Identification, Quantification, and Elimination of NOx and NH3 Impurities for Aqueous and Li-Mediated Nitrogen Reduction Experiments

Ammonia (NH3) ranks among the largest bulk chemical products in the world, with an annual production of 178 million tons and an estimated annual market growth of 3−5% to meet the global demand for fertilizer in the agricultural sector due to an increasing world population. The majority of NH3 is produced by the Haber−Bosch process, wherein elevated temperatures (300−500 °C) and pressures (200−300 bar) are required. In addition, the current process has a major environmental impact (∼1% of the global greenhouse emissions), mostly due to the production of hydrogen by steam-methane reforming. To meet the net-zero emissions goal by 2050, as established in the latest IPCC report, ammonia must be produced via a sustainable pathway. Direct electrocatalytic synthesis of ammonia from dinitrogen and water at mild conditions could potentially offer a carbon-free alternative, resilient to intermittent renewable energy generation. Despite the large research efforts on nitrogen electroreduction in aqueous electrolytes, current NH3 synthesis rates remain extremely low (0.003−14 nmol cm−2 s−1). This is mainly due to the lack of a suitable electrocatalyst and competition with the hydrogen evolution reaction (HER). Besides, the reliable quantification of these low ammonia yields has raised several concerns in the scientific community. The presence of trace amounts of extraneous N species (such as, NH3, NOx, N2O, NOx, and other, more labile forms of N) has led to an increasing number of reported false positives and non-reproducible results. Overall, the electrochemical reduction of nitrogen oxide species into ammonia is more facile than the nitrogen reduction reaction (NRR) on many transition metals. An exception is N2O, which has been proven to only electroreduce into N2 on several transition metals. This implies that N2O is not a concerning impurity source for the NRR. Numerous rigorous experimental protocols have been proposed to perform reliable quantification of NH3 produced by electrochemical N2 reduction. 18,19

A mmonia (NH 3 ) ranks among the largest bulk chemical products in the world, with an annual production of 178 million tons and an estimated annual market growth of 3−5% to meet the global demand for fertilizer in the agricultural sector due to an increasing world population. 1,2 The majority of NH 3 is produced by the Haber−Bosch process, wherein elevated temperatures (300−500°C) and pressures (200−300 bar) are required. 3 In addition, the current process has a major environmental impact (∼1% of the global greenhouse emissions), mostly due to the production of hydrogen by steam-methane reforming. 4 To meet the net-zero emissions goal by 2050, as established in the latest IPCC report, 5 ammonia must be produced via a sustainable pathway. 6 Direct electrocatalytic synthesis of ammonia from dinitrogen and water at mild conditions could potentially offer a carbon-free alternative, resilient to intermittent renewable energy generation. 7 Despite the large research efforts on nitrogen electroreduction in aqueous electrolytes, current NH 3 synthesis rates remain extremely low (0.003−14 nmol cm −2 s −1 ). 8 This is mainly due to the lack of a suitable electrocatalyst and competition with the hydrogen evolution reaction (HER). Besides, the reliable quantification of these low ammonia yields has raised several concerns in the scientific community. The presence of trace amounts of extraneous N species (such as, NH 3 , NO x , N 2 O, NO x − , and other, more labile forms of N) has led to an increasing number of reported false positives and non-reproducible results. 9−13 Overall, the electrochemical reduction of nitrogen oxide species into ammonia is more facile than the nitrogen reduction reaction (NRR) on many transition metals. 14−16 An exception is N 2 O, which has been proven to only electroreduce into N 2 on several transition metals. 15,17 This implies that N 2 O is not a concerning impurity source for the NRR. Numerous rigorous experimental protocols have been proposed to perform reliable quantification of NH 3 produced by electrochemical N 2 reduction. 18,19 Ultimately, purified 15 N 2 -labeled gas is used to reliably confirm the electroreduction of 15 N 2 into the unambiguously traceable 15 NH 3 . 20 However, over recent years, a significant amount of publications, that implemented all recommended control experiments (including 15 N 2 gas), could not be duplicated. 21,22 A common issue is that the efficacy of the implemented purification methods, such as gas purification or N removal from lab materials, is often poorly assessed. Additionally, it remains challenging to identify the main sources of extraneous N and to what extent it contributes to elevated NH 3 background levels.
