DBD in burst mode: solution for more efficient CO₂ conversion?

The CO₂ gas flows through a cylindrical DBD reactor as shown in figure 1. The CO₂ flow rate is set at 200 ml_n · min⁻¹ and the operating frequency is fixed at 28.6 kHz. The central cylindrical copper electrode has a diameter of 22 mm and a length of 120 mm. This inner electrode is powered by an AC high voltage, whereas the outer electrode is grounded. The latter is a mesh, made of stainless steel, with length of 100 mm (thus defining the discharge length), and placed around a dielectric barrier (2 mm in thickness), made of alumina and with inner diameter of 26 mm, yielding a discharge gap of 2 mm, and thus a discharge volume of 15.1 cm³. This corresponds to a residence time of the gas equal to 4.5 s for the given gas flow rate of 200 ml_n · min⁻¹. The applied power is provided by an AFS generator G10S-V and a transformer, which amplify the sinusoidal signal of the alternating voltage and which allow us to pulse the voltage in a controlled manner.

Zoom In Zoom Out Reset image size

In the present study, the signal frequency remains fixed at f_signal  =  28.6 kHz (period of 35.7 µs). Furthermore, T_ON and T_OFF are defined as the time of plasma ignition and the time when no power is applied to the reactor. Except for the analysis of T_ON and T_OFF at constant duty cycle, the main results described in this article are related to the influence of the duty cycle. In this latter case, T_ON remains fixed at 1 ms while T_OFF is changed between 0 and 1.5 ms, corresponding to a duty cycle (D_cycle) between 100 and 40%. Therefore, the repetition frequency of the burst mode, f_repetition  =  1/(T_ON  +  T_OFF), is varied between 400 and 900 Hz. The voltage as a function of time is plotted in figure 2 for high voltages delivered either in a pure AC mode (a) or in a burst mode (b). As the total power applied to the plasma (P_applied) is fixed at 50 W in all cases, the power during the plasma ignition (P_plasma) in burst mode is always higher than in pure AC mode, since P_plasma (W)  =  P_applied (W)/D_cycle (%).

Zoom In Zoom Out Reset image size

The gaseous products from the post-discharge are analyzed by a mass spectrometer operating at atmospheric pressure (Hiden analytical QGA). The electron energy in the ionization chamber is set at 70 eV and a secondary electron multiplier (SEM) is used as a detector. The CO₂ conversion (${{\chi}_{\text{C}{{\text{O}}_{2}}}}$) is calculated according to the following equation, where I stands for the intensity of the signals:

$$\begin{eqnarray}&&{{\chi}_{\text{C}{{\text{O}}_{2}}}}\left(\%\right)=\frac{{{I}_{\text{C}{{\text{O}}_{\text{2}}}~\text{without} ~\text{plasma}}}-~{{I}_{\text{C}{{\text{O}}_{\text{2}}}~\text{with} ~\text{plasma}}}}{{{I}_{\text{C}{{\text{O}}_{\text{2}}}~\text{without} ~\text{plasma}}}}~\times 100\%\end{eqnarray}$$

The energy efficiency of the CO₂ conversion is calculated from ${{\chi}_{\text{C}{{\text{O}}_{2}}}}$, the enthalpy of the splitting reaction (CO₂  →  CO  +  ½ O₂), namely $\Delta H_{298\text{K}}^{0}$  =  279.8 kJ · mol⁻¹  =  2.9 eV · mol⁻¹ and the specific energy input (SEI), according to the following equation:

$$\begin{eqnarray}&&{{\eta}_{\text{C}{{\text{O}}_{2}}}}\left(\%\right)={{\chi}_{\text{C}{{\text{O}}_{2}}}}\left(\%\right)\cdot \frac{\Delta H_{298\text{K}}^{0}\left(\text{eV}\cdot \text{mo}{{\text{l}}^{-1}}\right)~}{\text{SEI}~\left(\text{eV}\cdot \text{mo}{{\text{l}}^{-1}}\right)}\end{eqnarray}$$

The SEI is defined as follows:

$$\begin{eqnarray}&&\begin{array}{*{20}{l}}\text{SEI}~\left(\text{eV}\cdot \text{mo}{{\text{l}}^{-1}}\right)\\ \quad =\frac{\frac{{{P}_{\text{applied}}}~\left(\text{J}\cdot {{\text{s}}^{-1}}\right)}{\text{Gas} ~\text{flow} ~\text{rate}~\left(\text{c}{{\text{m}}^{3}}\cdot {{\text{s}}^{-1}}\right)}\times 6.24\times {{10}^{18}}\left(\text{eV}\cdot {{\text{J}}^{-1}}\right)\times 24\,500~\left(\text{c}{{\text{m}}^{3}}\cdot \text{mo}{{\text{l}}^{-1}}\right)}{6.022\times {{10}^{23}}\left(\text{molecule}\cdot \text{mo}{{\text{l}}^{-1}}\right)}\end{array}\end{eqnarray}$$

In principle, GC analysis is more accurate for determining the CO₂ conversion, but it was not available for this study. We believe that MS analysis is also sufficiently accurate at the current conditions.

