Results and Discussion

To examine the SPARC reaction as a function of temperature and pressure, we constructed a fixed-bed, tubular flow reactor in which the catalyst bed could be both heated and irradiated, as shown schematically in Fig. 1. During the reaction, CO₂ and steam were flowed at 40 standard cubic centimeters per minute (sccm) over the 5% cobalt on a TiO₂ catalyst bed, which was heated via an internal electric heater and irradiated with four surrounding 250-W Hg lamps. The products were collected by passing the hot effluent gas through a condenser unit at 0 °C to capture condensable products, through a back-pressure regulator to drop the pressure to 1.0 bar, and then through a sampling loop of an automated online gas chromatograph for real-time gas analysis (full details in the SI Appendix).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Schematic diagram of photothermal flow reactor with cartoon picture of a single Co/TiO₂ particle undergoing catalysis and TEM picture of cobalt on P25 TiO₂ catalyst.

As shown in Fig. 2, the condensable (liquid) productivity increases upon increasing the temperature from 110 to 200 °C (1 bar, P_H2O/P_CO2 = 1.2, UV irradiation) and the product distribution shifts to C₂ (CH₃CH₂OH and CH₃COOH) and C₃ (propanol) products. Temperatures higher than 200 °C were not explored due to reactor limitations. Below 150 °C, methanol is the exclusive condensable product, whereas at 150 °C or above, C₂ and C₃ products become more prevalent and the methanol diminishes. Upon going from 180 to 200 °C, there was doubling of the mass productivity but a drop in propanol production and a large increase in acetic acid formation. Upon lowering the P_H2O/P_CO2 to 0.6 (1 bar, 200 °C, UV irradiation) the amount of propanol increased 14-fold and pentanol (C₅H₁₂O) was also detected. The appearance of the heavier alcohols at or above 150 °C is consistent with an FTS-like mechanism becoming active as the thermal energy approaches the activation energy needed for chain formation (37, 38, 40). In a related study of the effect of temperature on the photochemical reduction of CO₂ to methane over TiO₂, Saladin et al. showed that the rate of methane formation increases upon going from 20 to 200 °C, indicating the rate-determining step is a thermal process (41, 42). Significantly, no higher C_n products were reported, which we suspect is because they lacked a proper FTS cocatalyst, such as cobalt. In our current system, the kinetics of the carbon-chain-forming reactions seem to be rate-determining at ∼200 °C. Higher temperatures may yield further improvement but it should be noted that the selectivity of cobalt-based FTS catalysts for higher C_n products peaks in the 200–220 °C region (43). Temperatures higher than this quickly lead to losses in the desired selectivity and corresponding increases in methanation.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Mass productivity and C_n selectivity as a function of photothermal reactor temperature at 1 bar and P_H2O/P_CO2 =1.2, and 40 sccm.

