Report

Efficient Dehydrogenation of Formic Acid Using an Iron Catalyst

See allHide authors and affiliations

Science  23 Sep 2011:
Vol. 333, Issue 6050, pp. 1733-1736
DOI: 10.1126/science.1206613

Abstract

Hydrogen is one of the essential reactants in the chemical industry, though its generation from renewable sources and storage in a safe and reversible manner remain challenging. Formic acid (HCO2H or FA) is a promising source and storage material in this respect. Here, we present a highly active iron catalyst system for the liberation of H2 from FA. Applying 0.005 mole percent of Fe(BF4)2·6H2O and tris[(2-diphenylphosphino)ethyl]phosphine [P(CH2CH2PPh2)3, PP3] to a solution of FA in environmentally benign propylene carbonate, with no further additives or base, affords turnover frequencies up to 9425 per hour and a turnover number of more than 92,000 at 80°C. We used in situ nuclear magnetic resonance spectroscopy, kinetic studies, and density functional theory calculations to explain possible reaction mechanisms.

Hydrogen is of critical importance in the chemical industry and might play a key role in the future in renewable energy technologies. As a potential secondary energy vector it would enable clean energy storage and transduction. If H2 is combusted in engines or fuel cells, only water emerges as benign exhaust (13). Although synthetic catalytic transformations of hydrogen mainly rely on expensive, low-abundant precious metal catalysts (4, 5), iron-based enzymes efficiently metabolize H2 as an energy source in many organisms. For example, natural hydrogenases binding iron or iron and nickel centers catalyze the formation or activation of H2 with turnover frequencies of 103 to 104 per second at 30°C (6), far exceeding currently known industrial hydrogenation/dehydrogenation catalysts. Besides the [FeFe]- and [FeNi]-metalloenzymes, the [Fe]-hydrogenase, a monomeric species bearing one iron center, is known to facilitate the oxidation of an organic hydride donor to give metal hydrides that can subsequently react with protons to produce H2 (7). Despite these impressive properties of iron-based enzymes, there are only few examples, mostly artificial water reduction catalysts, wherein structural analogs have been successfully applied in synthetic catalysis (810).

Today, the industrial production of hydrogen is mainly achieved by steam reforming and coal gasification, which are based on limited fossil resources such as natural gas, coal, and oil. On a mid- to long-term basis, there is an increasing demand for alternative feedstocks to generate hydrogen in a more sustainable manner. Recently, formic acid (FA), a major product of biomass processing and formally an adduct of H2 and CO2, has attracted considerable attention as a suitable liquid source for hydrogen and as a potential hydrogen storage material (11). A sustainable and reversible energy storage cycle can be easily envisioned by applying CO2 use of which as a C1-source is highly desirable as a supplement or alternative to underground sequestration (12). Again looking to nature for inspiration, the enzyme formate dehydrogenase (FDHH) in Escherichia coli bacteria catalyzes the oxidation of formate ions to carbon dioxide, a process coupled with the reduction of nicotinamide adenine dinucleotide/nicotinamide adenine dinucleotide phosphate (NAD+/NADP+) to their reduced forms (NADH/NADPH) that can go on to liberate H2 through hydrogenases (1315). However, successful demonstrations of bio-inspired metal catalyst systems for hydrogen evolution from FA/formates are scarce (16, 17).

Here, we describe well-defined iron catalyst systems capable of dehydrogenating HCO2H to CO2 in acidic media. H2 liberation takes place at near-ambient conditions with unexpectedly high catalyst activities comparable to the best-known noble metal catalysts. Contrary to the established catalytic FA dehydrogenation systems, this iron-based catalyst system needs no additional base or further additives, nor a highly sophisticated reaction setup (18, 19).

Only a few non-noble metal-based catalysts are known for the selective dehydrogenation of FA. All these catalyst systems showed activity only in the presence of visible light and additional base, hindering large-scale application. We sought high activity in nontoxic solvents under more readily scalable conditions. Through extensive explorative testing of commercially available iron precursors and various stabilizing ligands, we discovered the combination of cationic iron(II) sources, for example, Fe(BF4)2·6H2O, and the tetradentate ligand tris[(2-diphenylphosphino)ethyl]phosphine [P(CH2CH2PPh2)3, PP3] as an active catalyst system (Table 1). We applied the nontoxic and fully biodegradable propylene carbonate (PC) as a solvent (20). In general, catalytic activity toward FA dehydrogenation was investigated quantitatively with gas evolution measurements using automatic burets and high-pressure nuclear magnetic resonance (NMR) spectroscopy, and qualitatively by gas chromatography (GC) analysis. [For detailed instrumentation setup, see (21, 22).]

