Ambient Ammonia Electrosynthesis from Nitrogen and Water by Incorporating Palladium in Bimetallic Gold–Silver Nanocages

Bimetallic porous Au-Ag nanocages with an LSPR peak position at 650 nm are synthesized by the galvanic replacement method by adding HAuCl₄ (aq.) solution to a solution of solid silver nanocubes (AgNCs) (Figs. 1a and 1b).^21–24 Trimetallic Au-Ag-Pd nanostructures are synthesized by adding K₂PdCl₄ (aq.) solution to the porous bimetallic Au-Ag nanocages dispersed in DI water with the LSPR peak at 650 nm. The volume of the Pd²⁺ precursor added to the Au-Ag template, and subsequent reduction of Pd²⁺ to Pd⁰ (two Ag atoms are replaced with one palladium atom), is controlled by measuring the redshift in the LSPR peak positions of nanoparticles. As the amount of Pd salt solution increases, the amount of redshift and the LSPR bandwidth increases (Fig. 1a). Excessive addition of Pd salt results in aggregation and collapsing of the nanocages. The LSPR peak of the resulting trimetallic Au-Ag-Pd redshifts from 650 nm to 850 nm; indicating the replacement of the remaining Ag atoms in the Au-Ag nanocages with Pd atoms (galvanic replacement) or the growth of Pd on Au (Fig. 1). Due to the lattice mismatch between Pd (3.890 $\mathring{\rm{A}}$) and Au (4.079 $\mathring{\rm{A}}$), when Pd²⁺ precursor is added to the solution of Au-Ag nanocages, islands of palladium are formed on the Au surface. By further addition of the Pd salt solution, the islands are grown more, building a highly rough and porous surface (Figs. 1c and 1d).

Zoom In Zoom Out Reset image size

Scanning transmission electron microscopy (STEM) and energy dispersive X-ray (EDX) spectroscopy are performed on a single trimetallic nanostructure to determine the structure, the elemental composition, and distribution of Au-Ag-Pd-750 and Au-Ag-Pd-850 nanoparticles (Fig. 2). For Au-Ag-Pd-750, Pd islands are distributed in the Au-Ag nanocages structure, indicating the successful incorporation of Pd into bimetallic Au-Ag nanocages, and the synthesis of trimetallic Au-Ag-Pd nanoparticles (Figs. 2a–2f). By increasing the amount of Pd salt solution and red shifting the LSPR peak from 750 nm to 850 nm, a continuous porous layer of Pd is formed at the exterior surface of the nanoparticles (Figs. 2g–2l). In addition, inductively coupled plasma emission spectroscopy (ICPES) measurements reveal that by adding the Pd salt solution into the Au-Ag nanocages, the Au content remains relatively unchanged while the Ag content decreases from 2.68 mg l⁻¹ to 1.35 mg l⁻¹ when the LSPR peak redshifts from 650 nm to 850 nm (Table I). This indicates that the galvanic replacement of Ag atoms with Pd ions is one of the mechanisms for the reduction of Pd²⁺ to Pd⁰ in which the concentration of Pd increases to 1.23 mg l⁻¹ and 2.77 mg l⁻¹ as the LSPR peak redshifts to 750 nm and 850 nm. The stoichiometric balance between Ag and Pd (Pd²⁺_(aq.) + 2Ag⁰_(s)$\to$ Pd⁰_(s) + 2Ag⁺_(aq.)) and the Pd and Ag contents in Table I reveal that the reduction of Pd²⁺ to Pd⁰ is also accomplished through the island-growth mode on Au atoms (Fig. 2i).^30,31

