Figure 1a shows a schematic illustration of our open-channel, droplet-based microfluidic platform, which enables the individual control of droplet motions including moving, merging, mixing and analysis via the simple formation of a local dimple structure. PDMS was used as a substrate due to its excellent flexibility, superior chemical stability, and biocompatibility27. A wafer-scale, thin PDMS substrate with micro-pillar arrays was successfully fabricated through a single moulding step (Supplementary Fig. 1). Figure 1b illustrates a typical scanning electron microscope (SEM) image of the regular PDMS micro-pillar arrays with 4.5 μm periods. The micro-pillar arrays that were designed with round tips could effectively reduce the contact area between the substrate and the water droplets, which induced superhydrophobicity with an extremely large WCA of 151 ± 3° (Fig. 1c,d). Owing to the thin (275 μm thick) and flexible nature of the PDMS substrate, the fabricated substrate with micro-pillar array could provide excellent reversible stretchability without residual distortion (Fig. 1e).

To manipulate individual water droplet motions on the superhydrophobic PDMS substrate, a millimeter-scale dimple structure was utilized that generated a local deformation of the substrate. Figure 2a includes photographic images and schematic illustrations of the formation of the local dimple structure on the PDMS micro-pillar arrays using a vacuum tip with a diameter of 2.85 mm, which directly contacted the underside of the PDMS substrate. Without the application of vacuum pressure, the pressure inside the vacuum tip (p_int) was equal to the external atmospheric pressure (p_ext) of 101 kPa. When the vacuum pressure was applied, a significant pressure difference, ∆p = p_int − p_ext, was generated across the PDMS substrate. Due to the generated pressure difference, the PDMS substrate was stretched and deflected downward, forming the local dimple structure. In this work, ∆p was fixed at −81 kPa for the sake of convenient analysis. Considering the circular shape of the vacuum tip, it can be assumed that the pressure and corresponding strain stress were uniformly distributed across the entire surface of the dimpled PDMS substrate28. To characterize the structural changes of the vacuum-induced dimple structure, duplicated mould of the dimple structure were used (see Methods and Supplementary Fig. 2). Figure 2b,c show the optical photographic and SEM images of the negative replica’s cross-sectional image and duplicated dimple structure, respectively. The dimple structure caused the local stretching of the PDMS substrate as the substrate deflection formed a hemispherical shape. The degree of substrate stretching can be obtained by calculating the ratio of the relaxed (L_R) to the stretched (L_S) characteristic length of the dimple structure, as measured from the cross-sectional image of the negative replica. When the diameter of the vacuum tip was 2.85 mm, the measured substrate stretching (100 × (L_S − L_R)/L_R) was 20.7%. Due to the hemispherical shape of the dimple structure and the thickness of the PDMS substrate, negative and positive curvatures were generated at the bottom and border of the dimple structure, respectively. This local stretching of the substrate and the generation of the negative and positive curvatures enabled the dynamic changes in the distance between adjacent micro-pillars (ε_dist), which is proved by the magnified SEM images (Fig. 2d) of the micro-pillar arrays taken from the flat region (Fig. 2di), the negative curvature region (Fig. 2dii) and the positive curvature region (Fig. 2diii). On the flat region (ε_dist = 0%), the distance between adjacent pillars was 2 μm. On the local dimple structure, ε_dist could be dynamically changed due to the combined effects of the substrate stretching and the positive/negative curvatures; the maximum value of ε_dist is 49% for the positive curvature region and the minimum value is −11% for the negative curvature region.

