Single Catalyst Particle Diagnostics in a Microreactor for Performing Multiphase Hydrogenation Reactions

Please note that technical editing may introduce minor changes to the text and/or graphics, which may alter content. The journal’s standard Terms & Conditions and the Ethical guidelines still apply. In no event shall the Royal Society of Chemistry be held responsible for any errors or omissions in this Accepted Manuscript or any consequences arising from the use of any information it contains. View Article Online View Journal


Introduction
In the search for more effective catalysts and reaction conditions, the use of high-throughput experimentation to monitor catalyst performance on a single particle level has gained increasing interest over the last decades. [1][2][3][4] A growing field of research that has shown great potential for these type of experiments is Lab-on-a-Chip, and in particular droplet-based microfluidics. [5][6][7] When using two immiscible fluids and specific geometries such as a T-junction or flow focusing geometry, droplets can be created by a combination of surface tension and shear forces. 8 Due to the low Reynolds numbers in microfluidic channels, the droplets are typically monodisperse and equally spaced, and do not merge when traveling through the microreactor. With all kinds of external forces (e.g. electric, magnetic, acoustic) the droplets can be manipulated, offering possibilities for more complex chemistry or droplet sorting. 9 Furthermore, the use of droplets compartmentalizes reagents and products, and thus prevents Taylor dispersion or lateral mixing. The droplet compartments prevent dispersion along the length of the channel due to the parabolic flow profile, as observed in single phase flow. 8 In these droplets, single catalyst particles can be encapsulated to perform high-throughput screening of their catalytic activity. Droplet microfluidics aimed for single particle or single cell diagnostics has mainly been used for biochemical essays, such as the study of single cells or rapid screening for the discovery of new enzymes. [10][11][12] One of the main advantages of droplet microfluidics is that the before-mentioned droplets with a single cell or particle, act as pico-/nanoliter sized reaction vessels i.e. "nanoreactors". The use of optically-transparent reactor materials, such as polydimethylsiloxane (PDMS) or glass, 13 allows for in situ micro-and/or spectroscopic measurements within these droplets. When we perform a catalytic reaction inside droplets that contain a catalyst particle, we can detect the activity of each catalyst particle, by measuring for example induced fluorescence, a specific color change or distinctive Raman or IR bands in the droplet. With an optical or spectroscopic detection method along the channel, the products formed per single particle can be analyzed by measuring all droplets passing the detector. 14 This opens possibilities for the characterization of catalyst particles in liquid-phase reactions. However, a large class of heterogeneous catalytic reactions involve not only a liquid phase and a solid phase, but also a gas phase. This class of reactions brings an extra challenge to the field of microfluidics: not much work has been done on the combined control of both liquids, solids and gases. Recent work on the control of multiphase reactions in microfluidic devices has focused on the immobilization of a heterogeneous catalyst either as a packed-bed or as a coating on the channel interior. [15][16][17][18] Mixed feeds with gas slugs in a liquid flow were used to create a multiphase system. However, the precise control of gas slugs in a liquid flow has proven to be difficult, due to different forces other than the shear force and interfacial tension, that contribute to the formation and stabilization of liquid droplets. 19 Although, adding a surfactant to the liquid phase helps to increase flow stability, 20 it is still difficult to get monodisperse gas slugs or droplets at fixed flow rates for both phases. 21 This motivates the use of a microreactor with gas-permeable channel walls. In this way, we can form stable liquid "nanoreactor" droplets with encapsulated catalyst particles in an immiscible and non-reactive oil phase. Gases can diffuse through the permeable reactor into the droplets, thereby creating a multiphase (gas/liquid/solid) environment. Gas permeable liquids with low solubility for gases, such as fluorinated oil, can be used as the continuous phase to increase the gas flux towards the droplet. In this work, a polydimethylsiloxane (PDMS) microreactor has been developed with three parallel channels, as schematically shown in Figure 1, to perform gas/liquid/solid catalytic reactions for the diagnostics of single catalyst particles. PDMS is a flexible polymer that is often used for the fabrication of microfluidic systems, since the fabrication of PDMS microreactors utilizes soft lithography which is relatively cheap and easy. 22 Furthermore, it is known that PDMS has a high permeability to gases, such as H2 and O2, which is used to our advantage. The outer channels, as shown in Figure 1, are used to flow H2 that permeates through the PDMS walls and reaches the centered liquid channel. In this liquid channel droplets of methylene blue (MB) dissolved in ethanol, containing single particles of a Pd-catalyst, flow in a continuous fluorinated oil phase. As a proof of principle reaction, the hydrogenation of MB to leuco-MB that discolors from blue to colorless is performed. 23 This reaction is catalyzed by 40 µm sized hollow spherical Pd/SiO2 particles encapsulated in the MB/ethanol droplets. The droplets that flow through the reactor channel, have identical residence times at any chosen position along the length of the channel. This allows an in-situ analysis of the ongoing reaction, by measuring the color of the droplet at various spots along the length of the channel.

