formate yield observed during the third cycle. In their approach, Heldebrant and co-workers used a switchable ionic liquid comprising DBU and methanol to hydrogenate captured CO2 into an alkyl formate in the presence of known ruthenium catalysts.136 Switchable ionic liquids have the ability to reversibly bind carbon dioxide, resulting in a signiﬁcant change in polarity. In the presence of [RuCl2(PPh3)3] under 20 bar H2, CO2 in the form of [DBUH+] methyl carbonate was converted to [DBUH+] formate and methyl formate with a TON of up to 5100 after 40 h at 140 °C. At lower temperatures and with excess base, [DBUH+] formate was the only hydrogenation product, with a conversion of 24% (TON = 510) after 16 h at 110 °C. Higher temperatures and an excess of methanol, on the other hand, promoted the esteriﬁcation reaction toward methyl formate with concomitant release of the base and resulted in the hydrolysis of DBU. The ruthenium PNP pincer catalysts shown in Figure 9 were tested for H2 storage in organic solvent/H2O solutions of

Figure 9. Structure of ruthenium pincer catalysts reported by Olah and Prakash and co-workers for CO2 capture and hydrogenation.139

formate/bicarbonate/carbonate salts by the groups of Czaun, Prakash, and Olah.139 Previously, 16 had been reported to be active in CO2 reduction to methanol.140 The H2 storage step was realized under variable conditions, that is, in the presence of alkali hydroxides, carbonates, and bicarbonates and, in certain cases, under CO2 pressure in the temperature range of 70−85 °C.139 Formate yields greater than 90% were reached with 15 (Figure 7) and 16 when sodium bicarbonate in THF/ H2O served as the substrate under 40 bar H2 at 80 °C, corresponding to TOFs of 1688 and 958 h−1, respectively. Complex 16 also exhibited high activity in a LiOH solution under 60 bar H2/CO2 (1:1) and in a Na2CO3 solution under 38 bar H2/12 bar CO2 with approximate TOFs of 2000 h−1 in both cases (T = 75−79 °C). The NH functionality in the ligand backbone was not essential for catalysis; 17 transformed CO2 into formate in a 1,4-dioxane/H2O solution of NaOH 380

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

resulted in a maximum TON of over 7500 at 100 °C when the bicarbonate concentration was increased, whereas the TOF peaked at a value of 770 h−1.145 Exchanging the bicarbonate substrate for CO2 gas (30 bar, 1 equiv of H2) and performing the reaction in a MeOH/NEt3 solution resulted in the formation of both FA and methyl formate with a total TON of almost 1700. In the presence of water, the production of methyl formate was suppressed, and that of FA was slightly enhanced (TON = 1897). Furthermore, using dimethylamine instead of triethylamine yielded DMF as the main reaction product along with FA, with a total TON of 5104. In 2011, Milstein and co-workers reported the ﬁrst iron PNP3 pincer catalyst (21) that is active in both bicarbonate and CO2 hydrogenation under low hydrogen pressures (Figure 11a).146 With bicarbonate as the substrate, the highest TON of

Figure 11. Structures of iron pincer catalysts active in CO 2 hydrogenation in basic media developed by (a) Milstein and coworkers146 and (b) Gonsalvi and co-workers.147 Figure 10. (a) Structures of catalysts with N-heterocyclic carbene ligands tested for reversible H2 storage and (b) proposed reaction mechanism for bicarbonate hydrogenation. (Adapted with permission from ref 141. Copyright 2016 John Wiley & Sons.)

320 (32% formate yield) was obtained at 80 °C with just 8.3 bar H2 in aqueous solution containing an essential THF cosolvent. Enhanced catalytic activity was exhibited in the CO2 hydrogenation reaction, with a maximum TON of 788 and TOF of 156 h−1 under 10 bar pressure (H2/CO2 = 2:1) in an aqueous NaOH/THF solution. Subsequently, the groups of Kirchner and Gonsalvi demonstrated that iron PNNNP1 pincer complexes with either NH or NMe spacers could also catalyze the CO2/ bicarbonate transformation under comparable reaction conditions (Figure 11b).147 A TON of 140 (14% conversion of bicarbonate to formate) was achieved with 22 under 8.5 bar H2 pressure at 80 °C for 16 h in a H2O/THF (4:1) solvent mixture. This value increased to 1964 (98% conversion) and 4560 (23% conversion) after 24 h under a higher H2 pressure of 90 bar with catalyst loadings of 0.05 and 0.005 mol %, respectively. In neat THF, no reaction was observed, suggesting that a protic solvent was imperative, presumably because of the H-bond stabilization of catalytic intermediates. The hydrogenation of CO2 in basic media (0.5 M NaOH) proceeded with TONs of up to 1220 and quantitative yields in the presence of 22 under 80 bar pressure at 80 °C. Exchanging NaOH for DBU, dimethyloctylamine (DMOA), or NEt3 in ethanol as the solvent resulted in either complete inhibition of the reaction or very low formate formation. Complex 23, featuring tertiaryamine pincer arms, was less active than 22 when NaOH was used as the basic additive in H2O/THF (TON = 680) but exhibited improved performance in a DBU/EtOH mixture (TON = 1153). In the latter solvent, even at ambient temperature (25 °C), 856 turnovers were obtained after 21 h, corresponding to 86% conversion to formate. At a lower catalyst loading of 0.01 mol %, quantitative formation of formate was reported at 80 °C, corresponding to a TON of approximately 10 000.

accounting for inferior hydrogenation rates at reduced metal concentrations (Figure 10b). 2.2.2. Non-Noble-Metal-Based Catalysts. In the ﬁrst pioneering report describing homogeneous CO2 hydrogenation to FA, Inoue et al. obtained low activities (TON = 7) in the presence of [Ni(DPPE)2] and triethylamine at ambient temperature.74 In 2003, Jessop and co-workers screened a series of non-platinum-group metal salts (Co, Cr, Fe, In, Mo, Nb, Ni, W) alone or in combination with the phosphine and nitrogen donor ligands PPh3, DPPE, DCPE (Cy2PCH2CH2PCy2), TMEDA (N,N,N′,N′-tetramethylethylenediamine), and 2,2′-bipyrindine.142 The TON peaked at a value of 4400 (TOF = 20 h −1 ) when pretreated [NiCl2(DCPE)] (1 h under 40 bar H2) was used for the reaction in a DMSO/DBU solution, yielding an AAR of 0.37 after 216 h (40 bar H2, 160 bar CO2). In 2010, Beller and coworkers presented the ﬁrst example of iron-catalyzed bicarbonate hydrogenation in the absence of additional CO2 gas.143 Tris[2-(diphenylphosphino)ethyl]phosphine (PP3) was the only ligand found to aﬀord an active catalyst in combination with Fe(BF4)2·6H2O. The highest TON for formate formation was 610 after 20 h at 80 °C in methanol and under 60 bar H2 pressure. Under identical conditions, the equivalent cobalt catalyst, namely, Co(BF4)2·6H2O/PP3, exhibited a similar activity (TON = 645),144 although in a subsequent study the catalyst formed from Fe(BF4)2·6H2O/tris[(2diphenylphosphino)phenyl]phosphine (L2)145 was found to be superior, with a TON of almost 1600. In the former case, increasing the temperature to 120 °C yielded 3877 turnovers at an average rate of 190 h−1.144 The Fe(BF4)2·6H2O/L2 system 381

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 12. (a) Structures of iron and molybdenum pincer catalysts 24−26 (for 24 and 25, P = PiPr2) and (b) proposed mechanism for CO2 hydrogenation with 24, accounting for the beneﬁcial eﬀect of LiOTf.152,154

reaction under similar conditions, namely, in the presence of a Lewis acid (LA).152 Catalysis was realized in a DBU/THF solution at 80 °C under 69 bar H2/CO2 (1:1). Iron(II) carbonyl hydrides with either a NH- or NMe-containing PNP supporting framework were found to be active catalysts even in the absence of a Lewis acid, although addition of the latter signiﬁcantly improved the reaction rate and overall yield. Upon optimization of the catalyst structure, addition of lithium triﬂate (LiOTf) provided the best results, with TON and initial TOF (1 h) values of 58 990 and 23 190 h−1, respectively, for 24 and 25 (Figure 12a). For the catalysts incorporating a secondary amine, the LA-promoted enhancement in activity was attributed to the disruption of intramolecular hydrogen bonding between the NH proton and a coordinated formato ligand, thus accelerating the rate-determining formate elimination step. The authors had previously identiﬁed the same Hbond stabilization eﬀect to be favorable for CO2 insertion into the metal−hydride bond of an analogous iridium pincer catalyst that exhibited the opposite trend of assisting CO2 hydrogenation.122 Finally, for iron complexes bearing a NMe moiety as part of the pincer structure, the main role of the LA was proposed to be the cation-assisted substitution of a formate ligand by dihydrogen (Figure 12b). More recently, the group of Bernskoetter synthesized the equivalent cobalt catalyst, namely, [(PNP5)Co(CO)2]Cl (PNP5 = MeN(CH2CH2(PiPr2))2), which was then tested together with LiOTf for CO 2 hydrogenation in DBU/organic solvent mixtures.153 The highest TON of 29 000 (TOF1 h = 5700 h−1) was achieved in acetonitrile after 16 h at 45 °C under 69 bar H2/CO2. Higher temperatures favored the reaction rate, with a TOF of 12 000 h−1 at 80 °C, albeit at the expense of catalyst stability. An analogous complex with a cyclometalated N-methyl group and a κ2-formate ligand (26, Figure 12a) constitutes the only example of homogeneous molybdenum-catalyzed CO2 hydrogenation in basic media.154 Treatment of 26 with 50 equiv of LiOTf in a DBU/1,4-dioxane solution under 69 bar H2/CO2 at 100 °C yielded a maximum TON of 35. In addition to the iron pincer catalyst discussed above,147 Gonsalvi and co-workers studied the tetradentate 1,1,4,7,10,10hexaphenyl-1,4,7,10-tetraphosphadecane phosphine (P4) both in situ with Fe(BF4)2·6H2O and integrated into well-deﬁned iron structures for bicarbonate hydrogenation (Figure 13).155 Ligand P4 exists commercially as a mixture of meso and rac diastereoisomers (meso/rac = 3), which can lead to various

