Abstract
Ammonia production has increased from few thousand tons in 1908 to above 200 million tons per year today, revolutionizing the fertilizer industry thanks to the Haber-Bosch (HB) process. However, the HB process is highly energy intensive consuming about 1.4% of fossil energy generated worldwide and releasing 1.87 tons of CO2 per ton of ammonia produced. This further reduces ammonia's scope as a carrier fuel for the hydrogen economy. Hence, finding alternative energy efficient ways to synthesize ammonia is important from more than one perspective. Ammonia synthesis from its constituent nitrogen and hydrogen gases is mainly hampered by the nitrogen reduction reaction (NRR) due to the strong N≡N bond (945 kJ mol−1). Electrochemical synthesis (ES) routes in this regard offer a milder approach. However, ES of ammonia under different temperatures, utilizing different electrolytes and catalysts has not yet reliably produced ammonia at viable rates and efficiencies. We report an origami-like Mo2C cathode catalyst for NRR that achieved a maximum synthesis rate of 2.16 × 10−11 mol cm−2 s−1 and a faradaic efficiency of 1.8% at 30 °C using Nafion-212 as electrolyte. Origami-like morphology containing numerous kinks appears to improve electrocatalytic activity and show a promising route for fabricating NRR catalysts with higher catalytic activity.
Export citation and abstract BibTeX RIS
Ammonia is one of the most produced chemicals in the world (currently ranked 2nd) thanks to the invention of the Haber-Bosch (HB) process.1 The HB process has increased the production of ammonia from a few thousand tons per year in 1908 to above 200 million tons in 2018.2 This is due to ammonia's utility in various industries such as fertilizers, transportation and explosives production.3 Ammonia as a fuel has a promising future due to its very high energy density of 11.5 MJL−1 in comparison to hydrogen that has an energy density of 4.5 MJL−1 at 690 bar and 15 °C compressed state and 8.5 MJL−1 in liquid state.4 Ammonia is easy to transport and already has an established pipeline network of 4830 km in the USA with frequent ammonia retail locations.5 Ammonia has generated substantial interest as a hydrogen storage material alone or in combination with borane as ammonia-borane.6 This material consists of central atoms from 1st row elements and hence has the added advantage of being lightweight with significantly increased hydrogen storage capacity per weight (19.6 wt%).7,8 Despite these advantages, ammonia's potential as a clean energy storage medium is largely left unexplored due to the energy intensive nature of the HB process that requires about 485 kJ mol−1 of ammonia produced and accounts for about 1.4% of the natural gas consumption worldwide along with releasing 1.87 tons of CO2 per ton of ammonia synthesized.9 On the contrary, natural ammonia synthesis by nitrogenase enzymes needs only 244 kJ mol−1 of ammonia produced indicating a possible pathway that is less energy intensive and modular for synthesizing ammonia.10 The challenge associated with ammonia synthesis is largely due to the strong N≡N triple bond (dissociation energy = 945 kJ mol−1) which is difficult to activate due to the lack of a permanent dipole that necessitates extreme conditions such as higher temperatures and pressures.4,10–12
Electrochemical synthesis (ES) of ammonia offers a promising way forward as the kinetics can be tuned not just by the temperature, but also by the operating potential and has many advantages over catalytic synthesis and recent reviews discuss the advantages and disadvantages of different electrochemical systems.11 Since hydrogen is supplied electrochemically as protons, there is no need for high purity hydrogen, which increases the cost of traditional catalytic processes.13–15 In 2009, using SmFe0.7Cu0.3−xNixO3 as working electrode and Nafion as electrolyte, Xu et al. reported a synthesis rate of 1.13 × 10−8 mol cm−2 s−1 with a current efficiency as high as 90.4% at an operating temperature of 80 °C.16 This has created huge interest in this field as a rate of 1 × 10−7 mol cm−2 s−1 is expected to meet the requirements for commercial production.17 However, these results could not be reproduced and most of the recent literature reports ES of ammonia at rates on the order of only 1 × 10−11 to 1 × 10−9 mol cm−2 s−1. The first electrochemical synthesis of ammonia was reported in 1985, and currently, a variety of electrolytes ranging from room temperature polymer electrolytes to high-temperature ceramic electrolytes are being studied.18–22 Similarly, significant effort is being placed on the electrocatalyst development for the nitrogen reduction reaction (NRR). Among