Vapor-fed direct ethanol FCs
Though liquid-fed operation of direct alcohol FCs is well-studied in the literature, little data exist for operation using a humidified ethanol vapor feed at near-ambient temperatures. Ghumman and Pickup used such a system to examine the effect of voltage pulses on carbon dioxide generation [57]. They used a low flow rate of 27 mL/min, a moderate ethanol concentration of 6% (1M), and flowing H2 at the cathode to maintain a stable reference. We can find no other studies of ambient-temperature ethanol vapor-fed FCs in the literature. We therefore first systematically studied the performance of this system, separate from any fermentation.
In vapor-fed FCs, the ethanol partial pressure should influence the FC performance in a manner similar to ethanol concentration in the liquid phase [58, 62, 65]. To examine the role of ethanol partial pressure, different gas mixtures were run through the FC anode at very high flow rates (1 L/min). Over the range of ethanol–water mixtures examined (0.5–15% ethanol), the partial pressure of water in the vapor feed remains roughly constant (within 5%) at 20.5 mmHg, while the 1 L/min flow rate ensures that the FC is consuming only a small fraction of the ethanol passing through (<5%). In this way, the formation of products and their effect on the reaction kinetics is small, and the concentrations of ethanol and water in the FC can be assumed to be constant.
Figure 2 shows the resulting current output of the FC, with a voltage fixed at 200 mV, as a function of ethanol partial pressure. A second x-axis at the top of the graph shows the solution ethanol concentration for reference. At low partial pressures, the current density increases nearly linearly with ethanol concentration, consistent with a pseudo-first-order reaction with respect to ethanol (rate constant 36 min−1). At higher ethanol partial pressures, the current density peaks near 3.5 mmHg at a value of 13.6 mA/cm2, slightly higher than the peak current density obtained when these same FCs are run on liquid ethanol–water mixtures at the same temperatures and voltage [39, 43, 48]. At higher ethanol partial pressures, there is a large run-to-run variability, along with an apparent slight decrease in current density. Similar decreases are observed in FCs running on liquid fuels [39, 67]. The decrease in performance likely arises from ethanol crossing over through the Nafion separator, which produces a mixed potential and lowers performance [36, 68]. Overall, these results demonstrate that FCs running on ethanol–water vapor can achieve power densities comparable to those for liquid-fed FCs, over a wide range of ethanol partial pressures.
To investigate the impact of ethanol partial pressure on performance at different operating voltages, the FCs were fed with the vapor from ethanol–water mixtures at a high flow rate (1 L/min) as shown in Fig. 2, but the potential was varied between 0 and 500 mV. The potential was stepped and held for 20 min to allow the FC to reach a steady state. The polarization curves and the calculated power curves for various ethanol concentrations are shown in Fig. 3.
As shown in Fig. 3a, as the ethanol concentration increases, current output reaches a maximum and then decreases, though the optimal concentration depends on voltage. At high potentials (near 500 mV), the current output shown in Fig. 3a peaks at a low ethanol partial pressure of 1.7 mmHg (3% ethanol in bubbling solution). Since the reaction is limited by charge transfer at high potentials, ethanol crossover is a significant impairment to current output, especially at higher ethanol concentrations. For example, at 5.5 and 8.0 mmHg (10 and 15% ethanol solutions, respectively), the crossover is significant enough that it has lowered the open-circuit potential to 400‒500 mV (as compared to 600‒700 mV at lower concentrations) resulting in zero net current. In contrast, near 0 mV, there is little resistance to charge transfer and the reaction is concentration or mass-transfer limited, so current increases with increasing ethanol partial pressure up to 5.5 mmHg. The small decrease in current output between 5.5 and 8.0 mmHg of ethanol (10 and 15% in solution) may be related to ethanol reducing H+ conductivity in the Nafion [69].
When considering the operating voltage at a particular concentration, the power density (Fig. 3b) is often more relevant than the current. As shown in Fig. 3a, the impact of ethanol crossover at high voltage is very apparent and causes a very large shift in the peak power towards lower voltages at higher ethanol partial pressures: 500–150 mV between 0.3 and 8.0 mmHg of ethanol. Interestingly, while peak power is low at both high and low ethanol concentrations, at intermediate ethanol partial pressures (i.e., between 1.7 and 5.5 mmHg) the peak power density changes by <25%, even though the optimal voltage shifts between 150 and 300 mV.
