Energy-Efficient CO₂ Conversion for Carbon Utilization Using a Gliding Arc/Glow Discharge with Magnetic Field Acceleration—Optimization and Characterization

Abstract

The dry conversion of CO₂ into CO and O₂ provides an attractive path for CO₂ utilization which allows for the use of the CO produced for the synthesis of valuable hydrocarbons. In the following work, the CO₂ conversion is driven by an arc discharge at atmospheric pressure, producing hot plasma. This study presents a series of experiments aiming to optimize the process. The results obtained include the energy efficiency and the conversion rate of the process, as well as the electrical parameters of the discharge (current and voltage signals). In addition, optical emission spectroscopy diagnostics based on an analysis of C₂’s Swan bands are used to determine the gas temperature in the discharge. The data is analyzed according to several aspects—an analysis of the arc’s motion based on the electrical signals; an analysis of the effect of the gas flow and the discharge current on the discharge performance for CO₂ conversion; and an analysis of the vibrational and rotational temperatures of the arc channel. The results show significant improvements over previous studies. Relatively high gas conversion and energy efficiency are achieved due to the arc acceleration caused by the Lorentz force. The rotational (gas) temperatures are in the order of 5500–6000 K.

1. Introduction

According to the Global Carbon Budget 2024 [1], atmospheric carbon dioxide concentrations have increased by approximately 52% since the onset of the Industrial Revolution, rising from 278 ppm in 1750 to 422 ppm in 2024. This significant increase is primarily attributed to deforestation, land-use changes, and the combustion of fossil fuels—namely coal, oil, and natural gas. Notably, due to intensified wildfire activity in both North and South America, global CO₂ emissions in 2024 are estimated to be 11–32% higher than the average recorded between 2014 and 2023.

In contrast, fossil fuel emissions in the European Union are projected to decline by 2.8% in 2024, driven by regulatory efforts aiming to achieve net-zero carbon emissions by 2050. To reduce the reliance on fossil fuels, various renewable energy sources are being integrated into the electric power system. Additionally, technologies such as Carbon Capture and Storage (CCS) and Carbon Capture, Utilization, and Storage (CCUS) have been developed to reduce the amount of CO₂ released into the atmosphere. CCS focuses on the long-term storage of captured CO₂, while CCUS seeks to convert CO₂ into usable products, thereby creating a closed carbon loop [2,3,4].

However, CCUS alone cannot resolve the carbon emissions challenge, largely due to the significant energy demand associated with the utilization processes. Furthermore, if the energy used in these processes is derived from carbon-intensive sources, this can additionally contribute to emissions [2]. Nevertheless, the utilization phase is critical to achieving a sustainable CO₂ cycle [5].

CO₂ utilization methods include photocatalytic, chemical, electrochemical, thermochemical, and plasma-based approaches. Among these, photocatalytic methods remain far from industrial deployment due to their low product yields [4]. Electrochemical processes (non-plasma based) often result in a mixture of products, complicating the selectivity and reducing the energy efficiency. As of now, industrial scalability has not yet been achieved in electrochemical CO₂ dissociation.

The thermochemical conversion of CO₂ is also limited by high operational costs and considerable energy requirements which negatively impact its overall efficiency. In contrast, plasma-based technologies provide a promising alternative, offering a more energy-efficient pathway for CO₂ dissociation. Various types of non-thermal plasma discharges have been explored, including dielectric barrier discharges (DBDs), microwave (MW) discharges, radio-frequency discharges, inductively coupled plasma, and gliding arc discharges (GADs) [3,4,6,7]. Most of these systems operate at sub-atmospheric pressures, which require vacuum equipment that adds both complexity and energy costs. While DBDs and MW discharges can function at atmospheric pressure, they typically exhibit lower energy efficiencies under such conditions.

GADs stand out as a particularly suitable option for atmospheric-pressure CO₂ conversion, combining simplicity with relatively high energy efficiency. Their ability to operate without vacuum infrastructure not only streamlines the system but also lowers both the setup and the operational costs. Moreover, they offer flexibility in handling various gas compositions and compatibility with continuous-flow systems. These features make GADs well suited to practical, application-driven research, especially in the context of sustainable CO₂ utilization technologies.

However, their CO₂ conversion rates remain limited—typically not exceeding 10–15%—due to the short residence time of gas within the arc [3,6]. The conventional GAD setup consists of two diverging electrodes [8], in some cases placed between dielectric walls. The gas flow inlet is near the shortest electrode gap, and the gas velocity distribution is along the electrode’s length. An arc is initiated at the shortest electrode gap, travels downstream, and expands (elongates) due to the gas flow friction until reaching a critical length, after which point it extinguishes and reignites. To improve conversion rates, strategies such as applying a magnetic field for arc acceleration or stabilization have been proposed, as this can alter the relative motion of the arc and the gas flow, increasing the residence time of CO₂ within the active plasma region [7,9].

In addition to CO₂ decomposition, GADs have been extensively investigated over the past five years as an innovative physicochemical technique for the treatment of gases, waste, biomass, organic compounds, and water at atmospheric pressure [10]. A particularly promising application of GADs is the generation of NO and NO₂ from N₂ and O₂, which are widely used in medicine and agriculture [11,12]. Another practical use is plasma-activated water (PAW), which has gained attention in various fields. In medicine, PAW has been explored for applications such as sterilization, wound healing, infection control, and even cancer therapy [13,14,15,16]. In the agricultural and food industries, PAW—enriched with nitrites and nitrates—has been utilized to wash food products and plants, as well as for spraying crops and seeds [17].

