Structural characterization of fabricated devices
Figure 3 shows scanning electron microscope (SEM) images for a 3D nanomolded PMMA substrate in which a sinusoidal microfluidic channel was decorated with nanopillars. The corners of the microchannel were slightly rounded, which was inevitably produced during the 3D nanomolding process. However, nanopillars were well formed on the entire surface of the microfluidic channel, which indicates that 3D nanomolding is a feasible method to produce nanostructures on the walls of microfluidics.
Figure 4a shows SEM images of a 3D molded PMMA substrate in which ratchet microgratings were formed on the entire surface of the substrate. The 3D microchannel had the depth and width of 65.0 ± 2.0 μm and 129.5 ± 4.9 μm, respectively. The ratchet structures in the bottom center of the microchannel had the height of 3.5 ± 0.2 μm and were aligned at a ~45° angle with respect to the direction of the microchannel.
It is challenging to form an enclosed fluidic chip via bonding a cover plate to the 3D molded substrate due to the presence of microscale ratchet gratings on the top surface of the substrate. For this purpose, solvent-assisted bonding is preferred to thermal bonding which is the bonding method for most polymer microfluidic applications. In thermal bonding, an elevated temperature close to the PMMA glass transition temperature (Tg) of 105 °C is used [27]. At this temperature, PMMA chains are mobile and thus deformation of the molded microstructures can occur. In solvent-assisted bonding, on the other hand, a temperature lower than Tg of PMMA (85–95 °C) is used. Even though exposure to a solvent enhances softening of PMMA at the surface, only a portion of the exposed PMMA cover plate in contact with 3D molded PMMA substrate can deform during bonding, as demonstrated by Brown et al. [27]. Thus, ratchet gratings formed on the microchannel walls can survive during bonding. Figure 4b, c show the cross-section of a 3D microchannel after bonding to a plain PMMA cover plate. The ratchet gratings on the bottom surface and sidewalls of the microchannel were clearly visible, indicating that the solvent-assisted bonding process did not produce significant deformation on the integrated ratchet structures. The height and width of the 3D channel after solvent-assisted bonding were measured to be 60.5 ± 0.7 μm and 121.5 ± 1.0 μm, respectively, and the height of the integrated ratchet gratings in the bottom center of the channel was 3.7 ± 0.3 μm. Compared to the dimensions of the 3D microchannels prior to bonding, the dimensional variations occurring during the bonding process were <8%. Figure 4d shows a photograph for a complete microfluidic chip with the 3D microchannel after connecting and gluing the capillary tubes to the chip. Leak test results showed no leakage around the 3D microchannel, which in turn confirms that solvent-assisted bonding is a suitable method to form an enclosed fluidic system for 3D molded PMMA substrates.
Fluid mixing in 3D microchannels
After fabrication, DI water and a solution of a fluorescein dye in DI water were injected from separate inlets of the micromixers and their mixing behavior was studied using confocal microscopy. Figure 5 shows cross-sectional confocal microscopy images taken at different locations of 1, 3, 5, 10, 15, 20 and 28 mm from T-junction for four different mixing microchannels: (1) plain channel and 3D channels with (2) one side patterned (top side), (3) three sides patterned (bottom and side walls) and (4) four sides patterned (top, bottom and side walls). The volumetric flow rate in the mixing microchannel was 10 μL/min which corresponds to a Reynolds number of 1.85. For the plain channel, the dyed water and pure water moved side by side along the channel and thus mixing occurred mainly by diffusion, as can be seen in Fig. 5a. Advection was along the channel and was not useful for transversal mixing.
When one or more sides of microchannel walls were patterned with slanted micro ratchet gratings, the mixing mechanism became a combination of diffusion and advection. Ratchet gratings formed at a ~45° angle relative to the microchannel direction produced a transversal component of advection for fluids adjacent to the ratchet gratings, which in combination with the flow along the channel led to the formation of a helical flow. This can be clearly seen in Fig. 5b–d. The transversal component of flow increased the interfacial area between fluids and cut down the diffusion length for complete mixing. As a result, mixing was improved compared to that for the plain channel.
The helical flow became even more pronounced when three sides of the microchannel walls were patterned with continuous slanted ratchet gratings. Here, “continuous” means that line gratings were formed when two or more consecutive sides of a microchannel were virtually unfolded on a flat surface. Then, the gratings on two opposite sidewalls formed by single 3D molding were perpendicular to each other (Fig. 5c). The enhanced helical flow occurred because continuous ratchets on the left (right) sidewall in Fig. 5c also created transversal downward (upward) flow, which helps the fluids rotate faster as they move along the channel.
For the microchannel with all sidewalls patterned, the ratchet gratings on the top and bottom surface were parallel to each other while those on the two sidewalls were perpendicular to each other. Thus, the continuity of gratings was broken on the top surface. Consequently, an initial stretch of dyed fluid toward the pure water side occurred at both top and bottom surfaces, forming a stack of two helical flows one above the other in opposite directions. The helical flow formed on the bottom of the channel was stronger than the one on the top as it was strengthened by sidewall patterns. In this microchannel, almost complete mixing was achieved at a distance of 10 mm from T-junction.
We also studied the mixing behavior at a higher flow rate of 40 μL/min (Re = 7.4) and the fluorescence micrographs are shown in Fig. 6. A similar flow behavior was observed but the degree of mixing was lower compared to the results obtained at a 10 μL/min flow rate.
The degree of mixing can be quantified by taking the standard deviation of the normalized intensity from cross-sectional confocal images at different locations in the mixing microchannel. The standard deviation values for different 3D microchannels at 10 μL/min were shown in Fig. 7a. In general, the degree of mixing was in the increasing order of plain microchannel < microchannel with one side patterned < microchannel with three sides patterned < microchannel with four sides patterned.