Similar effects have been noted in previous work with liquid-fed direct ethanol FCs, as well as work by others with direct methanol FCs. LSV data for a liquid-fed direct ethanol FC are shown in Additional file 1: Figure S5 to examine whether there are differences in optimal FC operating voltages between liquid-fed and vapor-fed. When comparing these identical FCs, the optimal voltages for a given ethanol solution are comparable up to 6% ethanol, though the instantaneous open-circuit voltages are slightly higher for vapor-fed direct ethanol FCs. The current and power densities, obtained at ambient temperature with an air-facing cathode, are broadly similar to other direct ethanol FCs, but it is likely that further enhancements could be obtained with additional engineering of the FCs, especially its cathode [70, 71]. The energy from ethanol oxidization to acetic acid is converted to electrical energy at efficiencies typical for direct ethanol and direct methanol FCs (approaching 40% at 500 mV) [25], but the overall process efficiencies are much lower, in large part due to the inability to fully oxidize ethanol to carbon dioxide. Additionally, only a small fraction of the ethanol is converted at high flow rates, although as discussed below higher single-pass conversions can be obtained by reducing the flow rate. Direct methanol FCs are known to suffer severe performance drops from methanol crossover, though systems that feed methanol as a vapor have been shown to mitigate crossover [72]. For both methanol and ethanol, vapor-fed operation may offer a means of reducing crossover, and control of the voltage allows for higher-power operation across a range of partial pressures.
The behavior of the FC versus flow rate was as expected from the current dependence on concentration (Fig. 2). We determined the dependence of current output on flow rate at constant potential (200 mV) and constant composition of the vapor stream (1.7 mmHg). Figure 4 shows how the current density varies with flow rate. An ethanol partial pressure of 3.1 mmHg falls in the pseudo-first-order region for ethanol partial pressure (see Fig. 2), allowing the current densities to be fit using a continuous stirred-tank reactor (CSTR) model with first-order reaction kinetics (for the equations used, see Additional file 1: § CSTR Model). The CSTR model was chosen because of the rapid mixing of gas expected in the fuel reservoir at the anode of the FC. Based on this model, and the rate constant determined based on the data shown in Fig. 2, a fit was obtained (red dashed line in Fig. 4) that is in excellent agreement with the flow rate current densities. At high flow rates (≥250 mL/min), the current densities were comparable to currents achieved with batch liquid-fed operation [39, 48, 63]. At low flow rates, the current drops substantially, since the effective ethanol concentration in the FC is lower due to conversion of the ethanol to acetic acid and other products.
The current data shown in Fig. 4 can also be used to determine an ethanol conversion that can be compared with results from gravimetric analyses. Figure 5 shows the coulombic conversion of ethanol in the vapor-fed FCs as a function of flow rate, for a fixed voltage (200 mV) and ethanol partial pressure (1.7 mmHg). The dashed line shows the expected conversion based on a CSTR model. Conversion to acetic acid (4 electrons per ethanol molecule) was used as the basis for the figure, because acetic acid is the primary product of ethanol oxidation [57], and because it represents a more complete oxidation of ethanol than conversion to acetaldehyde, the most common secondary product. Based on gravimetric analysis of the cold-trap condensate of the FC effluent gas, acetic acid is almost exclusively produced at high flow rates (~20:1 acetic acid-to-acetaldehyde ratio in products at 1000 mL/min). At lower flow rates the percent of acetaldehyde in the products increases, reaching 12% at 200 mL/min and 21% at 100 mL/min. In all cases, the mass balance based on the inlet vapor and condensate approximately closes (within 10%), as does the electron balance from the current output and condensate (assuming 2 electrons per molecule of acetaldehyde, 4 electrons per molecule of acetic acid, and no Faradaic losses). Ethanol conversions >60% are observed at our lowest tested flow rate of 10 mL/min.