A Tektronix DPO 3032 oscilloscope is used to perform electrical measurements (during T_ON). The first channel is dedicated to the monitoring of the high voltage (V₁) with a Tektronix P6015A probe, whilst the second channel allows the measuring of the voltage passing through a capacitor, to obtain the power absorbed by the plasma via the Lissajous method [54–58], or through a Rogowski coil (Pearson 2877) to probe the current. The capacitor or the Rogowski coil is placed in series with the DBD. The obtained current profiles as a function of time, also called oscillograms, correspond to 2 different types of current: the plasma current and the dielectric current. These two can be simply distinguished as the dielectric component is a sinusoidal-like signal, whereas the plasma component corresponds to hundreds of microdischarges occurring at each half period, since an atmospheric pressure DBD operating in CO₂ generally operates in a filamentary regime [24, 59–61]. Analysis of these microdischarges is important to better understand the CO₂ splitting process. A statistical study will be carried out for several periods in order to evaluate the magnitude of different electrical parameters of these microdischarges. By applying a numerical method described in our previous article [61], the number of microdischarges (N_md), their lifetime (L_md), the plasma charge accumulation (Q_pl) and the resulting plasma current (i_pl) can be achieved. These data are collected for several periods and then averaged over a single period in order to have statistically valid results. In particular, N_md has been measured during T_ON  =  1 ms and multiplied by D_cycle in order to consider the same residence time. Likewise, the applied voltage (V_DBD) can also be expressed as the sum of two voltages: the dielectric voltage (V_diel) and the plasma voltage (V_pl), which typically represents 80% of the total V_DBD [61, 62].

The OES measurements are performed with an Andor Shamrock-500i spectrometer including an Andor DU420A-OE CCD camera. Each spectrum is acquired for an exposure time of 5 s and 7 accumulations. Of particular interest is the CO emission band headed at 483.3 nm from the CO Angstrom system (B¹ Σ⁺  →  A¹ Π; ν'  =  0  →  ν''  =  1) shown in figure 3. Its ro-vibrational structure is calculated and fitted to the experimental spectra in order to determine the rotational temperature (T_rot) and hence the gas temperature (T_gas), assuming T_rot  ≈  T_gas [63–65]. For the sake of comparison, the gas temperature is also measured using another band: the first positive system (FPS) of N₂ is analyzed by adding 10% of nitrogen to CO₂. In this case, the temperature is measured via a line-ratio peak formula, specific to this system [66, 67], using the peaks at 775.3 and at 773.9 nm.

Zoom In Zoom Out Reset image size

The CO₂ conversion and energy efficiency as a function of the duty cycle of the burst mode are depicted in figure 4(a). The conversion obtained in the pure AC mode (D_cycle  =  100%) is 16.1%, which is significantly lower than in the burst mode at 40% duty cycle (i.e. 25.8%). Decreasing the duty cycle at constant applied power thus results in a higher CO₂ splitting. As evidenced in figure 4(b), smaller duty cycles yield higher peak powers, e.g. 100 W for D_cycle  =  50% versus 50 W for D_cycle  =  100%. These higher peak powers delivered within shorter times can explain the higher CO₂ conversion. However, we will also show at the end of the paper that the enhancement of the burst mode is not only attributed to the higher peak power, but also to other effects (see below). The corresponding energy efficiency also increases upon decreasing duty cycle from 14.5% (at D_cycle  =  100%) to 23.1% (at D_cycle  =  40%).

Zoom In Zoom Out Reset image size

On the other hand, keeping the duty cycle constant but increasing the burst width from 1 ms to even more than 20 ms has no effect on the CO₂ conversion and energy efficiency, as illustrated in figure 5. Therefore, in the next sections we will only investigate the influence of T_OFF (or duty cycle or repetition frequency) on the electrical properties of the discharge, and more specifically on the microdischarge features. First we will show the effect of the duty cycle on the evolution of the applied voltage with increasing process time, to check the time-stability of the discharge. Next, the influence of the duty cycle on the current oscillograms and on the microdischarge properties, as well as on the gas temperature will be studied. Finally, the burst mode and the normal AC mode will be compared at the same delivered power, to illustrate that the enhancement of the burst mode is not only attributed to the higher peak power, but also due to the microdischarge properties.

