Development of Microdroplet Generation Method for Organic Solvents Used in Chemical Synthesis

Recently, chemical operations with microfluidic devices, especially droplet-based operations, have attracted considerable attention because they can provide an isolated small-volume reaction field. However, analysis of these operations has been limited mostly to aqueous-phase reactions in water droplets due to device material restrictions. In this study, we have successfully demonstrated droplet formation of five common organic solvents frequently used in chemical synthesis by using a simple silicon/glass-based microfluidic device. When an immiscible liquid with surfactant was used as the continuous phase, the organic solvent formed droplets similar to water-in-oil droplets in the device. In contrast to conventional microfluidic devices composed of resins, which are susceptible to swelling in organic solvents, the developed microfluidic device did not undergo swelling owing to the high chemical resistance of the constituent materials. Therefore, the device has potential applications for various chemical reactions involving organic solvents. Furthermore, this droplet generation device enabled control of droplet size by adjusting the liquid flow rate. The droplet generation method proposed in this work will contribute to the study of organic reactions in microdroplets and will be useful for evaluating scaling effects in various chemical reactions.


Introduction
Microfluidic technology allows the manipulation of minute volumes of fluids in micro/nanoscale flow channels. On the micro/nanoscale, certain physical properties become more influential, causing fluids to exhibit behavior distinct from that of macroscopic fluids. In particular, the characteristics of micro/nanoscopic fluids are dominated by three phenomena, namely, highly efficient mass and heat transfer, a strong influence of viscosity, and pronounced surface effects [1][2][3]. In addition, microfluidic systems permit systematic control over and manipulation of fluids and fluid interfaces [4]. Consequently, microfluidic technology has recently attracted considerable attention, especially in the fields of chemistry [5,6], biology [7,8], medicine [9], and engineering [10].
Channel-based droplet microfluidics is a subfield of microfluidics involving the generation and manipulation of droplets in microchannels. Such droplets are typically generated using immiscible multiphase fluids composed of a continuous phase (carrier) and dispersed phase (droplets) and are systematically manipulated via passive or active methods [11,12]. To date, a variety of droplet manipulation techniques have been devised to permit sorting [13], merging [14], splitting [15], mixing [16], and trapping [17] of droplets. The combination of these methods has led to the development of microscale droplet reactors with numerous potential biomedical [18][19][20] and chemical [21] applications. These reactors enable laboratory operations such as mixing reagents, transporting samples, and performing observation or analysis of target samples on a micro scale.
For chemical reactions and analysis, microscale droplet reactors are effective and efficient experimental tools. They show enhanced mixing and mass transfer within short diffusion distances and therefore can shorten reaction times [22][23][24][25]. They also allow for fast heat transfer and can reduce response times for controlling reaction temperatures. Droplets serve as small-volume containers in which reactions can be isolated. They are monodispersed and can be generated extremely rapidly. Therefore, microscale droplet reactors are suitable for quantitative studies as well as high-throughput experiments. In addition, they offer the possibility of performing many reactions in parallel using a single chip [26]. However, microdroplet techniques have predominantly been applied to aqueous-phase reactions rather than reactions in organic solvents. High reactivity organic reactions require strict temperature control and low reactivity organic reactions requires long experimental time of several to dozens of hours [27]. Microdroplet methods have the potential to eliminate the need for temperature control due to fast heat diffusion and accelerate reactions due to fast mass transfer. Although the advantages of organic-phase reactions using microdroplets have been demonstrated [28], examples of such reactions remain very limited.
