Beyond acceptable limits: intrinsic contamination in commercial ¹⁵N₂ impedes reliable N₂ reduction experiments

Abstract

The validity of N₂ reduction experiments relies on control experiments with isotopically labeled ¹⁵N₂. Here we discovered significant ¹⁵NH₃ contamination (73 ppm) in commercial ¹⁵N₂ and also observed the prescence of oxidized ¹⁵N in amounts comparable to the typical reported ammonia yields.

---

Introduction

The Haber–Bosch (HB) process, with its high carbon intensity (4.21 tCO₂/tNH₃), accounts for about 1.3% of the world's CO₂ emissions through the production of ammonia.^1,2 Over the past few years, various alternative synthesis routes for ammonia have been proposed. Amongst them is the direct nitrogen reduction reaction (NRR) in aqueous conditions (eqn (3)). This approach omits the use of a carbon-based hydrogen source and can operate at pressures and temperatures close to ambient. However, irrefutable experimental evidence for actual (photo)electrochemical ammonia formation at the cathode is typically elusive, this is mainly due to the extremely low reported ammonia production rates (10⁻¹²–10⁻⁹ mol s⁻¹). Moreover, the omnipresence of reactive nitrogen species, which are more easily converted to ammonia than the extremely stable dinitrogen, has led to concerns about false positives in many studies. In response, validation strategies as well as rigorous experimental protocols have been implemented. While most studies stress the use of high-purity reagents, blank tests, controls, etc., the cornerstone of NRR validation lies in the use of isotopically labelled nitrogen gas (¹⁵N₂). Detection of ¹⁵NH₃ post reaction is typically taken as an indicator of NRR occurrence. However, until now, there has been a lack of comprehensive information regarding the purity of commercial ¹⁵N₂ gas, and the impact of activated, ¹⁵N labelled impurities on the measured ammonia yield. Furthermore, there is a notable absence of precise, quantitative comparison between NRR experiments conducted under ¹⁴N₂ and ¹⁵N₂.

False positives and contaminants in the ¹⁴N₂ feed gas

In a typical electrochemical NRR experiment, nitrogen gas is bubbled through the cathode compartments’ aqueous electrolyte solution, where it serves as nitrogen source to generate ammonia. Depending on the pH, either gaseous NH₃ or dissolved NH⁺₄ is formed (eqn (1.1) and (1.2)). The required protons originate from the oxidation of water, conducted in a separate anode compartment (eqn (2.1) and (2.2)). The cathode's production rate and selectivity are typically determined by quantifying the ammonia yield over the course of a few hours.

Cathode:

pH = 0

N₂ (g) + 6e⁻ + 8H⁺ (aq) → 2NH⁺₄ (aq) E° = +0.275 V (1.1)

pH = 14

N₂ (g) + 6e⁻ + 6H₂O (l) → 2NH₃ (g) + 6OH⁻ (aq) E° = −0.772 V (1.2)

  Anode:

pH = 0

2H₂O (l) → O₂ (g) + 4e⁻ + 4H⁺(aq) E° = −1.229 V (2.1)

  pH = 14

4OH⁻ (aq) → O₂ (g) + 4e⁻ + 2H₂O (l) E° = −0.401 V (2.2)

Overall:

2N₂ (g) + 6 H₂O (l) → 4 NH₃ + 3 O₂ E°_cell = −1.17 V (3)

Considering that the reported ammonia production rates are typically small (10⁻¹²–10⁻⁹ mol s⁻¹ cm⁻² of cathode), the challenge lies in verifying that the measured ammonia indeed originates from N₂ reduction and not from more easily reducible, activated nitrogen species such as NO⁻₃, NO⁻₂, NO, NO₂, NH₂OH, etc. These latter species are often found as contaminants in the reaction system, for instance, being inherently present in the electrolytes, catalyst materials and/or the feed gasses.^3,4 While the concentration of these activated nitrogen species in the reaction system can be reduced, they cannot be entirely eliminated. Notably, even under optimal conditions, their amounts match typical ammonia yields.

