Microfluidic fabrication of microparticles with structural complexity using photocurable emulsion droplets

Polymeric microparticles with hexagonal surface patterns comprising of colloids or dimples were fabricated using photocurable emulsion droplets. Colloidal silica particles within the interior of the photocurable emulsion droplets formed two-dimensional (2D) crystals at the droplet surface by anchoring on the emulsion interface, and the resulting composite structures were captured by rapid photopolymerization. A microfluidic device composed of two coaxial glass capillaries was used to generate monodisperse microparticles, with the evolution time determining the area of the anchored colloidal silica particles on the microparticle that was exposed to the continuous phase. The exposed region of silica particles could be modified by the introduction of desired functional groups such as dye molecules through simple chemical reaction with a silane coupling agent. This ability to modify the surface should prove useful in many applications such as chemical or biomolecular screening and colloidal barcoding systems.


Introduction
Emulsion droplets or bubbles are encountered in many aspects of everyday life, and multiphase systems have fascinated scientists and engineers due to their complex physics and chemistry as well as their wide applications in areas ranging from foods, cosmetics and drugs to printing and polymer synthesis [1,2]. Recent advances in microfluidics have enabled the production and manipulation of emulsion droplets in a simple and facile manner, thereby expanding the potential of such droplets to advanced applications [3,4]. One particularly promising area is the use of emulsion droplets as spherical molds or containers of colloidal particles to fabricate spherical particles. By exploiting the volume shrinkage of emulsion droplets due to diffusion and evaporation, colloidal crystals [5]- [8], colloidal clusters [9] and microparticles [10] have been molded by confining colloidal particles inside the three-dimensional (3D) space of emulsion droplets. In addition, when colloidal particles are confined at the interface of emulsion droplets, spherical cage structures are produced [11]- [19]. Moreover, repulsive colloidal particles anchored to the interface of a droplet have provided a visual clue for solving Thomson's problem, a longstanding puzzle related to the ground state of repulsive species on a spherical surface [20,21].
Recently, many researchers have used photocurable emulsion droplets to fabricate monodisperse microparticles with controlled shapes and functionalities due to their high productivity and simple and rapid consolidation process. Also, quantum dots, nanoparticles or molecules have been incorporated into emulsion droplets to endow them with magnetic or optical functionalities [22]- [24]. In these cases, the emulsion droplets were confined and deformed into prolates or oblates by shallow microfluidic channels and the resulting microparticles had nonspherical shapes after photopolymerization. On the other hand, photocurable droplets containing 3D colloidal arrays were prepared in cocurrent-flow microfluidic devices to make spherical photonic crystals [25]. In addition to simple emulsion droplets, Janus droplets or double emulsion droplets have also been generated in elaborate microfluidic devices for electro-responsive particles [26,27], microparticles of peculiar shapes [28]- [30] and polymeric microcapsules [31]- [33]. However, the introduction of functional groups or submicron-scale surface complexity into microparticles is still challenging. 3 In this paper, we report on the microfluidic fabrication of microparticles with complex surface morphologies using photocurable suspension droplets containing colloidal silica particles. We used a microfluidic device composed of two coaxial glass capillaries for drop generation. Using the dripping mode of drop breakup in co-flowing streams [34,35], we generated monodisperse emulsion droplets of photocurable resin with controlled size and generation frequency. The structural complexity of the resulting microparticles derived from the interfacial confinement of suspended silica particles that had migrated from inside the droplets. Therefore, the present system is anti-Bancroft type [36]. Droplets decorated with a hexagonal array of silica particles could be captured by rapid photopolymerization induced by 1 s of UV irradiation, leading to the formation of 'raspberry-like' particle-decorated microparticles [37,38]. Here, colloidal silica particles on the microparticle surface were partially exposed to the continuous phase, with the exposed area of particles being controlled by varying the evolution time (the time interval between droplet generation and UV irradiation). Through simple chemical reaction with the silanol groups on the silica surface, the exposed region can be selectively modified by introducing desired functional groups. By appropriately modifying the surface, we can then bind desired chemical or biological molecules onto the microparticles. Moreover, the increased surface area of the particles resulting from their complex morphology can enhance the loading of molecules, which is especially important for biomolecular screening or colloidal barcode systems [39]- [41]. In addition, the microparticle surface morphology could be changed to a 'golf ball-like' dimpled structure by removing the silica particles from the polymeric matrix.
