Everything in its right place: controlling the local composition of hydrogels using microfluidic traps

Michael Kessler a, Hervé Elettro b, Isabelle Heimgartner a, Soujanya Madasu a, Kenneth A. Brakke c, François Gallaire b and Esther Amstad *a
aSoft Materials Laboratory, Institute of Materials, École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland. E-mail: esther.amstad@epfl.ch
bLaboratory of Fluid Mechanics and Instabilities, Institute of Mechanical Engineering, École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland
cMathematics Department, Susquehanna University, Selinsgrove, PA 17870, USA

Received 7th July 2020 , Accepted 27th October 2020

First published on 28th October 2020

Many natural materials display locally varying compositions that impart unique mechanical properties to them which are still unmatched by manmade counterparts. Synthetic materials often possess structures that are well-defined on the molecular level, but poorly defined on the microscale. A fundamental difference that leads to this dissimilarity between natural and synthetic materials is their processing. Many natural materials are assembled from compartmentalized reagents that are released in well-defined and spatially confined regions, resulting in locally varying compositions. By contrast, synthetic materials are typically processed in bulk. Inspired by nature, we introduce a drop-based technique that enables the design of microstructured hydrogel sheets possessing tuneable locally varying compositions. This control in the spatial composition and microstructure is achieved with a microfluidic Hele-Shaw cell that possesses traps with varying trapping strengths to selectively immobilize different types of drops. This modular platform is not limited to the fabrication of hydrogels but can be employed for any material that can be processed into drops and solidified within them. It likely opens up new possibilities for the design of structured, load-bearing hydrogels, as well as for the next generation of soft actuators and sensors.


Many natural materials display unique mechanical properties that are, at least in parts, a result of their locally varying composition. A prominent example is the mussel byssus cuticle, a stiff, but extensible coating that is five-fold harder than the fibrous core and hence protects it from abrasion.1 The cuticle is composed of granules that are embedded in a matrix. These granules are responsible for the abrupt changes in the composition and hence, the excellent mechanical properties of the cuticle.2–4 The ability of nature to controllably and abruptly change the composition of materials is related to their processing: natural materials are often produced from compartmentalized reagents that can be released into well-defined and spatially confined regions.5 Inspired by nature a large variety of bottom-up processes have been developed to fabricate synthetic soft materials with locally varying compositions. Well-defined micrometer-sized structures with feature sizes down to 10 μm can be introduced into soft materials using direct laser writing6 or lithography techniques that take advantage of digital micromirror devices (DMD's).7 However, these techniques are often limited to sample sizes below 100 μm. Macroscopic materials possessing feature sizes down to 200 μm can be produced with optical projection lithography, using a laser and DMD.8 Similarly, direct ink writing allows the fabrication of macroscopic samples that optionally can be composed of multiple materials with a resolution of 100–200 μm.9 Resolutions of 200 μm,10 100 μm11 or even 35 μm12 can be obtained if supporting matrices consisting of granular microgels or shear-thinning self-healing hydrogels are employed. However, the composition of materials made with these methods cannot be abruptly changed. Abrupt changes in the composition have been reported for 3D-printed continuous filaments, yet with a reduced resolution.13 Methods to fabricate composites made of multiple different materials that possess well-defined microstructures and whose composition can be abruptly and controllably changed remain to be established.

A possibility to gain true control over the local composition of soft materials is the immobilization and solidification of reagent-loaded drops in obstacle-free microfluidic devices. Drops can be immobilized in obstacle-free Hele-Shaw cells for example using acoustic waves14 or electrostatic potentials.15,16 However, such active methods rely on complicated microfabrication methods and require external energy sources, such that up-scaling becomes tedious. An elegant, passive method to immobilize drops in an obstacle-free Hele-Shaw cells is the use of microfluidic traps. This method relies on the minimization of the surface energy of drops: when drops with a diameter larger than the channel height are squeezed into the cell, they deform into pancakes, thereby increasing their surface area and hence total energy, E = γA; here γ is the interfacial tension and A the drop surface area. Squeezed drops can relax if they encounter a trap which is a restricted area where the channel height is locally increased. In this case, the surface energy of drops decreases. The gradient in surface energy between the squeezed and partially relaxed state of drops leads to a trapping force that opposes the drag force and hence, enables immobilization of drops. Studies conducted in Hele-Shaw cells demonstrated that larger traps produce larger trapping forces while larger drops experience larger drag forces.17 Devices that immobilize drops in microfluidic traps have been employed to monitor the evolution of cells,18,19 chemical reactions,18 or to investigate cross-talks between emulsion drops.20,21 However, this strategy has never been used to design macroscopic materials with well-defined locally varying compositions.

In this paper we introduce a microfluidic trapping device that enables the design of hydrogel sheets with locally varying compositions. To achieve this goal, we study the influence of the geometry of traps on their trapping strengths. We demonstrate that the trapping strength depends on the depth, area, and in-plane geometry of the trap. Based on these experimental results we introduce a mathematical model and perform simulations to predict the trapping strength. These tools facilitate the design of Hele-Shaw cells containing traps with different trapping strengths that enable the immobilization of different types of drops at well-defined locations. To demonstrate the potential of this device, we fabricate different microstructured hydrogel sheets with locally varying compositions.


