Product Distributions and Efficiencies for Ethanol Oxidation in a Proton Exchange Membrane Electrolysis Cell

A commercial fuel cell (5 cm² active area; Fuel Cell Technology Inc.) was operated as an electrolysis cell by supplying 0.1 M ethanol (Commercial Alcohols Inc.) solution to the anode at 0.2 or 0.5 mL min⁻¹ with a syringe pump and N₂ to the cathode at 35 mL min⁻¹. The cathode acts a dynamic hydrogen electrode (DHE) and provides a relatively stable reference potential. The anode inlet and both outlets of the cell were modified with stainless steel tubing that connected directly to the graphite flow field plates. The flow field channels were sealed with ethyl-2-cyanoacrylate³³ in order to minimize absorption of ethanol and reaction products into the graphite plates.³¹ The cell was operated at 80°C in all experiments.

Anodes were prepared using the following commercial catalysts. The carbon supported Pt catalyst (Pt/C) was HiSPEC 13100, 70% Pt on a high surface area advanced carbon support (Alfa Aesar; Lot# M22A026). The carbon supported PtRu alloy catalyst (PtRu/C) was HiSPEC™ 12100, 50% Pt and 25% Ru on a high surface area advanced carbon support (Alfa Aesar; Lot# P17B047). The carbon supported PtSn alloy catalyst (PtSn/C) was 40% HP PtSn alloy (3:1 atom%) on Vulcan XC-72 (BASF Fuel Cell Inc.; Lot# F0060910). Catalyst suspensions were prepared by sonication of the catalyst (21–39 mg) in ca. 0.2–0.4 mL of a 1:1 mixture of 1-propanol and Nafion solution (Dupont; 5% Nafion) for 3 h at ambient temperature. An appropriate amount of the suspension was spread onto Toray carbon fiber paper (CFP; TGP-H-090) with a spatula to give a metal loading of 3.2 mg cm⁻². Cathodes consisted of 4 mg cm⁻² Pt black on TGP-H-090.

Membrane and electrode assemblies were prepared by pressing (room temperature; ca. 1.5 MPa) an anode and cathode onto a Nafion 115 membrane in the cell.³⁴ Electrochemical measurements were made under steady state conditions at constant cell potentials using a Hokuto Denko HA-301 potentiostat.

The product collection apparatus is shown schematically in Fig. 1. Ethanol and reaction products from the anode and cathode were combined before analysis. Liquids were collected in a trap cooled with a mixture of ice and dry ice, while CO₂ remaining in the N₂ stream was measured in real time with a commercial non-dispersive infrared CO₂ monitor (Telaire 7001).³⁵ The current and CO₂ reading were allowed to stabilize, and then averaged over a period of at least 100 s.

Zoom In Zoom Out Reset image size

For analysis by ¹H-NMR, 400 μL samples collected from the trap were mixed with 100 μL of D₂O containing 32 mM fumaric acid as an internal standard, which gives a singlet peak in the spectra at 6.72 ppm.³² Spectra were recorded on a Bruker AVANCE III 300 spectrometer. The D₂O in the sample provided the field frequency lock and spectra were referenced to sodium 3-(trimethylsilyl)-2,2,3,3-tetradeuteropropionic propionate at 0 ppm. The residual ethanol concentration was determined from the triplet at 1.10 ppm. The only products detected in the exhaust solution were acetic acid (singlet at 2.01 ppm) and acetaldehyde (doublet at 2.15 ppm). Acetaldehyde forms a dimer under the conditions of these experiments,³⁶ as indicated by a doublet at 1.24 ppm, and so the integral of the 1.24 ppm peak was added to give a single acetaldehyde concentration.