The reason why microdroplet methods have not yet been widely used in studies of organic-phase reactions is that the physical and chemical properties of organic solvents are not compatible with conventional microfluidic devices. There are two key challenges in the generation of organic solvent droplets. The first is damage to device materials by organic solvents. The second is the difficulty of selecting a combination of carrier and organic solvents that can form stable monodisperse droplets. With respect to the chemical resistance of device materials, microfluidic devices are typically fabricated using soft lithography techniques and are composed of polydimethylsiloxane (PDMS) [29]. However, PDMS swells upon contact with organic solvents, potentially damaging the channel pattern [30]. To avoid swelling, resin devices with high chemical resistance, including a thiol-ene-based resin device, have been developed [31]. However, swelling in these devices was caused by some organic solvents, such as acetone and dichloromethane, despite the application of long-term heat treatment to increase their chemical resistance. In addition, thiol-ene-based resin shrinks during the curing process and expands during heating [32,33], which lowers the accuracy of the channel pattern. Consequently, devices fabricated from physically stable materials with high chemical resistance such as silicon and glass are essential. With respect to the combination of carrier and organic solvents, two important properties are low solubility of the organic solvent in the carrier and the relative wettability of the carrier and organic solvents on the channel surface. Even if the carrier and the organic solvent are immiscible, if the organic solvent has higher wettability than the carrier on the channel surface, then droplets will not form. Therefore, for each organic solvent, the carrier should be carefully selected with consideration given to its miscibility with the organic solvent and its wettability on the channel surface. For these reasons, it has not yet been possible to apply microdroplet techniques to organic-phase reactions despite their expected utility in the field of organic chemistry. Therefore, the development of methods for generating droplets of organic solvents is urgently needed.
In this work, droplet generation was successfully achieved for five typical organic solvents frequently used in organic synthesis, namely, toluene, chloroform, methanol, tetrahydrofuran (THF), and dimethyl sulfoxide (DMSO). Droplet generation for some organic solvents incompatible with resins, namely, acetone, acetonitrile, and dichloromethane, was also confirmed. Combinations of organic solvents and continuous-phase solutions that do not mix with each other and are suitable for droplet generation were determined. These combinations were selected with consideration of miscibility and ease of modifying wettability on the channel surface and were then used to generate stable droplets of the organic solvents in a silicon/glass-based device. The inertness of the device materials precluded the possibility of channel swelling. Droplet size and droplet generation rate are important parameters for the application of microdroplet methods to chemical reactions and should thus be controllable. Accordingly, the variation of droplet size and droplet generation rate under different flow rates of the dispersed and continuous phases were also evaluated. Figure 1 presents images of the cross-junction area of the silicon/glass-based devices and shows the flows of the five combinations of dispersed-phase and continuous-phase solutions. For all five organic solvents, the dispersed-phase solution flowed through the channel under laminar flow when the carrier solution did not have enough encapsulation ability, as shown in Figure 2 for toluene as a representative example. In contrast, when the carrier solution had sufficient encapsulation ability, monodisperse droplets of the organic solvents were continually generated. The flow rate at which droplets were generated was dependent on the combination of the dispersed and continuous phases ( Table 1). The standard deviation of droplet length was as low as 4.3% at most and the droplet generation rate ranged from 9.55 to 23.9 Hz. No deformation of the device channels was observed, and stable droplet generation of toluene and chloroform continued for several hours whereas that of methanol and DMSO continued for about an hour. THF droplet generation continued for tens of minutes. However, methanol, DMSO, and THF gradually wetted the channel surface. This was due to a change in wettability. Wetting by methanol and DMSO was likely because the hydrophobic treatment wore off over time. Wetting by THF was likely because THF and saturated saline separate into two phases but are not completely immiscible (Table 3), and THF gradually contacted the channel surface. Therefore, an improved hydrophobic treatment method and a dispersed phase solution with lower solubility in THF are needed for long-term methanol, DMSO, and THF droplet generation. The generated droplets of toluene and chloroform were passed through the channel to the outlet without collapsing. Toluene and chloroform droplets could be transported to the pool consisting of carrier solution and oil and did not collapse for at least 30 min. These results indicate that a surfactant was helpful for generating and stabilizing organic solvent droplets, as in the case of water-in-oil and oil-in-water droplets [34]. However, methanol, DMSO, and THF droplets sometimes merged at the outlet or collapsed after transportation. This was due to insufficient optimization of the surfactant concentration. Accordingly, stable droplet generation was successfully realized for five common organic solvents using immiscible continuous-phase solutions and surfactants with appropriate hydrophile-lipophile balance (HLB) values.