The fabrication of materials whose composition can be locally tuned with a tight control over the type and location of the different components requires means to selectively immobilize compartments such as drops. To achieve this goal, we design traps with different trapping strengths. Different types of traps with varying trapping strengths can then be arranged on the same Hele-Shaw cell to selectively immobilize different types of drops, as schematically shown in Fig. 1a. We take the minimum flow rate where immobilized drops are released from their trap, the critical flow rate QC, as a measure of the trapping strength. To quantify QC, we design a 5.1 mm long, 1.5 mm wide, and 55 μm high microfluidic chamber that contains four traps in a row, as schematically shown in Fig. 1b. Water-in-oil drops with an average volume of 0.9 ± 0.04 nL are produced in a microfluidic flow focusing junction that is located upstream of the main chamber. We produce drops with a diameter much larger than the channel height h. These drops remain squeezed and attain a pancake-shape when entering the main chamber. When they encounter a trap, they relax into it to minimize their surface energy, as schematically shown in the steady-state Surface Evolver simulation22 in Fig. 1c. As a result of the decreased surface energy, the drops remain trapped. When all four traps are occupied with drops, we stop their production and set the flow of the continuous phase to zero, as shown in the optical micrograph in Fig. 1di. When we increase the flow rate of the surrounding oil but keep Q < QC, drops deform in the direction of the flow but find an equilibrium position such that they remain trapped, as shown in the Fig. 1dii. When Q > QC, drops are removed from their trap, as shown in Fig. 1diii and in ESI Movie S1.
image file: d0lc00691b-f1.tif
Fig. 1 Microfluidic traps with different trapping forces. a) Schematic illustration of a microfluidic device that contains two types of traps possessing different trapping strengths to selectively immobilize different batches of drops. b) Sketch of a microfluidic device used to measure the critical flow rate of traps, composed of a flow focusing junction that is connected to a main chamber containing four identical traps. c) Side view of a trap occupied by a drop that is simulated with Surface Evolver. The relevant dimensions are defined as channel height h, trap depth p, trap width W, radius of the drop in the channel R, and radius of curvature of the drop r. d) Time-lapse optical microscopy images of a trapped drop (i) under static conditions and when the surrounding oil flows at a rate (ii) below and (iii) above QC.

Influence of trap width

To design traps with varying trapping strengths, we investigate the influence of the dimensions of traps on their trapping strength. We employ circular traps, which have been shown to efficiently trap drops,17,23 and vary their width W. The critical flow rate increases with increasing trap diameter for traps with W ≤ 100 μm and reaches a plateau thereafter, as summarized in Fig. 2a. To investigate the origin of the independence of the critical flow rate with respect to the trap diameter for large trap diameters, we image the drops in the traps. Only a fraction of the projected drop area occupies the trap if W ≤ 100 μm. This is in stark contrast to traps with diameters exceeding 100 μm where the entire projected drop area is contained within the trap such that the trapping force, Fγ, remains within experimental error constant, as shown in Fig. 2a and b.
image file: d0lc00691b-f2.tif
Fig. 2 Trapping regimes. a) Critical flow rates QC as function of the trap diameter W for trap depth p = 30 μm. b) Optical microscopy images of drops that are immobilized in traps with increasing W. The traps are highlighted with a red circle for better visibility. c) Surface Evolver renderings for the three trapping regimes.

Simulations of drop shapes at equilibrium

The magnitude of the trapping force is related to the shape drops attain when relaxing into traps.23 To investigate this parameter, we simulate trapped drops in steady state using Surface Evolver. At equilibrium, the pressure jump ΔP across the interface of a drop must be constant throughout the drop surface. According to the Laplace equation, the curvature scales inversely with ΔP such that ΔP must also be constant throughout the drop surface. The equilibrium curvature of the in-trap drop interface is hence equal to the curvature of the drop interface inside the channel. For Rh, this can be approximated as r = 8hR/(8R + πh);24–26 with R the radius of the pancake in the channel. We herein consider this approximation to be also valid for Rh.

Our simulations reveal three different trapping regimes that depend on the ratio of W to the radius of curvature r: if r > W/2, the part of the drop that relaxes into the trap forms a spherical cap, as shown in Fig. 2c to the left; we color-coded this regime orange. In this case, the deformation of the drop into the trap is limited by r, as has been described earlier.23 By contrast, if r < W/2, the trap is wide enough for the drop to fully enter such that it touches the trap ceiling and flattens against it. The part of the drop contained in this type of traps attains a shape what we call a dome, shown in Fig. 2c in the middle; we color-coded this regime blue. In this regime, the deformation of the drop into the trap is limited by the trap depth p. If R < W/2, the projected area of the drop fully lies inside the trap and the drop attains the shape of a pancake that is much more relaxed compared to that in the channel, as shown in Fig. 2c to the right; we color-coded this regime green.

