DROPLET FORMATION AND ENTRAINMENT IN LIQUID-GAS MICROFLUIDIC SYSTEMS

DROPLET FORMATION AND ENTRAINMENT IN LIQUID-GAS MICROFLUIDIC SYSTEMS A Thesis Presented By Pooyan Tirandazi to The Department of Mechanical and Industrial Engineering in partial fulfillment of the requirements for the degree of Master of Science in the field of Mechanical Engineering Northeastern University Boston, Massachusetts August 2017

ii TABLE OF CONTENTS LIST OF FIGURES...iii LIST OF TABLES. vi ABSTRACT vii Chapter 1. Introduction.. 1 1.1 Microfluidics and Droplet Microfluidics... 1 1.2 Droplet Generation in Microfluidic Networks... 3 1.3 A Guide Through the Thesis... 7 Chapter 2. Gas-Liquid Droplet Microfluidics.. 9 Chapter 3. Experimental.. 12 3.1 Fabrication of Microfluidic Devices... 12 3.2 Control and Modification of Microchannel Surfaces... 15 3.3 Experimental Setup... 19 Chapter 4. Flow Regime Mapping.... 22 Chapter 5. Dripping Regime Characterization 27 5.1 Generation Frequency... 27 5.2 Droplet Morphology... 29 5.3 Droplet Monodispersity... 35 Chapter 6. Conclusions... 38 REFERENCES... 40

iii LIST OF FIGURES Figure 1. A comparison of mixing reaction in a standard pressure-driven continuous-flow microfluidic platform (a) and using a droplet-based microfluidic system (b). In a continuous-flow scheme due to wall effects and different residence times mixing happens slower. However, by digitizing the mixing reaction inside distinct droplets, each droplet is mixed much faster and the total mixing time is reduced significantly as a result [13].... 2 Figure 2. Schematic representation of the most frequently used geometries in microfluidics for droplet formation. (A) T-junction or cross-flow geometry, (B) co-flow geometry, and (C) flowfocusing geometry.... 5 Figure 3. Representation of dripping and jetting inside microfluidic capillary tubes [23]. (A) Dripping regime; here droplet formation occurs at the tip of the inner capillary in the dripping regime. (B) Jetting regime; in this case instabilities of an extended liquid jet creates droplets which are commensurate to the jet diameter and are usually smaller in comparison to the dripping regime.... 5 Figure 4. Axisymmetric flow-focusing architecture for controlled production of droplets using a focusing air in a non-microfluidic format [52]. Here, the applied air pressure across the circular orifice results in breakup of the liquid jet into individual droplets.... 10 Figure 5. Schematic representation of the fabrication procedure for production and assembly of microfluidic devices. (1) Required channel geometry is designed and printed on transparency sheets. (2) UV photolithography process is performed using the printed mask to define the patterns on a negative photoresist that is coated on a silicon substrate. (3) After mold fabrication soft lithography using PDMS prepolymer is performed to replicate the channels from the mold into the elastic polymer. (4) After PDMS being solid, each device is cut and punched for the inlets/outlets. (5) PDMS chips are bonded to a clean glass slide after being plasma cleaned. (6) Final microfluidic device for water droplet formation in air. (7) Image of the flow-focusing junction under microscope. Liquid water is injected through the middle channel and meets the two side air streams at a flow-focusing junction. The microchannels depth is 40μm.... 14 Figure 6. Schematic of the method used treating the microchannel walls with fluorosilane chemical. All the setup are placed inside a vacuum chamber in order to prevent the deposition of toxic fluorosilane vapor into the environment.... 16

iv Figure 7. Interaction of liquid and gas in different channel conditions. (A) After plasma treatment; in this situation channels are hydrophilic and as a result the liquid water tends to adhere to the side walls of the microchannel once it enters the junction. (B) After modifying the channel walls with vaporized fluorosilane. In this case, the droplets could form in the channel. However, due to the challenges in the coating process the generation was not reproducible. (C) After post-baking step; here the PDMS has regained its hydrophobic nature and droplets maintain their morphology after generation... 17 Figure 8. Comparison of the surface tension forces between a glass-pdms microchannel (left) and a PDMS-PDMS microchannel (right). Using the provided contact angles in the table, the net surface tension force for a glass-pdms combination is less. Therefore, it provides less resistance during the course of droplet generation and movement inside the channel which results in a more reproducible formation process.... 19 Figure 9. Schematic of the flow circuit for liquid and gas control.... 20 Figure 10. Actual image of different parts of the experimental setup. (1) Air desiccator, (2) manual coarse pressure regulator, (3) voltage-controlled valves, (4) mass flow sensors, (5) needle valves, (6) liquid syringe pump, (7) white light source for the microscope, (8) high-speed camera connected to the microscope, and (9) microfluidic chip placed on the microscope with all the ports connected to their corresponding tubing.... 21 Figure 11. Flow regime map of water-in-air droplet formation in a planar flow-focusing microfluidic device. Different flow regions are distinguished with different colors. The transition between the flow regimes is also represented as a finite shaded region rather than a solid line to account for the uncertainties associated with the experiments and calculations of the We values.... 25 Figure 12. Comparison of the flow map for chips with different liquid channel sizes. The area for the Dripping region (the green region) is reduced as the liquid channel size increases. However, the Threading and Co-flow regions have expanded as a result of increase in the surface tensions forces that holds the thread and prevents the detachment at lower gas flow rates.... 26 Figure 13. Experimental data of droplet generation frequency as a function of air flow rate for different liquid flow rates. We can see that there is a relatively linear correlation between the frequency value and the air flow rate.... 28 Figure 14. Experimental data of droplet generation frequency at increasing liquid flow rate for two different gas flow rates. We can see that there is a relatively linear correlation between the frequency value and the liquid flow rate.... 28 Figure 15. Experimental data of the non-dimensioalized droplet generation frequency. Generation frequency in this system can be scaled as ƒ Q GQ L.... 29 Figure 16. Experimental data of droplet width inside the microchannel for different liquid and gas flows.... 30

