8.2.1 Droplet Generation

Currently, most planar microfluidic droplet generators are designed using T-junction [14,15] and flow focusing [16,17] geometries, in which small droplets are spontaneously formed at an intersection taking advantage of the interface instability between oil and aqueous streams. Using these approaches, droplets can be generated over a broad range of frequencies ranging from ~0.1 Hz to 10 kHz and using flow rates on the order of 0.1–100 μL/min [29]. Droplet volume and generation frequency depend on several factors, including the physical properties of the immiscible phases, flow rates, intersection geometry, and so on. For a given geometry and solvent composition, flow focusing and T-junction interfaces exhibit interdependence between flow rate and droplet generation frequency and cannot be easily modulated over short timescales.

For lower frequencies and applications for which the ability to rapidly change droplet size and generation frequency is desirable, on-demand droplet generation strategies become more favorable, as they ensure precise control and fine manipulation of individual droplets. Various approaches have been developed to generate droplets on demand, for example, by carefully balancing the pressure and flow in the system [27], as well as electrical [30] or laser pulsing [31] and piezoelectric actuation [32]. Pneumatic valving has also been explored and has been found to provide easy, independent control over both droplet size and generation frequency [33,34,35 and 36]. Galas et al. utilized a single pneumatic valve that was embedded in an active connector and assembled close to a T-junction to regulate the flow of the dispersed phase [33]. Constant pressures were applied on the inlets of two immiscible liquids to drive the flow in the microchannel. Individual droplets were created by briefly opening the valve. The aqueous droplet size depended on the valve actuation time and frequency, as well as the pressure applied at the oil inlet. Therefore, the droplet volume, spacing, and speed could be controlled accurately and independently. This device not only generated periodic sequences of identical droplets, but also enabled the production of nonperiodic droplet trains with different droplet sizes or spacing. Lin and coworkers also reported a similar platform for pneumatic valve-assisted on-demand droplet generation [34]. Negative pressure was applied at the outlet of the device to drive the flow of the two immiscible liquids through the microchannel. The dependence of droplet size on the valve actuation time and applied pressure was investigated. In addition, they utilized several aqueous flow channels, each with independently controlled microvalves, to generate arrays of droplets containing different compositions by alternately actuating the valves.

We have also investigated valve-controlled on-demand droplet generation. To minimize the dead volume and control droplet volume precisely, the pneumatic valve was placed over the side channel exactly at the T-junction (Figure 8.1a). Carrier oil flow was driven by a syringe pump, and the dispersed aqueous phase was injected by finely controlled air pressure. Figure 8.1a shows the generation process of an individual droplet. The valve is initially closed and the aqueous fluorescein solution is confined in the side channel. When the valve is opened briefly, a small volume of aqueous solution is dispensed into the oil channel to form a droplet, which is then flushed downstream by the carrier oil flow. Compared with conventional microfluidic droplet generation techniques based on a T-junction or flow focusing, the valve-integrated system can generate droplets with precise control over droplet volume, generation frequency, and velocity. Droplet velocity is determined by the syringe pump driving the oil stream, while the droplet generation rate is controlled by the valve actuation frequency as defined in the software. The droplet spacing is determined by the interval between valve openings and the oil flow velocity. Droplet volume depends on several parameters including the valve actuation time, the pressure of the aqueous solution, the oil phase flow rate, and the valve control pressure (Figure 8.2).

8.2.2 In-Droplet Reagent Combination and Mixing

Besides controlled droplet generation, droplet fusion is of crucial importance in relation to the development of microreactors because it allows precise and reproducible mixing of reagents at well-defined points to initiate, modify, and terminate reactions [6]. Ismagilov and coworkers carried out the pioneering work to combine different reagents into individual droplets by causing two reagent solutions to flow in a micro-channel as two laminar streams [37]. To prevent the prior contact of two reagents before droplet generation, an inert center stream was used to separate them. Thus, three streams were continuously injected into an immiscible carrier oil phase to form droplets. The gradient droplets in the reagent concentrations were achieved by varying the relative flow rates of the three streams [11,38]. A subsequent winding channel was designed to accelerate mixing by chaotic advection [20,37]. This approach has been widely employed to control networks of chemical reactions [37], study reaction kinetics [11], screen protein crystallization conditions [38], and investigate single-cell-based enzyme assays [39] and protein expression [12].