To address the rapidly increasing global energy demand, GADs have also been applied in the decomposition of biomass, offering a sustainable energy source with near-zero net carbon emissions [18,19,20]. Moreover, methane (CH₄), another potent greenhouse gas, can be effectively converted using GADs into carbon and hydrogen. This process not only reduces methane emissions but also yields valuable by-products that contribute to cleaner energy solutions [21,22,23,24].

In order to optimize the conversion process for thermochemical systems, in general, it is desirable to operate with high gas flows at atmospheric pressure. This avoids the storage of bulky components, reduces costs, and saves energy for compression stages [25]. Previous studies [8] have also indicated that non-equilibrium plasmas like that created in GADs, namely at a gas temperature T_g much lower than the electron temperature, are preferable for obtaining higher energy efficiencies. It is also established that depending on the value of the gas temperature, the conversion of CO₂ into CO is realized through different mechanisms [26,27,28,29], affecting energy efficiency. Therefore, it is crucial to have tools for assessing the gas temperature, as this is one of the most important parameters for this process.

One of the molecular bands widely used for emission diagnostics of carbon-containing discharges at high pressures is that in the C₂ (d³∏_g − a³∏_u) Swan system. The main reason is that they are intense and can be clearly identified over the visible range. However, to obtain reliable and accurate information about plasma, a precise analysis of the C₂ emission spectra is needed. Usually, it is assumed that there is an equivalence between the rotational temperature of the emitting species and the gas temperature. Most often, it is assumed that the partition functions of a rovibrational distribution follow the Maxwell –Boltzmann distribution, and the rotational (T_rot) and vibrational (T_vib) temperatures can be obtained using so-called Boltzmann plots. Nevertheless, overpopulation of the high-vibrational levels, resulting in “high-pressure bands” in the C₂ system, in combustion studies [30] makes analyses of molecular spectra more complicated. In these cases, the assumption that the vibrational levels follow the Boltzmann distribution function introduces systematic errors into the determination of the rotational temperature.

In this work, results from optical emission spectroscopy (OES) diagnostics are presented that determine the gas temperature in a magnetically accelerated gliding discharge (MAGD, arc or glow). The analysis of registered emission spectra using the open source MassiveOES software (https://bitbucket.org/OES_muni/massiveoes/src/master/ accessed on 13 July 2025) and the commercial Specair 3.0 software provides information about the vibrational and rotational temperatures. Under our experimental conditions, these temperatures appear to coincide well within their uncertainties and can be used to assess the gas temperature of the plasma. More details are presented in Section 3.3.

Overall, this study aims to build upon and expand our previous research [7,9] by investigating the performance of a MAGD under a higher electric current, an improved power supply, and additional characterization. Significant improvements have been achieved over the previous results, as will be summarized in Section 3.2 and Section 3.4. As a result, the CO₂ conversion setup based on the gliding arc discharge discussed in this work provides significant advantages: (1) low initial system costs due to the simplicity of the setup and the use of abundant materials and technologies, including a simple discharge design and a robust, low-cost DC power supply; (2) low operational expenses (e.g., occasional electrode replacement and cleaning after extended operation); (3) small volume; and (4) atmospheric-pressure operation.

Although our goal is to optimize and characterize the discharge for CO₂ conversion, this remains a complex, high-dimensional optimization problem. While numerical modeling could support optimization, developing predictive models for gliding discharges is highly challenging. Developing a model that captures the 3D, time-dependent, and stochastic nature of the discharge—especially with CO₂ chemistry—requires time that often exceeds the duration of experimental campaigns. Given these limitations, we have focused on systematic experimental investigation. This approach ensures reliable results for practical configurations. However, because the setup is designed for real-world applications, featuring, for example, active cooling for long-term stability, diagnostic access is also restricted. The cooling elements obstruct optical paths, limiting our ability to collect detailed emission data from the entire plasma volume, which is a drawback. However, designing a more research-oriented setup would compromise its practical relevance, so our priority has been to develop a stable, application-ready system and to extract key insights for further optimization and development.

This article has the following structure. In Section 2, we describe the experimental setup in detail, including the calculation of the quantities of interest and their associated uncertainties. In Section 3, the results obtained are presented and discussed, including the time evolution of the arc’s quasi-periodic motion based on an analysis of the electrical characteristics—the current and voltage signals (Section 3.1). In Section 3.2, selected results on the conversion rate and the energy efficiency of the CO₂ dissociation are presented. Section 3.3 includes the results on the vibrational and rotational gas temperatures obtained through emission spectroscopy. The final subsection presents a comparison of the discharge performance obtained here with relevant results from the literature.

2. Materials and Methods

2.1. The Experimental Setup

The experimental setup used in this work is shown schematically in Figure 1a. To prevent CO from entering the laboratory, the discharge device is placed in a sealed glass chamber, which remains at atmospheric pressure, along with the rest of the system. The CO₂ gas is controlled using the Bronkhorst (Ruurlo, The Netherlands) EL-FLOW F-201CM Mass Flow Controller (MFC) and is supplied to the discharge through a nozzle with a square cross-section of 5 × 3 mm (see Figure 2a). The gas flow rate is set and measured in “normal” L/min, referenced to conditions of 760 Torr and 273.15 K, as defined by the MFC specifications. Corrections for the actual operating pressure and temperature are applied afterward. The gas mixture produced by the discharge is analyzed using an infrared Fourier spectrometer. The gas is sampled using a small pump and fed to a Specac (Orpington, UK) Storm 10 gas cell (with a 10 cm optical path) installed in a Perkin Elmer (Shelton, CT, USA) Frontier MIR/NIR FT-IR spectrometer. It measures the concentration of CO gas by analyzing the CO absorption line at 2209 cm⁻¹ (4527 nm). Based on the CO concentration obtained and assuming negligible pure carbon quantities, the CO₂ dissociation is calculated, as explained in Section 2.4.