v Figure 17. Experimental data of droplet length inside the microchannel for different liquid and gas flows.... 30 Figure 18. Snapshots of droplets moving inside microchannel under different gas flow rates.... 31 Figure 19. Corrected length used in non-dimensionalizing droplet morphology data In high liquid flows, droplet exhibit a tail-shape which deviate from their common circular morpholy.... 31 Figure 20. Schematic of the Flow-Focusing geometry and the nomenclature used in scaling analysis for droplet morphology.... 34 Figure 21. Droplet length normalized with the minimum width of the droplets in the Dripping regime as a function of gas Weber number (We G). Minimum droplet width and length are assumed to be equal values that are obtained based on the minimum volume of the prefilled droplet during the formation process at high gas flow rates where droplet tends to have a more circular morphology. This minimum is dependent on the geometry of the flow-focusing microchannel and is calculated to be about 86μm for the geometry used in out experiments.... 34 Figure 22. Droplet length normalized with the minimum length of the droplets in the Dripping regime as a function of gas Weber number (We G). Minimum droplet width and length are assumed to be equal values that are obtained based on the minimum volume of the prefilled droplet during the formation process at high gas flow rates where droplet tends to have a more circular morphology. This minimum is dependent on the geometry of the flow-focusing microchannel and is calculated to be about 86μm for the geometry used in out experiments.... 35 Figure 23. (Left) Experimental images of the collected droplets in the immiscible liquid solution. (Right) detected droplets using image analysis with MATLAB.... 36 Figure 24. Actual images and size distributions of samples of 500 collected droplets within the Dripping and Multi-Satellite Formation regimes. Liquid flow rate for both cases is 5µL/min. Multi-Satellite Formation is obtained at higher We G according to the flow map. In the Dripping region by having a constant liquid and gas flow rate the generated droplets are uniform in size. In Multi-Satellite Formation regime, however, small daughter droplets are generated beside the main droplet which results in polydisperse distribution of the droplets.... 37

vi LIST OF TABLES Table 1. Measured contact angles for 10 samples of PDMS and glass substrates with the same preparation conditions of plasma treatment followed by a prebake process. The prebake process results in regaining the hydrophobic nature of PDMS. The glass surface also becomes less hydrophilic after the prebake process. Smaller contact angle hysteresis of the glass substrate results in easier droplet movement during formation. Therefore, it will help to obtain a wider Dripping regime and higher monodispersity.... 17

vii ABSTRACT Controlled generation of droplets in microfluidic networks is a promising tool for numerous applications in biochemistry and material sciences. Droplet-based microfluidic systems are typically utilized for creating dispersions of a liquid or gas within a second continuous liquid phase, such as creating uniform liquid emulsions or gas bubbles (foams). In this work, we have studied some of the fundamental aspects of liquid droplet formation within a continuous gas flow for creating uniform aerosols inside microfluidic channels. We have experimentally investigated interactions of liquid water within a high-speed air flow inside confined flow-focusing microchannels in which we identify six distinct flow regimes: Co-flowing, Threading, Plugging, Dripping, Multi-Satellite Formation, and Jetting. These flow regimes and their transitions are plotted and characterized based on the Weber number (We) of the two phases. Generation frequency, morphology, and monodispersity of the droplets are characterized in more detail for the Dripping regime and scaling laws are provided to better elucidate the mechanism of droplet formation for this system. Results of this work establish a relationship between generation frequency and the product of the liquid and gas flow rates. However, droplet morphology (length and width) is exclusively dependent on the air flow rate. Finally, we demonstrate the production of monodisperse droplets (d<100μm and σ/d<0.05) at khz formation rates in liquid-gas microfluidic systems. The results of this work provide practical and useful guidelines for precise oil-free delivery of ultra-small volumes of fluid which can be integrated in Lab-on-a-Chip (LOC) and micro-total-analysis Systems (μtas) for a variety of applications in biochemical research and material synthesis.

1 Chapter 1. Introduction 1.1 Microfluidics and Droplet Microfluidics Microfluidics has been increasingly developed over the past two decades. A consequence of this galloping growth is transforming the way in which many chemical and biological experimentations are performed [1] [4]. Thus, it has become an essential tool in a wide range of medical researches and technical applications. Despite the wide spectrum of microfluidic devices and schemes, they can be categorized into two major groups, continuous-flow and droplet-based systems [5]. Continuous-flow scheme is characterized by manipulation of a continuous fluid stream inside usually-confined microchannels. In a droplet-based system, however, one fluid is broken up within a second phase to create discrete droplets. The development of microfluidic platforms in different science and technologies has been driven by a number of fundamental characteristics related to system miniaturization [6]. As such, many advantages are associated with these miniaturized platform, including the small sample volumes consumption, low-cost fabrication methods, enhanced operational flexibility, and the ability to integrate different components within more complex networks for different analytical schemes [7]. Droplet-based microfluidics enables even more capabilities while overcoming some of the inherent limitations of the continuous-flow systems. In a continuous-flow scheme, intimate contact of the continuous fluid stream with microchannel walls is associated with parabolic velocity profile and therefore, different residence times [8]. Moreover, near-zero velocities at the wall result in fouling and contamination of the flow and the whole process as a consequence [9]. Creating microdroplets has emerged as a potential and appealing approach to perform a wide range of chemical and biological processes in a peerless fashion [10], [11]. The use of picoliter to nanoliter scale droplets further reduces sample consumption and increases surface-are-to-volume ratio which corresponds to much faster diffusion rates [5]. More importantly, discrete droplets are ideally suited to isolate and compartmentalize a

2 sample or reaction which obviate issues of a continuous-flow scheme (Figure 1). Therefore, unlike continuous-flow systems, this approach allows precise definition of the components of a process inside distinct droplets. In addition, droplet-based systems are intrinsically associated with high throughputs and parallel processing since typical droplet-based platforms can generate hundreds to thousands of drops per second allowing for any post process to be performed in parallel on individual droplets [12]. Figure 1. A comparison of mixing reaction in a standard pressure-driven continuous-flow microfluidic platform (a) and using a droplet-based microfluidic system (b). In a continuous-flow scheme due to wall effects and different residence times mixing happens slower. However, by digitizing the mixing reaction inside distinct droplets, each droplet is mixed much faster and the total mixing time is reduced significantly as a result [13]. Microfluidic technologies provide tremendous advantages for precise and rapid manipulation of fluids at sub-microliter scale inside micrometer channels. All the advantages of microfluidic systems come at the cost of understanding physics of the flow at microscale [14], [15]. These physics can be better understood in terms of a few dimensionless numbers. Typically, gravity

3 effects are neglected in the microfluidic domain. The importance of gravity to surface tension forces is expressed in term of Bond number (Bo); Bo= ρgl2 σ (1.1) where Δρ is the density difference between the dispersed and continuous phase, g is the gravitational acceleration, D is the characteristic length scale, and σ is the interfacial tension. Due to micrometer dimensions of the channels in microfluidic systems the effect of interfacial tension is more pronounced and as a result Bo<<1. The Reynolds number (Re) compares the relative importance of inertial and viscous forces and is defined as; Re= ρud μ (1.2) where U a characteristic velocity, and μ is the fluid viscosity. Similarly, the Weber number defined as; We= ρu2 D σ (1.3) compares inertia to surface tension. Capillary number (Ca) which relates the ratio of viscous to surface tension forces is the most commonly used dimensionless parameter in droplet microfluidic systems [15] and is defined as; Ca= μu σ (1.4) Expressing the flow in terms of the above dimensionless numbers facilitates better identification of the key parameters in droplet generation and breakup, thereby unifying the behavior of a wide variety of microfluidic systems. 1.2 Droplet Generation in Microfluidic Networks Droplet formation and entrainment processes are the essential steps for using any droplet-based system in applications which rely on discretizing a fluid sample. Many techniques for droplet formation have been presented which can be classified in various ways [16]. Droplets in

