Ammonia synthesis by N2 and steam electrolysis in molten hydroxide suspensions of nanoscale Fe2O3

See allHide authors and affiliations

Science  08 Aug 2014:
Vol. 345, Issue 6197, pp. 637-640
DOI: 10.1126/science.1254234

Taking carbon out of the ammonial recipe

The reaction used to make ammonia for synthetic fertilizer requires hydrogen. Nowadays, that hydrogen is stripped from methane, creating CO2 as a by-product. Licht et al. demonstrate a relatively efficient electrochemical process in which water and nitrogen react directly to form ammonia. The approach removes the need for an independent hydrogen generation step. The process takes place in molten hydroxide salt and requires a nanostructured iron oxide–derived catalyst. Although the catalyst suspension is currently only stable for a few hours, the protocol points to a way to produce ammonia from purely renewable resources.

Science, this issue p. 637


The Haber-Bosch process to produce ammonia for fertilizer currently relies on carbon-intensive steam reforming of methane as a hydrogen source. We present an electrochemical pathway in which ammonia is produced by electrolysis of air and steam in a molten hydroxide suspension of nano-Fe2O3. At 200°C in an electrolyte with a molar ratio of 0.5 NaOH/0.5 KOH, ammonia is produced at 1.2 volts (V) under 2 milliamperes per centimeter squared (mA cm−2) of applied current at coulombic efficiency of 35% (35% of the applied current results in the six-electron conversion of N2 and water to ammonia, and excess H2 is cogenerated with the ammonia). At 250°C and 25 bar of steam pressure, the electrolysis voltage necessary for 2 mA cm−2 current density decreased to 1.0 V.

The Haber-Bosch process annually hydrogenates over 120 million metric tons of N2 from the atmosphere (1, 2) to produce ammonia for fertilizer (Eq. 1) (3). Today, hydrogen for ammonia synthesis is produced primarily through steam reformation, which consumes 3 to 5% of the world’s natural gas production and releases large quantities of CO2 to the atmosphere (1):N2 + 3H2 → 2NH3 (1)CH4 + 2H2O → 4H2 + CO2 (2)The ammonia hydrogenation reaction is separate from the steam-reforming reaction (Eq. 2) that generates the hydrogen. Renewable energy–driven water splitting could provide an alternative H2 source, but economic, non–CO2-emitting sources of H2 have yet to be proven on the industrial scale. Although ammonia hydrogenation is exothermic, it is kinetically disfavored at ambient temperature and pressure. In the Haber-Bosch process, this kinetic limitation is overcome via an iron-based catalyst, repeated cycling, high pressure, and elevated temperature. The last-named conditions are energy-intensive and consume 2% of the world’s energy production (1).

Several electrochemical processes have been pursued to provide alternative syntheses of NH3. The electrolytic formation of nitrides from nitrogen gas in molten alkali chlorides was studied before 1980 (4, 5), including subsequent reactions with hydrogen to yield ammonia, but such studies have not led yet to commercial production of ammonia because of challenges including the competing back-reaction of nitride to nitrogen (610). In 1985, room-temperature electrolytic synthesis of ammonia was introduced, albeit at low rates, via protolysis of W(N2)2(PMe2Ph)4 (11). Ammonia was synthesized from H2 and N2 in aqueous sulfate solutions using metal-phthalocyanine (C32H18N8) complexes loaded into carbon black as a cathode and Pt loaded into carbon black as an anode, again at a low rate; note that tin, rather than iron, phthalocyanines exhibited the highest efficiency and stability (12). Density functional calculations have been used to evaluate hydrogen and nitrogen adsorption and/or reduction on a variety of transition metals as possible electrocatalysts in ammonia formation (13). The solid-state electrochemical synthesis of ammonia in two-compartment cells with either proton or oxide ion conductors has been reviewed, with the highest rate reported at 80°C and 2 V using a Nafion membrane (14).