To obtain high activities in CO2 hydrogenation to formate, Linehan and co-workers used a rational design approach, based on the assumption that the reaction mechanism involved three essential steps: (1) hydride transfer from the metal to the substrate releasing formate, (2) coordination of a hydrogen molecule, and (3) deprotonation of a metal dihydride to regenerate the starting complex.148 They therefore searched for a catalyst featuring a weak metal−hydride bond that would facilitate formate formation through step 1 and a basic additive that would be suﬃciently strong to promote the deprotonation in step 3. Indeed, using [Co(DMPE)2H] [DMPE = 1,2bis(dimethylphosphino)ethane] in the presence of Verkade’s base (pKa = 33.6) aﬀorded high reaction rates of up to 6400 and 74 000 h−1 under 1.8 and 20 bar of H2/CO2, respectively, in THF at 21 °C. Evidence of the underlying reaction mechanism was provided in a later study.149 In the presence of Verkade’s base, the reaction rate was found to be independent of both the base concentration and the hydrogen pressure and to increase linearly with the CO2 gas pressure, in agreement with step 1 being rate-determining. In addition, starting from any of the three proposed intermediates (steps 1− 3) aﬀorded similar TOFs. The pKa value of the base dictated the rate-limiting step, which became the deprotonation of the proposed metal dihydride in the case of DBU. As a result, experimentally, the reaction showed a ﬁrst-order dependence on the base concentration. The authors also observed that, at CO2/H2 ratios above 1, carbon monoxide formation occurred, presumably through the reverse water−gas shift reaction. The generated water, in turn, yielded bicarbonate that competed with the production of formate. Under these conditions, the catalytically inactive [(Co(DMPE)2)2(μ-DMPE)]2+ and [Co(DMPE)2CO]+ species formed and were identiﬁed by NMR spectroscopy. In a collaborative eﬀort, the groups of Muckerman, Himeda, and Fujita synthesized a series of cobalt(III) complexes bearing the proton-responsive dihydroxybipyridine ligand.150 Reactions were carried out in a sodium bicarbonate solution under an equimolar H2/CO2 pressure, resulting in TOFs of up to 39 h−1 at 100 °C when [Cp*Co(4,4′-DHBP)(H2O)](PF6)2 was used under 40 bar pressure. A series of iron PNP pincer complexes that had been previously shown to exhibit remarkable activities in FA dehydrogenation151 were further evaluated by the groups of Hazari and Bernskoetter for the reverse CO2 hydrogenation 382

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

obtained. Similar iron-based catalysts were either less active or inactive for bicarbonate hydrogenation. When CO2 was employed as the substrate in the presence of NaOH in the same solvent, only traces of formate were formed. The authors reasoned that the diﬀerence in pH under these conditions (7.45 compared to 8.80 in the case of bicarbonate) inhibited the formation of an active iron hydride intermediate (Knölker complex) through the Hieber base reaction. Following adaptation of the reaction protocol (a NaOH solution was ﬁrst pressurized with CO2 to favor formation of bicarbonate, after which the catalyst was introduced), carbon dioxide could be hydrogenated in the presence of sodium formate with a TON of 307 (120 °C, 30 bar H2, 20 bar CO2). The groups of Enthaler and Junge evaluated the [(PCP1)Ni(H)] catalyst (30, Figure 15) for the hydrogenation of

Figure 13. Structures of the P4 isomers and P4-based iron catalysts tested for bicarbonate hydrogenation.155 Figure 15. Structure of the nickel pincer catalyst that is active in bicarbonate hydrogenation as reported by the groups of Enthaler and Junge.157

octahedral catalyst conﬁgurations in solution, potentially exhibiting diﬀerent activities. When the commercial P4 was used in situ together with Fe(BF4)2·6H2O in methanol under 60 bar H2 at 80 °C, 15% of the sodium bicarbonate (TON = 154) was converted into formate after 24 h. Interestingly, the well-deﬁned iron complexes synthesized from the reaction between the iron precursor and either the meso or rac isomer exhibited quite diﬀerent activities: Under the aforementioned reaction conditions in the presence of PC, [Fe(rac-P4)BF4]BF4 (27, Figure 13) and its meso-P4 analogue aﬀorded 58% (TON = 575) and 6% (TON = 62) conversions, respectively. An increase in the substrate-to-catalyst ratio from 1000 to 3000 in the presence of 27 resulted in a minor increase in the turnover number to 723, albeit at the expense of the formate yield, which dropped to 24%. Catalyst 28 (Figure 13) led to an improved bicarbonate conversion of 76% in the absence of PC, whereas the TON peaked at a value of 1229 at a higher substrate-tocatalyst ratio of 10 000, corresponding to 12% bicarbonate conversion. High-pressure NMR studies pointed toward [Fe(rac-P4)H]+ as a key intermediate when starting with either 27 or 28, and an inner-sphere reaction mechanism necessitating two available cis positions was proposed to be responsible for the better activity of the Fe(rac-P4) catalysts. In the approach of Zhou and co-workers, the phosphine-free, air- and moisture-stable iron catalyst 29 (Figure 14) was found to be active for bicarbonate and CO2 hydrogenation in basic media.156 In the former case, the highest activity was reached in ethanol/water under 30 bar H2 pressure at 120 °C, giving a TON of 447 after 24 h (yield = 44.7%). Remarkably, even at a H2 pressure as low as 5 bar, 163 turnovers could still be

CO2.157 Although CO2 could not be reduced, bicarbonate was converted to the corresponding formate salt under 55 bar H2 pressure in methanol at 150 °C, with a TON of over 3000 (TOF = 150 h−1). In their work on amine-promoted CO2 capture and in situ hydrogenation to formate, the groups of Olah and Prakash studied the iron pincer catalyst 31 (Figure 16),108 previously

Figure 16. Structure of iron pincer catalyst 31 reported by the groups of Olah and Prakash for CO2 capture and hydrogenation.108

identiﬁed as an active catalyst in methanol dehydrogenation by Beller and co-workers.158 Hydrogenation of PEHA-absorbed CO2 (i.e., bicarbonate/carbonate and carbamate) gave a formate yield of 53% (TON = 255, TOF = 25 h−1) in 1,4dioxane/water under 80 bar H 2 at 50 °C. Recycling experiments in a biphasic medium consisting of 2-methyltetrahydrofuran/31 and bicarbonate/carbonate salts of DABCO in water aﬀorded TOFs of 115−145 h−1 (4−5 times lower than for the ruthenium catalyst; see section 2.2.1), improved yields of 95−96%, and a total TON of 3580 (ﬁve cycles). The metal content in the aqueous phase was found to be less than 2 ppm, and as for the ruthenium complex, the formato species [(PNP2)FeH(OOCH)(CO)] was identiﬁed as the catalyst resting state. In 2015, the groups of Watari and Ikariya reported copper catalysts that are active in CO2 hydrogenation in the presence of a suitable organic base.159 Under an equimolar H2/CO2 pressure of 40 bar in a DBU/1,4-dioxane solution, Cu(OAc)2· H2O (DBU/Cu = 500) aﬀorded TONs of up to 167 after 116 h

Figure 14. Structure of iron catalyst 29, which is active in bicarbonate hydrogenation as reported by Zhou and co-workers.156 383

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Table 2. Overview of Selected Catalytic Systems That Are Active in CO2/Bicarbonate Hydrogenation catalyst precursora

solvent

[RuH2(PPh3)4]74 [cis-Ru(DPPM)2Cl2]80 [RuH2(PMe3)4]81 [RuCl2(PMe3)4]81 [RuH2(PMe3)4]82

C6H6/H2O MIBC/H2O scCO2 scCO2 scCO2

[RuCl(OAc)(PMe3)4]84 [RuCl(OAc)(PMe3)4]85 [Ru(Cl2BPY)2(H2O)2](CF3SO3)286 [TpRuH(PPh3)(CH3CN)] (1)88 RuCl3·xH2O/PPh389 [Ru2(μ-CO)(CO)4(μ-DPPM)2]90 [RuCl2(mTPPMS)2]2/mTPPMS97 RuCl3·xH2O/DPPBTS105 [(C6Me6)Ru(DHPHEN)Cl]Cl109 [(PNP3)Ru(H)Cl(CO)]e (10)120 [(η6-p-cymene)Ru(η2-HN,H,HO)(H2O)]BF4e 123 [RuCl2(C6H6)]2/DPPM125 [(η6-p-cymene)Ru(bis-NHC1)Cl]PF6e (13)129 [(PNP4)RuH(Cl)(CO)]e (15)133 [(PNP4)RuH(H-BH3)(CO)]e (16)139 [(PNN1)RuH(CO)]f 171 [(RuCl2(benzene))2]/DPPM172 [Cp*Ir(DHPHEN)Cl]Cl (2)109 [Cp*Ir(4,4′-DHBP)Cl]Cl (3)111 [Cp*Ir(DHPHEN)Cl]Cl (2)111 [(Cp*Ir)2(THBPM)(H2O)2](SO4)2 (4)113 [Cp*Ir(THBPM)(H2O)]SO4 (5)114 [(PNP1)IrH3]e (8)118 [(PNP2)IrH2(OOCH)]e (9)122 [Ir(bis-NHC2)(AcO)I2]e (14)130 [RhCl(TPPTS)3]76 [Rh(hfacac)(dcpb)]77 [RhCl(PPh3)3]/PPh391 RhCl3·3H2O/CyPPhz2135 [Cp*Rh(NHC)Cl]Nae (18)141 K[Ru(EDTA-H)Cl]·2H2O162 [(η6-C6Me6)Ru(4,4′-DMBPY)(OH2)]SO4163 [RuCl2(PTA)4]e 164 [Cp*Ir(L3)Cl]Cle (32)165 [Rh(NBD)(PMe2Ph)3]BF4166 [Ru(η5-C5H4(CH2)3NMe2)(DPPM)]BF4167 [Rh(COD)Cl]2/DPPB56 [Ru(acriphos)(PPh3)(Cl)(PhCO2)] (33)168 Fe(BF4)2·6H2O/PP3143 [(PNP3)Fe(H2)(CO)]e (21)146 [(PNNNP1)Fe(H)Br(CO)]e (23)147 [(PNP5)Fe(H)(OOCH)(CO)]e (25)152 [Fe(rac-P4)(CH3CN)2](BF4)2e (28)155 [Fe] complexe (29)156 [Ni(DPPE)2]74 [NiCl2(DCPE)]142 [(PCP1)Ni(H)]e (30)157 Cu(OAc)2·H2O159 [Cu(triphos)(MeCN)]+ 160 Co(BF4)2·6H2O/PP3144 [Co(DMPE)2H]148 [Cp*Co(4,4′-DHBP)(H2O)](PF6)2150 [(PNP5)Co(CO)2]Cle 153 [(PMeNP4)Mo(C2H4)(OOCH)]e (26)154

scCO2 NEt3 EtOH CF3CH2OH EtOH/H2O Me2CO H2O H2O H2O DMF H2O H2O/THF H2O MeOH/H2O H2O/THF diglyme DMF H2O H2O H2O H2O H2O H2O/THF H2O H2O H2O DMSO MeOH/DMSO MeOH H2O H2O H2O H2O H2O THF THF DMSO DMSO/H2O MeOH H2O/THF EtOH THF MeOH EtOH/H2O C6H6 DMSO MeOH 1,4-dioxane CH3CN MeOH THF H2O CH3CN 1,4-dioxane

additive NEt3 Aminosol CST 115 NEt3/H2O NEt3/H2O NEt3/DMSO (or MeOH) NEt3/C6F5OH NEt3 NEt3 NEt3 NEt3 NEt3 NaHCO3 NHMe2 KOH DBU NEt3 NaHCO3 KOH KOH/KHCO3 Na2CO3 K2CO3 NEt3 KOH KOH KOH KHCO3 KHCO3 KOH KOH KOH NHMe2 NEt3 NEt3 PEI600 KHCO3 − − − − − − − − NaHCO3 NaOH DBU DBU/LiOTf NaHCO3 NaHCO3 NEt3 DBU NaHCO3 DBU DBU NaHCO3 Verkade’s base NaHCO3 DBU/LiOTf DBU/LiOTf 384

