Edited by Richard Eisenberg, University of Rochester, Rochester, New York, and approved July 24, 2015 (received for review April 26, 2015)
Natural photosynthesis, a process of solar-to-chemical conversion, uses light, water, and carbon dioxide to generate the chemical products needed to sustain life. Here we report a strategy inspired by photosynthesis in which compatible inorganic and biological components are used to transform light, water, and carbon dioxide to the value-added product methane. Specifically, this solar-to-chemical conversion platform interfaces photoactive inorganic materials that produce hydrogen from water and sunlight with microorganisms that consume this sustainably derived hydrogen to drive the transformation of carbon dioxide to methane with high efficiency. This system establishes a starting point for a broader materials biology approach to the synthesis of more complex chemical products from carbon dioxide and water.
Natural photosynthesis harnesses solar energy to convert CO2 and water to value-added chemical products for sustaining life. We present a hybrid bioinorganic approach to solar-to-chemical conversion in which sustainable electrical and/or solar input drives production of hydrogen from water splitting using biocompatible inorganic catalysts. The hydrogen is then used by living cells as a source of reducing equivalents for conversion of CO2 to the value-added chemical product methane. Using platinum or an earth-abundant substitute, α-NiS, as biocompatible hydrogen evolution reaction (HER) electrocatalysts and Methanosarcina barkeri as a biocatalyst for CO2 fixation, we demonstrate robust and efficient electrochemical CO2 to CH4 conversion at up to 86% overall Faradaic efficiency for ≥7 d. Introduction of indium phosphide photocathodes and titanium dioxide photoanodes affords a fully solar-driven system for methane generation from water and CO2, establishing that compatible inorganic and biological components can synergistically couple light-harvesting and catalytic functions for solar-to-chemical conversion.
Methods for the sustainable conversion of carbon dioxide to value-added chemical products are of technological and societal importance (1⇓–3). Elegant advances in traditional approaches to CO2 reduction driven by electrical and/or solar inputs using homogeneous (4⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓–16), heterogeneous (17⇓⇓⇓⇓⇓⇓⇓⇓–26), and biological (7, 27⇓⇓⇓–31) catalysts point out key challenges in this area, namely (i) the chemoselective conversion of CO2 to a single product while minimizing the competitive reduction of protons to hydrogen, (ii) long-term stability under environmentally friendly aqueous conditions, and (iii) unassisted light-driven CO2 reduction that does not require external electrical bias and/or sacrificial chemical quenchers. Indeed, synthetic homogeneous and heterogeneous CO2 catalysts are often limited by product selectivity and/or aqueous compatibility, whereas enzymes show exquisite specificity but are generally less robust outside of their protective cellular environment. In addition, the conversion of electrosynthetic systems to photosynthetic ones is nontrivial owing to the complexities of effectively integrating components of light capture with bond-making and bond-breaking chemistry.
Inspired by the process of natural photosynthesis in which light-harvesting, charge-transfer, and catalytic functions are integrated to achieve solar-driven CO2 fixation (32⇓⇓–35), we have initiated a program in solar-to-chemical conversion to harness the strengths inherent to both inorganic materials chemistry and biology (36). As shown in Fig. 1, our strategy to drive synthesis with sustainable electrical and/or solar energy input (37) interfaces a biocompatible photo(electro)chemical hydrogen evolution reaction (HER) catalyst with a microorganism that uses this sustainably generated hydrogen as an electron donor for CO2 reduction. Important previous reports have shown the feasibility of electrosynthesis (38⇓⇓⇓–42) but have not yet established solar-driven processes. We selected methane as an initial target for this approach owing to the ease of product separation, the potential for integration into existing infrastructures for the delivery and use of natural gas (of which CH4 is the principle component), and the fact that direct conversion of CO2 to CH4 with synthetic catalysts remains a formidable challenge due to large overpotentials and poor CH4/H2 selectivity. Two of the most active and selective direct electrocatalysts for CO2 to CH4 conversion reported to date produce methane with 61% (43) and 76% (44) Faradaic efficiencies, but require overpotentials of η = 1.28 V and η = 1.52 V, respectively. Promising advances in photothermal reduction of CO2 to CH4 also have been recently reported (45). In comparison with fully inorganic catalysts, a distinct conceptual advantage of this hybrid materials biology approach, where the materials component performs water splitting to generate hydrogen and the biological component uses these reducing equivalents for CO2 fixation, is that one can leverage the fact that biological catalysts operate at near thermodynamic potential (46). As such, the only overpotential involved is associated with hydrogen evolution from water, a more facile reaction to catalyze via sustainable electrochemical and photochemical means compared with CO2 reduction. Coupled with the diversity of potential chemical products available via synthetic biology, the marriage between artificial and natural platforms can create opportunities to develop catalyst systems with enhanced function over the individual parts in isolation.
General scheme depicting a hybrid bioinorganic approach to solar-to-chemical conversion. Sustainable energy inputs in the form of electrical potential or light can be used to generate hydrogen from water using inorganic HER catalysts; biological hydrogen-driven CO2 fixation can subsequently generate value-added products such as methane. This materials biology interface can be generalized to other chemical intermediates and end products by mixing and matching different compatible inorganic and biological components.
In developing hybrid bioinorganic platforms for solar-to-chemical conversion of CO2, we drew inspiration from both tandem organometallic–microbial systems (47, 48), in which products of microbial metabolism are further transformed by organometallic catalysts, as well as biological electrosynthesis, in which organisms accept reducing equivalents from an electrode either in the form of soluble electron carriers (for example, H2 or formate) (41, 49, 50) or via direct electron transfer (36, 51⇓–53). Engineered strains of Ralstonia eutropha have been used for the aerobic production of isobutanol and 3-methyl-1-butanol (41), and isopropanol (42). However, owing to the oxygen requirements of this organism and the relative inefficiency of its carbon fixation pathways (54), product titers and production efficiencies are relatively modest, and generation of reactive oxygen species is a serious concern. In addition, to our knowledge, no photosynthetic systems of this type have been reported. As such, we turned our attention to the use of a pure culture of Methanosarcina barkeri, an obligately anaerobic archaeon that fuels its metabolism via the 8-proton, 8-electron reduction of CO2 to CH4 (55). Prior studies have reported methanogenic electrosynthesis (51, 53, 56); however, a fully light-driven system remains to be realized. Additionally, mixed cultures and multiple possible sources of reducing equivalents have complicated Faradaic efficiency measurements in previous studies (51, 53, 56). Through the design of our hybrid system, we sought to surmount some of these aforementioned challenges.
Here we report an integrated bioinorganic catalyst platform for solar-to-chemical CO2 conversion using sustainable inorganic hydrogen generators in conjunction with CO2-fixing archaea. Under electrosynthetic conditions with a platinum cathode, a culture of M. barkeri shows chemoselective conversion of CO2 to CH4 with high Faradaic efficiencies (up to 86%) and low overpotential (η = 360 mV). The system is also capable of high yield production, cumulatively generating 110 mL (4.3 mmol) of methane over 7 d. Isotope labeling with 13CO2 establishes that CH4 is uniquely derived from CO2 for cultures in both rich media and minimal, carbon-free media. Replacement of Pt with an earth-abundant α-NiS electrocatalyst allows for CH4 generation at similar titers. Moreover, using a photoactive silicon cathode reduces the overpotential to 175 mV upon irradiation with 740-nm light. Unassisted light-driven methane generation was achieved using tandem solar absorption by a photoactive n-TiO2 anode and p-InP cathode assembly. Taken together, the results demonstrate the feasibility of combining compatible inorganic and biological systems to achieve solar-to-chemical conversion from light, H2O, and CO2, affording a starting point for the realization of sustainable fixation of CO2 to value-added molecules.