Further experiments were run at 200 °C at three total pressures (1.0, 2.7, and 6.1 bar) and two partial pressure ratios: P_H2O/P_CO2 = 0.6 and 1.2. These data are collected in Table 1, which reports the mean productivity for each run, in units of mass productivity (μg_prod·g⁻¹_cat⋅h⁻¹) and molar productivity (μmol_e·g⁻¹_cat⋅h⁻¹, where μmol_e = μmol electrons stored in the product). The products are divided into O₂, H₂, C₁, C₂–C₄, and C₅+, with certain higher C_n products explicitly listed. For data in which SDs are reported, the values are the mean of three independents runs, whereas those without are single-run data. The full product distribution and productivity data are given in SI Appendix, Tables S4 and S5. Fig. 3 is a bar graph comparing the mass productivity and selectivity as a function of pressure and feedstock ratio. The most compelling data are the observation that the C₂+ hydrocarbons and oxygenates are not minor products but instead constitute 74% by mass of the products under one of the best set of conditions screened (6.1 bar, P_H2O/P_CO2 = 0.6). Converted to a molar basis (moles of reducing electrons stored in products), this amounts to 68% of all electrons stored in products in C₂+ hydrocarbons. Of this, 22% by mass and 14% of the electrons are stored in liquid C₅+ products. The H₂ and C₁ products account for 35% of product mass and 29% of all electrons stored, but they are not the dominant products. In both feedstock ratios, we clearly observe an increase in overall productivity with increasing total pressure. However, only at the lower P_H2O/P_CO2 of 0.6 do we observe any C₅+ selectivity which becomes more favorable with increasing pressure. Clearly, the feedstock ratio is an important parameter that merits further investigation and optimization. These data reveal, to a first approximation, that the SPARC reaction responds to temperature and pressure in a manner similar to the FTS reaction and yields similar products. It is well-known that the FTS reaction responds to increases in pressure with an increase in productivity (gram product per gram catalyst per hour) and a shift in product distribution (selectivity) toward heavier hydrocarbons (40), and that the optimized industrial conditions are generally given around 20 atm, 220 °C, with a 2:1 ratio of H₂ and CO for a cobalt-based catalyst (44, 45). Because of this, we believe it is reasonable to assume some shared underlying chemical mechanisms and assume that higher pressures will shift the selectivity in SPARC chemistry to favor higher C_n products. Our current reactor could not safely go beyond 6.1 bar; however, we are fabricating a next-generation photoreactor able to withstand both higher pressures and temperatures.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Mass productivity and selectivity of SPARC reaction at different pressure and partial pressure ratios at 200 °C and 40 sccm.

The percentage of oxygenates in our product distribution at 70–90% by mass is much higher than found in typical FTS chemistry. This is, perhaps, not too surprising in that SPARC reactants are considerably richer in oxygen content. As water is a reactant in SPARC chemistry, P_H2O is expected to have a strong influence on the observed chemistry compared with the secondary effects which arise from water in FTS chemistry. At high P_H2O the SPARC reaction can presumably produce a great deal of H₂ or hydride equivalents, which should be advantageous, but water is also better able to bind to the surface than CO₂ due to its polar nature and strong H-bonding properties; therefore, high P_H2O can saturate the surface, blocking effective CO₂ adsorption and reducing hydrocarbon production. Just recently, Rani et al. showed that (Cu, Pt)–TiO₂-based photoreduction of CO₂ to CH₄ and CO at ambient temperature was maximized at 50% humidity and dropped at higher levels, presumably due to water-blocking surface binding of the CO₂ (46). FTS is technically run with zero P_H2O but because water is a product, some is always present. At low P_H2O, the SPARC reaction may be starved for hydride equivalents, leading to incomplete reduction of the CO₂. Surprisingly, a number of FTS studies have shown that deliberate addition of water to the feed can be beneficial in terms of shifting the selectivity toward C₅+ products (14⇓⇓–17), which indicates that the presence of water, in itself, is not detrimental to the hydrocarbon-chain-forming process. It is not yet clear how the selectivity of the SPARC reaction can be tuned toward less-oxygenated products, but the feedstock ratio is likely to be one key factor. In any case, most of these oxygenates are high molecular weight molecules with a greater fuel value on a mass basis than lighter alcohols or oxygenates.

Isotopic labeling experiments performed with ¹³CO₂ and D₂O feedstocks show incorporation of both labels into all of the reduced products (SI Appendix). The O₂ yields, shown in Table 1, were calculated by dividing all of the electrons released to form O₂ from H₂O by all of the electrons stored in the identified products at the expected stoichiometry for the SPARC reaction and range from 64% to 150%. The agreement is acceptable considering the large SDs in these measurements, which reflect difficulties in obtaining good peak integration due to large adjacent CO₂ and N₂ peaks in the chromatogram (SI Appendix, Fig. S6).

