Towards green magnesium preparation using a recyclable argon plasma anode for continuous electrolysis in molten chlorides

Abstract

During the magnesium electrolysis process, graphite anodes are prone to consumption and fragmentation, disrupting normal operations and releasing greenhouse gases. In this study, we introduce a continuous, carbon-free approach to magnesium electrolysis using a stable argon plasma as the anode. The argon plasma anode operates via two distinct electrochemical stages: argon ionization at an average potential of 347.3 ± 95.3 V and chlorine evolution at 124.5 ± 9.0 V. A protective boron nitride (BN) coating on the tungsten collector filament significantly reduces corrosion. Atomic emission spectroscopy confirms the presence of Ar⁺ ions, whose concentration increases with current. Thermodynamic calculations and reaction analysis demonstrate that Ar⁺ facilitates chloride oxidation via the reaction: 2Ar⁺ + 2Cl^- → 2Ar + Cl₂. This research demonstrates the feasibility of argon plasma as an inert anode for chloride-based molten salt electrolysis, offering a sustainable alternative for magnesium production and potential applications in other high-temperature corrosive environments.

Introduction

Molten salt electrolysis has long served as a foundational technology for the extraction of active metals^1,2,3, such as aluminum, magnesium, and rare earth elements. In recent years, its applications have expanded into emerging fields, including the recovery of inert metals via molten chloride electrolysis^4,5,6, iron production through molten oxide electrolysis^7,8,9, and the synthesis of carbon materials using molten carbonate electrolysis^10,11,12. Conventional processes for aluminum and rare earth metal electrolysis typically employ oxide-containing solutes and carbon-based anodes. However, these systems have significant environmental impacts: the anodic reactions release oxygen and fluoride compounds, thereby emitting potent greenhouse gases. including CO₂, CF₄, and C₂F₆¹³. Not only do these processes consume large quantities of high-purity carbon materials, but they also significantly increase the carbon footprint of metal production, thereby exacerbating global warming and conflicting with goals for carbon peaking and carbon neutrality.

To address these environmental challenges, adopting non-consumable anodes has become a crucial strategy in research and industry for maintaining continuous production and lowering greenhouse gas emissions. Consequently, considerable efforts have focused on developing inert anodes for molten salt electrolysis^2,8,10. These initiatives have motivated the exploration of various solid anode materials, including metallic^8,10,14, ceramic¹⁵, and cermet^16,17, as well as gas-phase anodes^18,19,20.

In magnesium electrolysis with chloride-based molten salts, graphite anodes—though thermodynamically stable—are susceptible to mechanical degradation during operating conditions. Frequent failures often require anode replacement within ten months or less²¹, leading to excessive graphite consumption, operational disruptions, and higher costs. The degradation mechanisms are multifaceted: firstly, hydrolysis of MgCl₂ produces MgOHCl or MgO, which engage in chlorination reactions that corrode the anode and release CO₂ (2MgOHCl + 2Cl₂ + C = 2MgCl₂ + 2HCl + CO₂). Moreover, oxidation of graphite caused by air infiltration at operating temperatures (650–700 °C), coupled with mechanical erosion by the electrolyte, further accelerates anode wear.

Therefore, developing durable inert anodes for chloride-based magnesium electrolysis is a key focus for research and industry. For instance, Mousa et al.²² developed a ceramic anode via solid-phase synthesis of Nb₂O₅ and TiO₂, yet observed increased porosity and performance decay after 16 h of electrolysis at 0.75–1.25 A·cm⁻², highlighting the challenges of achieving chlorine-resistant materials. Indeed, resistance to chlorination-induced corrosion presents a much greater challenge than oxidation resistance^10,23,24.

The need to immerse solid inert anodes in molten salt electrolytes makes complete corrosion prevention very difficult. An appealing alternative is to physically isolate the anode from direct contact with the corrosive melt and chlorine, thereby avoiding chlorination attacks. In this context, plasma-based non-contact anodes are a highly promising solution. For example, Feng et al. demonstrated the application of an argon plasma anode in both potassium cryolite²⁵ and a molten oxide electrolysis²⁶ systems, where oxygen evolution occurred at the plasma anode. At the same time, aluminum or iron was produced at the cathode.