Careful organism selection is critical to the successful realization of an integrated bioinorganic system. The autotrophic obligate anaerobe M. barkeri (55) is amenable to integration with inorganic catalysts for a variety of reasons. M. barkeri can use H2 as a source of reducing equivalents for the reduction of CO2 to CH4; the cathode of a water-splitting device could serve as a potential source of this H2. Owing to the anaerobic metabolism of the organism, oxygen is not required at the cathode, thereby improving Faradaic efficiency for the product of interest, simplifying gas delivery to the culture, and preventing generation of potentially harmful reactive oxygen species. Furthermore, CH4 is generated with high efficiency as a byproduct of normal metabolism. Finally, M. barkeri requires no added sources of reduced carbon and can produce CH4 in minimal media containing only supplemental vitamins and minerals.
Initial experiments were performed using a platinum cathode to electrochemically generate H2, which was subsequently used in situ by M. barkeri to reduce CO2 to CH4. Fig. 2A shows a general schematic of the gas-tight, two-compartment electrochemical cell that was specially fabricated for batch-mode electrolysis and subsequent headspace analysis by gas chromatography (GC). (See Fig. S1 for electrolysis cell photo.) Separation of the cathodic and anodic chambers with an ion-permeable membrane prevented any noticeable diffusion of O2 into the culture. After inoculation of the cathodic chamber with M. barkeri (130-mL final volume, OD600 nm = 0.35) and saturation of the carbon-free catholyte with pure CO2, galvanostatic electrolysis was performed at a current of 2.5 mA (j = 0.29 mA/cm2, η = 360 mV) for 11.5 h, followed by 0.5 h of sampling. (See the Supporting Information for calculation of overpotentials.) The headspace was analyzed by GC and replaced by sparging with fresh CO2 before restarting the electrolysis. (See Fig. S2 for GC calibration curves.) Methane production was linear under these conditions and cumulatively resulted in 16.8 ± 0.6 mL CH4 (0.660 ± 0.024 mmol) over 3 d (all volumes reported in this work assume 1-atm pressure and a temperature of 310 K) with an average Faradaic efficiency of 81 ± 3% (n = 3). (All errors represent standard deviation unless otherwise noted.) (Fig. 2B). Electrolysis was subsequently conducted in rich media containing yeast extract and casitone; variation of the applied current from 1 mA (j = 0.12 mA/cm2) to 7.5 mA (j = 0.88 mA/cm2) shows that CH4 generation is proportional to applied current, suggesting that the system is operating in a hydrogen-limited regime (Fig. 2C). The highest Faradaic efficiency, 86 ± 3%, was observed with an applied current of 2.5 mA (j = 0.29 mA/cm2). A slight decrease in Faradaic efficiency was observed at higher current densities. We speculate that increased rates of HER cause H2 to be less effectively delivered to the culture solution with loss to the headspace.
Electrocatalytic reduction of carbon dioxide to methane with a hybrid platinum/M. barkeri platform. (A) Schematic of generalized electrolytic setup showing in situ generation of hydrogen at the cathode followed by hydrogen-driven reduction of carbon dioxide to methane by the M. barkeri biocatalyst. (B) Cumulative methane generation and associated average Faradaic efficiency in minimal media. (C) Cumulative methane generation and associated average Faradaic efficiencies at various currents in rich media. (D) Cumulative long-term methane generation and associated average Faradaic efficiency in rich media. (E and F) High-resolution mass spectrometry of headspace gases after electrolysis under an atmosphere of (E) 12CO2 and (F) 13CO2 in rich media. Error bars represent SD with n = 3 independent experiments in all cases.
Electrolysis cell photographs. The electrolysis cell as described above (see Design of Photo/Electrochemical Cells) viewed from different angles. Important design features are highlighted: (a) anode chamber with platinum anode visible, (b) cathode chamber with platinum cathode and Ag/AgCl reference electrode visible, (c) electrical connection to cathode, (d) electrical connection to reference electrode, (e) CO2 line for solution sparging, (f) GC connection for headspace sampling, and (g) resealable septum for manual gas injection/sampling.
GC calibration curve for H2 and CH4 quantification. Hydrogen and methane calibration curves were generated by injecting known amounts of H2 and CH4 into an electrochemical cell prepared identically to those used in gas measurement experiments. Headspace samples were introduced onto a GC by first evacuating the sample loop (1 mL) and subsequently opening it to the electrochemical cell. Peak areas determined by GC are reported as a ratio compared with the He internal standard (1 mL). Data are mean ± SD (n = 3).
To test the limits of CH4 production of the system, the organisms were electrolyzed at 7.5 mA for 7 d (Fig. 2D). The headspace of the cathodic chamber was sampled daily and subsequently exchanged with fresh CO2. During each 24-h period (23 h electrolysis and 1 h sampling), 15.6 ± 0.7 mL CH4 (0.613 ± 0.028 mmol) were produced; this value corresponds to cumulative methane production of 109 ± 5 mL (4.28 ± 0.20 mmol) over 7 d at an average Faradaic efficiency of 75 ± 4% (n = 3). Moreover, the hybrid bioinorganic system showed no loss in efficiency over the course of the experiments, which were terminated due to time constraints rather than a decrease in performance. Taken together, these data establish that there are no viability concerns on the timescale tested and presage the possibility of extended operation with a single inoculation of biomass.
A series of isotopic labeling experiments was conducted to show that the observed methane was derived from carbon dioxide. Electrolysis cells containing nitrogen-sparged rich media were inoculated with M. barkeri. The headspace of one cell was filled with 12CO2 and another with 13CO2, and both were subjected to galvanostatic electrolysis at a current of 2.5 mA. High-resolution GC-mass spectrometry was used to show that only 13CH4 was observed when electrolyzed under an atmosphere of 13CO2 (Fig. 2F), whereas only 12CH4 was observed in the 12CO2 control experiment (Fig. 2E). Based on the limit of detection and amount of methane generated, <4% of the observed methane could potentially come from sources other than CO2. Similarly, only 13CH4 was observed when the electrolysis was performed in minimal media (Fig. S3). Based on the limit of detection and amount of methane generated, <7.5% of the observed methane could potentially come from sources other than CO2 under these conditions.
High-resolution mass spectra of headspace gases after electrolysis under 12CO2 or 13CO2 atmospheres. Experimental setup is as described. After setup under N2 in the specified media, the headspace of the electrolysis cell was replaced with 12CO2 or 13CO2. He (1 mL) was added as an internal standard for final methane quantification. Each experiment was electrolyzed at 2.5 mA using a platinum cathode and anode. After 3 d, the headspace was analyzed using high-resolution mass spectrometry, with representative mass spectra presented below. (A) 12CO2 headspace, carbon-free media, (B) 13CO2 headspace, carbon-free media, (C) abundances of parent ions and fragments in carbon-free media, and (D) abundances of parent ions and fragments in rich media (see Fig. 2 E and F for corresponding spectra).
