The development and numerical simulation of a plasma microreactor dedicated to chemical synthesis

2.1 Microfabrication processes

The plasma microreactors were made by soft imprint lithography based on methods described in the works of Bartolo et al. [9] and Priest [10] with the thiolene-based resin Norland Optical Adhesive 81 (NOA 81, Norland Products Inc., Cranbury, NJ, USA). This optical glue was chosen over traditionally used polydimethylsiloxane (PDMS) for its good physical, chemical, electrical, and optical properties: (1) unlike PDMS devices, NOA 81 microreactors are impermeable to gas such as air and water vapor, which ensures a enclosed environment for plasma [9]; (2) cured NOA 81 has a high elastic modulus (typically 1 GPa), which avoids sagging effects [9]; (3) NOA 81 appears less sensitive to solvent swelling effects than PDMS [9]; (4) the dielectric constant of NOA 81 being 4.05 at 1 MHz, it is as well a performant insulating material; (5) with a high transmittance in the visible and near-UV range, NOA 81 makes it possible for in situ discharge diagnostics (optical emission spectroscopy or ultra-rapid camera measurements). Owing to these properties, NOA 81 is a good candidate for the fabrication of plasma microreactors.

Another important factor for plasma microreactors is the electrodes fabrication. Here in our study, we used electrodes using microsolidics technique [11]. Figure 1 shows the main steps for the fabrication of a plasma microreactor with solidified electrodes. First, using photolithography, we obtained a micropatterned SU-8 thin layer on a Si wafer. The first replica of these micropatterns was realized by PDMS (RTV-615, MG Chemicals, Surrey, British Columbia, Canada) molding (Figure 1A). These PDMS molds were later used to make NOA 81 stickers with microstructured patterns. Two main techniques were employed for the fabrication of NOA stickers, either by stamp technique or by capillarity technique (Figure 1B) [9]. For the stamp technique, several drops of NOA 81 liquid were firstly deposited on a rigid substrate (glass slide, cover slide, Si wafer, etc.). Then the PDMS mold was placed on the NOA 81 liquid, followed by a short UV illumination to partially insulate the NOA 81 while keeping a thin layer of uncured polymer on the surface. As for the second technique, the PDMS mold was firstly put in contact with the rigid substrate, while the NOA 81 liquid was deposited around the PDMS mold afterwards. The polymer liquid would fill in the microstructures of PDMS mold, driven by capillary forces. A short UV illumination was also used for the partial insulation of NOA stickers. Three assembly methods were envisaged for the microfluidic device fabrication. NOA stickers made by either stamp technique or capillary technique may be sealed with another pre-drilled rigid substrate (Figure 1C (II)(III)), whereas these NOA stickers may be also sealed with another pre-drilled NOA sticker (Figure 1C (I)). Once the microfluidic device was closed, it was put on a hot plate for the injection of metallic electrodes (Figure 1D). The material used for the electrodes was an alloy with 52% In and 48% Sn (Indium Corporation of America, Utica, NY, USA).

Figure 1:

Fabrication steps of a plasma microreactor with solidified electrodes. (A) Photolithography and fabrication of PDMS stamp; (B) fabrication of micro-patterned NOA 81 stickers; (C) assembly of a NOA 81 microfluidic device; (D) injection of metallic electrodes.