the electrocatalysts studied, computational studies indicate that Fe, Mo, Rh, and Ru are best suited for the catalysis of NRR although Pt continues to be a comparative standard.23–28 For example, Lan et al. pre-exchanged the Nafion membrane with NH3 up to saturation and used it as the electrolyte for ammonia synthesis measurements using Pt as electrocatalyst, where an ammonia synthesis rate of 1.14 × 10−9 mol cm−2 s−1 was observed at an applied potential of 1.6 V.9 However, post-mortem analysis of the membrane was unavailable, and no information on its ammonia content was provided. On alternative catalyst research, Ru supported on carbon-felt as a catalyst and a combination of Nafion and 2 M KOH as electrolyte produced ammonia at a rate of 3.4 × 10−12 mol cm−2 s−1 at 20 °C and an applied voltage of −1.1 V.29 Further, it is also documented that Ru with higher steps dominate N2 dissociation even on a flat Ru (0001) surface.30 Mo-based catalysts are another promising class of catalysts for NRR. For example, Mo13 nanoclusters are theoretically demonstrated to catalyze NRR at potentials about −0.5 V and show preference towards NRR over hydrogen evolution reaction (HER) up to potentials of −0.8 V.31 Molybdenum carbides of different morphologies have recently been reported to show differing catalytic activity towards NRR.25,32–34 Further, among a variety of MXenes (M2X: M = Mo, Ta, Ti and W; X = C and N), Mo2C and W2C are theoretically shown to be good catalysts for ammonia synthesis.21 Despite all these advancements, the experimental rates for ES of ammonia ranges in 1 × 10−9 to 1 × 10−11 mol cm−2 s−1 as reported by Grigorii Soloveichik, program director of Advanced Research Projects Agency—Energy (ARPA-E).17 Hence, in this work, we describe our objective of electrochemical ammonia synthesis using origami-like Mo2C nano-flakes as NRR catalyst in a cell operating at 30 °C with Nafion 212 as the electrolyte and Pt/C as anode electrocatalyst. The origami structure, defined as a structure with lots of folded edges, may help to improve the electrocatalytic activity towards NRR.35 The durability of the Mo2C catalyst is also studied by continuous ammonia synthesis measurements for 50 h and a comparison of other molybdenum carbides tested in similar working environment is also provided.
Experimental
Synthesis and characterization of Mo2C
Origami structured Mo2C nanoflakes were prepared by a previously reported method.36 In short, molybdenum trioxide powder was placed over a Si substrate pre-coated with 10 nm Al and 1.5 nm Fe that was placed into a CVD furnace. Xylene (12 ml h−1) was passed over using Ar:H2 (85%:15%, flow rate 100 sccm) at 790 °C for 1.5 h. The purging speed of xylene solution was controlled by a syringe pump. Ar:H2 gas and xylene solution purged into the CVD furnace at the same time with separate lines as shown in the schematic in Fig. 1a. X-ray diffraction (XRD) measurements were carried out in a Siemens Diffractometer D5000 and thermogravimetric analysis (TGA) was done using a TGA Q50 TA Instruments at a heating rate of 10 °C/minute. For comparison to a baseline, we used a nanoscale MoC supplied to us by our collaborators from University of New Mexico (UNM) that was prepared by a previously reported procedure.37
Figure 1. (a) Schematic diagram of the Mo2C synthesis setup. (b) Schematic diagram of the ammonia synthesis setup.
Download figure:
Standard image High-resolution imageMembrane electrode assembly
Gas diffusion electrodes (GDEs) were fabricated by painting catalyst ink on Sigracet® 29BC (5 cm2, Gas diffusion layer, SGL carbon Inc.). The anode and cathode catalyst inks were prepared from 5 mg of Pt/C (60 wt% Pt on high surface area carbon, HiSPEC® 9100, Johnson Matthey, USA) and 5 mg of Mo2C nanoflakes catalyst (20 wt% Mo), respectively. Further, 18 mg of 5 wt% Nafion solution and 500 mg of 80:20 isopropanol-water solution was added to the catalyst followed by ultrasonication in a water bath for 1 h at room temperature. The well-dispersed catalyst ink was then brush painted on gas diffusion layers on a vacuum table at 60 °C. After painting, electrodes were left for 1 h on the vacuum table for drying out completely. Due to catalyst material losses associated with brush painting, the metal loading on the electrodes were measured by X-ray Fluorescence spectroscopy, which revealed a Pt loading of 0.6 mgPt cm−2 in the anode, and Mo2C loading of 0.1 mgMo cm−2 in the cathode. Nafion® 212 membrane was boiled in 0.5 M sulfuric acid solution to protonate the membrane followed by boiling in water. Prior to MEA assembly, the Nafion® membrane was sandwiched between GDEs and fixed in a fuel cell set-up to from a MEA without hot pressing.
