For internal gelation, cross-linkers come from inside the alginate droplets and are either soluble or insoluble/slightly soluble in water. In this approach, cross-linkers are always introduced in the microfluidic device.
With water-soluble cross-linkers such as barium chloride (BaCl2) and calcium chloride (CaCl2), alginate is crosslinked directly at the interior of droplets. These agents can be mixed with Na-alginate before or after droplet generation.
Mixing cross-linkers before droplet generation
A first category of strategies is based on mixing water-soluble cross-linkers with Na-alginate before droplet generation. The cross-linkers used in these studies are BaCl2 and CaCl2; these strategies are summarized in Table 1.
Trivedi et al. worked on cell encapsulation by alginate hydrogel microparticles . For the preparation of microparticles, an aqueous solution of cell-containing Na-alginate (1%) and a solution of BaCl2 (50 mM) were injected into the capillary and mixed via a Y-shaped junction. At the exit from the mixing region, highly viscous silicone oil (10 cSt) without surfactant was injected by flow-focusing in order to generate droplets. However, the mixing of Na-alginate and barium cations triggered ionic crosslinking, causing gelation in the mixing region which impacted droplet generation. Finally, instead of generating droplets as expected, a jet of gel was produced with a partially formed droplet head and a long gelatinous tail.
To deal with this issue, the mixing region can be reduced before droplet generation, as Zhang et al. did  Using a 5-channel microfluidic device, they mixed Na-alginate fluid (0.5 wt%), CaCl2 fluids (0.1 wt%) and mineral oil fluids with a surfactant (Span 80, no concentration mentioned) as shown in Fig. 5. Droplets were generated by co-flow. However, instead of producing discrete droplets, a line of knots connected with each other was formed. This phenomenon persisted with a wide range of flow rates of oil due to viscosity which increased instantly when Na-alginate and CaCl2 were mixed, because of rapid gelation. It was therefore impossible to generate droplets at the junction, despite the use of surfactant.
The problem can be solved by using low concentrations of Na-alginate and CaCl2 solutions. In this case, gelation proceeds after droplet generation and is enhanced by using partially miscible fluids. Rondeau and Cooper-White used Dimethyl carbonate (DMC) as the continuous fluid  (Fig. 6). The solubility of water in DMC is about 3 wt% at room temperature . Aqueous solutions of Na-alginate (0.5 wt%) and CaCl2 (0.25 wt%) were injected respectively from inlets A and B (Fig. 6a). After a short pre-gelation channel, DMC was injected from inlet C. Na-alginate/CaCl2 droplets were generated in DMC (no mention of surfactant usage) by flow-focusing. Along the serpentine channel, because of the low solubility of water in DMC, water diffused gradually from droplets into DMC, causing the shrinkage of droplets along the channel. Internal gelation occurred at the same time. Microparticles with a diameter of 20 μm were observed at the outlet of the channel and collected in an aqueous solution of CaCl2 (2 N) to reinforce the gelation (Fig. 6b). The diameter of Ca-alginate hydrogel microparticles was influenced by the experimental parameters such as the initial concentration of Na-alginate, flow rates of fluids and channel size. To be precise, smaller Ca-alginate hydrogel microparticles can be obtained by using a less concentrated Na-alginate solution, a higher flow rate ratio between the continuous fluid and the dispersed fluid, or a narrower channel. However, DMC is also slightly soluble in water, with a solubility of 12.7 wt% at 20 °C . Thus, during diffusion of water from droplets into DMC, DMC can also diffuse into droplets. This means that, after gelation, DMC can be captured inside alginate hydrogel microparticles. Additional work measuring the amount of DMC residue within microparticles could open the way to further applications.
Following the work of Rondeau and Cooper-White, we tested, in a T-junction (Fig. 7), the direct generation of droplets of Ca-alginate in DMC without surfactant from a mixture of more diluted Na-alginate and CaCl2 solutions (both at 0.06 wt% after mixing). However, this solution was not clear and local gelation was occasionally observed with the naked eye. When these gels entered the channel, droplets were generated in a discontinuous way. This indicates that, even at very low concentrations, thorough mixing of Na-alginate and CaCl2 solutions leads to gelation, disturbing droplet generation.