4 microfluidic networks can be produced using either active or passive techniques. Active droplet generation is achieved with the aid of additional energy inputs. Some of the commonly used methods in this category utilize electrical, magnetic, mechanical, centrifugal, and thermal mechanisms to actuate and control droplet breakup [17]. In passive methods, on the other hand, there are no external actuation mechanisms and droplet formation solely relies on the interaction of fluids interfaces inside special microgeometries. In comparison to active methods, passive techniques are usually less complex and easier to implement. Geometry of the microchannel, especially in passive formats, is of great importance for controlling droplet generation. Various microfluidic device geometries have been demonstrated for this purpose among which T-junction (cross-flow), co-flow, and flow-focusing configurations are the most prevalent formats (Figure 2) [18]. In a T-junction geometry, one channel intersects a main channel perpendicularly where one phase disperses within a continuous flow of a different phase [19]. In a co-flow configuration, the continuous and dispersed phase flow in parallel channels whereas the continuous phase surrounds the dispersed fluid [20]. This geometry is usually fabricated using inserted capillary tubes and is challenging to fabricate and implement in quasitwo-dimensional microfluidic formats. Flow-focusing geometries utilize a constriction in their junction where both phases are hydrodynamically focused through a small orifice [21]. Since the number of geometrical parameters in a flow-focusing geometry is higher in comparison to a T- junction and co-flow configurations, flow-focusing systems offer a more versatile generation scheme for controlling the droplet size and generation frequency. Several works have studied these geometries and other variations to obtain small droplets with high degree of monodispersity.

5 Figure 2. Schematic representation of the most frequently used geometries in microfluidics for droplet formation. (A) T-junction or cross-flow geometry, (B) co-flow geometry, and (C) flow-focusing geometry. Despite the different configurations for droplet formation, breakup process in microfluidic systems can be divided into a few fundamental modes. Dripping and jetting formation of droplets are usually the two most desired modes of breakup inside microfluidic channels to produce drops [22], [23]. Dripping mode is characterized by formation and breakup of droplets very close to the dispersed channel inlet, whereas in the Jetting regime, a liquid jet elongates in the microchannel outlet where the breakup occurs due to growing instabilities in the liquid column (Figure 3). Dripping regime is associated with higher monodispersity yet bigger droplets at slower frequencies comparing to the jetting regime. By increasing either the continuous phase or the dispersed phase flow rates, transition from dripping to jetting regime could occur. The region associated with these regimes and their transitions to other ones is usually mapped in a plot that usually consist of capillary numbers of the two phases in the system [24]. Figure 3. Representation of dripping and jetting inside microfluidic capillary tubes [23]. (A) Dripping regime; here droplet formation occurs at the tip of the inner capillary in the dripping regime. (B) Jetting regime; in this case instabilities of an extended liquid jet creates droplets which are commensurate to the jet diameter and are usually smaller in comparison to the dripping regime. Another major classification in microfluidic production of drops is by considering the phases involved in the generation process (liquid or gas). In this sense, three combinations are possible i.e. liquid-in-liquid or emulsion generation, gas-in-liquid or bubble generation, and liquid-in-gas or aerosol generation. Creating uniform microemulsions is the most widely used class in microfluidics. In this case, a continuous liquid phase meets the dispersed liquid in one of the

6 aforementioned geometries where individual droplets are pinched off as a result of interfacial tension and shear forces induced by the continuous phase. Droplet sizes can be accurately controlled in this scheme and a variety of techniques have been proposed for subsequent manipulation of the drops in other system processes such as fusion [25] [27], fission [28], sorting [29], [30], and mixing [31], [32] in liquid-liquid environments [33]. Therefore, numerous applications in microbiology, drug discovery, and particle synthesis exist that highly rely on controlled formation on liquid droplets within a continuous liquid stream [34] [36]. For instance, these microdroplets are ideal for encapsulating single biological cells where specific dynamics and metabolites can be conveniently probed within each droplet [37], [38]. Another huge area of liquidliquid digital microfluidic droplets is performing Polymerase Chain Reaction (PCR) in a digitized format [39] [42]. This process which is known as ddpcr (digital droplet PCR) has attracted tremendous attentions during recent years as a highly-sensitive approach for DNA amplification. In this technique, by partitioning the sample into millions of droplets, PCR can be performed on individual drops and the number of positive compartments in which a target sequence is detected can be counted and subsequently related to the initial number of targets in the original sample. Research on incorporating a gaseous phase in microfluidic systems has mostly focused on generation of uniform microbubbles and segmented gas slugs within a liquid phase. This process has been driven by a different set of applications mainly in food industry, pharmaceutical sciences, and medicine. For example, microfluidic formation of bubbles is proven in creation of polymerlipid microbubbles (PLBs) as a template for fabrication of 3D porous materials and scaffolds [43]. Moreover, using encapsulated microbubbles as ultrasound contrast agents (UCA) in medical imaging has gained a lot of attention for their potential in early detection and characterization of diseases [44]. Well-established microfluidic techniques have also been developed in this class for production of highly-monodisperse bubbles in liquid environments using similar device geometries as liquid-liquid systems [45], [46]. However, there is a paucity of studies regarding microfluidic

7 techniques for formation of liquid droplets inside a gaseous flow. Previous works related to liquid droplet formation in a gaseous flow is provided in the next chapter. The major challenge addressed in this thesis concerns investigation of droplet formation in gaseous microflow inside similar planar microfluidic networks as the previous classes. The objective of this study is to gain a better understanding of the fundamental physics behind this process with the ultimate goal of creating uniform micron-sized droplets completely in air in microfluidic devices. 1.3 A Guide Through the Thesis In Chapter 2, an overview of droplet formation in gaseous environments is presented. Previous researches in this regard are mostly geared towards droplet generation using an axisymmetric focusing air in unconfined nozzle-like geometries. In Chapter 3, we discuss the experimental aspects of this work. The methodology for microfluidic chip fabrication is described. We examine droplet formation in a gas flow inside a microfluidic flow-focusing device (MFFD) with mixed hydrophobicity surface conditions. We discuss preparation of the microchannels for obtaining monodisperse droplets in confined microchannels. Unlike previous studies that used different surface chemical coatings to render the microchannel fully hydrophobic, we show that incorporating a mixed hydrophobic/hydrophilic PDMS-glass hybrid structure without chemical coatings results in the most robust droplet generation in these systems. In Chapter 4, we provide the experimental results of different breakup scenarios that occur in the liquid-gas domain inside the microchannel. We identify six distinct regimes which are: Co-flowing, Threading, Plugging, Dripping, Jetting, and Multi-Satellite Formation [47]. The effect that microchannel size has on droplet formation is investigated towards the development of these richer flow maps.