Ammonia synthesis measurements
Electrochemical ammonia synthesis measurements were carried out in a fuel cell type setup shown in Fig. 1b by passing humidified ultra-high purity (UHP) 6% H2 balanced in N2 to the anode and dry UHP N2 to the cathode at a flow rate of 100 sccm. The low partial pressure 6% H2 was used to reduce the hydrogen redox kinetics since the currents involved in electrosynthesis measurements did not require pure H2. The potentials were applied using a PARSTAT 2273 Potentiostat/Galvanostat by Princeton Applied Research where working electrode and sensing electrode leads were connected to the nitrogen end while reference electrode and counter electrode leads were connected to the hydrogen end. The potentials applied in this study are −0.1, −0.25, −0.5 and −1.0 V. The cathode outlet was tested for ammonia by indophenol method. Impedance measurements were carried out between 1 MHz to 0.1 Hz at an amplitude of 10 mV while cyclic voltammetry measurements were carried out from 0.05 V to 1.0 V at a scanning rate of 50 mVs−1.
Indophenol measurements for ammonia quantification
Ammonia quantification was carried out using the indophenol method of Weatherburn and is summarized here.38 Solution A was prepared by dissolving 5 g of phenol in 250 ml of DI water and dissolving 25 mg of nitroprusside in 250 ml of DI water. These two solutions are mixed and stored in an amber bottle. Solution B was prepared by mixing 2.5 g of NaOH in 250 ml of DI water. To this 4.2 ml of sodium hypochlorite with 5% available chlorine was added and made up to 500 ml. This was stored in another amber bottle. Both the solutions are kept in a refrigerator when not in use. Indophenol measurements were carried out by two different methods. For short time sample collection, the cathodic outlet gas stream is directly passed through a mixture of 10 ml of Solution A and 10 ml of Solution B. The mixture is immediately measured via UV–vis to quantify ammonia. For calculating the background, 5 ml of Solution A and 5 ml of Solution B is mixed in a vial and left for the same amount of time (45 min) as the ammonia collection measurement and tested in absorbance through UV–vis spectra. For long duration sample collections, the cathodic outlet gas stream was passed through 30 ml of 0.5 M H2SO4, which in turn was tested for ammonia content. For this, 40 μl of the H2SO4 is mixed with 5 ml of Solution A and 5 ml of Solution B and heated in a water bath kept at 37.5 °C for 20 min followed by measuring UV–vis Spectra. The peak intensity of the background (Solution A + B) were subtracted from the sample peak intensity for concentration calculations. Solution A and B were prepared fresh and used within one week of their preparation.
Membrane ammonia exchange measurements
Acid treated Nafion® 212 membranes were thoroughly rinsed with DI-water in boiling conditions before exchanging with ammonia. The prepared membranes were soaked in 20% ammonia solutions for 16 h followed by boiling in 200 ml of DI water twice before storing them in DI water for further measurements. The amount of ammonia exchanged into the membrane was calculated by immersing the membrane in 10 ml of 0.5 M H2SO4 followed by sonication for 2 min. The membrane was removed from the sulfuric acid and 40 μl of the sulfuric acid solution was tested for ammonia by the indophenol method as described above.
Post mortem analysis of the membranes
To calculate the amount of ammonia exchanged with the Nafion® 212 membrane during ammonia synthesis measurements, the membrane was peeled off from the electrodes after the measurement and immersed in 10 ml of 0.5 M H2SO4 followed by sonication for 2 min. The sulfuric acid solution was tested for ammonia by the above described indophenol method.
Control measurements
The rates and efficiencies obtained in electrochemical ammonia synthesis measurements are often confounded by contributions from precursor materials, ammonia present in ambient air and heterogeneous catalysis.39 To avoid this, all the MEA components, Nafion® 212 membrane, anode and cathode are all pre-checked for any ammonia uptake from ambient atmosphere by indophenol method. Further, indophenol measurements were calibrated with a wide range of concentrations (0.001 M to 0.0001 M of (NH4)2SO4), that were further used for calculating the concentration of unknown solutions. Since our catalyst do not have a nitride structure, it rules out the possibility of ammonia evolving from the catalyst. Further, ammonia synthesis measurements were also carried out under Ar gas flow in the cathode for background measurements. Additionally, a cell with PtC anode, Mo2C cathode and N212 electrolyte was maintained under open circuit conditions for 24 h and the MEA was analyzed for ammonia uptake in post-mortem analysis.