Droplet Generation of Organic Solvents Incompatible with High Chemical Resistance Resins
To confirm the high chemical resistibility of silicon/glass-based devices, fluid experiments with acetone, acetonitrile, and dichloromethane, which swelled thiol-ene-based devices, were also conducted. Figure 3 shows the successful generation of acetone, acetonitrile, and dichloromethane droplets, which were generated at the cross-junction area of the channel, as is the case with typical organic solvents. The standard deviation of droplet length was as low as 3.1% at most and the droplet generation rate ranged from 27.3 to 47.6 Hz. The droplet generation of these solvents was stable and continued for several hours. Dichloromethane droplets were passed through the channel without collapsing or merging at the outlet. However, acetone and acetonitrile droplets sometimes merged in the channel due to the absence of surfactants. Table 2 shows the combinations of organic solvents and carrier solutions for acetone, acetonitrile, and dichloromethane droplet generation. For dichloromethane droplet generation, 1 wt % PVA solution was used as the carrier solution, which prevented the droplets from merging. However, surfactants could not be used for generating droplets of acetone and acetonitrile because they are miscible with many solvents. Accordingly, there were few carrier solution candidates. Although carrier liquids were found that were immiscible with acetone and acetonitrile and could generate droplets, surfactants that are soluble in the carrier liquids and have sufficient encapsulation ability have not yet been found. To prevent droplet merging, a proper surfactant is required, and thus further investigation is needed. Nevertheless, it was confirmed that the silicon/glass-based device has high chemical resistance and is capable of generating droplets of various organic solvents, according to these droplet generation results.
Molecules 2020, 25, x FOR PEER REVIEW 5 of 12 To confirm the high chemical resistibility of silicon/glass-based devices, fluid experiments with acetone, acetonitrile, and dichloromethane, which swelled thiol-ene-based devices, were also conducted. Figure 3 shows the successful generation of acetone, acetonitrile, and dichloromethane droplets, which were generated at the cross-junction area of the channel, as is the case with typical organic solvents. The standard deviation of droplet length was as low as 3.1% at most and the droplet generation rate ranged from 27.3 to 47.6 Hz. The droplet generation of these solvents was stable and continued for several hours. Dichloromethane droplets were passed through the channel without collapsing or merging at the outlet. However, acetone and acetonitrile droplets sometimes merged in the channel due to the absence of surfactants. Table 2 shows the combinations of organic solvents and carrier solutions for acetone, acetonitrile, and dichloromethane droplet generation. For dichloromethane droplet generation, 1 wt % PVA solution was used as the carrier solution, which prevented the droplets from merging. However, surfactants could not be used for generating droplets of acetone and acetonitrile because they are miscible with many solvents. Accordingly, there were few carrier solution candidates. Although carrier liquids were found that were immiscible with acetone and acetonitrile and could generate droplets, surfactants that are soluble in the carrier liquids and have sufficient encapsulation ability have not yet been found. To prevent droplet merging, a proper surfactant is required, and thus further investigation is needed. Nevertheless, it was confirmed that the silicon/glass-based device has high chemical resistance and is capable of generating droplets of various organic solvents, according to these droplet generation results.