After demonstrating CO2-to-CH4 conversion by M. barkeri during in situ electrolysis using a platinum HER cathode, we sought to replace the precious metal catalyst with an earth-abundant alternative. Several HER catalysts containing only first-row transition metals have been recently reported (57⇓⇓–60); among these examples, metal chalcogenides have featured prominently (61⇓⇓⇓⇓⇓–67). In addition to low overpotential and long-term stability, another requirement for a catalyst in a biomaterials hybrid system is that it must operate in aqueous media within a biologically relevant pH range (pH 5–8) and be nontoxic to the organism. Relatively few published examples meet all of these criteria. As such, we developed nanoparticulate α-NiS as a biocompatible HER catalyst for integration with M. barkeri cultures.
Nanoparticulate nickel sulfide was prepared by microwave irradiation (250 °C, 30 min) of an aqueous solution of nickel chloride, thioacetamide, and ammonium hydroxide (Fig. 3A). The resulting black powder was rinsed with water and isopropanol and dried under vacuum overnight. The newly synthesized catalyst was stored in a capped vial in air and retained activity for several months. Characterization by powder X-ray diffraction (pXRD) confirmed that the product is crystalline and identified the primary phase as α-NiS (Joint Committee on Powder Diffraction Standards number 77–1624) (Fig. 3B). Transmission electron microscopy (TEM) images show polydisperse hexagonal particles of 20–100-nm diameter (Fig. S4). The nickel-to-sulfur ratio was verified to be 1:1 by energy-dispersive X-ray (EDX) spectroscopy based on the nickel L and sulfur K peaks (Fig. 3C), a result which was further corroborated by inductively coupled plasma-optical emission spectroscopy (ICP-OES) measurements. High-resolution TEM images taken along the [010] axis confirm the assignment of single-crystalline α-NiS based on the (100) and (001) lattice planes (Fig. 3D). Thermogravimetric analysis (TGA) shows negligible loss of mass until 600 °C, suggesting the absence of surface-bound organics (Fig. S5).
Synthesis and characterization of a biocompatible, earth-abundant HER catalyst for use in hybrid bioinorganic systems. (A) Synthesis of α-NiS nanoparticles. (B) pXRD pattern of α-NiS. (C) EDX spectrum of α-NiS. (D) HRTEM images of α-NiS. (E) Polarization curve (RDE) of α-NiS on glassy carbon compared with glassy carbon background, 1 M KPi (pH 7), scan rate 5 mV/s, rotation speed 1,500 rpm. (F) Corresponding Tafel plot. (G) Cumulative long-term methane generation and associated average Faradaic efficiency using an α-NiS/carbon cloth cathode is comparable to results obtained with platinum cathodes. Error bars represent SD with n = 3 independent experiments.
TEM characterization of α-NiS nanoparticles. TEM image of α-NiS nanoparticles showing polydisperse hexagonal particles 20–100 nm in diameter.
TGA of α-NiS catalyst. The absence of a significant loss in mass until ∼450 °C indicates a lack of surface-bound ligands on the α-NiS nanoparticles. Experiment was performed under an atmosphere of N2 to prevent sample oxidation.
Electrocatalytic properties of the α-NiS particles were evaluated using rotating disk electrochemistry and chronopotentiometry. As shown in the polarization curve (Fig. 3E), crystalline α-NiS is an active HER catalyst in pH 7 potassium phosphate buffer and achieves current densities of 1 and 5 mA/cm2 at overpotentials of η = 275 and 350 mV, respectively. The observed Tafel slope of 111 mV/dec is similar to the predicted 118 mV/dec for a process in which the Volmer step (formation of an adsorbed H intermediate) is rate-determining (68). The exchange current density, j0, is 3.5 × 10−2 mA/cm2 based on extrapolation of the Tafel plot (Fig. 3F). These parameters compare favorably to other heterogeneous, first-row transition metal HER catalysts, especially given the ease of synthesis (Table S1).
Comparison of various first-row transition metal HER catalysts at neutral pH
A suspension of α-NiS powder in 3:1 ethanol:water with Nafion binder was deposited on conductive carbon cloth to fabricate larger-scale electrodes for hydrogen generation and long-term stability measurements. Galvanostatic experiments at a current of 2.5 mA were run for 23 h; the Faradaic efficiency for hydrogen generation was found to be 95 ± 4% (n = 3) (Fig. S6). As shown in Fig. S7B, the potential applied during these experiments does not significantly decay over time, indicating that the catalyst is stable under such conditions.
FEs for H2 generation for electrolysis, photoassisted electrolysis, and unassisted water splitting with Pt, α-NiS/C, n+/p-Si/NiMo, and p-InP/Pt cathodes. Experimental setup is as described, except that no M. barkeri was inoculated into the electrolysis cell. A galvanostatic experiment was conducted using Pt, α-NiS/C, or illuminated n+/p-Si/NiMo by passing a current of 2.5 mA for 23 h (n = 3 independent experiments). The p-InP/Pt cathodes were illuminated at 2.2 W/cm2 for 1–6 h (n = 4 independent experiments). H2 in the headspace was quantified by GC using He as an internal standard. FEs for H2 were calculated according to the following equation:
Representative galvanostatic and potentiostatic traces for (photo)electrochemical and photochemical experiments with a M. barkeri culture. For each of the galvanostatic experiments shown above (A–C), a current of 2.5 mA was passed for 23 h. A standard three-electrode cell setup was used, where the anode is Pt mesh, and the cathode is as listed below. For the potentiostatic experiment (D), a two-electrode cell setup was used in which the counter and reference leads were shorted on the illuminated n-TiO2 anode, and the working lead was attached to the illuminated p-InP/Pt cathode. The full cell potential was maintained at 0 V for 69 h. (A) Pt cathode, (B) α-NiS/C cathode, (C) n+/p-Si/NiMo cathode illuminated at λ = 740 nm, and (D) n-TiO2 and p-InP/Pt linked assembly illuminated with full spectrum and λ > 455 nm, respectively.
Having established the efficacy and stability of α-NiS as a catalyst for HER at biologically relevant pH, we sought to interface it with a M. barkeri culture for electrochemical methane production. A piece of conductive carbon cloth with 1.4 mg deposited α-NiS catalyst was used as an earth-abundant replacement for platinum. In a galvanostatic experiment at 2.5 mA (javg = 0.28 mA/cm2, η = 695 mV), 5.1 ± 0.2 mL CH4 (0.20 ± 0.01 mmol) were generated in 24 h, cumulatively generating 15.4 ± 0.8 mL CH4 (0.605 ± 0.031 mmol) over 3 d with an average Faradaic efficiency of 73 ± 5% (n = 3) (Fig. S8). These values are comparable to those obtained for the corresponding platinum experiments. When the current was increased to 7.5 mA (javg = 0.83 mA/cm2), experiments using α-NiS/C were virtually indistinguishable from Pt in terms of daily and cumulative methane, producing 108 ± 4 mL CH4 (4.24 ± 0.16 mmol) over 7 d with an average Faradaic efficiency of 74 ± 2% (n = 3) (Fig. 3G).
Galvanostatic experiment (2.5 mA) with α-NiS/C cathode. Experimental setup and electrode preparation is as described. Cumulative methane generation and associated average FE in rich media are shown compared with a nonelectrolyzed control. Error bars represent SD with n = 3 independent experiments.