XPS measurements
XPS data were collected using a Physical Electronics VersaProbe II system with a base pressure below 1 × 10−7 Pa at room temperature. A variable-size, monochromated Al kα X-ray source (1487 eV) was used throughout, and photoelectrons were energy sorted using a hemispherical analyzer. XP spectra are reported in terms of binding energy (BE) and instrument calibration was performed in accordance with ASTM procedure. Elemental composition was determined using survey scans at a pass energy of 117 eV. A pass energy of 29 eV was used for high-resolution scans to determine chemical valence state. Charge neutralization for insulating samples is accomplished by focusing low energy ions and electrons at the spot of X-ray impingement.
Results and Discussion
We fabricated an origami type interconnected molybdenum carbide using CVD techniques with a macrostructure that behaves like a porous sponge to assist adsorption of nitrogen for ammonia synthesis. The resulting powder had Mo2C in the hexagonal lattice structure with the space group P63/mmc along with excess carbon material whose peak is observed at a 2θ value of 25° (Fig. 2a). Whole profile peak fitting (JADE software) yielded a crystallite size of 12 nm. TGA in N2 did not show any significant change in weight up to 550 °C, indicating its suitability to operate under ammonia synthesis conditions over a wide temperature range (Fig. 2b). However, TGA analysis in air revealed that about 80 wt% of the prepared material was comprised of excess carbon as it oxidizes at around 500 °C. This excess carbon helped to improve the electronic conductivity as well as the overall surface area of the material, as a combined surface area (carbon + Mo2C) of 298 m2g−1 via Brunauer-Emmet-Teller (BET) analysis was observed for our catalyst in comparison to a reported surface area of 64 m2g−1 for Mo2C with a crystallite size of 10 nm with no carbon support.36,40 Nevertheless, the origami type Mo2C prepared here had larger particle sizes (15−25 nm) as seen in the TEM images in Fig. 2d due to aggregation of the crystallites to form the nanoflakes. An X-ray photoelectron spectroscopy (XPS) survey spectrum presented in Fig. 2c confirmed that the prepared molybdenum is in the carbide state as clearly indicated by a doublet peak at 228 eV. Further, there is no peak for either oxygen or nitrogen and only a strong carbon peak is observed indicating the absence of oxides and nitrides of molybdenum. The TEM image shown in Fig. 2d indicates the formation of nanoflakes while the HR-TEM image of the Mo2C in Fig. 2e clearly shows the interconnected structure with plenty of edges. Fig. 2f indicates the structure of a MoC supported in carbon supplied by UNM where spherical nanoparticles of ∼5 nm (without edges) were observed.
Figure 2. (a) XRD pattern obtained for as prepared Mo2C powders (b) TGA plots obtained with as prepared Mo2C under air and nitrogen atmospheres (c) XPS survey scan spectrum of Mo2C indicating only carbon and Mo are present with no evidence of oxygen or nitrogen, (d) TEM image indicating the Mo2C nanoflakes (e) HR-TEM image of Mo2C showing the edge like morphology and (f) HR-TEM image of MoC-C supplied by UNM.