Tb (°C)

timec (h)

25/25 30/60 120/85 120/85 120/85

RT 70 50 50 50

20 NA 1 1 0.5

87 18 000 3700 7200 2000

4 41 000 1400 1040 4000

120/70 60/40 30/30 25/25 30/30 35/35 35/60 25/25 30/30 10/30 50/50 0/50 20/20 − 12/38 10/30 30/30 30/30 30/30 30/30 25/25 0.5/0.5 30/30 28/28 30/30 20/20 20/20 40/20 40/40 0/50 17/3 25/55 50/50 25/25 48/48 40/40 20/20 40/40 0/60 3.3/6.7 40/40 35/35 0/60 0/30 25/25 160/40 0/55 20/20 20/20 0/60 10/10 20/20 35/35 35/35

50 50 150 100 60 RT 80 60 120 120 100 70 200 150 79 200 100+RT 120 120 120 80 25 120 185 200 RT RT RT 60 100 40 40 60 80 40 80 RT 60 80 80 80 80 80 120 RT 50 150 100 140 120 21 100 45 100

0.3 0.5 8 16 5 21 NA 15 24 0.1 10 2 75 36 1 48 4 1 NA NA 1 1 48 2 75 12 NA 20 32 72 NA 55 NA 0.08 48 16 6.5 16 20 5 21 1 24 24 20 216 20 116 2 20 NA 1 1 16

31 200 NA 5000 1815 200 2160 NA 4980 15 400 200 000 400 1400 23 000 18 420 2200 23 000 6400 21 000 190 000 222 000 79 000 193 3 500 000 37 300 190 000 3400 NA 2500 852 3600 NA 55 159 1100 130 8 2000 h−1) at pH 2.8 (5 M Fa/ NaOOCH) at room temperature.116 Himeda and co-workers were investigating the transfer hydrogenation of ketones with FA as the hydrogen donor already in 2003;183 however, FA dehydrogenation in the framework of chemical H2 ﬁxation was only reported in 2009.206 Subsequently, a plethora of water-soluble metal 387

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 20. Structures of iridium complexes tested in aqueous formic acid dehydrogenation.206,211−215

Activities of up to 12 000 h−1 were reached during the ﬁrst 10 min of reaction at 90 °C with an optimum NaOOCH/ HCOOH ratio of 5. In the absence of base, the catalyst remained active, albeit with a lower initial TOF of 3750 h−1. Carbon monoxide-free hydrogen could also be generated under pressurized conditions (>260 bar) at a reduced rate compared to that observed under atmospheric conditions, and in an independent experiment, the catalyst reached a TON of 350 000 over 35 h. Using a similar approach, the well-studied [Cp*Ir(4,4′-DHBP)(H2O)]SO4 catalyst (36) was employed in selective FA dehydrogenation under evolved H2/CO2 pressures of up to 1200 bar at 80 °C.216 Under the speciﬁc reaction conditions employed, the generated gas mixture existed in the supercritical phase and, therefore, allowed for H2 puriﬁcation through a simple reduction of the temperature. Cooling to −51 °C resulted in the separation of two phases, a CO2-rich liquid and an 85 mol % pure H2 gas phase. The authors concluded that H2 release from FA in this manner could suﬃciently feed a fuel-cell vehicle that typically stores the high-purity carrier at 700 bar, thus obviating the need for a costly high-pressure pump. On the basis of their previous experience in transfer hydrogenation,217−219 Wills and co-workers also investigated FA dehydrogenation in the presence of triethylamine. Out of a number of ruthenium catalysts tested, RuCl2(DMSO)4, RuCl2(NH3)6 and anhydrous RuCl3 exhibited the highest and essentially the same activity with TOFs of up to 18 000 h−1 after four cycles at 120 °C.220 Because of the observed increase in activity in the ﬁrst three cycles, the authors speculated that the common binuclear species Ru2(HCO2)2(CO)4 was formed during the activation process at temperatures of >100 °C. A drawback of the elevated reaction temperature was the occurrence of FA decarbonylation resulting in CO formation (200 ppm), which would prohibit long-term coupling to a fuel cell. To maintain a constant catalytic activity, a continuous hydrogen generation setup was established wherein triethylamine was replaced with the higher-boiling-point dimethyloctylamine.221 The rate of FA dosage was adjusted to achieve a stable (maximum) hydrogen output, based on a temperature or impedance feedback mechanism. Similarly to the previously reported batch experiments, the production of carbon monoxide also occurred during the continuous process.

Figure 21. Pendent-base eﬀect in hydrogen generation from an iridium complex bearing an ortho-substituted bipyridine ligand. (Adapted with permission from ref 209. Copyright 2014 John Wiley & Sons.)

iridium catalysts bearing functionalized pyrimidyl imidazoline ligands, which, upon optimization, displayed activities of up to 322 000 h−1 at 100 °C in the presence of formate (38, Figure 20c).212 Moreover, the potential of a practical FA dehydrogenation setup was demonstrated through the evolution of 1.02 m3 of H2 gas (pressurized, Pmax = 10 bar), corresponding to a TON of 2 000 000, over a period of 363 h, with the optimized iridium complex 39 featuring a nonfunctionalized pyridyl imidazoline ligand (Figure 20d).213 In this case, the catalyst was highly stable in an acidic environment (pH 1.7) with an average TOF of 6250 h−1 without the need for base-assisted activation. In a further advancement toward a real-life application where the catalyst is readily removable from the reaction solution and reactants/products, the groups of Kawanami and Himeda employed the homogeneous iridium complex 40 featuring a 1,10-phenanthroline-4,7-diol ligand (Figure 20e) in a “heterogeneous mode”,214 similar to a previous report on bicarbonate hydrogenation.110 Owing to its pH-dependent water solubility, the catalyst precipitated from the aqueous medium following complete FA consumption (6.5 M) under high pressure (≤250 bar) at 50 °C. In this case, the pH shifted from an initial value of 0.9 to 1.9 at the end of the reaction. The precipitated complex was then recovered by ﬁltration (loss of ≤6 mol %) and recycled more than 10 times over a period of 200 h at high pressure without a decrease in activity. In an independent experiment performed at 60 °C, the catalyst retained its stability and activity for as long as 3.5 months, yielding 100% FA decomposition (10 M) into CO-free hydrogen, or a TON of 5 000 000 (TOF = 3010 h−1). In a continuous quest for more eﬃcient catalysts, the groups of Huang and Himeda recently developed ruthenium complex 41 bearing (η6)p-cymene, chloride, and 2,2-biimidazoline ligands (Figure 20f).215 388

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 22. Ionic liquids reported as cocatalysts for the dehydrogenation of formic acid.224−227

the IL and additional stabilization through Coulombic FA− solvent interactions, the equilibrium was largely shifted toward FA. As a result, FA could be formed at mild temperatures (≥40 °C) and pressures (3−23 bar) without necessitating a base, albeit inevitably constraining the H2 release capacity of the system under isochoric conditions. Another ruthenium precursor, namely, RuCl3, and several IL cocatalysts were tested for H2 production from FA/NaOOCH by Wasserscheid and co-workers.227 Because of its low viscosity, its thermal stability, and its broad liquid range, EMim was chosen as the starting point of the investigations and screened with several anions. Signiﬁcant CO quantities were detected in the cases of [PF6], [N(CN)2], [EtSO4], [OctSO4], [MePO 2 (OMe)], and [HPO 2 (OMe)], whereas selective H 2 formation occurred with either [NTf2] or [OAc] counterions. The more basic acetate anion was further evaluated with 1ethyl-2,3-dimethylimidazolium (EMMim) as the cation because of suspected in situ N-heterocyclic carbene formation in the presence of 1,3-dialkylimidazolium salts.228 Indeed, the best catalytic activities were obtained with [EMMim][OAc] (see Figure 22), namely, 91% FA conversion after 84 h at 80 °C compared to 59% attained with [EMim][OAc]. An induction period was observed during the ﬁrst cycle, whereas subsequent recycles yielded higher and stable TOFs of 150 h−1 at 80 °C. Under these conditions, complete FA conversion was achieved within 20 h. In agreement with Laurenczy and co-workers,191 the authors proposed that the activation period was related to the reduction of Ru(III) to Ru(II) active species, which was conﬁrmed by further experiments with a Ru(II) precursor. The signiﬁcant deactivation observed upon recycling RuCl3/ [EMim][NTf2] but not the respective [EMim][OAc]-based system was explained by the formation of metal nanoparticles in the former case that were identiﬁed by transmission electron microscopy. Further evidence was provided by independently synthesized Ru nanoparticles which resulted in very sluggish FA dehydrogenation. In 2011, the groups of Prakash and Olah reinvestigated H2 evolution from FA/NaOOCH mixtures in water with RuCl3 as the catalyst precursor without any ligand at 109 °C.229 Under these conditions, FA decarboxylation and decarbonylation occurred (0.21% CO). Metal nanoparticles were identiﬁed in solution during the second run when Ru concentrations of over 12 mM were utilized, which coincided with the formation of 0.3% methane, likely by CO2 hydrogenation. An independent experiment in the presence of preformed Ru nanoparticles revealed that the latter were much less active in FA dehydrogenation than the ruthenium trichloride salt. Without sodium formate, a signiﬁcantly lower reaction rate was obtained, in agreement with Laurenczy and co-workers’ previous ﬁndings,187 whereas subsequent base addition allowed the initial activity to be recovered and for the isolation and characterization of a tetranuclear [Ru4(CO)12H4] species. This complex was an active catalyst for FA/NaOOCH dehydrogenation in DMF but not in water because of insolubility. The authors also reported that, under atmospheric pressure, H2