Influence of trap depth

To test the validity or our categorization of the trapping regimes, we vary the trap depth p and evaluate the critical flow rate QC as a function of W. We expect QC to be independent of p if the drops form a spherical cap where 2r > W. By contrast, we expect QC to increase with the trap depth p if 2r < W and 2R < W as in these cases the degree of relaxation of drops into traps increases with increasing trap depth. Indeed, if 2r > W, we do not observe any influence of p on QC, as shown by the orange curves in Fig. 3a. By contrast, if 2r < W and 2R < W, QC increases with increasing p, as shown by the green and blue curves in Fig. 3a. These results are in excellent agreement with our expectation and support our proposed trapping regimes.
image file: d0lc00691b-f3.tif
Fig. 3 Influence of trap depth p on trapping. a) Experimentally measured critical flow rates as a function of p for W = 40 (orange triangles) and 70 μm (orange circles), where trapped drops form spherical caps, W = 100 μm, where trapped drops form domes (blue), and W = 150 μm (green circles) and 200 μm (green triangles), where trapped drops form pancakes. The solid lines represent linear fits to the results. b) Volume fraction of the drop in the trap as a function of p for W = 70 μm (orange) and 100 μm (blue). Filled symbols represent experimental results, empty symbols results from Surface Evolver simulations, and crosses results from the mathematical model for a drop volume that is equal to the one used for simulations. The solid lines are linear fits to the experimental results. Insets: Surface Evolver renderings of trapped drops for p = 20 μm (left) and 40 μm (right) and W = 100 μm (top) and 70 μm (bottom).

Our and others23 results suggest that the trapping force is influenced by the decrease of the surface energy of a drop upon its relaxation into a trap. An interesting physical proxy to quantify the drop deformation into the trap is to determine the volume fraction of the drop contained in the trap. Although this parameter has a priori no direct physical meaning and its relation to QC cannot be generalized, it allows to determine to what extent a drop relaxes into a trap of a given depth. This information is acquired by imaging a drop when it is squeezed in the Hele-Shaw cell and while it is trapped. We calculate the volume of the pancake contained in the Hele-Shaw cell with our model by measuring its radius R from optical microscopy images and taking the known value of the channel height h, as detailed in the ESI. The volume of the drop contained in the trap is calculated from the volume difference of the pancakes located in the channel between the trapped and untrapped states. For 2r > W, the drop volume contained in traps is independent of p, as shown by the orange solid symbols in Fig. 3b. By contrast, if 2r < W, the drop volume contained in traps increases with p, as shown by the blue solid symbols in Fig. 3b. To test our findings, we evaluate the volume fraction of drops contained in traps with varying p using Surface Evolver simulations and a mathematical model, as detailed in the ESI. These results are in excellent agreement with our experimental ones, as shown in Fig. 3b by the empty symbols for simulations and crosses for the modelling. They confirm that the extent of relaxation of drops in traps is independent of p if 2r > W, but increases with increasing p if 2r < W. Further, for a fixed trap in-plane geometry, the volume fraction of the drop in the trap scales with the change of its surface area. As a result, for a fixed in-plane trap geometry, the critical flow rate and the volume fraction of the drop in the trap show similar slopes, as exemplified in blue for large, circular traps in Fig. 3a and b. From this finding, we deduce that the range over which the trapping strength can be varied is much wider if the trap diameter exceeds 2r. Note that this condition is not valid for very shallow traps, as discussed in the ESI.

Calculations of surface area change upon trapping

To facilitate the design of traps possessing different trapping strengths it would be beneficial to predict QC for different trap geometries and drop sizes. This would require means to estimate the change in surface area that drops undergo when they enter a trap. To do so we initially only consider the untrapped, squeezed pancakes located in the channel. Different models have been proposed to calculate the surface area and volume of drops squeezed into a pancake shape within a Hele-Shaw cell.27,28 These models provide accurate results of the surface area either for heavily27 or minimally squeezed drops,28 but not for both cases. To predict QC, we need a model that captures the drop shape for a wide range of aspect ratios. To achieve this goal, we describe the drop as a cylinder surrounded by a half-torus, as sketched in the inset of Fig. 4a. The volume of the drop is calculated as
image file: d0lc00691b-t1.tif(1)
and the surface area is calculated as

image file: d0lc00691b-t2.tif (2)
here, r is defined as the channel height divided by two, h/2, and xa as the difference between R and r. For simplicity, we consider a constant radius of curvature r. Our model is in excellent agreement with the surface area data obtained from Surface Evolver simulations: the deviation of these two data sets is below 0.3% for channel heights varying between 10 μm, where drops are deformed to extremely high aspect ratios, and 120 μm, where drops are not confined and attain a spherical shape, as shown in Fig. 4b. Indeed, this simple model shows excellent agreement over a wide range of channel heights that has not been achieved with the previously reported models, as shown in Fig. 4a.

image file: d0lc00691b-f4.tif
Fig. 4 Surface area model for drops squeezed in a Hele-Shaw cell. a) Surface area as a function of the channel height for the model proposed by Lv27 using elliptic integrals (green), the model proposed by Nie et al.28 using a truncated sphere (blue), the model proposed in this work using a disk with surrounding half-torus (black), and Surface Evolver simulation data (red, empty symbols). Inset: Schematic illustration of a drop squeezed in a channel of height h with indicated radius of the resulting pancake R, the length of contact between the drop and the channel xa, and the radius of curvature r. b) A comparison of our model (black, solid symbols) with simulation data obtained from Surface Evolver (red, empty symbols) shows excellent agreement of the two data sets with errors below 0.3% over a wide range of channel heights.

