Passive double-emulsion generation
As the first step in ATPS droplet formation, the generation of water-water-oil double emulsions determines the final droplet size and morphology. To generate monodisperse double-emulsion drops, we passively break up the aqueous flow into droplets by oil pinching, which can be achieved in a typical multi-inlet flow-focusing channel23. Different from the water-oil-water (W-O-W) double-emulsion drops that are formed by two steps of emulsification24, water-water-oil (W-W-O) double-emulsion drops are formed in a single step at the second cross junction: The two aqueous phases flow laminarly side-by-side after meeting in the first junction, followed by pinch-off of the water–water compound jet by the oil flow to form W-W-O double-emulsion drops (Fig. 1a and Movie S1). To demonstrate the efficiency of this method, the two most commonly used ATPS polymers, polyethylene glycol (PEG) and dextran (DEX) that are both biocompatible, are introduced. To prevent coalescence and destabilization of the droplets during subsequent droplet extraction, two oppositely charged polyelectrolytes, poly(allylamine hydrochloride) (PAH) and poly(sodium 4-styrenesulfonate) (PSS), are dissolved into the innermost (DEX-rich) and intermediate (PEG-rich) solutions to stabilize the water-water interface through interfacial complexation. However, due to the fast reaction, premature precipitation prior to droplet breakup tends to gradually block the channel and cause a transition from the original dripping regime to an unstable jetting regime within 3 min. Thus, a spacing stream of pure PEG-rich solution is introduced at a flow rate of 100 μL/h to separate the two reagent streams, thereby enabling stable droplet generation for over 1 h and forming monodisperse double emulsions (Fig. 1b, c).
To optimize the microfluidic operations for controllable double-emulsion generation, the droplet size and generation frequency at different flow conditions are systematically studied. In this flow regime, the two streams of PEG-rich phases with and without PSS converge into one stream before droplet generation, and varying the flow rate ratio between the two PEG-rich phases does not lead to any observable effect on the resultant droplet size (Fig. S1). Thus, in the following analysis, we use the combined flow rate to characterize the flow behaviors. First, by regulating the flow rates of the aqueous phases, the formed core-shell structure can be controlled. We change the flow rate ratio of the DEX-rich phase and PEG-rich phase from 1:7 to 7:1 while keeping the oil flow rate constant. As a result, the inner droplet diameter increases while the outer droplet diameter remains unchanged, with the diameter ratio ranging from 50 to 96% and the resultant shell thickness varying from 26 μm to ~2 μm (Fig. 1d). The flow rate of oil is a key parameter that affects both the double-emulsion drop size and generation frequency. To quantify the effect of flow rate on the double-emulsion formation, we set the flow rates of DEX-rich and PEG-rich phases constant at 200 μL/h and 600 μL/h, respectively, while tuning the flow rate of oil from 2000 to 10,000 μL/h. For oil flow rates below 2000 μL/h, the aqueous compound jet cannot be pinched off. Experimentally, the core diameter gradually drops from 73 μm to ~43 μm as the oil flow rate increases, and the generation frequency increases from 270 Hz to ~1200 Hz (Fig. 1e). Since the two aqueous phases are formed by phase separation of PEG and DEX solutions, the adopted polymer concentrations determine the interfacial tension and viscosity of the two aqueous phases, thus affecting the interface pinch-off and droplet breakup. To understand the effect of polymer concentrations and test the feasibility of this approach in more viscous flows, we increase the PEG and DEX concentrations from 10 to 20% and measure the resultant core size and generation frequency at constant flow conditions (200 μL/h for the DEX-rich phase, 600 μL/h for the PEG-rich phase and 3000 μL/h for the oil phase). The results indicate that as the polymer concentration increases, the size of the core droplets increases from 65 μm to 77 μm, and the frequency decreases from 385 to 240 Hz (Tables 1 and 2). In summary, this method enables high-throughput generation of W-W-O double emulsions with controlled structures.
Active double-emulsion generation
In addition to the demonstrated passive generation, double-emulsion drops can be actively generated by injecting the core droplet into the shell droplet one-by-one by pico-injection. Here, a typical pico-injection channel is employed for the two-step generation25. In the flow-focusing junction, the PEG-rich solution is broken up into droplets under the shear effect of the oil stream. Then, when the droplets flow by the pico-injector, a high voltage is applied to the electrodes to destabilize the water-oil interface, thus introducing the DEX-rich solution into the shell drop to form a double emulsion (Fig. 2a). This method relies on accurate control over the applied electric field, where an optimal pulse with a frequency of 10 kHz and peak-to-peak voltage of 500 Vp-p is applied to achieve a stable injection with identical volumes injected into each droplet25.
In comparison with the passive generation that simultaneously pinches off two aqueous streams, only one stream of the PEG-rich phase is segmented in the active operations, thus allowing more frequent double-emulsion generation. To demonstrate the higher efficiency, a similar test is performed to explore the effect of the oil flow rate on the double-emulsion generation. We change the flow rate of the oil phase from 500 to 10000 μL/h, while the flow rates of the droplet phase (PEG-rich) and injection phase (DEX-rich) are maintained at 600 and 200 μL/h, respectively. The results indicate that under the same flow conditions, smaller droplets can be generated at a higher frequency in the active mode. For instance, at the same oil flow rate of 3000 μL/h, the frequency of active generation (750 Hz) is approximately two times that of passive generation (380 Hz) (Fig. 2b). Moreover, double-emulsion formation at different polymer concentrations is studied. We set the flow rates of DEX-rich, PEG-rich and oil phases at 200, 600 and 3000 μL/h, respectively. Experimentally, when the concentrations of the two polymers increase from 10 to 20%, the resultant core size increases from 52 to 61 μm, and the generation frequency decreases from 750 to 480 Hz (Table 3).
