Water splitting–biosynthetic system with CO2 reduction efficiencies exceeding photosynthesis

See allHide authors and affiliations

Science  03 Jun 2016:
Vol. 352, Issue 6290, pp. 1210-1213
DOI: 10.1126/science.aaf5039

Artificial photosynthesis steps up

Photosynthesis fixes CO2 from the air by using sunlight. Industrial mimics of photosynthesis seek to convert CO2 directly into biomass, fuels, or other useful products. Improving on a previous artificial photosynthesis design, Liu et al. combined the hydrogen-oxidizing bacterium Raistonia eutropha with a cobalt-phosphorus water-splitting catalyst. This biocompatible self-healing electrode circumvented the toxicity challenges of previous designs and allowed it to operate aerobically. When combined with solar photovoltaic cells, solar-to-chemical conversion rates should become nearly an order of magnitude more efficient than natural photosynthesis.

Science, this issue p. 1210


Artificial photosynthetic systems can store solar energy and chemically reduce CO2. We developed a hybrid water splitting–biosynthetic system based on a biocompatible Earth-abundant inorganic catalyst system to split water into molecular hydrogen and oxygen (H2 and O2) at low driving voltages. When grown in contact with these catalysts, Ralstonia eutropha consumed the produced H2 to synthesize biomass and fuels or chemical products from low CO2 concentration in the presence of O2. This scalable system has a CO2 reduction energy efficiency of ~50% when producing bacterial biomass and liquid fusel alcohols, scrubbing 180 grams of CO2 per kilowatt-hour of electricity. Coupling this hybrid device to existing photovoltaic systems would yield a CO2 reduction energy efficiency of ~10%, exceeding that of natural photosynthetic systems.

Sunlight and its renewable counterparts are abundant energy sources for a sustainable society (1, 2). Photosynthetic organisms harness solar radiation to build energy-rich organic molecules from water and CO2. Numerous energy conversion bottlenecks in natural systems limit the overall efficiency of photosynthesis (3). Most plants do not exceed 1%, and microalgae grown in bioreactors do not exceed 3%; however, efficiencies of 4% for plants and 5 to 7% for microalgae in bubble bioreactors may be achieved in the rapid (short-term) growth phase (3). Artificial photosynthetic solar-to-fuels cycles may occur at higher intrinsic efficiencies (47), but they typically terminate at hydrogen (8), with no process installed to complete the cycle via carbon fixation. This limitation may be overcome by interfacing H2-oxidizing autotrophic microorganisms to electrodes that generate hydrogen or reducing equivalents directly (914).

We recently developed a hybrid inorganic-biological system that uses the catalysts of the artificial leaf (15, 16) in combination with the bacterium Ralstonia eutropha (17) to drive an artificial photosynthetic process for carbon fixation into biomass and liquid fuels (18). In this system, water is split to oxygen by a cobalt phosphate (CoPi) catalyst and hydrogen is produced by a NiMoZn alloy at applied voltages of Eappl = 3.0 V. Because the maximum energy efficiency is limited by the value of Eappl relative to the thermodynamic potential for water splitting (= Eappl/1.23 V), a reduction in Eappl leads to biomass and liquid fuel efficiencies that surpass those of previous integrated bioelectrochemical systems and are commensurate with natural photosynthetic yields (18). However, reactive oxygen species (ROS) produced at the cathode were detrimental to cell growth. Because hydrogen peroxide (H2O2), as well as short-lived superoxide (O2) and hydroxyl radical (HO) species, are thermodynamically favored against H2 production at pH = 7, ROS production dominated at or below the potential to generate H2. When Eappl reached a sufficient overpotential to drive water splitting, H2 production to support cell growth outweighed the toxic effects of ROS (18). In addition, leaching of Ni from the NiMoZn alloy into solution inhibited microbial growth.

To develop a biocompatible catalyst system that is not toxic to the bacterium and lowers the overpotential for water splitting, we used a ROS-resistant cobalt-phosphorus (Co-P) alloy cathode (Fig. 1A, pathway 1). This alloy drives the hydrogen evolution reaction (HER) while the self-healing CoPi anode (19, 20) drives the oxygen evolution reaction (OER). The electrode pair works in concert to maintain extraneous cobalt ions at low concentration and to deliver low Eappl that splits water to generate H2 for R. eutropha, which supports CO2 reduction into complex organic molecules at high efficiency. The Co-P alloy, which is known to promote HER under alkaline solutions (21), exhibits high HER activity in water at neutral pH with minimal ROS production. X-ray photoelectron spectroscopy of Co-P thin films supports the elemental nature of the alloy (fig. S1), and energy-dispersive x-ray spectroscopy (fig. S2) establishes a phosphorus composition of 6 weight percent, which we have found to exhibit optimal HER activity in water at neutral pH with a faradaic efficiency of 99 ± 2% (fig. S3). Moreover, the activity of this Co-P alloy surpasses the activity of the Earth-abundant NiMoZn and stainless steel (SS) cathodes used previously (18) (Fig. 1B). At constant voltage, a stable HER current is maintained for at least 16 days (Fig. 1C). Negligible H2O2 is produced during HER (Fig. 1D), in contrast to that of simple metallic cathodes of Pt and SS.