Droplet Size and Generation Rate Variation
Given that scaling effects are important in chemical reactions involving microdroplets, control over droplet size is desirable. Therefore, the influence of the flow rates of the two phases on droplet size was investigated using toluene as a representative example. Figure 4a shows the variation of the mean toluene droplet length under different flow rates of the dispersed phase (toluene; 0.5, 1.0, or 1.5 µL/min) and continuous phase (water containing 4 wt % Tween 20; 0.5, 1.0, 1.5, 2.0, 3.0, 4.0, or 5.0 µL/min). The mean droplet length ranged from 110 to just over 280 µm. As shown in the plot, the mean droplet length decreased with increasing flow rate of the aqueous phase. The flow rate of the organic phase had the opposite effect on the droplet length, with an increased flow rate affording

Droplet Size and Generation Rate Variation
Given that scaling effects are important in chemical reactions involving microdroplets, control over droplet size is desirable. Therefore, the influence of the flow rates of the two phases on droplet size was investigated using toluene as a representative example. Figure 4a shows the variation of the mean toluene droplet length under different flow rates of the dispersed phase (toluene; 0.5, 1.0, or 1.5 µL/min) and continuous phase (water containing 4 wt % Tween 20; 0.5, 1.0, 1.5, 2.0, 3.0, 4.0, or 5.0 µL/min). The mean droplet length ranged from 110 to just over 280 µm. As shown in the plot, the mean droplet length decreased with increasing flow rate of the aqueous phase. The flow rate of the organic phase had the opposite effect on the droplet length, with an increased flow rate affording larger droplets. Uniform droplet generation was also confirmed, especially for high total flow rates (i.e., the sum of the flow rates of the aqueous and organic phases); the standard deviation of the droplet length ranged from 10% at low total flow rates to 1% at high total flow rates. However, the variation of the mean droplet length was irregular when the toluene flow rate was 1.5 µL/min, which was ascribed to facile wetting of the device surface by toluene and inhibited droplet formation when the toluene flow rate was too high. Figure 4b-d presents images of toluene droplets in the cross-junction area for various aqueous flow rates and a fixed toluene flow rate. It can be clearly seen that the flow rate of the continuous phase affected the toluene droplet length. These results indicate that the toluene droplet length could be manually controlled by adjusting the flow rates of the organic and aqueous phases.
Molecules 2020, 25, x FOR PEER REVIEW 6 of 12 larger droplets. Uniform droplet generation was also confirmed, especially for high total flow rates (i.e., the sum of the flow rates of the aqueous and organic phases); the standard deviation of the droplet length ranged from 10% at low total flow rates to 1% at high total flow rates. However, the variation of the mean droplet length was irregular when the toluene flow rate was 1.5 µL/min, which was ascribed to facile wetting of the device surface by toluene and inhibited droplet formation when the toluene flow rate was too high. Figure 4b-d presents images of toluene droplets in the crossjunction area for various aqueous flow rates and a fixed toluene flow rate. It can be clearly seen that the flow rate of the continuous phase affected the toluene droplet length. These results indicate that the toluene droplet length could be manually controlled by adjusting the flow rates of the organic and aqueous phases. The toluene droplet generation rate also changed as the flow rates were changed. Figure 5a,b shows the variation in toluene droplet generation rate under different flow rates of dispersed and continuous phases. It was shown that the droplet generation rate increased with the increasing toluene flow rate or aqueous flow rate. The generation rate ranged from about 3.3 to 2833.3 Hz. Sufficient throughput was achieved at high total flow rates, especially when the flow rate of continuous-phase was high. However, the generation rate of toluene droplets was lower than that of water-in-oil droplets. This was due to the low capillary number of the toluene/water system. The droplet generation rate increases as capillary number (Ca), which can be defined by the following Equation (1), increases [35].