The high-voltage power supply (HV PS) is based on simple classical 50 Hz transformers connected as shown in Figure 1b. It consists of 3 groups of 3 high-voltage “neon” transformers, connected to a 3-phase line supply through variable series inductances with a maximum value of 20 mH. The transformer groups have different maximum current nominals—group 1: 3 × SIET (Rome, Italy) Metalbox 15 kV/30 mA; group 2: 3 × SIET Metalbox 15 kV/60 mA; and group 3: 3 × SIET Metalbox 10 kV/100 mA. The transformers have a magnetic shunt in the magnetic core to limit the output current and energy, even in the case of an output short circuit. This design allows them to operate in a regime close to that of a current source. However, the operating current is not firmly fixed at a certain value but partially depends on the load (the discharge), which, in our case, dynamically varies over time. The output current and voltage can be controlled by the additional series inductance L_s, which modifies both the voltage and the current—both are reduced when the inductance is increased. The three groups of transformers provide approximate maximum voltage and current values as follows: (1) 10.5 kV/100 mA, (2) 10.5 kV/250 mA, and (3) 7 kV/500 mA. Additionally, a ballast resistance R_b is installed in series with the discharge, with values of 5 kΩ, 2.5 kΩ, and 1 kΩ for groups 1 to 3, respectively. The purpose of R_b is not to control the current but to reduce the ringing induced in the circuit due to continuous discharge breakdowns and the presence of significant inductance in the circuit.

The use of this 3-phase transformer circuit, instead of a single-phase one, is obviously motivated by the need to produce voltage and current closer to those in a DC power supply without output filters, which are more difficult to implement at high voltages. Due to the lack of an output filter, there is a ripple in the PS at 300 Hz, but it is relatively minor and does not change the discharge behavior. Of course, a modern switching PS could be used instead, offering better efficiency and flexibility in general. However, the switching HV PSs available in our lab are not suitable for this kind of discharge and require a very high ballast resistance to operate. Based on our experience, most commercially available switching PSs are not well suited to loads such as GADs and often need certain customizations to operate properly. For example, the typical use of output RC filters is not suitable for loads with a steep negative slope in the current–voltage characteristics (i.e., the discharge voltage decreases as the current increases) like those seen for low-current GADs. Moreover, frequent breakdowns caused by arc gliding lead to rapid current peaks, which require a very fast active current feedback control circuit in the PS. Therefore, standard HV PSs with RC filters can mainly be used with considerable series ballast resistors R_b when powering GADs. In our previous works [7,9], we used a single-phase transformer with an inductive filter. However, the filter inductance is not sufficient to provide proper filtering at high currents above 150–200 mA. For completeness, in Section 3.2, we also show a comparison of the results obtained with a 3-phase PS and a single-phase PS.

The current and voltage signals are recorded on a Rohde & Schwarz^® (Munich, Germany) RTB2004 Digital Oscilloscope (200 MHz), equipped with a Tektronix (Beaverton, OR, USA) P6015A HV probe and a clamp-on Pintek (New Taipei City, Taiwan) PA-699 current probe (Figure 1a).

2.2. The Discharge Design and the Magnetic Field Distribution

A detailed, expanded view of the discharge is given in Figure 2a. The electrodes are classical GAD knife-shaped copper electrodes with a thickness of 3 mm, a height of 72 mm, and a width of 34 mm. The shortest interelectrode distance is 3.5 mm. The gas flow is channeled between the electrodes and the quartz glasses. Water-cooled aluminum plates and two magnets, sized 40 × 7.4 × 60 mm, are placed on the glass panels. The distribution of the magnetic field is shown in Figure 2b, calculated using a 3D FEM model, and verified through measurements at specific points. The distance between the magnets is 37 mm, and their position relative to the lower ends of the electrodes is 27 mm. Note that an additional experimental campaign was conducted to roughly optimize the magnets’ position in the z-direction for a limited set of conditions to maximize the dissociation rate. Although the corresponding results are not presented in this article to avoid including data of limited significance, the conversion rate and the energy efficiency were obtained for three magnet positions, spaced 8 mm apart. For the different magnet positions, the conversion rate varied between 9 and 11% for the given test conditions. The 3D FEM model used to compute the magnetic field distribution shows a different location for the maximum compared to that in the previous study [9], with the maximum now occurring further along the arc’s movement. Moreover, the use of active cooling of the quartz glasses and the relatively small distance between them (3 mm) contribute to faster quenching of the processed gas and thus to a reduction in the reverse processes, thus enhancing the conversion rate and the energy efficiency.