A significant obstacle in chloride molten salt electrolysis is the poor corrosion resistance of traditional solid inert anode materials, and developing such materials has been so difficult that research progress has largely stalled. In this work, we present the use of a non-contact argon plasma anode for chloride-based molten salt electrolysis to produce magnesium. We systematically investigate the electrochemical behavior of argon plasma under varying current conditions and elucidate the anodic reaction mechanisms in a NaCl-KCl-MgCl₂ electrolyte at 953 K. The catalytic function of argon plasma in promoting the anode reaction is identified and thoroughly discussed. The comprehensive experimental details are provided in Fig. 1, and the detailed configuration of the DC discharge reactor is illustrated in Supplementary Fig. 1.

Fig. 1

Schematic diagram of the experimental setup.

Results and discussion

Electrochemical characteristics of argon plasma anodes

The working electrode potential indicates the potential at which electrochemical reactions occur during electrolysis. Based on the temporal potential variations and corresponding gas evolution profiles at different currents (Fig. 2), the discharge process can be divided into two distinct phases: (I) the argon ionization phase, occurring between 400 and 800 V, and (II) the chlorine evolution phase, observed between 90 and 200 V. Argon ionization exhibited considerable potential fluctuations within the 400–800 V range, whereas chlorine evolution occurred more stably, typically between 100 and 150 V (Fig. 2J).

Fig. 2: Plasma anode potential and anode products for different current values.

Relationship between the potential of the plasma anode and time at A 0.1 A, B 0.2 A, C 0.3 A, D 0.4 A, E 0.5 A, F 0.6 A, G 0.7 A, H 0.8 A, I 0.9 A. J The average potential of each reaction of Ar and Cl for the function of different current values. The error bars represent the standard deviations of the potentials. K Average potential for different reactions. The error bars denote the standard deviations of the potentials. L Experimental, theoretical yield, and current efficiency of chlorine as a function of current.

The reaction potentials were evaluated separately across all applied currents (Fig. 2K), yielding average values of 347.3 ± 95.3 V for argon ionization and 124.5 ± 9.0 V for chlorine evolution.

The theoretical minimum energy required to produce 1 kg of magnesium via electrolysis can be calculated using Faraday’s law and the operating voltage of the cell.

Based on the reaction MgCl₂ → Mg + Cl₂ (2e⁻ per atom), the charge required was calculated using Eq. (1):

$$Q=\,\frac{{2\times 96385C{{\cdot }}\,{mol}}^{-1}}{{0.024305\,{kg}\,{{\cdot }}{mol}}^{-1}}\,\approx {7.94\,\times \,{10}^{6}C{{\cdot }}{kg}}^{-1}$$

(1)

The electrical energy (in kWh·kg⁻¹) was then calculated using Eq. (2)

$${E}_{{kWh}}=\frac{{E}_{J}}{{3.6\,\times \,{10}^{6}\,J{{\cdot }}{kW}}^{-1}{{\cdot }}{h}^{-1}}=\frac{V\,\times Q}{{3.6\,\times \,{10}^{6}\,J{{\cdot }}{kW}}^{-1}{{\cdot }}{h}^{-1}}\,\approx 2.205\,\times V$$

(2)

Where V is the average cell voltage (in volts).

Given the high operating voltage (V) of 115.5–133.5 V in this system, primarily driven by the anode potential, the theoretical specific energy consumption ranges from 254.7 to 294.4 kWh·kg⁻¹.

This value significantly exceeds the typical energy consumption for magnesium electrolysis in industry (13.2–15.4 kWh per kg of Mg, corresponding to a cell voltage of 6.0–7.0 V). The main cause of this large discrepancy is due to fundamentally different anode reactions. The traditional process uses carbon anodes for the oxidation of chloride ions (2Cl⁻ → Cl₂ + 2e⁻) and operates at a relatively low overpotential. In contrast, this work employs an argon plasma anode, where the energy-intensive processes of gas ionization and plasma maintenance require a very high voltage. Thus, the increased energy use indicates the inherent cost of driving inert gas ionization rather than traditional chloride oxidation. Despite this, the technology provides a clear environmental advantage by eliminating the direct CO₂ emissions associated with carbon anode consumption (≈5~10 kg CO₂ per kg of Mg in traditional methods), paving the way for green metallurgy.