Fig. 2 compares polarization curves for the oxidation of 0.1 M ethanol at the Pt/C, PtRu/C, and PtSn/C anodes. As expected from many previous studies,^10,12 the current for ethanol oxidation at low potentials was much higher at the PtRu/C anode than the Pt/C anode. However, the current at the PtRu/C anode levelled off at potentials above 0.5 V, while the current at the Pt/C anode increased sharply to a peak at ca. 0.6 V that was approximately double the current at the PtRu/C anode. The plateau in the current at the PtRu/C anode suggests that it became limited by ethanol diffusion through the CFP backing layer, in which case this mass transport limited current (i_lim) can be described by Eq. 5.³⁷

Where A is the area of the electrode, D is the diffusion coefficient, C is the concentration of ethanol, and l is the thickness of the CFP. Since changing the catalyst can only change n_av in Eq. 5, the higher currents at the Pt/C anode at potentials ≥ 0.45 V in Fig. 2 suggest that n_av was much higher than at the PtRu/C anode over this potential range.

Zoom In Zoom Out Reset image size

The currents at the PtSn/C anode in Fig. 2 were similar to those at the Pt/C anode at potentials below 0.35 V, but lower than those for both the Pt/C and PtRu/C anodes at higher potentials. The relatively poor performance of this electrode may be due to the lower loading of metal on the carbon support (40% vs. 70% and 75% for Pt/C and PtRu/C, respectively). The lower limiting current relative to the PtRu/C anode may indicate a lower n_av or may be due to an additional diffusion limitation in the catalyst layer, which was thicker because of the larger amount of carbon support. The PtSn/C catalyst employed here is same type as the PtSn₃/C catalysts from E-TEK employed and characterized by several other groups.³⁸ The 20% PtSn₃/C catalyst has been shown to be superior to 20% Pt/C from E-TEK for ethanol oxidation in a DEFC.³⁹ However, it is clearly inferior to the more advanced Johnson Matthey Pt/C and Pt/Ru/C catalysts from Alfa Aesar. The lower masses of Pt in the PtSn/C and PtRu/C anodes relative to the Pt/C anode was not a significant factor here, since the loading of the Pt/C catalyst could be decreased to match their Pt masses without a significant decrease in performance.

If it is assumed that there were no losses of ethanol from the cell or during collection of the liquids from the combined anode (liquid + CO₂) and cathode (N₂ + CO₂ + condensed liquids) exhausts, n_av can be obtained directly from the concentration of ethanol in the exhaust liquid (C_out) by using Eq. 6:

where i is the average current, and u and C_in are the flow rate and concentration, respectively, of the ethanol solution entering the cell. In these experiments 3.2% to 65.8% of the ethanol entering the cell was consumed electrochemically. Results for ethanol oxidation at the Pt/C, PtRu/C, and PtSn/C anodes are plotted as a function of potential in Fig. 3, together with values calculated from the product yields by using Eq. 4. The two methods gave similar values, with small differences that can be attributed to experimental uncertainty.

Zoom In Zoom Out Reset image size

It can be seen from the data in Fig. 3 that n_av was much higher at the Pt/C anode than at the PtRu/C and PtSn/C anodes. The maximum n_av was 7.9 at 0.5 V, while values of 5.8 and 4.5 were obtained for PtSn/C and PtRu/C at this potential. These differences are consistent with the higher yields of CO₂ that have generally been reported for ethanol oxidation at Pt relative to PtRu and PtSn.^7,13,35

Interestingly, the potential dependences were significantly different for the three catalysts. Whereas n_av varied over only a narrow range at PtRu/C, from 3.2 to 4.5, PtSn/C gave a significant peak at ca. 0.55 V, while there were two peaks for Pt/C. The peak at 0.5 V for Pt/C can be attributed to CO₂ production via the oxidation of adsorbed CO (CO_ads),^40–42 while the peak of n_av = 6.1 at 0.35 V suggests that there may be a pathway to CO₂ that does not involve CO_ads as an intermediate.