In a microfluidic device (Fig. 8) of similar design to Zhang et al. , we were able to generate discrete droplets by using extremely diluted solutions of Na-alginate (0.006 wt%) and CaCl2 (0.002 wt%). The continuous fluid was DMC without surfactant. Droplets were observed after the cross-junction (point A in Fig. 8a). Since they were relatively close to each other in the channel, causing coalescence at the outlet (point B in Fig. 8a), a second flow of DMC was introduced as a spacer using a T-junction. When the second DMC flow rate was relatively low, the generation of droplets upstream was not disturbed, so that droplets were uniform (Fig. 8b). However, the coalescence at the outlet persisted. Thus, high second DMC flow rates were applied to sufficiently increase the distance between droplets. Nevertheless, this quickly disturbed the generation of droplets upstream, as indicated by heterogeneities in droplet size and frequency (Fig. 8c). Using surfactant would prevent droplet coalescence.
To summarize (Table 1), authors mixed highly concentrated solutions of Na-alginate and water-soluble cross-linkers before droplet generation to make them gelate. However, droplet generation was hindered by rapid gelation and microparticles were difficult to obtain. To delay gelation, less concentrated solutions of Na-alginate and water-soluble cross-linkers were mixed before droplet generation, in the microfluidic device or off-line. Then, concentrations were increased after droplet generation by diffusion of water from the droplets to the continuous fluid, due to their partial miscibility with water. Hence, gelation proceeded slowly with droplet shrinkage. However, mixing cross-linkers with Na-alginate before droplet generation led to heterogeneities in droplet size and frequency.
Mixing cross-linkers after droplet generation
To delay gelation, water-soluble cross-linkers need to be mixed with Na-alginate only after droplet generation. Studies doing so, and which also use BaCl2 and CaCl2 as cross-linkers, are summarized in Table 2.
Xu et al. prevented rapid gelation by delaying the direct contact between Na-alginate and calcium cations . In a first cross-junction, two face-to-face channels were used to introduce CaCl2 (2 wt%) and Na-alginate (2 wt%) solutions (Fig. 9a) perpendicularly to a flow of water. Thus, after the first cross-junction, a flow of water (acting as a buffer) separates the flows of Na-alginate and CaCl2. Then octyl alcohol oil (no mention of surfactant) was injected at a second cross-junction. Droplets of Na-alginate/CaCl2 were generated by flow-focusing. In the “synthesizing channel” (Fig. 9a), within each droplet, mixing Na-alginate and CaCl2 induced internal gelation. In this way droplets were transformed into Ca-alginate hydrogel microparticles (Fig. 9b). For this strategy, the size of Ca-alginate hydrogel microparticles is entirely dependent on the experimental conditions, such as flow rates of fluids and channel size. Manipulation of microparticles is difficult if their diameter is smaller than 10 μm.
Another strategy to delay gelation was carried out by Liu et al.  involving coalescence of Na-alginate droplets with CaCl2 droplets generated separately. First, on a microfluidic chip (Fig. 10a), Na-alginate (2 wt%) droplets (Fig. 10b) and CaCl2 (2 wt%) droplets (Fig. 10c) were generated in soybean oil without surfactant by flow-focusing using two independent cross-junctions. Then droplets converged via a T-junction (Fig. 10d) followed by two successive circular expansion chambers (Fig. 10d, e). Thus, droplets could collide either at the T-junction or in circular chambers. Within the coalesced droplets, Na-alginate was crosslinked by calcium cations forming Ca-alginate hydrogel microparticles. With different flow rates and channel geometries, various shapes and sizes of microparticles could be produced (Fig. 10f). Nevertheless, the design of circular expansion chambers gives rise to local changes in flow velocity. Droplet circulation can be disturbed, thereby affecting homogeneity in droplet shape, size and frequency.
Droplets could also be coalesced by exploiting physicochemical parameters between the continuous fluid and the dispersed fluid. In the work of Trivedi et al., droplets of Na-alginate (1 wt%) containing cells were generated upstream in a highly viscous silicone oil (10 centistoke) by flow-focusing without surfactant [47, 48]. An aqueous solution of BaCl2 (50 mM) was injected downstream by a T-junction. With the help of dye, observations at the T-junction indicated that BaCl2 fluid merged spontaneously with Na-alginate/cells droplets, instead of forming independent BaCl2 droplets. However, this strategy lacks flexibility. The expected coalescence happens only when appropriate solvents are used. For instance, when using low-viscosity and low-interfacial energy γCD soybean oil, independent droplets of BaCl2 were observed. They coalesced downstream with Na-alginate/cells droplets. This implied that successful coalescence of droplets could only take place with appropriate interfacial energy and viscosity .