To address the high energy demand, future research must prioritize lowering the anode overpotential via two main strategies: electrode/interface optimization and system-level integration. The former involves selecting gas systems with lower ionization energies and employing interface engineering to improve charge transfer. The latter entails coupling the process with renewable energy. This integration not only enhances the net environmental benefit but also creates a pathway to monetize carbon savings, significantly improving the technology’s economic feasibility and application prospects.

As shown in Fig. 2A–I, during the initial stage of electrolysis, argon gas between the tip of the tungsten anode and the electrolyte breaks down at potentials above 400 V. The breakdown voltage varied significantly with current, reaching as high as 800 V, likely due to differences in the anode-to-electrolyte distance across experiments. After plasma formation, the anode potential stabilized below 150 V. At most current levels, argon ionization and chlorine evolution alternated, causing noticeable potential fluctuations. However, a stable chlorine evolution reaction occurred when the current exceeded 0.8 A. Therefore, a current of 0.8 A was chosen for subsequent electrolysis experiments.

Once argon plasma was established, electrolysis continued through the oxidation of chloride ions, producing chlorine gas. The concentration of chlorine gas at different currents was measured using a chlorine sensor (Fig. 2L). The detected chlorine concentration increased linearly with current at a rate of 104.1 vol% per ampere, which contrasts with the theoretical increase of 278.6 vol% per ampere. This discrepancy may be attributed to the following factors:

(a) Incomplete mixing of chlorine with argon in the reactor can lead to an underestimation of chlorine concentration.

(b) Chlorine may react with the tungsten anode at operating temperatures, as supported by the negative Gibbs free energy change that suggests spontaneous tungsten chlorination, thereby reducing detectable chlorine gas yield.

The anode current efficiency increased linearly with current at a rate of 17.5% per ampere, reaching a maximum efficiency of 35.5%.

The current study is limited to laboratory-scale, single-electrode experiments. Since scaling up by increasing the current is impractical, we have designed a multi-anode system (Fig. 3) to enable larger-scale experimentation. Future work will involve a more systematic investigation based on this setup.

Fig. 3

Schematic of magnesium production by chloride molten salt electrolysis with multi-argon plasma anodes.

Given magnesium’s lower density relative to aluminum and iron, the liquid metal readily floats, which, in a conventional single-chamber cell, exposes it to the anode gas environment and promotes oxidation. To address this, a dual-chamber setup was used in the multi-anode experiments, featuring separate cathode and anode chambers. This physical barrier effectively prevented contact. Liquid magnesium was subsequently safely extracted from the cathode chamber using a vacuum transfer ladle, resulting in significant improvements in product purity and current efficiency. Furthermore, the gas produced at the anode was purified, enabling argon recovery and improving the economic and environmental performance of the experimental protocol.

Tungsten concentrations in the electrolyte after electrolysis were measured to evaluate corrosion behavior under various current conditions (Fig. 4A). No evident relationship was found between tungsten dissolution and current intensity, suggesting that the collector tungsten filament does not actively engage in electrochemical reactions. Optical images of tungsten collectors, with and without BN protection, were recorded under various current conditions during the 20-min electrolysis process (Supplementary Fig. 2).

Fig. 4: Corrosion of tungsten collectors during the electrolysis process.

A Tungsten content in electrolytes with or without BN protection under different currents ranging from 0.1 A to 0.9 A during 20 min electrolysis. B Gibbs free energy change for tungsten and BN chlorination as a function of temperature (HSC Chemistry 6.0).

A non-uniform, irregular deposit is found on the surface of the bare tungsten anode collector. Analysis of the electrolyte’s tungsten content indicates that this deposit likely serves as a protective layer, preventing direct interaction between tungsten and the anodic chlorine gas and thereby preventing corrosion. Thermodynamically, as shown in Fig. 4B, the chlorination of both tungsten (to WCl₆) and boron nitride (to BCl₃) is spontaneous (ΔG < 0) under the experimental conditions. However, due to the high volatility of WCl₆, the observed deposit is mainly caused by the condensation of volatile species from the electrolyte, rather than solid corrosion products.