Table 1

Selective iron-catalyzed hydrogen evolution from FA. Conditions: 5.3 μmol (0.01 mol % [Fe]), as stock solution or powder, plus 2 equiv. PP3 were added to 5 mL PC and 2 mL HCO2H in an argon-purged reaction vessel, with temperature kept constant at 40°C; volumes of evolved gases (mL) were measured with automatic gas burets and analyzed by GC (H2:CO2 1:1); <10 ppm CO was detected. In the case of [FeCl(PP3)]BF4, neither H2 nor CO2 could be detected. Results varied between 1 and 15% over at least one repeated experiment. [FeH(H2)(PP3)]+X (X = BF4, BPh4) (31), [FeH(PP3)][BF4] (32), and [FeCl(PP3)][BPh4] (33) were synthesized according to literature protocols. V, gas volume of H2 and CO2; 2h, 2 hours; 3h, 3 hours.

View this table:
Fig. 1

Continuous iron-catalyzed hydrogen production from FA. In a typical experiment, we applied 74 μmol Fe-precursor and 4 equiv. of PP3 ligand in 50-mL PC with a constant dosage of FA (0.27 ± 0.04 mL min−1) at 80°C. For further details, see (22).

In the presence of only 0.01 mole percent (mol %) of the precursor Fe(BF4)2·6 H2O together with 1 equivalent (equiv.) of PP3, constant hydrogen evolution [CO < 10 parts per million (ppm)] was observed at 40°C, with a turnover number (TON) of more than 825 after 3 hours (22). Using a ligand-to-iron ratio of two enhanced the catalytic activity (TON = 1942) considerably. Previously, we applied the well-defined [FeH(PP3)]+ complex for the hydrogenation of carbon dioxide (23). Therefore, we also investigated the activities of various presynthesized iron-hydride complexes posited as the active catalysts (Table 1). Interestingly, the monohydride complex [FeH(PP3)]BF4 gave slightly higher activity compared with the in situ assembled system: When a second equivalent of PP3 was added, comparable activity was observed to the in situ system employing 2 equiv. PP3. The [FeH(H2)(PP3)]+ complex with BF4 and BPh4 as counter ions showed similar activities as [FeH(PP3)]BF4, whereas the catalytic activity of [FeH(H2)(PP3)]BPh4 was only slightly lower than [FeH(H2)(PP3)]BF4. All tested presynthesized iron-hydride complexes showed notably higher activity compared to the in situ assembled system with only one equiv. PP3. Except in the case of [FeCl(PP3)]BF4, which showed no activity, full conversion (>95%) was reached in less than 24 hours.

Without the Fe-precursor/ligand or the presynthesized catalyst, no gas evolution was observed. To determine the influence of the evolved product gases (H2 and CO2), we applied isochoric reaction conditions in sapphire tubes (fig. S3). In contrast to the experiments under isobaric conditions, a nonlinear gas evolution curve was observed, indicating an influence of increasing product concentrations in the liquid phase during catalysis. Therefore, we performed several experiments with different initial pressures of H2 and CO2. The catalyst system showed a high tolerance toward CO2, with no significant influence up to 30 bar of CO2 (~23 mmol CO2). However, at moderate H2 pressures (~4 mmol H2 at 20 bar H2), the activity dropped significantly, from TON 325 (without additional pressure) to TON 136 (20 bar H2) after 3 hours (table S1). In addition, we investigated the stability of the derived catalytic system toward water, air, and other traces of impurities typically present in commercially available FA (tables S2 and S3). No influence of air (solvents and FA not degassed) or traces of water (reagents not dried or distilled) was observed, a robustness vital for the long-term applicability of this system. However, at water contents as high as 3.7 and 6.9 M, decreases in activity of about 30 and 50%, respectively, compared with standard conditions, were seen. The catalyst system proved sensitive toward traces of chloride; adding chloride to the reaction mixture at a Cl to Fe ratio of 1:1 resulted in a significant decrease of activity (TON of 91). With 15 equiv. Cl present, the activity dropped nearly to zero (table S3).