In addition to the higher efficiency, the active generation method allows injection of different ingredients into the core droplets, thereby enabling time-controlled reactions within the droplets. We achieve this aim in a dual-injector device, where the two injectors are closely located at a spacing of 500 μm to efficiently merge the injected cores before the assembly of polyelectrolytes to form a membrane at the interface. To demonstrate the efficiency of this approach, a fluorescently labeled DEX-rich solution is introduced from the first injector, while a conventional solution without fluorescent dye is introduced from the second injector (Fig. 2c). As the flow rate ratio of the two injection flows is tuned from 1:7 to 7:1, double emulsions with different fluorescence intensities can be formed, achieving good mixing to allow accurate control of the fluorescent polyelectrolyte concentration (Fig. 2d).
To extract the encapsulated core droplets for ATPS droplet formation, the carrier oil must be removed to break the double-emulsion drops. Typically, the droplets are broken by first adding the demulsifier to destabilize the surfactant layer and then centrifuging the samples to separate the water and oil phases based on their density difference26. However, this method involves severe flow disturbance, which tends to disrupt both the water-oil interface and water-water interface, thus inducing undesirable coalescence of the ATPS droplets. Therefore, gentler extraction should be employed to break the outer droplet interface while keeping the inner droplets intact. Here, we introduce a noninvasive method that hinges on evaporation of the oil phase to break the double emulsions. This method is performed by transferring the collected emulsions into a Petri dish filled with the PEG-rich solution. When the emulsions fall onto the solution, the interfacial tension keeps the oil phase on the top, with the double-emulsion drops aggregating at the center. As the surrounding oil spreads and forms a thin layer, its surface area increases. Consequently, the oil layer evaporates in ~20 s, and the shell phase merges with the PEG-rich solution underneath; hence, the cores of the DEX-rich phase are released to form monodisperse DEX-in-PEG droplets (Fig. 3a, b).
In this method, the surfactant concentration is a key factor in implementing core-droplet extraction. To elucidate the effect of the surfactant concentration on the extraction efficiency, droplets generated using different concentrations of surfactant, ranging from 0.005 to 1%, are tested. The results indicate that when the surfactant concentration is below 0.02%, droplet coalescence occurs due to the unstable water-oil interface; when the surfactant concentration is above 0.2%, the droplets remain separated from the PEG-rich solution after complete oil evaporation (Fig. 3c, d). Thus, the surfactant concentration should be kept at appropriate ranges to stabilize the interface and release the core droplets. Using 0.05% surfactant as a demonstration, the core droplets can be efficiently extracted within 40 s. Moreover, with this evaporation-driven extraction, the inner core droplets are free from external disturbance and can maintain their uniform shapes and sizes (coefficient of variation of approximately 2%).
Single-cell-laden microgel sorting
Droplet-based hydrogel microspheres, which can provide both mechanical stability and cell-binding properties, are promising scaffolds for 3D cell culture and single-cell analysis27,28. However, the conventional fabrication method based on a water-oil system requires strong mechanical and chemical treatment to extract the formed hydrogels from organic solvents, possibly resulting in loss of cell viability29. In comparison, the demonstrated ATPS droplet facilitates biocompatible microgel fabrication in an all-aqueous environment. We achieve this aim by incorporating alginate, a biocompatible polymer that can be physically crosslinked with divalent ions30, into the DEX-rich phase. Through collecting the double emulsions into a calcium chloride (CaCl2) solution bath, the released core droplets can be rapidly crosslinked to form spherical and uniform-sized hydrogels (Fig. 4a, b).
Moreover, the transient use of the oil phase enables inclusion of a droplet sorting step after double-emulsion generation and before extraction of the core droplets to produce single-cell-laden hydrogels, which is difficult to achieve in other ATPS droplet generation approaches31. We demonstrate this process in an integrated channel composed of both generation and sorting parts (Fig. 4c). Here, a spacing oil flow is introduced to further gap the continuously generated droplets for accurate sorting, and the two branch channels are designed with precise differences in hydraulic resistances. As a result, the droplets are guided into the lower channel when no electric field is applied and are driven into the upper channel when an electric field is applied (Fig. 4d). Through setting the flow rates of the core, shell, generation and spacing oil phases at 200, 600, 3000 and 3000 μL/h, respectively, high-throughput sorting can be achieved at frequencies above 300 Hz. To demonstrate the sorting-based fabrication of single-cell-laden microgels, the average cell number in each droplet is tuned to 0.1 so that only approximately 10% of the generated droplets contain single cells following the Poisson distribution. With a sorting step to remove the negative droplets, the single-cell encapsulation ratio increases to over 90%, thus allowing a high-throughput analysis of the encapsulated single cells (Fig. 4e, f). Moreover, to verify the biocompatibility of this technique, a live/dead assay is performed on the encapsulated cells and shows high viability, with over 95% of the cells remaining alive. This result suggests that the encapsulated cells can be further cultured for tissue development and engineering applications.