Having performed electrochemical conversion of CO2 to CH4, we next sought to develop a photoelectrochemical system in which a portion of the potential required for water splitting is contributed by light. Indeed, transforming electrochemical systems to photochemical ones remains challenging. To achieve this goal, we sought to use semiconductor photocathodes coated with a thin-film HER catalyst. The overall performance of such an assembly is determined by a tradeoff between efficient catalysis and light absorption: thicker films confer superior electrocatalytic activity and stability, whereas thinner and more transparent films allow for greater photon capture. Although α-NiS proved effective as an earth-abundant HER electrocatalyst, previously published work from our laboratories on the electrodeposition of a related cobalt sulfide film on planar n+/p-Si showed that the current density under illumination drastically diminished as the thickness of the film increased (62). For this reason, we chose to use a nickel–molybdenum alloy, a previously characterized earth-abundant HER catalyst, which has shown favorable performance under photocatalytic conditions despite having a slightly larger overpotential than platinum (69). Photoactive cathodes were prepared by sputtering a thin layer of Ni–Mo alloy atop TiO2-passivated n+/p-Si (Fig. 4A). To confirm hydrogen evolution, abiotic galvanostatic experiments at a current of 2.5 mA (javg = 0.36 mA/cm2) were run for 23 h while illuminating the cathode with 740-nm light (20 mW/cm2); Faradaic efficiencies for H2 were 103 ± 3% (Fig. S6). A light toxicity control experiment with M. barkeri showed that methane generation is not affected by illumination at this wavelength (Fig. S9A). Biological galvanostatic electrolysis at 2.5 mA with the photocathode (740-nm illumination) generated 17.6 ± 2.1 mL CH4 (0.692 ± 0.083 mmol) with a Faradaic efficiency of 82 ± 10% (n = 3) with only 175-mV overpotential (Fig. 4B).
Photoelectrochemical and unassisted solar-driven conversion of carbon dioxide to methane with hybrid bioinorganic catalysts. (A) Schematic of electrode setup for photoelectrochemical water splitting. (B) Cumulative photoelectrochemically derived methane and associated average Faradaic efficiency with an n+/p-Si/NiMo photocathode illuminated with 740-nm light in rich media. Error bars represent SD with n = 3 independent experiments. (C) Schematic of electrode setup for unassisted solar water splitting. (D) Average methane produced after 3 d under illuminated (n = 2) and nonilluminated (n = 3) conditions using an n-TiO2 photoanode and a p-InP/Pt photocathode with no applied potential.
Evaluation of microbial light toxicity. Experimental setup is as described. (A) Illuminated (λ = 740 nm) and nonilluminated galvanostatic electrolyses (2.5 mA, Pt cathode) show comparable methane production, indicating that illumination with this wavelength does not affect methane production over the course of 72 h (n = 1). (B) Illuminated (full-spectrum light) and nonilluminated galvanostatic electrolyses (2.5 mA, Pt cathode) were performed. When n-TiO2/FTO was used as the only light filter, a lower than expected FE for methane was observed in the illuminated experiment. Upon addition of a 455-nm filter after the n-TiO2/FTO, FEs for methane were similar between illuminated and nonilluminated experiments (n = 1).
Finally, we sought to construct a fully light-driven hybrid bioinorganic system for CO2-to-CH4 conversion through the use of a tandem semiconductor assembly. In this setup, full-spectrum light first impinges on a large bandgap anode [nanowire n-TiO2 on fluorine-doped tin oxide (FTO)], where water oxidation generates oxygen (37). The filtered, lower-energy light subsequently illuminates a smaller bandgap cathode (p-InP coated with Pt), where water reduction generates hydrogen (Fig. 4C). The electrochemical cell design described above required slight modifications for unassisted photochemical experiments: a 1-in-diameter quartz window was added to the anodic chamber to prevent initial filtering of the full-spectrum light by glass, and the anode and cathode compartments were separated by an anion exchange membrane to minimize pH changes. This linked two-electrode assembly generates nonzero photocurrent under illumination (iavg = 0.17 mA, javg = 0.057 mA/cm2 during the first hour) (Fig. S7D). The Faradaic efficiency for hydrogen generation in abiotic experiments is 100 ± 8% (Fig. S6) (n = 4). Before introduction of M. barkeri into such a system, galvanostatic control electrolyses of a M. barkeri culture were performed using Pt electrodes under illuminated and nonilluminated conditions. When only the n-TiO2/FTO anode was used as a light filter, lower than expected Faradaic efficiencies for methane were observed in the illuminated experiment (Fig. S9B); this is in agreement with literature concerning the photosensitivity of methanogenic archaea to blue light (70, 71). Installation of a 455-nm filter directly after the photoanode restored the Faradaic efficiency for methane to expected levels (Fig. S9B). Experiments to produce methane using this hybrid bioinorganic system were conducted similarly to those described earlier except headspace analysis was performed only once per experiment after 3 d. On average, 1.75 mL CH4 (68.8 nmol) (n = 2) were produced in illuminated experiments, whereas only 0.58 mL CH4 (22.8 nmol) (n = 3) were produced in identical nonilluminated controls (Fig. 4D). We note that higher background levels of methane production were observed in these unassisted photochemical experiments compared with the electrochemical and photoelectrochemical experiments and speculate that this may be due to biological corrosion of the multicomponent photocathode, a phenomenon previously documented in other systems (53, 72). Subtraction of the background methane results in an average Faradaic efficiency of 74%. This result clearly demonstrates the successful conversion of CO2 to CH4 using light as the sole energy input.
In summary, we have established a hybrid bioinorganic approach to solar-to-chemical conversion by transforming carbon dioxide and water to the value-added chemical product methane. The present integrated system couples inorganic hydrogen generation catalysts powered by sustainable electrical and/or solar energy inputs with a biological catalyst that can harvest reducing equivalents from H2 to convert carbon dioxide to methane with up to 86% Faradaic efficiency. The data show that the low aqueous solubility and mass transfer rate of hydrogen do not preclude it from being an effective molecular redox carrier and offer a starting platform for potential integration with sustainable sources of electricity (e.g., solar, wind, hydrothermal, nuclear, etc.). Indeed, a solar-to-chemical efficiency of 10% and an electrical-to-chemical efficiency of 52%, which compare favorably to previously reported systems (73), are possible given caveats of materials and biological integration, assuming efficiency values of 20% for solar-to-electrical conversion using a commercially available photovoltaic and 70% for electrical-to-hydrogen conversion in an optimized system (74), as well as 86% energetic efficiency for H2 to CH4 based on thermodynamic values. Furthermore, we have developed an earth-abundant nanoparticulate nickel sulfide HER catalyst that operates effectively under biologically compatible conditions. At an applied current of 7.5 mA, use of the α-NiS catalyst results in 110 mL (4.3 mmol) of methane over 7 d with a Faradaic efficiency of 74%, comparable to Pt. Moreover, we have demonstrated that use of a photoactive silicon cathode results in an overpotential of only 175 mV, allowing for greater methane generation from a defined quantity of electrical energy. Finally, combining an n-TiO2 photoanode with a p-InP photocathode allows for fully unassisted light-driven conversion of CO2 to CH4. By realizing an effective artificial photosynthesis platform for the production of a value-added chemical product from light, CO2, and water, these results provide a starting point for achieving sustainable chemistry with materials biology.
Reagents were purchased from commercial sources as noted in the Supporting Information and used without further purification. Ultrahigh-purity gases purchased from Praxair were used for all anaerobic manipulations. 13C-labeled CO2 was purchased from Cambridge Isotopes. Distilled water (dH2O) was deionized to a resistivity of 18.2 MΩ·cm using a Millipore Milli-Q UF Plus system (ddH2O).