Zoom In Zoom Out Reset image size

Figure 6 shows the applied voltage as a function of the process time for several different duty cycles. The lower the duty cycle, the more stable and the higher the voltage is as a function of the process time. For D_cycle  <  70%, the voltage remains quite stable over time, e.g. at D_cycle  =  50% it is close to 4400 V for all process times investigated (with only a small variation within the first 60 s). Therefore, the discharge appears as very stable in time. However, this time-stability gradually vanishes upon increasing D_cycle and in the pure AC mode (D_cycle  =  100%) the time dependency is quite significant: a drop of approximately 500 V is reached in 90 s, from 4400 V to a plateau at 3900 V. This transient regime is clearly visible in the pure AC mode and is due to the dielectric properties of the barrier: it behaves as a capacitor with a dielectric relaxation time that is much longer than the period associated with the applied voltage. The dielectric relaxation time of the barrier t_relax is defined as the product of its dielectric permittivity (ε  =  ε₀ · ε_r) with its electrical resistivity (ρ), i.e. t_relax  =  ε₀ · ε_r.ρ, where ε₀  =  8.854 · 10⁻¹² F · m⁻¹ and, in the case of alumina, ε_r  =  9 and ρ  ≈  10¹⁴ Ω⁻¹. cm⁻¹ [68]. A rough estimation leads to t_relax  ≈  0.8 s. This value is much higher than the period of the applied voltage, which is 35.7 µs (since f_signal  =  28.6 kHz). In the first half period, the electrical charge deposited on the barrier cannot be entirely evacuated and remains partially in the second half period. This charge accumulation is negligible at the very beginning of the discharge ignition, but becomes significant in less than 30 s of operation (see figure 6 for D_cycle  =  100%). Once the DBD is ignited, the barrier potential gradually increases, hence inducing a gradual decrease in V_DBD observed in figure 6. This phenomenon is clearly less pronounced in the burst mode, where the T_ON times are too short to induce significant charge accumulations on the barrier. Moreover, the T_OFF times are much longer so that the sustaining voltage can remain closer to the ignition voltage [69]. The applied voltage (V_DBD) is expressed as the sum of two voltages, V_DBD  =  V_pl,eff  +  V_diel, where V_pl,eff is the effective plasma voltage (in the gap) and V_diel is the voltage through the dielectric barrier. In a previous work [61], we have demonstrated that varying V_DBD for a fixed power induces variations of V_pl,eff while V_diel remains almost unchanged. Here, decreasing the duty cycle at P_applied  =  50 W only induces an increase in V_DBD, hence of V_pl,eff. As the gap is fixed, we can thus conclude that a stronger electric field is created as the plasma voltage is higher, and therefore there will be more energetic electrons. As a result, the burst mode can enhance the rates of electron impact excitation, ionization and dissociation of CO₂ and thus improve the conversion and energy efficiency of the CO₂ splitting process, as illustrated in figure 4(a). Hence, besides the higher peak power (due to the shorter duty cycle), this can also explain the higher CO₂ conversion in the burst mode.

Zoom In Zoom Out Reset image size

The influence of the duty cycle on the gas temperature is obtained from OES measurements. Ro-vibrational spectra of the Angstrom CO band are investigated, as well as the first positive system (FPS) of N₂. In figure 7, a linear decrease in gas temperature is observed when reducing the duty cycle, i.e. when increasing the T_OFF times. This is logical because in the burst mode, the gas can cool down. In the pure AC mode, where the CO₂ conversion is the lowest (16.1%), the gas temperature is the most elevated, suggesting that in general, a larger fraction of the power is used for gas heating, while a smaller fraction of the power leads effectively to CO₂ splitting. It should be noted that the gas temperature as obtained from the FPS of N₂ is about 100 K higher than the gas temperature obtained from the Angstrom CO band, over the entire range of duty cycles investigated. At this point, it is rather difficult to explain this difference since no other independent method (e.g. Doppler broadening, thermocouple, etc.) was used for the gas temperature determination in the plasma phase. Besides this, it is also worth mentioning that the line-ratio and synthetic spectra calculation methods are prone to errors as they assume a Boltzmann distribution and the validity of this assumption cannot be directly checked. Nevertheless, despite the difference of about 100 K, the trend of decreasing gas temperature with decreasing duty cycle is obvious from both curves, as shown in figure 7.