2.2 T-junction geometry for gas/liquid emulsion

As for the geometry of the plasma microreactor, a T-junction design was used for the generation of bubbles. Garstecki et al. [12] have employed this geometry to controllably generate monodispersed bubbles with and without surfactants at low-to-moderate Reynolds numbers (Re between 0.01 and 10). Figure 2A shows the design of a plasma microreactor with T-junction. With this fabrication process, electrodes were located on the sides of the gas inlet channel, which ensured the possibility of optical observation of plasma discharges. Figure 2B shows a microfluidic plasma reactor with solidified electrodes. For the studied plasma microreactor, d_g=120 μm, d_l=240 μm. The height of the microchannels was h=50 μm. The main channel length (from the T-junction to the outlet) was 2 cm. The liquid phase was 2-propyl alcohol (IPA, RECTAPUR, VWR International S.A.S, Fontenay-sous-Bois, France) or deionized water (Milli-Q water purification system, Merck Millipore, Molsheim, Alsace, France), while the working gas was argon (Alphagaz 1 Argon, Air Liquide, Paris, France). The pressure controller (Fluigent, MCFS-EZ system, Paris, France), regulated by the computer, gave rise to a stable bubble generation within the flow-focusing geometry. The flow rate of the liquid phase varies between 0 and 100 μl/min for IPA flow and between 0 and 50 μl/min for deionized water. The flow rates of the gas phase are estimated by post-treatments of bubble images and vary from 10−² to several sccm. Figure 3 shows the segmented gas-liquid flow generated by the T-junction. Argon gas was dispersed into the main channel with the liquid phase as the continuous phase. As shown in Figure 3A and B, the gas/liquid flow was more stable with IPA than with deionized water, indicated by more regular bubble forms and more stable bubble generation frequencies for the IPA flow. This difference in flow behaviors may result from the difference in surface tension: IPA has a lower surface tension than water. No surfactant was added in water deliberately since surfactant molecules may interfere with the radical diffusion at the gas/liquid interface. As presented in the work of Garstecki et al. [12], it is somehow difficult to control bubble sizes when the discretized phase adheres to the channel wall (the case for water free of surfactant with NOA channel walls).

Figure 2:

Schematic drawing (A) and image (B) of a T-junction plasma microreactor.

Figure 3:

Flow dynamics of the gas/liquid segmented flow: the gas phase is argon; the liquid phase is 2-propyl alcohol (A) and deionized water (B).

2.3 The plasma generation and observation

The two electrodes of the microreactor are connected, respectively to a high-voltage source and the ground. Sinusoidal waves generated by a function generator are magnified by a signal amplifier (Trek 20/20C, Trek Inc., Lockport, NY, USA) to several tens of thousands of volts. Figure 4 illustrates the experimental set-up for the observation of plasma discharges. A current probe (Pearson Current Monitor, model 2877, 200 MHz, Pearson Electronics, Palo Alto, CA, USA) is placed between the microreactor and the ground connection, in order to electrically detect plasma discharges. In this experiment, plasma discharges were observed when the high-voltage source achieved 9 kV as peak-to-peak value at 500 Hz. An intensified charge-coupled device (iCCD) camera (Pimax4, Princeton Instruments, Trenton, NJ, USA) is used to collect images of plasma discharges through a macroscope (Leica Z16 APO, Leica Microsystemes SAS, Nanterre, France). A dark environment is indispensable for the detection of weak optical emission of electrical discharges.

Figure 4:

Schematic illustration of the experimental setup for optical measurement of the plasma microreactor.

2.4 Electron spin resonance spectroscopy measurements

Since radicals are important initiators for chemical synthesis, the diffusion of radicals (hydroxyl radicals in this study) generated by plasma into the liquid phase was studied by the electron spin resonance (ESR) spectroscopy.

Wu et al. [13] demonstrated the generation and diffusion of hydroxyl radicals (·OH), superoxide anion radicals (·O₂−), and singlet oxygen (¹O₂) in a gas/liquid plasma reactor with ESR spectroscopy. The use of spin trapping agents such as 5,5-dimethyl-1-pyrroline N-oxide (DMPO, Interchim, Montluçon, France) avoids the diagnosis problem due to the short life span of radical species (Figure 5A).

Figure 5:

Spin trapping reaction between DMPO and hydroxyl radicals. (A) The chemical equation of the spin trapping reaction; (B) the geometry configuration of the gas/liquid flow; (C) schematic illustation of the spin-trapping process.