Zoom In Zoom Out Reset image size

X-ray photoelectron spectroscopy (XPS) results of Au-Ag-Pd-850 nanoparticles reveal a spin–orbit doublet for Au 4 f at 84.0 eV and 87.7 eV which is consistent with our prior study, indicating that the zero-valence state of Au (Au⁰) is preserved after the addition and reduction of the Pd²⁺ precursor in bimetallic Au-Ag nanocages (Fig. 3a).²⁴ The Ag 3d doublet energy peaks at 367.9 eV and 373.9 eV slightly shifts to lower binding energy (0.2 eV) compared to the Ag 3d doublet in bimetallic Au-Ag nanocages; this is attributed to the interaction and charge distribution of metallic Ag⁰ and Pd precursors that occur after galvanic replacement (Fig. 3b).³² The Pd 3d spectrum is deconvoluted into two pairs of doublets. The doublet peaks at 335.8 eV, and 340.8 eV correspond to Pd at zero-valence state, suggesting the successful incorporation of Pd⁰ in the bimetallic Au-Ag nanocages. The doublet peaks at higher binding energies and with lower intensities than those of zero-valence Pd metal (337.1 eV and 342.5 eV) are attributed to the oxidized Pd states (Pdⁿ⁺) (Fig. 3c).³³ The X-ray diffraction (XRD) patterns of Au-Ag-Pd and Au-Ag nanocages indicate the presence of metallic Pd which is evident from the three distinct peaks at higher 2 $\theta$ angles (40.4^o, 46.9^o, 68.4^o) in Au-Ag-Pd compared to those of the Au-Ag nanocages (38.3^o, 44.6^o, 64.7^o) (Fig. 3d).

Zoom In Zoom Out Reset image size

The electrochemical surface area (ECSA) of the Au-Ag-Pd nanoparticles is determined using a three-electrode cell with a rotating disk electrode (RDE) and performing cyclic voltammetry (CV) in an Ar-saturated 0.1 M LiOH (aq.) at zero rotation rate at a scan rate of 50 mV s⁻¹. The ECSA_Pd is calculated from the reduction peak of Pd oxide after double-layer correction and a charge density of 424 μC cm⁻²_Pd (Fig. 4a and Table I). The ECSAs of the nanoparticles are normalized to the Pd and Au loading on the working electrode, obtained from ICPES measurements (Table I). Au-Ag-Pd-850 has a higher surface area (345.4 m² g⁻¹) compared to the Au-Ag-Pd-750 (291.4 m² g⁻¹). This can be attributed to the existence of a thick porous layer of Pd at the exterior surface of the Au-Ag-Pd-850 nanoparticles, which increases the ECSA for catalytic reactions. It is well known that rough surfaces are catalytically more active than smooth surfaces, as atoms present on rough surfaces are more thermodynamically active.^30,34 The ECSA_Pd obtained using our trimetallic Au-Ag-Pd-850 nanoparticles is approximately five times higher than the ECSA_Pd of the commercial Pd/C catalysts.^17,35 Although the reduction peak area of Pd oxide is remarkably higher than that of Ag oxide, due to the comparable reduction potentials of these two metal oxides, the reduction peak area of Ag oxide is subtracted from the actual reduction peak area centered at around 0.5 V vs RHE to determine the ECSA_Pd (Fig. 4a). The observed negative shift in the electrochemical reduction of Pd oxide (∼0.1 V) in our trimetallic nanostructures, compared to the pure Pd, is attributed to the slight alloying of Pd and Ag at their interface during the galvanic replacement reaction.³⁶ The ECSAs_Au of trimetallic Au-Ag-Pd nanoparticles is comparable (∼0.93) to the ECSAs_Au of bimetallic Au-Ag nanocages (Table I).²² This indicates that the incorporation of Pd in bimetallic Au-Ag nanocages does not result in blocking the Au active sites for electrochemical NRR and confirms the preservation of the hollow structure in trimetallic nanoparticles where Au, Ag, and Pd can act as catalytically active centers for the reaction.