Figure 7b compares the standard deviation versus location from the T-junction for the plain channel and the 3D channel with three sides patterned at two different flow rates of 10 and 40 μL/min. An increased flow rate for both cases increased the standard deviation (or decreased the degree of mixing) at the same location of the microchannels from T-junction. However, the degree of mixing in the 3D microchannel with three sides patterned at 40 μL/min was still significant and comparable to that in the plain microchannel at 10 μL/min, indicating that 3D microchannels are particularly useful for high flow rate microfluidic applications.
Comparing the position of dyed water front in the microchannel with three sides patterned for two different flow rates (Figs. 5c, 6c), the degree of the spiral rotation induced by the surface ratchet gratings was not much changed by varying the flow rate. At a high flow rate, the time for fluids to reach a location in a microchannel, i.e. the time for diffusion to occur, was short for both plain and 3D channels. However, as a result of the enhanced interfacial area between two fluids by transversal fluid motion in 3D microchannels, the reduction of mixing by an increased flow rate will not be significant compared to that in the plain microchannel.
Comparison with simulation results
The experimental results were compared with results from numerical simulations. The differences in the models used for simulations from the actual structures used for experiments are described in Sect. 2.6. Despite the differences, numerical simulations provide qualitative comparisons with the corresponding experiments. Figure 8 shows the concentration profile images for mixing of two water-based liquids at different locations along the plain and various 3D microchannels. Qualitatively, the simulation results were in good agreement with the experimental results in that surface ratchet gratings induced transversal motion of fluids. The rotation of fluids was enhanced when more sidewalls were patterned with continuous ratchet gratings (Fig. 8b, c). When ratchet gratings on the top surface were formed in parallel to ratchet gratings on the bottom surface, stretching of fluids occurred at both top and bottom surfaces, in agreement with experimental results (Fig. 8d). However, the degree of stretching on the bottom surface relative to that on the top surface was significantly reduced. This can be seen when the flow patterns for microchannels with four sides patterned (see at 3 mm in Figs. 5d, 6d and at 1390 in Fig. 8d) are compared at the distance showing a similar flow pattern for microchannels with three sides patterned (see at 3 mm in Figs. 5c and 6c and at 1390 µm in Fig. 8c). Thus, in the simulated case, the top helical flow seems to hinder the stretching of fluids on the bottom surface, which will be further discussed in the next.
Figure 9 shows the standard deviation of the concentration profiles shown in Fig. 8. Addition of ratchet gratings improved mixing significantly. Due to smaller dimensions of the mixing microchannel, larger size of integrated structures and a slower fluid velocity, mixing occurred in a short length compared to what we observed in the experiments. Mixing was most efficient when the sidewalls and bottom of the microchannel were patterned. However, incorporation of ratchet gratings to the top surface (microchannels with four sides patterned) did not improve mixing further with respect to the microchannel with three sides patterned, which is different from experimental results. We attribute this to the use of a diffusion coefficient value of 10−10 m2/s for the dyed water, which is lower than ~10−9 m2/s for a small molecule in water at room temperature [29]. Thus, mixing by diffusion at the interface of stretched liquids seems to be a rate-limiting process over the transverse flow of liquids induced by surface ratchets. In this case, the helical flow formed by the top surface ratchets prevents the interface area of two liquids from further expanding, resulting in a detrimental effect on mixing.
The deformation of microstructures on the surface of microchannel walls during fluidic experiments may be an issue due to a strong shear force applied. We have not performed an SEM investigation after the fluidic experiments. However, an inspection with an optical microscope before and after the fluidic experiments does not indicate any hint that the microstructures at the microchannel walls were deformed. We calculated the shear stress applied of the fluid using a simple Poiseuille model with a parallel plate with infinite aspect ratio in the cross-sectional dimensions where the shear stress is a function of volumetric flow rate, Q, channel dimensions (height h, width w, and length L), and fluid viscosity μ, as follows:
$$\tau = - \frac{12Q\mu }{{h^{2} w}}.$$
Putting the experimental conditions used in this study to the equation (Q = 40 μL/min; μ ~1 Pa s; h = 65 μm; and w = 130 μm) shows a shear stress in the range of ~73,000 Pa. The actual shear stress at the microchannel wall should be even smaller than this value. Even though it is a rough estimation, this value is significantly lower than the tensile stress values of PMMA, which is in the range of 48–76 MPa. Therefore, under the experimental conditions used in this study, it is not expected that the microgratings are deformed during the fluidic experiments.
Finally it should be noted that in most cases microfluidic designs are limited to planar, layer-by-layer geometries that are imposed by current lithography based techniques of microfabrication [20]. Using the 3D molding process, 3D patterns can be imprinted easily in a wide range of thermoplastic polymers used for low cost lab-on-a-chip applications and enclosed microfluidic devices with 3D patterns can be formed via solvent-assisted bonding. The current 3D molding process time is limited by curing of PDMS to form an intermediate stamp. However, the process time can be significantly used by using other UV curable polymers with similar cross-linking densities to have similar elastic properties since UV curing time is much shorter than thermal curing needed for PDMS. Various structures such as hierarchical micro and nanostructures with different geometries and dimensions can be patterned on the walls of microchannels and the cover plate enabling manipulating flow patterns. The direction of the patterns can also be controlled by setting different angle between brass mold protrusions and micropatterns on the surface of the thermoplastic polymer in the modified 3D molding process. Such advantages make the 3D molding process a suitable and powerful technique for fabricating micromixers.