2.5 Numerical simulation of OH radical trapping in the plasma microreactor

A simplified 1-D model (COMSOL Multiphysics, COMSOL France, Grenoble, France) was established to estimate the quantity of spin-adduct in the liquid phase with various experimental parameters (the initial concentration of DMPO, the bubble size, and the gas/liquid ratio). The studied zone is represented as a red and blue line in Figure 5B, with the red segment representing the gas phase and the blue segment representing the liquid phase. We supposed that the species distribution is symmetric in the gas bubble as well as in the liquid slug. Therefore, we only studied the diffusion and reaction in half a bubble and half a liquid slug.

In the present model, mass transfer is supposed to occur purely by diffusion in both phases using 10−⁵ and 10−⁹ m² s−¹ as diffusion coefficients in gas and liquid phases, respectively. The basic species transport equations solved in the liquid and gas were therefore as follows:

Where c is the concentration of chemical species i (mol/m³), t is the time (s), and D_g and D_l are the diffusion coefficients (m²/s) in the gaseous phase and in the liquid phase, respectively. In these equations, chemical reactions are modelled as source terms R_i.

Reactions in the gas phase and in the liquid phase (Figure 5C) are described in Tables 1 and 2.

Based on the spin-trapping reaction (Reaction 12 in Table 2), the molar reaction rate of DMPO-OH creation is given by

Where k₁₂ is the reaction rate constant of Reaction 12 (Table 2), [DMPO]_l is the concentration of DMPO in the liquid phase, and [OH]_l represents the hydroxyl radical concentration in the liquid phase.

We assumed that each bubble, when squeezed into the main channel, carried 36 ppm of hydroxyl radicals in the gas phase [17]. This concentration was hypothesized assuming that argon is saturated in water vapor and in accordance with the works of Hibert et al. [17] and Bruggeman and Schram [18]. The size of gas slugs and liquid slugs are represented by r_bubble and l_liquid.(Figure 5B). In the present model, mass transfer is supposed to occur purely by diffusion in both phases using 10−⁵ and 10−⁹ m² s−¹ as diffusion coefficients in gas and liquid phases, respectively. It is also assumed that there is a thermodynamic equilibrium between interfacial phase concentrations for ·OH and H₂O₂ species. The Henry constant of hydroxyl radicals in the gas phase is 760 mol m−¹ Pa−¹ [19].

3.1 Plasma diagnosis

Due to the small size of the microreactor and the weak intensity of plasma discharges, it was extremely complicated to observe plasma discharges by eye directly. Two aspects of plasma discharges are presented here. Figure 6A shows the electrical measurement of plasma discharges. A discharge was generated every half period as indicated by the current peak visible in Figure 6A.

Figure 6:

Electrical diagnosis (A) and optical diagnosis (B) of the plasma.

Figure 6B shows the electrical discharges in the vicinity of the T-junction using iCCD camera. The internal trigger mode was used with a gate delay of 25 ns and a gate width of 2 ms. We observed an emission of photons of the gas phase between the two electrodes, which proved the ignition of the plasma. We can as well figure out the outlines of the two electrodes in the iCCD image because of their reflecting surfaces.

3.2 Study of spin trapping reaction

An aqueous solution of DMPO was prepared with a concentration of 0.05 mol/l for the spin trapping reaction. The pressures for the gas and the liquid inlet were both 90 mbar. After the T-junction, the gas/liquid segmented flow continued travelling several minutes before reaching a reservoir for ESR spectroscopy. It is to be noted that the production of hydroxyl radicals in the gas phase should only happen before the formation of gas slugs since no discharges were possible at the downstream of the T-junction. The gas slug length was 2r_bubble=870 μm, and the liquid slug length was l_liquid=1900 μm. However, no DMPO-OH signal was detected by the ESR spectroscopy.

We used mathematical tools to estimate the concentration of DMPO-OH at the outlet of the microreactor. For the experimental condition (r_bubble≈500 μm, l_liquid≈2000 μm), the final concentration of DMPO-OH calculated by the model was inferior to 10−⁸ mol/l. Because the sensitivity of our ESR spectroscopy apparatus goes down to ~10−⁶ mol/l, there was no way that a spin adduct signal could be detected experimentally.

