First, a bench scale single bed PSA system was used to extract CO₂ from a simulated flue gas stream containing 12.5% CO₂ and 87.5% N₂. This composition sits well within the range expected as an output from coal fired power stations, for example, found in the UK (Naims, 2016; von der Assen et al., 2016). This PSA system makes use of a cheap and robust solid sorbent, which is based on previously reported poly-ionic liquid sorbents by this group (Supasitmongkol and Styring, 2010; Dowson et al., 2016), which selectively adsorbs CO₂ on pressurization of the gas stream. As the pressure is released, the N₂ and CO₂ desorb from the sorbent surface at different rates, thus creating two separate streams; one rich in N₂ followed by one rich in CO₂. An experimental diagram of the process is shown schematically in Figure 2. High-pressure adsorption experiments were carried out in a bespoke stainless steel packed-bed adsorber column constructed using Swagelok™ 1/2″ tubing 100 mm in length. A cross section of which is shown in Figure 3. The column is compression sealed and can withstand pressures up to 200 bar. The sorbent is packed inside with quartz wool filters used to avoid sorbent carry over during testing.

Flue gas streams were pressurized to 15 bar and allowed to adsorb for 15-min cycles. Full methodology was as described previously for solid ionic liquid sorbents (Dowson et al., 2016). A two-stage depressurization of the adsorber leads to two gas streams. This CO₂-rich stream would be sent directly to the corona reactor for reduction to CO.

A simplified experimental diagram of the downstream plasma utilization section is presented in Figure 4. Experiments were performed at atmospheric pressure and ambient temperature. The reactor used was a wire-to-cylinder corona discharge reactor constructed using a stainless steel cylinder as the outer ground electrode (R = 17 mm), and a narrow tungsten wire as the inner live electrode (r₀ = 125 µm). The total reactor volume was approximately 300 cm³.

Plasma ignition was provided by means of a high voltage pulse generator (NPG18-3500N Megaimpulse Ltd.) (Lyublinsky et al., 2013). Electrical measurements were taken using a high voltage probe (Tektronix P6015A) connected to the live electrode to measure the voltage and a wide band current transformer (Pearson 6595) to measure the current. These two parameters enabled the energy and power used during the utilization process to be calculated. The unique aspect of this power supply is the ability to send a series of short pulses approximately 40 ns in width over a microsecond at a repetition rate that can be varied up to 3,200 Hz. These short pulses remove the need for a dielectric layer, prevent gas heating (thus increasing the efficiency of CO₂ splitting), and reduce the possibility of plasma arc development. The applied voltage signal is shown in Figure 5. It is characterized by a primary short pulse of very high amplitude proceeded by a series of smaller short pulses of lower amplitude. The corresponding output current is also shown in Figure 6.

Following capture from the PSA rig, the CO₂-rich flue gas was passed to the pulsed plasma reactor wherein it was subjected a corona discharge for varying residence times. The residence time was altered by changing the gas flow rate (Q) into the plasma reactor between 100 and 400 mL/min by using mass flow meters (Bronkhorst) which corresponds to residence times of 37.5, 75, and 150 s. Gaseous products from the corona reactor were measured by means of a Fourier transform infrared (FTIR) spectrometer (Varian 660-IR Agilent) post discharge. No liquid products were observed. A typical absorbance spectrum is shown in Figure 7. The method used to quantify the production of CO was performed by calibrating the FTIR with known quantities of CO to generate a calibration curve. Before plasma ignition a background spectrum of the CO₂ or CO₂/N₂ mixture was taken. After being exposed to the plasma, another spectrum was taken, and the background spectrum subtracted to determine the absolute IR absorbance. This absorbance was then compared to a known value along the calibration curve, and the quantity of CO was found. The main products detected were CO, NO, NO₂, N₂O, N₂O₅, and N₂O₃ with CO the dominant product. As the most important species as far as utilization is concerned, only the conversion of CO₂ to CO was calculated. In addition, the values of other products remained in the parts per million range so were viewed as minor importance. Consideration into separation of all the product gases requires further work and is outside the scope of the article. This is in contrast to the works of Heijkers et al. (2015) who observed effective conversions up to 15% of N₂ to other products, although in that work the specific energy input (SEI) peaked at 7.1 eV/molecule compared to the maximum of 0.43 eV/molecule used in this work.

To evaluate the effectiveness of CO₂ utilization, two important and commonly used parameters will be deployed. These are the conversion factor (X) and energy efficiency (η) as defined by Eqs 1 and 2, respectively. The conversion is defined as the change in CO₂ concentration over the initial CO₂ concentration. As the production of other carbon containing species is negligible, this can be said to approximate the CO yield. From this, effective conversion (X_eff), which is also used in literature, can be found as a product of X and the relative CO₂ content in the feed. For example, if a gas contains 80% CO₂, the conversion is multiplied by 0.8. (1)Xabs(%)=[CO2]in−[CO2]out[CO2]in×100≈[CO]out[CO2]in×100

The energy efficiency is defined as a ratio between the dissociation enthalpy of CO₂ (ΔH_R) as shown in Eq. 3 to the energy penalty to produce one molecule of CO (E_CO). (3)CO2 → CO + 12O2, ΔHR=2.9 eV/molecule = 283 kJ/mol

Equation 2 is usually expressed in the literature in terms of the SEI, or the ratio of discharge power to the gas flow rate (SEI = W/Q) (J/cm³) (Eq. 4). Discharge power is defined as the total energy supplied over the residence time (W = E/t_g) (J/s). For calculation of total energy, see Eq. 6.