Recently, Weitz and colleagues presented a robust picoinjector to add reagents to droplets in microfluidic systems [40]. The picoinjector was controlled by an electric field to trigger the injection of a controlled volume of reagents into each droplet. The injection volume was precisely controlled by adjusting the droplet velocity and injection pressure. Selective injection was realized by switching the electric field on and off at kilohertz frequencies.

In-channel droplet fusion is another attractive approach to combine different reagents in individual droplets to initiate or terminate the confined reactions. The process of droplet merging introduces convective flows into the system, resulting in far more rapid mixing than relying on diffusion alone [41]. In-channel droplet fusion is readily achieved by bringing two or more surfactant-free droplets into contact. Both passive and active methods have been developed to control droplet fusion. For passive fusion devices, droplet coalescence is usually initiated by utilizing specially designed fusion elements in the channel. For example, Bremond et al. incorporated an expanded coalescence chamber in the channel network in which two droplets were brought into close proximity and merged together before they entered a narrow channel [42]. Fidalgo et al. reported a method for droplet fusion based on a surface energy pattern inside a microfluidic channel where the segmented flow was disrupted and the droplets were trapped and fused together [43]. In this case, full control of droplet fusion could be achieved by varying channel and pattern dimensions, as well as the fluid flow. This surface-induced droplet fusion method enabled the merging of multiple droplets containing different reagents to form a large droplet. However, this approach could potentially cause cross-contamination between droplets from the patterned surface. Niu et al. developed a pillar-induced droplet merging device, in which rows of pillars were constructed in the channel network serving as passive fusion elements or chambers [44]. The pillar array trapped droplets and drained the carrier oil phase through the apertures between pillars. The first trapped droplet was suspended and merged with succeeding droplets until the surface tension was overwhelmed by the hydraulic pressure. The merging process depended on the droplet size, and the number of droplets that could be merged relied on the mass flow rate and volume ratio between the droplets and merging chamber.

Active fusion methods that can be controlled externally and selectively have also been developed using, for example, electric fields [45,46,47 and 48] and laser pulses [49] to trigger coalescence. To perform active droplet fusion effectively, synchronization of droplets is a key factor because fusion efficiency relies on the droplets being in very close proximity [50]. Currently, special designs are often employed to synchronize droplets in two parallel channels, which then merge into a single channel downstream to realize droplet coalescence [34,42,51]. However, this system can potentially be disturbed by a few factors such as flow rate and back pressure in the channel, which may reduce the fusion efficiency. Recently, Jambovane et al. used valve-based droplet generation for multiple reagents to perform controlled reactions and establish chemical gradients among arrays of droplets [52]. Droplets were generated at a valve-controlled side channel and then different reagents were added to the droplets as they passed by similar side channels downstream.

An efficient method for reagent combination that we have recently developed employed two pneumatic valves integrated at a double-T intersection (Figure 8.3a). Reagents were introduced through the different side channels, each controlled by a separate valve, and simultaneous opening of the valves resulted in the creation of an aqueous plug containing both reagents. Upon actuation, the oil between the two side channels is quickly displaced, and the two aqueous streams collide and combine. No sample cross-contamination was observed because of the applied pressures, the rapid valve actuation, and the offset between the two side channels. The two liquids mixed together by a combination of diffusion and convection caused by an equalization of internal pressures following the combining of the aqueous streams. The linear dependence of droplet volume on valve actuation time and the independent valving for the two aqueous streams provide a high degree of control over droplet composition. Figure 8.3b shows arrays of six droplets containing different ratios of two dyes that were created by controlling the operation of two valves. Such control should be useful for optimizing or screening reactions and for studying reaction kinetics.