The application of ionic liquids as solvents and cocatalysts in transfer hydrogenation using formic acid/triethylamine mixtures as hydrogen donors was already pursued in the mid-2000s by Geldbach and Dyson222 and Ohta and co-workers.223 However, the groups of Shi and Deng ﬁrst reported the use of an amine-functionalized ionic liquid (see Figure 22) as a base substitute for FA dehydrogenation with the commercially available [RuCl2(p-cymene)]2 catalyst precursor.224 Through the use of an IL, the contamination of gaseous products associated with the discharge of volatile solvents/additives, leading to deactivation, could be avoided. Even though low activity was obtained in the presence of sodium formate, i Pr2NEt, and the ILs Et2NEMimCl (EMim = 1-ethyl-3methylimidazolium), i Pr 2 NEMimBF 4 , i Pr 2 NEMimOTf, i Pr2NEMimNTf2, and iPr2NEMimCl at 60 °C, a TON of 240 was reached with a 1:1 ratio of NaOOCH to iPr2NEt2 in BMimCl (BMim = 1-butyl-3-methylimidazolium) after 2 h. Furthermore, combining sodium formate and iPr2NEMimCl, doubling the amount of the IL to 10 mmol, and reducing the catalyst loading yielded a maximum TON of 1267 in the same time. The reaction rate increased with increasing formate amount up to 7 mmol and subsequently stabilized, which was attributed to the high viscosity and steady ionic polarity of the medium. A water content of up to 12% (w/w) was well tolerated by the system, with higher concentrations causing a drop in catalytic activity. Major limitations of the system were related to incomplete FA conversion (at the end of the reaction, an FA amount equal to the amount of the IL remained), as well as catalyst deactivation already after the ﬁrst run. In a subsequent study, the group of Dupont employed the same ruthenium catalyst ([RuCl2(p-cymene)]2) and a similar IL, Et2NEMimCl (Figure 22), for FA dehydrogenation.225 At ambient temperature, the reaction did not take place, possibly because the active catalytic species was not formed. Nevertheless, at 80 °C, over 90% of the FA was dehydrogenated within 7 h, with an initial TOF of 1540 h−1 (calculated for 20% FA conversion). The authors reported a detrimental eﬀect of water and phosphine ligands on the catalysis, likely because of competing coordination to the metal and/or solvation processes (for water). The IL concentration also aﬀected the reaction rate, with high amounts yielding improved initial TOFs but lower overall TOFs. This was attributed to high viscosity and tentative blockage of important ruthenium coordination sites by the IL amine functionalities. In contrast to Deng and co-workers’ catalyst,224 Dupont’s system was stable and could be recycled with only a minor decrease in activity. Under the reaction conditions used, traces of carbon monoxide ( 220 000 or 129 L of H2. Finally, in a proof-of-principle experiment, this setup was coupled to a PEM demonstration kit fuel cell and successfully reached a constant power output of over 10 h. Recently, a number of metal compounds bearing pincer-type ligands was demonstrated to exhibit high activities in the selective FA dehydrogenation reaction (see refs 120, 139, 151, 157, and 243−247). Following the ﬁrst examples by Nozaki and co-workers121 and Milstein and co-workers (see section 2.3.2),243,244 Pidko and co-workers investigated FA dehydrogenation in the framework of reversible H2 ﬁxation.120 Among a number of ruthenium PNP-type pincer catalysts tested, the 394

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

on a metal−ligand cooperativity similar to that previously reported by Xiao and co-workers,231 albeit in the absence of basic promoters. Initial tests based on a continuous FA dosage for 10 min at 35 °C revealed that the amidoiridium complex 52 (Figure 32) catalyzed selective H2 generation with a TON of

complex shown in Figure 3c, bearing hydrido, carbonyl, and chloro ligands, exhibited the best performance in the presence of a base promoter in DMF. A TOF of 256 000 h−1 and more than 706 500 turnovers were reached at 90 °C, without CO contamination, when trihexylamine was employed. The authors demonstrated that with the less-nucleophilic DBU base, the signiﬁcant activity ﬂuctuations linked to the reaction temperature (65−90 °C) could be used to control the reaction rate. In this case, the reaction was ﬁrst-order with respect to the FA concentration. However, in the presence of triethylamine instead of DBU, an enhancement of the dehydrogenation rate was reported upon FA consumption. Because of these observations, it was concluded that the rate-limiting step diﬀered depending on the nature of the basic additive: With the weaker NEt3, the formation of H2 from the combination of a metal hydride and the protonated base was favored compared to CH bond cleavage, making the latter rate-determining. However, in the presence of the stronger DBU, the rate was suggested to be dependent on the H2 formation step. The groups of Zheng and Huang tested three ruthenium PNNNP2 pincer catalysts with a dearomatized pyridine group and an imine arm for H2 evolution from FA.247 Complex 51 (Figure 31) exhibited the best activity in DMSO, namely, a

Figure 32. Structures of complexes 52 and 53 and proposed waterstabilized transition state for H2 formation (light gray). (Adapted with permission from ref 232. Copyright 2015 John Wiley & Sons.)

255 after 1 h in 1,2-dimethoxyethane (DME), whereas in a 50:50 (v./v) H2O/DME solution, the activity was 1 order of magnitude higher, yielding an initial TOF of 4990 h−1 and a TON of 1910 after 1 h. In pure water, the catalyst performance deteriorated, possibly because of lower catalyst solubility. Replacing either the substituents on the sulfonyl group with electron-releasing donors or the NH moiety with the dimethyl amine analogue was accompanied by almost complete catalyst deactivation. Addition of 5.3 equiv of FA to 52 at −80 °C yielded an iridium formato intermediate with a protonated amido functionality, whereas reaction with excess FA at temperatures above −30 °C generated the hydrido amine complex 53 through release of CO2 (Figure 32), with a characteristic resonance at −10.64 ppm in the 1H NMR spectrum. The latter aﬀorded initial TOF and TON values (6090 h−1 and 2340, respectively) similar to those of 52 under identical reaction conditions. The authors concluded that hydrogen bonding of water onto the acidic amine proton stabilizes a six-membered transition state and therefore accelerates the rate-limiting H2 formation step, as part of a proton-relay mechanism. The beneﬁcial role of electronwithdrawing sulfonato substituents was rationalized in terms of the resulting acidity enhancement of the NH proton. The two iridium catalysts [Cp*Ir(L41/L42)Cl]Cl (54/55, Figure 33), bearing either a 2,2′-bi-2-imidazoline (L41) or a

Figure 31. Proposed mechanism for the dehydrogenation of formic acid with ruthenium pincer catalyst 51 [P = P(tBu)2]. (Adapted with permission from ref 247. Copyright 2016 John Wiley & Sons.)

TON and TOF of 95 000 and 2380 h−1 respectively, without a basic additive at 50 °C. Conveniently, exactly the same activity was obtained after the catalyst had been aged for 1 month under aerobic conditions. The reaction was dependent on the solvent, with signiﬁcantly lower catalyst stabilities exhibited in toluene, acetonitrile, DMF, and THF. Introduction of triethylamine and heating to 90 °C enhanced the reaction rate by over 3-fold (TOF = 7333 h−1), with a prolonged stability of 1 100 000 turnovers. The use of diﬀerent bases or higher FA concentrations reduced the catalyst lifetime due to hydrolysis of the NP bonds. Following the evolution of the reaction by NMR spectroscopy and solution color changes under various conditions, the authors observed the (de)protonation of the imine group and (de)aromatization of the pyridine ring, which allowed them to suggest a mechanism for FA dehydrogenation (Figure 31). The ﬁrst step consists of NH protonation coupled to pyridine rearomatization and formate coordination onto the metal center, followed by CO2 release to yield a dihydride species (observed only upon pressurization of the catalyst in DMSO with H2) and eventually elimination of dihydrogen to close the cycle. The latter step is accompanied by regeneration of the starting dearomatized pincer catalyst. Matsunami, Kayaki, and Ikariya investigated FA dehydrogenation with half-sandwich iridium(III) complexes bearing noninnocent ligands with a crucial NH functionality,232 relying

Figure 33. Structures of iridium catalysts 54 and 55 that were tested in the dehydrogenation of FA.248

2,2′-bi-1,4,5,6-tetrahydropyrimidine (L42) ligand and exhibiting an unprecedented activity for FA dehydrogenation without added base, were investigated by Li and co-workers in 2015.248 At 90 °C, initial TOFs of 487 500 and 375 000 h−1 and FA conversions of 94% and 87% were reached with 54 and 55, respectively. The reaction rate was determined to be ﬁrst-order in complex 54, whereas a positive fractional order of 0.41 was found with respect to FA, pointing toward a mononuclear 395

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

active catalyst species and an equilibrium between Ir + HCOO− and an IrOOCH intermediate. Addition of variable amounts of sodium formate to the system revealed a maximum TOF of 72 500 h−1 at pH 2.8. Considering that the pKa1 value of catalyst 55 is approximately 9.0, the authors excluded a connection between the (de)protonation of the ligand and the observed activity. The decreasing TOFs between pH values of 2.8−8.0 and those lower than 2.8 suggested that both HCOO− and H3O+ moieties were relevant to catalysis, whereas through a series of KIE experiments, the generation of IrH was found to be turnover-limiting. The catalyst generated in situ from [Cp*IrCl2]2 and L42 was additionally evaluated for its longterm stability, aﬀording a total TON of 2 400 000 within 14 h, without inhibition at elevated H2/CO2 pressure. During the past two years, the group of Joó has reported iridium-based catalysts containing the water-solubilizing mTPPMS ligand (Figure 34).249,250 Unprecedented activities

in the pH of the solution. Finally, complex 57 was stable throughout ﬁve subsequent runs, with an independent experiment performed at 115 °C over 40 h yielding a TON of 674 000. A novel approach regarding homogeneous FA/formate dehydrogenation was introduced in 2016 by Williams and coworkers.251 Despite being soluble in FA, the homogeneous iridium-NP1 [NP1 = 2-(di-tert-butylphosphinomethyl)pyridine] complex 58 (Figure 35) precipitated following complete substrate consumption, which allowed for its straightforward recovery at the end of the reaction in a manner similar to that use for heterogeneous catalysts. A basic cocatalyst was necessary even though its nature was not of great importance because similar activities were obtained with 5 mol % NaOOCH, KOOCH, KOH, NaOH, LiOH, or nBu4NOH or 2.5 mol % Na2CO3 or K2CO3. The catalyst could be recycled up to 50 times under aerobic conditions, yielding over 66 000 turnovers and 89% FA conversion at 90 °C without loss of activity. However, after a longer period of 4 months and a TON of 2 160 000, the catalyst retained just 15% of its initial activity despite protection against oxygen. Carbon monoxide formation, which occurred in neat FA likely as a result of its thermal decomposition, was suppressed in the presence of 10 vol % water or at a higher sodium formate concentration (50 mol % with respect to FA). In solution in the presence of FA or H2, complex 58 dimerized, releasing CO2 and cyclooctene and thus yielding a species with two bridging hydrides. On the other hand, in a buﬀered FA solution, a formate-bridged dimer was generated. The reaction was found to be ﬁrst-, one-half-, and inverse-order in iridium, base, and FA, respectively, suggesting stability of the formed dimeric iridium structure, activation of two metal centers by 1 equiv of base, and ﬁnally inhibition by FA, for which, however, no clear explanation was presented. Kühn and co-workers introduced a rhodium complex bearing a bis-N-heterocyclic carbene that catalyzed selective FA dehydrogenation in aqueous solution and in the absence of additives (Figure 10a).141 Even though the related iridium and ruthenium complexes are only moderately active at 80 °C, catalyst 18 exhibited a TOF of 90 h−1 already at ambient temperature (25 °C), one of the few examples where a rhodium catalyst was superior. Optimization of the solution pH and a gradual decrease of the catalyst loading led to a maximum TOF of 39 200 h−1 at 100 °C. Under these conditions, the catalyst was stable over 50 h, yielding 449 000 turnovers. To further demonstrate the catalytic stability, the rhodium complex was tested in continuous FA dehydrogenation at 80 °C over a period of 72 h, resulting in a constant TOF of 4300 h−1. 2.3.2. Non-Noble-Metal-Based Catalysts. In 2010, the ﬁrst iron-based catalyst operating with a combination of nitrogen and phosphine ligands was presented for FA dehydrogenation by Beller and co-workers.252 From initial screening studies, [Fe3(CO)12] in the presence of PPh3 and 2,2′:6′,2″-terpyridine (TPY) (Fe/PPh3/TPY = 1:1:1) emerged