To calculate E_CO (J), the total energy (E) must be calculated during the residence time of the gas as a ratio to the number of CO molecules that have been subjected to corona discharge (N_CO), (4)ECO=E/NCO

where N_CO can be expressed as follows: (5)NCO=Xeff[CO2]inQ tg.

The total energy deposited is calculated as shown in Eq. 6. It is the product of the energy applied during a single pulse (E_pulse) (J), the gas residence time (t_g) and the pulse repetition frequency (f) (Hz). E_pulse can be further calculated from the integral of the applied voltage and current over the duration of a pulse (τ). (6)E=ftgEpulse=ftg∫0τV(t)I(t)dt

From previous work (Moss et al., 2017), it is known that the shape of the applied voltage waveform heavily influences the energy deposited within the plasma reactor and it can affect the chemical kinetic pathway of CO₂ utilization. Under all experimental conditions, the repetition frequency was fixed at 1,700 Hz, as higher values led to instabilities in the discharge, increasing the temperature of the gas and thus reducing the energy efficiency of utilization. In this work, the energy deposited in the gas was altered by modifying the residence time of the gas only and estimated to be between 1 and 2 mJ across all gas mixtures.

To generate the product streams for the corona reactor, the PSA system uses a poly-ionic liquid solid sorbent, which allows separation of CO₂ from a simulated flue gas stream due to the difference in desorption rate of CO₂ and N₂ from the sorbent. N₂ is desorbed instantly, whereas CO₂ is held for longer periods of time giving a clear separation of the two gas streams. Although the exact mechanism of this desorption behavior is not fully understood, the physical nature of the adsorption in this case leads to weak physical interactions with the CO₂ molecule. Although there is some interaction with N₂, it is almost negligible and decreasing the system pressure fully reverses it, allowing almost instantaneous release of N₂ from the system. It is likely that CO₂ desorption rate is a function of the system pressure, and as the pressure drops, the weak interactions formed between the molecules and the surface of the sorbent are continually broken as the pressure is reduced with the most weakly bound molecules leaving the system first. Desorption rate was determined for each gas individually by first allowing gas adsorption for 15 min at 15 bar. Previous work (Dowson et al., 2016) has shown that after this 15-min period, the sorbent will be fully saturated at 100% and will adsorb no more as shown in Figure 9. In fact, adsorption rates for some of the ionic liquid sorbents were found to be mere seconds. The pressure was then released from the adsorber and the sorbent weight was tracked until a steady state was reached and desorption was completed.

As can be seen from Figure 10 during certain points of the desorption step, both gasses will be released at the same time therefore, after one pass, it is clear that a 100% stream of both N₂ and CO₂ cannot be achieved. Further passes would be required to gradually increase the CO₂ concentration to higher percentages. It is also thought that the instantaneous desorbing N₂ will in fact pull some of the weakly bound CO₂ with it as it flows through the packed bed. To determine the CO₂ composition of the gas in the outlet, a mixed gas stream of 12.5% CO₂ and 87.5% N₂ was produced using Bronkhorst mass flow controllers. This gas mixture was then pressurized using a Midwest Pressure Systems 40 series air driven gas booster to 15 bar and left to adsorb for 15 min with the reactor held at a steady temperature of 25°C. The pressure within the adsorber was then released gradually using a JASCO MODEL BP-1580-81 programmable electronic back pressure regulator with a gas flow rate of approximately 50 mL/min and allowed to flow through a COZIR Sprint IR inline infrared sensor to determine the CO₂ composition.

These initial tests show that there are two distinct stages within the pressure drop, as the pressure drops from 15 down to approximately 5 bar, the CO₂ concentration is low and N₂ desorption dominates, this stage would is vented as a “waste” gas. At 5 bar, there is a sharp gradient shift as CO₂ starts to desorb, this stage is product gas. The peak concentration of the product gas is around 55% CO₂ with an average composition of around 44% CO₂ in the total volume of gas released. This product gas can then be sent to the corona reactor directly, or can be recycled for a second pass of the adsorber. Figure 11 shows measured gas compositions with ranging CO₂ inputs and outputs to show the flexibility of the system. To generate these data, the pressure was dropped instantaneously using the back pressure regulator to mimic the two stages previously mentioned allowing for a combined adsorption/desorption cycle time of around 5 min to be achieved. The output results shown are the average composition from stage 2 of the pressure drop.

Higher concentrations of product gas were obtained by doing multiple passes of the packed-bed adsorber. It is shown that with some optimization, a product output over 90% CO₂ could be achieved with just two passes of the adsorber if the utilization system required it.