8 In Chapter 5, particular attention is paid to the Dripping regime for the generation of uniform droplets. Droplet morphology, monodispersity and generation frequency are characterized as functions of gas and liquid flow rates in this regime for the first time. Scaling laws are also provided for the droplet morphology and generation frequency to non-dimensionalize the results. In Chapter 6, we provide conclusions of this work and the future avenues that can be pursued for implementation of this system in next generation Lab-on-a-Chip (LOC) technologies.

9 Chapter 2. Gas-Liquid Droplet Microfluidics Spontaneous generation of liquid droplets within a gaseous flow in microfluidic networks is a newer approach to the conventional liquid-in-liquid systems. One of the main advantages of this approach for droplet formation is the possibility of creating uniform particles purely in air (monodisperse aerosols) and without the presence of a second liquid carrier. Therefore, the final product can be readily attained obviating extra washing steps to remove the carrier oil [35]. Moreover, generation of droplets in a liquid medium is usually facilitated by adding surfactants to the continuous liquid carrier. However, it is believed that using certain surfactants may impose usage limitations mainly due to cross-contamination of the droplet contents [48], [49]. Surfactants are clearly not required for droplet generation in a gaseous phase, ensuring preservation of the droplets contents. Some early works have investigated the breakup of a liquid jet into droplets using a focusing air in unconfined nozzle-like geometries as in Figure 4 [50] [54]. Here, a flowing liquid through a capillary tube is forced through a coaxial round orifice located downstream of the tube. The liquid stream is drawn by a focusing gas stream discharging through the nozzle into an infinitely large chamber. In this scheme, drop breakup occurs as a result of the instability of a capillary jet and it has been demonstrated that under certain conditions [54] [56], monodisperse droplets can be achieved from an unstable liquid jet.

10 Figure 4. Axisymmetric flow-focusing architecture for controlled production of droplets using a focusing air in a non-microfluidic format [52]. Here, the applied air pressure across the circular orifice results in breakup of the liquid jet into individual droplets. Generation of monodisperse drops within a microfluidic system is an entirely different process due to confinement induced effects [57]. Interaction of liquid and gas inside confined microchannel geometries has also been widely studied in the operation of Proton Exchange Membrane Fuel Cells (PEMFC) [58] [67]. However, droplet generation in gaseous microfluidic systems that employ similar architectures as those used in conventional oil-based systems has been the subject of a very few studies during recent years. The gas flow velocities required for droplet generation are at least an order-of-magnitude higher than those employed when using a highly viscous oil phase. As such, inertial effects become more prominent and relevant in the fluid-fluid interactions of these systems. Droplet detachment has been numerically and experimentally investigated in a T-shaped junction under the introduction of a high-speed gaseous flow [68], [69]. It was demonstrated that the mechanism responsible for the breakup of the drops transitions from hydrodynamic pressure difference (arising from the microchannel confinement) at lower Re to inertial drag at higher Re as the gaseous flow increases. Confined generation of aqueous droplets has also been shown in a

11 circular capillary co-flow system using gas as the continuous phase. Under relatively low liquid flow rates (less than 1 μl/min) and gas velocities (below 3 m/s) uniform droplets between 250μm and 320μm were obtained within concentric hydrophobic glass capillaries [70]. More recently, confluence of liquid in a gaseous stream in various planar confined flow-focusing microgeometries was studied. It was shown that over a wide range of liquid and gas flow rates three main regions (i.e. Dripping, Jetting, and Stratified Flow) are observed within confined PDMS microchannels [71]. While this study provided useful insights regarding the dynamics of gas-liquid interactions in planar flow-focusing geometries, it did not address several key aspects of this system. For example, the monodispersity and characteristics of the droplets generated within the Dripping region was not presented. This is very relevant information towards the design of these systems since the Dripping regime is one of the two desirable modes of operation for droplet generation. Furthermore, neither study looked at the effect that flow conditions has on the droplet morphology or generation frequency in this flow regime.

12 Chapter 3. Experimental 3.1 Fabrication of Microfluidic Devices We fabricated the microfluidic devices using standard photolithography process with SU-8 and soft lithography with polydimethylsiloxane (PDMS) [72]. First, the flow-focusing configuration was designed in a CAD software (SolidWorks 2014) and then printed out on high-resolution transparencies (CAD/Art services). A 4-inch single side polished silicon (Si) substrate is used as the master mold. The mold is prepared using standard photolithography process with a negative near-uv photoresist (SU-8 2050 Microchem Corp.). Initially, the substrate is cleaned with acetone and isopropyl alcohol, and dried with a nitrogen gun. The clean wafer is spin-coated (Laurell Technologies) with the photoresist. The spinning speed is initially set to 500rpm for complete spreading of the photoresist followed by a final speed of 3000 rpm for 1 min to obtain 40μm thickness. Residues of the photoresist also known as edge beads were removed from the sides of the substrate using a razor blade to ensure a flat coated surface for better mask attachment and exposure process. The coated wafer is then baked for solvent removal on a hot plate at 65 C and 95 C respectively. The prebaked mold is left at room temperature for about 5 min to reduce thermal stresses in the photoresist layer. The transparency mask sheet which contains the chip geometries is attached to a borosilicate glass sheet to provide a rigid support for the mask when covering the wafer. The mask and the wafer are loaded in the mask alignment equipment (Quintel 4000 Mask Aligner) and exposed to UV light through the printed mask for about 45s. The exposed wafer is then post-baked with similar temperatures as the prebake step for 5 min and 10 min respectively. Once the mold reached the room temperature after the post-bake, it was soaked in a solution of SU- 8 developer for defining the micropatterns on the photoresist. A hard-bake process at 120 C is performed for improving the adhesion of the photoresist to the wafer as well as balancing the thermal stresses in the photoresist film. Once the mold fabrication is complete, microfluidic chips

13 were casted out of the SU-8 master molds by performing standard soft lithography method. Polydimethylsiloxane (PDMS) prepolymer mixture (Sylgard 184, Dow Corning) is prepared from the elastomer base and curing agent at a 10:1 weight ratio. The solution is then placed onto the Silicon mold and is degassed in a vacuum chamber. The assembly is cured at 80 C for approximately 2 hours on a leveled hotplate. Hardened PDMS is carefully peeled from the mold. The available area on a 4-inch silicon mold allows a number of devices to be replicated in a single lithography process. Each microfluidic chip is cut out of the PDMS layer. After punching the required flow inlets/outlets, each chip is cleaned with DI water and dried with a nitrogen gun. Each piece is then bonded to a pre-cleaned standard microscope glass slide (Thermo Scientific) using an oxygen plasma cleaner (Harrick Plasma). Finally, the chips are post-baked for 4 hours at 200 C and can be used for experiments after being cooled down to room temperature. The fabrication process and final device and the microchannel configuration is depicted in Figure 5.