M. barkeri (ATCC 43241) was purchased from the American Type Culture Collection (ATTC). All culture manipulations were performed inside a Vacuum Atmospheres Nexus One glovebox with an atmosphere of 90% nitrogen and 10% hydrogen. Oxygen levels and humidity levels within the box were controlled using a STAK-PAK palladium catalyst and desiccant system in a Coy Laboratory Products unheated fan box. M. barkeri primary cultures were propagated in 18 × 150-mm Balch tubes with butyl rubber stoppers and aluminum crimp seals using ATCC medium 1043 with deoxygenated methanol (1% vol/vol) as the growth substrate. Secondary cultures were started by diluting late exponential phase primary cultures 1:100 into 1 L of ATCC medium 1043 in a 2-L anaerobic media bottle with deoxygenated methanol (1% vol/vol) as the growth substrate. All cultures were grown at 37 °C with shaking (200 rpm). The primary culture was propagated by diluting 1:100 into fresh ATCC 1043 medium every 4 d for ∼3–4 mo, or until changes in growth patterns were observed. Glycerol stocks were used for long-term culture storage.
A M. barkeri secondary culture was harvested after 4 d of growth by centrifugation at 6,000 × g for 7 min at 4 °C using O-ring–sealed centrifuge bottles. After centrifugation, the pellets were combined and washed twice with methanol-free media (200 mL). The final washed pellet was resuspended in fresh methanol-free media (20 mL). At this point, the OD600 nm of the culture was determined by diluting the cell suspension 1:100 into 10 mM Tris buffer (pH 8.5).
The electrolysis cells were assembled inside an anaerobic glovebox. The appropriate media was added to the cathodic (130 mL) and anodic (70 mL) compartments as specified in the Supporting Information. Additionally, a magnetic stirbar was added to the cathodic compartment. The resuspended cell pellet was used to inoculate the cathodic side of the electrolysis cell to a final OD600 nm of 0.35. Appropriate electrodes were added to the cathodic and anodic compartments, and the tightly sealed electrolysis cells were removed from the anaerobic chamber. Once outside, the cathodic compartment was sparged with CO2 for 5 min using the internal sparge line. Helium (1 mL) was injected into the cathodic headspace as a gaseous internal standard. The electrolysis cells were submerged to the horizontal flanges in water baths set to 37 °C. The cultures were stirred at 300 rpm for the duration of the experiment. In experiments with 24-h (12-h) timepoints, the experimental interval consisted of 23 h (11.5 h) of electrolysis followed by a 1-h (0.5-h) period for headspace analysis and sparging. GC analysis was performed at each timepoint by direct introduction of the headspace into a GC sampling loop as described in the Supporting Information. Immediately after sampling, the headspace of the cathodic chamber was exchanged by sparging with fresh CO2 for 5 min via the internal sparge line, followed by injection of a helium (1 mL) internal standard.
Freshly prepared solutions of NiCl2·6H2O (0.2 M), thioacetamide (0.4 M), and 5% (vol/vol) ammonium hydroxide were prepared using ddH2O. Nickel chloride solution (2 mL) and thioacetamide solution (2 mL) were combined in a 10-mL microwave reaction vial and sparged with N2 for 10 min. Upon addition of ammonium hydroxide solution (40 µL), a small amount of yellow precipitate was observed. The reaction tube was then capped under N2 and microwaved for 30 min at 150 °C with a 5-min temperature ramp using a CEM microwave reactor. Following the reaction, the supernatant was decanted and the black particles were washed three times with ddH2O and once with isopropanol by centrifuging the sample, decanting the supernatant, and resuspending the particles. Drying under vacuum overnight yielded a fine black powder (22 mg, 61% yield). The product was characterized by rotating disk electrochemistry (RDE) and galvanostatic experiments, pXRD, EDX, TEM, high-resolution TEM (HRTEM), ICP-OES, and TGA as described in the Supporting Information. The catalyst was stored in air for 3–4 mo, after which time it was found to be no longer active for HER.
All electrochemical experiments were performed using a BASi Epsilon potentiostat. Pt gauze (electrochemical grade, purchased from Alfa Aesar or Sigma-Aldrich) was used as the anode in all experiments except the unassisted full-splitting, where an n-TiO2-coated FTO anode was used instead. The reference electrode was an aqueous Ag/AgCl electrode (3.5 M KCl) purchased from BASi. All reference electrodes were externally calibrated to potassium ferricyanide in pH 7 phosphate buffer (Fe3+/2+ couple is 0.436 V vs. standard hydrogen electrode) before and after each experiment. No iR compensation was applied.
Sodium chloride, magnesium sulfate heptahydrate, ammonium chloride, potassium phosphate monobasic, potassium phosphate dibasic, sodium phosphate monobasic, sodium phosphate dibasic, manganese(II) chloride tetrahydrate, methanol, and pressure-release aluminum crimp seals (20 mm) were purchased from Fisher Scientific. Calcium chloride dihydrate, iron(II) sulfate heptahydrate, nickel(II) chloride hexahydrate, boric acid, cobalt(II) chloride hexahydrate, copper(II) chloride dihydrate, sodium molybdate dihydrate, sodium sulfide, l-cysteine, biotin, folic acid, pyridoxine hydrochloride, thiamine hydrochloride, riboflavin, nicotinic acid, calcium d-(+)-pantothenate, vitamin B12, p-aminobenzoic acid, and thioctic acid were purchased from Sigma-Aldrich. Yeast extract was purchased from EMD Biosciences. Casitone (pancreatic digest of casein) was purchased from BD. Zinc sulfate heptahydrate was purchased from Mallinckrodt Chemicals. Thioacetamide was purchased from Alfa Aesar. Resazurin was purchased from Eastman Kodak Company. Balch tubes (18 × 150 mm), anaerobic media bottles (250 mL and 2 L), and butyl rubber stoppers (20 mm) were purchased from Chemglass.
ATCC 1043 medium was used as the standard medium for the cultivation of M. barkeri ATCC 42431. To prepare the medium, all components (listed below) except iron sulfate, sodium bicarbonate, and reducing agent were dissolved in the appropriate volume of ddH2O. The solution was brought to a vigorous boil while sparging with an 80% N2/20% CO2 gas mixture. Once boiling, iron sulfate and sodium bicarbonate were added and boiling was continued until the solution became a bright pink color. The solution was moved to an ice bath and cooled to room temperature under constant sparging with the above gas mixture. The medium was subsequently transferred to a N2-degassed serum bottle using a modified Drummond Original Pipet-Aid Pipet Controller with the gas inlet line dispensing an 80% N2/20% CO2 gas mixture. The serum bottle was sealed with a butyl rubber stopper and aluminum crimp, and 10 mL reducing agent (see preparation below) were injected per liter of medium using a N2-flushed syringe. The sealed medium was autoclaved, resulting in a light yellow solution.
A stock of reducing agent was prepared by adding a 0.01% (wt/vol) resazurin solution (200 μL) to ddH2O (200 mL) and sparging the solution with N2 for 20 min. Under continuous N2 flow, l-cysteine (6 g) was added, followed by sodium sulfide nonahydrate (6 g). The N2 sparge was continued until the sodium sulfide had dissolved and the solution was completely colorless. After preparation, the reducing agent was stored under N2 atmosphere in a 250-mL serum bottle sealed with a butyl rubber stopper and aluminum crimp.