2.3. The Experimental Arrangement for OES

An integrated optical emission from the plasma is collected in the x-direction (see Figure 1a) on the outer side of the glass chamber, within the numerical aperture (0.22NA) of a light guide (with a 400 nm inner diameter), which is positioned slightly above the top of the electrodes (see Figure 1a and Figure 2a). The collected light is delivered to the entrance slit (approximately 50 µm wide) of a Horiba (Kyoto, Japan) iHR550 Imaging Spectrometer. Using a NARVA (Plauen, Germany) magnesium hollow cathode lamp, the instrumental line profile was determined and was well approximated using a Gaussian function with a full width at half maximum of about 27 pm. An illustration of the most dominant emission feature in the visible range, in the C₂ (d³∏_g − a³∏_u) Swan band system, is shown in Figure 3. Transitions between the two triplet electronic states are notated according to the difference between the vibrational quantum numbers of the upper (v′) and lower (v′′) states. The data is normalized to the band head of the ∆v = v’ – v′′= 0 (516.5 nm) C₂ Swan band. More information on the methodology for analyzing the molecular spectra and deriving the vibrational and rotational temperatures of CO₂ is provided in Section 3.3.

2.4. Calculation of the Conversion Rate and Energy Efficiency and Their Uncertainties

The main quantities of interest when discussing CO₂ conversion are the conversion rate and the energy efficiency. The conversion rate is the relation of the difference between the inlet ($n_{{CO}_2}^{in}$) and the outlet ($n_{{CO}_2}^{out}$) gas concentrations to the inlet concentration:

$$X_{{\mathrm{C}\mathrm{O}}_2}\left[\mathrm{\%}\right]={\displaystyle \frac{n_{\mathrm{C}{\mathrm{O}}_2}^{\mathrm{i}\mathrm{n}}-n_{\mathrm{C}{\mathrm{O}}_2}^{\mathrm{o}\mathrm{u}\mathrm{t}}}{n_{\mathrm{C}{\mathrm{O}}_2}^{\mathrm{i}\mathrm{n}}}}\times 100\mathrm{\%}.$$

(1)

The conversion of CO₂ is assumed to result mainly in oxygen and CO. In the current experiment, the fraction of CO is measured through the relative absorbance of a sample of the processed gas using a Fourier spectrometer, and based on that, the CO₂ conversion is calculated using the expression

$$X_{{\mathrm{C}\mathrm{O}}_2,\mathrm{\%}}={\displaystyle \frac{Y_{\mathrm{C}\mathrm{O}}^{\mathrm{o}\mathrm{u}\mathrm{t}}}{{1-\frac{1}{2}Y}_{\mathrm{C}\mathrm{O}}^{\mathrm{o}\mathrm{u}\mathrm{t}}}}\times 100\mathrm{\%},$$

(2)

where

$$Y_{\mathrm{C}\mathrm{O}}^{\mathrm{o}\mathrm{u}\mathrm{t}}={\displaystyle \frac{I_{\mathrm{C}\mathrm{O}}^{\mathrm{o}\mathrm{u}\mathrm{t}}{p}^{\mathrm{c}\mathrm{a}\mathrm{l}}{T}^{\mathrm{o}\mathrm{u}\mathrm{t}}{C}_{\mathrm{C}\mathrm{O}}^{\mathrm{c}\mathrm{a}\mathrm{l}}}{I_{\mathrm{C}\mathrm{O}}^{\mathrm{c}\mathrm{a}\mathrm{l}}{p}^{\mathrm{o}\mathrm{u}\mathrm{t}}{T}^{\mathrm{c}\mathrm{a}\mathrm{l}}}}=\alpha I_{\mathrm{C}\mathrm{O}}^{\mathrm{o}\mathrm{u}\mathrm{t}}{\displaystyle \frac{{p^{cal}T}^{out}}{p^{out}{T}^{cal}}}$$

(3)

Here, α is defined as ${\alpha =C}_{\mathrm{C}\mathrm{O}}^{\mathrm{c}\mathrm{a}\mathrm{l}}/I_{\mathrm{C}\mathrm{O}}^{\mathrm{c}\mathrm{a}\mathrm{l}},$ where $C_{CO}^{cal}$ and $I_{CO}^{cal}$ are the relative content of CO and the measured FTIR intensity of a CO absorption line for the calibration gas mixture, respectively. As mentioned above, the absorption line at 2209 cm⁻¹, corresponding to the CO molecular band, is used for diagnostics. The coefficient α is determined through a separate series of measurements with statistical uncertainty of 1%, using a calibration gas mixture containing 20% CO and 80% CO₂. In our experiment, $C_{CO}^{cal}=0.2$. The choice of the calibration mixture’s composition is motivated by the assumption that CO₂ is primarily converted into CO and O₂, with a negligible amount of carbon C and ozone O₃. $p^{cal}$ and $T^{\mathrm{c}\mathrm{a}\mathrm{l}}$ denote the pressure and temperature during the calibration procedure, while $p^{out},T^{\mathrm{o}\mathrm{u}\mathrm{t}},\mathrm{a}\mathrm{n}\mathrm{d}{I}_{CO}^{out}$ represent the pressure, temperature, and CO line intensity of the outlet gas, respectively, after treatment of the CO₂ gas (99.9% purity) with the arc plasma. Within the CO₂ conversion rate typical for a GAD (about 5 to 25%), the dependence between the line intensities and the CO density is linear for our FTIR. Expression (2) accounts for the expansion in the gas volume resulting from the splitting of CO₂ into CO and ½ O₂ [31].

Another crucial parameter which allows us to compare the properties of the different plasma devices for gas treatment is the specific energy input (SEI). This quantity is the energy delivered per mole of gas, as follows:

$$SEI\left[{\displaystyle \frac{kJ}{L}}\right]={\displaystyle \frac{P[kJ/s]}{MFR[L/s]}},$$

(4)

where MFR is the mass flow rate of the gas in “normal” L/s.