Here, and are the dynamic viscosity and the velocity of the continuous phase, respectively. σ is the surface tension between the dispersed and continuous phases. The viscosity of water with the surfactant was about 0.9 mPa s and the interfacial tension between toluene and water with the surfactant was 10 mN/m [36]. In contrast, the viscosity of mineral oil is 57.2 mPa s and the interfacial tension between water and oil with Span 80 is 5 mN/m. Therefore, the capillary number of a toluene-in-water droplet is significantly, approximately 100 times, lower than that of a water-in-oil The toluene droplet generation rate also changed as the flow rates were changed. Figure 5a,b shows the variation in toluene droplet generation rate under different flow rates of dispersed and continuous phases. It was shown that the droplet generation rate increased with the increasing toluene flow rate or aqueous flow rate. The generation rate ranged from about 3.3 to 2833.3 Hz. Sufficient throughput was achieved at high total flow rates, especially when the flow rate of continuous-phase was high. However, the generation rate of toluene droplets was lower than that of water-in-oil droplets. This was due to the low capillary number of the toluene/water system. The droplet generation rate increases as capillary number (Ca), which can be defined by the following Equation (1), increases [35].
Here, µ c and u c are the dynamic viscosity and the velocity of the continuous phase, respectively. σ is the surface tension between the dispersed and continuous phases. The viscosity of water with the surfactant was about 0.9 mPa s and the interfacial tension between toluene and water with the surfactant was 10 mN/m [36]. In contrast, the viscosity of mineral oil is 57.2 mPa s and the interfacial tension between water and oil with Span 80 is 5 mN/m. Therefore, the capillary number of a toluene-in-water droplet is significantly, approximately 100 times, lower than that of a water-in-oil droplet when the velocity of the continuous phase was the same. To improve throughput, a carrier solution having high viscosity and low interfacial tension with organic solvents will be needed. It was also found that droplet generation destabilized when the total flow rate was too high. This was because the capillary number increased with increasing aqueous flow rate and the generation regime changed from a dripping regime to a jetting regime [35]. Increasing the velocity of the carriers helps to increase throughput, but it must be increased with care to ensure that the generation regime does not change.
Molecules 2020, 25, x FOR PEER REVIEW 7 of 12 solution having high viscosity and low interfacial tension with organic solvents will be needed. It was also found that droplet generation destabilized when the total flow rate was too high. This was because the capillary number increased with increasing aqueous flow rate and the generation regime changed from a dripping regime to a jetting regime [35]. Increasing the velocity of the carriers helps to increase throughput, but it must be increased with care to ensure that the generation regime does not change.  Table 3 summarizes the combinations of continuous-phase and dispersed-phase solutions used in the microfluidic experiments. Organic solvents are typically classified into three types: nonpolar, protic polar, and aprotic polar. This classification scheme is important for chemical reactions and the appropriate organic solvent should be used for a particular target reaction. In this work, five common organic solvents belonging to these three classes and possessing different dipole moments were selected as the dispersed phase to evaluate the versatility of our droplet generation method; toluene and chloroform were used as nonpolar solvents, methanol was used as a protic polar solvent, and THF and DMSO were used as aprotic polar solvents. These solvents are used for some scientifically and industrially important reactions and a wide range of applications is expected. For example, diethyl ether and THF are used as solvents for the Grignard reaction, and toluene is used for the Wittig reaction. Methanol is used as a solvent for synthesis of metal complex proteins that are used as drugs. Furthermore, three organic solvents causing strong damage to resins (acetone, acetonitrile, and dichloromethane) were also selected as the dispersed phase to confirm the high chemical resistance of the silicon/glass-based device. All eight solvents were purchased from Suzuki Chemical Industry (Okazaki, Japan). Liquids that were immiscible with the dispersed-phase solutions were selected as the continuous phase. To select the carrier, simple preliminary experiments