Zoom In Zoom Out Reset image size

Linear sweep voltammetry (LSV) tests are performed in N₂-saturated 0.5 M LiClO₄ (aq.) electrolyte in an H-type cell, where a proton exchange membrane separates anodic and cathodic compartments. A neutral pH electrolyte was selected in this study based on our optimization of electrolytes' pH to obtain the highest electrocatalytic NRR activity.²² A remarkably higher current density is achieved using Au-Ag-Pd-750 and Au-Ag-Pd-850 nanoparticles compared to the bimetallic Au-Ag-715 nanocages (the best bimetallic Au-Ag nanocatalyst obtained in our previous study²²) (Fig. 4b). For instance, at an applied potential of −0.3 V vs RHE, the current density increases ∼4.8 and ∼3.7 times for Au-Ag-Pd-850 and Au-Ag-Pd-750 compared to the Au-Ag-715. This observation is consistent with significantly higher ECSA_Pd compared to the ECSA_Au (Table I) in trimetallic Au-Ag-Pd nanoparticles, which provides more active sites for nitrogen adsorption and reduction. Although achieving high current density at low overpotentials does not necessarily indicate the high selectivity and activity of an electrocatalyst toward NRR, it is the primary step toward the high production rate that is mandatory for reaching the overarching goal of commercializing electrochemical NRR for sustainable ammonia production.

The selectivity performance ($\tfrac{{I}_{{N}_{2}}-{I}_{Ar}}{{I}_{{N}_{2}}}\times 100$) of trimetallic Au-Ag-Pd nanoparticles is evaluated toward NRR. For both electrocatalysts (i.e., Au-Ag-Pd-750 and Au-Ag-Pd-850), the higher current density was achieved in N₂-saturated electrolyte compared, to the Ar-saturated electrolyte, within the wide potential window (Fig. 4c). Unlike the bimetallic Au-Ag nanocages demonstrated in our previous study,²² the selectivity performance of electrocatalysts improves at low overpotentials (0 V to −0.3 V vs RHE) toward NRR after the incorporation of Pd into the bimetallic Au-Ag nanocages (Fig. 4c). In addition, both electrocatalysts are selective toward NRR until the high negative potential of ∼−0.75 V vs RHE, even with an enhanced current density that is achieved using trimetallic nanostructures. This expands the selectivity of an electrocatalyst toward NRR in negative potentials by 0.15 V over that of Au-Ag nanocages. Moving toward more negative potentials than −0.75 V, a hydrogen evolution reaction (HER) becomes dominant. Here, we focus on the potential window of −0.3 to −0.5 V vs RHE, which is a compromise of obtaining high current density and selectivity. It is important to note that achieving a high current density for NRR is crucial to decrease the capital cost, leading to the commercialization of electrochemical NRR. At all three applied potentials, the selectivity is higher for Au-Ag-Pd-850 than for the Au-Ag-Pd-750; this is attributed to the higher Pd content and lower Ag content of Au-Ag-Pd-850 nanoparticles (Table II). Even though the selectivity is not substantially higher in trimetallic Au-Ag-Pd nanostructures than in the bimetallic Au-Ag nanocages, due to the enhanced current density obtained (${I}_{{N}_{2}}$) using Au-Ag-Pd nanoparticles, the corresponding current density toward NRR (${I}_{{N}_{2}}-{I}_{Ar}$) is much more pronounced after incorporation of Pd into Au-Ag nanocages.

Chronoamperometry (CA) tests are performed at a series of applied potentials using Au-Ag-Pd-750 and Au-Ag-Pd-850 nanoparticles to determine the ammonia yield rate and FE (Figs. 5a and 5b). The highest ammonia yield rate (13.74 μg cm⁻² h⁻¹) is obtained at −0.4 V corresponding to an FE of 44.11% while the highest FE (48.94%) is achieved at −0.3 V (ammonia yield rate = 5.80 μg cm⁻² h⁻¹) using Au-Ag-Pd-850. The higher ammonia yield rate but lower FE at −0.4 V, compared with −0.3 V, is attributed to the compromise between increasing current density and competitive selectivity toward HER rather than NRR. For all potentials tested, both the NH₃ yield rate and the FE are higher when Au-Ag-Pd-850 is used than when Au-Ag-Pd-750 is used. This result suggests that Pd content and ECSA_Pd of nanoparticles have a role in promoting the electrocatalytic NRR activity (Fig. 5b). In addition, greater Pd content in Au-Ag-Pd-850 compared to that of Au-Ag-Pd-750 results in an upshift of the d-band center (E − E_f) from −4.45 eV to −3.73 eV, which was determined via ultraviolet photoelectron spectroscopy (UPS) measurements (Fig. S4, available online at stacks.iop.org/JES/167/054511/mmedia). This leads to improved binding strength of N-containing adsorbates with the catalyst surface, which is a crucial step for engineering selective and active NRR catalysts in an aqueous solution where selectivity is a major challenge.