Fig. 1 Active water-splitting catalyst pair with minimal biological toxicity.

(A) Reaction diagram and scanning electron microscopy images for Co-P alloy cathode and CoPi anode. The main water-splitting reaction is shown in black; the side reactions that yield toxicants are in red. Scale bars, 10 μm. (B) Current-voltage (I-V) characteristics of different HER catalysts (10 mV/s). (C) Stability of Co-P cathode, as demonstrated by 16-day chronoamperometry. (D) Assay of H2O2 accumulation for various cathodes combining with CoPi anode: yellow, Pt; blue, stainless steel (SS); red, Co-P alloy. Eappl = 2.2 V. Error bars denote SEM; n = 3. (E) Cyclic voltammetry of Co2+ and Ni2+ in the presence of phosphate (Pi). Metal concentrations are both 0.5 mM; 50 mV/s. The current for Ni2+ is magnified by a factor of 50.

The Co-P HER and CoPi OER catalysts work in synergy to form a biocompatible water-splitting system that salvages Co2+ cations leached from the electrodes (Fig. 1A, pathway 2). In the cyclic voltammogram of Co2+ in the phosphate buffer (pH = 7) (Fig. 1E), a pre-wave to the catalytic water-splitting current corresponds to the oxidation of Co2+ to Co3+, which drives deposition of the catalyst. The CoPi catalyst is also known to exhibit a deposition rate that is linearly proportional to Co2+ concentration (22). The self-healing property of CoPi is derived from this interplay of the potential at which OER occurs versus the potential at which the catalyst deposits (20). In concert, the Co-P and CoPi catalysts preserve extremely low concentrations of Co2+ in solution through activity derived from the self-healing process. Inductively coupled plasma mass spectrometry (ICP-MS) analysis of a Co-P|CoPi catalyst system (Eappl = 2.2 V) (23) reveals submicromolar levels of Co2+ in solution after 24 hours. This concentration of Co2+ (0.32 ± 0.06 μM) is well below the concentration of Co2+ (half-maximal inhibitory concentration IC50 ≈ 25 μM) that is toxic to R. eutropha (fig. S4). When diffusion between the two electrodes is impeded by a porous glass frit, Co2+ concentrations rise to ~50 μM. We note that for the NiMoZn cathode, Ni2+ concentrations are not regulated by self-healing, as NiPi cannot form from Pi (24), and the deposition to NiOx occurs at >1.5 V versus normal hydrogen electrode (NHE) (Fig. 1E; see fig. S5 for comparison with potentials of relevant redox processes).

Interfacing the biocompatible Co-P|CoPi water-splitting catalysts with R. eutropha results in a system capable of CO2 fixation. The CoPi catalyst was deposited on a high–surface area carbon cloth as the electrode support (Fig. 1A and fig. S6), resulting in high currents (fig. S7) and a faradaic efficiency of 96 ± 4% (fig. S8). CO2 reduction proceeded under a constant voltage within a batch reactor (fig. S9), which was half-filled with a solution containing only inorganic salts (mostly phosphate) and trace metal supplements (23).

The CoPi|Co-P|R. eutropha hybrid system can store more than half its input energy as products of CO2 fixation at low Eappl (Fig. 2A and table S1). Entries 1, 2, 3, and 5 show that ηelec increases with decreasing Eappl under 100% CO2 until Eappl < 2.0 V. Below Eappl = 2.0 V (entry 8), a higher salt concentration (108 mM phosphate buffer) is required to facilitate mass transport and attendant current (fig. S10). However, high salt concentrations are undesirable for R. eutropha metabolism. Thus, a concentration of 36 mM phosphate and Eappl = 2.0 V resulted in optimal ηelec; the highest ηelec achieved for biomass production was 54 ± 4% (entry 5, n = 4) over a duration of 6 days. Our CO2 reduction efficiency from H2 is comparable to the highest demonstrated by R. eutropha during H2 fermentation (25). This biomass yield is equivalent to assimilating ~4.1 mol (180 g) of CO2 captured at the cost of 1 kWh of electricity. The amount of captured CO2 is 10% of the amount caught by amine-based carbon capture and storage (~2000 g at the cost of 1 kWh) (26), whose processed product cannot be used as fuel. Enlarging the batch reactor volume by a factor of 10 did not perturb the efficiency (entries 4 and 6), indicating that the system is scalable and the reactor volume does not pose immediate limits. Note that ηelec under air (400 ppm CO2) is 20 ± 3% (entry 7, n = 3), which is lower than for pure CO2 by a factor of only 2.7, although the partial pressure of CO2 is reduced by a factor of 2500. This indicates that CO2 is not a limiting reagent (see below). The ~20% of ηelec for biomass is equivalent to assimilating ~1.5 mol of CO2 captured from about 85,200 liters of air at ambient condition with the cost of 1 kWh of electricity.