8.2.3 Droplet Incubation

Many biological assays involving, for example, enzymatic reactions, have relatively slow kinetics, requiring microdroplets to be incubated for minutes to hours for efficient reaction. Similarly, studies involving cell incubation or protein expression require extended incubations. A straightforward method for microdroplet incubation is to simply increase the channel length following droplet generation [53,54], but increased back pressure and disruption of droplet formation can quickly become an issue. Frenz et al. incorporated deeper and wider delay lines following the droplet generation section, which enabled the length of times for reactions in the droplets to increase from 1 min to >1 h [21]. Similarly, Kennedy and coworkers have interfaced capillaries or Teflon tubing with the droplet generation devices to collect and store sample plugs for 1–3 h [55]. For longer-term online incubation, droplets can be stored in reservoirs, traps, or dropspot arrays. For example, Courtois et al. fabricated a large reservoir for storing droplets for periods of up to 20 h to study the retention of small molecules in droplets [23]. Huebner et al. designed a droplet trapping array to store and incubate picoliter-sized droplets for extended periods of time to investigate the encapsulated cells and enzymatic reactions [25]. Weitz and colleagues introduced a “Dropspots” device to immobilize and store thousands of individual droplets in a round chamber array over a 15 h incubation period [56]. Droplets can also be incubated off-chip for periods ranging from minutes to several days when appropriate surfactants are used to stabilize the droplets [57,58]. The incubated droplets can then be reinjected into microfluidic devices for further processing and detection.

8.2.4 Droplet Readout Strategies

To date, in-droplet fluorescence detection remains the most widely used method for analyzing the contents of droplets due to its ability to measure in real time and with high sensitivity. Fluorescence detection has been implemented to study enzyme kinetics within droplets [11,59,60], characterize the behavior of encapsulated single cells [12,27], detect PCR products [61,62], and investigate the interactions between biological samples [63]. Fluorescence detection is ideally suited to rapid, sensitive detection of a small number of distinct species. For cases in which a large number of analytes need to be detected and identified (e.g., proteomics and metabolomics) and where fluorescent labeling is not desirable, alternative measurement strategies are needed.

In-droplet Raman spectroscopy has recently been used to detect and analyze droplet contents [64,65]. It is a nondestructive and label-free detection approach with high molecular selectivity which can track the droplets in real time to determine fundamental droplet properties and chemical contents, including droplet sizes, encapsulated species, structures, and concentrations. Surface-enhanced Raman spectroscopy (SERS) can offer higher sensitivity and reproducible quantitative analysis of the droplets due to the enhancement of the Raman signal intensity [66]. Electrochemical detection is an inexpensive and label-free approach to collect information on the physical and chemical properties of droplets and can monitor droplet production and measure droplet length, frequency, and velocity [67]. It can provide chemical information when the reaction within the droplets involves an electrochemically active reactant or product [68]. Another advantage of electrochemical measurement is its compatibility with alternative chip materials, including opaque substrates, which are difficult to implement for conventional optical detection strategies such as fluorescence. Nuclear magnetic resonance (NMR) has also been used for droplets or segmented flow analysis. Karger and coworkers developed a microcoil NMR probe for high-throughput analysis of sample plugs in dimethyl sulfoxide (DMSO) [69].

While the above detection strategies can be employed in situ, others require the contents to be removed from the droplet for subsequent analysis. Once extracted to an aqueous stream, the droplet contents can be analyzed using more information-rich techniques including LC, CE, and MS. MS is an especially attractive technique for in-depth, label-free biological analysis because of its ability to identify and provide structural information for hundreds or more unique species in a given analysis [70]. Below we detail methods used for droplet extraction and subsequent analysis.

Ismagilov and colleagues used a microfluidic system to screen and optimize organic reaction conditions in microdroplets detected using matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) [71]. The incubated reaction plugs were deposited onto a sample plate for MALDI-MS analysis. Kennedy and coworkers directly pumped nanoliter plugs of sample into a mass spectrometer for analysis through a metal-coated capillary nanospray emitter, separating the analyte from the carrier at the emitter itself [72,73]. Teflon tubing was placed close to the emitter tip to siphon the accumulated oil away from the tip, and could maintain stable electrospray at flow rates as high as 2000 nL/min. However, it is generally necessary to extract the aqueous droplet from the oil phase for further separation or on-line MS analysis to avoid contamination of the mass spectra with peaks from the oil and to maintain the electrospray Taylor cone in the most efficient cone-jet mode of operation.