Zoom In Zoom Out Reset image size

Here, we also report the two critical parameters (i.e., production energy efficiency (%) and energy input (MWh _Elec./ton _NH3)) in evaluating the performance of the N₂ electrolysis system, which is neglected in most electrochemical NRR studies in the literature. The highest production energy efficiency (28.9%) is achieved using Au-Ag-Pd-850 at −0.3 V, which corresponds to the electrical energy input of 19.1 MWh/ton _NH3. Moving toward more negative potentials, the production energy efficiency decreases to 24.8% at −0.4 V and 8.3% at −0.5 V, mainly due to the significant increase in current density at more negative potentials, which increases the electrical energy input (Fig. 5c). It is noted that the increase in the production energy efficiency is in line with the decrease in the electrical energy input for various applied bias. Similar to the electrocatalytic NRR activity results (Fig. 5b), both production energy efficiency and energy input deteriorate for Au-Ag-Pd-750 compared to Au-Ag-Pd-850 in various applied potentials (Fig. 5c). Currently, a significant portion of the input electrical energy ($\gt$75%) for all conditions is consumed at the anode side, where oxygen evolution reaction (OER) takes place. Alternative use of organic-based electrolytes (e.g., glycerol) in the anodic half-reaction could remarkably lower the electricity consumption.³⁷ It is worth mentioning that the-state-of-the-art thermochemical process (Haber-Bosch) for ammonia synthesis consumes the energy of 7.8 MWh/ton _NH3 (based on natural gas), and the target production energy efficiency of ARPA-E program (REFUEL) for electrochemical fuel production is greater than 60%.⁵

A CA test is performed at −0.4 V for 72 h to evaluate the stability of Au-Ag-Pd-850 (Fig. 5d). The electrocatalyst could maintain the NH₃ yield rate (13.62 μg cm⁻² h⁻¹) and FE (43.5%) over 72 h, which is very close to the NRR activity of the catalyst in the initial 12 h test (Fig. S3a). This performance corresponds to the turnover frequency (TOF) of 59 h⁻¹ as per active Pd, Au, and Ag sites (see supporting information for the detailed calculation). The SEM images before and after the stability test, reveal that the nanoparticles are firmly attached to the substrate, and they are not washed away in long-term testing under stirring and applied bias (Fig. S1, Table SI). In addition, the TEM images after the stability test show the minor morphology change of nanoparticles after 72 h of the CA test (Fig. S2). Extensive control experiments are carried out following the instructions in the literature and our prior studies.^{21,24,38–40} Control experiments (i.e., Ar, N₂ at OCV, N₂ with no catalyst) with conditions similar to those in Fig. 5b, yield remarkably smaller amounts of NH₃ (Ar: ∼7.3% of N₂, N₂ at OCV: ∼6.1% of N₂, N₂ with no catalyst: ∼2.5% of N₂) (Fig. S3b). The amount of ammonia produced in an isotopic labeling experiment using ¹⁵N₂ gas after a 4 h electrolysis test is close to that of ¹⁴N₂ (73.6 $\mu$M, ∼91% of ¹⁴N₂) (Fig. S3c). Furthermore, the doublet and triplet couplings of ¹⁵N₂ and ¹⁴N₂ obtained from ¹H NMR measurement confirms that the supplied N₂ is the major source of ammonia formation in the system (Fig. 5e). The amounts of ammonia measured using ¹H NMR are similar to the indophenol method (71.9 $\mu$M for ¹⁵N₂ and 78.4 $\mu$M for ¹⁴N₂). Calibration curves of the ¹H NMR signal and indophenol blue method for standard solutions of ¹⁵NH₄⁺ and ¹⁴NH₄⁺ were provided in our prior studies.^21,24