In this Viewpoint, we present a systematic impurity screening of the most common used lab materials and gases in the aqueous and non-aqueous lithium-mediated NRR field. Not only does this give new insights into the origin of an impurity, but it also highlights the severity of specific sources for an impurity. More importantly, the effectiveness of earlier proposed cleaning strategies for gases, cell components, materials, and lab consumables are re-evaluated and further optimized.
We discover by using sensitive in-line gas detection methods that 14 N 2 and Ar feed gases are free of NH 3 and NO x impurities and do not require excessive N purification. Only 15 N 2 is contaminated and must be purified with a certified or pre-assessed gas filter. Often-used in-house-made scrubbers or liquid traps have a much lower N trapping efficiency and should not be implemented. The accumulation of atmospheric N species on ambient exposed cell components, chemicals, lab consumables, and other labware is inevitable and is most likely the main source of elevated NH 3 background levels. This can be significantly reduced by our recommended pre-treatment procedures. For Li-NRR systems, trace amounts of nitrate might be present in Li-salts and can interfere with the genuine NH 3 quantification, especially at low concentrations. Therefore, we recommend to determine a nitrate background concentration since it cannot be removed from the salt. Ultimately, this work will equip the experimentalist with specific guidelines and tools to perform more reliable NRR measurements.
Impact of Atmospheric NO x and NH 3 Species. One potential source of the extraneous N species can stem from the accumulation of atmospheric NH 3 or NO x on exposed materials. The presence of NH 3 in the atmosphere is primarily caused by emissions from the agricultural sector, where NH 3 volatilization occurs due to intensified herbivore production and field-applied manure. 23 These emissions vary regionally and depend on multiple factors, such as wind direction and speed, humidity, and usage of N fertilizers. The monthly average atmospheric NH 3 concentration in The Netherlands varies between 2 and 44 ppb, 24 which might seem negligible. However, it is expected that long-term atmospheric exposure of chemicals, consumables, and glassware employed in NRR experiments will lead to an unavoidable introduction of contaminants due to the release of adsorbed NH 3 . The majority of atmospheric NO x emissions are derived from industrial and automotive combustion of fossil fuels. 25 Atmospheric NO x concentrations in our laboratory were measured with a chemiluminescent NO x analyzer (details available in the Supporting Information). Our results show that the concentrations fluctuated over the course of five consecutive days, with a maximum atmospheric concentration of 27 ppb ( Figure 1a). However, the uptake rates during 24 h of both atmospheric NO x and NH 3 in water and freshly prepared 0.1 and 1 M KOH solutions were negligible ( Figure  S1). This indicates that short-term atmospheric exposure is not an issue. Long-term accumulation of NO x impurities was monitored for both low-and high-purity grade KOH (85% and 99.99%), and it was found to depend solely on the storage conditions ( Figure S2). KOH bottles stored in a chemical safety cabinet, hence exposed to the laboratory environment for a considerable time period (10 months), contained 4.4 μmol NO 3 − L −1 in a freshly prepared 1 M KOH solution, while NO 2 − concentrations were negligible (<0.2 μmol NO 2 − L −1 ). Remarkably, storing the KOH pellets in a vacuum desiccator for approximately 9 months reduced the NO x impurities to negligible levels (<0.3 μmol NO 3 − L −1 ). Therefore, it is strongly advised to store chemicals in controlled environments such as desiccators or Ar gloveboxes. Impurity Assessment of the Feed Gases. Feed gases are suspected to contain ppm levels of NO x that can be continuously introduced in the electrolyte during reactant gas saturation. We used a commercially available NO x analyzer to assess our high-purity (99.999%) He, 14 N 2 , and Ar gases (see Supporting Information, Figure S3). Additional effort was made to screen the gases for trace levels of NH 3 with our recently developed gas chromatograph (GC). 26 Our analysis reveals that the NH 3 and NO x impurities in all the gases are extremely low. NH 3 concentrations do not exceed the lower detection limit (LOD NHd 3 < 150 ppb) of the GC, and the NO x content falls in the instrument's LOD (1 ppb). High-purity 14 N 2 and Ar gases are manufactured by cryogenic distillation of air. Low concentrations (ppb level) of atmospheric NH 3 and NO x can end up in the process but will be separated because of their significantly higher boiling point. This justifies our observation, while it is in contradiction with earlier claims. If in-line gas detection methods are not available or used, it remains challenging to adequately quantify impurities in the gas stream due to interference from other sources.