There are few reports in the literature about using water or steam as a reactant in lieu of hydrogen for the electrosynthesis of ammonia, as one means to avoid the CO2 emissions of Eq. 2. The rate of ammonia formation is lower by several orders of magnitude, and the coulombic efficiency drops to less than 1%, when water, rather than H2, is used (14). One study used a strontia-ceria-ytterbia oxide proton-conducting solid electrolyte at 450° to 700°C and a Ru-based catalyst but reported that the conversions with respect to nitrogen or steam were low, primarily because of the poor conductivity of the working electrode (15). Using a Nafion separator in aqueous 2 M KOH with a Ru on C cathode enabled ammonia synthesis from water and nitrogen at a rate and maximum coulombic efficiency of 2.8 × 10−12 mol NH3 s−1 cm−2 and 0.9% at 20°C and, at 90°C, a maximum rate of 2.1 × 10−11 mol s−1 cm−2 at 0.2% efficiency (16). Using Pt/C on a gas diffusion layer at both electrodes and room temperature Nafion as the electrolyte yielded NH3 at a higher rate of 1.1 × 10−9 mol s−1 cm−2, which consumed water at the anode and air at the cathode at 0.6% coulombic efficiency (17). Recently, ammonia was formed at 0.8% coulombic efficiency and a similar rate of 0.9 × 10−10 mol s−1 cm−2 by using Pt/C electrodes and a Nafion membrane (18). The related literature for alkaline electrochemical water splitting (1921) is substantially larger than for water-air ammonia electrosynthesis. We became intrigued by a molten hydroxide (a NaOH-KOH eutectic electrolyte) ammonia fuel cell (22) in which NH3-gas (as fuel) and air reacted to produce electricity and unspecified products. A subsequent paper used the NaOH-KOH eutectic cell to split water (23) into hydrogen and oxygen in a manner similar to our earlier molten NaOH water splitting protocol (24):H2O → H2 + 1/2O2 (3)By effectively reversing the NH3 fuel cell, we present an electrochemical pathway to produce ammonia from air and steam at 200°C with simple materials (molten hydroxide, Ni electrodes, and nano-Fe2O3), in one pot without a separator. We reasoned that combining Eq. 3 with Eq. 1 in a highly conductive molten hydroxide for the hydrogen generation, in the presence of an appropriate ammonia-generating catalyst such as iron, should provide a one-pot medium for the electrolytic synthesis of ammonia from air and water:

xN2 + yH2O (6x + y)e-/h+, iron/iron oxide, molten hydroxide) > 2xNH3 + (y – 3x)H2 + y/2O2 (4)

The thermodynamic potentials for water splitting and the reaction of nitrogen with water are plotted in Fig. 1; they exhibit similar redox potentials at room temperature. We also calculated alternative ammonia-producing reactions, such as the reaction of N2 with several water molecules, shown in Fig. 1 that could occur in a molten hydroxide medium. In each case, the electrolysis potential is calculated from the known temperature variation of the entropy and enthalpies of the reactants and products by using the convention to describe the positive potential necessary to drive a nonspontaneous potential, ET = ΔG(T)/nF (2528), where the Gibbs free energy is calculated at a temperature (T) and n is the electrons transferred and F is Faraday’s constant. On the basis of these calculations, when three H2O molecules, rather than H2, act as the hydrogen source for the NH3 (Eq. 5), the potential decreases up to the water boiling point, but then increases with temperature.

Fig. 1 Thermodynamic electrolysis potentials for water splitting and for water-based ammonia syntheses.

Calculations of potentials are based on the temperature variation of the individual species’ thermochemical data. The upper curves are calculated at unit activity, whereas the lower curve is calculated at a high ratio of reactants to products given by Q = 4 = log(10,000), that is aNH32aO23/2 /aN2aH2O3 = 10,000. Electrolysis provides control of the relative amounts of reactant and generated product in a system. A substantial activity differential (Q > 1) can also lower E(V) to drive STEP improvement at elevated temperature.

N2 + 3H2O → 2NH3 + 3/2O2; E(25°C, 100°C, or 750°C) = 1.18 V, 1.13 V, or 1.22 V (5)

With more than three H2O equivalents per N2, as in Eq. 6, hydrogen is cogenerated as a product along with ammonia, and as seen in Fig. 1, the electrolysis potential decreases with increasing water.