To further identify the local deformation of the PDMS film with micro-pillar array, we controlled the bending radius of positive/negative curvature (R_p/R_n) and the slope of the dimple structure (θ_slope), by varying the sizes of the vacuum tips. Figure 3a represents a series of cross-sectional profiles of the dimples as a function of tip diameter under the same pressure condition (∆p = −81 kPa). Due to the high mechanical strength and elasticity of PDMS, the substrate was stretched from ~7 to ~32%, when the diameters of the vacuum tips were increased from 2 to 3.75 mm, forming positive/negative curvatures with corresponding bending radii R_p/R_n. Figure 3b shows the results of the extracted values of the bending radii and slopes of the dimple structures from the cross-sectional profiles in Fig. 3a. R_n and R_p indicate the minimum radii at the center and border of the dimple, respectively, and θ_slope represents the steepest angle on the side of the dimple. The absolute value of R_p decreased from ~0.8 to ~0.4 mm as the diameter of the tip became larger, while the absolute value of R_n gradually increased from ~1.3 to ~1.8 mm as the tip diameter increased. The structural changes in the dimple structures that were dependent on tip size directly affect ε_dist of micro-pillars. To investigate ε_dist between the neighboring pillars depending upon R_p/R_n, we measured ε_dist of micro-pillars from the SEM images taken from positive and negate curvatures of the duplicated dimple structures (Supplementary Fig. 3). As shown in Fig. 3c, even though the substrate was deformed only between ~7 and ~32% by tensile stress caused by substrate stretching, the lateral distance between the neighboring pillars that we can adjust would be enlarged from −15 to 61% due to additionally generated deformations by a micro-pillar placed on the positive/negative curvature. For example, when the diameter of the tip is 2 mm, the distance between adjacent pillars at the positive curvature will be 28% broader than in the flat state, whereas the negative curvature has a 15% denser pillar interval. The grey, blue and orange colored region represents variations of substrate stretching, ε_dist at positive and negative curvature, respectively. The experimental results showed good agreement with the calculated ε_dist, that considered the substrate stretching, R_p, and R_p. (Supplementary Fig. 4).

Since the changing effects of ε_dist has a sensitive correlation with the water adhesion force26, we investigate the dynamic F_adh changes on the dimple structure. Figure 4a depicts the force-distance curves for the PDMS substrate obtained with different values of ε_dist. For these measurements, a water droplet suspended at a metal ring was attached and detached from the flat PDMS micro-pillar arrays to measure the force between the water droplet and the substrate. The measured force was gradually increased after contact and reached a maximum just before the contacted droplet was separated from the substrate. F_adh of the PDMS substrate with micro-pillar arrays was significantly decreased from 67 to 49 μN as ε_dist increased up to 50%. Figure 4b shows the measured F_adh as varying ε_dist. It could be clearly observed that the F_adh of the PDMS micro-pillar arrays was gradually decreased as ε_dist increased. These F_adh changes according to ε_dist could be attributed to the variations in contact area between the micro-pillar arrays and the water droplet on the substrate. Theoretically, the wetting on the superhydrophobic PDMS micro-pillars without stretching is similar to the Cassie-Baxter wetting model, where air pockets exist between a droplet and a rough surface29; in this model, F_adh is proportional to the number of pillars in contact with the droplet303132. Therefore, when the stretching-induced strain stress was applied to the PDMS substrate, the number of micro-pillars that were directly contacting the water droplet could be reduced, resulting in decrease of F_adh.

Figure 4c is a schematic illustration of the forces on the surface of the dimple structure affecting the water droplet motions (left) and time-sequential photographic images of a blue-dyed, 10 μl moving water droplet on the PDMS micro-pillar arrays being moved via the controllable dimple structure (right). When the dimple structure was formed, the water droplet could be put into the structure and it seemed that the water droplet was spread along the dimple structure. Thus, the overall contact area of the water droplet with dimple structure was increased and the corresponding adhesion force between the water droplet and the surface was also increased. However, at the border of the dimple structure, the water droplet can be easily detached from the substrate due to the decreased local F_adh and generated sloped wall-induced gravitational force (F_g = mgsin θ_slope). Here, m is the mass of the water droplet and g is the gravitational acceleration constant. When F_g became larger than the local F_adh at the boundary of dimple, the droplet detachment from the substrate continuously occurred during the horizontal moving of the dimple structure. Consequently, the water droplet could be moved along with the moving dimple structure (Supplementary Fig. 5). We observed that the controllable droplet volume could be determined by the diameter of vacuum tip (Supplementary Fig. 6). A larger vacuum tip could manipulate smaller water droplet, which can be attributed to the smaller F_adh and larger θ_slope of the generated dimple structure. When the volume of water droplets were larger than the minimum volume, the motions of the water droplets could be successfully manipulated even when the diameter of a water droplet was larger than the tip size due to the surface tension of the water droplet.