Water-insoluble or weakly soluble cross-linkers
To delay gelation, Na-alginate can be mixed with water-insoluble or weakly soluble cross-linkers, in water. This will not lead to instant gelation since there are no available cations in water. In the case of cross-linkers which are pH-sensitive, such as calcium carbonate (CaCO3) and calcium-ethylenediaminetetraacetic acid (Ca-EDTA) complex, an acid is used in the continuous fluid to release the cations from inert cross-linkers. Therefore, gelation by the available cations happens after droplet generation. These strategies are summarized in Table 3.
In the work of Zhang et al. , fine particles of CaCO3 (0.1 wt%) were dispersed in an aqueous solution of Na-alginate (2 wt%). Soybean oil with a surfactant (Span 80, 3 wt%) and containing acetic acid (5 wt%) was used as the continuous fluid (Fig. 11a). Droplets of Na-alginate/CaCO3 were generated by flow-focusing in soybean oil/acetic acid (Fig. 11b). Droplets pH decreased because of the acetic acid in the oil. As a result, calcium cations were released from CaCO3, causing internal gelation of the alginate. Finally, Ca-alginate hydrogel microparticles were collected in oil (Fig. 11c). However, when collected on a substrate, they had a “pancake” shape and were soluble in aqueous solution owing to insufficient gelation. No improvement was observed from increasing the concentration of acetic acid or that of CaCO3. Moreover, a higher concentration of CaCO3 particles would give rise to their aggregation in the channel . The mechanical properties of the microparticles could not therefore be improved.
The same principle was also applied by Akbari and Pirbodaghi to prepare cell-encapsulating microparticles (Fig. 12) . At a first T-junction, a fluid of Na-alginate (1.5 wt%) containing cells flowed into the middle channel (Fig. 12a), while the Na-alginate fluid (1.5 wt%) containing CaCO3 nanoparticles (35 mM) was introduced by two side channels (Fig. 12b). This geometry was used to create a coaxial stream while avoiding direct mechanical contact between cells and the potentially damaging CaCO3 particles. At a second T-junction, fluorocarbon oil with surfactant (fluorinated surfactant, 1 wt%) was injected. Droplets of Na-alginate/cells/CaCO3 were then generated by flow-focusing. After droplet collection, acetic acid (0.1 vol%) dissolved in oil was added to release calcium cations within droplets, causing gelation of alginate. Droplets were thus transformed into Ca-alginate hydrogel microparticles, some with cells encapsulated (Fig. 12c). However, the mixture of CaCO3 and Na-alginate was not homogeneous, which can be seen from Fig. 12b. Thus, the varying amounts of CaCO3 influenced the degree of gelation in each droplet, yielding Ca-alginate hydrogel microparticles with different mechanical properties. This issue is not discussed by Akbari and Pirbodaghi . Furthermore, not all microparticles encapsulated cells, for reasons not explored in the article. Sorting is therefore required after the preparation of microparticles, which complicates the procedure.
Combining the strategy of Zhang et al. and Akbari and Pirbodaghi to conduct gelation both in the microfluidic device and in the collection bath, Yu et al.  produced Ca-alginate hydrogel microparticles for protein encapsulation. First, from inlet 4 (Fig. 13a), an aqueous solution of antigen or protein was injected. It co-flowed with another aqueous solution of alginate (2 w/v%) mixed with CaCO3 particles (200 mM) and injected from inlet 3. Mineral oil with Span 80 added was injected from inlet 2 as a continuous fluid. In the flow-focusing channel, droplets containing alginate, CaCO3 and protein were formed. From inlet 1, another continuous fluid, mineral oil containing Span 80 and acetate acid, was introduced. When the acetate acid diffused into droplets, calcium cations were released. The alginate was then crosslinked, leading to preliminary gelation. The droplets were collected in an aqueous solution of CaCl2 (0.27 M) to enhance gelation. In the end, spherical hydrogel microparticles were formed, with protein encapsulated (Fig. 13b-c). According to the authors, the preliminary gelation in the microchannel prevented the deformation that occurs when droplets are collected directly in an aqueous solution of CaCl2. As mentioned above, however, since CaCO3 is not soluble in water, a high concentration of CaCO3 will clog the microchannel. Thus, the scope for preliminary gelation is limited. Moreover, it takes time (in this case, overnight) to obtain a mixture where CaCO3 particles are well dispersed.