A key distinction exists between oxidized nitrogen species like NO⁻₃ or NO_x and NH₃/NH⁺₄. While background levels of NH₃ (g) and NH⁺₄ (aq.) in the experimental setup can be factored in via blank tests and careful analysis of the reagents and electrodes, identifying contributions from oxidized nitrogen species poses a more significant challenge. Moreover, the ubiquitous and heavily fluctuating presence of ammonia in the atmosphere and on labware, complicate the accurate quantification of a background level.^5–7

One contamination source of oxidized nitrogen species is the feed gas. Our mass spectrometry results (ESI Fig. 2a†) reveal that a commercial, high-purity (N6) ¹⁴N₂ gas cylinder contains approximately 52 ppb of ¹⁴NH₃, 254 ppb of ¹⁴NO, and 22 ppb of ¹⁴NO₂ (Table 1). The ¹⁴NO_x and ¹⁴NH₃ contamination levels found in a lower purity (N3) ¹⁴N₂ gas was found to be comparable and in the same order of magnitude as the 220 ppb of ¹⁴N₂O reported previously by Choi et al. in N5 purity ¹⁴N₂.⁵ In these feed gasses, the background level of ammonia is approximately 5–6 times lower than the total amount of oxidized nitrogen species. For typical NRR experiments with ¹⁴N₂ feed gas flowing through the cell at 100 ml min⁻¹, the total amount of oxidized nitrogen species introduced from high purity ¹⁴N₂ gas (N6) can be expected in the order of 10⁻¹¹mol cm⁻² s⁻¹ (for a cathode of 1 cm²). As this molar flux of contaminants matches the ammonia production rate of many reported electrocatalysts, it is important to consider that reduction of contaminants in the feed gas alone could potentially account for the entire ammonia output, overshadowing the reduction of N₂.

Table 1 Absolute concentrations of NH₃, NO and NO₂ in commercially obtained high purity (N6) and low purity (N3) ¹⁴N₂ gas as well as ¹⁵N₂ (≥98 at%), determined by SIFT-MS

	Background NH3	Oxidized nitrogen species
	^14NH3 (ppb)	^14NO (ppb)	^14NO2 (ppb)	^14N2O (ppb)
^14N2 (N6 purity)	52	254	22	220^a
^14N2 (N3 purity)	76	220	34	NA

---

	^15NH3 (ppb)	^15NO (ppb)	^15NO2 (ppb)	^15N2O (ppb)
a N2O concentration reported by Choi et al. for commercially obtained ^14N2 (N5) and ^15N2 (N3).
^15N2 (98 at%)	73 293	659	20	230^a

---

The credibility of a given NRR study is primarily based on the ammonia production rate of the cathode. Consequently, thresholds to determine the plausibility of electrochemical NRR have been suggested by Choi et al.⁵ Given typical contamination levels, electrodes exhibiting production rates lower than 10⁻¹⁰ mol cm⁻² s⁻¹ are not considered promising, while promising NRR experiments should present ammonia production rate exceeding 10⁻⁸ mol cm⁻² s⁻¹. Taking this threshold and the above mentioned feed-gas contaminant quantification into account; at modest flow rates, the risk of measuring false positives due to contaminated feed gas is rather limited if high purity (N6) ¹⁴N₂ gas is used. Obviously, this needs to be assessed for each individual case and researchers should always establish background contamination levels prior to performing NRR experiments.