Figure 34. Structure of mTPPMS-containing iridium catalysts reported by Joó and co-workers.249,250

in sodium formate dehydrogenation were exhibited with an iridium(I)-N-heterocyclic carbene phosphine complex that also catalyzes the reverse bicarbonate hydrogenation reaction. The limited solubility of 56 in the basic solution was improved by addition of 2 equiv of excess mTPPMS. Under these conditions, a 5 M NaOOCH solution was dehydrogenated at a rate of 15 110 h−1 at 80 °C. In addition, when the reaction solution containing 56 was stored under 100 bar H2 pressure for a period of 71 days, hydrogenation/dehydrogenation proﬁles identical to those of the freshly prepared mixtures were obtained. On the other hand, the iridium hydride 57 was evaluated for FA dehydrogenation in the presence of formate and yielded among the highest activities to date.250 Complete FA conversion and initial TOFs of over 17 800 h−1 (at 70 °C) were reached for a 0.72 M FA solution containing approximately 7 equiv of NaOOCH, with the latter not being converted. The reaction rate increased linearly with the iridium concentration, decreased upon addition of mTPPMS, and plateaued at FA concentrations above 0.5 M. Furthermore, the TOF peaked at a pH value of 3.75, which corresponds to the pKa of formic acid. Under isochoric conditions (Pfin = 70.9 bar), the maximum rate for H2 evolution was noted as 298 000 h−1 at 100 °C. At a constant sodium formate concentration, higher HCOOH-to-catalyst ratios led to a linear increase in ﬁnal pressures but to lower reaction rates, likely as a result of a shift

Figure 35. Activation of iridium catalyst 58, developed by Williams and co-workers.251 396

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

contrast to other systems,200,252 no enhancement of activity was observed upon irradiation with visible light.256 Hydrogen evolution was inhibited upon addition of stoichiometric amounts of amine or strong acids and in pure FA (i.e., in the absence of solvent), with the latter related to the inability of the ligand to coordinate to the iron center under the reaction conditions. Similarly, several phosphorus- (mono-, bi-, and polydentate), nitrogen-, and sulfur-containing donors yielded either low or no catalytic activity. More recently, air-stable in situ generated iron(II) catalysts were adapted for use in aqueous solution by sulfonation of the multidentate PP3 ligand (sPP3), resulting in selective FA dehydrogenation with TOFs of up to 240 h−1 at 80 °C (FeCl2/sPP3 = 1:2).257 The reaction was not inhibited by either H2 or CO2 gas pressure, and the catalyst was recycled several times under both isochoric and atmospheric conditions without loss of activity. This constitutes the ﬁrst example of an iron catalyst that operates in aqueous media without additives. Milstein and co-workers published the ﬁrst example of selective FA dehydrogenation with the iron PNP3 pincer complex 59 in the presence of NEt3 (Figure 36a).243 Among

as the most active for the dehydrogenation of FA/NEt3 (5:2) adducts in DMF at temperatures above 100 °C. Interestingly, it was noted that, under visible light irradiation (300 W xenon lamp, UV and IR regions of the spectrum were ﬁltered out) the in situ generated catalyst became active even at ambient temperature. Further tests under the aforementioned conditions showed that several non-noble-metal carbonyls [Mo(CO)6, Mn2(CO)10, Cr(CO)6, Co2(CO)8] could generate active catalysts, although the performance of Fe3(CO)10 was clearly superior. Visible light was necessary both for the formation of the active iron species [FeH(CO)3(PPh3)]− and for the progression of the catalytic cycle; that is, halting the irradiation led to inhibition of H2/CO2 production. The inexpensive and commercially available PPh3 ligand gave the best results among various phosphine ligands tested. In regard to the nitrogen donors were concerned, improved TONs (2 h) of approximately 50 were attained with either 6,6″-(bromo)2,2′:6′,2″-terpyridine or 6,6″-(phenyl)-2,2′:6′,2″-terpyridine (PhTPY) in place of TPY (TON = 31). Despite its poor stability, the catalyst generated from [Fe3(CO)12]/PPh3/ PhTPY yielded a maximum initial TOF of 200 h−1 at 60 °C. Based on spectroscopic studies and DFT investigations, the authors proposed that the main role of the N-ligand was the temporary stabilization of the metal center, that is, the delay of CO dissociation from the active species, which would inevitably cause deactivation. A more stable analogue was later reported to be formed by substituting the PPh 3 ligand with benzylphosphine;253 as a result, a TON of 1266 was achieved in the presence of [Fe3(CO)12]/benzylphosphine/TPY within 51 h under otherwise-identical reaction conditions. The authors reported a potential connection between the improved stability and the formation of a ﬁve-membered metallacycle through ortho metalation. In collaboration with the group of Laurenczy, the group of Beller developed a highly active catalyst comprising Fe(BF4)2· 6H2 O and PP 3 in environmentally friendly propylene carbonate, yielding TOFs of up to 9425 h−1 and TONs of over 92 000 at 80 °C.254 A reaction mechanism was proposed that rationalized the inhibiting eﬀect of H2 pressure on the catalytic activity. Well-deﬁned platinum catalysts that incorporated the PP3 ligand and were active in FA dehydrogenation had been reported earlier in the framework of mechanistic investigations by Prosenc and co-workers.255 In recent years, further studies focused on non-noble-metal-catalyzed FA dehydrogenation.256 Through the screening of iron-, cobalt-, and manganese-based precursors, ligands, solvents, additives, and metal-to-ligand ratios, as well as the optimization of reaction conditions, the previously reported Fe(BF4)2·6H2O/ PP3 in propylene carbonate254 was conﬁrmed to exhibit optimal performance without additional base (water traces originating from the solvent did not aﬀect catalysis).256 Both Fe(II) and Fe(III) metal precursors, with the exception of those incorporating chloride, exhibited comparable activities in FA dehydrogenation in the presence of PP3, whereas Fe(0)-based catalysts were 2 orders of magnitude less active.256 Employing either Fe(BF4)2·6H2O or [FeH(PP3)](BF4)2 with 2 equiv of PP3 yielded a TON (3 h) of approximately 1900 at 40 °C, whereas the corresponding cobalt [Co(BF4)2·6H2O] and related manganese [Mn(acac)2] catalysts were almost inactive. According to the authors, the PP3 ligand had a dual functionality: It stabilized the metal center and served as a basic additive. (Similar activity was exhibited when PP3 was replaced by 50 equiv of ammonium formate.) Furthermore, in

Figure 36. Structures of non-noble-metal-based pincer catalysts that are active in FA dehydrogenation.151,243,244,246

the solvents tested in 2FA/1NEt3 dehydrogenation, THF, 1,4dioxane, and DMSO led to the highest catalyst activities, whereas in EtOH and MeCN, a decrease in the reaction rate after the ﬁrst hour pointed toward catalyst deactivation. Water was also found to be a poor solvent, likely because of reduced catalyst solubility. The base concentration was crucial to achieve high catalytic activity, with a TOF of 500 h−1 obtained in the presence of 50 mol % amine at 40 °C. With a catalyst loading of only 0.001 mol %, 100 000 turnovers were realized over a period of 10 days, translating into the consumption of 1 mol of FA. In addition, no inhibition of FA decarboxylation occurred with initial H2/CO2 pressures in the range of 0−10 bar, indicating that the reaction was irreversible under the speciﬁc set of conditions. The rhenium catalyst 60 shown in Figure 36b was synthesized by the same group through deprotonation of the cis-[(PNP3)Re(CO)2Cl] precursor, yielding its cis-[(PNP3)Re(CO)2] analogue with an unsaturated pincer ligand arm, followed by reaction with FA.244 This rhenium formato complex underwent β-hydride abstraction at temperatures above 75 °C, aﬀording the corresponding 397

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 37. (a) Proposed transition state for the β-hydride elimination step and (b) suggested FA dehydrogenation mechanism with Myers and Berben’s aluminum catalyst. (Adapted with permission from ref 246. Copyright 2014 Royal Society of Chemistry.)

four runs at 80 °C with a catalyst concentration of just 0.0001 mol %, translating into a TON of over 980 000. The group of Gonsalvi also examined base-free H2/CO2 generation from formic acid with Fe(II)/P4 catalysts in propylene carbonate (Figure 13).155 Initial tests revealed that the well-deﬁned iron complex 28 (Figure 13) was inactive for FA decomposition at 40 °C, whereas in situ formed Fe(BF4)2· 6H2O/P4 (meso/rac = 3) yielded only 4% FA conversion after 6 h. When rac-P4 was isolated by fractional crystallization from the commercial P4 mixture and utilized at an Fe/ligand ratio of 1:1, 60% of the FA substrate was dehydrogenated at an initial rate of 35 h−1. Complete FA conversion was accomplished in the presence of 2 equiv of the ligand per metal center with a higher TOF of 139 h−1. However, recycling tests at 40 °C showed a 70% decrease in catalytic activity already during the third cycle. Improved values, namely, TON = 6061 and initial TOF = 1737, were reached at a higher Fe(II)/rac-P4 (1:4) ratio of 10 000. In addition, the pure meso-P4 ligand was signiﬁcantly less active regardless of its ratio with respect to iron. Enthaler et al. synthesized a [(PCP)Ni(H)] catalyst (30, Figure 15) from the respective bromo complex through its reaction with LiBH4 in THF, either at 50 °C or after a prolonged reaction time at ambient temperature.157 In the latter case, the intermediately formed species [(PCP)Ni(H2BH2)] was isolated and characterized by X-ray diﬀraction. Upon addition of an FA/NEt3 (1:1) mixture to 30, the hydrido ligand was substituted by a formate, as indicated by a characteristic triplet resonance at 8.08 ppm (4JP−H = 2.15 Hz) and a singlet at 65.9 ppm in the 1H and 31P{1H} NMR spectra, respectively. Following complete consumption of formic acid, the initial NiH was regenerated. No activity was observed in the absence of amine, whereas use of dimethyl-N-octylamine (FA/amine = 11:10) in place of triethylamine in propylene carbonate as the solvent at 80 °C provided an improved TON of 626 after 3 h. Because of solvent cleavage under these reaction conditions an increased CO2 to H2 ratio was detected in the gas phase, upon which the authors replaced propylene carbonate with diglyme or 1,4-dioxane. In both cases no solvent degradation occurred, although catalyst activities were reduced by up to 65%. Similarly when the bromide, formate and borane analogues of 30 were tested for FA dehydrogenation in propylene carbonate at 60 °C, the TONs decreased to 85%, 39% and 36% of the initial value (i.e., TON = 122). Selective FA dehydrogenation in the absence of a metal center but with dialkylborane derivatives was investigated by the group of Cantat (Figure 38).259 Their approach was based