Although n_av is the key parameter that determines the fuel efficiency of a DEFC or EEC, accurate determination of the product distribution is also very important. In addition to verifying the accuracy of n_av values obtained from Eq. 6, it provides crucial information for analysis of the reaction kinetics, and the quantities of byproducts produced by the cell. Since acetaldehyde is an intermediate that can be recycled through the cell, whereas acetic acid is a terminal product, knowledge of their ratio is necessary to determine the overall system efficiency, and the final amount of acetic acid produced.

Fig. 4 shows faradaic yields of CO₂ vs. potential for the three different anodes. These results parallel the n_av data in Fig. 3 because changes in n_av are dominated by changes in the production of CO₂, due to the higher number of electrons transferred. However, the faradaic yields exaggerates the amount of CO₂ produced for the same reason. They are shown here to allow comparisons with literature reports, where faradaic yields have normally been presented.

Zoom In Zoom Out Reset image size

Fig. 5 shows chemical yields of CO₂, acetic acid, and acetaldehyde as a function of potential. These are based only on the measured quantities of these three products, and have been calculated by using Eq. 7:

where N_i is the number of moles of ethanol required to produce the amount of product i measured in the exhaust from the cell. The measured amount of ethanol consumed by the cell was not used here in order to minimize the uncertainty, and to obtain yields that would produce independent values of n_av when used in Eq. 4. However, it should be noted that the amount of ethanol required to form the measured products agreed very well with the measured decrease in ethanol concentration. The sum of ethanol plus products exiting the cell was 99.3 ± 1.1% of the ethanol entering the cell averaged over the three data sets. This indicates that there was not a significant loss of ethanol or products, and that no other products were formed in significant (>1%) quantities. This is an important result because it means that NMR analysis alone can provide the product distribution, with the yield of CO₂ calculated from the mass balance. Other minor products that have been reported include ethyl acetate,^23,28 ethane,⁴⁰ methane,⁴⁰ ethane-1,1-diol,²⁸ and ethoxyhydroxyethane.²⁸ None of these products were detected in this work.

Zoom In Zoom Out Reset image size

It can be seen from the data in Fig. 5 that the highest chemical yields of CO₂ were 50% for Pt/C at 0.5 V, 7% for PtRu/C at 0.45 V, and 27% for PtSn/C at 0.6 V. At lower potentials, which are necessary for the voltage efficiency to be reasonable, the CO₂ yield remained substantial for Pt/C but dropped to very low values (≤5% for potentials of ≤0.35 V) for PtRu/C and PtSn/C. At low potentials, Pt/C also gave high yields of acetaldehyde, which can be further oxidized to CO₂ and acetic acid if recycled and thereby increase the overall faradaic efficiency.

PtRu/C provided high selectivity for the oxidation of ethanol to acetic acid, which can be attractive for production of H₂ from ethanol electrolysis because acetic acid is an important commodity.² At low potentials, large amounts of acetaldehyde were also produced. PtSn/C also produced predominantly acetic acid, but the sample used here did not produce high enough currents at low potentials to be useful. The Pt/C catalyst was superior for faradaic efficiency, while the PtRu/C provided better voltage efficiency and better efficiency for acetic acid production.

Comparative product distributions from the literature have previously been tabulated and reviewed.^9,32 Many new catalysts that show improved activities and/or selectivity for the complete oxidation of ethanol at ambient temperature in liquid electrolytes have been developed.^9,11 However, we are not aware of any reports of product distributions for ethanol oxidation at elevated temperature or in polymer electrolyte cells for these advanced catalysts.

Ethanol electrolysis cell

The energy efficiency for production of hydrogen from the electrolysis of ethanol involves a number of factors, including the electrical energy required for the electrolysis and the various losses from incomplete oxidation of the ethanol and crossover to the cathode.⁵ However, in theory all of the residual ethanol and products can be collected without a significant loss of energy. Consequently, only the electrical energy input and the energy density of the ethanol (8.0 kWh kg⁻¹) will be considered here.