Below we begin by describing the experimental details of the preparation of microparticles with complex surface patterns and discuss the microfluidic droplet generation and morphology of the resulting photopolymerized microparticles. In particular, we examine in detail the morphological evolution of the microparticles as a function of aging time for different sized colloidal silica particles. In addition, we modified the surfaces of the silica particles exposed to the continuous phase with dye molecules to demonstrate the surface functionalization through a simple chemical reaction.

Preparation of the photocurable silica suspension
To create polymeric microparticles with structurally complex surfaces, silica particles of 235 nm and 1 µm in diameter were synthesized by the Stöber-Fink-Bohn and seeded growth methods [42,43]. The silica suspensions were washed twice with ethanol and dried at 70 • C. The mass of dried powder was measured and then the powder was redispersed in ethanol under sonication. A solution of the photocurable monomer, ethoxylated trimethylolpropane triacrylate (ETPTA; Aldrich), containing 2-hydroxy-2-methyl-1-phenyl-1-propanone (Darocur 1173; Ciba Chemical) as a photoinitiator, was added to the ethanolic silica suspensions. Here, the volume fraction of silica particles in ETPTA was kept at 5% (v/v) in ethanol-free base. The ethanol was removed selectively from the mixture by heating at 70 • C in a convection oven for 1 day.

Generation of droplets and photopolymerization
Monodisperse emulsion droplets were prepared using a microfluidic device composed of two glass capillaries, as shown schematically in figure 1(a). Glass capillaries with the desired diameters and geometries were fabricated by the heating and pulling method using commercial glass tubes (Sutter Instrument Company no. BF100-50-10 for the inner capillary and a disposable glass Pasteur pipette for the outer capillary). The inner and outer capillaries were then assembled coaxially using optical adhesive (NOA 63; Norland) under microscope observation. Here, the end of the inner capillary was positioned at the narrowest region of the outer capillary and sealing with optical adhesive was applied three times to prevent leakage.
To generate the suspension of photocurable silica droplets with a narrow size distribution, the silica-ETPTA suspension and aqueous solution with 1 wt.% Pluronic F108 (ethylene oxide-propylene oxide-ethylene oxide tri-block copolymer surfactant; BASF) were introduced through the inner and outer capillaries, respectively. The flow rates of the suspension and aqueous solution were controlled using syringe pumps (model 781200; KD Scientific) for stable generation of emulsion droplets in the dripping regime. The monodisperse droplets from the end of the outer capillary were collected in a glass Petri dish.
Photopolymerization of emulsion droplets was carried out by 1 s UV irradiation with a high-pressure mercury lamp. The rapid consolidation that occurred during photopolymerization made it possible to capture the complex structures without deformation. As schematically illustrated in figure 1(b), silica particles in a droplet move to the interface and are anchored. To determine the time required for particle anchoring, UV irradiation was applied at a specific time after droplet generation at the end of the inner capillary. The resulting microparticles, which have a raspberry-like structure, were rinsed in deionized water several times to remove the residual surfactant molecules.