To probe the possible reaction mechanisms and track the intermediate species in electrocatalytic NRR, operando SERS spectra are collected during the CV tests on the SERS active substrate, which is comprised of Au-Ag-Pd nanoparticles deposited on the Au thin film working electrode. Detailed information regarding the spectro-electrochemical setup and CV curves is provided in the supporting information (Figs. S5 and S6). The SERS spectra in N₂− and Ar-saturated LiClO₄ (aq.) solution contain the vibrational band located at 932 cm⁻¹, which is attributed to the stretching mode of the ClO₄⁻ anion (Figs. 6a and 6b). In addition, the wide vibrational band centered at 3415 cm⁻¹ corresponds to O–H stretching. As the potential is swept to negative values in the reductive pathway (0.97 V to −0.23 V vs RHE), the faint evolution of three vibrational modes at 1094, 1393, 1601 cm⁻¹ is observed, reaching their highest intensity at −0.23 V; these can be classified as N–N stretching, H–N–H bending, and N–H wagging, and these suggest the formation of N₂H₄ as an intermediate species (Figs. 6a, 6c, 6d).²⁹ Two more peaks centered at 2937 cm⁻¹ and 3605 cm⁻¹ are evolved at −0.23 V, corresponding to the N–H asymmetric and symmetric stretching modes, respectively. The evolution of these two peaks strongly supports the formation of NH₄⁺ during the reductive potential sweep.^29,41,42 By moving toward more negative potentials during the reductive CV scan (−0.23 v to −0.63 V), the intensity of peaks at 1094, 1393, 1601 cm⁻¹ (intermediates) decreases while the peak at 2937 cm⁻¹ and the shoulder at 3605 cm⁻¹ (NH₄⁺) reach their highest intensities. This strongly suggests the formation of NH₃ from the intermediate species (N₂ $\to$ N₂H₄ $\to$ NH₃) (Figs. 6a, 6c, 6d). Since the literature values used to assign these peaks were obtained using infrared (IR) spectroscopy, the wavenumbers are not expected to align precisely, as the excitation mechanisms are different in each type of spectroscopy. The intensity of the peaks ascribed to intermediates and ammonium decreases during the oxidative pathway due to the oxidation of the N-containing species (Figs. 6a, 6c, 6d). The SERS spectra of the standard ammonia and hydrazine solutions in 0.5 M LiClO₄ (aq.) confirm the band assignment of various vibrational modes in NRR operando SERS measurements (Figs. S7 and S8). Furthermore, operando SERS measurements in Ar-saturated electrolyte with the electrocatalyst and N₂-saturated electrolyte without the electrocatalyst reveal no pronounced peaks, implying the peaks observed in N₂-saturated electrolyte with the electrocatalyst are solely related to the formation of N-containing species at the presence of the electrocatalyst (Fig. 6b, Fig. S9).

Zoom In Zoom Out Reset image size

The role of incorporating Pd in bimetallic Au-Ag nanocages on the electrocatalytic NRR selectivity and activity was investigated. TEM, EDX, ICPES measurements, and stoichiometric balance revealed that the reduction of Pd²⁺ to Pd⁰ is accomplished through both an island-growth mode on Au as well as the galvanic replacement of Ag. The increase in the Pd content is controlled by monitoring the localized surface plasmon resonance (LSPR) peak position of the resulting trimetallic nanoparticles. The ECSA_Pd achieved with our synthesized Au-Ag-Pd-850 nanoparticles is approximately 5 times higher than that of commercial Pd/C catalyst, which is mainly attributed to the formation of a rough and porous layer of Pd in trimetallic nanostructures. The correlation between the position of the d-band center of trimetallic nanoparticles and the Pd content was explored. Furthermore, the effect of Pd content on NRR current density, selectivity, and activity was discussed. The highest electrocatalytic NRR activity (FE = 48.94% and NH₃ yield rate= 5.80 μg cm⁻² h⁻¹) was achieved at −0.3 V vs RHE. This corresponds to the production energy efficiency of 28.9% and the electrical energy input of 19.1 MWh/ton _NH3. Operando SERS revealed that electrochemical NRR takes place on the catalyst surface through an associative mechanism with N₂H₄ as an intermediate species.