N2 + 10H2O → 2NH3 + 5O2 + 7H2; E(25°C, 100°C, or 750°C) = 1.21 V, 1.15 V, or 1.06 V (6)

The free energy and, hence, potential variation with activity, a, of the reaction isΔG(T,a) = ΔG°(T) + RTln[∏i=1 to x a(Ri)ri / ∏i=1 to y a(Ci)ci ]= ΔG°(T) + 2.303RTQ; where Q = log[(∏i=1 to x a(Ri)ri / ∏i=1 to y a(Ci)ci] (7)R is the gas constant, T is the Kelvin temperature, and i and Ci represent the products and reactants. Q in Eq. 7 combines the log of the reaction quotient and n into a single term to assess the magnitude of this Nernst effect and reflects the relative activity of the reactants compared with the products. Reaction 5 yields Q = log(aNH32aO23/2/aN2aH2O3)/6. As shown in Fig. 1, with appropriate choice of medium, this Eq. 7 Nernst effect can generate a dramatic energy decrease in the required electrolysis potential, for example, in molten electrolytes when the water reactant and nitrogen concentrations are high and the product concentration is relatively low. The effect is enhanced proportionally to the relative increases in the Kelvin temperature. Compared with Q = 1 in Eq. 5:

Q = 4: N2 + 3H2O → 2NH3 +3/2O2; E(25°C, 100°C, or 750°C) = 0.94 V, 0.83 V, or 0.39 V (8)

As measured at 200°C, the molten hydroxide electrolyzer efficiently splits water, bubbled in as steam. As expected, H2 was produced at a 2:1 ratio to O2 when the electrolyzer did not contain nitrogen or an effective ammonia generation surface. A range of LiOH, NaOH, KOH, and CsOH eutectic mix (29) electrolytes (such as a molar ratio of 0.5 NaOH/0.5 KOH) were effective for water electrolysis below 300°C. The pure alkali hydroxides each melt only at temperatures above 300°C. Based on common materials, the NaOH-KOH eutectic is of particular interest and melts at 170°C. At 200°C, this electrolyte approached 100% of the electrolysis efficiency for water splitting. The variation of the water-splitting electrolysis voltage as a function of current density and temperature at 1 atmosphere was measured between planar Ni electrodes and is presented in Fig. 2. Alternative, textured, and/or alloyed electrodes and modifications of the cell configuration have been widely studied (1921) and can decrease the electrolysis voltage at higher water-splitting current densities.

Fig. 2 Measured electrolysis potential.

The measured electrolysis potential at 200°C of a molar molten mix of 0.5 NaOH/0.5 KOH at 1 atmosphere between two 2.5- by 1.5-cm planar nickel electrodes. Steam is bubbled into the cell to saturate each electrolyte with water. For example, the 200°C molten NaOH-KOH electrolyte contains ~8% weight water. At 200°C, the coulombic water-splitting efficiency approaches 100% as measured for currents of 25 to 1000 mA cm−2. Water-splitting coulombic efficiency in the open air drops with increasing temperature as the molten electrolyte dehydrates with increasing temperature. At 300°C, the coulombic efficiency has dropped to ~90%, and by 500°C, the coulombic efficiency has decreased to 25%. Note that measured potentials are ~0.1 V lower when lithiated Ni anodes and monel mesh cathodes are used in lieu of planar nickel (30).

Experimentally, we observed high rates of ammonia generation when the 200°C molten hydroxide (NaOH-KOH) electrolyte was mixed with high–surface area Fe2O3 to provide iron as a reactive surface and when nitrogen and water vapor were in the cell. The medium was electrolyzed between a planar nickel anode and a mesh nickel–monel cathode. Initially, the H2-evolving mesh cathode had been used to enclose the iron oxide, but the mesh openings were too large to contain the nano-Fe2O3. Therefore, nano-Fe2O3 was simply added to the electrolyte. Both water-saturated nitrogen and CO2-scrubbed air (bubbled through a 1 M NaOH solution to remove CO2) yielded similar efficiencies of ammonia generation. In lieu of air, 99.999% nitrogen was saturated with water at room temperature by bubbling the nitrogen through doubly deionized water en route to the electrolyzer. Unlike the water-splitting electrolysis, the efficiency of the ammonia generation by electrolysis was lower at higher current densities.