Download figure:
Standard image High-resolution imageElectrochemical ammonia synthesis measurements were carried out in a typical fuel cell set-up where humidified 6% H2 balanced in N2 was supplied to the anode while ultra-high purity (UHP) dry nitrogen was supplied to the cathode. The conductivity of the Nafion® 212 membrane was comparatively lower [3 mS cm−1, Fig. 3a] owing to the partial humidification (only anode gas supply was purged through H2O kept at room temperature) of the supplied gases although it remained stable during the entire operation. Ammonia synthesis measurements on this cell (cell-1) were carried out by applying different potentials such as −0.1 V, −0.25 V, −0.5 V and −1.0 V for 45-minutes duration each, and the exhaust N2 line was passed through an indophenol solution mixture to measure the produced ammonia. Since any trace amount of water crossing over from the anode can interact with the ammonia trapping solution, the nitrogen outlet was first passed through a moisture trap before entering the ammonia trap solution. Ultraviolet-visible (UV–vis) spectroscopy results obtained in these measurements, presented in Fig. 3b, indicate that the maximum synthesis rate was observed at −0.5 V while ammonia synthesis reaction onset was observed at as low an over potential as −0.1 V. Applied potentials above −0.5 V did not increase the synthesis rate although the overall current increased indicating the increased preference for hydrogen evolution reaction over NRR below −0.5 V (Fig. 3c). This correlates well with previous reports on Mo2C that show Mo2C as a good catalyst for HER especially at higher over-potentials although inferior to Pt.36,41,42 At low over-potentials the currents reached stable values swiftly while at higher over-potentials, the double layer charging transient took longer duration. The NH3 measured in the outlet cathodic gas stream yielded a maximum rate of 7.7 × 10−12 mol cm−2 s−1 and a corresponding faradaic efficiency of 0.4%. However, Nafion membranes, being acidic in nature, interact with ammonia, a basic gas, resulting in some of the produced ammonia being exchanged with the protons of the sulfonic acid groups. Nafion's ability to react readily with ammonia was previously demonstrated when the performance of the Nafion based polymer electrolyte fuel cell was reported to drop significantly after exposure to ammonia concentrations as low as 48 ppm.43 Hence, after ammonia synthesis measurements, the Nafion 212 membrane was peeled off from the electrodes and analyzed for ammonia content by washing the membrane with 0.5 M H2SO4 and quantifying the ammonia in washed solution via the indophenol method. After 6 h of operation, the measured ammonia content of the membrane was found to be 1.1 × 10−7 mol cm−2 corresponding to an overall ammonia synthesis rate of 2.16 × 10−11 mol cm−2 s−1 with a faradaic efficiency of 1.8%. Further, we also ran a cell with Nafion 212 membrane pre-exchanged with ammonia with Pt/C as both anode and cathode to observe ammonia leaching from the electrolyte membrane but did not observe any ammonia in the cathodic gas stream resulting from either electro-synthesis or leaching from the membrane. Post-mortem analysis of this membrane revealed an ammonia content of 2.5 × 10−6 mol cm−2, indicating that the exchanged ammonia did not leave the membrane easily at these operating conditions.
Figure 3. (a) Impedance Nyquist plot obtained for the electrochemical ammonia synthesis cell (cell-1) in the frequency range of 1 MHz to 0.1 Hz with PtC as anode and Mo2C as cathode at 30 °C over a period of 6 h, (b) UV–vis plots of the Solution A + Solution B mixtures used for collecting ammonia at the cathode outlet under various applied potentials and (c) I-V plots obtained with Mo2C as cathode catalyst under various applied over-potentials.
Download figure:
Standard image High-resolution imageWe extended our study with Mo2C catalyst further by applying −0.5 V for extended periods on a fresh cell (Cell-2) and monitored the ammonia synthesis by trapping ammonia in the cathodic gas stream using 0.5 M sulfuric acid. This experiment revealed after post-mortem analysis a synthesis rate of 2 × 10−12 mol cm−2 s−1 and a faradaic efficiency of 0.4%. However, after approximately 50 h of operation, ammonia was no longer detected in the cathode gas stream. To diagnose the possible reason we monitored the cell before and after the ammonia synthesis measurements with impedance measurements and in situ cyclic voltammetry (Figs. 4a and 4b). Impedance Nyquist plots indicate no drop in membrane conductivity while the charge transfer resistance associated with the low frequency semi-circle increased significantly. Further, the CV data observed on the ammonia synthesis cell at the start and after 50 h of operation clearly indicates a change in the catalyst surface as the 3-redox couple behavior obtained at the start of the measurement changed into a simple double layer charging behavior indicating either a loss of catalytic activity or detachment of the Mo2C catalyst. Mo2C has been reported to go through a similar 3-redox couple process although the nature of redox couple is not clear.44,45 Nevertheless, ammonia synthesis measurements using spherical MoC catalyst, supplied by UNM, as cathode did not provide any ammonia production as can be seen from Fig. 4c clearly indicating the advantage of the origami-inspired Mo2C catalyst.
Figure 4. (a) Impedance Nyquist plot obtained on cell-2 (PtC∣N212∣Mo2C) in the frequency range of 1 MHz to 0.1 Hz at different times during ammonia synthesis measurements. (b) in situ cyclic voltammetry plots observed for Mo2C during ammonia synthesis measurements at the start of measurements and after 50 h of operation. (c) UV–vis plot obtained for MoC catalysts supplied by UNM where no ammonia production was observed over the entire potential window.