Conversely, a 15 N 2 isotopologue is commercially available at a lower purity level than the conventional 14 N 2 ; thus it might contain a higher concentration of contaminants. As such, we measured up to 9.8 ppm of ammonia contained in a 15 N 2 gas bottle (99% purity, Sigma-Aldrich), as reported in Figure S4a. By using isotope-sensitive GC-MS, 27 we found that the totality of the measured ammonia is in the form of 15 NH 3 ( Figure  S4b). The presence of 15 NH 3 presumably derives from traces of unreacted 15 NH 3 used during the catalytic oxidation process for the production of 15 N 2 gas from isotopically enriched 15 NH 3 . 28 Although not measured by us, different 15 NO x species were previously detected in various 15 N 2 gas bottles and can be derivatives from the production process (Table S1). It should be noted that measuring gaseous NH 3 can be subject to underestimation, due to ammonia physisorption. To avoid this, it is recommended to use a direct gas analysis method in combination with inert materials for all the surfaces that are in contact with the gaseous sample. In fact, Figure S4a shows that no ammonia was detected when the same 15 N 2 gas was dosed via a non-passivated mass flow controller. Prolonged 15 N 2 bubbling into the electrolyte is often necessary to reach saturation, which means that the use of cumulative quantification methods requires several hours of reaction time to collect significant amounts of 15 NH 3 . 27 This issue can be partly circumvented by adopting a gas recirculation setup in combination with a suitable gas filter to save costs and minimize accumulation of impurities. 29 From our analysis, it seems that, especially for the execution of 15 N 2 control experiments, the implementation of a gas purifier is strictly necessary. The complete data set with flow rates from 1 to 50 mL min −1 is given in Figure S8.
Feed Gas Purification Methods. Strategies to purify the feed gases are based on catalytic reduction or scrubbing using commercially available certified gas filters (<1 ppb), 21,30 inhouse-made catalytic filters (e.g., based on a Cu-Zn-Al oxide), 31 or scrubbers containing a liquid trap. 32−34 The latter are, to some extent, more economic and are therefore more common. However, it is especially important for uncertified filter systems, such as in-house-developed scrubbers or catalytic filters, to assess their N removal functionalities.
Here, the NO x and NH 3 removal efficiency is examined for a set of commonly used filters by purging them with 50 ppm of NO in He or 13.8 ppm of NH 3 in 14 N 2 for 3 h at experimentally relevant flow rates. We first tested two standard 20 mL scrubbers with a glass frit (Supelco Analytical, 6-4835) connected in series ( Figure S5). The poor solubility of NO in aqueous media results in less than 25% NO removal efficiency when using Milli-Q water ( Figure S6). Alkaline solutions are a common choice because gaseous NO x can be trapped in the form of NO x − . 35,36 Substituting water with 0.1 M KOH already enhances the NO removal efficiency up to 78%.
Previous studies recommended the use of strong oxidizing agents, such as KMnO 4 or NaClO 2 , to convert NO directly into soluble NO 2 − or NO 3 − and improve the overall filter performance. 8 NaClO 2 was mentioned as one of the most effective oxidants and is evaluated in the present work. 37 A solution of 0.1 M NaClO 2 in 0.1 M KOH removed 88% of NO after 3 h purging time (Figure 1b). Additionally, the scrubbing efficiency can be increased by optimizing the gas residence time and the bubble contact area between the gas−liquid interface. As such, inert polytetrafluoroethylene (PTFE) beads were inserted into a 30 cm long, 25 mL in-house-made scrubber (see Figure S7). This results in a further improvement in the removal efficiency, up to 98% over the course of 3 h at 10 mL min −1 (Figure 1b). However, the trapping efficiency drops drastically at higher flow rates (>10 mL min −1 ), as is illustrated in Figure 1c, which limits this purification strategy only to lower flow rates. Remarkably, the commercially certified gas filters (Agilent OT3-4 and Entegris GPUS35FHX) show a consistent unity removal efficiency, within the 1−50 mL min −1 range (Figure 1c and Figure S8). NH 3 was completely eliminated by both commercial filters and our scrubber containing a 0.1 M NaClO 2 and 0.1 M KOH solution ( Figure S9), which was expected due to the high ammonia solubility in water (∼500 g L −1 ). This analysis shows that certified commercial filters are the most efficient and durable solution for feed gas purification. Furthermore, both filters have been extensively used in our laboratories for over 1 year without showing any sign of decay in performance. Moreover, they do not require extensive cleaning and preparation procedures. Lastly, commercial filters are widely accessible and affordable, often with the possibility of being conveniently regenerated via thermal H 2 treatments.