This work highlights the importance of engineering the morphology and composition of nanocatalysts to improve the electrocatalytic NRR activity. In particular, it informs the crucial role of Pd with enhanced ECSA for increasing the rate of ammonia electrosynthesis. It also demonstrates the use of operando SERS as a powerful technique for unraveling reaction mechanisms for the electrocatalytic phenomenon.

Nanoparticle synthesis

Production energy efficiency calculation

The production energy efficiency (${\eta }_{PEE}$)of the electrolysis system is calculated according to the following equation:

$$\begin{eqnarray}&&{\eta }_{PEE}\left( \% \right)=\displaystyle \frac{{\rm{\Delta }}G\,for\,ammonia\,generation\,\left(J\,mo{l}^{-1}\right)\times ammonia\,generated\,\left(mol\right)}{\displaystyle \int IVdt\left(J\right)}\times 100\end{eqnarray} \tag{ 1 }$$

The free energy for NH₃ generation (${\rm{\Delta }}G$) is 339 kJ mol^–1 and $\displaystyle \int IVdt\left(J\right)$ is the electricity consumed in the process (energy input). Here, $V$ (V) is the full cell potential and is calculated according to the following equation:

$$\begin{eqnarray}&&V=\left({{\rm{E}}}_{{\rm{anode}}}-{{\rm{E}}}_{{\rm{cathode}}}\right)+IR\end{eqnarray} \tag{ 2 }$$

The cathode potentials (NRR) were measured in the three-electrode set up for each condition (Fig. 5c). The anode potential is estimated as the sum of the thermodynamic potential for water oxidation (1.23 V vs RHE) and OER overpotential (∼0.45 V on the Pt mesh).⁴³

The resistance (R) between anode and cathode in our setup is measured to be 75 Ω. This indicates that even for the case with the highest average current density (∼0.4 mA cm⁻² at −0.5 V vs RHE using Au-Ag-Pd-850), the resistive losses (IR) will be ∼30 mV.

Determination of the electrochemical surface area (ECSA) of the nanoparticle

The ECSA of the catalyst is determined from the charge associated with the reduction peak of Pd oxide (PdO) after double-layer correction and is normalized to the Pd loading on the working electrode and the charge density of 424 μC cm⁻² according to the following equation³⁵:

$$\begin{eqnarray}&&ECSA\left(\displaystyle \frac{c{m}_{Pd}^{2}}{{g}_{Pd}}\right)=\displaystyle \frac{Q\left(\mu C\,c{m}^{-2}\right)}{424\,\mu C\,c{m}_{Pd}^{-2}\times electrode\,loading\left({g}_{Pd}\,c{m}^{-2}\right)}\end{eqnarray} \tag{ 3 }$$

where $Q$ (μC cm⁻²) is the charge associated with the reduction peak of Pd oxide after double-layer correction and is calculated according to the following equation:

$$\begin{eqnarray}&&Q=\displaystyle \frac{\displaystyle \int iV}{v}\end{eqnarray} \tag{ 4 }$$

where $i$ is the current density (μA cm⁻²), $V$ is the potential (V), and $v$ (V s⁻¹) is the scan rate.

The authors gratefully acknowledge Amazon Catalyst at ECS grant for supporting this work. This work was performed in part at the Georgia Tech Institute for Electronics and Nanotechnology, a member of the National Nanotechnology Coordinated Infrastructure (NNCI), which is supported by the National Science Foundation (Grant ECCS-1542174). Additional financial support is acknowledged from the National Science Foundation under Grant No. 1904351.