Fig. 2 Energy efficiencies ηelec and kinetics of the hybrid CO2 reduction device.

(A) ηelec values for the production of biomass and chemicals at different values of Eappl and various configurations (table S1). Solid bars are 5- to 6-day averages; hatched bars are 24-hour maxima. Error bars denote SEM; n ≥ 3. (B and C) Optical density at 600 nm (OD600; indicator of biomass accumulation) and amounts of electric charges that were passed, plotted versus the duration of experiments with 100% CO2 (B) and air (C) in the headspace at 1 atm pressure. Eappl = 2.0 V. Error bars denote SEM; n = 4 for (B) and n = 3 for (C). (D) A microbial growth model predicts linear correlation between electric charges and biomass accumulation, when the H2 generation rate by water splitting (I/2FV) is smaller than the maximum rate of H2 consumption by active biomass (rmaxXa) (23) (fig. S12). Dashed line indicates Michaelis constant of hydrogenase for H2. (E) Real-time monitor of biomass accumulation under “day”/“night” cycle test.

We also isolated a ROS-resistant variant of R. eutropha from one SS|CoPi water-splitting reactor after 11 consecutive days of operation (Eappl = 2.3 V) with a H2O2 generation rate of ~ 0.6 μM/min. Genome sequencing found several mutations between the strain (BC4) and the wild type (H16) (table S2). In the presence of paraquat as a ROS inducer (27), the IC50 of paraquat for BC4 is almost one order of magnitude higher than that of the wild type (fig. S11). There is no obvious benefit of the BC4 strain with regard to ηelec (table S1), further confirming the absence of ROS in our system (see above). Nonetheless, BC4 should find great utility for achieving high ηelec in systems where ROS is problematic.

We found that biomass accumulation scales linearly with the amount of charge passed under pure CO2 (Fig. 2B) or ambient CO2 levels (Fig. 2C). The linear growth is accounted for by a model that combines governing equations for H2 generation from water splitting and biomass accumulation from carbon fixation (23). The model predicts a linear correlation between biomass and charge passed after an induction period of low population density of bacteria and high H2 concentration (Fig. 2D and fig. S12), which is consistent with the data shown in Fig. 2, B and C, where the induction period is too short to be observed. Gas chromatography measurements revealed a H2 concentration in the reactor headspace of 0.19 ± 0.04% (n = 3) in 100% CO2 and 0.10 ± 0.05% (n = 3) in air, corresponding to 1.5 ± 0.3 μM and 0.8 ± 0.4 μM, respectively, in water. These concentrations of H2 are well below the Michaelis constant of ~6 μM for membrane-bound hydrogenases in R. eutropha (28), which suggests that H2 is facilely consumed by R. eutropha. Moreover, similar linear growth conditions for both pure and ambient CO2 atmospheres provide evidence that H2 oxidation rather than CO2 reduction is rate-limiting for biosynthesis. Lastly, R. eutropha halted growth during “night” cycles and continued CO2 reduction 12 hours later upon resumption of the water-splitting reaction (Fig. 2E), confirming the intrinsic dependence of R. eutropha on H2 generation. These data also reveal that the CoPi|Co-P|R. eutropha hybrid system is compatible with the intermittent nature of a solar energy source. Direct CO2 reduction from air highlights the relatively high affinity of R. eutropha for CO2 at low pressures and at high O2 concentrations, in contrast to results reported for synthetic catalysts (29), individual enzymes (30, 31), and strictly anaerobic organisms such as acetogens and methanogens (1114) (table S3).