To describe the total surface area of trapped drops, that includes the surface area of the pancake and the cap that resides in the traps, we extend our model using geometric arguments, as detailed in the ESI and Fig. S1. The calculated values are in excellent agreement with values obtained from Surface Evolver simulations, as shown in Fig. S1. These estimations greatly facilitate the design of optimized traps. The agreement between our model and simulations further indicates that the approximation for r is justified to describe the relaxation of drops into traps, even for cases where Rh.

Influence of the in-plane trap geometry

We expect the in-plane trap geometry to be another parameter that critically influences the trapping strength. To test this hypothesis, we quantify QC as a function of the number of corners of the trap, N. We design five different polygonal traps that have identical surface areas and N = 3, 4, 5, 6, or ∞ corners, as shown in the optical micrographs in Fig. 5a. We consider the initial circular trap as a limiting case with an infinite number of corners. For traps with a depth of p = 15 μm, the critical flow rate increases by 39% from 45.3 ± 4.8 μL h−1 for N = 3 to 63.1 ± 3.0 μL h−1 for N = ∞, as indicated by the green symbols in Fig. 5b. The increase in QC with increasing N becomes even more pronounced for deeper traps: if p is increased to 40 μm, QC increases by 60% from 86.9 ± 8.3 μm to 139.4 ± 15.6 μm for N = 3 and ∞ respectively, as shown in Fig. 5b. To test if the stronger increase in QC for traps with p = 40 μm is related to the higher absolute values of QC of deeper traps, we normalize all values of QC with the corresponding one measured for circular traps, QC/QC. Indeed, all normalized data falls onto a master curve where the critical flow rates QC of triangular and square traps are approximately 30% lower than those of circular and hexagonal counterparts, as shown in Fig. 5c. Triangular and square traps possess large gutters that strongly influence QC. By contrast, hexagonal and circular traps possess very small or no gutters that do not significantly influence QC. Pentagonal traps have intermediate gutters that influence QC, but only weakly.
image file: d0lc00691b-f5.tif
Fig. 5 Influence of the in-plane geometry on the critical flow rates of microfluidic traps. a) Optical micrographs of polygonal traps with N = 3, 4, 6, and ∞ sides, each one immobilizing a drop. b) QC as a function of N for trap depths p = 15 μm (green), 20 μm (red), 30 μm (blue), and 40 μm (black). c) QC normalized by QC for different trap heights. d) Surface Evolver renderings of a triangular and a spherical trap holding drops in steady state. Drops in triangular traps retract from the sharp corners. The renderings are for illustration purposes only and are not to scale. e) Surface Evolver simulation data of the surface area difference between the trapped and untrapped state normalized by the value obtained for the circular trap for p = 15 μm (green), 20 μm (red), 30 μm (blue), and 40 μm (black). f) QC as a function of the modified number of sides N* for traps with a depth p = 30 μm with (black) and without (blue) additional gutters. The black dotted line provides guidance to the eye. Insets: Optical micrographs of traps with (black) and without (blue) additional gutters for N* = ∞.

To investigate the reason for the increase in QC with increasing number of corners of the traps, we simulate the relaxation of drops into traps with different in-plane geometries in steady state using Surface Evolver. Drops that are immobilized in triangular traps strongly retract from the corners to maximize the radius of curvature and hence, to minimize the Laplace pressure within the drops, as shown in Fig. 5d. The interface position can be understood as an equilibrium location in a geometrical singularity, in this case the corners, where the Laplace pressure within the drop is minimized. As a result of this retraction, the decrease in surface area is much smaller for drops immobilized in triangular traps than that of drops immobilized in circular traps, as shown in Fig. 5e. The difference in surface area between trapped and untrapped drops increases with increasing N and is most likely a contributing factor for the observed increase in QC.

We hypothesize that another contributing reason for the observed decrease in QC with decreasing N could arise from increased drag forces. The size of gutters that form when drops retract from the trap corners increases with decreasing number of sides. As the size of gutters increases, more continuous phase can flow into them. We expect this effect to impart higher drag forces on the drop, thereby decreasing QC. To test this hypothesis, we add gutters to the polygonal traps with N = 4, 6, ∞ by increasing their length threefold, as exemplified in the inset of Fig. 5f for the circular trap. For these trap geometries, we use N* to refer to the number of corners of the polygons and modified, elongated geometries. We orient the elongated traps to minimize the modification of the drop relaxation compared to the regular polygonal case. Additional gutters do not measurably influence QC of square traps. By contrast, they significantly decrease QC of hexagonal and circular traps, as shown in Fig. 5f. Triangular and square traps possess relatively large gutters and hence increased drag forces compared to circular traps, such that the addition of even larger gutters does not measurably alter QC. By contrast, hexagonal and circular traps possess gutters that are too small to significantly increase the drag force exerted on drops. If large gutters are added to these geometries, the drag force increases and hence, QC decreases. These results indicate that the lower values of QC for triangular and square traps can at least partially be assigned to the drag force exerted on drops by the continuous phase flowing into the gutters. Note that the orientation of triangular and square traps relative to the flow of the continuous phase does not measurably influence QC, as shown in Fig. S2. These results demonstrate that QC can also be tuned over a considerable range with the in-plane geometry of traps.