Per liter, the medium contains:
Per liter, the SL-6 trace elements solution contains:
To prepare a minimal (carbon-free) version of ATCC 1043 medium, yeast extract, casitone, and sodium bicarbonate were omitted from the standard recipe above. The sodium bicarbonate was replaced with a phosphate buffering system composed of Na2HPO4∙7H2O (4.84 g/L of media) and NaH2PO4∙H2O (1.88 g/L of media). This medium was used for initial electrolysis experiments with Pt. Additionally, a N2-sparged version of this medium was used for isotopic labeling studies.
Slight modifications were made to the minimal medium used for unassisted photochemical experiments. To prepare the catholyte, yeast extract, casitone, and resazurin were omitted from the standard medium recipe, although trace resazurin was still present from the reducing agent. In addition to sodium bicarbonate, the phosphate buffering system described above was added to increase ionic strength. The same medium was used to prepare the anolyte, except the reducing agent was omitted.
Gas evolution experiments were conducted in custom-made two-compartment glass electrochemical cells (Adams and Chittenden Scientific Glass) (Fig. S1). The anode chamber (80 mL) (a) was closed at the top with a GL45 media bottle cap with silicone septum. The cathodic chamber (150 mL) (b) was sealed with a gas-tight lid possessing five distinct ports: (i) electrical connection to cathode (c), (ii) electrical connection to reference electrode (d), (iii) CO2 line for solution sparging (e), (iv) connection to GC for headspace sampling (f), and (v) resealable septum for manual gas injection/sampling (g).
Ports d, e, and f on the lid (size Ace #7) were sealed with front-sealing PTFE bushing closures (5846-44) and port c (size Ace #15) was sealed with a rear-sealing PTFE bushing closure (5846-48) (Ace Glass). Port g (size Ace #7) was sealed with a front-sealing nylon bushing closure (5846-04, Ace Glass). Bushings for ports c–f were modified in the following manner: a 1/16-in hole was drilled all of the way through the center of the bushing, and 1/4-in -28 screw threads were drilled in the center, 8 mm deep from the outer side of the bushing. A 4-in-long stainless steel rod (1/16-in o.d.) was inserted through the bushings for ports c and d and further sealed using a 1/16-in Tefzel ferrule and flangeless male nut (P-200 and XP-235X, respectively; Upchurch Scientific) that could be screwed directly into the newly made opening on the bushing. Ports e and f were constructed in a similar manner, but replacing the stainless steel rods with 1/16-in-o.d. PEEK tubing (1531, Upchurch Scientific). The PEEK tubing was connected to Swagelok ball valves (40 series 1/8-in tube fitting; Swagelok) using PEEK unions (P-703, Upchurch Scientific), 1/8-in-o.d. stainless steel tubing, and 1/16-in and 1/8-in Tefzel ferrules and flangeless male nuts (P-300X and P-335X, respectively; Upchurch Scientific). A piece of 1/8-in-o.d. stainless steel tubing was used to connect the ball valve of port f to the stem of a Quick-Connect (SS-QM2-S-200, Swagelok), which could be attached directly to the GC inlet that had been modified with a Quick-Connect body (SS-QM2-B-200, Swagelok). The bushing for port g was modified as follows: a 1/16-in hole was drilled all of the way through the center of the bushing, and a #8–32 screw thread was drilled in the center, 1/2-in deep from the outer side of the bushing. An 8-mm-diameter Teflon-coated silicone septum (Ace Glass) was placed behind the modified bushing of the manual gas injection port, and a 1/2-in #8–32 sealing socket head cap screw (95198A535, McMaster-Carr) was used to block access to the port when not in use. All electrodes used in the cathodic chamber were entirely contained within the chamber for the duration of the experiment to minimize leaks and were connected to a potentiostat via the stainless steel feed-through rods.
The headspace volume of the cathodic chamber was 100 mL. For electrochemical and photoelectrochemical experiments, the anode and cathode chambers were separated by a cation exchange membrane (Nafion 117, Sigma-Aldrich); for experiments with no externally applied potential, an anion exchange membrane was used (AMV membrane; AGC Engineering Co., Ltd.) to minimize pH changes.
Potentials during galvanostatic experiments were measured with respect to a Ag/AgCl reference electrode that had been externally calibrated to K3[Fe(CN)6] in pH 7 phosphate buffer [E1/2 = 0.437 V vs. standard hydrogen electrode (SHE)]. Overpotentials were determined from the potentials measured at t = 5 min; we surmise that the decay in applied potential over time does not indicate instability of the catalyst but rather is caused by drift of the Ag/AgCl reference due to sulfide poisoning. Taking into consideration that CO2-sparged 1043 medium has a pH of 6.5, overpotentials for CO2 reduction to CH4 (E = -0.21 V vs. SHE at pH 6.5, 25 °C, 1 atm) were calculated as follows:
Headspace samples were analyzed with a multiple gas analyzer 8610C GC system equipped with a Haysep D column (1/8 in × 6 ft) and a 13X Mol Sieve column (1/8 in × 6 ft) (SRI Instruments). All gases were detected with a thermal conductivity detector with Ar as the carrier gas at a setpoint flow rate of 23 mL/min. The oven program was as follows: hold at 35 °C for 7.4 min, ramp to 60 °C (40 °C/min) followed by a hold for 4 min, and ramp to 220 °C (40 °C/min) followed by a hold for 2 min. Events were set as follows: valves 1 and 2 inject at 0.5 min; stop-flow solenoid is on at 3.2 min and off at 8.4 min, valve 1 returns to load at 8.4 min, and valve 2 returns to load at 15.5 min.
Helium (1 mL) was injected into the headspace of each experiment as an inert internal standard; He and H2 peaks are adequately baseline-separated using the aforementioned method. Representative elution times for gases of interest are as follows: He standard (1.93 min), H2 (2.16 min), CH4 (5.98 min), CO2 (9.31 min).
Headspace samples were introduced onto the GC by first using an evacuated 350-mL Strauss flask to evacuate the GC sample loop (1 mL), and then opening the sample loop to the headspace using the Swagelok ball valve located on the cap of the electrochemical cell. This procedure allowed for direct sampling of the headspace with little oxygen contamination.
Hydrogen and methane were quantified according to calibration curves prepared by injecting known volumes of H2 and CH4 into an electrochemical cell prepared identically to those used in gas measurement experiments (Fig. S2). At the start of each set of experiments, one sample of known H2 and CH4 concentration was run to ensure that the GC remained properly calibrated.
All CH4 Faradaic efficiency calculations are based on cumulative measured CH4 (VT) over the duration of the experiment (3 d or 7 d). For each biological methane production experiment, a corresponding nonelectrolyzed/nonilluminated control with identical electrodes was performed to account for differences in residual methanol (a potential growth substrate) after washing of the culture. The cumulative amount of CH4 measured in each control experiment is denoted as VC. Faradaic efficiency (FE) for methane was calculated as follows:
The high-resolution GC-MS data were collected using an Agilent 7890A chromatograph and an AutoSpec Premier mass spectrometer (Waters) equipped with an electron impact ion source and Masslynx software. For gas chromatography, an HP-5 column (0.0250 mm × 30 m, 0.25-μm film thickness; Agilent) was used. The carrier gas was helium and the oven temperature was maintained at 50 °C. Samples were introduced directly to the column via a splitless manual injection using a 1-mL gas-tight Hamilton syringe (Hamilton Company) that had been flushed with N2. For methane detection, 500 μL of electrolysis cell headspace was injected onto the instrument. The source temperature of the mass spectrometer was maintained at 150 °C and electron energy was set at 70 eV. For methane detection, the instrument was tuned to 10,000 resolution using the N2 parent ion (28 m/z) and fragment (14 m/z). Methane was detected using voltage scanning from 8 to 33 m/z with water (18 m/z), an O2 fragment (16 m/z), or a N2 fragment (14 m/z) used as reference peaks as appropriate.