The catalyst system is also remarkably stable at higher temperatures. For example, using 0.01 mol % (5.3 μmol) Fe(BF4)·6H2O together with 4 equiv. PP3 in PC, we measured TOFs of 2018 hour−1 at 60°C and 8136 hour−1 at 80°C. Using only 0.005 mol % (50 ppm) of iron catalyst, we observed a TOF of 9425 hour−1 at 80°C. To investigate the long-term stability of our system, we used a device for continuous hydrogen production (22). After a short induction period, a constant gas flow of 325.6 mL/min (H2 + CO2) was detected over a time period of 16 hours without significant decrease of activity (Fig. 1). Then, due to chloride and/or water accumulation in the batch reactor, deactivation occurred. Nevertheless, a total TON of 92,417 was achieved. In addition, the CO content did not exceed 20 ppm.

Next, we studied the rate dependence on temperature, FA, and catalyst concentration (24, 25). The temperature dependence of the initial rates (conversion <20%) followed clean Arrhenius behavior (Fig. 2). Activation energies derived for HCO2H dehydrogenation in THF (EA = 82.1 ± 1.3 kJ mol−1) and PC (EA = 77.0 ± 0.4 kJ mol–1) indicate a minor influence of the solvent.

Fig. 2

Arrhenius plots of TOF values for hydrogen generation from HCO2H and DCO2D in PC and THF. Stock solution [5.3 μmol (0.01 mol %) Fe(BF4)2·6H2O] plus 2 equiv. PP3 were added to 5 mL PC or THF and 2 mL HCO2H in an argon-purged reaction vessel, respective reaction temperature was kept constant during the reaction, and volumes of evolved gases were measured with automatic gas burets and analyzed by GC (H2 or D2:CO2 1:1); <10 ppm CO was detected.

These observed activation energies are in the typical range of noble metal–catalyzed hydrogen generation from FA (26). When HCO2H was replaced by DCO2D (EA = 85.4 ± 3.3 kJ mol−1 in THF), a small primary deuterium kinetic isotope (KIE) effect of 3.00 ± 0.49 at 40°C was observed [see section S4 of (22)].

Subsequently, we determined the dependence of the reaction rate on the catalyst concentration with Fe(BF4)2·6H2O / 2 equiv. PP3. The concentration of the precatalyst Fe(BF4)2·6H2O and 2 equiv. PP3 was varied between 0.76 and 3.00 mM. To keep the initial concentration of FA (7.6 M) and the amount of solvent (5 mL) constant, we kept the conversion below 20%.

From a double logarithmic plot of the initial rate against the catalyst concentration, a linear dependence on the iron concentration was determined (Fig. 3A). Thus, the reaction proceeded first order (0.92) in iron concentration, which implies that Fe(BF4)·6H2O / 2 PP3 is rapidly converted into the active species during the induction period (<1 min) and that dimeric catalyst species are not involved in the reaction. Hence, the active catalyst must be homogeneous. To determine the reaction order in FA, we varied the concentration of FA between 0.5 and 4.4 M, keeping the catalyst concentration constant at 0.76 mM. The reaction rate showed a positive broken reaction order of 0.44 at FA concentrations up to 2.4 M. This result implies that only one FA molecule coordinates to the iron catalyst, which probably occurs before the rate-determining step, and that an equilibrium is involved between [Fe-catalyst] + HCO2H and an assumed [Fe-catalyst-HCO2H] complex. With FA concentrations exceeding 2.4 M, we observed no further increase of the reaction rate (Fig. 3B).

Fig. 3

(A) Plot of ln(rate of H2 evolution) versus ln([catalyst]). We used 0.76 mM to 3.00 mM Fe(BF4)2·6H2O and 2 equiv. PP3 in 5 mL PC with 2 mL FA, reaction temperature was kept at 40°C, gas evolution was monitored with automatic gas buret, and gases were analyzed by GC. (B) Plot of ln(rate of hydrogen evolution) versus ln([FA]). We used 0.76 mM Fe(BF4)2·6 H2O and 2 equiv. PP3 in 5 mL PC and 0.5 to 4.4 M FA; otherwise, conditions and methods were as in (A).