Contamination levels of commercial ¹⁵N₂ gas

While the contamination level in high-purity (N6) ¹⁴N₂ gas is relatively minor compared to the literature-reported ammonia production thresholds, contamination in the form of NH₃/NH⁺₄ and oxidized nitrogen (¹⁴N) species such as NO⁻₃, NO⁻₂, NH₂OH, etc., present in the used electrolytes, chemicals, and labware is significantly higher – orders of magnitude above the threshold for plausible NRR. As an example, Li et al. measured a NO⁻₃ concentration of 11 ppm in an aqueous 0.5 M Li₂SO₄ electrolyte solution.⁷ Distinctly separating the reduction of these contaminants to ammonia from the NRR proves even more challenging than eliminating the contribution of background NH₃/NH⁺₄ in the system. As a consequence, adopting isotopically labelled N₂ gas (¹⁵N₂) and subsequent quantification of the produced ¹⁵NH₃ has been proposed as a valid strategy to confirm ammonia production truly originates from N₂.^5,8 By switching to ¹⁵N₂ feed gas, the interference of previously mentioned contaminants decreases substantially, as the natural abundance of the ¹⁵N isotope is only 0.33–0.42%.⁹

Previous research by Choi and Dabundo pointed out the existence of contaminants such as ¹⁵NO_x in commercial ¹⁵N₂ gas. However, these studies were limited in their capability to provide precise quantifications of ¹⁵N labelled contaminants.¹⁰ Dabundo's approach relied on indirect means of measurement, identifying the presence but not the exact amounts of reactive ¹⁵N species in the gas.¹⁰ In contrast, our study employs mass spectrometry to explicitly quantify both oxidized ¹⁵N species (¹⁵NO and ¹⁵NO₂) and ¹⁵NH₃ in the gas phase. Alarmingly, our findings show that commercial ¹⁵N₂ is considerably contaminated with ¹⁵NH₃ (Table 1, ESI Fig. 3†) (approx. 73 ppm), consistent with dissolved ¹⁵NH⁺₄ concentrations reported by Dabundo (34–1900 ppm). Herein lies a profound risk of measuring false positives: using as-obtained ¹⁵N₂ gas bubbled at a 100 ml min⁻¹ flow rate either during the pre-experimental electrolyte saturation step or consistently throughout the entire experiment introduces sufficient ¹⁵NH₃ to account for an apparent ammonia yield of approximately 1 × 10⁻⁸ mol s⁻¹. This exceeds most reported ammonia production rates by factors of 100–1000. It is important to note that despite the evident risks, many recent NRR studies continue to use unpurified ¹⁵N₂, even though pre-experimental gas purification has been suggested.^5,8,11–13

NRR experiments using ¹⁵N₂ and ¹⁴N₂

To effectively demonstrate the impact of ¹⁵N labelled impurities on the reported ammonia yield, we conducted NRR experiments at a constant potential (−0.91 V vs. RHE), using bismuth, reportedly a highly active element for N₂ reduction, as cathode.¹⁴ We performed two sets of experiments under ¹⁴N₂ and ¹⁵N₂ saturated conditions (Fig. 1). Regularly sampling the catholyte revealed a measurable increase in ammonium content in the ¹⁵N₂ experiment. Specifically, after four hours, the ¹⁵N₂ experiment showed 8.6 μmol of NH⁺₄, compared to only 0.68 μmol NH⁺₄ in the ¹⁴N₂ experiment (Fig. 2). This result clearly demonstrates that when using unpurified ¹⁵N₂ even at low flow rates (20 mL min⁻¹), a considerable amount of ammonia will be detected, primarily originating from the ¹⁵N labelled impurities rather than from the direct reduction of ¹⁵N₂. If the ¹⁵N₂ contamination were overlooked, an apparent ammonia yield of 6 × 10⁻¹⁰ mol s⁻¹ would be calculated for the ¹⁵N₂ experiment. This figure is over ten times higher than the ammonia yield (5 × 10⁻¹¹ mol s⁻¹) of the same experiment under ¹⁴N₂.

Fig. 1 Schematic representation of the setup used for studying the NRR under ¹⁴N₂ or ¹⁵N₂ atmospheres. NRR experiments were conducted in an H-cell configuration. Feed gases (¹⁴N₂ and ¹⁵N₂) were continuously bubbled into the cathode compartment at a predefined flow rate of 20 mL min⁻¹. An acid trap was employed at the exit to completely isolate the cathode compartment from the surrounding atmosphere and to trap any gaseous ammonia.