Microreactor Design and Fabrication
The PDMS microreactor, as shown in Figure 2, was designed to benefit from the gas-permeability of PDMS: a thin wall of 50 µm between the gas and the droplet channel allows for fast diffusion of H2 to the liquid droplets with reactant and catalyst. The hydrogenation reaction can only take place when the three phases meet (the solid catalyst particle, the liquid MB in ethanol and the gaseous H2) and therefore starts when the catalyst particle-containing droplets enter the liquid microchannel (250x250 µm width and depth) surrounded by two H2 gas channels with similar cross-sectional dimensions. Ethanol droplets (containing MB) For the hydrogenation to take place, the gaseous H2 has to travel from the gas channels through the PDMS wall and the fluorinated oil (FC-40) continuous phase before it reaches the ethanol droplet. It is important that the 50 µm thick PDMS wall does not interfere too much with the H2 diffusion in ethanol, to prevent a change in reaction kinetics due to mass transport limitations. Figure 3a schematically shows which factors influence the concentration of H2 in the ethanol droplet. For a flow with spherical droplets, the continuous phase is wetting the channel walls, whereas the droplet phase is non-wetting. This leads to the formation of a thin stagnant film of oil between the ethanol droplet and the PDMS wall. Figure 3b shows that the droplets in the microreactor system are slugs rather than spherical droplets. 24 Based on the small contact angle between the non-wetting slug and the wall, we assume that there is a nanometer thick (but invisible) layer of FC-40 in between the droplet and the PDMS. Since FC-40 has a high H2 gas permeability (similar to PDMS) 25 , consequently there is no diffusion limitation caused by the FC-40 film. From this, we concluded that H2 must first diffuse into and through the PDMS wall before it can enter the ethanol droplet containing the catalyst. From Table 1, it can be observed that the diffusivity (D) of H2 through PDMS is a factor 100 higher than of H2 through ethanol. Inside the PDMS wall and the ethanol droplet, the solubility follows Henry's law. The solubility (S) parameter of H2 is 200 times higher in ethanol compared to PDMS. A high diffusivity, but low H2 solubility in PDMS results in fast transport of H2 through the PDMS membrane. In ethanol, the H2 diffusivity is lower, but its solubility is higher compared to PDMS, leading to H2-saturated ethanol droplets. Therefore, it is  Please do not adjust margins Please do not adjust margins assumed that the PDMS is not introducing significant diffusion limitations. Within the droplets, chaotic advection causes fast mixing 26 , leading to a homogeneous and constant H2 concentration. Figure 3a sketches the situation where the ethanol droplet is completely saturated with H2 and homogeneously mixed, therefore there is a discontinuity in the simplified representation at the PDMS-droplet interface. When the ethanol concentration in the droplet becomes lower because the H2 reacts with MB on the Pd catalyst, H2 can diffuse from the PDMS and dissolve into the ethanol droplet again.

Catalytic Hydrogenation of Methylene Blue
The developed PDMS microreactor was tested for the hydrogenation of MB (blue) into leuco-MB (colorless). As catalyst, we have opted for Pd supported on hollow SiO2 spheres of 40 µm in diameter. We have selected Pd as active metal because of its known activity towards hydrogenation reactions. An added benefit is its stability in air. After a reduction step, the formed Pd particles are not prone to oxidation and can, therefore, easily be transferred through or stored in air. Our microreactor setup lacks options for in situ catalyst reduction steps, thus a stable and robust catalyst system is vital. The hollow core of the silica particles lowers their weight by a factor 10 approximately (bulk density of 0.2 mg/L vs. skeletal density 2.5 mg/L), to improve their flow through the microchannel. In general, catalyst particles with much higher densities than the surrounding liquid have difficulty moving inside microfluidic channels, as they tend to sink to the bottom due to gravity. 30 Once stuck at the channel bottom, heavy catalyst particles cannot easily be removed because it is expected that they stick to the PDMS. Since catalyst support materials often have a higher density compared to the liquid reactants, this sinking is an important factor in microfluidic single particle diagnostics. The prepared catalyst was characterized with SEM, XRD, TEM and N2-physisorption. SEM and TEM images can be found in Figures 4a and b. With SEM, the external surface and morphology was investigated. It was found that the structure and shape of the spherical SiO2 particles was intact after the preparation of the catalyst. The 40 m porous SiO2 microparticles were too big to analyze with TEM, but after crushing them, the remaining fragments could be analyzed to determine the Pd dispersion and nanoparticle size. On average, the Pd nanoparticles are ~ 25-30 nm. The particle size distribution is large, but their dispersion over the SiO2 support is homogeneous, as can be seen in Figure 4d. In the XRD (Figure 4c), diffraction peaks allocated to crystalline Pd were found, and with the Scherrer equation using the Full Width Half Maximum (FWHM) of the XRD peaks, an average crystallite size of ~ 25 nm could be determined. The BET surface area of the particles was calculated to be ~ 180 m 2 /g. For performing the hydrogenation reaction inside the droplet microreactor, a setup was built with a vertically mounted syringe pump to benefit from gravitational forces acting on the catalyst particles. A shaker was connected to the tip of the syringe, inducing vibrations in the tubing and a stir bar was added to the syringe