In situ manipulation of droplet motions including moving, merging and mixing was demonstrated on the superhydrophobic, droplet-based microfluidic platform. Since the motions of water droplets on the platform are controlled by the vacuum-induced dimple structure, the moving path can be freely designed without additional patterning process (Fig. 5a). These individually controllable droplet movements can be used for merging and mixing operations of droplet (Fig. 5b and Supplementary Movie 1). The capture and release of water droplets during the operations could be controlled by adjusting the applied pressure of the vacuum tip. When transported droplets were brought into proximity, they merged and mixed slightly as forming one large droplet. The complete mixing of the water droplet could be achieved by moving the water droplet back and forth on the superhydrophobic PDMS micro-pillar arrays. The mixing process could be attributed to the rolling effect on the dimple structure, which may generate internal hydrodynamic flows in the moving water droplets. As previously mentioned, the water droplet manipulation could be achieved by the slope angle and variations in F_adh of the dimple structure; therefore, a water droplet on the moving dimple structure could be continuously rolled along the slope wall of the dimple structure, following the trace of the dimple structure. Our method for the manipulation of droplet motions takes advantage of the fact that the pathway of the water droplet motion is freely designed and additional additives are not required to be added into the water droplets during the operations.

The simple and intuitive mechanism of our system enables an efficient reaction and analysis capability using microliter-sized analytes or reagents. The programmable bio-chip platform could be applied to SERS measurements, which enable the detection of a few molecules in a highly diluted solution33. Figure 5c shows the schematic of the droplet-based SERS measurement system and in situ/ex situ Raman spectra of Rhodamine 6G (R6G) with concentrations ranging from 10⁻³ to 10⁻¹⁵ M. For the in situ SERS measurement, individual water droplets containing the target molecules and Ag nanoparticles (Ag NPs) which exhibit a wide resonance spectrum wavelength range were merged at the detection spot using the dimple structure, and then signals were collected from the droplet mixture. As shown in Fig. 5c, several typical peaks for R6G can be identified such as the C−C−C ring in plane bending (612 cm⁻¹), C−H out of plane bending (773 cm⁻¹), aromatic C−H bending (1183 cm⁻¹), C−O−C stretching (1312 cm⁻¹) and C−C stretching (1363, 1511, 1575 and 1650 cm⁻¹)34. We note here that the locations of observed peaks were identical to the previous work (Supplementary Fig. 7)34. In addition, droplet mixture could be removed without any residues remaining on the substrate, which enable the highly repeatable and reproducible in situ SERS measurements on the same substrate (Supplementary Fig. 8). However, in situ SERS measurements on the droplet-based microfluidic platform had a detection limit of 10⁻⁵ M due to freely diffusing Ag NPs and target molecules in the water droplets, which hindered the binding of the molecules to the surface of the Ag NPs33. To overcome the detection limit, the droplet mixture was evaporated on the surface over time until water was fully evaporated35. On a conventional hydrophilic SERS substrate, solution with target molecules could be spread out along the surface during the evaporation process and only few target molecules are located on the detection spot. In contrast, on the superhydrophobic surface, the contact area between the droplet and the surface can be minimized during the evaporation process due to the large contact angle35. As evaporation proceeded, the R6G/Ag NP solution became more and more concentrated and the target molecules and Ag NPs accumulated within an area of hundreds of square micrometres. After full evaporation, the target molecules were highly enriched on the Ag NPs detection spot, which permitted extremely high sensing capability even at femtomolar levels (10⁻¹⁵ M) for the ex situ SERS measurements. This detection limit is a hundred-fold lower than conventional SERS measurements on the flat substrate34. The obtained detection limit is poorer than the previously reported SERS measurements on superhydrophobic nanosensors35, due to the water adhesive property of the PDMS micro-pillar arrays. It is believed that the detection limit of our platform could be improved by using the water-repellent superhydrophobic surface since it enables the more enrichment of target molecules within smaller area.