To determine the limit of detection of the high-resolution GC-MS, successively diluted samples of methane gas were injected onto the instrument until a methane signal could no longer be detected. The last concentration at which a methane peak was detectable was 0.5% (vol/vol). For each isotopically labeled electrolysis experiment, the sample was also analyzed using the SRI Instruments GC (as described above in Headspace Analysis by GC) after analyzing the headspace using the high-resolution GC-MS instrument. Using this method, the total volume of methane generated during the 3-d experiment could be quantified. To determine the percent of unlabeled methane potentially present in the system but undetectable by the high-resolution GC-MS, the volume equivalent to 0.5% of the 100-mL electrolysis cell headspace (500 µL) was divided by the total amount of detected methane.
pXRD was performed on a Bruker GADDS Hi-Star D8 diffractometer using Co Kα radiation (1.790 Å). TEM and EDX spectroscopy were performed with a Hitachi TEM using copper grids (Ted Pella). HRTEM was performed with a 200-kV FEI monochromated F20 UT Tecnai instrument. TGA was performed with a TA Instruments Q5000 TGA.
ICP-OES measurements were performed on a Perkin-Elmer Optima 5300 DV instrument with an internal standard containing 50 µg Sc3+/mL. A calibration curve was made using [Ni] (NiCl2·6H2O) and [S] (H2SO4) concentrations of 1 μM, 10 μM, 100 μM, and 1 mM. α-NiS samples were dissolved in concentrated HNO3 and then diluted to 7% (vol/vol) HNO3 using ddH2O. This solution was filtered before analysis using a 0.45-μm PVDF syringe filter (Whatman). The nickel to sulfur ratio was found to be 0.80, with respective concentrations of 0.144 and 0.180 mM.
For RDE experiments, a 1-M sodium phosphate buffer, pH 7 (10 mL), was degassed by sparging with N2 for 15 min. A glassy carbon electrode (A = 0.071 cm2) was polished before use, then coated with a 3-μL drop of NiS ink (see Preparation of Pt and α-NiS/C Electrodes below) and allowed to dry for 20 min in air. The rotation speed was set to 1,500 rpm. The reference electrode was an aqueous Ag/AgCl electrode (3.5 M KCl) purchased from BASi, which was referenced to K3[Fe(CN)6] after the experiment. The counter electrode was a Pt wire, which was polished before use.
Platinum cathodes were fabricated from Pt foil (2.5 cm × 1.7 cm × 0.125–0.135 mm) (Sigma-Aldrich) and connected to stainless steel feed-through rods with Pt wire (Sigma-Aldrich). Before each experiment, Pt cathodes were cleaned by electrooxidation at 0.6 V vs. SHE for 2 min in 1 M HCl, followed by repeated rinsing with ddH2O.
α-NiS/C cathodes were prepared by deposition of an ink of α-NiS on carbon cloth. α-NiS powder (5 mg) was weighed into a 4-mL sample vial using a microbalance. Ethanol (600 μL), ddH2O (200 μL), and a solution of 5% (wt/vol) Nafion in aliphatic alcohols (Sigma-Aldrich) (40 μL) were added, and the vial was sonicated for 20–30 min. Meanwhile, Pt wire (0.5-mm diameter, Sigma-Aldrich) was threaded through the top of a 1.5 cm × 4.5-cm strip of carbon cloth (Fuel Cell Earth). This Pt wire was used only as an inert conductive material and was never submersed in electrolyte. α-NiS ink (240 μL) was applied to the bottom 3 cm of carbon cloth and allowed to dry for 2–4 h before use. Because both sides of the Pt and α-NiS/C electrodes are catalytically active, their geometric surface areas are 8.5 cm2 and 9.0 cm2, respectively.
Four-inch planar <100> p-silicon wafers (10–30 Ωcm) were cleaned in piranha solution and a 1:10 buffered hydrofluoric acid (HF) etch, followed by rinsing with dH2O and centrifuge drying at room temperature. Arsenic doping was performed using a rapid temperature annealing (RTA) process. In preparation for this procedure, 6-in silicon handle wafers were spin-coated with arsenic dopant solution at 3,000 rpm for 1 min, followed by baking on a hotplate at 150 °C in air for 30 min. Immediately after cooling, a 6-in handle wafer and freshly cleaned 4-in wafer were placed together such that the arsenic-coated side was in contact with the wafer to be doped. After one or two dummy runs to ensure reliable heating and cooling profiles, the wafers were placed in the RTA chamber (Allwin21 Rapid Thermal Processing System) and purged with Ar for 5 min. Optimal doping conditions were found to vary from batch to batch, so normally 3–4 wafers were prepared using slightly different conditions and their photoactivity was evaluated at a later point. Standard RTA conditions were 900 °C for 3–4 min or 1,000 °C for 60–90 s. Following As doping, atomic layer deposition (ALD) of crystalline TiO2 was performed to protect the photocathode from corrosion during illumination in neural aqueous media. Immediately preceding ALD, the doped silicon wafer was pretreated with 1:10 buffered HF to remove native oxide. ALD was performed using a home-built setup with TiCl4 and H2O as precursors; a typical recipe is 600 cycles at 300 °C.
A thin layer (10–20 nm) of nickel–molybdenum alloy was sputtered on top of the TiO2 layer to act as an HER catalyst. For this step, it is imperative that the Ni–Mo layer not be too thick, as light transmission to the semiconductor may be impeded. Dummy runs were performed in advance to establish a suitable recipe. Rectangles of TiO2-coated silicon (2.5 × 3 cm2) were cleaned for 2 min with oxygen plasma to remove organics, then taped to a handle wafer around the edges using Kapton tape (to prevent side deposition, which may shunt the electrode). Sputtering was performed at 50-W dc (Ni target) and 150-W rf (Mo target) for 8 min with the Ni shutter open and the Mo shutter closed.
Electrical contact with the finished piece of silicon was made by first gently scratching a 2 × 2-cm2 region on the backing of the silicon with a diamond scribe and then applying a thin layer of gallium–indium eutectic (Sigma-Aldrich). A roughly 1 × 1-cm2 piece of conductive double-sided carbon tape was placed on the short edge of a 2.5 × 6-cm2 piece of titanium foil (Sigma-Aldrich). Subsequently, a thin layer of silver paste (SPI Supplies) was applied on top of the GaIn eutectic, and the silicon chip was gently pressed onto the carbon sticker and Ti foil. Once dry, epoxy resin (Loctite Hysol 1C) was applied to the front and back of the silicon–titanium assembly, taking care to leave no gaps where water could enter the device. The electrode was allowed to dry at ambient temperature in air for at least 24 h before use. The photoactive geometric surface area of the finished cathodes was 7.0 cm2. Before use in a biological experiment, the fabricated electrode was soaked in 1043 media for 24 h to remove any soluble fabrication materials that might be toxic to the cells.