were conducted. One of the five organic solvents and a carrier candidate were poured into a glass vial. If the carrier candidate and organic solvent were immiscible, they would clearly separate into two phases. After separation, the glass vial was capped and stirred. If the carrier candidate had a sufficient encapsulation ability, the organic solvents became emulsions. Deionized (DI) water, mineral oil (M8410, CAS 8042-47-5; Sigma-Aldrich, St. Louis, MO, USA), diethyl ether (Suzuki Chemical Industry), fluorinated oil (Novec TM 7300; 3M Japan Limited, Tokyo, Japan), and perfluorocarbon  Table 3 summarizes the combinations of continuous-phase and dispersed-phase solutions used in the microfluidic experiments. Organic solvents are typically classified into three types: nonpolar, protic polar, and aprotic polar. This classification scheme is important for chemical reactions and the appropriate organic solvent should be used for a particular target reaction. In this work, five common organic solvents belonging to these three classes and possessing different dipole moments were selected as the dispersed phase to evaluate the versatility of our droplet generation method; toluene and chloroform were used as nonpolar solvents, methanol was used as a protic polar solvent, and THF and DMSO were used as aprotic polar solvents. These solvents are used for some scientifically and industrially important reactions and a wide range of applications is expected. For example, diethyl ether and THF are used as solvents for the Grignard reaction, and toluene is used for the Wittig reaction. Methanol is used as a solvent for synthesis of metal complex proteins that are used as drugs. Furthermore, three organic solvents causing strong damage to resins (acetone, acetonitrile, and dichloromethane) were also selected as the dispersed phase to confirm the high chemical resistance of the silicon/glass-based device. All eight solvents were purchased from Suzuki Chemical Industry (Okazaki, Japan). Liquids that were immiscible with the dispersed-phase solutions were selected as the continuous phase. To select the carrier, simple preliminary experiments were conducted. One of the five organic solvents and a carrier candidate were poured into a glass vial. If the carrier candidate and organic solvent were immiscible, they would clearly separate into two phases. After separation, the glass vial was capped and stirred. If the carrier candidate had a sufficient encapsulation ability, the organic solvents became emulsions. Deionized (DI) water, mineral oil (M8410, CAS 8042-47-5; Sigma-Aldrich, St. Louis, MO, USA), diethyl ether (Suzuki Chemical Industry), fluorinated oil (Novec TM 7300; 3M Japan Limited, Tokyo, Japan), and perfluorocarbon (PFC)-based solvent (CT-SOLV180; AGC Inc., Tokyo, Japan) were used as the continuous phases for the toluene, chloroform, and dichloromethane droplets, methanol droplets, DMSO droplets, acetonitrile droplets, and acetone droplets, respectively. Saturated saline solution was used as the continuous phase for the THF droplets because these two liquids are immiscible due to salting out. Three nonionic surfactants possessing different HLB values, which indicate the balance in size between the hydrophilic head and hydrophobic tail, were used for stable droplet generation. Surfactant molecules adsorbed to the interface of the dispersed and continuous phases to stabilize it and prevent wetting of the channel by the dispersed-phase solution. A surfactant with a low HLB value is helpful for generating aqueous-in-organic droplets. In contrast, a surfactant with a high HLB value is helpful for generating organic-in-aqueous droplets ( Figure 6). Tween 20 (HLB: 16.7, P1379, CAS 9005-64-5; Sigma-Aldrich) was mixed into the DI water. Surflon S-656 (HLB: 8-9; AGC Seimi Chemical Co., Ltd., Chigasaki, Japan) and Span 80 (HLB: 4.3, 37408-32, CAS 1338-43-8; Kanto Chemical Co., Inc., Tokyo, Japan) was mixed into the mineral oil and diethyl ether. Polyvinyl alcohol (PVA, approximate molecular weight 86000, CAS 9002-89-5, Acros Organics; Thermo Fisher Scientific, Waltham, MA, USA) was also used as a surfactant (HLB: 18.0) and channel modifier. The PVA (1, 3 wt %) was added to DI water and the resulting solution was stirred at room temperature for 1 h. The temperature was then gradually increased to 90 • C and the solution was stirred for an additional 1 h. Finally, the temperature was gradually decreased to 65 • C and the solution was left to stir overnight. After dissolution, DI water was added to compensate for any water loss due to evaporation. To prepare saturated saline solution, sodium chloride (Suzuki Chemical Industry) was dissolved in DI water.