The superhydrophobic surface allowing for the intuitive movement of droplets was further tested by generating uniform complexes composed of gene and vector for intracellular gene transfer. For decades, a great effort has been made to develop non-viral gene delivery vectors that can replace viral vectors, which have inherent safety concerns including tumorigenicity, immunogenicity, and insertional mutation3637383940. Most of the non-viral gene delivery vectors based on cationic polymers or lipids have relied on manual pipetting or vortexing to formulate transfection complexes via electrostatic charge interactions between anionic genetic materials and cationic delivery reagents364142. However, the manual handling of the droplets often results in the inefficient formulation of transfection complexes and loss of the samples43. Thus, we hypothesized that the homogeneous mixing of the droplets of genes and delivery vectors by automated operation on the superhydrophobic surface may be able to minimize the loss of the materials and also generate uniform complexes with higher transfection efficiency by increasing the chance of interaction between anionic and cationic materials. To test the potential application of our surface in the efficient formulation of transfection complexes, the surface was used to induce the formation of the complexes with lipidoid, a potent lipid-like material for gene delivery, and siRNA. Lipidoid has been identified as a highly effective siRNA delivery vector with a higher transfection efficiency and lower cytotoxicity than currently available transfection reagents36. Lipidoid can mediate siRNA transfer into various types of cells and tissues, and thus it has shown great potential for therapeutic applications in diverse diseases4445464748.

The transfection efficiency of the lipidoid-siRNA complexes generated on our surface was compared with that of the complexes prepared by the conventional mixing method using manual pipetting. The complexes composed of lipidoid (ND98) and green fluorescent protein-siRNA (siGFP) for the transfection were prepared on the PDMS micro-pillar arrays by simply merging and mixing the solutions of lipidoid and siRNA via microfluidic operations (Fig. 5d). Two days after the transfection into GFP-HeLa cells, ~70% silencing of GFP expression (70.0 ± 1.9%) was observed in the cells transfected with the complexes prepared by the conventional mixing method (Fig. 5e). Interestingly, the formulation of the complexes on our surface further increased GFP silencing up to ~80% (78.8 ± 0.5%) (Fig. 5e). The automated precise handling of the droplets of lipidoid and siRNA solutions on our surface may induce the efficient formation of more homogeneous complexes, ultimately leading to the enhancement of siRNA transfection efficiency and GFP silencing. Considering the high intercellular gene transfer efficiency of GFP-HeLa cell36, ~10% of improvement in the transfection is meaningful and it might be further increase in the cells usually exhibiting low transfection efficiency such as stem cells or primary cells. Given that the superhydrophobic PDMS micro-pillar array enables the formation of homogeneous complexes with enhanced transfection capability and minimal loss of the materials, and is also compatible with an intuitive automated operation, this system would be useful as a high-throughput platform to screen the material candidates for effective gene delivery and produce genetic therapeutics for disease treatment.

The 4-inch wafer-scale Si mould with micro-holes (2.5 μm radius, 4 μm height) was fabricated by conventional photolithography and subsequent reactive ion etching. The surface of the Si mould was modified with a hydrophobic self-assembled monolayer (dodecyltrichlorosilane: DTS, Aldrich) by immersing the mould in a 3 mM solution of DTS dissolved in toluene for 30 minutes at room temperature. Then, the DTS-coated Si mould was rinsed with ethanol and baked at 130 °C for 1 h to obtain a dense DTS layer. Next, 5 g of polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning), mixed with a curing agent at a volume ratio of 10:1, was spin-coated onto the mould (500 rpm, 10 seconds). To ensure the fine replication of the micro-pillar structures, the remaining air bubbles between the PDMS and the Si mould were removed in a vacuum chamber. The PDMS was cured at 70 °C for 2 h and carefully peeled off. The obtained PDMS micro-pillar arrays showed superhydrophobicity without further surface modification processes.

The obtained PDMS substrate with micro-pillar arrays was suspended on a sample holder with an 85 mm-diameter central hole, without sagging of the substrate. The central hole, the actual area for the water manipulation, is smaller than the PDMS substrate because the adhesion area between the substrate and the mould is required to mount the substrate on the sample holder. A vacuum tip, which created a local dimple on the PDMS substrate, was placed beneath the substrate for the manipulation of droplet motions. The location of the vacuum tip was precisely controlled by a programmed translational stage.