Given the stochastic nature of electrolyte evaporation and the subsequent deposition on the tungsten wire surface, this study applied a boron nitride (BN) coating to the tungsten wire three times uniformly before experimentation to reduce oxidation of the tungsten wire by chlorine gas (the anode product). The BN-coated tip of the tungsten wire was gently polished with 3000-mesh sandpaper to ensure good electrical contact. After conducting electrolysis experiments under different current conditions, the average tungsten concentration was 0.001 wt. % with BN-coated tungsten filaments, compared to 0.004 wt. % with uncoated filaments, demonstrating the effectiveness of BN coating in reducing chlorine-induced corrosion.

However, post-electrolysis inspection of the tungsten anode collector showed partial degradation of the BN coating. Two main mechanisms account for this damage:

(a) Mechanical failure: Under high-temperature experimental conditions, the BN coating failed to form a strong interface with the tungsten wire. During sample extraction, mechanical stress caused part of the coating to detach from the substrate.

(b) Chemical consumption: Thermodynamic calculations (Fig. 4B) show that BN can be oxidized by chlorine gas to form boron trichloride (BCl₃), a volatile compound that gradually depletes the coating.

Notably, electrolyte evaporation occurred during the experiments, resulting in surface deposits composed of an electrolyte-BN mixture. Although the BN coating provides measurable corrosion protection, its overall effectiveness is compromised by the combined effects of high temperature and a highly oxidative environment. Future research will focus on exploring additional protective coatings for tungsten wires to further improve corrosion resistance.

Analysis of cathodic products

Electrolysis was carried out using argon plasma as the anode at a steady current of 0.8 A for 6 h. Afterward, the cathodic products were collected and thoroughly analyzed. The results are shown in Fig. 5.

Fig. 5: Analysis of the cathodic product flowing out of the corundum tube in NaCl – KCl – MgCl₂ molten salt at 953 K after 6 h of electrolysis with an Ar plasma anode at a current of 0.8 A.

A Optical graphs of the cathodic product. B XRD pattern of the cathodic product. C SEM images and element mapping of the cathodic product. D SEM image and EDS spectrum of the cathodic product.

As shown in Fig. 5A, an irregularly shaped deposit with distinct black spheres was observed on the surface of the corundum tube. A metallic luster was evident around a small opening, initially suggesting the presence of metal. A section of this material was extracted for further analysis. X-ray diffraction (XRD) confirmed that the spherical particles on the corundum tube consist of pure magnesium (Fig. 5B).

To further characterize the cathodic product, the extracted sample was examined using scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM-EDS), as shown in Fig. 5C. Elemental analysis confirmed that the sample primarily consisted of magnesium, with small amounts of potassium and chlorine detected, indicating the presence of residual KCl-based molten salt electrolyte adhering to the magnesium metal. Two distinct regions of the sample were chosen for point-specific elemental analysis (Fig. 5D). In the metallic area with less electrolyte residue, the magnesium content reached about 95 wt. % after excluding any incidental gold (Au) from sample preparation. In contrast, the region with higher electrolyte retention had a magnesium content of around 93 wt. %.

Cathodic products, primarily composed of magnesium, were observed emerging from a small opening on the corundum tube. Because magnesium has a lower density than the molten salt electrolyte, it generally floats on the surface of the bath. It is also known that some liquid magnesium may dissolve into the molten electrolyte. To reduce such dissolution losses, the cathode assembly was quickly removed at the end of electrolysis.

The corundum tube was then cut through diametrically at the aperture, revealing a nearby region with a metallic shine. A transverse cross-section was made 2 mm above this aperture, exposing magnesium and trapped electrolyte inside the tube (Supplementary Fig. 3A). SEM-EDS analysis was conducted separately on the metal and electrolyte phases in this region (Supplementary Fig. 3B, C). Supplementary Fig. 3B confirmed that the metallic portion was primarily magnesium with traces of occluded electrolyte, consistent with the results in Fig. 5. Supplementary Fig. 3C showed the coexistence of magnesium and salt, indicating possible dissolution of magnesium into the electrolyte—a phenomenon due to the relatively high solubility of liquid magnesium in molten salts.

Additional XRD (Supplementary Fig. 4) and SEM-EDS (Supplementary Figs. 5–8) analyses performed both inside and outside the corundum tube consistently detected metallic magnesium within the electrolyte phase. These results collectively support the conclusion that liquid magnesium partially dissolves into the molten electrolyte during the electrolysis process.