The voltage response of the vapor-fed FC depends on the flow rate as shown in the variable voltage data plotted in Fig. 6. Figure 6a shows current as a function of voltage, while Fig. 6b shows power as a function of voltage. At very low flow rates, (<50 mL/min) the current and power increases for all voltages as flow rate increases. However, as the flow rate increases further, the current output increases only at low voltages (<400 mV). In this regime, the reaction is likely mass-transfer limited with charges created at the anode having little resistance flowing to the cathode [28]. In contrast, at higher voltages, the vapor-fed FC is likely limited by charge transfer between the anode and the cathode. These effects also shift the optimal voltage to higher voltages as the flow rate decreases. Though there is a 55% drop in current output at 200 mV when running the FC at 1000 mL/min as opposed to 50 mL/min, the peak power only drops 28% between 1000 and 50 mL/min. There is a trade-off between low flow rates that produce low power densities with high conversions and high flow rates that produce higher-power densities but may require the fuel to be recycled due to their low single-pass conversion. Since operating at lower flow rates may be desirable to reach higher conversions, the high power density available across a range of flow rates should be highly beneficial. Additionally, running the vapor-fed FCs at higher voltages should enable more of the ethanol’s energy to be captured as electrical power with minimal losses in the power density, especially at low flow rates.
Bio-hybrid fuel cells
In “Vapor-fed direct ethanol FCs” section, we showed that vapor-fed FCs have similar performance to liquid-fed direct ethanol FCs when operating on ethanol–water mixtures. Here we show one of the major advantages of vapor-fed FCs: the ability to operate on complex mixtures such as fermentations that often contain components that poison FCs. To demonstrate these capabilities, the vapor-fed FC was run as a bio-hybrid FC with an ongoing yeast fermentation as the ethanol source. The fermentation was run with a rich medium (YPD) that is known to rapidly poison FCs [38]. Even with an RO separation membrane to help purify the ethanol from the fermentation, liquid-fed bio-hybrid FCs have a >90% reduction in power density over a period of 1 week when running on rich media [48].
Because ethanol was constantly being removed from the fermentation, the bubbled fermentation appeared to be maintainable indefinitely simply by adding sugar and water periodically. A nitrogen flow rate of 30 mL/min was used to ensure reasonable ethanol conversion by the FC; this flow rate was sufficient to keep the ethanol concentration of the 125 mL fermentation between 4 and 8% ethanol when 3.75 g of glucose was added twice per week (i.e., every 3–4 days). A preliminary 1-month bio-hybrid FC run at higher flow rates is shown in Additional file 1: Figure S5.
The vapor-fed FC can operate with an ongoing fermentation for at least 5 months while maintaining a power density >0.3 mW/cm2. The current density and fermentation ethanol concentration for the first 90 days are plotted in Fig. 7. The baseline vapor-fed FC characterization was used to provide an expected current density for the bio-hybrid FC based on the average ethanol concentration (4‒8%), the flow rate (30 mL/min), and the operating voltage (200 mV). Over the first 30 days, the ethanol concentration was roughly 7%, with fluctuations between 6 and 10%; the current density averaged 3.5 mA/cm2 while the expected current density for this ethanol concentration and flow rate was roughly 5 mA/cm2. This modest difference between the expected current density and the actual current density may be linked to the extended operation of the FC in the bio-hybrid FC configuration (months in Fig. 7 vs. hours in Figs. 2, 3, 4, 5). In month-long runs, current density can decline due to build up of reaction products on the catalysts and limit on the lifetime of the FC [73, 74]. Despite these issues, the FC runs nearly as well over the last 60 days, as shown in Fig. 7, as over the first 30 days, with a modest drop in current attributable to the decline in the fermentation ethanol concentration from roughly 7 to 4.5%. The hourly and daily average current densities for the entire 5-month run are shown in Additional file 1: Figure S6, § Averaged 5-Month Data. These results clearly demonstrate that the FC performance is maintained for at least 150 days when running as a bio-hybrid FC.
One interesting effect of the fermentation on the FC is that whenever sugar is added to the fermentation, spikes appear in the current density. These dramatic increases take several minutes to build up after reconnecting the gas flow. This time frame suggests the current density increases are caused by disrupting the gas flow and fermentation headspace. Purely electrical changes in the system would only require seconds to take effect. Increased ethanol in the fermentation would require hours, so that also cannot be the cause (although it may lengthen the effect). We can reproduce the effect by briefly substituting air for the ethanol/water vapor, but not by briefly cutting off the vapor flow, strongly suggesting oxygen as the cause. Addition of oxygen to the anode fuel stream induces a similar effect in hydrogen-fed proton-exchange membrane FCs, where it is thought to react with adsorbed carbon monoxide and reactivate the poisoned catalyst [75]. We therefore hypothesize that, after long-term operation of the FCs, oxygen clears catalytic sites of adsorbed CO species and/or other non-reactive molecules that are intermediates or products in ethanol electro-oxidation. Regardless of the reason, the increase in power lasts for days, which is 1000× longer than the several-minutes-long interruption. A practical device might do well to purposefully incorporate oxygen addition, either through inclusion of a small amount of air or through an alternating flow scheme.