Coupling of dye molecules and particle dissolution
The surfaces of silica particles that are partially exposed from the ETPTA matrix can be selectively functionalized using a silane coupling agent. Here, we demonstrated this by treating the surface with dye molecules [44]. For the coupling with red dye molecules, tetramethyl rhodamine isothiocyanate (TRITC; Aldrich) molecules were first bonded with 3-(aminopropyl)trimethoxysilane (APTMS; Aldrich). TRITC (0.0015 g) was dissolved in 1.8 ml of ethanol and 0.09 ml of APTMS was dropped into the previous ethanolic solution and reacted for 24 h with gentle stirring. On the other hand, 6.5 ml of ethanol containing microparticles was mixed with 0.551 ml of ammonia for 10 min; the microparticles in this solution had been previously immersed in 0.1 wt.% NaOH solution for 10 min to activate the silanol groups. Subsequently, 2 µl of the TRITC-APTMS ethanolic solution was added to the microparticle suspension. Then, 4 µl of tetraethoxysilane (TEOS; Aldrich) was added and the mixture was allowed to react for 48 h. In a similar manner, fluorescein isothiocyanate (FITC; Aldrich) was used as a green dye molecule. FITC (0.0030 g) was dissolved in 2 ml of ethanol and reacted with 0.474 ml of APTMS for 24 h. The resulting solution was subjected to the same procedures as those followed for red dye coupling. After successful coupling of the dye molecules, the microparticles were washed with ethanol several times and then redispersed in water. In other experiments, internal doping of microparticles with dye molecules was achieved by using an ETPTA suspension containing 10 −4 M rhodamine B isocyanate (Aldrich).
Golf ball-like microparticles were fabricated by dissolving the silica particles on the raspberry-like particles, as shown in figure 1(b). To remove the silica particles, the microparticles were dispersed in 1 wt.% NaOH solution (Junsei) or 5% HF solution (50%; Sigma-Aldrich). We found that 235 nm silica particles dissolved completely in 1 wt.% NaOH solution within 24 h, whereas 1 µm silica particles required a few days. Therefore, 5% HF solution was used for dissolution of the large silica particles; in this solution, the particles dissolved within a few minutes.

Characterization
The generation and flow of emulsion droplets in the glass capillary device were observed using an inverted optical microscope (TE2000-U; Nikon) equipped with a high-speed video camera (Motionscope M1; Redlake). The surface morphologies and sizes of the microparticles were observed by optical microscopy (L150; Nikon) and scanning electron microscopy (XL30; Philips) after Au coating. In addition, dye-treated microparticles were observed by laser scanning confocal microscopy (LSM 510; Carl Zeiss). For excitation of TRITC/rhodamine B isocyanate and FITC, 543 and 488 nm lasers were used, respectively.

Generation of monodisperse emulsion droplets
The microfluidic device composed of glass capillaries generated monodisperse emulsion droplets of photocurable silica suspension, as shown in figure 2. The silica-ETPTA suspension was forced to flow through the inner capillary (outer diameter: 78 µm), whereas an aqueous solution containing surfactant molecules was introduced through the outer capillary (inner diameter: 239 µm). Typically, the volumetric flow rate of the continuous phase was approximately 100 times that of the dispersed phase, and the suspension was emulsified in dripping mode. Immediately after the suspension flowed out through the end of the inner capillary, the outer continuous stream elongated the suspension and induced the formation of a neck, ultimately leading to the formation of a spherical droplet. The droplets were generated at a constant frequency and flowed through the wide channel of 2 mm diameter. Figures 2(b)-(f) show optical microscope images of different regions of the glass capillary device. Because the droplets are produced at a constant frequency, they form a train until they arrive at the expansion area, as shown in figure 2(c). Near the entrance of the expansion area, the distance between neighboring droplets diminishes until the droplets finally come into contact (see figure 2(d)). Because the droplets were stabilized with surfactant molecules, they flowed through the wide channel without coalescing, as shown in figures 2(e) and (f). Even though the droplets contain particles at 5% (v/v), they are transparent due to negligible scattering as a result of refractive index matching between the photocurable monomer (n ETPTA = 1.4689) and silica particles (n silica = 1.45). The droplet generation process and flow near the expansion area are shown in detail in the movie clip available from stacks.iop.org/NJP/11/075014/mmedia. Satellite droplets were also produced even in the dripping mode. At the high flow rate of the continuous phase, the viscous suspension formed a long neck whose length and diameter were about seven times longer and 13 times smaller than the diameter of a main