Device Design and Fabrication
The channel design is depicted in Figure 7a. The dispersed-phase solutions and continuous-phase solutions were injected via inlets (a) and (b), respectively. The channel possessed a tapered shape at the cross-junction area to facilitate droplet formation. The rectangles adjacent to the main channel were designed to allow measurement of the droplet size.
The device fabrication process is presented in Figure 7b. The devices were fabricated via lithography and dry-etching techniques. First, a silicon wafer with a thickness of 525 µm and soda-lime glass with a thickness of 500 µm were cut to obtain square substrates (30 × 30 mm 2 ). Positive photoresist (OFPR-800LB, 200cp; Tokyo Ohka Kogyo, Tokyo, Japan) was spin-coated onto the silicon substrate and the channel design was patterned onto the resist via ultraviolet lithography (MA/BA6; SUSS MicroTec, Garching, Germany). The silicon substrate with the patterned resist was dry-etched via deep reactive ion etching (RIE-400iPB; Samco, Kyoto, Japan) to fabricate the channels. After the first etching step, the channel was partially covered with polyimide tape to restrict the etching area prior to a second cycle of etching to form the through holes. After completion of the etching processes, the resist was removed and the glass substrate was bonded to the silicon substrate by anodic bonding to seal the channels. Finally, eyelets were glued onto the substrate to provide the inlets and outlet. additional 1 h. Finally, the temperature was gradually decreased to 65 °C and the solution was left to stir overnight. After dissolution, DI water was added to compensate for any water loss due to evaporation. To prepare saturated saline solution, sodium chloride (Suzuki Chemical Industry) was dissolved in DI water.   The channel design is depicted in Figure 7a. The dispersed-phase solutions and continuousphase solutions were injected via inlets (a) and (b), respectively. The channel possessed a tapered shape at the cross-junction area to facilitate droplet formation. The rectangles adjacent to the main channel were designed to allow measurement of the droplet size.

Device Design and Fabrication
The device fabrication process is presented in Figure 7b. The devices were fabricated via lithography and dry-etching techniques. First, a silicon wafer with a thickness of 525 µm and sodalime glass with a thickness of 500 µm were cut to obtain square substrates (30 × 30 mm 2 ). Positive photoresist (OFPR-800LB, 200cp; Tokyo Ohka Kogyo, Tokyo, Japan) was spin-coated onto the silicon substrate and the channel design was patterned onto the resist via ultraviolet lithography (MA/BA6; SUSS MicroTec, Garching, Germany). The silicon substrate with the patterned resist was dry-etched via deep reactive ion etching (RIE-400iPB; Samco, Kyoto, Japan) to fabricate the channels. After the first etching step, the channel was partially covered with polyimide tape to restrict the etching area prior to a second cycle of etching to form the through holes. After completion of the etching processes, the resist was removed and the glass substrate was bonded to the silicon substrate by anodic bonding to seal the channels. Finally, eyelets were glued onto the substrate to provide the inlets and outlet.

Channel Surface Treatment
The channel surface was modified to alter its wettability to hydrophilic or hydrophobic depending on the type of continuous-phase solution [37,38]. For the microfluidic experiments involving toluene, chloroform, THF, and dichloromethane, a hydrophilic surface was required. This was accomplished by first passing PVA solution through the device channel for 10 min using a glass syringe. The device was then heated at 80 °C for 3 min and the PVA solution remaining in the channel was flushed out using DI water. The device was then heated at 80 °C for 1 h to remove any residual moisture. A hydrophobic surface was needed for the microfluidic experiments involving methanol, DMSO, acetone, and acetonitrile. This surface was achieved by first passing a methanolic solution of trichloro(1H,1H,2H,2H-perfluorooctyl)silane (448931-10G, CAS 78560-45-9; Sigma-Aldrich) through the device channel for 10 min using a glass syringe. Then, methanol was then passed through the device for 10 min and the residual methanol was removed by blowing air through the channel followed by heating at 80 °C for 1 h.