Figure 5. Schematic representation of the fabrication procedure for production and assembly of microfluidic devices. (1) Required channel geometry is designed and printed on transparency sheets. (2) UV photolithography process is performed using the printed mask to define the patterns on a negative photoresist that is coated on a silicon substrate. (3) After mold fabrication soft lithography using PDMS prepolymer is performed to replicate the channels from the mold into the elastic polymer. (4) After PDMS being solid, each device is cut and punched for the inlets/outlets. (5) PDMS chips are bonded to a clean glass slide after being plasma cleaned. (6) Final microfluidic device for water droplet formation in air. (7) Image of the flowfocusing junction under microscope. Liquid water is injected through the middle channel and meets the two side air streams at a flow-focusing junction. The microchannels depth is 40μm. 14

15 3.2 Control and Modification of Microchannel Surfaces The wetting behavior of the microchannel walls plays a critical role in the performance of dropletbased microfluidic systems that use a planar architecture. In conventional liquid-in-liquid systems the surface of the microchannels is modified to reduce droplet interactions with the walls. In other words, stable water-in-oil emulsions can be formed in hydrophobic microchannels and on reverse emulsions (oil-in-water) the microchannels should be hydrophilic (oleophobic). Although PDMS is intrinsically hydrophobic, the use of oxygen plasma for bonding renders them hydrophilic which makes liquid handling difficult within microchannels. Moreover, compatibility issues of PDMS surfaces with organic solvents [73], including a variety of the oils used in droplet-based systems, pose additional challenges for droplet generation in liquid-liquid systems. Obtaining the conditions that suit the generation process is usually achieved by manipulating the microchannel surfaces through chemical coatings. A variety of techniques and chemicals are employed to achieve the desired hydrophobicity as well as solvent compatibility [74]. However, the chemicals used for this purpose are typically hazardous to work with. Moreover, applying a reproducible and uniform coating especially in confined microchannels is a challenge from the preparation perspective. Having a gaseous phase as the carrier fluid entirely eliminates the compatibility issue of PDMS in our system. Figure 7A represents the behavior of droplet formation after plasma treatment in this liquid-gas system. Droplets in this case show an affinity to ride on the side walls due to their hydrophilic condition after exposure to plasma. As a result, droplets are not capable of keeping their morphology during formation and tend to spread and form a liquid film on the side walls. To overcome this issue, we implemented two methods for modify the properties of the microchannel walls. In the first approach, we used a fluorosilane to coat the microchannels. Because of the silanol groups existing on the surface of the PDMS, a wide spectrum of silanes can be used for surface modification[74], [75]. We tested Trichloro(1H,1H,2H,2H-perfluorooctyl)silane (FDTS 97%, Sigma-Aldrich) to modify the surface of the PDMS using vapor deposition [76]. Since coating

16 process prior to bonding caused some issues and prevented strong adhesion of the PDMS chips to the glass substrate, we performed the deposition after the bonding step. The process for treating the microchannels inner surfaces involves passing the vaporized FDTS into the bonded PDMS microchannel as shown in Figure 6 [77]. For this purpose, a microdroplet of FDTS is placed on a slide next to the previously-bonded chip, while the outlet of the chip is connected to the vacuum port of a micro diaphragm pump (Parker Corp.) All the parts, including pump, chip, and slide are placed inside a vacuum chamber. By turning on the pump the vaporized silane is directed towards the inlets of the chip and is forced to pass through the microchannels and being deposited on the channel walls. After 1 hour of coating process, the chip was disconnected from the vacuum pump, and used for the droplet generation experiments. Results of droplet formation after applying silane on the walls is shown in Figure 7B. Although, droplet generation after silane coating showed improvements, the results were not consistent due to the challenges associated with the experimental procedure of applying the chemical on the confined structure. Therefore, we tried a different and more robust approach to tackle this issue. Figure 6. Schematic of the method used treating the microchannel walls with fluorosilane chemical. All the setup are placed inside a vacuum chamber in order to prevent the deposition of toxic fluorosilane vapor into the environment. In the second approach, a high-temperature post-baking step was used instead of chemical treatment to achieve the surface conditions suitable for droplet generation in our system. Table 1 shows the measured contact angles for PDMS and glass substrates after the post-bake process. This

17 technique ensures that all the PDMS walls regain their hydrophobic nature in a fast and uniform manner, minimizing the tendency of the droplets to adhere to the walls as shown in Figure 7C. This process is very reproducible and does not involve the challenges of the previous technique. Figure 7. Interaction of liquid and gas in different channel conditions. (A) After plasma treatment; in this situation channels are hydrophilic and as a result the liquid water tends to adhere to the side walls of the microchannel once it enters the junction. (B) After modifying the channel walls with vaporized fluorosilane. In this case, the droplets could form in the channel. However, due to the challenges in the coating process the generation was not reproducible. (C) After post-baking step; here the PDMS has regained its hydrophobic nature and droplets maintain their morphology after generation. Table 1. Measured contact angles for 10 samples of PDMS and glass substrates with the same preparation conditions of plasma treatment followed by a prebake process. The prebake process results in regaining the hydrophobic nature of PDMS. The glass surface also becomes less hydrophilic after the prebake process. Smaller contact angle hysteresis of the glass substrate results in easier droplet movement during formation. Therefore, it will help to obtain a wider Dripping regime and higher monodispersity. Static Angle (θ S) Advancing Angle (θ A) Receding Angle (θ R) Glass 60±4 69±5 52±5 PDMS 102±4 116±4 75±5 In confined microfluidic channels droplet morphology is bounded by the microchannel dimensions. Thus, if the diameter of the generated droplet exceeds the height of the microchannel the droplet will be in contact with the top and bottom walls which result in additional interactions that influence

18 the formation process. This condition occurred throughout the Dripping regime of our system. It has been observed that using a non-hydrophobic glass substrate as the bottom wall enables a more robust and reproducible dripping over a wider range of the flow map with higher monodispersity. We believe this behavior can be related to the smaller contact angle hysteresis of a glass slide in comparison to a PDMS surface. Based on the measurements of the static, advancing, and receding contact angles for the glass and PDMS (see Table 1) surfaces we can schematically represent the droplet inside the microchannel from a side view as shown in Figure 8 for each case. From these measurement two things become apparent. First, post-bake treatment of the glass substrate substantially reduces its hydrophilicity, rendering it partially hydrophilic. Secondly, the droplet holding or restraining force from surface tension effects is smaller for the hybrid (glass substrate- PDMS) structure than for the PDMS-PDMS one, due to the smaller overall contact angle hysteresis effect. This can be quantified by comparing the added top and bottom hysteresis of the hybrid P G structure ((γcosθ R - γcosθ AP ) + (γcosθ R - γcosθ AG )) to the hysteresis from the PDMS-PDMS structure (2(γcosθ RP -γcosθ AP )). This results in a surface tension force for the hybrid structure that is less than half that of the PDMS-PDMS structure. As a result, especially in lower flow rates of the gaseous phase, we speculate that the higher surface tension force from the PDMS-PDMS structure creates an added hindrance to the motion of the droplet while being formed and generated. Therefore, obtaining a reproducible formation process within the flow conditions of the Dripping regime in these structures becomes a challenging task, with the delineation of the boundary between the Dripping regime and its neighboring regimes becoming more blurred.