Ammonia generation by electrolysis here refers to the global reaction of nitrogen, water, and electrons to form ammonia (and oxygen), and the efficiency is calculated based on the moles of electrons consumed compared with the equivalents of ammonia (3e/NH3) generated. This efficiency was >30% at 20 mA through 10 cm2 electrodes compared with ~7% at 250 mA. This suggests that the conversion efficiency is not limited by the available hydrogen, but rather by the available surface area of the nano-Fe2O3 to promote the nitrogen and hydrogen conversion to ammonia.

The measured efficiency of ammonia evolution in time in 200°C NaOH-KOH molten electrolyte under a variety conditions is shown in Fig. 3, including a constant current of either 0 or 20 mA between the 10 cm2 Ni electrodes. Evolved ammonia was collected and measured in a room-temperature water trap. The constant current of electrolysis was measured. The three electron equivalents of ammonia, measured as described in the supplementary material, were divided by this integrated electrolysis charge to determine the electrolysis efficiency to synthesize ammonia. The electrolysis efficiency to produce ammonia was high, in excess of 30% when we used either wet air or wet nitrogen reactants and with nano-Fe2O3 to provide iron as a reactive surface (30). This observed >30% efficiency for the conversion of electrons, air, and water to ammonia compares with the highest values of <1% previously noted for the generation of ammonia from air or nitrogen and water (1618). The cogeneration of H2, as measured by a Micro IV hydrogen analyzer (GfG Instrumentation) is consistent with the remaining (~65%) electrolysis current (30). At 20 mA applied current, the cogeneration of ammonia and hydrogen is consistent with the net reaction: N2 +8H2O → 2NH3 +4O2 +6H2, but, as will be shown, the ratio of H2 to NH3 grows with increasing applied current. As seen in Fig. 3, ammonia is not generated if current is not applied. Iron oxide–iron mixes are catalysts for the traditional chemical synthesis of ammonia. The high surface area of the nano-Fe2O3 in the new electrochemical synthesis appears critical to the process. As seen in Fig. 3, the cell with no Fe2O3, or conventional (99.4%, J.T. Baker), rather than nanoscopic, Fe2O3, did not generate discernible ammonia. The 20- to 40-nm Fe2O3 remains colloidal throughout the electrolysis, whereas the conventional Fe2O3 descends and collects at the bottom of the electrolysis cell. After we milled conventional Fe2O3 at 300 rpm for 2 hours in a Retsch PM100 ball mill, discernible ammonia still was not generated during electrolysis, and the Fe2O3 still collected at the bottom of the cell. However, ball milling at 600 rpm for 2 hours, which decreased the particle size to ~200 nm, sustained a colloidal suspension throughout the electrolysis and generated a small, but discernible, quantity of ammonia at ~0.03 times the rate of the 20- to 40-nm Fe2O3 electrolysis cell. As delineated in the supplementary materials, when the 20- to 40-nm Fe2O3 was placed above (in the headspace), rather than in, the electrolyte, ammonia was not generated. Ammonia also was not generated when 20- to 40-nm Fe2O3 was first heated under argon to desorb any nitrogen before its addition to the molten electrolyte, and when argon (saturated with water vapor), rather than nitrogen, was bubbled into the cell during the electrolysis. However, as seen in Fig. 3, when wet (water-saturated) argon, rather than nitrogen, was bubbled into the cell, and nano-Fe2O3 without this desorption pretreatment was added to the electrolyte, a low level of ammonia was initially generated until nitrogen, evidently preadsorbed onto the nano-Fe2O3, was depleted.

Fig. 3 Efficiency of current conversion of ammonia product.

Experimental quantification of ammonia from air or nitrogen, either saturated with water or dry, by one-pot synthesis. The input gas was added to molten hydroxide at 200°C and electrolyzed in the presence of nano- or micron-sized Fe2O3. The indicated constant current was applied between 10 cm2 Ni electrodes. The product gas was bubbled through a water trap quantitatively analyzed for ammonia and compared with the applied, integrated electrolysis charge to determine the electrolysis efficiency.

Ammonia was also initially generated under conditions without water vapor in the nitrogen. However, as seen in Fig. 3, this ammonia production tapered off rapidly in time. This appears to be consistent with consumption of the molten hydroxide as an alternate source of water:2MOH → M2O + H2O (M = alkali) (9)In accord with Eq. 9, ammonia production continued until the molten hydroxide became dehydrated and oxide enriched.