The trapping force, Fγ = γΔA/d, scales linearly with the change in surface area between trapped and untrapped drops, ΔA; here d is the characteristic length scale over which the surface area changes.23 To test if this scaling applies to the traps investigated here, we calculate ΔA using the model derived in the ESI and divide it by d, which we take as the length of the trap in the direction of flow for drops that are not entirely contained in the traps. For drops whose projected area is fully inside the trap, we take d as the sum of the pancake radius R of a drop inside the trap and that of a drop squeezed in the channel. The surface tension γ is kept constant and measured to be 4.5 mN m−1 for all experiments. Hence, we only consider the geometric components of Fγ = γΔA/d. All our measured data points fall reasonably well onto a Master curve when we plot QC as a function of ΔA/d, as shown in Fig. 6, justifying our approach. This result indicates that we can predict QC using simple geometric arguments, thereby significantly facilitating the design of microfluidic trapping devices.
image file: d0lc00691b-f6.tif
Fig. 6 Influence of the trapping force on the critical flow rate QC. QC as a function of the surface area change ΔA divided by the characteristic length over which the change in surface energy takes place, d, for circular traps with W = 40 μm (dark orange circles), 70 μm (light orange circles), 100 μm (blue circles), and 150 μm (green circles), triangular traps (magenta triangles), and square traps (black squares). Filled symbols represent ΔA/d values obtained from the model and empty symbols represent ΔA/d results obtained from Surface Evolver simulations.

Application to the design of hydrogels with locally varying compositions

To demonstrate the potential of microfluidic trapping to design materials with locally varying compositions, we fabricate a microfluidic device with a 4 mm wide and 19.3 mm long Hele-Shaw cell that contains 570 traps. The device features two different types of traps, larger hexagonal traps with a critical flow rate QC,H = 100 ± 15 μl h−1 and smaller square traps with a critical flow rate QC,S = 34 ± 4 μl h−1. We form a first batch of aqueous drops that is labelled with a red fluorescing dye and immobilize the drops in all traps, as shown in Fig. 7a. We subsequently rinse the Hele-Shaw cell at a flow rate QC,S < Q < QC,H, so that all drops that are immobilized in hexagonal traps remain trapped while the ones immobilized by square traps are removed, as shown in Fig. 7b and ESI Movie S2. We convert the trapped drops into hydrogel particles by exposing the device to UV light. A second batch of aqueous drops labelled with fluorescein is formed and these drops are immobilized in the remaining traps, as shown in Fig. 7c. Untrapped drops are removed by rinsing the Hele-Shaw cell with Q < QC,S. The trapped drops are again converted into hydrogel microparticles through UV illumination, as shown in Fig. 7d. Note that our microgels are rather stiff such that we do not observe any deformation of these particles during the introduction of the matrix precursor. Finally, we exchange the continuous oil phase with an aqueous phase containing poly(ethylene glycol)diacrylate (PEGDA) and a photoinitiator, as shown in Fig. 7e and ESI Movie S3. We solidify this solution through UV illumination to obtain an intact hydrogel sheet whose composition can be deliberately changed with a high spatial precision, as shown in Fig. 7f and g. Hydrogel sheets fabricated in such microfluidic devices are mechanically stable and can be further processed and tested, as exemplified in Fig. 7h.
image file: d0lc00691b-f7.tif
Fig. 7 Fabrication of a hydrogel sheet with locally varying compositions. Optical micrographs of a microfluidic trapping device (a) as it is filled with a first batch of drops containing a red fluorescing label, (b) after the excessive drops had been rinsed out and the trapped drops have been converted into particles, (c) as it is filled with the second batch of drops, (d) after the excessive drops were rinsed out and the trapped drops are converted into particles, (e) as the continuous oil phase is replaced with an aqueous phase containing precursors of the matrix and (f) after the matrix material has been solidified to result in an integral hydrogel with locally varying compositions. (g) Fluorescence micrograph of an integral hydrogel sheet encompassing two different types of microgels. (h) Free-standing macroscopic hydrogel sheet.