In-situ detection of argon plasma anode

After stabilization at various currents, plasma emission spectra were recorded within a two-minute window and compared in Fig. 6. Six characteristic spectral lines were identified across all tested currents (from 0.1 A to 0.9 A). Four spectral regions were magnified for more precise visualization, as shown in Fig. 6B–E.

Fig. 6: Optical emission spectra collected from the DC argon discharge plasma during magnesium electrolysis.

A 200–1100 nm; B 550–630 nm (enlargement of A); C 670–710 nm (enlargement of A); D 720–810 nm (enlargement of A); E 810–830 nm (enlargement of A); F Dependence of Ar II peak intensity on discharge current. All experiments were performed with a constant argon flow rate of 250 mL·min⁻¹ and discharge currents ranging from 0.1 to 0.9 A.

The observed wavelengths—589.38 nm, 691.4 nm, 693.17 nm, 767.01 nm, 770.07 nm, and 819.25 nm—represent six distinct emission lines. By comparing with standard atomic spectral data from the National Institute of Standards and Technology (NIST)²⁷, the lines at 589.38 nm, 691.4 nm, and 819.25 nm were assigned to Ar II (Ar⁺). In contrast, those at 693.17 nm and 770.07 nm were attributed to W I (neutral tungsten). The line at 767.01 nm corresponds to Ar I (neutral argon). Notably, the intensities of all characteristic emission lines increased steadily with increasing current. The lack of observable chlorine atomic lines in the spectrum can be attributed to the timing of the detection. Since the measurement was taken early in the experiment, the chlorine concentration in the argon flow was still limited, preventing its spectral signature from being clearly detected as it emerged. Lu et al.^28,29 developed a micro-plasma spectroscopy system for real-time analysis (MIPECA) that successfully excited the characteristic spectral signals of multiple elements under different operating conditions. Notably, the intensities of all characteristic emission lines steadily increased with rising current.

Building on our previous research^25,26, it should be noted that neither Ar I nor W I participates in the chemical reactions at the plasma anode. Therefore, we examine the relationship between the intensity of the characteristic Ar II spectral lines and the applied current, as shown in Fig. 6F. Across all monitored wavelengths, the emission intensity increases with increasing current.

To further explore this behavior, the functional dependence of the Ar II line intensity on current was examined in detail (Fig. 7). Based on the observed trends, the current range was divided into three distinct regimes: stage I (0.1–0.5 A), stage II (0.5–0.8 A), and stage III (0.8–0.9 A).

Fig. 7

The relationship between the increase in Ar II (Ar⁺) spectral line intensities and current.

In stage I, the increases in Ar II spectral line intensities at 589.38 nm, 691.4 nm, and 819.25 nm remain nearly constant at 250.2 a.u., 30 a.u., and 106 a.u., respectively, exhibiting little dependence on current. In this stage, the intensity (y₁) varies linearly with current (x₁) according to: y₁ = k₁·x₁ + b₁.

In stage II, the intensity increases at each wavelength show linearly with current, with rates of 1256.7 a.u.·A⁻¹, 290 a.u.·A⁻¹, and 770 a.u.·A⁻¹, respectively. Here, the intensity (y₂) follows a quadratic relationship with current (x₂): y₂ = k₂·x₂² + b₂.

In stage III, the intensity increases again exhibit linear dependence on current, but at much higher rates of 17,190 a.u.·A⁻¹, 1730 a.u.·A⁻¹, and 5310 a.u.·A⁻¹, respectively. The intensity (y₃) maintains a quadratic dependence on current (x₃): y₃ = k₃·x₃² + b₃.

A summary of the Ar II spectral line intensities at 589.38 nm, 691.4 nm, and 819.25 nm as functions of current is provided in Table 1.

Table 1 Intensity of Ar II characteristic lines at 589.38 nm, 691.4 nm, and 819.25 nm as a function of current

In atomic emission spectroscopy, the intensity of characteristic spectral lines generally reflects the concentration of the corresponding species in the plasma. Therefore, our results show that the concentration of Ar II in the plasma increases steadily with increasing current. The variation in the growth rate of Ar II content over different current intervals may be due to changes in the plasma discharge mode³⁰.