Screening of Lab Consumables. Besides the impurity contributions from atmospheric N species and 15 N 2 gas, there are additional concerns regarding lab consumables because significant NO 3 − concentrations have been observed earlier. 38,39 Therefore, we screened various consumables from our laboratory supply cabinets, including polypropylene 0.1−1 mL pipet tips, 1.5−12 mL sample tubes, and latex and nitrile chemically resistant gloves. For the analysis of the polypropylene consumables, the pipet tips and tubes were submerged and sonicated in 0.1 M KOH for 15 min. This procedure was repeated five times while reusing the same alkaline solution (more details in the Supporting Information). Remarkably, the N content per item is negligible (3−7 nmol), which was unexpected due to continuous ambient exposure. Nevertheless, several 1.5 mL sample tubes that were directly analyzed after arrival were completely free of any N impurities ( Figure S10). This demonstrates that accumulation of adsorbed atmospheric N is inevitable, as was earlier observed for our KOH salts, but is to some extent less severe, and the N species can simply be removed with water.
Patches of latex and nitrile gloves (6 cm × 6 cm) were screened by cutting the patches in little chunks and sonicating them collectively in 0.1 M KOH for 15 min. The latex gloves released reproducible quantities of 5.1 ± 0.7 nmol NH 3 cm −2 and 31.7 ± 2.2 nmol NO 3 − cm −2 , while the nitrile gloves released 3.7 ± 0.5 nmol NH 3 cm −2 and 90.8 ± 1.3 nmol NO 3 cm −2 . These significant NO 3 − concentrations are most likely remaining trace impurities from the calcium nitrate used as coagulant material to harden the gloves during the manufacturing process. Not all manufacturers use calcium nitrate as a coagulant, which can explain the NO x − variations reported in the literature. 19 Regardless, direct contact with electrolyte-exposed surfaces, such as membranes, electrodes, glassware, etc., should be avoided as much as possible. To demonstrate the impact, we performed a qualitative assessment (see the Supporting Information) by rubbing a nitrile glove over the Celgard membrane and observed that reproducible amounts of N species (0.6 ± 0.1 nmol NH 3 cm −2 , 0.6 ± 0.2 nmol NO 2 − cm −2 , 12.2 ± 2.1 nmol NO 3 − cm −2 ) were released (Figure 2a). This shows that especially NO 3 − can be unintentionally introduced during cell assembly.
Encountered Impurities in Commonly Used Cell Materials. Nafion membranes are notorious for their initial NH 4 + uptake and release during NRR experiments. Here, the buildup of atmospheric NH 4 + appears to be the main issue, 40 and it remains difficult to remove because of its ion-selective and porous properties. Impurity effects in other commonly used membranes and electrode materials are, to some extent, unexplored. This motivated us to review other types of membranes, carbon paper (often used as a support), Pt foil, and a Cu electrode prepared by electrodeposition (Cu EL). A pre-defined geometrical area (indicated below) of each particular component was sonicated in 0.1 M KOH for 15 min either as received or after a treatment step for the quantification of trapped N impurities.