The electrical energy consumed (W_e) is given by Eq. 8,⁵

where E_cell is the applied cell potential, the energy density of the hydrogen produced is 33 kWh kg⁻¹ and the thermodynamic potential for electrolysis of water is 1.23 V. Consequently the energy consumed is proportional to the cell potentials vs. DHE in Fig. 2. The PtRu/C anode therefore produced the highest electrical efficiency at currents up to ca. 100 mA while the Pt/C anode was more efficient at higher currents and could produce hydrogen at higher rates. Pt/C provides the added advantages of lower ethanol consumption and lower amounts of byproducts that would need to be utilized to maintain the overall efficiency of the system. At 0.5 V and 35 mA cm⁻² the cell with the Pt/C anode consumed ca. 13 kWh kg⁻¹ of electrical energy to provide hydrogen with an energy density of 33 kWh kg⁻¹. Based on the measured n_av of 7.94, this would require ca. 5.8 kg of ethanol per kg of hydrogen, corresponding to 46 kWh kg⁻¹ of chemical energy. The total energy input was therefore ca. 59 kWh kg⁻¹ with concurrent production of ca. 3.5 kg of acetic acid. Electrolysis of water typically requires 50 kWh kg⁻¹, making this a potentially attractive technology for co-production of hydrogen and acetic acid, if the current density can be increased to be competitive with the ca. 1 A cm⁻² level currently used for water electrolysis.⁵

Direct ethanol fuel cell

Since n_av cannot be accurately determined in a DEFC because of the chemical reaction of ethanol with oxygen, values from ethanol electrolysis (Fig. 3) are used here to estimate the efficiencies of DEFCs. This, for the first time, allows catalysts to be compared based on the overall efficiency that they will provide. If the loss of ethanol due to crossover is neglected for simplicity, the overall efficiency of a DEFC (ɛ_DEFC) can be calculated by using Eq. 9:

where ɛ_rev is the theoretical efficiency of 96% at 80°C, ɛ_E is the potential efficiency (ɛ_E = E_cell/E_rev, where E_rev is the reversible cell potential of ca. 1.15 V), and ɛ_F = n_av/12. The potential efficiency includes losses due to the anode (η_anode) and cathode (η_cathode) overpotentials as well as the ohmic resistance of the cell (R) (Eq. 10).

If it is assumed that the overpotential of the DHE cathode was negligible, (η_anode + iR) is given approximately by the difference between the cell potential in Fig. 2 and the standard potential of +0.084 V for Reaction 1. In order to assess the effects that the different anodes would have on the efficiency of a DEFC through Equations 9 and 10, values of η_cathode were modelled by fitting a representative⁴³ iR corrected cathode polarization curve to a Tafel relationship (Eq. 11), with b = 47 mV (the low value can be attributed to ethanol crossover⁴⁴) and a current of 1 mA at 0.81 V vs. DHE.

The average n_av at each potential in Fig. 3 was used to calculate ɛ_F. Fig. 6 shows the calculated DEFC efficiencies as a function of the current density.

Zoom In Zoom Out Reset image size

It can be seen from the data in Fig. 6 that the highest efficiency for a DEFC would have been obtained with the Pt/C anode over the whole current range. The PtSn/C anode would give the lowest efficiency at all currents, while PtRu/C would give slightly higher efficiencies and somewhat higher current densities. These results highlight the importance of developing catalysts with high selectivity for the oxidation of ethanol to CO₂, since this is the dominant factor influencing DEFC efficiency.

It should be noted that the comparison of efficiencies calculated using Eq. 10 depends on the cathode overpotentials employed. If the cathode overpotentials are higher, the PtRu/C anode can provide the highest overall efficiency in the intermediate current region, but the efficiencies with all anodes would be lower than shown in Fig. 6. It should also be noted that the PtSn/C catalyst employed here had a significantly lower Pt loading than the Pt/C and PtRu/catalysts, which were also newer and more advanced. Consequently, the comparison of its efficiency with these state of the art catalysts does not accurately reflect the potential for advanced PtSn/C catalysts to contribute to DEFC and EEC technologies.