Microfluidic Experiment Setup
The device inlets were connected to syringes (1725CX or 1750CX; Hamilton, Reno, NV, USA) via

Channel Surface Treatment
The channel surface was modified to alter its wettability to hydrophilic or hydrophobic depending on the type of continuous-phase solution [37,38]. For the microfluidic experiments involving toluene, chloroform, THF, and dichloromethane, a hydrophilic surface was required. This was accomplished by first passing PVA solution through the device channel for 10 min using a glass syringe. The device was then heated at 80 • C for 3 min and the PVA solution remaining in the channel was flushed out using DI water. The device was then heated at 80 • C for 1 h to remove any residual moisture. A hydrophobic surface was needed for the microfluidic experiments involving methanol, DMSO, acetone, and acetonitrile. This surface was achieved by first passing a methanolic solution of trichloro(1H,1H,2H,2H-perfluorooctyl)silane (448931-10G, CAS 78560-45-9; Sigma-Aldrich) through the device channel for 10 min using a glass syringe. Then, methanol was then passed through the device for 10 min and the residual methanol was removed by blowing air through the channel followed by heating at 80 • C for 1 h.

Microfluidic Experiment Setup
The device inlets were connected to syringes (1725CX or 1750CX; Hamilton, Reno, NV, USA) via PTFE tubing (500 µm, i.d.) or PVC tubing (1000 µm, i.d.). The dispersed-phase and continuous-phase solutions were injected into the device using syringes and syringe pumps (KD Scientific Inc., Holliston, MA, USA). To reduce instability, the flow rate of the syringe pumps was set to 0.1 µL/min or more and the soft tubes were cut as short as possible. Flow in the device channel was observed by optical microscopy (IX71; Olympus, Tokyo, Japan) and images and movies of the fluids were captured at 2000 fps using a high-speed camera (Fastcam Mini AX; Photron, Tokyo, Japan). When the flow rates of the syringe pumps were changed, videos and photos were captured after waiting a few minutes for the flow rates to stabilize.

Conclusions
In this work, we have demonstrated droplet generation for wide varieties of organic solvents. Droplets of five common organic solvents (toluene, chloroform, methanol, THF, and DMSO) and three organic solvents highly harmful to resins (acetone, acetonitrile, and dichloromethane) were successfully generated by using appropriate solutions as the continuous phase. The fabricated silicon/glass-based microfluidic device was found to be superior to devices based on resins such as PDMS and thiol-ene-based resin owing to the high resistance of the constituent materials to many organic solvents. The developed device is also expected to be applicable to the generation of droplets of organic solvents with high dissolving power toward resins, such as chlorinated solvents, without any swelling of the device materials. The device also realizes a minimum droplet length of 50 µm and a maximum droplet length of 300 µm. Furthermore, the size of the organic solvent droplets could be controlled by adjusting the flow rates of the two phases, and uniform droplets were obtained for high total flow rates. This latter property is expected to be advantageous for evaluating scaling effects in organic reactions involving microdroplets. The throughput of droplet generation could be increased with optimization of the flow rates, and the device realizes high throughput of droplet generation (2833.3 Hz) at high total flow rate. It was suggested that the addition of a surfactant to the continuous phase and channel surface treatment were helpful for stable organic solvent droplet generation and stabilizing the generated droplets. Comprehensive characterization of the fabricated devices will require further investigations, including contact angle and viscosity measurements. Optimization of surface treatment will be considered as needed. It is necessary to investigate the effect of surfactants on chemical reactions and to optimize the surfactant concentration. The proposed method is anticipated to be applicable to a variety of organic reactions involving microdroplets. Therefore, the results of this study should prove valuable for research into chemical reactions in microdroplets and contribute to the development of novel chemical synthesis strategies.