droplet, respectively. This neck was divided into a few small droplets due to the plateau-Rayleigh instability. Such satellite droplets are indicated with black arrows in figures 2(c) and (g). However, these small droplets were spatially isolated from the main droplets without performing an additional separation step. As can be seen in figure 2(g), which was taken at the same position as figure 2(f) but with a different focal plane, the small satellite droplets flowed through an edge of the midplane in the cylindrical channel, while the main droplets flowed through the bottom side of the channel. A similar separation phenomenon of large and small particles in an expanding rectangular channel was reported by Yamada and coworkers [45]. Here, as we use a cylindrical channel with an expanding geometry, the efficiency of separation is enhanced because the gravitational force causes the main droplets to flow along the bottom side of the cylinder. Finally, when we connected the exit of the outer capillary of 1 mm diameter to a glass Petri dish filled with continuous phase, which is the second expansion geometry, the satellite droplets were piled far from the position of the main droplets. This spatial separation arises because the satellite droplets from an edge of the mid-plane region of the channel are forced to travel forward to the side direction at the end of the outer capillary due to the curved stream line near the edge. In addition, because the satellite droplets are smaller and thus induce a smaller gravitational force, they can move much longer distances from the exit than the main droplets moving at the same exit velocity. Figures 3(a) and (b) show optical microscope images of the collections of main and satellite droplets on the same Petri dish. These images clearly demonstrate the separation of the droplets with bimodal size distribution into two distinct regions: the separation distance between the two regions of differently sized droplets was approximately 2 cm.
In our system, the size of the main droplets is determined by the flow rate of the continuous phase for a given device geometry. Because the drag force and capillary force are balanced at the moment of droplet generation in dripping mode, the droplet size decreases as the velocity of the continuous phase increases. Figures 4(a)-(c) show optical microscope images of monodisperse microparticles with three distinct sizes obtained by photopolymerization of droplets produced at three different velocities of the continuous phase. Figure 4(d) shows the variations in microparticle diameter and Reynolds number as a function of the average velocity of the continuous flow, where the latter was calculated as the volumetric flow rate per crosssectional area of the channel. These data disclose that, as the velocity of the continuous phase increases, the droplet size approaches asymptotically a value equal to the outer diameter of the inner capillary. At too low a velocity, on the other hand, the suspension burst out into the outer capillary and the droplet breakup point was shifted to an unfixed position on the outer capillary, leading to a broadening of the droplet size distribution (data not shown).
The flow rate of the suspension does not affect the size of the main droplets in the stable dripping regime. Instead, it determines the droplet generation frequency. Because all of the flowing suspension is emulsified into droplets, the frequency of droplet generation is determined by the volumetric flow rate per volume of a single droplet (if we ignore the volume of the satellite droplets). Figure 5 shows optical microscope images taken at the same position of the device for different flow rates of the suspension, where the flow rate of the aqueous phase was the same in all cases. From the images, we can discern an increase in the generation frequency with increasing volumetric flow rate of the suspension. By examining high-speed video, we found that the generation frequencies in the systems shown in figures 5(a)-(c) were approximately 40, 80 and 120 Hz, while the frequency of the system in figure 5(d) may be close to 600 Hz although we could not confirm it experimentally.

Structural complexity of microparticles
Within the 3D space of each ETPTA droplet, the silica particles move around through Brownian motion. However, once the particles contact with the free interface, motion in the radial direction is limited because the interface holds the particles for a while to reduce the total interfacial energy of the system. In particular, when the interface is covered compactly with anchored particles, the particles are immobilized and form a regular arrangement on the spherical surface due to interparticle interactions in the direction parallel to the interface. Therefore, we can obtain spherical structures decorated with hexagonal arrays of colloids if the ETPTA droplets are solidified by photopolymerization after sufficient times required for the formation of 2D crystals.