Celgard (3401) microporous membranes are considered cleaner alternatives to ion-exchange membranes. 20 From our analysis, we confirm that NH 3 levels for a 2.5 cm × 2.5 cm Celgard membrane are negligible (<1.5 nmol cm −2 ), as shown in Figure 2b. However, we found a relatively high amount of NO x − species of around 7.5 nmol cm −2 . According to the manufacturer, no sources of NO x reactants were used during the production process, hence it is likely that physisorption of atmospheric NO x occurred and accumulated over time. Yet, simply rinsing with water reduces impurity levels to <1 nmol cm −2 . Anion-exchange membranes (AEMs), also commonly used in the NRR field, are mostly used with alkaline electrolytes and have the lowest ammonia crossover rates. AEM ionomers consist of positively charged quaternary ammonium functional groups that give the membrane its anion-selective properties. One could expect that, due to degradation and protonation of the N-functional groups, spontaneous ammonia formation occurs. 9,10,41 However, we did not observe any sign of ammonia leaching from a 2.5 cm × 2.5 cm AEM (Figure 2b), even after 1 h of sonication ( Figure  S11). Additionally, the amount of NO x − species was negligible, which is most likely related to the wetted and sealed storage of the membrane.
Catalyst and electrode materials can also be a potential source of N contaminants. Electrocatalysts prepared by using concentrated ammonia solvents or nitrate compounds should ideally be avoided. If usage is necessary, then additional pretreatment steps and careful examination of the removal effectiveness are advised. Herein, an example is discussed where a 1.13 cm 2 copper electrode (Cu EL) was prepared by electrodeposition using 0.5 M Cu(NO 3 ) 2 on carbon paper. 42 From Figure 2b, it becomes clear that a freshly prepared Cu EL released enormous amounts of NO 3 − (1499 ± 186 nmol cm −2 ). Left-over NO x − can ideally be electroreduced with cyclic voltammetry by scanning the Cu EL between −0.2 and −0.7 V vs RHE in 0.1 M KOH (see the Supporting Information). More than 98% of the initial N-content was removed by this strategy, although the remaining ∼30 nmol is still significant ( Figure S12). Alternatively, metal nitrate hydrates can be thermally decomposed into metal oxides, water, and gaseous NO x . The Cu EL was kept at 200°C overnight because supported Cu(NO 3 ) 2 hydrate decomposition starts at 175°C. 43 The thermal decomposition strategy was able to remove 99.3% of the initial N-content, indicating that it is more efficient than cyclic voltammetry. Moreover, this method was applied earlier to remove NO x − species from commercial metal oxide powders, and similar removal rates were reported. 12 Platinum foil is commonly used as an anode material due to its high stability. After excessively rinsing a 2.5 cm × 2.5 cm Pt foil with H 2 O, approximately 6 nmol cm −2 of NO x − was released. This quantity is comparable with that found with the untreated Celgard membrane, which suggests that atmospheric adsorbed NO x species on the Pt are more stable, forming most likely Pt mononitrosyls. 44 Flame annealing is an often used technique to remove organic impurities and to pre-oxidize the Pt surface. Interestingly, the flame annealing step provokes an increase in the N impurities ( Figure S12). Sonicating the Pt foil in 0.1 M KOH or applying the thermal decomposition method was sufficient to reduce impurities to a bare minimum. NO 3 − Assay of Common Used Lithium Salts in Li-NRR. NRR with electroplated lithium as a N 2 activator (Li-NRR) has recently gained significant scientific interest. There are, however, various concerns about high NO 3 − concentrations in Li-salts, 45 which can easily be converted to NH 3 in these extremely reduced environments. Herein, LiClO 4 , LiBF 4 , LiPF 6 , and lithium bis(trifluoromethanesulfonyl)imide (LiTF-SI, also abbreviated as LiNTf 2 ) are screened with dualwavelength ultraviolet (UV) spectroscopy for NO 3 − quantification. 46 Figure 3 shows that LiClO 4 and LiPF 6 are free of NO 3 − . Clear UV absorbance at 210 nm (associated with NO 3 − ) was measured for LiBF 4 and LiTFSI. Any organic interference at 210 nm was compensated by subtracting 2 times the absorbance at 270 nm (elaborated in the Supporting Information). After this correction, LiTFSI has no noteworthy NO 3 − absorbance, while LiBF 4 in Figure 3f shows a clear upward trend in NO 3 − levels as a function of the salt concentration. It is important to note that NO 3 − quantities can vary with different purities, suppliers, and batches. 45 Therefore, it is recommended to analyze Li-salts with this spectrophotometry method. NO 2 − concentrations in all Li-salts were quantified by ion chromatography (IC) and remained negligible (<1 μmol L −1 ). Ethereal solvents that are stable during Li-NRR, such as tetrahydrofuran, 1,2-dimethoxyethane, and 2-methoxyethyl ether, were screened by IC. Ethanol was also evaluated, since it is often used as a proton source for Li-NRR. None of the organic solvents showed any NO x − -related peak ( Figure S13).