19 Figure 8. Comparison of the surface tension forces between a glass-pdms microchannel (left) and a PDMS- PDMS microchannel (right). Using the provided contact angles in the table, the net surface tension force for a glass-pdms combination is less. Therefore, it provides less resistance during the course of droplet generation and movement inside the channel which results in a more reproducible formation process. 3.3 Experimental Setup A custom microfluidic test setup has been developed for the experimental investigation of the gasliquid droplet generation. The gas used as the continuous phase in all the experiments is air at room temperature which is supplied from the facility compressed air source. The air is first dried by passing through a desiccator and cleaned using 0.2μm filters (Wilkerson Corp.). Air is regulated in multiple steps before being routed into the microfluidic chip inlets. A manual coarse pressure regulator (SMC corp, up to 50 psi) reduces the pressure of the supply line. A voltage-controlled pressure regulator (Proportion-Air, up to 18±0.04 psi) is used to deliver a precise air pressure into the system. For real-time monitoring of the air flow rate, an in-line mass flow meter (Sierra Smart- Trak2) is placed in the setup. All pressure regulators and flow sensors are powered using dedicated 24V DC power supplies. A data logger/switch unit (Agilent 34970A) is also used for data acquisition which is controlled by a customized LabVIEW interface for controlling the valve and collecting data from the flow meter. A needle valve (Swagelok) is also placed upstream of the chip for a finer flow regulation prior to injection. The air stream bifurcates after passing through the needle valve and is connected to the corresponding inlets of the chip through 1/16-inch PEEK tubes. Liquid flow control is provided by a glass syringe using a high precision syringe pump (PHD2000

20 Harvard Apparatus). The accuracy of the injected flow rates are within 0.35% of the specified value for the flow. Imaging and visualization of the formation process is facilitated using a high-speed CMOS camera (Photron SA5) which is capable of providing a 7500 frames per second (fps) at the resolution of 1024 1024. The camera is mounted to an inverted microscope (Nikon Ti-U) and illumination is provided by a metal halide white light (Prior Scientific). Schematic and actual images of the experimental setup are shown in Figure 9 and Figure 10 respectively. Figure 9. Schematic of the flow circuit for liquid and gas control.

21 Figure 10. Actual image of different parts of the experimental setup. (1) Air desiccator, (2) manual coarse pressure regulator, (3) voltage-controlled valves, (4) mass flow sensors, (5) needle valves, (6) liquid syringe pump, (7) white light source for the microscope, (8) high-speed camera connected to the microscope, and (9) microfluidic chip placed on the microscope with all the ports connected to their corresponding tubing. In order to analyze the population of the generated drops, they are collected off the chip after generation in air, inside a different liquid phase. For this purpose, n-hexadecane (Sigma-Aldrich Corp.) which is almost insoluble in water, is used as the collection medium. In order to prevent droplets from merging inside the solution 2% of the nonionic surfactant span 80 (Sigma-Aldrich Corp.) is mixed with the oil. The outlet of the microfluidic chip is cut and submerged in the oil bath [78]. As a result, the generated droplets are collected downstream of the device outlet once they transition into the oil medium.

22 Chapter 4. Flow Regime Mapping Unlike liquid-liquid droplet microfluidic systems that are usually characterized by the Capillary number (Ca) of the two fluids [24], [79], Ca in our experiments always remained below 0.02. In contrast, the Reynolds number (Re) for the continuous gaseous phase was in the range between 10 and 600 indicating the relevance of inertia in this system. However, such high Re and low Ca for all the flow conditions of the system do not clearly illuminate the fundamental mechanisms behind the different flow regimes. Therefore, we employed the Weber number (We) to parametrize the droplet breakup processes as it changed in a moderate range in our experiments (We Re.Ca). Utilizing We led to a better characterization of the flow regimes presented in this system. The We number for both phases is defined as: We = ρu2 D H σ (4.1) where ρ and σ are the fluid density and interfacial tension respectively. Air is used as the gaseous phase whose density throughout the experiments were calculated considering its compressible behavior inside the microchannels. Distilled water serves as the liquid phase. U is the average fluid velocity inside the channel which is calculated by dividing the flow rate by the cross section area, and D H represents the hydraulic diameter of the rectangular cross section of the microchannel. The We for the gaseous phase (We G) is defined based on the throat cross section at the flow-focusing junction, and for the liquid phase We L is defined based on the liquid channel dimensions. Using the aforementioned definitions, We L is in the range 0.0001-0.3 and We G in the range 0.008-25. These wide range of flow conditions result in various breakup modes which are explained in the following section. Experiments on each chip were performed by initiating a gaseous flow inside the device to prevent the liquid from flooding the microchannels. Liquid water is subsequently injected into the flow-

23 focusing junction. Mapping experiments were conducted by setting a specified flow rate and increasing the gas flow rate. As such, the maps consist of multiple horizontal lines of a fixed flow rate (We L), with transitions occurring as the flow rate (We G) is changed (increased). We conducted several experiments using multiple chips to account for wettability changes associated with the chips being exposed to large amount of water for a long period of time. Over a wide range of flow conditions we observed six distinct regimes which are defined as: Co-Flowing ( ), Threading ( ), Plugging ( ), Dripping ( ), Jetting ( ), and Multi-Satellite Formation ( ). A typical flow regime map depicting these different breakup modes is shown in Figure 11 based on We G and We L. In this map the region associated with each flow regime is distinguished with a different color. The gray areas in the plot represent the transition regions across which the transition between two regimes takes places. We used these finite transition regions rather than infinitesimally thin solid lines to depict the change in the flow regime more realistically. This is more consistent with the observations from experiments since flow regime transition could occur anywhere within the shaded regions from a statistical point of view, which is determined by the microfluidic chips and flow rate measurements in the experiments. However, dashed lines are also included in the flow map as a reference and guideline to distinguish between the actual flow regimes and intermediate transition regions. Furthermore, these lines can be considered as approximate boundaries between each two flow regime regions. At low values of We G, the flow is characterized by the formation of liquid threads that extend along the microchannel outlet. We observed that at low values of liquid flow (for We L<10-2 ), the formed thread does not remain stable along the outlet. In this region which is referred to as the Threading regime, an unstable thread extends downstream of the junction where it breaks up intermittently at different locations and results in multiple plugs and droplets. In the Co-flowing regime, which occurs at higher values of We L, the liquid thread is stable and extends throughout the microchannel outlet. As the value of We L increases, the thickness of this continuous liquid thread increases until