The energy efficiency is defined as

$$\eta \left[\%\right]={\displaystyle \frac{X_{{CO}_2}\times ∆H_R}{SEI}}\times 100\%,$$

(5)

where ∆H_R = 279.8 kJ/mol is the reaction enthalpy for the CO₂ splitting reaction and $X_{{\mathrm{C}\mathrm{O}}_2}=X_{{\mathrm{C}\mathrm{O}}_2,\%}/100\%$. $\eta$ is the fraction of the energy required for the conversion of CO₂ into the SEI.

The absolute uncertainty of the conversion rate is calculated by

$$∆X_{{CO}_2}={\displaystyle \frac{4}{{\left(2-Y_{CO}^{out}\right)}^2}}{∆Y}_{\mathrm{C}\mathrm{O}}^{\mathrm{o}\mathrm{u}\mathrm{t}}.$$

(6)

The relative uncertainty in the energy efficiency is calculated using the formula

$$\left|{\displaystyle \frac{∆\eta }{\eta }}\right|=\left|{\displaystyle \frac{∆X_{{\mathrm{C}\mathrm{O}}_2}}{X_{{\mathrm{C}\mathrm{O}}_2}}}\right|+\left|{\displaystyle \frac{∆SEI}{SEI}}\right|,$$

(7)

where |∆SEI/SEI| ≈ |∆MFR/MFR| + |∆P/P|, and P is the discharge power.

3. Results and Discussion

This study is mainly experimental and provides results on and an interpretation of the dependencies of the most important quantities of interest on the different parameters (the discharge current, gas flow, etc.). The results presented are for the conversion rate of CO₂, the energy efficiency of the conversion process, the vibrational and rotational temperatures of the MAGD, and electrical parameters such as the discharge voltage and current signals, as well as their Fast Fourier Transform (FFT) spectra.

The following test conditions were examined and varied within the presented result data of the MAGD study:

• The average discharge current (I_av) and power (P_av)—due to the intrinsic non-stationary nature of the discharge and the limitations of the power supply, the discharge current cannot be fixed at a precise value, as it has experiences variations during the discharge operation and at different gas flow rates (see Figure 4a). Therefore, we provide (see

  Table 1. Main parameters and their ranges/values.

  Parameter	Values/Ranges
  Gas flow rate	2–12 L/min (normal)
  Averaged discharge current Iav	MAGD: 100 mA (95–111 mA) MAGD: 255 mA (230–312 mA) MAGD: 470 mA (455–477 mA) GAD: 280 mA (273–308 mA)
  Average discharge power Pav	MAGD: 240 W at 100 mA (235–245 W) MAGD: 335 W at 255 mA (300–350 W) MAGD: 520 W at 470 mA (511–527 W) GAD: 390 W at 280 mA (350–410 W)
  Average discharge voltage Uav	MAGD: 2.45 kV at 100 mA (2.13–2.67 kV) MAGD: 1.65 kV at 255 mA (1.1–1.94 kV) MAGD: 1.15 kV at 470 mA (1.12–1.2 kV) GAD: 1.62 kV at 280 mA (1.15–2.50 kV)
  Channel thicknesses	3 mm
  SEI	1.2–15.6 kJ/L
  Rb	5 kΩ at 100 mA 2.5 kΩ at 255 mA 1 kΩ at 470 mA
  Ls	0 H at 100 mA 0 H at 255 mA 2 mH at 470 mA

  ) the averaged values for the current and the corresponding power. The average current was set to three approximate values—100 mA, 255 mA and 470 mA—corresponding to average powers of 240, 335, and 520 W, respectively.
• The CO₂ mass flow rate in the range of 2–12 L/min.

All relevant parameters with their respective values and ranges are given in Table 1.

3.1. Electric Properties

The electrical parameters are crucial for understanding the behavior of the low-current arc discharges. As introduced earlier in this work, we use a three-phase power supply, which can provide a more stable (constant) and higher current but a slightly lower voltage compared to those with the single-phase supply used in our previous studies [7,9]. In Figure 4, the discharge current (a) and the voltage (b) are shown for a flow rate of 10 L/min and an applied power of P_av = 520 W (I_av = 470 mA). The voltage exhibits a periodic sawtooth shape, typical for this type of GAD. The current remains relatively constant during the arc’s movement downstream, and only when it reaches its maximum length, i.e., extinguishment and reignition at a maximum arc voltage of around 2 kV, does it become unstable. At that point, ringing is excited in the circuit, which is typical for circuits with significant inductance, as in our case. The ringing becomes more intensive if the ballast resistance R_b is reduced. The value of R_b is a compromise between the damping of the circuit ringing and the additional unwanted power losses in the resistor. The breakdown voltage for the reignition of a new arc is around 4–5 kV.