Firstly, a negative replica of the dimple structure, made of ultraviolet-curable photoresist (SU-8, Microchem), was obtained by ultraviolet light exposure for 30 min. Then, the negative replica was detached from the substrate and PDMS mixed with curing agent was poured onto the negative replica to fabricate the duplicated mould of the dimple structure.

Rhodamine 6G (R6G, Aldrich) was used as a probing molecule for the SERS measurements in solution of various concentrations (10⁻³, 10⁻⁶, 10⁻⁹, 10⁻¹² and 10⁻¹⁵ M). An aqueous nanoparticle suspension containing Ag NPs ~80 nm in diameter was synthesized using the polyol process49. Then, 5 μl of the Ag NP suspension and R6G solution were dropped onto the PDMS substrate and each droplet was moved to the detection spot for SERS measurement using the droplet motion manipulating system. A focused He-Ne laser (633 nm, 2 mW) was used as the Raman excitation light source. The signals from each sample were collected for 1 s using a 10 × microscope objective (Olympus) and analyzed using a Raman spectrometer (Horiba-Jobin-Yvon, LabRam HR).

GFP-expressing HeLa (GFP-HeLa) cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco BRL) supplemented with 10% (v/v) fetal bovine serum (FBS, Gibco BRL), penicillin (100 U/mL), and streptomycin (100 mg/mL) in humidified air with 5% CO₂ at 37 °C. Lipidoid (ND98), a potent siRNA transfection reagent, was synthesized as previously described3648. The lipidoid-siRNA complexes prepared by the conventional mixing method with pipetting36 were used for control transfection. To prepare the lipidoid-siRNA complexes on the PDMS micro-pillar arrays, ND98 lipidoid and GFP-siRNA (siGFP) were dissolved in the same volume of 25 mM sodium acetate (NaOAc) buffer solution (Sigma-Aldrich, pH 5.2). The droplets of each solution were placed apart on the PDMS micro-pillar arrays and then merged using the microfluidic operations. Subsequently, the merged droplets were further mixed by moving them side to side within a range of 30 mm for 5 min to induce complex formation. The mixed droplets were transferred into a tube right after the mixing and further incubated at room temperature for 15 min. The ratio of siGFP to ND98 in the formed complexes was 5:1 (w/w). The cells (2.5 × 10⁴ cells/cm²) were then transfected with the complexes (0.2 μg siGFP/cm²). To evaluate the GFP silencing two days after the siRNA transfection, GFP expression in the transfected cells was observed using a fluorescent microscope (IX71, Olympus). The fraction of GFP-positive cells was quantified by flow cytometry analysis. For the flow cytometry analysis, the cells were collected by trypsin treatment, washed with 1 × phosphate buffered saline (PBS, Sigma-Aldrich), resuspended in 2% (v/v) FBS (in PBS), and analyzed by a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA, USA) (n = 3).

The surface morphologies of the PDMS micro-pillar arrays and duplicated PDMS structures were characterized using a field emission scanning electron microscope (JSM-6360, JEOL). Static WCA measured using a contact angle measurement system equipped with a dynamic image capture camera (Phoenix 300, SEO Co., Ltd.). To characterize structural properties of the dimple structure, cross-sectional images of negative replicas were obtained using the contact angle measurement system and values of the substrate stretching, R_p, R_n, and slope angles were measured by Image J software. ε_dist values were measured from the SEM images taken from the duplicated PDMS structures and compared with the calculation that considered the substrate stretching, R_p, and R_n. The water adhesion force of the PDMS with micro-pillar structures was measured by a home-made micro-electromechanical balance system. A 10-μl water droplet suspended from a hydrophobic metal ring was moved toward and retracted from the sample at a speed of 0.01 mm s⁻¹. After the droplet was contacted to the substrate, it was dragged back from the substrate. The value of the measured force reached a maximum just before the contacted droplet separated from the substrate. To measure water adhesion force under stretched condition, micro-pillar arrayed PDMS substrates with different ε_dist (0, 25, 50, 75%) were fabricated using ε_dist tuned Si moulds.