24 it obstructs the microchannel outlet. A periodic breakup takes place at higher We G where each formed thread is broken up into a single plug whose length is bigger than its width (L Droplet>2W Droplet). This corresponds to the Plugging region, with threads that break up very close to the flow-focusing junction at a distance L breakup<10w outlet measured from the junction. The Dripping regime is observed beyond the Plugging regime when the inertia forces start balancing the surface tension forces at the junction (around We G 0.3). In this regime, circular drops are formed at the junction which are highly-reproducible and uniform over a wide range of frequencies and sizes. Increasing We G, however, creates a chaotic condition that involves unsteady and random droplet breakup dynamics. It can be seen from the flow map that a transition from the Dripping to the Multi-Satellite Formation regime occurs around We G 6. Beyond this point a polydisperse spray of the liquid is generated inside the microchannel. The Multi-Satellite Formation regime has not been previously reported in liquid-liquid systems since its appearance is mostly due to the highly inertial nature of the continuous phase which is usually negligible in oil-water systems. We observed that at higher liquid flows a long liquid jet is formed after the junction which breaks into individual droplets due to Rayleigh capillary instability [80]. In this Jetting regime, characterized by droplet tip streaming of an extended liquid jet, droplet generation is less controlled. However, generated droplets are in the order of the liquid jet diameter and much smaller than the droplets from the Dripping regime.

25 Co-Flowing Dripping Threading Multi-Satellite Formation Plugging Jetting Figure 11. Flow regime map of water-in-air droplet formation in a planar flow-focusing microfluidic device. Different flow regions are distinguished with different colors. The transition between the flow regimes is also represented as a finite shaded region rather than a solid line to account for the uncertainties associated with the experiments and calculations of the We values. In addition to the flow conditions, the geometry of the microchannel flow-focusing section plays an important role in the flow map characteristics of this system. We were interested on the influence that liquid microchannel size has on the Dripping regime in particular. Three different chips with liquid channel widths of 25μm, 50μm, and 100μm were considered for this part. Flow mapping experiments were conducted for the wider liquid channels (50μm and 100μm) and the results are compared against the original flow map with the narrower liquid channel (25μm) as shown in

26 Figure 12. For these maps we have only included the approximate dashed lines to better compare the flow regimes layout together. It can be seen that increasing the liquid channel size results in a smaller Dripping region, which is highlighted with the green color in the flow map. Increasing the size of the liquid channel results in larger surface tension forces that hold the droplet during formation. Therefore, higher detaching forces are required to pinch-off the thread at the junction to create discrete droplets. This will shift the transition from the Threading and Plugging regimes to the Dripping regime to higher values of We G. Transition to the Satellite Formation regime is less dependent on the liquid channel size since it is mostly due to the inertial effects of the gaseous phase. According to the flow maps, the Satellite Formation transition typically occurs around We G 6-7 for all the chips. Consequently, increasing the size of the liquid channel would shrink the Dripping region by expanding the Threading and Co-flow region. 25μm 50μm 100μm 0.1 0.1 0.1 WeL 0.01 0.01 0.01 0.001 0.1 1 10 0.001 0.1 1 10 WeG 0.001 0.1 1 10 Figure 12. Comparison of the flow map for chips with different liquid channel sizes. The area for the Dripping region (the green region) is reduced as the liquid channel size increases. However, the Threading and Co-flow regions have expanded as a result of increase in the surface tensions forces that holds the thread and prevents the detachment at lower gas flow rates.

27 Chapter 5. Dripping Regime Characterization 5.1 Generation Frequency In this chapter, three important aspects of the Dripping regime are investigated by characterizing the generation frequency, morphology of the droplets inside the microchannel, and monodispersity of the generated droplets within this region. The frequency of droplet formation (ƒ Generation) is obtained by manually counting the number of droplets generated over a given time interval from the high-speed videos. To do this, the sequential images taken from the high speed camera were analyzed in a lower speed to count the generated droplets with naked eye. Figure 13 and Figure 14 show the frequency data for the Dripping regime at different liquid and gas flow rates. It can be seen that the frequency varies between less than 30 up to 1000 Hz (drops per second) over the entire Dripping regime. Results show that frequency of the generation increases as either the liquid or gas flows increase. Moreover, experimental plots show that frequency values are proportionate with both flow rates. Therefore, we scaled the frequency in this system as the product of the two flow rates (i.e. ƒ Q GQ L). In order to non-dimensionalize the frequency results, flow rates are nondimensionalized by characteristic inertial velocities (U * =(σ/ρd H) 1/2 ) obtained from a Weber number of unity for both phases, and a characteristic timescale (τ * =D H/U * ). Therefore, generation frequency can be non-dimensionalized as ƒ(τ * Gτ * L) 1/2. The results of this scaling are shown in Figure 15.

ƒ Generation (Hz) ƒ Generation (Hz) 28 1200 A 800 400 0 0 10 20 Q G (ml/min) Figure 13. Experimental data of droplet generation frequency as a function of air flow rate for different liquid flow rates. We can see that there is a relatively linear correlation between the frequency value and the air flow rate. 1200 B 800 400 0 0 20 40 Q L (μl/min) Figure 14. Experimental data of droplet generation frequency at increasing liquid flow rate for two different gas flow rates. We can see that there is a relatively linear correlation between the frequency value and the liquid flow rate.

29 0.01 ƒ(τ * Gτ * L) 1/2 0.001 0.0001 0.01 0.1 1 Q G /(U * GD 2 H,G) Q L /(U * LD 2 H,L) Figure 15. Experimental data of the non-dimensioalized droplet generation frequency. Generation frequency in this system can be scaled as ƒ Q GQ L. 5.2 Droplet Morphology Images of the droplets moving inside the microchannel are captured and analyzed to determine their corresponding length and width. Experiments are conducted multiple times for each flow condition to verify the reproducibility of the measured values. In the presence of a continuous gas flow we always find a finite gap between the droplets and the walls within the Dripping regime. Therefore, both the width and length of the droplets change as a function of flow rate. Figure 5 shows the droplet length and width as a function of gas flow rate (Q G), for different liquid flow rates (Q L). Snapshots of the droplets inside the microchannel for different flow conditions shown in Figure 18 clearly demonstrate these changes in length and width. It can be seen that a higher air flow rate results in decreasing both length and width. However, liquid flow rate does not have an apparent effect on the droplet width. Moreover, at high liquid flow rates, droplets exhibit a long tail

L Droplet (μm) W Droplet (μm) 30 which deviated from their common circular morphology. In order to take into account this length for later scaling analysis, we used a correct length for this purpose as shown in Figure 19. 180 130 80 5 10 15 20 Q G (ml/min) Figure 16. Experimental data of droplet width inside the microchannel for different liquid and gas flows. 200 180 160 140 QL=1 μl/min QL=5 μl/min QL=10 μl/min QL=20 μl/min QL=30 μl/min QL= 40 μl/min 120 100 80 5 10 15 20 Q G (ml/min) Figure 17. Experimental data of droplet length inside the microchannel for different liquid and gas flows.