to disperse the particles in the ethanol solution. Both measures are a precaution against particle sedimentation in the syringe or tubing. A detailed image of the setup is given in Figure 5. The droplets were followed with optical microscopy to monitor a color change of all droplets that contained catalyst particles. As part of the experimental setup, the catalyst particles were added to the MB-ethanol syringe prior to starting the flows. Since no reaction can happen before the particles have entered the microreactor, due to the absence of H2 in the glass syringe, we have a well-defined starting time for the reaction (t0). Interestingly, after adding the Pd/SiO2 to a 20 ppm solution of MB in ethanol, a rapid color change of the fresh catalyst was observed from brown to blue, due to the adsorption of MB. Figure 6 shows two microscopy images of the catalyst before and after adsorption of MB. This color change is advantageous for the catalyst characterization: the blue particles should return to their original brown color upon hydrogenation. This color change effect is expected to be more pronounced compared to a discoloration of the light blue color of the MB droplet itself and therefore easier to detect. As a proof of concept experiment, droplets containing blue Pd/SiO2 particles were followed over time, without any liquid flow, leading to stagnant droplets. It can be seen in Figure 7 that after starting the H2 flow the Red/Green/Blue (RGB) values, representing the color of the particle, reach a plateau over time. During the experiments a H2 sensor (Industrial Scientific Pro Gas Badge) is put next to the microreactor for safety reasons. The minimum amount it can detect is 1 ppm and during all experiments, this level is never reached. These results show the working principle of the PDMS hydrogenation microreactor: H2 can diffuse through the PDMS wall and dissolves in the reactant droplet. However, for single particle diagnostics, these experiments should be performed in flow. Therefore, the same hydrogenation of MB was performed at a total flow of 7.5 µL/min, leading to a droplet formation rate of 1  droplet/s and a residence time of 45 s in the microchannel. The analysis of particles in flow is more difficult compared to the static experiment. Therefore, three monitoring regions for characterization were defined at the start, middle and the end of the microchannel (Figure 8a). In this way, all particles can be analyzed for three different reaction times. Figure 8c shows two sets of images of Pd/SiO2 in a MB-ethanol droplet recorded at the start and end regions. Indeed, the blue color of the particles at the start of the channel has disappeared towards the end of the channel. From these results, we conclude that H2 also permeates the PDMS wall, dissolves in the ethanol droplet and participates in the reaction when the droplets experience a flow. Thus, our microreactor design is working: within the 45 s residence time (tr) in the microchannel, the MB hydrogenates to the colorless leuco-MB. To support these findings, control experiments without H2 or active catalyst are performed: the particles stay blue during their residence in the microchannel. These results are shown in Figure S1. The color change is most visible for the more concentrated MB, adsorbed on the particles. Empty droplets contain the same initial concentration of MB as droplets containing a particle. However, the concentration in the droplets is too low to be visible in the images and both empty droplets and particle containing droplets appear 'white' and have no discernable color change in the RGB level. With the presented setup, the encapsulation of particles in droplets was relatively straightforward. However, fluctuating concentrations of particles near the inlet of the microreactor often led to multiple particles per droplet, as can be seen in movies M1-M4. For the flow experiments, a statistically relevant analysis was performed by monitoring the red-green-blue (RGB) levels of the acquired movies in Matlab. The movies were screened for the presence of catalyst particles in a droplet using a small window at three positions on the microreactor (start, middle, and end) as indicated in Figure 8a. Figure S2 shows three examples of particles found at these different locations. Matlab processing was used to find the frames containing particles. The RGB values of all particles were collected using a line profile positioned perpendicular to the flow direction. From these line profiles, it can be clearly seen that the particles have lower RGB values than the background. The lowest 10 RGB values of each particle are summed and averaged to obtain the RGB value for a particle at each position. This tracking allowed for an evaluation of the RGB levels for each particle. Figure 8b shows a boxplot of red levels for multiple particles at the start, middle and at the end of the flow channel. The datapoints obtained from the Matlab script are plotted on top of the boxplot. Only the red value is shown, as that is the value with the most distinctive change upon discoloration, as shown in Figure 7. The values are normalized with respect to a reference position on the PDMS, denoted as background. From this plot, the color change of Pd/SiO2 catalyst particles flowing through the microchannel is clearly visible. At the start, the blue particles show low red values, whereas the brown particles at the middle and end show high red values. For the middle region, only a few particles (i.e. 6) could be measured, leading to less sound statistics. However, due to the similar red color value in the box plot at the middle compared to the end position, it is concluded that the hydrogenation has already been completed at the middle position (54 mm after droplet formation and exposure to H2).