rhenium hydride, which further reacted with FA, eliminating molecular hydrogen. Therefore, at 0.03 mol % catalyst loading, FA dehydrogenation was demonstrated in 1,4-dioxane inside a Fischer−Porter tube, with a TOF of 3300 h−1 at 180 °C. The only study of an aluminum complex that is active in FA dehydrogenation was reported by Myers and Berben in 2014 (61, Figure 36c).246 A cooperative metal−ligand hydrogenbonding framework was suggested to be implicated in the C H cleavage step. Initial TOFs of up to 5200 h−1 were obtained with a 0.006 mol % catalyst loading in THF with an azeotropic FA/NEt3 substrate (5FA/2NEt3). Interestingly, the reaction rate was independent of the FA concentration when the latter exceeded 7.6 M. Addition of stoichiometric or excess FA to the catalyst led to the formation of an aluminum formato complex with a singly or doubly protonated ligand structure (at the amido position), respectively, and concomitant release of hydrogen gas. The authors also characterized an aluminum formato complex, obtained by insertion of CO2 into the AlH bond, by single-crystal X-ray diﬀraction and 1H NMR and IR spectroscopies. This insertion step was reversible, and acidassisted CO2 release led to the recovery of a monoprotonated aluminum hydride. The authors therefore concluded that βhydride abstraction from a doubly protonated Al−(OOCH)2 species was also feasible, reasoning that a second protonation occurring in the ligand carbon backbone would not aﬀect the electronic properties of the complex. They proposed an “outersphere” β-hydride elimination pathway based on a transition state incorporating a six-membered ring, tentatively stabilized through hydrogen bonding involving the noninnocent ligand (Figure 37a). FA dehydrogenation was suggested to proceed through catalyst activation, that is, formation of the doubly protonated Al−(OOCH)2 species as the catalyst resting state, followed by CO2 and H2 release by β-hydride elimination and further FA-enabled protonation (Figure 37b). Further insights into the reaction mechanism were recently provided through DFT computations conducted by Lu et al.258 In 2014, the groups of Hazari and Schneider demonstrated that the iron PNP2 pincer complex 62 catalyzed FA dehydrogenation with high activity and stability in the presence of a Lewis acid cocatalyst and 1,4-dioxane as the solvent, without necessitating a base (Figure 36d).151 Their studies showed an enhancement in activity in the presence of Lewis acids exhibiting a higher aﬃnity toward carboxylate, which they attributed to the stabilization of a H-bound formate intermediate. With an optimal Lewis acid (LiBF4) loading of 10%, a TOF of greater than 196 000 h−1 was accomplished over 398

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 38. Structures of organoborane derivatives catalyzing selective FA dehydrogenation.259

susceptible to reduction further enhanced FA/NEt3 (1:1) dehydrogenation, with a maximum attained conversion of 66% after 45 h or a TON of 72. Based on their previous work on molybdenum catalysts,185 Neary and Parkin investigated how electronic and structural changes in a series of [CpRMo(PMe3)3‑x(CO)xH] [CpR = Cp (cyclopentadienyl), Cp*; x = 0, 1, 2, or 3] complexes aﬀected their activity in FA dehydrogenation without base.261 Interestingly, even though the tricarbonyl and tris(phosphine) derivatives [CpMo(CO)3H] and [CpMo(PMe3)3H] were practically inactive, in the presence of the hybrid complex [CpMo(PMe3)2(CO)H], hydrogen evolution proceeded at a rate of 31 h−1 at 100 °C in benzene. Replacing the Cp ligand for a more sterically hindered Cp* group had only a minor inﬂuence on FA decomposition with a slightly increased TOF of 54 h−1. The reaction mechanism was proposed to comprise (1) catalyst protonation yielding a metal dihydride species [rather than a metal−(H2) complex], (2) elimination of H2 with concomitant coordination of formate, and (3) CO2 release from the coordinated formato ligand to regenerate the starting MoH species. Evidence for protonation (i.e., step 1) was provided for the molybdenum tris(phosphine) analogue, with [CpMo(PMe3)3H2][HCO2] identiﬁed by 1H NMR spectroscopy and the chloro equivalent [CpMo(PMe3)3H2][Cl] by Xray diﬀraction. Because [CpMo(PMe3)3H] does not catalyze FA decomposition, despite being readily protonated, the authors concluded that the inactivity was related to obstructed H2 release (i.e., step 2) from the protonated molybdenum center. In the case of the eﬃcient [CpMo(PMe3)2(CO)H] catalyst, on the other hand, not only its protonated product formed through step 1 but also [CpMo(PMe3)2(CO)(κ1OOCH)] formed by step 2 were observed. In addition, carbon dioxide dissociation from the latter yielding the initial molybdenum hydride complex was retarded in the presence of excess FA (i.e., during the initial reaction stages, the aforementioned step was rate-determining), whereas at higher FA conversions, [CpMo(PMe3)2(CO)H] became the catalyst resting state. Finally, the inactivity of [CpMo(PMe3)(CO)2H] and [CpMo(CO)3H] in FA dehydrogenation was rationalized by their inertness to protonation by formic acid, linked to the less basic metal center as a result of increased CO coordination. Notably, even though H2 and CO2 were the main reaction products, methanol was also formed under these conditions through FA disproportionation, with a maximum selectivity of 21% in the presence of the [CpMo(CO)3H] catalyst. In a subsequent study, the authors described the dehydrogenation of FA by [Ni(PMe3)4] in benzene at 80 °C with a modest TOF of 0.2 h−1 and a TON of 70 over 14 days without external base.262 In 2016, the groups of Kirchner and Gonsalvi studied the eﬀect of bases, additives, solvents, temperatures, and catalyst loadings on Fe-PNNNP1-catalyzed FA dehydrogenation (Figure 39).263 In the absence of base, no reaction occurred with any of the complexes, whereas with 1 equiv of NEt3 in

on dual FA activation: On one hand, a Brønsted base able to abstract the acidic OH proton was required, and on the other hand, a Lewis acid enabling β-hydride elimination and CO2 release from a coordinated formato moiety as part of the CH bond activation step had to be present. As a starting point, FA/ NEt3 (2:5) conversions of 48% and 26% were obtained with BBN-I and 63, respectively (Figure 38), upon heating at 130 °C for 19 h in THF. The nature of the solvent, the reaction temperature, and the concentration of the basic additive were crucial parameters for H2 generation. For example, in nonpolar (benzene, toluene) or polar protic (methanol) solvents, only low FA conversions were reported, whereas at temperatures below 120 °C, no catalytic activity was observed. A better FA conversion of 84% was reached in acetonitrile with the commercially available BBN-I. The bis(formyloxy)borate anion 64 was a common catalytic intermediate derived from both BBN-I and 63 under the reaction conditions and also active in FA dehydrogenation, as demonstrated upon its independent synthesis. When BBN-I was replaced by its dicyclohexylborane analogue, a higher FA conversion of 99% was attained within 19 h. Also in this case, the corresponding bis(formyloxy)borate anion 65 was identiﬁed to play a pivotal role in catalysis. The potential of copper-catalyzed hydrogen evolution from FA/amine mixtures was studied in a unique example by Zaccheria and co-workers.260 The nature of the copper precursor [i.e., Cu(OAc)2, Cu(OOCH)2, Cu(acac)2, Cu(NO3)2, CuO], with the exception of CuCl2 and Cu powder, which were almost inactive, had only a minor eﬀect on the exhibited activity with an FA/NEt3 (1:1) substrate at 95 °C (average TON = 20 and TOF = 1 h−1), most likely because of fast exchange of the ligand by either the amine or formate under the reaction conditions. In contrast, the basicity of the amine was a crucial parameter aﬀecting FA dehydrogenation, even though no linear correlation could be determined. Screening studies demonstrated that both the nucleophilicity and steric hindrance of the base were relevant for catalysis, with higher activities obtained by lowering the former and increasing the latter. The authors reported that strong chelation of the amine to the metal center, such as in the case of ethylenediamine, resulted in a blockage of the necessary coordination sites and therefore inhibition of FA dehydrogenation. The replacement of triethylamine with tributylamine increased FA conversion from 17.2% to 25.6% after 22 h at 95 °C with a Cu(OAc)2 loading of 0.93%. Reduction of the active Cu(II) species under the basic reaction conditions leading to deactivation could be reversed through simple exposure to air. In addition, the presence of a stoichiometric amount of strongly coordinating ethylenediamine relative to the metal successfully stabilized the catalyst in solution. Alternatively, formation of Ru(0) nanoparticles could be avoided by addition of 1 equiv of a secondary acid (acetic acid) with respect to FA, yielding a higher substrate conversion of 24.5% (vs 16.4%) after 22 h. The use of copper precursors such as CuI that were less 399

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Although a few examples of cobalt-catalyzed CO2 hydrogenation exist,73,142,144,148,153 no experimental study on the reverse FA dehydrogenation reaction has been reported. In 2016, Yang and colleagues proposed a series of cobalt catalysts bearing acylmethylpyridinol and aliphatic PNP-type pincer ligands as potentially active for reversible H2 storage in the formic acid/CO2 couple, based on DFT calculations (Figure 41).265 The total free energy barrier was calculated to be as low

Figure 39. Structures of iron-PNNNP1 pincer catalysts that are active in FA dehydrogenation.263 Figure 41. Structure of cobalt-based catalysts computationally designed by Yang and colleagues.265