Figure 12 shows the absolute and effective CO₂ conversion produced during pulsed corona discharge against the gas residence time for various gas mixtures of CO₂ and N₂. It is evident that for all gas mixtures increasing, the residence time has a positive effect on the conversion. Initial observation suggests that there is no immediate correlation between gas mixture and conversion; however, it can be seen that a small quantity of N₂ enhances CO₂ conversion, but further increase leads to a reduction in conversion. Examining effective conversion shows raw flue gas (12.5% CO₂) and 100% CO₂ perform the worst across all gas mixtures, which is highly beneficial for the proposed process as it proves raw flue gas to be unsuitable for utilization and that pure CO₂ is not required. Ideally, an upgraded flue gas feed is most suitable in terms of conversion for this non-thermal plasma system, which is exactly what the PSA system used in this work provides.

Peak effective conversions of approximately 8% were found in an 80% CO₂/20% N₂ mixture. Snoeckx et al. (2016) observed a similar effect using a dielectric barrier discharge (DBD) plasma reactor although the reported conversions were slightly higher, peaking at 12% for a 25% CO₂/75% N₂ mixture. However, it should be noted that the conversions reported in that work were achieved at significantly higher SEIs approximately one order of magnitude higher than this work. Transfer of energy from vibrationally excited N₂ has been proven to be effective in enhancing the conversion of CO₂ in DBD systems and it is theorized that the same effects are exhibited in a pulsed corona, although the extent of this requires further investigation. Pontiga et al. (2011) also investigated mixtures of CO₂ and N₂ in a DC corona discharge. However, in that work, the conversion to CO was only in the order of a few percent. They attributed the lower conversion upon increasing N₂ concentration to the formation of nitrous oxides (NO_x), which inhibit the formation of CO through reactions such as Eq. 7, which could prove a limiting factor. (7)NO3+CO→NO2+CO2

Heijkers et al. (2015) observed a large conversion of N₂, in addition to CO₂, in a microwave plasma. N₂ conversions up to 15%, at peak SEI of 7.1 eV/molecule, was observed, which is significantly higher than those reported in this work where the total NO_x products formed was in the order of hundreds of parts per million. This observation can be attributed to the lower specific input energy used during pulsed corona discharge operation which is typically one order of magnitude lower than other plasma technologies. As less total energy is applied the amount available to convert N₂ is also less; this combined with a higher energy threshold to ionize N₂ over CO₂ results in low NO_x production.

Figure 13 shows the energy efficiency of CO₂ conversion under plasma conditions as a function of the SEI. For all mixtures, except 60% CO₂/40% N₂, the efficiency decreases as the SEI is raised, i.e., the residence time is increased. As demonstrated in previous work (Moss et al., 2017), the efficiency is substantially higher than other reported works at low SEIs, and this remains the case in CO₂/N₂ mixtures, peaking at approximately 90% albeit with a large error. Again compared to the works of Snoeckx et al. (2016), the highest reported efficiency was 13% in a 75% CO₂/25% N₂ mixture. It is noted that the gas mixture containing 60% CO₂ exhibits a different trend compared to the others and this is attributed to a different kinetic pathway of reduction but to validate this would require an in-depth kinetical analysis.

Taking the case of pure CO₂, the energy efficiency is again much higher than other works with a maximum around 35% being achieved at the lowest SEI/longest residence time. Except for raw flue gas, N₂ addition has a positive effect on efficiency, and at higher values of SEI, as the N₂ content is decreased the efficiency increases. It is theorized that up to 20% N₂ content in the gas mixture has a positive effect on CO₂ conversion and efficiency owning to the formation of N₂ metastables, which are effective at transferring energy to lower energy states of CO₂. Further increase in N₂ quenches these metastables, and the beneficial effect is no longer observed. Furthermore, as the gas content contains more N₂, a higher energy requirement is needed to ignite and sustain the plasma, and more energy is directed toward reactions of N-containing species and diverted away from CO₂ dissociation. This effect becomes dominating as more N₂ is added as demonstrated in other works (Heijkers et al., 2015).

The main reason for higher efficiencies being achieved has been stated in previous work and proven by means of numerical modeling (Moss et al., 2017). The unique form of the applied voltage pulse, characterized by an initial large amplitude pulse followed by a series of smaller amplitude pulses, allows a two-step pathway of CO₂ utilization to be established. During the high amplitude pulse, ionization and vibrational excitation of the gas are the dominant reactions; however, during the series of smaller amplitude pulses, dissociation of neutral and vibrationally excited CO₂ reactions dominates, which is the most direct route to produce CO. In contrast to other works, the role of vibrational excitation and subsequent dissociation is not expected to play a major role in the formation of CO; as the relaxation times of such species are shorter than the inter-pulse duration. N₂ vibrational species may play a role in enhancing CO₂ conversion as shown in the literature but without a detailed kinetic model of this type of plasma discharge it cannot be conclusively stated. Small quantities of N-containing and O-containing molecules are expected to lower the overall conversion and efficiency as energy will be diverted toward reactions of these components but as the quantity of these is small their effects on conversion and efficiency are minimal.