Download figure:
Standard image High-resolution imageTo further understand the changes in the surface morphology of the Mo2C catalyst, we collected XPS data on the Mo2C electrode before and after (7 h and 50 h) the ammonia synthesis measurements (Figs. 5a–5f). Mo2C and MoN are known to oxidize to form MoO3 in the presence of oxygen and in the presence of moisture. XPS results on Mo2C further reiterate this as the Mo 3d peak intensity at 232 eV, that could be attributed to MoOx or molybdenum oxycarbide, has increased in the catalyst with increasing time under ammonia synthesis measurements while the peak corresponding to Mo2C at 228 eV decreased in intensity. A corresponding small increase in oxygen peak intensity is observed in the tested catalyst while no change is observed in the peaks associated with carbon (Figs. 5a and 5b). While the cathode is maintained under an inert or reducing environment (N2, Ar and cross-over hydrogen), the cross-over water from anode humidification to the cathode seems to have resulted in the oxidation of the Mo2C catalyst. Mo2C has been reported to form molybdenum oxycarbide upon exposure to water or oxygen and the increase in peak position associated with Mo6+ and Mo4+ in XPS indicates that the cross-over water under electrochemical cycling is enough to oxide the Mo2C.46,47 Further, no peaks for nitrogen were observed during the measurements suggesting either no nitride was formed during the ammonia synthesis measurements or that any nitride formed was instantaneously oxidized (Fig. 5c). From Fig. 5f, it is clear that most of the surface Mo2C is converted to either MoOx or molybdenum oxycarbide vindicating the observation of activity loss towards NRR after 50 h of operation.
Figure 5. XPS spectra obtained on Mo2C electrode before and after electrosynthesis measurements for (a) C1s (b) O1s (c) N1s; Molybdenum 3d XPS spectra obtained for Mo2C electrode (d) before ammonia synthesis measurements (e) after 7 h of ammonia synthesis measurements (f) after 50 h of ammonia synthesis measurements.
Download figure:
Standard image High-resolution imageThe catalytic activity of the Mo2C nano-flakes is attributed to the edges associated with the origami-like structure coupled with higher surface area due to excess carbon (Figs. 6a and 6b). We believe that this microporous morphology is advantageous for nitrogen reduction reaction as it brings the reactant close to the catalyst. The hexagonal Mo2C has been reported to have better catalytic activity towards ammonia decomposition at higher temperatures (at 600 °C) where the catalyst is completely converted to the hexagonal nitride MoN, however, no report is available on its ability to produce ammonia at low to intermediate temperatures.48 Nevertheless, in order for the ammonia decomposition reaction to be effective, the nitrogen atoms have to go through a re-combinative desorption process that suggests the ability of this surface to interact with nitrogen. Further, it was reported that the active site for ammonia synthesis might be a single step edge while the required hydrogen populates the terraces of metal particles in a highly mobile form (Fig. 6a).49 A recent DFT study on cubic MoC indicates that all crystallographic surfaces have high activity for both nitrogen and hydrogen adsorption and dissociation although the hexagonal Mo2C is not included.34 However, DFT studies on Mo2C by Zheng et al.48 indicate that the edge sites in Mo2C acts as the adsorption and desorption sites for ammonia decomposition while Ivana et al.50 reported theoretical calculations that indicate NRR activity for Mo2C. Hence we do not attempt a separate theoretical study on Mo2C here. However, we are the first to experimentally report Mo2C catalyst for electrochemical ammonia synthesis. This result is especially important considering that nanoscale (5 nm spherical particles) MoC did not produce any ammonia under identical conditions. While our synthesis rate of 2.16 × 10−11 mol cm−2 s−1 is still several orders of magnitude lower than what is required for commercialization or the highest reported rates,16 it still compares favourably with some recent reports. While a synthesis rate of 15 mgm−2 h−1 (2.31 × 10−11 mol cm−2 s−1) with a faradaic efficiency of 45% was recently reported for a Fe nanoparticle catalyst, it involves complex operation procedure containing ionic liquids that require careful removal of water vapour before ammonia synthesis measurements that may not be suitable for scale-up.51 A specially designed vanadium nitride (VN) catalyst has produced ammonia at 1.6% faradaic efficiency at 80 °C for 116 h, though this material also suffers from oxide formation.52 In comparison, our methodology uses a simple fuel cell type single cell setup that is readily available and produces a synthesis rate of 2.16 × 10−11 mol cm−2 h−1 with a faradaic efficiency of 1.8%. Nevertheless, the Mo2C catalyst used in this work as well as the vanadium nitride catalyst reported by Yang et al.,52 both suffer activity loss due to catalyst oxidation that indicates the propensity of the transition metal carbides and nitrides to oxidize if strict environmental control is not employed. A detailed analysis of VN catalyst stability by Manjunatha et al., indicates that depending upon the operating pH, VN catalyst formed different vanadium oxide ions such as vanadyl (VO2+), cyclic vanadate (c-V4O124−) and vanadate (VO43−).53 While our measurements were carried out under solid-gas interfaces with no direct immersion in acidic or alkaline liquid electrolytes, exposure to water vapour is seemingly responsible for the oxidation of Mo2C to higher oxidation states. While our results are consistent with the previous report, unlike the experiments on VN, our experiments were not performed in conditions where the dissolution of the catalyst is a possibility. These results also indicate that future research needs to focus on studying Mo2C stability with respect to pH and operating temperature to better understand the ideal operating conditions and improve the durability of catalysts for NRR.