THF, reaction rates between 610 and 720 h−1 (T = 60 °C) and full conversion of the 5 M FA solution were reached. Complex 66 showed only poor activity in DMOA and was inactive in DBU, unlike 67, which was not sensitive to the nature of the base. Replacement of the base by a Lewis acid completely inhibited FA dehydrogenation with all catalysts. In addition, it was observed that higher reaction rates and overall FA conversions could be obtained in aprotic solvents such as THF, PC, and 1,4-dioxane. In the case of 67, doubling the FA concentration from 5 to 10 M and reducing the catalyst loading to 0.01 mol % gave the highest TOF of 2635 h−1 and full conversion after 6 h at 80 °C. Both 67 and 68 could be recycled several times to aﬀord TONs of over 12 000, although the catalyst-to-acid ratio had to be increased with 67 to maintain its stability (from 1:5000 to 1:1000). In the framework of their research on small-molecule activation, Wang et al. reported that (bis-aryl)imine-supported iron catalysts exhibited moderate activity in FA dehydrogenation in the presence of 50 mol % NEt3 and 10 mol % LiBF4, which were needed to ensure substrate activation.264 Through variations of the nature, number, or position of the aryl substituents, complex 70 (Figure 40) emerged as the most eﬃcient, albeit exhibiting only moderate activity and stability at 80 °C; heating for 24 h in benzene led to a maximum FA conversion of 31% and TON of 594. Under the reaction conditions, when 69 was used as the catalyst, rapid conversion to 70 was observed by 1 H NMR spectroscopy with simultaneous gas evolution. However, prolonged heating of either 69 or 70 led to catalyst deactivation as a result of cleavage of the ﬂuoroaryl imine ligand, as conﬁrmed by mass spectrometry and NMR spectroscopy.

as 23.1 and 20.9 kcal mol−1 in water and THF, respectively, for R1 = Me, R2 = OH, and R3 = NH. The computational design of catalysts can indeed provide a solid basis for experimental studies as shown by comparison of the results obtained by the authors266 and Milstein.146 2.4. Reversible H2 Storage

The basis for the development of a reversible H2 storage device is the interconnection of CO2/bicarbonate hydrogenation to FA (derivatives) and the dehydrogenation of the latter in a continuous and reproducible manner. Furthermore, the systems should be characterized by practical energy densities, fast kinetics, and stable catalyst activities. The potential applicability is also directly related to the conversions attained in each step upon tuning of the reaction parameters. Even though several catalysts are known to be active for both steps independently, in certain cases under diﬀerent reaction conditions (see refs 90, 112, 116, 121, 141, 151, 152, 155, and 157), in this section, we summarize the experimental demonstrations of coupled H2 ﬁxation and release (see Table 3). The ﬁrst example of a such a reversible CO2-based H2 storage cycle dates back to 1994.56 In their study, Leitner, Dinjus, and Gatßner demonstrated that the FA adduct produced in the presence of [Rh(COD)Cl]2/DPPB in an acetone/NEt3 mixture under 40 bar H2/CO2 at ambient temperature could be directly dehydrogenated upon release of the overpressure. Hydrogen incorporation occurred at a rate of approximately 54 h−1 yielding 1.7 M FA, which could then be decomposed with a TOF of 30 h−1. Despite an observed

Figure 40. Deactivation pathway of the ﬂuoroarylimine-stabilized iron catalyst developed by Li and co-workers. (Adapted with permission from ref 264. Copyright 2016 John Wiley & Sons.) 400

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Table 3. Selected Catalysts Active in Formic Acid Dehydrogenation catalyst precursora

substrate

solvent

T (°C)

timeb (h)

TONc

TOFc (h−1)

[RuCl2(PPh3)3]188 [Ru(H2O)6](tos)2/mTPPTS191 [RuCl2(C6H6)]2/DPPE198 [RuCl2(C6H6)]2/DPPE172 [RuCl2(C6H6)]2/DPPE202 [(PNP3)Ru(H)Cl(CO)]d (10)120 [RuCl2(benzene)]2/meso-P4d242 [(PNNNP2)RuH2(CO)]d (51)247 [IrH3(PPh3)3]173 [(Cp*Ir)2(THBPM)Cl2]Cl2 (4)113 [(PNP1)IrH3]d (8)121 [Cp*Ir(4-(1H-pyrazol-1-yl-κN2)benzoic acid-κC3)(H2O)]2SO4 (35)116 [Cp*Ir(H2O)(BPM)Ru(BPY)2](SO4)2 (34)205 [Cp*Ir(4,4′-DHBP)(H2O)]SO4 (36)206 [Cp*Ir(TMBI)H2O]SO4 (37)211 [Cp*Ir(pyrimidyl imidazoline)H2O]SO4 (38)212 [Cp*Ir(THBPM)(H2O)]SO4 (39)209 [Cp*Ir(PHEN-diol)H2O]SO4 (40)214 [Cp*Ir(L30)Cl]d 231 [Cp*Ir(L41)Cl]Cld (54)248 [Cp*IrCl2]2/(L42)d (55)248 [Ir(COD)(EMIM)(mTPPMS)] (56)249 [cis-mer-IrH2Cl(mTPPMS)3] (57)250 [Ir(COD)(NP1)](TfO)d (58)251 [Cp*Rh(H2O)(BPY)]2+ 204 [Cp*Rh(bis-NHC)Cl]Nad (18)141 [Cp*RhCl2]2/KI235 [(PNC)Rh(CO)]d (43)237 [Fe3(CO)12]/PPh3/PhTPY252 [Fe3(CO)12]/benzylphosphine/TPY253 Fe(BF4)2/PP3254 [(PNP3)Fe(H)2(CO)]d (59)243 [(PNP2)Fe(H)(CO)(OOCH)]d (62)151 Fe(BF4)2·6H2O/rac-P4d155 FeCl2/sPP3d257 [(PNNNP1)Fe(H)(Br)(CO)]d (67)263 [Fe] complexd (70)264 cis-[(PNP3)Re(CO)2(OOCH)]d (60)244 Cu(OAc)2260 [(PhI2P2−)Al(THF)H] (61)246 [Cp*MoH(PMe3)2(CO)]261 [(PCP1)Ni(H)]d (30)157 [Ni(PMe3)4]262 Cy2B−I259

HCO2H/NEt3 HCOOH/NaOOCH HCO2H HCO2H HCO2H HCO2H/NHex3 HCO2H/DMOA HCO2H/NEt3 HCO2H HCOOH/NaOOCH HCO2H/NEt3 HCOOH/NaOOCH HCOOH/NaOOCH HCO2H HCO2H HCOOH/NaOOCH HCOOH/NaOOCH HCO2H HCO2H/NEt3 HCO2H HCO2H NaOOCH HCOOH/NaOOCH HCOOH/NaOOCH HCOOH/NaOOCH HCO2H HCO2H/NEt3 HCO2H HCO2H/NEt3 HCO2H/NEt3 HCO2H HCO2H/NEt3 HCO2H HCO2H HCO2H HCO2H/NEt3 HCO2H/NEt3 HCO2H HCO2H HCO2H/NEt3 HCO2H HCO2H/nOctNMe2 HCO2H HCO2H/NEt3

DMF H2 O Me2NHex Me2NHex DMOA DMF PC DMSO AcOH H2 O t BuOH H2 O H2 O H2 O H2 O H2 O H2 O H2 O − H2 O H2 O H2 O H2 O − H2 O H2 O − 1,4-dioxane DMF DMF PC 1,4-dioxane 1,4-dioxane/LiBF4 PC H2 O PC C6H6/LiBF4 1,4-dioxane NEt3 THF C6H6 PC C6H6 MeCN

40 120 40 80 25 90 60 90 118 90 80 25 25 90 80 100 80 60 40 90 80 80 100 90 25 100 60 75 60 60 80 40 80 60 80 80 80 180 95 65 100 80 80 130

0.3 90 264 NA 1080 3 48 150 NA 0.25 0.017 NA NA NA 0.17 0.17 0.08 2600 0.003 0.005 14 NA 0.2 2880 NA 50 1 000 000 706 500 >220 000 1 100 000 >11 000 165 000 5000 NA NA NA 10 000 68 000 11 000 5 000 000 NA 47 000 2 400 000 NA 60 000 2 160 000 NA 449‘000 700 NA NA 1266 >92 000 100 000 985 000 6000 13 000 10 000 594 3300 20 2200 NA 630 70 100

2688 460 900 47 970 1000 256 000 4580 7300 8900 228 000 120 000 >2000 426 14 000 34 000 322 000 39 500 1900 147 000 487 500 171 400 15 110 298 000 NA 28 9‘000 4400 170 200 25 5800 420 197 000 1700 240 2600 25 3300 1 5200 54 210 0.2 2.5

a

With ID number used in text, where applicable. bTime corresponding to TOF. cValues were either directly taken from the corresponding references or calculated based on experimental data provided within; insigniﬁcant digits were rounded. dFor complete structures, see main text.

induction period during the ﬁrst H2 storage run, a recycling experiment (repressurization with 40 bar H2/CO2) commenced without any delay and at a similar rate of 65 h−1. The H2 storage capacity of this system was 1.7 mol or 3.4 g of H2 per liter of solution. Beller and colleagues reported that the catalyst generated in situ from [(RuCl2(benzene))2]/DPPM was active both in the hydrogenation of bicarbonates, carbonates, and carbon dioxide in basic media and in the dehydrogenation of formate.126 Therefore, the reactions were independently optimized and subsequently interconnected. The maximum formate yield of 96% (TON = 1108) was reached in sodium bicarbonate hydrogenation in the absence of CO2 after 20 h at 70 °C. In this case, catalysis was realized in a THF/H2O solvent under 80

bar H2. Sodium formate dehydrogenation in DMF/H2O also yielded complete conversion (owing to the release of the produced H2 gas) and an initial TOF of almost 2600 at 60 °C. Finally, in a proof-of-principle experiment, the authors performed both the hydrogenation and dehydrogenation reactions with the same recycled catalyst. In this case, following formate dehydrogenation at 30 °C, the DMF/H2O solvent was removed, and the recovered catalyst and hydrogen carrier were dissolved in THF/H2O to perform bicarbonate hydrogenation at 80 °C, resulting in 80% conversion to formate. For this system, the H2 storage capacity was calculated as ∼1.6 g of H2 per liter of solution. In collaboration with the group of Laurenczy, Beller and co-workers further improved the hydrogen storage step by adjustment of the reaction conditions: 401

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

h−1 reported for FA dehydrogenation. An AAR of 1.6 corresponds to an FA concentration of approximately 1.5 M, equivalent to a H2 density of 3 g per liter of solution. This reversible H2 storage principle was also examined by Plietker and co-workers with a [Ru(PNNP)(CH3CN)Cl] pincer complex (72, Figure 42).245 With 0.0075 mol % 72, a

