A simple technique to design microfluidic devices for system integration

Mimi X. Yanga, Benjamin Wangb, Xiaolin Hua and H.-S. Philip Wong*a
aDepartment of Electrical Engineering, Stanford University, Stanford, USA. E-mail: mxy@stanford.edu
bDepartment of Mechanical Engineering, Stanford University, Stanford, USA

Received 12th September 2017 , Accepted 13th October 2017

First published on 16th October 2017


We present a robust method to fabricate a polydimethylsiloxane (PDMS) microfluidic device with critical channel features located near the periphery. The fabricated device has a window cutout, allowing close placement and integration of the channel with solid state circuits and devices. The flow characteristics of the fabricated device are shown to closely follow that of theoretical behavior, making it an appropriate alternative to a monolithic, cast-molded device. As an illustration of the use of this method, a novel integrated microfluidic-electronic system is demonstrated with the microfluidic channel bonded directly onto the surface of an integrated circuit while leaving access to probe pads on the chip's surface. This fabrication process has important implications in furthering the development of next generation microfluidic-based lab on chip systems.


1 Introduction

In recent years, lab on chip systems have found increasing use in personalized medicine, biosensensing and chemical synthesis.1–5 The microfluidic-based systems offer the powerful ability to precisely deliver reagents and samples directly to sensors, actuators, and transducers. Integration of microfluidics with chemical, mechanical, electrical, and other technology devices offers a range of solutions for various applications, provided that the fabrication and assimilation processes of the subsystems are compatible. Often, such systems require direct contact between sensing devices and the channel liquid.6

Versatile processes are required to build such integrated systems. In general, the fabrication requires the direct bonding of the channel devices onto the silicon or glass device surface. For systems with large, planar surfaces, directly bonding to the substrate may be appropriate, provided that the devices and sensors can be arranged to account for the irregular dimensions of hand-diced microfluidic chips and for the large tolerances of the placement process. However, microelectronic fabrication is capable of fabricating complex, high density, nanometer and micrometer scale devices on silicon substrates. Interfacing these devices with microfluidic systems typically requires far more sophistication and complication because of the different length scales involved. Typical silicon die area can be on the order of cm2 while microfluidic devices generally are on the order of mm2, so the area mismatch and physical requirements to be able to make electrical connections to the silicon chip can create difficult problems for integration and packaging.6

There have been several solutions built on introducing intermediate encapsulation and polyimide layers to allow localized exposure of the chip's surface to liquid flow. A common resolution involves aligning and attaching a sheet of polyimide layer (commonly a printed circuit board (PCB)) with a cut out opening on top of a silicon chip so that the electronic devices of interest are exposed through the opening.7–10 An additional interconnect stack with encapsulation material is generally needed between the polyimide layer and silicon chip to allow for electrical connections and for a liquid barrier for the unexposed portion of the chip. The top surface of the polyimide layer acts as a large, planar surface onto which a microfluidic channel is bonded. A variant of this solution involves milling orifices into a PCB to create a pocket for the silicon chip to sit in.6 Encapsulation material is needed to secure the chip flush with the PCB surface and to create a planarized, patterned mask on the chip's surface and surround area for selective liquid exposure. This solution is advantageous in that it can allow for wire bonding connections and is not limited to via interconnects. However, for all these solutions, the additional layers between the PDMS and device interfaces can add to the vertical dimension of the channel at the chip, therefore creating dead volumes and modifying the intended flow behavior at critical locations.6–10 Rather than developing a complex interface technique, an alternative solution is to design the microfluidic subsystem to allow for eventual bonding.

Depending on the experimental setups, it may be necessary to align equipment fixtures, wireless actuators, device features, cable connectors, or probe pads close to microfluidic channel features. For example, ongoing work by Hu et al. to develop an intracellular, radio frequency (RF) sensor platform requires fluid flow across the wireless electronic sensors.11 To maximize the electrical signal output, the probe pads should be located within a few hundred micrometers of the wireless sensor and must remain uncovered by the microfluidic device. In general, the exact alignment spacing depends on the packaging allowances and maximum permissible electrical signal attenuation, and systems with millimeters to hundreds of micrometer spacings have been demonstrated.11–14 Refining alignment techniques is an important step toward developing large scale integration (LSI) and microfluidic hybrid systems.15 However, the general bulkiness of these components and the required unobstructed access to them make system integration difficult and the development of such systems a one-off process. When the spacing requirement is on the order of the millimeter scale, the outline of the microfluidic device can be roughly cut or molded so that the channel runs closer to one edge of the device or so that window openings are created by removing PDMS material.11,12 However, when the component must be within a couple hundred micrometer of the channel edge, a more robust method of fabrication is needed for consistent results.11,14 This is exacerbated by the tendency of cured PDMS to incompletely cut through and tear during the dicing process.

This work presents a method of designing and fabricating a polydimethylsiloxane (PDMS) microfluidic device footprint to enable a simple and robust method of creating microfluidic integrated lab on chips. The fabricated microfluidic chip has critical channel features located between 50 μm and 250 μm from the periphery. As an illustration of the method described, we demonstrate a system that integrates the microfluidic device with a small, several millimeter by millimeter integrated circuits chip and enables liquid flow on the chip's surface while maintaining access to probe pads on the chip's surface.