In order to obtain a homogeneous internal structure of hydrogel microparticles, Utech et al. used a slightly water-soluble calcium-ethylenediaminetetraacetic acid (Ca-EDTA) complex as the cross-linker . An aqueous solution of Na-alginate (2 wt%) mixed with Ca-EDTA (50 mM) was first prepared. This homogeneous mixture was used as the dispersed fluid for the microfluidic system. The continuous fluid was a fluorinated carbon oil with a biocompatible surfactant (1 wt%) containing acetic acid (0.05 vol%). Droplets of Na-alginate/Ca-EDTA were generated in oil/acetic acid by flow-focusing (Fig. 14a). Due to the use of acetic acid, calcium cations were released from Ca-EDTA in each droplet (Fig. 14b), causing internal gelation of the alginate. The Ca-alginate hydrogel microparticles formed (Fig. 14c) had a homogeneous internal structure and were stable in an aqueous medium without dissolution. It should be noted that, the solubility of Ca-EDTA in water being low (0.26 M at 20 °C), the concentration of Ca-EDTA in the Na-alginate solution was limited in order to keep the solution homogeneous. Thus, this strategy is not appropriate when microparticles need to be highly crosslinked. Furthermore, care should be taken with Ca-EDTA, as EDTA is used to dissolve alginate hydrogel microparticles in the literature [29, 56].
In conclusion, internal gelation of alginate can be realized by using cross-linkers that are soluble or insoluble/slightly soluble in water. When water-soluble cross-linkers are used, the instant gelation disturbs droplet generation. The problem can be solved by using partially miscible fluids with limited mixing prior to droplet generation, and/or by using extremely diluted solutions and surfactant (Table 1). Mixing cross-linkers and Na-alginate after droplet generation involves merging droplets or flows of Na-alginate and water-soluble cross-linkers (Table 2). The resulting droplets are dependent on physicochemical properties like viscosity and interfacial energy. If water-insoluble/slightly soluble cross-linkers are used, they are mixed with alginate before droplet generation. For pH-sensitive cross-linkers, acid is then needed to release cations, after which internal gelation takes place (Table 3).
A homogeneous microparticle internal structure can be achieved by choosing appropriate cross-linkers. However, because of low solubility in water, it is important to limit the concentration of cross-linkers to avoid precipitates in the channel.
In external gelation, cross-linkers come from outside the alginate droplets and are diffused into the alginate droplets or the microparticles formed, inducing crosslinking. Unlike internal gelation, in which cross-linkers are always introduced “on-chip” (in the microfluidic device), in external gelation, cross-linkers can be introduced both “on-chip” and/or “off-chip” (outside the microfluidic device).
On-chip introduction of cross-linkers
For external gelation, several authors introduced cross-linkers “on-chip”. They used calcium acetate (Ca (CH3COO)2) or CaCl2 as cross-linkers, as summarized in Table 4.
Cross-linkers can be contained in the continuous fluid, as described in Zhang et al.  Ca (CH3COO)2 (2 wt%) was dissolved in soybean oil, the continuous fluid. In the microfluidic device detailed previously , Na-alginate (2 wt%) droplets were generated by flow-focusing (Fig. 15a) in oil/Ca (CH3COO)2, with surfactant (Span 80, 3 wt%). Ca (CH3COO)2 diffused and dissolved in Na-alginate droplets along the channel (Fig. 15b), causing external gelation on-chip. Finally, Ca-alginate hydrogel microparticles were collected in oil (Fig. 15c). They showed better stability in an aqueous medium and had a higher Young’s modulus compared with those produced by internal gelation (III.1.2.). Consequently, stronger gelation was achieved by external gelation. However, increasing the concentration of Ca (CH3COO)2 in soybean oil caused clogging in the microchannel . Thus, it is difficult to vary the rate of gelation of microparticles.