To demonstrate the power and versatility of our platform to fabricate functional hydrogel sheets possessing well-defined microstructures, we design hydrogel sheets encompassing polyelectrolyte microgels. To achieve this goal, we produce aqueous drops loaded with the monomer AMPS (2-acrylamido-2-methylpropane sulfonic acid), a crosslinker and a photoinitiator, and immobilize them in large hexagonal traps. Untrapped drops are removed before the immobilized drops are converted into PAMPS particles by illuminating them with UV light to initiate the polymerization reaction within them, as shown in Fig. 8a. The oil is subsequently replaced with an aqueous matrix solution containing PEGDA and a photoinitiator. During this solution exchange, the polyelectrolyte particles swell by at least a factor three in in-plane diameter, as indicated with the red circles in Fig. 8a and b. After the matrix precursor is polymerized, the integral hydrogel sheet is removed from the microfluidic chip. To demonstrate its functionality, we immerse the microstructured sheet in a solution containing cresyl violet perchlorate, a positively charged dye. The dye is selectively adsorbed within the polyelectrolyte particles, as indicated by their change in color. This example demonstrates the versatility of this device in terms of materials that can be used to introduce micropatterns and functionality into hydrogel sheets.

image file: d0lc00691b-f8.tif
Fig. 8 Fabrication of a selectively adsorbing hydrogel sheet. Optical micrographs of (a) polymerized PAMPS particles that are surrounded by an oil phase and (b) PAMPS particles swollen in the matrix precursor that was subsequently polymerized. (c) Integral hydrogel sheet that has been removed from the device and immersed in a cresyl violet-containing solution. Cresyl violet is electrostatically attracted by the anionic microparticles, as indicated by the selective staining. The red circles in (a) and (b) highlight the borders of the PAMPS microgels (a) before and (b) after they have been swollen in the matrix precursor solution.

Our results demonstrate that our device is well-suited to fabricate micropatterned hydrogel sheets. Importantly, this technology is not limited to trapping water-in-oil emulsions but can also be used to trap oil-in-water emulsions. To demonstrate this feature, we form drops composed of a fluorinated oil that encompass surfactants. These drops are dispersed in an aqueous solution containing PEGDA and a photoinitiator. After the oil drops are trapped and untrapped drops are removed, we polymerize the PEGDA to obtain a hydrogel sheet possessing well-defined highly regular micropores, as shown in Fig. 9.

image file: d0lc00691b-f9.tif
Fig. 9 Fabrication of a porous hydrogel sheet from oil-in-water drops. Optical micrograph of a porous hydrogel sheet after it has been removed from the microfluidic device with the surface profile of a pore in the inset.


We introduce a microfluidic Hele-Shaw trapping cell that enables the fabrication of hydrogel sheets with locally varying compositions. Importantly, our method enables tuneable, abrupt changes in material composition on the 100 μm length scale, in stark contrast to most additive manufacturing techniques that rely on the continuous deposition of filaments. The abrupt compositional changes are achieved by selectively immobilizing different types of drops at well-defined locations using traps possessing varying trapping strengths. We demonstrate that the trapping force strongly depends on the area, depth, and in-plane geometry of the traps and hence can be tuned over a wide range. Importantly, this method is not limited to the fabrication of hydrogels but can be extended to many other types of materials. Its facile and passive approach facilitates upscaling such that this platform opens new possibilities to study fundamentals of fracture and crack propagation in soft, structured materials or, by adding functionality to well-defined regions, to design the next generation of sensors and actuators.

Materials and methods

Fabrication of the microfluidic devices

Microfluidic devices are produced from poly(dimethyl siloxane) (PDMS, Sylgard 184, Dow Corning, USA) using soft photolithography.29 The PDMS mold is bonded to a glass slide using oxygen plasma (Plasma Harrick, UK) to close the chamber. To render the surfaces fluorophilic, we inject a HFE7500-based solution (Novec, 3M, USA) containing 2% (v/v) trichloro-(1H,1H,2H,2H-perfluorooctyl)silane (Sigma-Aldrich, USA). The solution is kept in the channels for 15 minutes before they are dried with compressed nitrogen. To render the surfaces hydrophilic we inject a polyelectrolyte solution containing 2% (w/w) PDADMAC poly(diallyldimethylammoniumchloride) (Sigma-Aldrich, USA) in a 2 M sodium chloride solution. The solution is kept inside the channels for 1 hour before the solution is removed with compressed nitrogen. The channel height h is kept constant at 55 ± 3 μm for all the experiments. Its variation is measured using a VK-X360K 3D Laser Scanning Confocal Microscope (Keyence, Belgium). To study the influence of the trap dimensions on the trapping strength, we employ a 1.5 mm wide and 5.1 mm long microfluidic channel. The center-to-center distance between the traps is 800 μm. Our polygonal traps have an area of 7854 μm2, equivalent to the area of circular traps with diameter W = 100 μm. To fabricate hydrogel sheets, we use a 4 mm wide and 19.3 mm long Hele-Shaw cell that contains 570 traps. Hexagonal traps with an area of 7854 μm2 and square traps with an area of 3849 μm2 are used for strong and weak traps, respectively. These areas are equivalent to the areas of circular traps with diameters W = 100 μm and 70 μm, respectively. For the second and third applications we employ hexagonal traps with an area of 7854 μm2 to ensure strong trapping.