In this study, Ar II in the plasma is reduced by Cl^- from the electrolyte, producing neutral argon (Ar), while electrons from the plasma migrate through the anode collector to the cathode, where they reduce Mg²⁺ ions. As a result, the plasma remains macroscopically neutral and primarily consists of electrons, Ar⁺, neutral Ar, and their excited ionic and atomic species.

Mechanism of argon anode reactions

Upon melting, NaCl and KCl fully dissociate into Na⁺, K⁺, and Cl^-. In contrast, the complete ionization of MgCl₂ is limited due to its mixed ionic – covalent nature. Under these conditions, MgCl₂ partially ionizes, producing ions such as Mg²⁺, Cl^-, MgCl₃^-, and MgCl₄^2-. Therefore, the electrolyte used in this study primarily consists of the following ionic species: Na⁺, K⁺, Mg²⁺, Cl^-, MgCl₃^-, and MgCl₄^2-31,32,33. Given the low concentration of MgCl₂, Cl^- remains the predominant anion, with smaller amounts of anionic complexes such as MgCl₃^- and MgCl₄^2-.

We employed Density Functional Theory (DFT) to model the plasma-assisted anodic reaction because no dedicated thermodynamic software is available for plasma conditions. The optimized DFT models for each species and the energy values of the reactants and products in the argon-plasma anode reaction are listed in Supplementary Tables S1, S2, respectively. The results, presented in Fig. 8, show that the electrochemical reaction ΔG for 2Cl^- - 2e^- = Cl₂ is positive in the absence of Ar⁺, indicating that the oxidation of Cl^- to Cl₂ is non-spontaneous and needs an external driving force. In contrast, when argon plasma is used as the anode, Ar⁺ actively participates in the reaction by interacting with Cl^- ions in the electrolyte. This process facilitates the oxidation of Cl^- and leads to the spontaneous formation of Cl₂. The reaction (2Ar⁺ + 2Cl^- = 2Ar + Cl₂) has a negative ΔG, indicating that the oxidation of Cl^- to Cl₂ occurs spontaneously in the presence of Ar⁺.

Fig. 8

ΔG(eV) for the oxidation of Cl^- to Cl₂ in the presence or absence of Ar⁺ for different temperatures.

Based on the above analysis, it can be concluded that Ar⁺ in the plasma facilitates the oxidation of Cl^- in the electrolyte. However, the Gibbs free energy change associated with the formation of argon plasma²⁵ indicates that generating argon plasma remains the most challenging step in the entire process when it is used as an anode for magnesium electrolysis.

During magnesium electrolysis, metallic magnesium—which has a lower density than the electrolyte—tends to float on the surface. At the same time, anode gases (such as Cl₂) evolve and rise upward, which can cause re-chlorination of the magnesium metal produced at the cathode. To address this issue, industrial magnesium electrolysis cells are often designed with a partition that separates the cathode and anode compartments.

In this study, argon plasma is used as the anode for magnesium electrolysis. A key design feature is the physical isolation of the cathode from the argon plasma reaction zone, accomplished by a quartz partition that divides the cathode and anode chambers. This separation prevents the metallic magnesium produced at the cathode from re-chlorination by the chlorine gas generated at the anode, as shown schematically in Fig. 9.

Fig. 9

Schematic illustration of magnesium electrolysis using argon plasma as an inert anode.

The partition clearly separates the anode and cathode reaction chambers. In this setup, the anode process involves the oxidation of Cl^- to Cl₂, aided by Ar⁺ from the argon plasma, as shown in Eq. (3)

$${2{{{\rm{Ar}}}}}^{+}+{2{{{\rm{Cl}}}}}^{-}\to 2{{{\rm{Ar}}}}+{{{{\rm{Cl}}}}}_{2}$$

(3)

In addition to the evolution of chlorine gas at the anode, Ar⁺ ions in the plasma are reduced to neutral argon (Ar), which can be re-ionized to Ar⁺ during electrolysis, enabling continuous recycling of the argon species. As a result, argon plasma functions as an inert anode in magnesium electrolysis, allowing a carbon-free process.