In a DMF/NEt3 solution, an AAR of 2.69 was attained under 60 bar CO2/H2 upon application of an appropriate temperature program, namely, 2 h at 100 °C and 2 h at room temperature.172 Because the intensive heating was necessary only for the formation of the catalytically active species, it could be circumvented by employing the preformed complex [RuH2(DPPM)2]2 instead of [(RuCl2(benzene))2]/DPPM. Thus, the feasibility of a reversible “hydrogen-battery” was demonstrated in eight continuous H2 storage/release cycles, where CO2 was converted into a triethylammonium formate salt that was then dehydrogenated at ambient temperature and pressure. Signs of catalyst deactivation became apparent during the eighth cycle. Based on their previous experience in CO2 hydrogenation in basic media,94−99,102 the groups of Joó and Laurenczy demonstrated that H2 could be reversibly stored in a sodium formate/bicarbonate couple through adjustment of the reaction pressure without removal of the solvent or hydrogen carrier or tuning of the pH.103 In the presence of [RuCl2(mTPPMS)2]2/ mTPPMS and under isochoric conditions at 83 °C, sodium formate dehydrogenation resulted in a H2 pressure buildup and approximately 50% conversion to bicarbonate, as dictated by the thermodynamic equilibrium under the speciﬁc reaction conditions. The obtained solution containing both the charged and discharged carriers was subsequently pressurized with 100 bar H2 gas and heated, yielding 90% conversion to formate. The overall ﬁxation/evolution sequence was repeated three times without any notable catalyst deactivation. Furthermore, venting of the produced H2 pressure during formate decomposition further promoted the reaction, thus providing a convenient way of controlling the energy output (i.e., H2 gas production). However, because of the low substrate concentration, the overall H2 storage density of the system was limited to 0.5 g of H2 per liter of solution. The binuclear iridium complex featuring four hydroxyl functionalities attached to a bridging bipyrimidine ligand (Figure 2c) reported in 2012 by the groups of Hull, Fujita, and Himeda exhibited remarkable activity in CO2/bicarbonate hydrogenation in basic solution and FA dehydrogenation under acidic conditions (see sections 2.2.1 and 2.3.1, respectively).113 The pH-dependent activity of 4 was further evaluated in a reversible manner. The catalyst was initially dissolved in a 2 M KHCO3 solution and subjected to a CO2/H2 gas ﬂow for 136 h, leading to the production of 0.48 M formate (H2 storage density of 1 g of H2 per liter of solution). The solution pH was then adjusted to 1.7 with sulfuric acid before formic acid dehydrogenation was initiated at 50 °C. When the reaction pressure leveled oﬀ at 23 bar, addition of KHCO3 was employed to restore the initial pH and initiate a new cycle. The ruthenium pincer complex 10 (Figure 3c) examined by Pidko and co-workers120 catalyzed CO2 hydrogenation and FA decomposition with rates of up to 1 100 000 h−1 (DMF/DBU, 40 bar, 120 °C) and 257 000 h−1 (DMF/NEt3, 90 °C), respectively. When triethylamine was exchanged for DBU, FA dehydrogenation occurred with a lower TOF of 93 100 h−1. In a ﬁnal demonstration of a closed H2 storage cycle, the authors performed 10 consecutive H2 ﬁxation/liberation cycles over a period of 1 week in a DMF/DBU solvent mixture. Hydrogen uptake was realized at 65 °C under either 40 or 5 bar pressure (H2/CO2 = 1:1), and dehydrogenation was realized at 90 °C under atmospheric pressure. Reproducible H2 gas volumes and AARs (1.6 and 1.1 for high and low pressure loadings, respectively) were obtained, with a maximum TOF of 150 000

Figure 42. Structure of ruthenium pincer catalyst 72 developed by Plietker and colleagues for reversible H2 storage.245

DBU−formate salt (1:1, 5 mmol) could be selectively decomposed under atmospheric pressure at a rate of 1140 h−1 in toluene at 100 °C. With the same catalyst loading in neat DBU, a formic acid−DBU adduct with a ratio of approximately 1.4:1 was obtained from CO2 hydrogenation, whereas with 0.015 mol % 72, only a 0.85:1 adduct was formed, corresponding to a TON of 5600. Interestingly, the addition of toluene had almost no eﬀect on the product yield with the higher catalyst loading but led to decreased activity (TON = 4000) in the case of the lower metal concentration. During these experiments, carbon dioxide was supplied in its solid form, and the reactor was pressurized with 70 bar H2 and subsequently heated to 100 °C, resulting in an initial reaction pressure of about 140 bar. The hydrogen charge and discharge steps were then coupled in a continuous manner, resulting in similar gas productions over a total of ﬁve cycles. Both the evolving gas pressure and the reaction temperature could be used to control the yield and rate of FA dehydrogenation. The groups of Czaun, Prakash, and Olah tested the aliphatic ruthenium PNP4 pincer complex 16 (Figure 9) in bicarbonate hydrogenation (at 80 °C under 53 bar H2) and H2 evolution from the resulting formate salt following decompression (at 60 °C) in THF/H2O.139 One cycle was performed with a reported formate dehydrogenation yield of over 90%, although no results were provided for bicarbonate hydrogenation. Catalyst 16 was also employed in CO2 hydrogenation (3H2/CO2, 75 bar) in the presence of NaOH at 70 °C and the subsequent dehydrogenation of the formed formate salt under atmospheric pressure in a 1,4-dioxane/H2O medium. The overall charge−discharge sequence was repeated six times, resulting in substrate conversions of over 90% and a ﬁnal TON of 11 500, without notable loss of catalytic activity. In this case, the H2 storage capacity can be calculated as 16.4 L of H2 or approximately 1.5 g of H2 per liter of solution. In the most recent publication regarding reversible and interconnected H2 storage/evolution, cesium formate and bicarbonate salts were proposed as hydrogen carriers because of their 20-fold-higher water solubility with respect to their sodium analogues, allowing for a more practical H2 storage density of 14 g L−1 (for a 11 mol kgwater−1 CsOOCH solution at 15 °C).267 Naturally, because the [RuCl2(mTPPTS)2]2/ mTPPTS catalyst was active in both bicarbonate hydrogenation and formate decomposition, the reaction yields were governed 402

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Figure 43. Proposed reaction mechanism for the dehydrogenation of formate with [RuCl2(mTPPTS)2]2 /mTPPTS. (Adapted with permission from ref 267. Copyright 2015 John Wiley & Sons.)

by thermodynamic equilibria that were dependent on the reaction conditions (temperature and pressure). In practice, that meant that, at atmospheric pressure, a formate dehydrogenation yield of 100% was possible, whereas increasing H2 pressures shifted the equilibrium toward the side of formate. As a result, the equilibrium conversion of a 15 mol kgwater−1 CsOOCH solution under isochoric conditions was just 20% (where this value was dependent on the equilibrating pressure). The directing eﬀect of pressure can be rationalized through the reversible formation of a pentacoordinated [RuH(H2O)(mTPPTS)3]+ intermediate (73). Nevertheless, ﬁve consecutive H2 storage and release cycles were demonstrated without any decrease in catalytic activity or cumbersome adjustment of the reaction environment; bicarbonate hydrogenation yields between 95 and 98% and formate dehydrogenation yields of about 62% were obtained. Through NMR investigations, a number of catalytically active species were identiﬁed, and a reaction mechanism for formate dehydrogenation was proposed (Figure 43). Species 73 reversibly takes up H2 and generates a nonclassical [Ru(H)(H2)(H2O)(mTPPTS)3]+ complex (73b), hindering progression to cycle II and, thus, formate dehydrogenation.

Figure 44. Reversible hydrogen storage cycle based on the methanol/ CO2 couple.

molecules of hydrogen (Table 4, eqs 10 and 11) or, in the presence of water, to CO2 and three molecules of hydrogen (Table 4, eqs 12 and 13). In the latter case, one H2 molecule arises from water. All of these processes are endothermic and endergonic. Considering hydrogen use and CO poisoning, the reactions given in eqs 9−11 (Table 4) are undesirable. On the other hand, eqs 12 and 13 (Table 4) and their reverse reactions are expected to be essential for a methanol economy. Currently, practical applications of methanol synthesis starting from CO2 and H2 are only based on heterogeneous catalysis. The existing three plants are (1) the George Olah CO2 to Renewable Methanol Plant in Iceland with a capacity of 3500 tons per year; (2) Mitsui Chemical in Japan, which aims for an annual production of 100 tons; and (3) Silicon Fire AG, which produces 50 tons per day.268 The CO2-to-methanol synthesis is mainly based on heterogeneous Cu/ZnO catalysts. The reverse reaction, the steam reforming of methanol, is applied to release hydrogen from methanol (Table 4, eq 12) and usually takes place at higher temperatures (200−300 °C) with heterogeneous catalysts.271,272 Copper-based catalysts such as Cu/Zn/ Al2O3 and CuO/ZnO/Al2O3 show excellent activity and are among the most frequently used catalysts for this reaction. However, high-temperature steam reforming of methanol also encounters the problem of CO generation in a non-negligible amount, poisoning the electrodes of reformed methanol fuel cells (RMFCs). This issue can be overcome by applying aqueous-phase reforming (APR) of methanol (Table 4, eq 13). The APR process was ﬁrst reported by Dumesic and coworkers and takes place at temperatures of 200−225 °C using a heterogeneous Pt/Al2O3 catalyst.273 In view of this state of the art, the homogeneously catalyzed aqueous-phase reforming of

3. METHANOL AS A HYDROGEN STORAGE MEDIUM 3.1. The Carbon Dioxide−Methanol Couple

In addition to FA, the simplest alcohol, methanol, is considered to be a promising chemical energy carrier because of its relatively high hydrogen content, low volatility under ambient conditions, and the possibility of using the existing infrastructure for its storage and transportation.268 Methanol is a colorless and odorless liquid that contains 12.6 wt % (99 g L−1) hydrogen and shows good miscibility with water. In the late 1960s, Asinger269 already proposed the idea of a methanol economy based on the carbon-neutral hydrogen loading of CO2 and unloading of the resulting methanol (Figure 44), and more recently, Olah has popularized this concept.270 Methanol dehydrogenation can proceed by three diﬀerent pathways categorized by the use of hydrogen atoms in the substrate molecule(s). When only two hydrogen atoms of methanol are used, dehydrogenation to formaldehyde and hydrogen occurs (Table 4, eq 9). If all four hydrogen atoms are released, methanol is dehydrogenated either to CO and two 403

DOI: 10.1021/acs.chemrev.7b00182 Chem. Rev. 2018, 118, 372−433

Chemical Reviews

Review

Table 4. Thermodynamic Data for Reversible H2 Storage Based on Methanol/CO2274,275 ΔH° (kJ mol−1)

ΔS° (J mol−1 K−1)

ΔG° (kJ mol−1)

equation

+129.8 +94.6 +127.9 +53.3 +130.7

+222 +219.1 +332 +176.8 +408.7

+63.5 +29.3 +29.0 +0.6 +8.9

(9) (10) (11) (12) (13)

CH3OH(l) → HCHO(g) + H2(g) CH3OH(g) → CO(g) + 2H2(g) CH3OH(l) → CO(g) + 2H2(g) CH3OH(g) + H2O(g) → 3H2(g) + CO2(g) CH3OH(l) + H2O(l) → 3H2(g) + CO2(g)

methanol at lower temperatures (

Homogeneous Catalysis for Sustainable Hydrogen ... - ACS Publications

Recommend Documents