Edgar et al. first reported the extraction of aqueous droplet contents into a channel for CE separation [74]. A femtoliter-volume aqueous droplet was directly delivered to fuse with the aqueous phase in the separation channel for CE separation. Niu et al. employed a similar method to inject the droplets in which the LC eluent was fractionated into a CE channel for comprehensive two-dimensional separations in both time and space [75]. A pillar array was constructed at the interface to evacuate the carrier oil phase prior to loading samples into the separation channel. In these two cases, it was very difficult to maintain a robust extraction because the segmented flow was perpendicular to the CE separation channel.

Kennedy and colleagues exploited a surface modification method to form a stable interface at the junction between two immiscible phases in the microchannel [76,77 and 78]. They selectively patterned glass surfaces in the segmented flow channel to be hydrophobic in order to stabilize the oil–water interface and facilitate droplet extraction. But in some cases only part of each droplet was extracted due to the presence of a “virtual wall,” and was not suitable for quantitative analysis because of irreproducibility and loss of information [76]. Fang and coworkers employed a similar surface modification technique to obtain a hydrophilic tongue-based droplet extraction interface that could control the droplet extraction by regulating the waste reservoir height [79]. The extracted droplet contents were then detected by MS through an integrated ESI emitter. More recently, Filla et al. used a corona treatment to hydrophilize a portion of a polydimethylsiloxane (PDMS) chip to establish an extraction interface [80]. Aqueous droplets were transferred into the hydrophilic channel when the segmented flow encountered the interface. The droplet contents were subsequently analyzed by electrochemistry or microchip-based electrophoresis with electrochemical detection.

Huck and coworkers employed electrocoalescence to control droplet extraction [81,82]. The segmented flow and continuous aqueous flow met at a rectangular-shaped chamber where an interface between two immiscible phases was built up. A pulsed electric field was applied over the chamber to force droplets to coalesce with the continuous aqueous stream, which then delivered the droplet contents to a capillary emitter for ESI-MS detection [82]. This droplet extraction approach required careful adjustments of the flow of the two immiscible phases to maintain a stable interface in the extraction chamber and avoid cross-contamination of the aqueous and oil streams. In addition, the severe dilution of the droplet contents resulted in high detection limits (~500 μM bradykinin). Lin and coworkers used an electricity-based method to control the droplet breaking and extraction at the stable oil–water interface [83]. One reported issue in this case was the difficulty of achieving complete extraction with high efficiency, which limited its compatibility with quantitative analysis.

Kelly et al. invented a droplet extraction interface that was constructed with an array of cylindrical posts to separate the segmented flow channel and the continuous aqueous phase channel [84]. When the aqueous stream and carrier oil phase flow rates were well controlled to balance the pressure at the junction, a stable oil–aqueous interface based on interfacial tension alone was formed to prevent bulk crossover of the two immiscible streams. The droplets could be transferred through the apertures to the continuous aqueous stream and finally detected by ESI-MS with virtually no dilution, enabling nanomolar detection limits.

Most of the reported methods and techniques for droplet extraction, as mentioned above, need to adjust two immiscible liquid flow rates to stabilize the interface and extract entire droplets. It is desirable to perform effective and complete droplet extraction independent of the flow rates, which would provide added flexibility for device operation. Recently, we have developed a robust interface for reliable and efficient droplet extraction, which was integrated in a droplet-based PDMS microfluidic assembly. The droplet extraction interface consisted of an array of cylindrical posts (Figure 8.4a), the same as was previously reported [84], but the aqueous stream microchannel surface was selectively treated by corona discharge to be hydrophilic. The combination of different surface energies and small flow-through apertures (~3 × 25 μm) enabled a very stable liquid interface between two immiscible steams to be established over a broad range of aqueous and oil flow rates. All aqueous droplets were entirely transferred to the aqueous stream (Figure 8.4b) and detected by MS following ionization at a monolithically integrated nanoelectro-spray emitter.

8.3 Perspectives for Droplet-Based Microfluidics

As mentioned above, droplet-based microfluidics have been employed for a wide range of analyses and due to their unique advantages, their use will undoubtedly grow. The following subsections outline a few promising applications that will leverage the strengths of the platform.