31 QL=5μL/min QG=8 ml/min QG=10 ml/min QG=12 ml/min QG=14 ml/min QG=16 ml/min 100μm Figure 18. Snapshots of droplets moving inside microchannel under different gas flow rates. Figure 19. Corrected length used in non-dimensionalizing droplet morphology data In high liquid flows, droplet exhibit a tail-shape which deviate from their common circular morpholy.

32 During the course of the generation process, a droplet starts forming from the liquid microchannel in the flow-focusing junction. As the liquid is injected into the droplet volume, the droplet grows in the plane at the junction. The droplet is attached to the liquid microchannel by a surface tension force that can be scaled as, Fσ ~ σd H,L (5.1) where D H,L is the hydraulic diameter of the liquid microchannel. As the droplet protrudes into the junction, it experiences inertial forces of the high speed gaseous medium that co-flows with the liquid droplet. The inertial force of the gaseous phase can be scaled as; F I ~ ρq 2 G 4 h(w*-w L ) cos θ (5.2) D H,G where D H,G is the hydraulic diameter of the gas microchannel, h is the droplet height which is equal to the channel height, θ is the angle at which the air enters the junction, W L is the liquid microchannel width, and W* is the droplet width during the formation stage (see Figure 20). When the aforementioned forces balance, the droplet width can be approximated as 4 W*~σD H,G /ρq 2 G hcosθ+w L. However, the morphology of the droplets was assessed from the images of the droplets moving inside the microchannel outlet after the droplets are detached. As such, a similar scaling dependence can be used for the measured width with respect to the inertial force but with a different proportionality factor and minimum width. To calculate this minimum width (W), we consider the limiting case of W * W L which occurs experimentally at high gas flow rates and also from the scaling equation. In this case, the prefilled droplet volume in the junction forms a rectangular liquid ligament whose width is in the order of the liquid microchannel width and the length is stretched to the throat of the junction ( Z). This volume results in minimum droplet width of 86μm after detachment, assuming a circular morphology for these droplets at high gas flow rates, which is a reasonable assumption based on the experimental data. Since the complete generation process always starts with the prefilling process, this obtained droplet size can be considered the

33 minimum width and length that can be achieved within the Dripping regime. Therefore, droplet width can be calculated as W=C 1 σ ρq G 2 +C 2, where C 1 takes into account the geometrical parameters and a fitting factor, and C 2 is the minimum droplet width that can be achieved in the Dripping regime of this system. The volume of the droplet in the dripping regime can be related to the generation frequency and the liquid flow rate via conservation of mass of the liquid flow, V droplet = Q L ƒ (5.3) If the droplet is assumed to have a relatively uniform width over its entire length, the volume of the droplet is proportional to the product of its dimensions. Therefore, the volume can be approximated as V droplet LWh, where L, W, and h are the droplet length, width, and height, respectively. In order to be consistent with the assumed (and mostly observed) droplet volume, experimental data for the measured length were corrected in some cases where the droplet contained a morphology with nonuniform width. By substituting the aforementioned scaling for the frequency (ƒ Q GQ L), droplet length can be scaled as L~1/hWQ G. Figure 24 shows the normalized droplet morphology using the minimum values obtained in the Dripping regime as a function of the We G. It can been seen that both droplet length and width decrease with increasing gas flow. However, the effect of liquid flow rate is less discernable in the morphology and more pronounced in the formation frequency.

34 Figure 20. Schematic of the Flow-Focusing geometry and the nomenclature used in scaling analysis for droplet morphology. W/W min 3 2 1 Scaling QL=1 QL=5 QL=10 QL=20 QL=30 QL=40 0 0 5 10 We G Figure 21. Droplet length normalized with the minimum width of the droplets in the Dripping regime as a function of gas Weber number (We G). Minimum droplet width and length are assumed to be equal values that are obtained based on the minimum volume of the prefilled droplet during the formation process at high gas flow rates where droplet tends to have a more circular morphology. This minimum is dependent on the geometry of the flow-focusing microchannel and is calculated to be about 86μm for the geometry used in out experiments.

35 L/L min 3 2 1 Scaling QL=1 QL=5 QL=10 QL=20 QL=30 QL=40 0 0 5 10 We G Figure 22. Droplet length normalized with the minimum length of the droplets in the Dripping regime as a function of gas Weber number (We G). Minimum droplet width and length are assumed to be equal values that are obtained based on the minimum volume of the prefilled droplet during the formation process at high gas flow rates where droplet tends to have a more circular morphology. This minimum is dependent on the geometry of the flow-focusing microchannel and is calculated to be about 86μm for the geometry used in out experiments. 5.3 Droplet Monodispersity Generated droplets are collected off the chip in a glass vial filled with a different liquid which is immiscible with the water droplets. Captured images from the collected droplets were processed using a customized MATLAB program to analyze the size of the drops and their distribution as shown in Figure 23. Since the droplets are spherical after transitioning into the liquid medium, we use the droplets diameter as the characteristic length scale in this part. Comparison between droplets in the Dripping and Multi-Satellite Formation regions is shown in Figure 24. Size uniformity of the collected droplets within the Dripping regime demonstrates the controlled generation of the droplets within this region. The generated droplets have a polydispersity index

36 (PDI) defined as σ/d (where σ represents the standard deviation of the droplet diameters and d is the average diameter of the generated drops), of less than 5%. Droplets generated in the Dripping regime can form a uniform closely-packed structure inside the liquid medium as shown in the figure. Satellite Formation regime, however, results in polydisperse droplets which are distributed over a wide range of diameters even for a fixed flow condition. Figure 23. (Left) Experimental images of the collected droplets in the immiscible liquid solution. (Right) detected droplets using image analysis with MATLAB.

37 Dripping regime Multi- Satellite Formation regime Figure 24. Actual images and size distributions of samples of 500 collected droplets within the Dripping and Multi-Satellite Formation regimes. Liquid flow rate for both cases is 5µL/min. Multi-Satellite Formation is obtained at higher We G according to the flow map. In the Dripping region by having a constant liquid and gas flow rate the generated droplets are uniform in size. In Multi-Satellite Formation regime, however, small daughter droplets are generated beside the main droplet which results in polydisperse distribution of the droplets.