Metabolic engineering of R. eutropha enables the renewable production of an array of fuels and chemical products (17). When R. eutropha confronts nutrient constraints coupled with carbon excess, the biosynthesis of poly(3-hydroxybutyrate) (PHB) is triggered in the wild-type H16 strain as an internal carbon storage pathway (17). As such, digestion is necessary for PHB collection (23). Under a constant rate of water splitting, PHB synthesis was not manifest until nitrogen became limiting (~2 days), as indicated by the cessation of biomass accumulation (Fig. 3A) as well as the ηelec measured every 24 hours (Fig. 3B and fig. S13). With a titer of ~700 mg/liter, the 6-day average for PHB synthesis was ηelec = 36 ± 3% (Fig. 2A, entry 9) with a 24-hour maximum of ηelec = 42 ± 2% (n = 3) (Fig. 3B). In engineered strains (32, 33), this PHB pathway could be modified to excrete the fusel alcohols isopropanol (C3), isobutanol (C4), and 3-methyl-1-butanol (C5), which possess energy densities of 24, 28, and 31 MJ/liter, respectively. The culture supernatant was then analyzed to quantify the secreted alcohols (23). The accumulation of these liquid fuels followed trends similar to those observed for PHB synthesis. As shown in Fig. 3, C and E, biomass production reached a plateau while isopropanol titers grew to ~600 mg/liter and C4 + C5 alcohol titers grew to ~220 mg/liter. An engineered R. eutropha strain produced isopropanol with a 6-day average ηelec = 31 ± 4% (Fig. 2A, entry 10) and a 24-hour maximum of ηelec = 39 ± 2% (n = 4) (Fig. 3D); a strain engineered to produce C4 + C5 alcohols averaged a 6-day ηelec = 16 ± 2% (Fig. 2A, entry 11) with a 24-hour maximum of ηelec = 27 ± 4% (n = 3) (Fig. 3F). The achieved titers are higher than previous reported values, and ηelec values have increased by a factor of at least 20 to 50 (10, 18). R. eutropha has demonstrated tolerance toward isopropanol (fig. S14), allowing for enriched product concentrations under extended operation.

Fig. 3 Efficient synthesis of selectively produced chemicals from CO2 and water.

(A to F) PHB [(A) and (B)], isopropanol (C3) [(C) and (D)], and C4 and C5 alcohols [(E) and (F)] were selectively produced from the hybrid device. In (A), (C), and (E), the OD600 values, concentrations of selective chemicals, and charges passed through the electrodes are plotted versus the duration of experiments. Shown in (B), (D), and (F) are averaged ηelec values for different products, measured at 24-hour intervals. Also shown are overall ηelec values combining biomass and chemical formation. The ηelec values for biomass, defined as intracellular organics excluding PHB, have been corrected to exclude the PHB interference in (B) (23) (see fig. S13 for values before correction). Error bars denote SEM; n = 3.

Our combined catalyst design mitigates biotoxicity at a systems level, allowing water-splitting catalysis to be interfaced with engineered organisms to realize high CO2 reduction efficiencies that exceed natural photosynthetic systems. Because Eappl required for water splitting is low (1.8 to 2.0 V), high ηelec values are achieved that translate directly to high solar-to-chemical efficiencies (ηSCE) when coupled to a typical solar-to-electricity device (ηSCE = ηsolar × ηelec). For a photovoltaic device of ηsolar = 18%, the Co-P|CoPi|R. eutropha hybrid system can achieve ηSCE = 9.7% for biomass, 7.6% for bioplastic, and 7.1% for fusel alcohols. This approach allows for the development of artificial photosynthesis with efficiencies well beyond that of natural photosynthesis, thus providing a platform for the distributed solar production of chemicals.

Supplementary Materials


Tables S1 to S3

Figs. S1 to S14

References (3449)

References and Notes

  1. See supplementary materials on Science Online.
Acknowledgments: We thank N. Li for ICP-MS measurement and reagents, and J. Torella, C. Myhrvold, C. Lemon, and M. Huynh for helpful discussions. C.L. acknowledges X. Ling at Nanyang Technological University. Supported by a Lee Kuan Yew Postdoctoral Fellowship (C.L.), a predoctoral fellowship from the NSF Graduate Research Fellowships Program (B.C.C.), Office of Naval Research Multidisciplinary University Research Initiative award N00014-11-1-0725 (P.A.S.), Air Force Office of Scientific Research grant FA9550-09-1-0689 (D.G.N.), the Wyss Institute for Biologically Inspired Engineering (P.A.S.), and the Harvard University Climate Change Solutions Fund. This work was performed under the First 100 W Program at Harvard University. C.L., B.C.C., M.Z., P.A.S., and D.G.N. are inventors on patent applications (62/218,131) filed by Harvard University and Harvard Medical School related to the technology described in this paper. The genome sequences are accessible in the NCBI SRA database under accession number SRP073266.
View Abstract

Navigate This Article