As seen in Fig. 4, no O₂ and only trace hydrocarbons are detected before the catalyst bed is irradiated. Although difficult to see, there is a small amount of H₂, CH_4, and C₂H₆, produced during the dark period, constituting less than 10% of the CH₄ and C₂H₆ observed once the light is on. These basal products are attributed to thermal reactions of H₂O and CO₂ with the freshly prepared and highly reactive nanosized cobalt islands (catalyst preparation and characterization are given in the SI Appendix), some of which are unstable toward oxidation. When the catalyst is irradiated there is a substantial increase in HC, H₂, and O₂ productivity consistent with a coupling of photochemical water oxidation with thermal CO₂ and proton reduction. Furthermore, the productivity is reasonably stable over the run period; however, transmission electron microscopy (TEM) analyses of the catalyst pre- and postrun reveal substantial agglomeration of the cobalt islands over this period (SI Appendix, Fig. S2). The initial cobalt islands averaging 4.5 ± 2.0 nm in diameter approximately doubled in size to 9.7 ± 2.5 nm over the 5-h run timeframe. Cobalt-based FTS catalysts undergo considerable reorganization in the first 24 h of on-stream use (47, 48) and it seems that a similar reorganization is occurring here. However, given the early stage of SPARC chemistry and our current limit on short time runs (5–8 h), there are not yet enough data to fully understand the catalyst behavior here. This will certainly be explored as we examine the catalyst stability and evolution over longer time periods.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Productivity of O₂ (blue circles) H₂ (red squares), and hydrocarbons (HC) (black diamonds) as a function of time at 200 °C, 6.1 bar, P_H2O/P_CO2 = 0.6 and 40 sccm.

Table 1 also lists the productivity data for select C₅+ products, whose structures are shown in Fig. 5. The structures were determined by National Institute of Standards & Technology (NIST) database matching of the mass spectrum fragmentation patterns and were fit with confidence level greater than 99% for nearly all compounds. C₅H₁₂O and C₈H₁₈ are two saturated hydrocarbons typical of an FTS-like mechanism. The alkylbenzenes or oxygenates thereof suggest formation of surface acetylides/acetylenes, which can easily undergo cyclotrimerization to lead to the observed products (49). This second C–C chain-forming pathway is reasonable if we consider that the SPARC reaction conditions are considerably less hydrogen-rich relative to normal FTS, which rarely yields alkynes. Whereas this unexpected mechanism seems to be responsible for the aromatic C₆+ products, the presence of C₈H₁₈ as well as C₇ and C₅ alkanes and olefins supports an FTS process is occurring concurrently.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Structures of select C₅+ products as determined from NIST mass spectral database.

The highest mass productivity obtained, 200 °C, 6.1 bar, P_H2O/P_CO2 = 0.6, corresponded to 175 μg·g⁻¹⋅h⁻¹, not including O₂, or a molar electron productivity of 35 μmol_e·g⁻¹⋅h⁻¹. For comparisons, this latter value is ∼20 μmol H_2(equiv)g⁻¹⋅h⁻¹ and is within the typical range of the 0.2–100-μmol product·g⁻¹⋅h⁻¹ (50, 51), reported for CO₂/H₂O semiconductor photocatalysis, although claims as high as 2,000 μmol ethanol g⁻¹⋅h⁻¹ exist (52). When examined with respect to the incident photon flux (λ ≤ 400 nm), an incident photon quantum yield (IPQY) of 0.02–0.05% is obtained on a per electron stored basis. These IPQYs are slightly higher if reported relative to the O₂ yield. Unfortunately, many studies of CO₂/H₂O semiconductor photocatalysis either do not report quantum yields or do not report them in a consistent manner. Where data exist, typical IPQY values are anywhere from 0.001% to 30%, although the higher values (>1%) were only seen with nonoxide catalysts (i.e., ZnS, CdS, GaP) (24). We suggest that our modest IPQY is more a reflection of the early stage of this work rather than any practical limitation and note that a significant (2.5-fold) jump in IPQY was observed upon increasing the pressure from 1 to 2.7 bar; however, no further increase was seen at 6.1 atm.