To evaluate the effect of the magnetic field on the behavior of the arc for all investigated conditions, a Fast Fourier Transform (FFT) analysis of the voltage signals was performed. An example of a Fourier spectrum of the voltage is shown in Figure 5. Overall, the spectrum contains frequencies with various values, and it is very far from a well-defined single discrete frequency. In all of the spectra for the current and voltage at P_av = 335 and 520 W (I_av = 255 mA and 470 mA), we observe a distinct spectral component at a frequency of 300 Hz, which is related to the power supply’s three-phase full-wave mid-point rectifier. Figure 6 shows the dominant frequencies in the FFT spectra for all of the experimental conditions as a function of the flow rate. To distinguish the effect of the magnetic field on the discharge, we also present the dominant frequency component for a GAD without a magnetic field in the case of 335 W/255 mA (a power supply with the group 2: 3 × SIET Metalbox 15 kV/60 mA). In this case, as expected, the increase in the gas flow leads to an almost proportional increase in the GAD’s frequency, as this corresponds to an increase in the gas velocity. When there is an accelerating magnetic field in the configuration of the MAGD under the same conditions (red dots in Figure 6, 335 W/255 mA), the dependence remains close to linear at higher flow rates above 10 L/min and stays almost constant at low and zero flow rates—around 220 Hz. The difference between the two dependencies, with and without a magnetic field, is approximately 200 Hz. This tendency, towards a higher frequency in the MAGD compared to a GAD, logically follows from the accelerating effect of the J × B force.

The data for the highest power and current of 520 W/470 mA deviate from the expected trend of an increasing frequency with the discharge current due to the greater accelerating J × B force.

One possible explanation for this effect is the appearance of back-breakdown events, which can take place in GADs [32,33,34]. These phenomena are stochastic in nature and correspond to breakdowns between the different parts of a highly bent arc column; i.e., normally, they do not include the dynamics of the cathode spot but only breakdowns between parts of the arc column/the positive column. As a result of these phenomena, the length of the arc and consequently the arc voltage are reduced. Due to these multiple length and voltage reductions, it takes a longer time until the maximum sustained voltage and length are reached, and thus the period of arc transition and reignition increases; i.e., the frequency decreases.

However, this hypothesis is refuted by the experiments for the following reasons. Closer examination of all oscillograms of the voltage for 520 W/470 mA shows indications of back-breakdown events like those in Figure 4b mainly at low gas flows of 2–6 L/min, also present in the spectra as weak higher-frequency components (Figure 5). However, at higher gas flows (8–12 L/min), there are no signs of sharp voltage reductions or back-breakdown events, while the discharge frequency remains lower compared to that under 335 W/255 mA. Moreover, the sharp drops in the voltage observed in Figure 4b could be due to prolonged attachment of the cathode spot. At a high current, the cathode spot is most likely sustained by thermo-field electron emissions, which can cause the spot to remain attached to the same point on the cathode for an extended period [35]. Eventually, a breakdown may occur between the arc column and another point on the cathode, leading to slight shortening of the arc’s length, potentially explaining the behavior observed in Figure 4b. Unfortunately, the current design of the discharge with an active cooling system is not suitable for direct visual observations using a high-speed camera to obtain additional proof of the arc’s behavior.

Another possible reason for the lower frequency in the highest-power case is the significant negative slope of the universal current–voltage characteristics of low-current arcs [36]—an increase in the arc’s current leads to a significant decrease in the reduced electric field and the arc’s voltage. This means that if the PS’s maximum voltage is the same, the length of the arc will be higher at higher currents, and thus it can travel and extend in length for a longer time and thus have a lower frequency.

An additional effect which can decrease the arc velocity is related to the drag force on the arc column due to friction with the surrounding gas. In the case of a magnetically accelerated arc, the arc moves faster than the gas, and it is subject to a friction force. In the case of a laminar flow (estimations show that this is valid for many of the conditions in our experiments), the latter is roughly proportional to the square of the relative gas velocity, to the gas’s density, to its viscosity, and to the size of the object (body). As a result of the increased deposited power at a higher current, we expect both the gas’s viscosity and the approximate arc diameter (i.e., the high-temperature region) to increase. This leads to greater drag, which in turn reduces the relative velocity, as determined by the balance between the drag and the Lorentz force. Typically, gas viscosity is a quantity that depends on the gas temperature. The increased power leads to a higher gas temperature, higher viscosity of the environment, and eventually an increased diameter of the plasma column.

The complexity of the above-mentioned mechanisms and the limited diagnostics do not allow for a clear interpretation of the results and definitive determination of the dominant effect.

The results on the frequency dependence under the lowest power/current studied here (240 W/100 mA) are close to, but lower than, those for higher-current MAGDs. This is likely due primarily to a weaker Lorentz force.

3.2. The CO₂ Conversion Rate and Energy Efficiency in the MAGD

Based on the analysis of the results from our previous investigations of the MAGD [7,9] and additional tests, certain parameters like the magnet size, the gas channel thickness (3 mm), and the electrode distance are fixed in the current study since they provide the optimum conditions.

The dependencies of the conversion rate and energy efficiency on the gas flow rate are shown in Figure 7. The observed tendency of decreasing conversion and an increasing energy efficiency with an increase in the gas flow rate has been confirmed by previous measurements [7,9,37]. For the MAGD, the values determined for the conversion rate are in the range of 6–17%, while for the energy efficiency, they are within the range from 13 to around 60%. The dissociation rate increases with an increase in power and current. For a clear demonstration of the effect of the magnetic field, we present the results for a GAD without a magnetic field for conditions of 335 W/255 mA. At high gas flow rates (10–12 L/min), the results are practically the same—the gas velocity is high enough to diminish the effect of the Lorentz force and to equalize the arc and gas velocities. The relative velocity between the arc and the gas seems to be the main parameter leading to a significant improvement in the CO₂ conversion performance in the setup [9]. At lower gas flow rates (below 10 L/min), the Lorentz force becomes dominant, leading to an increase in the arc–gas relative velocity. This increase significantly improves the dissociation rate.