Under the combined influence of the electric field and concentration gradients, residual Mg²⁺ ions in the anodic region migrate toward the cathode. At the cathode surface, they accept electrons and undergo reduction to form metallic magnesium, as described by Eq. (4)

$${{{{\rm{Mg}}}}}^{2+}+{2{{{\rm{e}}}}}^{-}\to {{{\rm{Mg}}}}$$

(4)

Conclusions

In summary, a stable argon plasma was successfully generated and used as an anode in magnesium electrolysis, enabling a carbon-free continuous process. The average potential required for argon ionization was measured at 347.3 ± 95.3 V, compared to 124.5 ± 9.0 V for the chlorine evolution reaction. Chlorine gas produced at the anode causes the tungsten filament collector to be consumed; however, this can be prevented by evenly coating the filament surface with boron nitride (BN). The existence of Ar⁺ in the plasma was confirmed by atomic emission spectroscopy, and its concentration was observed to rise as the current increased. Thermodynamic analysis further shows that Ar⁺ facilitates the oxidation of Cl^- through the reaction: 2Ar⁺ + 2Cl^- → 2Ar + Cl₂.

The argon plasma non-contact anode signifies a significant shift in inert anode technology, effectively addressing long-standing problems with traditional carbon-based anodes. The significance of this work extends beyond magnesium electrolysis, as the unique properties and catalytic capabilities of argon plasma may be helpful in other high-temperature, corrosive molten salt electrolysis processes. We expect this method to play a key role in advancing clean energy technologies and sustainable materials production.

Materials and methods

Preparation of electrolytes and electrodes

Sodium chloride (NaCl, 99.8%, GR grade, Aladdin), anhydrous potassium chloride (KCl, 99.5%, AR grade, Aladdin), and anhydrous magnesium chloride (MgCl₂, 99.9%, Aladdin) were stored in a glove box (MBRAUN MB 200B, Germany) with water and oxygen contents below 0.1 ppm. An electrolyte with the composition of 11.7 wt. % NaCl, 69.4 wt. % KCl, and 19.0 wt. % MgCl₂ (Supplementary Fig. 9) was accurately prepared. Following preparation, the prepared electrolyte was transferred into a custom quartz crucible and promptly sealed with aluminum foil in the glove box. It was then moved directly into the transparent electrolysis cell, which had been pre-heated to 473 K to ensure a dry environment. The temperature was then raised to 673 K and maintained for 48 h to reduce trace moisture from handling, thereby minimizing potential interference with the experimental process. Finally, it was gradually heated to the target operating temperature of 953 K.

Following complete melting of the electrolyte, the anode system is carefully adjusted to precisely position the plasma-generating tungsten wire (φ = 1 mm) above the melt. A quartz anode gas collector is then sealed onto the molten salt surface to establish a stable argon atmosphere. This design serves two purposes: it effectively prevents the leakage of anodic gas products. Also, it eliminates the risk of cathodic products entering the anodic reaction zone and being oxidized by the anodic gas. Throughout electrolysis, the stability of the plasma is maintained by fine-tuning the vertical position of the tungsten filament.

A tungsten cathode (φ = 2 mm) was used, protected by a sealed-end corundum tube with a small hole 2 mm from the bottom. This setup effectively prevented the dissolution of liquid magnesium from the cathode into the molten salt. An Ag/AgCl reference electrode was connected to the system via a separate tungsten wire (φ = 1 mm). Before each experiment, the tungsten electrodes (99.99%, Qinghe County Jinou Metal Materials Co., China) were polished sequentially with progressively finer sandpapers, rinsed thoroughly with distilled water, and air-dried.

Characterization techniques

Upon completion of the experimental setup, high-purity argon gas (99.99%, Shenyang Shuntai Gas Co., China) was introduced into the system at a flow rate of 1 L · min⁻¹ to purge residual air. The flow was then precisely adjusted to 250 mL · min⁻¹ using a mass flow controller (AST10-DL, 0–300 sccm, Beijing Asert Instruments Ltd., China).

A high-voltage DC power supply (JP10001D, 1A, 1000 V, Wuxi Annaisi Electronic Technology Co., Ltd., China) was used to conduct the electrolysis experiments. The voltage between the tungsten filament anode and the reference electrode was monitored with a Keysight digital multimeter (34461A) and recorded via BenchVue software.