Interestingly, this hydrogenation is a fast reaction and does not need the full residence time of 45 s (when using a total flow of 7.5 µL/min) in the microreactor to be completed. Due to the gradual color change from blue to brown, it is difficult to pinpoint the exact position in the microreactor where the hydrogenation is completed, but it is approximated that the reaction is completed within 10 s, because the red value of the particle does not change after this point, as shown in Figure 7, from frame F onwards. Movies M1 to M4 show hydrogenations at different total flow speeds and indicate the position where the hydrogenation is complete. For a total flow of 7.5 µL/min, ~ 1 droplet per second passes the analysis region. For high-throughput experimentation holds that it is better to use a higher flowrate used. Increasing the flowrate to 40 µL/min leads to larger, but less stable droplets (Movie M4), but the hydrogenation still takes place and the droplet size does not seem to affect the reaction kinetics. Figure 9 Figure 7: Separate frames (a; with an interval of 1.3 s each, going from A to H) show a rapid particle discoloration from blue to brown when the droplets are exposed to a H2 flow via the adjacent gas channels. The plot (b) shows the changes in the RGB channels per frame.
Please do not adjust margins Please do not adjust margins shows the red color values for the start and middle of the reaction channel, for four different flow rates. It can be seen that up to  For the fastest rate (total flow of 40 µL/min) also the end region is analyzed, because for this flow rate there is a clear difference between the end and middle. The graphical representation of the results b) shows that at higher flow rates, the reaction is completed more down-stream in the microchannel. The number of particles (n) analyzed for each flow rate and position are n=8 and n=5 at t1 and t2, respectively for 10 L/min; n=8 and n=9 at t1 and t2, respectively for 15 L/min; n=3 and n=4 at t1 and t2, respectively for 20 L/min; and n=8, n=20 and n=46 at t1, t2 and t3, respectively for 40 L/min. The horizontal spread in datapoints is done for visibility purposes only and contains no other information.
a total flow of 20 µL/min (10/10, w/o), the reaction has completed at the mid region of the microreactor. For a total flow of 40 µL/min (20/20, w/o), the residence time of a droplet in the microreactor is decreased to 8.5 s. At this flow rate, the full channel length is needed for completion of the reaction, meaning that the reaction time for the hydrogenation of MB, at these conditions, is ±8 s. Therefore, the end position is also measured at this flow rate.

Conclusions
In this work, we show the proof of concept of a polydimethylsiloxane (PDMS) microreactor for high-throughput, multiphase catalytic reactions, demonstrating the liquid-phase hydrogenation of methylene blue (MB). A liquid channel with Pd/SiO2 particles of 40 µm encapsulated in droplets containing 20 ppm MB in ethanol are flanked by two gas channels with H2, separated with 50 µm PDMS walls. Due to the high permeability of PDMS, H2 diffuses into the liquid channels and facilitates the Pd-catalyzed hydrogenation of MB at room temperature. In approximately 8 s the reaction completes, as concluded from the discoloration of the particles in droplets going from blue (due to adsorbed MB) to brown (their original color). Droplets containing single particles could be analyzed individually to allow for the diagnostics of multiple heterogeneous catalyst particles, as opposed to the use of packed-bed or wall-coated reactors. This newly developed microreactor has, therefore, proven to be useful for the single particle diagnostics of catalyst particles in complex multiphase reactions. With this we have a new diagnostics tool in hand to improve the statistical relevance of single catalyst particle analysis and bridge the gap towards bulk characterization. In addition, a safe and relatively easy method has been demonstrated to work with H2, as no H2 leak was detected by a commercial sensor during all the experiments and only small amounts of gas are involved. Furthermore, by placing the PDMS chip on a heating stage, reactions at elevated temperatures can be performed. The thermal stability of crosslinked PDMS allows for heating up to 400 °C before it starts to degrade 31 , but in this situation more problems rise, such as the boiling of solvents or a non-uniform heating of the microreactor. In addition, the optical properties of PDMS allow for monitoring reactions with visible (laser) light, therefore allowing to perform in-line UV-vis or Raman spectroscopy. These techniques should be optimized in a way that they can work fast and can detect individual droplets at high flow rates. Finally, more sophisticated microreactor designs with microfluidic droplet sorters can be implemented to sort the catalyst particles of interest.

Conflicts of interest
There are no conflicts to declare.