Well-defined iron(II)hydridophosphine complexes have been investigated in the past (27), and some reports on the redox properties of classical monohydride, dihydride iron(II) complexes and nonclassical iron(II)hydrides with PP3 have been published (28, 29). We also reported on some NMR studies on the carbon dioxide hydrogenation by applying Fe(BF4)2·6 H2O and PP3 (23). Based on these findings, the occurrence of various neutral and cationic Fe-hydride species can be assumed. Based on our observations in the kinetic studies and the catalyst tests of various well-defined iron species, we explored the present reaction mechanism in more detail using in situ high-pressure NMR spectroscopy and additional density functional theory calculations (22). Stirring a mixture of Fe(BF4)2·6 H2O and PP3 in HCO2H and THF resulted in clean formation of [FeH(H2)(PP3)]+ and [FeH(PP3)]+, which have been characterized previously by x-ray crystallography and NMR (23). Using high-pressure in situ NMR spectroscopy with 13C-enriched FA [for experimental setup, see (22, 30)], we could detect [FeH(CO2)(PP3)]+ with a broad singlet Fe-CO2 signal at 206.2 ppm in the 13C NMR spectrum. Moreover, the complex [Fe(HCO2)(PP3)]+ gave rise to a doublet of quartets at 164.5 ppm [J(H,C) = 206.9 Hz and J(P,C) = 4.9 Hz], where J is the spin-spin coupling constant between the H and the C nuclei. A broad doublet at 173.1 ppm [J(H,C) = 204.7 Hz] indicated the presence of [Fe(H2)(HCO2)(PP3)]+ or [Fe(H)(HCO2)(PP3)]. During catalysis (Fe:PP3 = 1:1), the 31P NMR spectrum exhibited two singlets at 89.9 ppm (3 P atoms) and 173.6 ppm (1 P atom), indicative of a fully coordinated PP3 ligand. However, when an additional PP3 was present, the 31P NMR spectrum exhibited additional signals at +19.2 ppm (multiplet, 1 P atom) and –12.5 ppm (doublet, 3 P atoms), which implies the presence of uncoordinated HPP3+ species. Based on these results, we assume two competing cycles I and II (Fig. 4). Initially, Fe(BF4)2·6 H2O / PP3 reacts with 1 equiv. FA to form [FeH(PP3)]+ (1). According to cycle I, coordination of FA to 1 and abstraction of H2 results in 4, or accordingly for the η2-coordinated HCO2, in species 4a. Subsequent β-hydride elimination from 4 in cycle I leads to species 5. From 5, CO2 elimination recovers the starting complex 1. Concerning cycle II, the coordination of formate to 1, leading to formation of species 2, is energetically favored. Through the following β-hydride elimination and protonation, accompanied with the subsequent release of CO2, species 3 is obtained. Subsequent elimination of H2 from 3 can recover the starting complex 1.

Fig. 4

Proposed mechanism for the selective iron-catalyzed hydrogen generation from FA with relative energies of complexes (kJ mol−1) calculated at the B3PW91/6-31G* and B3PW91/6-311G** level of theory. The underlined data indicates energies obtained for the second basis set used in the calculations. *, addition of polarization functions except for hydrogen. **, addition of polarization functions also for hydrogen.

Consistent with our results from kinetic measurements, the β-hydride elimination of CO2 from 2 is calculated to be the rate-determining step in cycle II. In cycle I, the rate-determining step is the β-hydride elimination too (from species 4 to 5). If instead of [FeH(PP3)]+, the complex [FeCl(PP3)]+ is introduced, the β-hydride elimination from the corresponding [FeCl(HCO2)(PP3)] species is calculated to be energetically unfavored (+424.95, +424.02 kJ mol–1), which is in agreement with experimental observations. Due to our pressure dependence measurements, the elimination of H2 plays an important role, which is well reflected by our calculations.

Supporting Online Material

www.sciencemag.org/cgi/content/full/333/6050/1733/DC1

Materials and Methods

SOM Text

Figs. S1 to S5

Tables S1 to S6

References (3438)

References and Notes

  1. Detailed experimental methods and computational details are available as supporting material on Science Online.
  2. Acknowledgments: This work has been supported by the state of Mecklenburg-Vorpommern, Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie, and Deutsche Forschungsgemeinschaft (Leibniz Prize). We thank A. Koch, S. Buchholz, C. Fischer, A. Lehmann, and A. Kammer for their excellent analytical and technical support. F.G. thanks the Fonds der Chemischen Industrie (FCI) for a Kekulé grant. We also thank École Polytechnique Fédérale de Lausanne and Swiss National Science Foundation for financial support. R.L. thanks the University of Rostock for continuous support. Parts of this work have been filed as a German patent application (DE 10 2011 007 661.1, submitted on 19 April 2011 by the Leibniz-Institute for Catalysis). All the data described in the paper are available in the supporting online material or from the authors.
View Abstract

Stay Connected to Science

Navigate This Article