The full cell voltage to drive molten hydroxide electrolysis of wet nitrogen or air to ammonia at 200°C in the presence of nano-Fe2O3 was 1.23 (±0.02) V when the applied current was 20 mA between the 10 cm2 Ni electrodes (2 mA cm−2) in the molten NaOH-KOH electrolyte; it increased to 1.44 (±0.02) V when the current increased to 250 mA (25 mA cm−2), and then to 2.4 V for 2000 mA (200 mA cm−2). At 2 mA cm−2, under these conditions, ammonia evolved at a rate of 2.4 × 10−9 mol s−1 cm−2, and hydrogen was cosynthesized at 6.6 × 10−9 mol s−1 cm−2. At 25 mA cm−2 the ammonia synthesis rate increased to 6.7 × 10−9 mol NH3 s−1 cm−2, and the rate of hydrogen coproduction increased to 1.0 × 10−7 mol H2 s−1 cm−2. Hence, with increasing current density, the observed produced ratio of H2 to NH3 increased, that is the hydrogen coproduction rate increased more rapidly than the ammonia production rate, and the global (NH3 + H2) coulombic efficiency remained high at current densities of both 2 and 25 mA cm −2. At the highest current density of 200 mA cm−2, the ammonia production rate was 1.0 × 10−8 mol NH3 s−1 cm−2. Cesium has been observed to enhance the catalytic activity of iron and iron oxides in ammonia production (31, 32). At 200°C when the NaOH-KOH electrolyte was replaced by a CsOH electrolyte (containing 30 wt% water to maintain the liquid-molten state at 200°C) the 25 mA cm−2 rate increased from 6.7 × 10−9 to 7.7 × 10−9 mol NH3 s−1 cm−2. At the measured current densities and temperature, the variation of the electrolyte cation has not been observed to affect the electrolysis potentials. The same electrolysis potentials were observed when the 0.5:0.5 NaOH-KOH molten electrolyte was replaced by a 0.48:0.52 molar ratio NaOH-CsOH, a 0.7:0.3 NaOH-LiOH, or a 0.7:0.3 KOH-LiOH eutectic electrolyte. Presumably, higher current densities and higher temperature potential will be affected by the cation, which will affect electrolyte conductivity and water retention.

The energy consumption of this ambient pressure process varies with rate (which affects the voltage) and the ammonia-only or global current efficiency (including both ammonia and hydrogen). Hydrogen cogenerated with the ammonia is not lost energy and is available for storage, or use as a fuel, or as a chemical reactant. As a first estimate of the energy consumption, we use the 1.2 V at 2 mA cm−2 (= 2.4 × 10−9 cm−2 MJ s−1). This produces 2.4 × 10−9 s−1 cm−2 mol NH3 (= 4.1 × 10−11 s−1 cm−2 kg NH3); that is, 59 MJ is consumed per kg NH3 produced. The energy content of the recovered H2 is 120 to 142 MJ/kg (with or without the heat of water vaporization). This energy consumption estimate does not include the associated engineering losses, or the energy of heating to 200°C, as the optimal operating temperature needs to be determined.