One way to limit channel clogging is to make the cross-linkers diffuse slowly in Na-alginate droplets. Thus, Liu et al. used emulsion fluids to introduce cross-linker s. A glass-based microfluidic device was used, with channels modified so as to be hydrophobic. Droplets of Na-alginate (3 wt%) were first generated in corn oil at the first flow-focusing channel (Fig. 16a). The emulsion of CaCl2, containing CaCl2 droplets in corn oil (with surfactant SY-Glyster CRS-75), was injected downstream of the cross-junction. The contact between CaCl2 and Na-alginate droplets caused ionic crosslinking, leading to gelation. Ca-alginate hydrogel microparticles were obtained. However, it was found that the microparticles could easily be deformed (Fig. 16b-A, b-C) by several parameters, such as the mass fraction of the aqueous CaCl2 solution in emulsion (W). Deformation occurred when the value of W was too high or too low, so that an optimal value of W was required for homogeneous spherical microparticles (Fig. 16b-B). The morphology and homogeneity of microparticles also varied with flow rates and surfactant concentrations. Lacking flexibility, this strategy is thus not appropriate for producing spherical hydrogel microparticles. Moreover, generating small particles requires reducing the channel size, involving a risk of droplet coalescence before gelation in CaCl2 emulsion.
To avoid reducing the channel size, partially miscible fluids can be used as the dispersed and continuous fluids. Sugaya et al. used methyl acetate as the continuous fluid . Na-alginate (0.025–0.15 wt%) droplets were generated in methyl acetate (no mention of surfactant usage) by flow-focusing. In the following channel, because of the solubility of water in methyl acetate (8 wt%), water dissolved gradually from the droplets into methyl acetate. Thus, the droplets shrank and became more concentrated downstream. CaCl2 solution (1 M) was then injected by side channels and flowed with the droplets by co-flow. Calcium cations diffused into the droplets, inducing on-chip external gelation of alginate. Finally, spherical Ca-alginate hydrogel microparticles with a diameter of less than 20 μm were obtained. In this strategy, after CaCl2 fluids were introduced, two competing processes occurred simultaneously in each droplet: gelation and shrinkage of droplets. The competition between gelation and shrinkage is not discussed in this article. However, the results indicate that extremely small droplets tend to approach the channel wall after shrinkage. With the CaCl2 fluid, after gelation, Ca-alginate hydrogel microparticles adhere to the channel wall.
Adhesion to the channel and coalescence of microparticles can be avoided thanks to progressive addition of the cross-linker. Pittermannová et al. used as continuous fluid 1-undecanol, whose water-solubility is 2.7 vol% . The experiment was carried out in a PDMS-based microfluidic device (Fig. 17a). An aqueous alginate solution (1 wt%) was first injected. After 1-undecanol (with 5 wt% surfactant Abil Em 90), shown as “oil” in Fig. 17a, was injected into the flow-focusing channel, droplets of alginate were formed. CaCI2 (2 wt%) was dispersed in another fluid, 1-undecanol with 5 wt% surfactant Abil Em 90, yielding an emulsion. This emulsion was injected after droplet generation, through successive channels (Fig. 17a). Hence, the droplets were increasingly separated from each other, avoiding coalescence. Moreover, they were surrounded by more and more CaCl2, increasing gelation, and by more and more 1-undecanol, increasing diffusion of water. Thus, gelation and shrinkage of droplets occurred gradually and simultaneously. According to the authors, this procedure avoids the droplet generation instability caused by pre-gelation. However, spherical hydrogel microparticles (Fig. 17b) were only obtained using certain flow rates and calcium concentrations. Otherwise, the microparticles were slightly deformed (Fig. 17c) or collapsed (Fig. 17d), which was explained using a core-shell model . Unfortunately, this explanation does not take into account the change in alginate concentration in the droplets due to water extraction, a factor which is bound to impact deformation. Simply prolonging the water extraction process before introducing cross-linkers, as done by Sugaya et al., could avoid deformation .
Off-chip introduction of cross-linkers
Other strategies of external gelation introduce cross-linkers “off-chip”, i.e., during droplet collection. Ca (CH3COO)2 or CaCl2 are used as cross-linkers and the collection bath procedure depends on the strategy, as summarized in Table 5.