A more comprehensive presentation of the results is provided when they are shown with respect to the SEI—Figure 8. The dependence on the SEI offers a more condition-independent view of the results and allows for an easier comparison, even between different gas discharges. In Figure 8, we observe that for a MAGD at 335 W/255 mA, both the conversion and the efficiency are higher compared to the results under the other experimental conditions for the same SEI. The relative advantage of the 335 W/255 mA case ranges from 5 to 15%.

Figure 9 presents the combined performance in terms of both the conversion rate and the energy efficiency. Here, we clearly see the advantage of the 335 W/255 mA system, providing combined improvement over the other two cases. The results are also compared with the results (hollow triangles in Figure 9) obtained using the experimental setup used in previous work [9], i.e., with a single-phase power supply. There is between a 10 and 50% relative improvement for a given conversion rate or energy efficiency. This result is worth highlighting because the main difference between the setups is the power supply (voltage and current signals) and the additional optimization of the magnet’s position.

3.3. An Analysis of the Emission Spectra and Derivation of the Vibrational, Rotational, and Gas Temperatures

The experimentally registered C₂ Swan bands under different conditions (average applied power P_av = 335 and 520 W and gas flow rates of 4, 6, 8, 10, and 12 L/min) are analyzed by fitting them with simulated ones using the open source MassiveOES [38] and Specair software. The intensity of the C₂ band at the lowest power and current (240 W/100 mA) was insufficient, and processing of the spectra was unreliable. Therefore, no data for this case is included here.

In Figure 10, a typical C₂ Swan spectrum of the (0,0) band is presented. It is dominated by lines of P- (∆J = J′ − J′′ = −1, where J′ and J′′ are the rotational quantum numbers of the upper and lower states) and R-branches (∆J = + 1), whereas those of the Q-branches (∆J = 0) are negligibly weak.

Under the assumption of a Boltzmann plot distribution for the rotational and vibrational distributions, the fit of the complete ∆v = 0 transition group, derived using the open source MassiveOES, reveals that T_rot and T_vib coincide well (Figure 11, filled symbols) within their uncertainties. According to the experimental data, neither temperature is significantly affected by the gas flow rate or the applied power values. The analysis of the ∆v = −1 transition group, considering the same assumption, affirms these values and the dependencies of T_rot and T_vib on the applied power and gas flow rates (see Figure 11, semi hollow symbols).

The T_rot and T_vib values and dependencies are also checked by fitting the experimental spectra using Specair software. The use of the latter reveals that the plasma is not in equilibrium and the best fit residuals are achieved at an electron temperature of about 1 eV and with values of T_rot and T_vib in the range of those obtained using MassiveOES software.

In the case of competition between production and destruction processes (dissociation and recombination), which often occurs in non-equilibrium plasma, the rotational and vibrational distributions do not always thermalize, and the molecular spectrum can then not be fitted using single rotational and vibrational temperatures. To assess possible deviations from thermal equilibrium of the rotational populations, the state-by-state fitting option in MassiveOES [39] is also used for the ∆v = 0 transition group. The analysis reveals that the rotational levels, providing the main contribution to the emission of the molecular band, appear to be thermalized. As can be seen from Figure 12a, the results for T_rot are in good agreement whether using Boltzmann fitting (BF) or state-by-state fitting (SSF).

According to the literature [40], to study so-called C₂ “high-pressure bands”, it is appropriate to fit the ∆v = + 1 transition group, as it is the most sensitive with respect to determination of the vibrational state densities and therefore to the vibrational temperature. The analysis reveals (Figure 12b) that for the present plasma conditions, there is no evidence of deviations in the thermal equilibrium of the vibrational populations. We attribute the higher uncertainties in the value of T_vib obtained using state-by-state fitting of the ∆v = + 1 transition group with respect to that obtained through Boltzmann-fitting ∆v = 0, −1 to the lower signal-to-noise ratio (see Figure 3).

Thermalization of the vibrational states of the C₂ (d³∏_g) level indicates that the dominant dissociation pathway under the investigated conditions is thermal dissociation. According to the literature, this could be expected at high pressures for CO₂ GADs [6] and MW plasmas [40].

Under our experimental conditions, it seems like there is an equivalence between the rotational and vibrational temperatures. In addition, during the optimization of the MAGD with respect to its configuration, we conducted experiments where a small amount of nitrogen (0.05%) was added. This resulted in a strong decrease in the Swan band’s emission intensity (expected according to the literature [40,41]), while the CN violet system increased drastically from barely being detectable. The admixture allowed us to fit the ∆v = 0 transition group of the CN (B²∑⁺) spectrum using the two independent software Specair 3.0 and Lifbase 2.1.1 and to determine the T_rot and T_vib values. The results (5500–6000 K) reveal that under the investigated conditions, there are no significant differences in both temperatures and between the temperatures determined from the CN and Swan molecular systems. To the best of our knowledge, contrary to the C₂ (d³∏_g) state, the CN (B²∑⁺) state does not have known mechanisms leading to non-equilibrium rovibrational emission spectra [42]. The shorter rotational relaxation times of this state than its radiative lifetime [43,44] provide the possibility of thermalization of the rotational and vibrational states before observing their emissions. The cross-validation between the rotational temperatures obtained from the C₂ Swan and CN molecular systems gives additional confidence that they can be considered a local measure of the translational temperature of the plasma, i.e., the gas temperature. Since the nitrogen admixture had a slight but negative influence on the energy efficiency and the conversion rate, systematic investigation of the MAGD’s characteristics was carried out in pure CO₂, as reported in this study.