A 5-nm layer of Zn and a 50-nm layer of Au were sequentially thermally evaporated onto the back side of an InP wafer, which was then subjected to an RTA process (450 °C for 30 min) to fabricate an ohmic contact. The annealing process transforms the Zn–Au layer into a Zn–Au alloy and a fraction of the Zn diffuses into the underlying InP layer, forming a p+-InP layer. The presence of the Zn–Au alloy layer prevents oxidation of metallic Zn. During this step, the color of the film changes from golden yellow to silver. Next, the wafer was sonicated sequentially in acetone and isopropanol and blown dry with N2. The wafer was etched in a 1:1 mixture of conc. HCl:conc. H3PO4 for 5–10 s, then rinsed with dH2O water three times and blown dry. Immediately after this step, the sample was placed into the chamber of a home-built ALD and coated with 7–10 nm of amorphous TiO2 at 150 °C (TiCl4 and H2O were used as precursors). The TiO2-passivated wafer was sputtered with a 5-nm layer of Pt to act as an HER catalyst.
An ∼1.5 × 2-cm2 piece of p-InP/Pt wafer was used to fabricate each electrode. A 1 × 1-cm2 piece of conductive double-sided carbon tape was placed on the short edge of a 2 × 6-cm2 piece of titanium foil (Sigma-Aldrich). Subsequently, a thin layer of silver paste (SPI Supplies) was applied to the back side of the indium phosphide and gently pressed onto the carbon sticker and Ti foil. Once dry, epoxy resin (Loctite Hysol 1C) was applied to the front and back of the InP/titanium assembly, taking care to leave no gaps where water could enter the device. The electrode was allowed to dry at ambient temperature in air for at least 24 h before use. The photoactive geometric surface area of the finished cathodes was 3.0 cm2. Before use in a biological experiment, the fabricated electrode was soaked in 1043 media for 24 h to remove any soluble fabrication materials that might be toxic to the cells.
Titanium dioxide nanowires were synthesized via hydrothermal methods according to published procedures (37, 75). A 3 × 4-cm2 piece of FTO-coated glass was cleaned by sonicating first in acetone and then three times in isopropanol, then blown dry. The freshly cleaned FTO plates were placed in a Teflon-lined autoclave container, conductive side facing down. It was critical to mechanically remove any residual TiO2 from the walls of the Teflon container by sonicating in dH2O for at least 2 h, followed by sonication in 6 M HCl for 30 min and multiple rinses with dH2O. In a typical synthesis, 0.5 mL titanium tetraisopropoxide (Sigma-Aldrich) was injected into 30 mL of 6 M HCl and shaken well before pouring into the Teflon container such that 75% of the FTO substrate was immersed. The assembled autoclave was placed into a preheated oven at 200 °C for 2–2.5 h. To terminate growth, the autoclave was removed from the oven and cooled to room temperature for 3 h before opening.
The coated FTO plates were removed and rinsed with copious dH2O, then blown dry before being annealed in air (30-min ramp to 450 °C, 30-min anneal, followed by natural cooling). They were subsequently introduced into the chamber of a home-built ALD and coated with 10 nm of amorphous TiO2 (precursors were TiCl4 and H2O).
To assemble the electrode, a 0.5 × 2-cm2 piece of conductive double-sided carbon tape was placed in the region at the top of the anode that was not coated with n-TiO2 nanowires (and hence remained conductive: average resistance should be < 100 Ω). Subsequently, a thin layer of silver paste (SPI Supplies) was applied to the conductive carbon and gently pressed onto a 2 × 4-cm2 piece of Ti foil. Once dry, epoxy resin (Loctite Hysol 1C) was applied around the junction with the Ti foil, taking care to leave no gaps where water could enter the device. The electrode was allowed to dry at ambient temperature in air for at least 24 h before use. The photoactive geometric surface area of the finished cathodes was 9.0 cm2.
Titanium wire (GalliumSource, LLC) and silver paste (SPI Supplies) were used to make an electrical connection between the titanium foil support of the photocathode and the stainless steel feed-through rod of the electrochemical cell. The entire photoactive region of the photocathode was submersed in media and illuminated with a 740-nm LED (Mightex Systems) with an intensity of 20 mW/cm2. The electrolysis cell was heated to 37 °C in a water bath and stirred at 300 rpm for the duration of the experiment.
The standard electrochemical cell design described above was slightly modified for unassisted photochemical experiments. A 1-in-diameter quartz window was added to the anodic chamber, 180° from the membrane, to prevent initial filtering of the full-spectrum light by glass. The n-TiO2 photoanode was placed in the center of the anodic compartment, and immediately behind it was added a 1-in-diameter 455-nm filter (Edmund Optics), which was necessary to prevent blue light from reaching the culture (70, 71). Instead of Nafion-117, an AMV anion exchange membrane was used (AGC Engineering Co., Ltd.). Titanium wire (GalliumSource, LLC) and silver paste (SPI Supplies) were used to make an electrical connection between the titanium foil support of the photocathode and the stainless steel feed-through rod of the electrochemical cell. A full-spectrum 300-W Xenon arc lamp (Newport Corp.) was used as a light source, with an intensity of 2.2 W/cm2. The light beam passed first through the n-TiO2 anode, then through the 455-nm filter and the membrane, before finally impinging on the cathode. The photoanode was illuminated with a 1-in-diameter light spot. The entire photoactive region of the photocathode was submersed in media and illuminated.
Photocurrent under illumination was measured with a BASi potentiostat using a standard two-electrode cell setup (reference and counter leads were shorted on the n-TiO2 anode, and working lead was attached to the p-InP cathode).
We thank Dr. Zhongrui Zhou for help with HR GC-MS analysis, Dr. Hans Carlson for helpful advice on methanogen culturing, and Prof. Jonah Jurss for help with the design of electrochemical cells. This work was supported by DOE/LBNL DE-AC02-05CH11231, FWP CH030201 (to C.J.C. and M.C.Y.C.), a Laboratory Directed Research and Development Seed Grant from LBNL (to C.J.C. and M.C.Y.C.), and DOE/LBNL DE-AC02-05CH11231, PChem ( to P.Y.). C.J.C. is an Investigator with the Howard Hughes Medical Institute. E.M.N. and J.J.G. gratefully acknowledge support from the National Science Foundation Graduate Research Fellowship Program (NSF GRFP). J.J.G. also acknowledges support from NIH Training Grant 1 T32 GMO66698. J.R. gratefully acknowledges the support of the NSF GRFP under Grant DGE-0802270, and the University of California, Berkeley Chancellor's fellowship. This work used the Vincent J. Proteomics/Mass Spectrometry Laboratory at University of California, Berkeley, supported in part by NIH S10 Instrumentation Grant S10RR025622.
↵1E.M.N. and J.J.G. contributed equally to this work.
Author contributions: E.M.N., J.J.G., C.L., P.Y., M.C.Y.C., and C.J.C. designed research; E.M.N., J.J.G., C.L., Y. Su, J.R., Y.Y., and Y. Sun performed research; E.M.N., J.J.G., C.L., P.Y., M.C.Y.C., and C.J.C. analyzed data; and E.M.N., J.J.G., C.L., P.Y., M.C.Y.C., and C.J.C. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1508075112/-/DCSupplemental.