With monodisperse droplets of diameter 112 µm containing 235 nm silica particles, we could fabricate monodisperse microparticles with a raspberry-like structure. To investigate the change in microparticle surface morphology as a function of time, the UV-induced photopolymerization was carried out after various times from droplet generation at the end of the inner capillary in the microfluidic device. Figures 6(a)-(d) show scanning electron microscope (SEM) images of the surfaces of microparticles prepared by polymerization after evolution times of 20 s, 7.5 min, 1 h and 2 h, respectively. For an evolution time of 20 s ( figure 6(a)), which meant that the microparticles were photopolymerized inside the microfluidic device, the particles are randomly attached on the surface of the droplets. On the other hand, microparticles with an evolution time of over 7.5 min show a hexagonal arrangement of colloids on their surfaces, as shown in figures 6(b)-(d). Here, the area of a silica particle exposed to the continuous aqueous phase increases with increasing evolution time. As shown in figure 6(b), more than half of the surface of a silica particle is embedded inside the ETPTA phase with an evolution time of 7.5 min. However, more than half of the surface was exposed to the continuous aqueous phase with an evolution time of 2 h (see figure 6(d)). This morphology evolution arises because the system is still approaching the equilibrium state (the state of minimum total interfacial energy). When the silica particles are introduced to the continuous phase initially, they do not anchor on the interface of pure ETPTA droplets [37]. This means that the presence of unanchored silica particles in the aqueous phase is favorable from the standpoint of interfacial energy due to the large interfacial energy between silica and ETPTA compared with that between silica and water.
The surface morphology of the raspberry-like microparticles could be changed to a golf ball-like dimpled structure by removing the silica particles from the ETPTA matrix. By placing the microparticles in NaOH aqueous solution, we could selectively dissolve the silica particles to create an array of dimples where the particle array occupied previously. Figures 6(e)-(h) show the surface morphologies of golf ball-like microparticles derived from microparticles prepared with evolution times of 20 s, 7.5 min, 1 h and 2 h. These images again demonstrate the evolution of the surface morphology with time. Figure 7 shows SEM images of raspberry-like and golf ball-like microparticles prepared with an evolution time of 7.5 min. As can be seen from the low-magnification image ( figure 7(a)), the fabricated microparticles have a narrow size distribution. At moderate magnification (figures 7(b) and (c) for the raspberry and golf ball-like microparticles, respectively), moiré fringes are observed on the particle surfaces, indicating the presence of a well-ordered array of colloids on the entire surface. In fact, the moiré fringes were caused by the interaction between the SEM scanning line pattern and the hexagonal lattices of colloids [46]. Therefore, the fringe appears only at specific magnifications and is clearer on the golf ball-like particles due to the higher contrast between the ETPTA matrix and dimples compared with that between the matrix and the silica particles on raspberry-like microparticles. The higher magnification image in figure 7(d) shows individual dimples instead of the fringe. Using the parts of the silica particles exposed to the continuous phase, we can introduce functional groups or chemical/biological molecules that are important in many applications of microparticles. The silica particles have silanol groups on their surfaces, which can be converted into other functional groups through simple chemical reactions using silane coupling agents. As an example, we treated the exposed areas of silica particles embedded in microparticles with a silane coupling agent bonded to a dye molecule, TRITC. Laser scanning confocal microscope images of the TRITC-treated microparticles are shown in figure 8. Because the dye molecules were bound to the exposed regions of the silica particles, the fluorescent signal is predominantly emitted from the microparticle surfaces. Even at higher magnification, we still could not see the signals from the individual silica particles due to their small size of 235 nm, which is less than the resolution of the confocal microscope.
In other experiments, 1 µm silica particles were introduced into emulsion droplets of diameter 210 µm. However, this system exhibited an evolution time scale different from that of the 235 nm silica particles on 112 µm droplets. Figure 9 shows SEM images of microparticles prepared with evolution times of 20 min, 1 h, 5 h and 14 h. As is evident in figure 9(a), 20 min was an insufficient time to create a high density of particles on the interface for the 1 µm silica particles; this stands in contrast to the behavior of the 235 nm silica particles, which required just 1 min. One hour after droplet generation, the 1 µm silica particles were packed hexagonally on the spherical interface of the 210 µm droplets and a small part of each particle was exposed to  the aqueous continuous phase. Similar to the 235 nm particles, the exposed area increased with increasing evolution time (see figures 9(c) and (d)). After an evolution time of 14 h, the droplets were partially stripped by the continuous phase, as shown in the inset of figure 9(d) where the dark hemisphere is bald. The relatively long times required for the complete adsorption of particles originate from the large sizes of silica particles and droplets. The diffusivity of a particle in dilute suspension is inversely proportional to the particle diameter (d) according to the Stokes-Einstein equation: D 0 = kT /3πηd, where η is the viscosity of the suspension, Figure 10. SEM image of microparticles with a 1 µm particle array in spherical 2D space. The microparticles were generated by photopolymerization 1 h after droplet generation. The scale bar is 50 µm.