Hu et al.  studied the influence of external gelation conditions on the shape of microparticles. Na-alginate (1.5 wt%) droplets were first generated in n-decanol with surfactant (Span 80, 5 wt%), using concentric glass capillaries in co-flow geometry (Fig. 18A). For off-chip external gelation, droplets were collected in a two-phase gelation bath: the upper phase of n-decanol with surfactant (Span 80, 5 wt%) containing CaCl2 (15 wt%) allowed for pre-gelation of alginate; the bottom phase, an aqueous solution of barium acetate (15 wt%), strengthened the gelation. Glycerol (0–70 wt%) was added to the bottom phase to regulate viscosity. Ca-alginate hydrogel particles of different shapes (Fig. 18B) were obtained by varying gelation conditions such as the interfacial energy γCD, the concentration and type of surfactant, the height h between the end of the capillary and the surface of the gelation bath, and the viscosity of the bottom phase in the gelation bath. The shape of microparticles was shown to depend on forces applied to the surface of droplets when they passed through the interface in the gelation bath. The force from γCD maintains the spherical form of droplets, while the viscous force causes deformation. The final shape resulted from the overall effect of these two forces . As can be seen in this strategy, droplet collection is accompanied by the consumption of the two different cations. Thus, to obtain a large quantity of microparticles, these cations should be replenished to ensure that each droplet undergoes sufficient gelation. However, in practice, when and how to replenish them remains an issue.
To avoid the problem of replenishing the bath with the two cations, we collected Na-alginate droplets in an aqueous solution containing CaCl2 (Fig. 19a) without pre-gelation. Na-alginate (0.006–1 wt%) droplets were first generated in DMC in a T-junction and Teflon-like capillaries (IDEX Health and Science), without using surfactant . Because water is slightly soluble in DMC, 3 wt%, water diffused gradually from droplets into DMC, causing the droplets to shrink as they passed through the channel (point A to B in Fig. 19a). Thus, droplet size reduced to below 100 μm. Furthermore, since alginate dissolution in the continuous fluid is negligible , with the loss of water, the alginate concentration in droplets increased. Then, the channel outlet (point B in Fig. 19a) was immersed in an aqueous solution of CaCl2 (0.1–1 wt%). An interface was created at the channel outlet (Fig. 19b) because of the non-total miscibility between DMC and water. Na-alginate droplets passed through the interface and entered the CaCl2 solution, leading to off-chip external gelation. After gelation, Ca-alginate hydrogel microparticles were droplet-shaped (Fig. 19c) and tadpole-shaped (Fig. 19d), as in Fig. 18B, b-c. The shape of the microparticles varied with the flow rates, the concentration of Na-alginate and that of CaCl2. It is likely that the deformation mechanism involved the forces applied to droplets at the interface, as explained by Hu et al. .
To improve the spherical shape of microparticles, we collected droplets in a bath of the continuous fluid, i.e., DMC. Hence, the droplets continued to shrink and were finally transformed into spherical condensed Na-alginate microparticles, not yet gelated. For the gelation of the microparticles, the bath of DMC was first evaporated. Then, an aqueous solution of CaCl2 (0.5–10 wt%) was added to the dried Na-alginate microparticles, inducing off-chip external gelation. Observations showed that this process was accompanied by the swelling of the microparticles without deformation (Fig. 20b-c). In the end, spherical Ca-alginate hydrogel microparticles were obtained. They were insoluble in water, indicating efficient gelation. Moreover, the concentration of CaCl2 had no significant effect on the size of the Ca-alginate microparticles. Since no surfactant is used in this method, no surfactant-removing step is needed, which simplifies the process. However, the quantity of microparticles produced is limited by the need to avoid droplet coalescence.
In conclusion, external gelation of alginate can be performed both on-chip and off-chip. For on-chip external gelation (Table 4), cross-linkers can be added to the continuous fluid, i.e., the oil. However, only limited concentrations can be used, since most are slightly soluble in oil. Therefore, introducing cross-linkers in emulsion form increases the quantity of alginate available for gelation. However, the particles are large and deformed. On the other hand, if partially miscible fluids are used, an aqueous solution of cross-linkers can be injected after droplet shrinkage. Particle size is reduced but the gelation is too rapid. Things can be improved by dissolving cross-linkers in an oil-based emulsion, introduced in small quantities but repeatedly.
For off-chip external gelation (Table 5), cross-linkers are introduced into the collection bath. A two-phase collection bath permits pre-gelation of Na-alginate droplets before gelation. However, the particles are large. Droplet size can be reduced to below the channel diameter by using partially miscible fluids for droplet generation, and the droplets can then be collected directly in the dispersed phase containing the cross-linker. However, the microparticles are deformed. To further reduce particle size and improve gelation, our solution is to perform two-step collection. Thus, off-chip external gelation can be used to produce shape-controlled and size-controlled microparticles.