Using the numerical simulation, a parametric study was carried out to study the influence of the initial concentration of DMPO ([DMPO]₀), the bubble size (r_bubble), and the liquid/gas ratio (l_liquid/r_bubble) on the final concentration of DMPO-OH. The objective was to determine the highest possible spin-adduct concentration under optimum operating conditions. Prior studies on the fluid dynamics of the gas/liquid flow showed that it was complicated to generate tiny liquid slugs with large gas slugs; therefore, the liquid/gas ratio range was chosen between 0.5 and 40. Figure 7A shows that the final concentration of DMPO-OH increases as the initial DMPO concentration increases, whereas the final [DMPO-OH] reaches a plateau level when the initial concentration of DMPO gets to around 10−³~10−² mol/l. In the liquid, the reaction of OH radicals with DMPO competes with the recombination of OH radicals into H₂O₂. The higher the concentration of DMPO, the more selective the reaction between OH radicals and DPMO, explaining the increase of DMPO-OH in the first part of the curve. When the DMPO concentration exceeds 10−³ mol/l, the selectivity is around 100%, and a plateau is reached. This plateau level is determined by the size of the plasma bubble and the liquid thickness. Figure 7B shows that the final concentration of DMPO-OH decreases as the l_liquid/r_bubble ratio gets higher, while [DMPO]₀ remains 0.01 mol/l and the gas slug size remains the same. Since the spin trapping reaction happened at the gas/liquid interface afterall, the more segmented the liquid phase was, the more OH radicals could be trapped into the liquid phase. For the gas bubble size, Figure 7C indicates that the final concentration of spin adduct decreases as the bubble size increases, while the initial DMPO concentration and the l_liquid/r_bubble ratio stay unchanged. This result can be explained as follows: the larger the bubble size, the higher the fraction of gaseous OH radicals which recombined in gaseous H₂O₂ before reaching the gas-liquid interface. Finally, the simulation work indicated an upper limit for the final concentration of spin-adduct in the liquid phase of around 10−⁶ mol/l, which corresponds to the sensitivity threshold of the ESR analysis. According to the simulation, ESR analysis is therefore not sensitive enough to be used with the plasma microreactor used in this study.

Figure 7:

Model predictions of the final spin adduct concentration: (A) at different initial DMPO concentrations; (B) at different lliquid/r_bubble ratios; (C) at different bubble sizes.

3.3 Outlook: Plasma ignition in moving bubbles

The low level of spin-adduct concentration was directly related to the low concentration of hydroxyl radicals generated in the gas bubble. The measurement of Hibert et al. [17] by resonant absorption spectroscopy indicates that the OH radical production in an argon-water mixture is around 36 ppm. The quantity of hydroxyl radicals delivered by each gas bubble to the liquid phase was quite limited, which led to a low spin-adduct concentration. In order to have more hydroxyl radicals diffused into the liquid phase, a solution is to ignite the plasma discharge several times during the passage of the bubble along the main channel, so that a higher amount of radicals can be generated and then transferred into the liquid.

To realize this, a similar microdevice with electrodes injected on each side along the main channel was fabricated (Figure 8). The main channel dimensions of this device are the following: w_g=w_out=240 μm; w_o=w_d=1/2w_g=120 μm; h=80 μm. With this device, it was possible to ignite several times the plasma discharge during the passage of the bubble in the channel using the following conditions: p(Ar)=86 mbar, p(H₂O)=88 mbar, high voltage=12.51 kV, frequency=200 Hz. An illuminated plasma bubble has been observed travelling through the main channel, which shows a continuous ignition of plasma in the reactor (Figure 9). More experiments are underway to quantify how many OH radicals can be injected in the liquid phase using this new microdevice. A European patent has been applied with the latter design [20].

Figure 8:

A plasma microreactor in which the plasma can be regenerated in gas bubbles along the main channel.

Figure 9:

Optical diagnosis for continuous ignition of a plasma bubble in the main channel (internal trigger, gate width 500 μs, gate delay 400 μs, 10 accumulations).