k the Boltzmann constant and T the temperature. On the other hand, the distance (l) to reach the interface is proportional to the droplet radius (R). Therefore, the time (t) to reach the interface by diffusion is given by Therefore, 1 µm silica particles confined in a 210 µm droplet will require about 15-fold more time to be completely adsorbed than 235 nm particles confined in a 112 µm droplet when we do not consider the number of particles required to complete adsorption and dispersed in a droplet. Consistent with the theory that the process is diffusion limited, we found that 90 µm droplets containing 1 µm particles required an evolution time of about 10 min for complete adsorption (data not shown). Particle removal was also carried out with a wet etching process, and the resulting surface morphologies are shown in figures 9(e)-(h). Here, the dissolution etchant was changed to HF because NaOH solution required too long a time to remove the 1 µm silica particles. Figure 10 shows an SEM image of 210 µm microparticles covered with 1 µm silica particles prepared with an evolution time of 1 h. At this magnification, we can see the individual particles forming hexagonal arrays on the center of the microparticle and moiré fringes on the north pole. These fringes are caused by the difference in lattice spacing between the center and pole regions due to the orthographic projection of the hexagonal lattice on the pole region.
When the 1 µm silica particles on microparticles were treated with TRITC, they also showed a fluorescent signal from the microparticle surfaces, as shown in figure 11. In contrast to the 235 nm silica particles, the 1 µm particles are large enough to see the signals from individual particles by high-magnification confocal microscopy. As can be seen from the six images in figure 11(c), each silica particle shows a fluorescent signal, whereas the ETPTA matrix does not, confirming that the dye molecules were selectively attached to the surfaces of the silica particles. Here, the six confocal microscope images show the same microparticle at six different focal planes, ranging from near the bottom plane at z ∼ 0.5 µm (image 1) to the plane at z ∼ 13 µm (image 6).
The interior of the microparticles can also be doped with dye molecules in a simple manner. If we use a dispersed phase that is premixed with dye molecules, we can obtain a fluorescent signal from the entire volume of the microparticle. Here, we used ETPTA resin containing 10 −4 M rhodamine B isocyanate, which can be excited by 543 nm laser irradiation. In addition, we treated the exposed surfaces of the silica particles on the microparticles with FITC using a silane coupling agent, which can be excited by 488 nm laser irradiation. As shown in figures 12(a) and (b), the microparticles showed fluorescence from their interior or surface depending on the laser beam excitation wavelength; the red signal comes from the interior of the particle due to the rhodamine B isocyanate under 543 nm excitation, whereas the green signal comes from the surface of the microparticle due to the FITC under 488 nm excitation (red and green are pseudo-colors). Figure 12(c) shows a merged image of figures 12(a) and (b).

Conclusion
We have demonstrated the self-organization of silica particles on the spherical interface of photocurable droplets, and the structural consolidation of the resulting microparticles by UV-induced photopolymerization. To do this, a microfluidic device composed of two coaxial glass capillaries was used to generate monodisperse droplets that were subsequently converted into monodisperse microparticles by UV-induced photopolymerization. Silica particles dispersed in each droplet of photocurable emulsion moved to the interface, where they were anchored eventually. Depending on the evolution time, the area of silica particles exposed to the continuous phase was controlled and the silanol groups on the exposed region could be converted into other functional groups through simple chemical reaction. In addition, the surface morphology could be changed to a dimpled structure by selectively dissolving the silica particles on the polymeric microparticles. The strategy used in this study will be useful for introducing functional groups or biomolecules and complex patterns on microparticles, which is important in a wide range of microparticle applications such as chemical/biological screening, colloidal barcoding and rheological studies.