The fermentation can also be run for very long times with little loss in ethanol production rate, allowing more ethanol to be produced than in a batch fermentation. Concentrations were obtained bi-weekly just prior to addition of glucose and water, via FTIR spectroscopy of small samples of the fermentation liquid, as explained in the “Fourier transform infrared (FTIR) spectroscopy” section. Since opening the fermentation causes current spikes as discussed above, samples were normally taken only when glucose and water were being added to the fermentation. Figure 8 shows the concentrations of the major components of the bubbled fermentation versus time, for about 90 days, while Additional file 1: Figure S7, § Control Fermentation shows the concentrations of major components of the control (non-bubbled) fermentation operating in the same conditions but without bubbling. In the control fermentation, ethanol builds up to 16% over 2 weeks, but then only another 2% is produced by the yeast. In contrast, when the fermentation is running with a bio-hybrid FC, the nitrogen bubbling prevents the ethanol concentration from building up beyond 10%, which would slow the fermentation and also be undesirable for FC operation. Eventually the ethanol concentration settles down to 4.5–5%, at the lower end of the concentration range for optimal FC operation. Since the rate of ethanol production in the control fermentation slows down noticeably once it has reached even modestly higher ethanol concentrations (ca. 10% and above), sugar accumulates in the fermentation. In contrast, because of the faster fermentation in the bubbled fermentation, the glucose concentration remains consistently low when samples are taken (normally <1%, although around day 25 some higher concentrations are observed). The low glucose concentrations indicate that the fermentation is sometimes glucose limited, and that more feeding would probably allow a higher ethanol concentration to be sustained, if desired. Even without optimizing the glucose addition rate, the yeast fermented 75 g of glucose over the course of 90 days. The gas leaving the fermentation, being saturated with ethanol, stripped 34 g of ethanol from the fermentation for use by the FC (45% of the glucose mass). Since the maximum theoretical yield is 39 g of ethanol (ignoring the yeast’s metabolic needs), the yield was 87%. By comparison, using the same strain and a fermentation in which a high initial concentration of glucose (30‒40%) is used, a yield of 66% is obtained (data not shown). Presumably the yeast grown for months with stripping produces higher ethanol yields than yeast grown for days in batch operation because the former are spending almost 100% of their time in stationary phase rather than growth phase. Similar yeast conditions can be obtained in batch operation, such as the technique used by Brazilian ethanol producers where yeast is collected and re-used at the end of a fermentation. This keeps the yeast in a metabolic state that produces ethanol, not biomass [15]. Yields of 91% are reported, comparable to our result of 87% [15].
The variations in the concentrations of the lesser components are also instructive. The concentrations of acetic acid (filled black triangles) and glycerol (filled green squares) both grow steadily for 30 days before plateauing and remaining stable for the next 60 days (note that the acetic acid concentration shown is ×10). Glycerol is commonly produced as a minor component by yeast, and builds up in the fermentation due to its low volatility. The acetic acid is probably also produced by yeast metabolism [76], rather than backflow from the FC, given its very low concentration in the fermentation. The low acetic acid concentration is crucial, since previous work has demonstrated that the amounts of acetic acid produced by a FC can wipe out a suitably sized fermentation within days if not excluded or suitably remediated [43]. Since the acetic acid is now a component of the output stream, it may be possible to break it down further in a separate reactor or microbial FC using bacteria that can consume acetate and produce fuel gases (e.g., methane), such as Shewanella oneidensis [77], Anaeromxyobacter dehalogenans [78], or Geobacter sulfurreducens [79]. Yeast extract and bacteriological peptone are low concentration components of the growth medium that provide yeast with material for cellular synthesis. Their combined concentration (Y + P, open cyan circles) is steady for about 25 days, and then gradually declines to about 3/5 of its original level. Concomitantly, a visible layer of dead yeast cells was observed at the bottom of the fermentation vessel, suggesting that some of these components were being trapped in dead cells. This decline in Y + P was slowing by 90 days suggesting that Y + P was being slowly released by deteriorating dead cells. Quick checks with light microscopy show many cells in the process of budding (data not shown). All of these observations indicate continued yeast cell formation, growth, and death, throughout the entirety of the 90 days.