As mentioned previously, the choice of a good metal for FTS catalyst is crucial and Fe, Co, and Ru are among the best known (39). Here, we chose cobalt due to its known FTS performance and superior stability relative to iron (44), and to demonstrate that the SPARC reaction could be achieved without the requirement of rare or precious metal cocatalysts. In one of the few studies of the CO₂/H₂O semiconductor photocatalysis at temperatures in excess of 100 °C, only C₁ products were found with copper and platinum zincates and titanate photocatalysts (53). However, when Fe was used in the same systems, ethanol in addition to the C₁ products was observed (54), suggesting the SPARC reaction was in operation. It is curious that even higher C_n products were not observed, but these photoreactions were conducted at temperatures in excess of 300 °C, which may well be too high for good operation. Oddly enough, cobalt-based catalysts yield primarily methane during the hydrogenation of CO₂ (55, 56). It was also shown that the increased water partial pressure stabilized the catalyst and it may be that the even higher P_H2O is influencing the SPARC product distribution here.

In general, elevated temperatures are avoided in photochemical reactions due to increased rates of charge recombination in the photocatalyst. For example, the photoluminescence intensity of TiO₂ drops by 50–60% upon raising the temperature from 20 to 200 °C (57, 58). At present, we observe an increase in productivity with temperature, showing a significant thermal component to the rate-determining kinetic step in this photothermochemical reaction. Nonetheless, the losses in charge carriers could be an important consideration upon development of an improved SPARC catalyst, where the thermal steps are no longer rate-determining.

At present, this gas-phase SPARC reaction is far from optimized and simply shows proof of principle. Higher productivities and better product distributions are likely to be realized as pressure, temperature, reactant ratio, space velocity, and catalyst are optimized. For example, the current photoreactor can only safely be operated at pressures less than 6.1 bar, whereas higher pressures are likely to dramatically affect product selectivity and productivity. Similarly, Brunauer–Emmett–Teller (BET) surface area measurements reveal that the Nafion polymer, used to bind the catalyst powder to the glass beads, effectively covers up 50–60% of the catalyst surface area (SI Appendix), revealing that we can likely do better using alternative methods to immobilize the catalyst. The current data also raise a number of important questions which must be addressed if this technology is to be further developed. As both H₂ and CO are observed in the product stream, it is not clear if they are responsible for hydrocarbon formation via the FTS upon readsorption or if they represent a small fraction of surface escaped species. The most commonly invoked mechanism for metal cocatalysts on photoactive TiO₂ is that the metal islands act as electron collectors upon which metal hydrides and CO₂ adsorption/reduction occurs, whereas the holes migrate to the TiO₂ surface where they oxidize water. In a gas-phase model, the protons must migrate over the surface of the TiO₂ to the cobalt islands, leading to questions about the degree of surface hydration. It is notable that in the hydrogenation of CO or CO₂ on cobalt metal, several common surface-bound intermediates are invoked, even though the former gives hydrocarbon chains and the latter gives methane (55, 56).

Although the current SPARC technology is currently impractical on a commercial scale, it does offer a conceptually new and commercially promising solar fuels technology that would be simple and inexpensive relative to most PV-EC and PEC systems. The direct production of the value-added hydrocarbons liquid fuel minimizes the number of unit operations involved and the associated efficiency losses and capital expenses of each. In a field operation, it is easy to imagine the use of parabolic mirrors to focus and concentrate sunlight onto a catalyst bed, providing both the photons required for photoexcitation and the thermal energy needed to run the reaction. Assuming such a system may require active cooling, the excess thermal energy could be used for downstream product separations or other applications in which relatively low-grade heat can be applied. In this respect, an SPARC process can realize greater efficiencies than process requiring ambient or near-ambient temperatures in that the low-energy photons are used to help heat the SPARC reaction and to heat a working fluid to a more useful temperature (i.e., 200 °C). Concentrated sunlight minimizes photocatalyst costs and reactor volume, the former of which is often one of the more expensive components of any such technology. Another significant economic consideration of the process SPARC is that it is not expensive to compress CO₂ and H₂O and a liquid product is produced. Solar biomass gasification and related processes which produce CO and H₂ will subsequently require expensive compression to convert these gases to usable solar fuels. Finally, the SPARC reaction, as realized here, is a true gas-phase operation operating at conditions more typical of an industrial operation. The elevated temperature, pressure, and gas-phase operation also open previously unidentified possibilities about the types of semiconductors catalysts that are needed and may be used.