Fig. 2 Cumulative ammonium production (¹⁵NH⁺₄ or ¹⁴NH⁺₄) in a four hour NRR experiment with a Bismuth cathode.

Mitigation strategies for the ¹⁵N contaminants

Our precise quantification reveals that the level of oxidized ¹⁵N species in this gas is nearly three times higher than in both high (N6) and low purity (N3) ¹⁴N₂ gases, exacerbating the risk of false positives in control experiments. Purifying commercial ¹⁵N₂ gas, which comes in extremely limited volumes of 500–1000 cc at 1 to 5 atm, presents significant technical and economic challenges for NRR research. Traditional methods for purification, such as bulky, home-made scrubbers or acid traps, are neither practical nor effective in removing all ¹⁵N contaminants, and even more advanced solutions like small, commercial gas filters require complex, leak-tight setups for recirculating the limited ¹⁵N₂ supply. Moreover, to ensure data reliability, continuous monitoring of activated ¹⁵N_x species using sensitive techniques like mass spectrometry is necessary, as is the tracking of ¹⁵NH⁺₄ levels in solution over time, rather than relying on a single post-reaction measurement, as is often performed nowadays.

Given these constraints, the risk of false positives in isotopic control experiments for NRR studies is significantly larger when using ¹⁵N₂ gas than ¹⁴N₂, regardless of purity level. This casts serious doubt on the reliability and cost-effectiveness of using ¹⁵N₂ gas as it currently stands and highlights the urgent need for alternative validation methods in NRR research that can reliably account for these challenges.

Conclusion

In recent years, the predominant strategy to exclude false positives in NRR experiments has centered on using isotopically labelled ¹⁵N₂ gas as a feed. While Choi et al. warned researchers about potential interferences from isotopically labelled reactive nitrogen species in these gasses, a comprehensive understanding of the precise contamination levels of these gaseous, ¹⁵N labelled contaminants remained elusive. In this work, we quantify the exact levels of ¹⁵N_x contaminants in commercially obtained ¹⁵N₂, typically used for performing NRR control experiments. Alarmingly, the contamination levels in these gasses make them unsuitable, increasing the chances of obtaining false positives. Thus, given the current gas purity, validation experiments with ¹⁵N₂ do not enhance the credibility of the NRR study when the production rate is low (e.g. <10⁻⁸ mol cm⁻² s⁻¹), neither do they confirm NRR is occurring. Moreover, the ¹⁵N labelling experiments might even mislead researchers into wrongly attributing the measured ¹⁵NH₃ to ¹⁵N₂ reduction at the cathode; as we clearly demonstrated in our comparative NRR experiments. Indeed, the ammonia production rates originating from ¹⁵N labelled impurities in the feed gas is in the order 10⁻⁹–10⁻¹⁰ mol s⁻¹, which is comparable to most reported NRR studies in aqueous conditions. Therefore, the most convincing proof for successful NRR is consistently measuring high ammonia yields (e.g. >10⁻⁸ mol s⁻¹), particularly when using purer ¹⁴N₂, which is more realistic to purify, over the scarce ¹⁵N₂.

Author contributions

M. D. R wrote the paper and assisted in performing the SIFT-MS measurements. M. B. J. R. participated in the writing and editing of the manuscript. L. H., J. A. M. and J. H. participated in the revision of the manuscript. T. L., B. M. N. and M. L. A. T. M. H contributed to scientific discussions and setting up the methodology.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We thank Tessa Vanempten for her assistance with the SIFT-MS measurements and subsequent quantification. M. D. R. acknowledges the Research Foundation Flanders (FWO) for his personal grant (1SA3321N). J. H. and M. B. J. R gratefully acknowledge the financial support from the Research Foundation-Flanders (FWO, Grant No. G0983.19N, G0F2322N, VS06523N), Interne Fondsen KU Leuven through project (C3/20/067 and CELSA/21/016). J. H. acknowledges the Flemish Government Through Long-Term Structural Funding Methusalem (CASAS2, Meth/15/04) and the Moonshot cSBO project P2C (HBC.2019.0108), and the MPI as MPI fellow.