Figure 6. (a) Scheme suggesting the nitrogen at edge sites combining with protons and electrons from flat surface sites to produce ammonia on Mo2C (b) HR-TEM image indicating the origami structure of Mo2C (c) TEM image of as-prepared Mo2C nanoflakes showing a thin layer of carbon cover (d) TEM image of Mo2C nanoflakes after ammonia synthesis measurements indicating no significant change in the carbon layer coverage.
Download figure:
Standard image High-resolution imageRegarding the mechanism of NRR on Mo2C catalysts, the mode of adsorption is not clear. However, experimental evidence indicates that there is no formation of a nitride intermediate as no indication of nitrogen in XPS measurements unlike the MoN formation reported during ammonia decomposition reaction with hexagonal Mo2C catalysts.48 Further, the XPS measurements taken at various stages of ammonia synthesis measurements indicate a transformation to an oxygen rich molybdenum structure (oxide or oxycarbide) as evidenced by a Mo 3d peak shift to higher binding energy, which is also confirmed by the increase in oxygen peak before and after testing (Fig. 5b). EDS measurements taken during TEM measurements further confirm this (data not shown), as no evidence of nitrogen is observed on the tested samples. The carbon layer associated with Mo2C also does not undergo any change as TEM images before and after ammonia synthesis measurements show no morphological changes (Figs. 6c and 6d). The MoN formation from Mo2C reported is a slow process even at a temperature of 600 °C, therefore forming MoN at room temperature operation seems highly unlikely, and we exclude this mode of the reaction mechanism.48 Further, the trace amount of oxygen in the UHP nitrogen and water crossover from the anode supply of humidified UHP hydrogen is likely to enable Mo2C oxidation. Therefore having an electrolyte with significant conductivity under dry conditions may limit catalyst deactivation. Additionally, increasing the operating temperature may further enhance the catalytic activity. While ammonia synthesis measurements using solution state electrochemistry with non-aqueous electrolyte systems are reported to produce high faradaic efficiency albeit similar synthesis rates,54 methodologies that rely on fuel cell type set up with continuous flow of nitrogen and gas phase reaction at the interface tend to produce very low faradaic efficiencies. Despite this, our measurements with a fuel cell type has shown a higher faradaic efficiency of 1.8% among similar synthesis methodologies. Table I compares some of the recently reported ammonia synthesis rates relevant to this work while a complete list of ammonia synthesis rates is detailed by Martin et al.55
Table I. Rates reported in some recent publications with as reported units and their converted units to traditional units (mol cm−2 s−1).