Zoom In Zoom Out Reset image size

Figure 8(a) shows a picture of the filaments in the DBD for two different duty cycles (50 and 100%). For this purpose, the alumina dielectric barrier is replaced by (transparent) borosilicate glass in order to visually observe these filaments. The pictures reveal an apparent larger number of filaments at D_cycle  =  50% than in pure AC mode (D_cycle  =  100%). To study this in more detail, these pictures are correlated with their corresponding current oscillograms in figure 8(b), illustrating the currents peaks (I_DBD  =  I_pl  +  I_diel) during one AC period, i.e. 35.7 µs. Note that the oscillogram corresponding to D_cycle  =  50% is a current profile related to P_plasma  =  100 W, as these oscillograms are recorded only during T_ON. A visual observation of these oscillograms indicates that the number of microdischarge pulses, as well as their lifetime and the current amplitude all increase from D_cycle  =  100% to D_cycle  =  50%. Based on a numerical method explained in [61], we can also obtain more quantitative information on these microdischarges, such as their number (N_md), lifetime (L_md), charge (Q_pl) and conduction current (i_pl), from these current peaks.

Zoom In Zoom Out Reset image size

These parameters are plotted as a function of duty cycle in figure 9, as obtained for one AC period of 35.7 µs. These results refer to the same applied power, and thus the peak power is higher in the burst mode. All these parameters are given in terms of (i) the same residence time (by taking into account T_OFF and thus, the electrical parameters are multiplied by D_cycle) and (ii) only the plasma ignition (without taking into account T_OFF). As observed in the figures, all of them are clearly changed in the burst mode and upon decreasing duty cycle. For instance in terms of same residence time, a reduction of D_cycle from 100 to 50% induces a decrease in the N_md from 390 to 340 per period. On the other hand, their mean lifetime (L_md) increases from 12.9–14.8 ns during the plasma ignition. A drastic rise of the plasma current i_pl (from 9–14 mA) is also observed when the duty cycle drops from 100 to 50%, which will result in a significant increase in electron density. The latter can be deduced from the rise in plasma charge Q_pl in terms of a given residence time (from 0.3–0.5 µC). As the oscillograms are recorded during T_ON, and the peak power becomes higher in the burst mode, the values of N_md, L_md, Q_pl and i_pl will increase upon increasing plasma power P_plasma (see our previous study on the effect of power on the electrical characteristics [61]). Therefore in the last paragraph, we will also compare the electrical characteristics in burst mode and pure AC mode, at the same delivered power during T_ON.

Zoom In Zoom Out Reset image size

Figure 10 depicts the number of microdischarges, their mean lifetime and the plasma charge and current for the pure AC (values adopted from our previous paper [61]) and burst mode, as a function of plasma power. Several observations are noteworthy:

• (i)  
  In pure AC mode (i.e. all corresponding to D_cycle  =  100%), the plasma power corresponds to the applied power (P_plasma  =  P_applied). Thus, the plasma power is increased by tuning the applied power.
• (ii)  
  The plasma power in burst mode corresponds to P_plasma  =  P_applied/D_cycle where P_applied is fixed (50 W). The plasma power (delivered during T_ON) is increased by tuning the duty cycle. For example, P_plasma  =  50 W at D_cycle  =  100% and ca. 100 W at D_cycle  =  50%. The corresponding duty cycles are indicated next to the black points in the figure.
• (iii)  
  All the aforementioned results in the article can only be compared to the black curves in figure 10, i.e. for P_applied  =  50 W.

Zoom In Zoom Out Reset image size

In figure 10(a), the number of microdischarges (N_md) is determined for T_ON  =  1 ms and multiplied by the duty cycle to always consider the same residence. Whatever the plasma power, N_md is higher in pure AC mode than in burst mode. For instance, at P_plasma  =  100 W, the burst mode (for 50% duty cycle) allows the formation of 340 microdischarges per period at P_applied  =  50 W, while 480 microdischarges are obtained at P_applied  =  100 W in the pure AC mode. The burst mode presents a lower number of microdischarges, but the applied power is twice as low, which is thus more interesting in terms of energy efficiency. For the same plasma power, a lower N_md means a higher plasma voltage (V_pl,eff), hence a higher electric field and thus higher electron temperatures. The filaments are generated in a higher electric field, leading to a higher efficiency of the CO₂ conversion.

Figures 10(b) and (c) give information on the microdischarge lifetime as well as the electrical charge and current versus plasma power. These electrical parameters obviously increase with plasma power and the differences between pure AC mode and burst mode are noteworthy: Q_pl is indeed lowered in the burst mode. In terms of plasma ignition, the two modes result in a quite similar plasma charge accumulation and plasma current, when operating at the same delivered power, as observed in figure 10(c). As the plasma charge is directly related to the plasma current, the electron densities might be very similar in both modes during T_ON.