Formation of water-in-oil emulsion drops

Water-in-oil drops are produced using a microfluidic flow focusing device whose main channel is 100 μm wide and 55 μm tall.30 We employ an aqueous solution containing 15% (w/w) poly(ethylene glycol) (PEG, Mw = 6 kDa) as a dispersed phase and HFE7500 as a continuous phase. To ensure that drops do not break during any of the liquid exchange steps, we add 1% (w/w) of a fluorinated triblock surfactant (DiFSHJeffamine900) to the oil phase.31 The interfacial tension between the water and the oil phase is 5 mN m−1, as measured with the pendant drop method (Krüss, DSA30, Germany). The two phases are injected into the device at 100 μl h−1 and 500 μl h−1 for the water and oil phase, respectively, using two volume-controlled syringe pumps (Cronus Sigma 1000, Labhut, UK).

Measuring critical flow rates of traps

For every data point, experiments are performed on at least 2 different microfluidic chips to exclude that observed differences in trapping strengths are related to surface treatments. In every device, at least 6 drops are measured per geometry, resulting in at least 12 data points from which we calculate the average and standard deviation.

Fabrication of structured hydrogel sheets

Hydrogel sheets with locally varying compositions are fabricated using aqueous solutions containing 50% (w/w) PEGDA (PEG700-DA, Mw ≈ 700 Da, Sigma-Aldrich, USA) and 2% (w/w) 2-hydroxy-2-methylpropiophenone (97%, Sigma-Aldrich, USA), a photoinitiator. 33% (v/v) of fluorescent, red food colorant (Migros, CH), or 0.8 mg ml−1 fluorescein disodium salt (Carl Roth, Germany) are added to label different batches of drops. Drops are trapped before reagents contained in them are polymerized by exposing the device to UV light (320 nm < λ < 500 nm) (Omnicure S 1000, Lumen Dynamics, Canada). We keep the distance between the optical fiber, which has a diameter of 8 mm, and the microfluidic device constant at 5 cm and expose the sample to UV light for 3 minutes to ensure complete curing of the hydrogel. The matrix, composed of the same polymer and photoinitiator, is inserted using a syringe pump and crosslinked with UV light as described above. The device is opened using a surgical blade and the resulting hydrogel sheet is removed from the device. Hydrogel sheets are observed under a fluorescent microscope (Eclipse Ti-S, Nikon, Japan).

For the sheets with selective absorption we use an aqueous solution containing 20% (w/w) AMPS (2-acrylamido-2-methylpropane sulfonic acid, Sigma-Aldrich, USA) as a monomer, 3% (w/w) N,N′-methylenebisacrylamide (Sigma-Aldrich, USA) as a crosslinker, and 11 μl ml−1 2-hydroxy-2-methylpropiophenone (Sigma-Aldrich, USA) as a photoinitiator. The sheets are stained using an aqueous solution containing 0.8 mg ml−1 cresyl violet perchlorate (Sigma-Aldrich, USA).

For the fabrication of porous hydrogel sheets we render all surfaces hydrophilic. We employ the fluorinated oil phase described above as the dispersed phase and an aqueous phase containing 50% (w/w) PEGDA700, 2% (w/w) 2-hydroxy-2-methylpropiophenone, and 1% (w/w) Pluronic F127 Krill, a surfactant, as the continuous phase. The surface profile of the pore is recorded using a VK-X360K 3D Laser Scanning Confocal Microscope (Keyence, Belgium).

Surface Evolver simulations

Equilibrium states of drops in traps in steady state are evaluated using the Surface Evolver software.22 Surface Evolver is an efficient numerical solver that relaxes the shape of fluid interfaces in space-dependent energy landscapes under different constraints, such as conservation of volume or presence of solid boundaries. The contact angle was put to 179° for non-wetting conditions and gravity was neglected. Drops have a volume of 0.9 nL except if stated otherwise. The channel height is kept constant at 55 μm for all simulations. The evolution of the simulations is monitored using a built-in surface energy evolution diagram.

Conflicts of interest

There are no conflicts to declare.


We would like to thank Nicolas Burnand for experimental help. This work was financially supported by the Swiss National Science Foundation (SNSF, 200020_182662).