In the absence of the nano-Fe2O3, water is simply electrolyzed into hydrogen at the cathode and oxygen at the anode in the 200°C molten hydroxide chamber. In the presence of nano-Fe2O3, two alternative mechanisms of the ammonia synthesis can be considered. In the first, electrochemical reduction of water to hydrogen occurs at the cathode, which then diffuses to react with adsorbed nitrogen on the nano-Fe2O3 surface to form ammonia. An alternative mechanism to consider is the electrochemical reduction of nitrogen and water at the nano-Fe2O3 to form ammonia. The latter mechanism would necessitate electron transfer from the nickel cathode to the dispersed, electrolyte-suspended nano-Fe2O3. The latter mechanism could be ruled out in the 200°C molten hydroxide electrolyte with the high–surface area Fe2O3 if ammonia were to be formed when no electrochemical current was applied, as in the case where H2 and N2 (rather than H2O and N2) were instead added as chemical reactants. That situation would preclude the electrochemical reduction of nitrogen and water at the suspended Fe2O3 and yet facilitate ammonia formation. It was interesting that little or no ammonia was formed in this case when gas phase H2 and N2, with or without O2, were introduced to this cell without current. As one experiment, H2 was used as the inlet reactant (with O2, N2, and H2O) after generation by room-temperature electrolysis at 2000 mA in an aqueous solution of 4 m NaOH and 4 KOH. This generated gas containing 2:1 H2 to O2, and forming 15 ml min−1 of H2, was mixed with 5 ml min−1 of water saturated N2 as the inlet gas. This procedure converts the one-pot synthetic chamber into a two-pot chamber (in which hydrogen is formed by electrolysis in the preliminary room-temperature pot and bubbled into the second molten electrolyte pot through the nickel tube and nickel mesh). In this case, ammonia was formed at a marginal, but discernible, rate compared with that observed when the 2000 mA of current was applied directly to the 200°C electrolysis chamber (0.2 × 10−9 versus 1.0 × 10−8 mol NH3 s−1 cm−2). As a second experiment, without any electrolysis, 15 ml/min pure H2 and 5 ml/min pure N2 gases were bubbled through water and used as the inlet gas. Again, ammonia was only formed at a marginal, but discernible, rate (0.2 × 10−9 versus 1.0 × 10−8 mol NH3 s−1 cm−2). We posit that these experiments provide supporting evidence that the second mechanism (electrochemical reduction of the nitrogen and water at the nano-Fe2O3) of ammonia synthesis dominates. However, alternative factors such as a (smaller, more reactive) H2 bubble size for the in situ generated (one-pot) versus ex situ hydrogen may contribute to the lack of the observed reaction of gas phase hydrogen to ammonia.

The simple dispersion of the nano-iron oxides in the electrolyte, as demonstrated in this study, was not conducive to long-term stability of the cell, as electrostatics tend to coagulate the nanoparticles over time. During the last 2 hours of a 200°C (NaOH-KOH) 6-hour, 2 mA cm−2 run, the ammonia production rate fell to 85% of its average value over the first 4 hours. Better mixing and excess nitrogen and water vapor significantly stabilized the rate. When the water-saturated nitrogen increased from 4 to 111 ml min−1 (retaining all other conditions at the 20 mA applied current), the ammonia production fell only 3% (to 97% of the average rate over the first 6 hours). We are exploring providing a rigid structure to immobilize the dispersed nano-iron oxides in a solid framework.

In this study, we also introduced a solar thermal water self-pressurizing, low electrolysis energy path system. Solar thermal energy is readily absorbed at conversion efficiency in excess of 65% (3335), and here provides an efficient energy source and mechanism to maintain a high reactant pressure. Specifically, the NaOH-KOH electrolyte under N2 gas is heated with varying amounts of water in a confined volume. As expected, heating in a constrained volume evaporates water and yields up to a demonstrated increase in water pressure from 0.03 bar at room temperature to 60 bar at 275°C, and a concurrent decrease in the ammonia electrosynthesis potential in hydroxide electrolytes as ammonia is formed, and as described in the supplementary materials (30). The generated high water pressure is in accord with improved high Q /low ammonia energy synthesis conditions theoretically predicted by Eqs. 7 and 8, as seen by the lower voltage curve of Fig. 1. At 250°C and 25 bar of steam pressure, the observed electrolysis potentials were 0.78V, 1.01V, and 1.31V, respectively, at 0.1, 2, and 25 mA cm−2. The last-named potentials are 0.2 V more favorable than observed at ambient pressure and 200°C. The measured 2 mA cm−2 rate of ammonia synthesis of 2.4 × 10−9 mol cm−2 s−1 at high pressure is similar to that observed at ambient pressure, but the coproduction of H2 is not observed at this lower potential (30).

There is ample room for advances of this pathway. Fe2O3 was utilized as the reactive surface, whereas today’s Haber-Bosch catalysts use Fe2O3 or ruthenium-based catalysts with a wide variety of carefully optimized additives (31, 32, 36), which may also improve this electrochemical process.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S5

Table S1


References and Notes

  1. Materials and methods are detailed in the supplementary materials on Science Online.
  2. Acknowledgments: Full materials and methods and a section on solar thermal–constrained volume pressurization are presented in the supplementary materials. The authors are grateful for partial support of this research by the Office of Naval Research (award N00014-13-1-0791). A related provisional U.S. patent has been filed.)
View Abstract

Navigate This Article