3.4. A Comparison with Other Experiments from the Literature

In this subsection, we present a comparative analysis of our device and its modifications in comparison with those from other recent studies (from within the last five years) investigating CO₂ conversion using similar configurations or discharge types. Specifically, we focus on glow- and arc-discharge-based systems. The findings from these studies are summarized and visualized in an energy efficiency versus conversion diagram for direct comparison in Figure 13.

Trenchev et al. [45] explored a dual-vortex gliding arc plasmatron (GAP) operating at atmospheric pressure, which sustained a rotating gliding arc discharge. Their study reported discharge currents ranging from 150 and 400 mA and gas flow rates between 5 and 12.5 L/min. The maximum CO₂ conversion rate achieved was 9.5%, with an associated high energy efficiency of up to 41%. The GAP demonstrates relatively high energy efficiency under substantial conversion rates, attributed to the arc rotation mechanism and the increased volume of treated gas. GAP types of gliding discharges are typically considered the most efficient type of GADs. It is worth noting that here, we achieve a similar or even a slightly higher performance in terms of the energy efficiency and conversion rates, but still, the GAP’s results are achieved at higher gas flow rates. Increasing the gas flow in our setup further would reduce the conversion rate and diminish the benefit of the magnetic field. In the next section, possible paths for scaling the discharge operation are discussed.

Renninger et al. [46] utilized a glow discharge system at atmospheric pressure to achieve efficient CO₂ dissociation. Their experiments were conducted at gas flow rates of 0.65 to 1.25 standard L/min (SLM) and using input powers between 100 and 200 W. The highest conversion rate was observed at a discharge power of 165 W with a gas flow rate of 0.65 SLM, while a peak energy efficiency of 32% was achieved at a higher flow rate of 1.25 SLM. While the conversion rate achieved in these experiments is impressive, the typical gas flow remains very low.

In a follow-up study, Renninger et.al. [47] investigated an optimized reactor operating in the glow-to-arc transition regime at atmospheric pressure. The system was tested under similar flow rates (0.65 to 1.25 SLM) and higher input discharge powers (150–300 W).

Li et al. [37] examined classic gliding arc discharges using converged electrodes under atmospheric conditions, evaluating the performance both with and without an applied external magnetic field. The discharge currents ranged from 210 to 240 mA, with gas flow rates between 1 and 8 L/min.

Li et al.’s findings [37] clearly demonstrate that the presence of a magnetic field enhances the system performance by improving both the conversion rate and the energy efficiency. This improvement is attributed to the elongation of the arc’s critical length and the expansion of the plasma area, which allows a greater amount of gas to be treated. Overall, our discharge system achieved higher conversion rates and energy efficiencies with a lower SEI under the same discharge current conditions, which was probably due to a better design and condition optimization.

Vertongen et al. [48] investigated the effect of gas quenching in an arc reactor. The graph above presents the benchmark results and their best results using the so-called “heat exchanger”—a system consisting of several channels enclosed in a tube that is actively cooled. The flow rates investigated were 10, 15, and 20 L/min. The discharge current varied from 0.7 to 1 A, and the discharge power ranged from 300 to 1500 W.

The addition of a gas quenching system significantly improved the conversion rate and the energy efficiency of the process, increasing the conversion rate from about 5% to approximately 10–15% and the energy efficiency from around 20% to 30%. In our system, gas quenching is achieved through active cooling of the quartz glasses. In our case, the combined effect of quenching and magnetic acceleration leads to similar conversion rates and efficiencies to those in Vertongen et al.

4. Conclusions

The current study presents further optimization and characterization of a magnetically enhanced classical gliding arc discharge. It provides results which are a significant improvement over those in our previous work and are on par with the best results available in the literature for similar types of discharges. The discharge’s simplicity and the use of abundant materials provide an excellent basis for real-life applications. Moreover, the use of active cooling provides very good stability and long-time operation.

The energy efficiency of CO₂ conversion reaches relatively high values of around 60% at a 7% conversion rate, while the maximum conversion rate is around 16% at a 20% energy efficiency. The latter can probably be improved through additional design optimization and an increase in the power and gas flow. It is important to note that this and other studies have shown that the discharge is rather sensitive to the power supply characteristics, and further improvements could be made with a deeper understanding of these effects.

Under experimental conditions of 335/255 mA and 520 W/470 mA and gas flow rates of 4, 6, 8, 10, and 12 L/min, the rotational and vibrational temperatures coincide well within their uncertainties. Due to the fast thermalization of the rotational states of the C₂ (d³∏_g) state with the background gas, the values of the rotational temperature determined can be considered as local measures of the gas temperature. The rotational (gas) temperature values obtained are in the range of 5500 K–6000 K.

From an industrial point of view, normally, one would expect to process higher gas flows and apply higher powers compared to the values considered in this study. The higher gas flow in the current discharge design in terms of the dimensions leads to a reduced dissociation rate, as we see from the results obtained. The discharge, gas flow, and power scaling can be examined in the future according to two directions where the considerable relative velocity between the arc and the gas is sustained: (1) a higher gas flow (gas velocity) combined with a higher current in a domain similar to that in the current design or (2) a discharge current with an order of magnitude similar to that in the current study but a higher gas flow in a larger domain, so that the gas velocity is similar to the value considered here, combined with a higher voltage and a longer arc length. The latter will require a modified design of the electrodes and the whole setup.