Notes and references

  1. M. J. Harrington, A. Masic and N. Holten-Andersen, Science, 2010, 328, 216–220 CrossRef CAS.
  2. L. V. Zuccarello, Tissue Cell, 1981, 13, 701–713 CrossRef.
  3. N. Holten-Andersen, T. E. Mates, M. S. Toprak, G. D. Stucky, F. W. Zok and J. H. Waite, Langmuir, 2009, 25, 3323–3326 CrossRef CAS.
  4. N. Holten-Andersen, G. E. Fantner, S. Hohlbauch, J. H. Waite and F. W. Zok, Nat. Mater., 2007, 6, 669–672 CrossRef CAS.
  5. T. Priemel, E. Degtyar, M. N. Dean and M. J. Harrington, Nat. Commun., 2017, 8, 14539 CrossRef.
  6. Y. Hu, Z. Wang, D. Jin, C. Zhang, R. Sun, Z. Li, K. Hu, J. Ni, Z. Cai, D. Pan, X. Wang, W. Zhu, J. Li, D. Wu, L. Zhang and J. Chu, Adv. Funct. Mater., 2020, 30, 1907377 CrossRef CAS.
  7. B. Özkale, R. Parreira, A. Bekdemir, L. Pancaldi, E. Özelçi, C. Amadio, M. Kaynak, F. Stellacci, D. J. Mooney and M. S. Sakar, Lab Chip, 2019, 19, 778–788 RSC.
  8. P. Kunwar, A. V. S. Jannini, Z. Xiong, M. J. Ransbottom, J. S. Perkins, J. H. Henderson, J. M. Hasenwinkel and P. Soman, ACS Appl. Mater. Interfaces, 2020, 12, 1640–1649 CrossRef CAS.
  9. W. Liu, Y. S. Zhang, M. A. Heinrich, F. De Ferrari, H. L. Jang, S. M. Bakht, M. M. Alvarez, J. Yang, Y. C. Li, G. Trujillo-de Santiago, A. K. Miri, K. Zhu, P. Khoshakhlagh, G. Prakash, H. Cheng, X. Guan, Z. Zhong, J. Ju, G. H. Zhu, X. Jin, S. R. Shin, M. R. Dokmeci and A. Khademhosseini, Adv. Mater., 2017, 29, 1–8 Search PubMed.
  10. T. J. Hinton, Q. Jallerat, R. N. Palchesko, J. H. Park, M. S. Grodzicki, H. J. Shue, M. H. Ramadan, A. R. Hudson and A. W. Feinberg, Sci. Adv., 2015, 1, e1500758 CrossRef.
  11. T. Bhattacharjee, S. M. Zehnder, K. G. Rowe, S. Jain, R. M. Nixon, W. G. Sawyer and T. E. Angelini, Sci. Adv., 2015, 1, 4–10 Search PubMed.
  12. C. B. Highley, C. B. Rodell and J. A. Burdick, Adv. Mater., 2015, 27, 5075–5079 CrossRef CAS.
  13. L. Ouyang, C. B. Highley, W. Sun and J. A. Burdick, Adv. Mater., 2017, 29, 1604983 CrossRef.
  14. A. Fornell, C. Johannesson, S. S. Searle, A. Happstadius, J. Nilsson and M. Tenje, Biomicrofluidics, 2019, 13, 044101 CrossRef.
  15. R. De Ruiter, A. M. Pit, V. M. De Oliveira, M. H. G. Duits, D. Van Den Ende and F. Mugele, Lab Chip, 2014, 14, 883–891 RSC.
  16. A. M. Pit, M. H. G. Duits and F. Mugele, Micromachines, 2015, 6, 1768–1793 CrossRef.
  17. P. Abbyad, R. Dangla, A. Alexandrou and C. N. Baroud, Lab Chip, 2011, 11, 813–821 RSC.
  18. E. Fradet, C. McDougall, P. Abbyad, R. Dangla, D. McGloin and C. N. Baroud, Lab Chip, 2011, 11, 4228 RSC.
  19. S. Sart, R. F. X. Tomasi, G. Amselem and C. N. Baroud, Nat. Commun., 2017, 8, 469 CrossRef.
  20. G. Etienne, A. Vian, M. Biočanin, B. Deplancke and E. Amstad, Lab Chip, 2018, 18, 3903–3912 RSC.
  21. P. Gruner, B. Riechers, B. Semin, J. Lim, A. Johnston, K. Short and J. C. Baret, Nat. Commun., 2016, 7, 10392 CrossRef CAS.
  22. K. A. Brakke, Exp. Math., 1992, 1, 141–165 CrossRef.
  23. R. Dangla, S. Lee and C. N. Baroud, Phys. Rev. Lett., 2011, 107, 1–4 CrossRef.
  24. C. W. Park and G. M. Homsy, J. Fluid Mech., 1984, 139, 291–308 CrossRef.
  25. R. Dangla, S. C. Kayi and C. N. Baroud, Proc. Natl. Acad. Sci. U. S. A., 2013, 110, 853–858 CrossRef CAS.
  26. M. Nagel, P. T. Brun and F. Gallaire, Phys. Fluids, 2014, 26, 032002 CrossRef.
  27. C. Lv, 2017, arXiv:1704.05830 [cond-mat.soft].
  28. Z. Nie, M. S. Seo, S. Xu, P. C. Lewis, M. Mok, E. Kumacheva, G. M. Whitesides, P. Garstecki and H. A. Stone, Microfluid. Nanofluid., 2008, 5, 585–594 CrossRef CAS.
  29. Y. Xia and G. Whitesides, Annu. Rev. Mater. Sci., 1998, 28, 153–184 CrossRef CAS.
  30. S. L. Anna, N. Bontoux and H. A. Stone, Appl. Phys. Lett., 2003, 82, 364–366 CrossRef CAS.
  31. G. Etienne, M. Kessler and E. Amstad, Macromol. Chem. Phys., 2016, 1–10 Search PubMed.


Electronic supplementary information (ESI) available. See DOI: 10.1039/d0lc00691b

This journal is © The Royal Society of Chemistry 2020