Gaseous products at the anode were qualitatively and quantitatively analyzed using a chlorine gas sensor (0–200 ppm, Shenzhen Xinchuang Andi Electronic Technology Co., Ltd., China). The composition of the anode plasma was characterized in situ using a fiber optic spectrometer (Maya 2000 Pro, Ocean Optics, USA).

Phase identification of the cathode products was performed by X-ray diffraction (XRD, Empyrean, PANalytical B.V., Netherlands). Morphological and elemental analyses were carried out using scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDX) on a Thermo Scientific Apreo 2 instrument.

Calculation methods

The anode gas produced during electrolysis contains highly corrosive components, making it unsuitable for direct analysis using conventional gas chromatography. To accurately determine the gas concentration, which exceeded the detection limit of the available equipment, the samples were diluted by a factor of 100 before measurement. The volume of chlorine gas produced was then calculated according to Eq. (5):

$${V}_{{{{{\rm{Cl}}}}}_{2}}={c}_{{{{{\rm{Cl}}}}}_{2}}\times 100\times {V}_{{Total}}$$

(5)

Where: ${V}_{{{{{\rm{Cl}}}}}_{2}}$ is the volume of chlorine gas obtained from the experiment (ml); ${c}_{{{{{\rm{Cl}}}}}_{2}}$ is the concentration detected by the chlorine sensor; 100 is the dilution factor applied to the anode gas; ${V}_{{Total}}$ is the total volume of the diluted gas (5000 ml in this study).

According to Faraday’s law, the electrochemical equivalent of chlorine gas (Cl₂) is 0.6614 g·A⁻¹·h⁻¹. Thus, the charge equivalent corresponding to the amount of anode product can be expressed by Eq. (6):

$${Q}_{{{{\rm{Experimental\; yeild}}}}}=\frac{{m}_{{{{{\rm{Cl}}}}}_{2}}}{0.6614}=\frac{n\,\times M}{0.6614}=\frac{{V}_{{{{{\rm{Cl}}}}}_{2}}\,\times M}{0.6614\,\times \,{V}_{{{{\rm{m}}}}}}$$

(6)

Where: M is the molar mass of Cl₂; V_m is the molar volume of gas, taken as 24.5 L·mol⁻¹ at 25 °C and 101 kPa.

The theoretical yield of anode gas for different current levels was determined based on the total charge input, as given by Eq. (7):

$${Q}_{{Theoretical\; yield}}={I \cdot t}$$

(7)

Where: I is the current levels (A); t is the electrolysis time (h).

The current efficiency at various applied currents was then calculated using Eq. (8):

$${Current}\,{Efficiency}=\frac{{Q}_{{Experimental}\,{yeild}}}{{Q}_{{Theoretical}\,{yeild}}}$$

(8)

Computational calculations

Given the relative simplicity of the systems involved (primarily monatomic and diatomic species), the initial geometries were constructed directly using Gaussview 5.0 software. Geometrical configuration optimization and frequency calculations for all reactants and products under various temperature conditions were performed at the B3LYP/6-311 G(d,p) level of theory^34,35,36 using Gaussian 09 software³⁷. The optimized structures and thermally corrected Gibbs free energies were obtained from these calculations. Single-point energy evaluations were carried out with the double-hybrid functional B2PLYP-D3^37,38,39 combined with the Def2-TZVPP basis set⁴⁰. The total Gibbs free energy G was calculated by using Eq. (9):

$$G={E}_{{{{\rm{TC}}}}}+{E}_{{{{\rm{SP}}}}}$$

(9)

Where: E_TC is the thermal correction to Gibbs free energy (eV); E_SP is the single-point energy (total electronic energy from DFT calculation) (eV).

The Gibbs free energy change ΔG for a reaction was computed using Eq. (10):

$$\varDelta G={\sum }_{i}{{v}_{i}G}_{i}-{\sum }_{j}{{v}_{j}G}_{j}$$

(10)

Where: $\varDelta G$ is the total Gibbs free energy change of the reaction (eV); ${v}_{i}$ and ${v}_{j}$ represent the stoichiometric coefficients of products and reactants, respectively; ${G}_{i}$ and ${G}_{j}$ denote the Gibbs free energies of products and reactants, respectively (eV).