| Catalyst | Rate with units as reported | Rate (mol cm−2 s−1) | Year | Reference |
|---|---|---|---|---|
| Fe/SS | 14 mg m−2 h−1 | 2.16 × 10−11 | 2017 | Energy Environ. Sci.,2017, 10, 2516 |
| Hollow Au nanocages | 3.9 μg cm−2 h−1 | 6.01 × 10−11 | 2018 | Nano Energy 49 (2018) 316–323 |
| Pd/C | 4.5 μg mg−1Pd h−1 | 2.08 × 10−11 | 2018 | Nature Communications, 2018 9, 1795 |
| Fe-N-C | 7.48 μg mg−1 h−1 | 1.15 × 10−10 | 2019 | Nature Communications 2019, 10, 341 |
| Fe2O3/TiO2/C | 2.70 × 10−10 mol mg−1 s−1 | 2.70 × 10−9 | 2019 | ACS Appl. Mater. Interfaces 2019, 11, 7981−7989 |
| Mo2C | 1.4 μg cm−2 h−1 | 2.16 × 10−11 | 2019 | This work |
Control Measurements
Despite increasing research interest in electrochemical ammonia synthesis, many reported results are hard to reproduce and also further complicated by ammonia adsorption from ambient conditions. Recently two reviews summarizes the necessary control measurements required to confirm the electrochemical ammonia synthesis especially the Ar based control measurement and 15N2 isotope label measurement.39,56 We carried out control measurements with Ar gas supply where we found no evidence of ammonia formation (Fig. 7a). Further, we fabricated an identical electrochemical cell as the one used for ammonia synthesis and operated for 24 h under open circuit conditions followed by post-mortem analysis of the MEA for any ammonia uptake. We did not observe any ammonia formation either on the outlet or on the membrane implying that ammonia is produced only under applied potential biases with nitrogen supplied at the cathode and not via heterogeneous reaction or via absorption from ambient atmosphere. Since our catalyst, Mo2C, does not have any inherent nitrogen content as evidenced by XPS measurements, it avoids the possibility of ammonia evolving from catalyst nitrogen source. Further, the total ammonia produced over 50 h of operation is about 5 times higher than that of possible ammonia production if the exposed Mo2C produced ammonia through Mo2N formation indicating catalytic production of ammonia. The next part of control measurement is demonstrating 15N isotope labelling measurements. To this purpose, we measured 1H NMR spectra from a 500 MHz Bruker Avance NMR spectrometer at room temperature for ammonium sulphate samples in 90:10 H2O-D2O mixture with total of 200 scans, and chemical shifts were referenced to the HDO residue peaks from H2O-D2O at 4.7 ppm. We measured 1H-NMR spectra for various concentrations (1 μg to 1 mg) of commercially available 15N ammonium sulfate since 15N give a doublet at a chemical shift of 7.0 ppm while 14N shows a triplet.38 As shown in Fig. 7b, the doublet corresponding to 15N is not observed for concentrations below 0.1 mg. The total produced ammonia after 50 h of operation is 0.032 mg, hence, detecting or quantifying it through NMR is challenging. Further, at a flow rate of 100 sccm, the total 15N2 required for 50 h operation would be 300 l, which is prohibitively expensive. Also, New Mexico, USA has one of the least reported levels of ammonia in ambient conditions (0.3–0.5 ppb) and do not provide measurable contributions during measurements as observed from our control measurements.57 The produced ammonia is quantified using indophenol method. Various known concentrations of ammonium sulphate were prepared and UV–vis plot obtained for them is given in Fig. 7c and the calibration plot obtained for the peaks at an absorbance frequency of 636 nm is given in Fig. 7d indicating the suitability of this methodology to quantify ammonia concentrations reported in this study.
Figure 7. (a) UV–vis plot obtained under different cathode gas supply and bias conditions, (b) 1H-NMR obtained for various amount of commercial 15N-ammonium sulphate dissolved in 1 ml of 90:10 H2O/D2O mixture, (c) UV–vis plots obtained for various concentrations of ammonium sulphate calibration solutions, (d) Standardization curve obtained for the calibration solutions at an absorbance wavelength of 636 nm.
Download figure:
Standard image High-resolution image
Conclusions
We have successfully demonstrated electrochemical synthesis of ammonia using Mo2C as NRR catalyst using Nafion-212 membrane as the electrolyte at 30 °C with selective humidification of the anode. While regular MoC structures did not produce any ammonia, the origami-like Mo2C produced ammonia at a maximum synthesis rate of 2.16 × 10−11 mol cm−2 s−1 with a corresponding faradaic efficiency of 1.8%. This provides the first experimental evidence to support DFT calculations showing high NRR catalytic activity of Mo2C edge sites and provides guidelines for the design of special structures to improve NRR catalyst activity. Nafion, an acidic proton conductor, is not the ideal choice for electrolyte in this application as it readily absorbs the produced ammonia, although it has some utility for testing new catalysts for short durations. Additionally, the need for humidification makes it an unsuitable electrolyte for catalysts such as Mo2C as they react with surface-bound water to form oxides or oxycarbides of Mo, which is catalytically inactive towards NRR. Nevertheless, our catalyst is the first of its kind to produce ammonia for close to 50 h of continuous operation, indicates a possible direction for future catalyst design, and holds promise for operation under hot and dry conditions.
Acknowledgments
The authors thank ARPA-E for funding this work (Award ID - DE-AR0000687) and Triad National Security, LLC, Operator of Los Alamos National Laboratory under U.S. Department of Energy Contract Number 89233218CNA000001 for providing the lab space to carry out the research work.