References

1. R. Nayak-Luke, R. Bañares-Alcántara and I. Wilkinson, Ind. Eng. Chem. Res., 2018, 57, 14607–14616.
2. P. Gilbert and P. Thornley, Energy and carbon balance of ammonia production from biomass gasification, ed. A. V. Bridgwater, Proceedings of the Bioten Conference on Biomass, Bioenergy and Biofuels, Host publication, 2010.
3. J. Kibsgaard, J. K. Nørskov and I. Chorkendorff, ACS Energy Lett., 2019, 4, 2986–2988.
4. L. F. Greenlee, J. N. Renner and S. L. Foster, ACS Catal., 2018, 8, 7820–7827.
5. J. Choi, B. H. R. Suryanto, D. Wang, H.-L. Du, R. Y. Hodgetts, Federico M. Ferrero Vallana, D. R. MacFarlane and A. N. Simonov, Nat. Commun., 2020, 11, 5546.
6. B. Hu, M. Hu, L. Seefeldt and T. L. Liu, ACS Energy Lett., 2019, 4, 1053–1054.
7. L. Li, C. Tang, D. Yao, Y. Zheng and S.-Z. Qiao, ACS Energy Lett., 2019, 4, 2111–2116.
8. S. Z. Andersen, V. Čolić, S. Yang, J. A. Schwalbe, A. C. Nielander, J. M. McEnaney, K. Enemark-Rasmussen, J. G. Baker, A. R. Singh, B. A. Rohr, M. J. Statt, S. J. Blair, S. Mezzavilla, J. Kibsgaard, P. C. K. Vesborg, M. Cargnello, S. F. Bent, T. F. Jaramillo, I. E. L. Stephens, J. K. Nørskov and I. Chorkendorff, Nature, 2019, 570, 504–508.
9. J. R. de Laeter, J. K. Böhlke, P. De Bièvre, H. Hidaka, H. S. Peiser, K. J. R. Rosman and P. D. P. Taylor, Pure Appl. Chem., 2009, 81, 1535–1536.
10. R. Dabundo, M. F. Lehmann, L. Treibergs, C. R. Tobias, M. A. Altabet, P. H. Moisander and J. Granger, PLoS One, 2014, 9, e110335.
11. Y. Ashida, Y. Onozuka, K. Arashiba, A. Konomi, H. Tanaka, S. Kuriyama, Y. Yamazaki, K. Yoshizawa and Y. Nishibayashi, Nat. Commun., 2022, 13, 7263.
12. P. L. Arnold, T. Ochiai, F. Y. T. Lam, R. P. Kelly, M. L. Seymour and L. Maron, Nat. Chem., 2020, 12, 654–659.
13. Q. Wang, J. Pan, J. Guo, H. A. Hansen, H. Xie, L. Jiang, L. Hua, H. Li, Y. Guan, P. Wang, W. Gao, L. Liu, H. Cao, Z. Xiong, T. Vegge and P. Chen, Nat. Catal., 2021, 4, 959–967.
14. Y.-C. Hao, Y. Guo, L.-W. Chen, M. Shu, X.-Y. Wang, T.-A. Bu, W.-Y. Gao, N. Zhang, X. Su, X. Feng, J.-W. Zhou, B. Wang, C.-W. Hu, A.-X. Yin, R. Si, Y.-W. Zhang and C.-H. Yan, Nat. Catal., 2019, 2, 448–456.

---

Footnote

† Electronic supplementary information (ESI) available: Materials and methods, technical drawing of the custom setup and obtained SIFT-MS data. See DOI: https://doi.org/10.1039/d3gc04272c

---