Implications of NO x Impurities for the Li-NRR Experimentalists. Other extraneous N sources from atmospheric exposure are limited in Li-NRR systems because most handling and storage of solvents, salts, and cell materials are conventionally done in a glovebox, with the main motivation to control moisture content. The content of N contaminations in our feed gases and lab consumables is negligible (except 15 N 2 ), thus only NO 3 − impurities in the Lisalt seem to be relevant for Li-NRR. It is important to note that NO 3 − (most likely present as LiNO 3 ) cannot simply be removed by a heat treatment, 45 since the decomposition temperature of LiNO 3 (≥500°C) is much higher than those of LiBF 4 , LiPF 6 , and LiTFSI. 47 With the hypothetical experimental conditions stated in Figure 4, roughly 107 nmol of NO 3 − can potentially be reduced into NH 3 during cell operation, leading to a yield of 0.12 nmol s −1 cm −2 . Our estimated NO 3 − content can differ significantly if higher salt concentrations are used or with different Li-salt batches that contain more NO 3 − . Nevertheless, it is not realistic to expect that NH 3 yields obtained by the electroreduction of NO 3 − will approach the recently obtained 2500 nmol s −1 cm −2 at 1 A cm −2 , 48 and 150 nmol s −1 cm −2 at a current efficiency near unity (at 15−20 bar). 49 This, however, might not be true when the Li-NRR reports lower NH 3 yield (e.g., when operating at ∼1 bar). Overall, we find that N impurities are less relevant for the Li-NRR field, although it remains good practice to assess the NO 3 − content in the Li-salts to be certain of the origin of NH 3 .
Estimation of a Minimum Background Level for Aqueous NRR Measurements. In the NRR, the atmospheric N contributions are more severe, as experiments are generally not performed in a controlled environment, including storage of chemicals and cell materials in ambient air. By combining the most important findings from this study, as illustrated in Figure 4, a background level of ∼140 nmol was estimated. By assuming that most NO x − species electroreduce into NH 3 , an obtained yield of 0.16 nmol s −1 cm −2 is already enough for a NRR catalyst to be labeled as plausible. 8 Approximately 84% of these impurities can be avoided by applying the most effective cleaning procedures. These are material dependent and include alkaline washing for membranes and electrodes, heat treatment for the Pt foil, desiccator storage for salts, and rinsing lab consumables with ultrapure water. Important factors such as catalyst impurities and the influence of gloves are excluded from this analysis because they may vary between studies. Extra care must be taken when validating electrocatalytic NRR activity with 15 N 2 gas, since ppm levels of 15 NH 3 were detected by our GC-MS and 15 NO x by others. Cleaning the feed gases is not straightforward, since our analysis shows that commonly adopted liquid scrubbers do not properly eliminate the NO x contaminations, due to limited mass transport and reactivity. More importantly, the trapping efficiency should be evaluated at conditions close to experimental conditions, as we show that factors such as flow rate and duration of the experiment highly affect the removal efficiency. For these reasons we strongly recommend the application of commercial gas purifiers that exhibit the best performance at all relevant conditions. An absolute minimum background level is rather difficult to assess because of the large variety of experimental approaches within the research community. Nevertheless, we provide experimentalists with recommendations and various cleaning procedures in order to reduce the effect of impurities to an acceptable minimum.    Figure S12 and Tables S1− S4, assuming the N 2 flow (20 mL min −1 , 99.999%), membrane area (Celgard, 10 cm 2 ), working electrode (carbon paper, 1 cm 2 ), counter electrode (Pt foil, 4 cm 2 ), electrolyte (1 M KOH, 10 mL), 1 pipet tip, and 1 tube with a total experiment time of 15 min. For Li-NRR, only 14 N 2 and electrolyte impurities were considered. The applied cleaning procedures for NRR were as follows: alkaline wash for Celgard 3401 membrane and carbon paper, heat treatment for Pt foil, KOH desiccator storage, and rinsing lab consumables with water.
Materials and methods, experimental details, quantification methods and calibration curves, schematic of the setups, electrolyte storage data, 15