2 Fabrication of reduced footprint microfluidic devices

Fig. 1 shows the process flow of fabricating a PDMS microfluidic device to have a reduced and specific shaped footprint with a window cutout. The microfluidic channel fabricated in this study is a simple straight channel terminated with inlet and outlet ports for fluid flow. The channel is flanked by parallel and unconnected spacing markers to aid in alignment. As shown in Fig. 2, the PDMS device is designed to open up a window in the footprint to allow for electronic chip integration. This microfluidic chip can be bonded onto the surface of an integrated circuits chip with the channel located directly on device features while leaving the majority of the chip uncovered and allowing electrical probing or wired connections to be made on the chip's surface. Additionally, the integrated microfluidic system can be arranged so that any bulky and nonplanar components, such as connectors, wireless actuators, and optical components, can be located at a minimum of 50 μm away from the channel feature. The intention of this study and design is to minimize the channel-edge distance, the distance between the edge of the microfluidic channel feature (the edge closest to the window) and the outer PDMS edge, while maintaining ideal operation of the microfluidic device.
image file: c7ay02177a-f1.tif
Fig. 1 Schematic of the fabrication process flow. (a) Prepare a SPR 220-7.0 master mold. Apply surfactant solution on the edge of the wafer to aid the film release. (b) Spin coat uncured PDMS mixture on the master mold. Bake to cure. (c) Using a flexible Parafilm M backing, release the patterned film from the mold and transfer to a blank silicon or glass substrate. (d) Prepare a laser cut acrylic mold by coating the top surfaces with surfactant solution. (e) Pour uncured PDMS mixture in the acrylic mold well and cover with a surfactant-coated glass slide. Bake to cure. (f) In a water bath, release the molded bulk PDMS. The molded piece will have an attached thin layer of excess PDMS, which can be roughly trimmed. (g) Transfer and oxygen plasma bond the bulk PDMS to the PDMS film. (h) Transfer the bonded device to an acrylic substrate. (i) Laser ablate along the perimeter of the microfluidic device to remove the thin film of excess PDMS. (j) Clean the device and repeat with additional layers or bond to the final desired substrate.

image file: c7ay02177a-f2.tif
Fig. 2 Final microfluidic device to be integrated into a system. The cut out window allows the straight microfluidic channel to be placed in close proximity to a sensor subsystem. The channel features are located on the back side of the PDMS device, and minor laser ablation damage is visible on the front side of the device. The channel-edge distance is defined relative to the back surface (see Fig. 1j), on which channel features are located. It is defined as the distance between the channel edge (the edge closest to the window) and the outer PDMS edge.

2.1 Step 1: master mold fabrication

Photoresist features on silicon wafers are used as master molds for the fabrication of the microfluidic channel. First, clean wafers are coated with hexamethyldisilazane (HMDS) to promote photoresist adhesion. The wafers are spin coated with 15 μm of Shipley SPR 220-7 and exposed with a channel pattern using a Karl Suss MA6 Aligner. The patterns are developed as appropriate. While this process uses a standard photoresist and wafer master mold method, using a laser cutting to engrave channel features on PDMS films can further simplify the process and eliminate the use of expensive cleanroom equipment and processes.

2.2 Step 2: bulk mold fabrication

The bulk of the PDMS device is cast molded with an acrylic mold. The desired device footprint is laser cut into a quarter inch thick piece of acrylic. The thickness of this acrylic determines the approximate thickness of the final PDMS device, and a quarter inch height gives sufficient structural stability to port connections. The patterned acrylic is then acrylic cemented to a blank sheet to produce a mold well. The patterned acrylic is oriented so that the laser kerf profile, the profile of the track of material removed during the cutting process, allows the exposed well opening to be larger than the bonded opening. This facilitates the later extraction of the cast molded PDMS bulk piece. The mold is then annealed to release stress in the cemented joints and to prevent further outgassing during subsequent PDMS curing.

2.3 Step 3: release layer preparation and application

This process utilizes a non-toxic surfactant solution to aid the release of the PDMS film from the master mold. To prepare the surfactant solution, a 5% Dawn soap in DI water is sonicated for 15 minutes. Both the surfactant and concentration affect the PDMS surface morphology and can be further optimized for a specific application.16 To assist the film release, surfactant solution is applied along the edge of the wafer, and the master mold is dried in air.

As with the master mold, the surfactant solution is applied within the wells and on the top surface of the mold. For a strong bond between the bulk PDMS and subsequent PDMS layers, the bonding surface of both layers should be smooth and flat. A glass slide will be placed on the mold opening to ensure that the PDMS bulk piece is planar and does not have a meniscus. Surfactant is applied on one surface of a clean glass slide to aid in the separation of the PDMS from the slide surface. Both the mold and glass slide are dried in air.

2.4 Step 4: PDMS film and bulk fabrication

A Sylgard 184 silicone elastomer mixture is prepared with a base[thin space (1/6-em)]:[thin space (1/6-em)]curing agent ratio of 10[thin space (1/6-em)]:[thin space (1/6-em)]1. After carefully pouring a dollop of the PDMS mixture in the center of the wafer, any bubbles that appear are popped with a nitrogen gun. The wafer is spin coated at 600 rpm for 30 seconds and the coating cured at 70 °C for 20 minutes to produce a film of approximately 200 μm in thickness. Thinner films can be easily made by increasing the rotation speed, but transferring and handling the films becomes more difficult due to a tendency for the film to stick and tear. Alternative release layers include silane and poly vinyl alcohol, which may aid the handling of thinner films.17 To aid film release and prevent the PDMS from folding over, a Parafilm M backing is applied to the exposed PDMS surface. The wafer stack is placed in a DI water bath and the PDMS is carefully demolded by pulling and peeling at the Parafilm layer. Using a razor blade, the film is roughly diced to separate the desired channel features. The film pieces are applied to glass microscope coverglasses, feature side down.

Silicone elastomer mixture is poured in the acrylic mold, making sure to overfill. After pulling vacuum on the mixture to remove any air bubbles, an air gun can be used to gently pop any residual bubbles drawn to the surface. Finally, without introducing any additional air pockets, the surfactant-coated glass slide is applied to the well opening to close off the well. A weight is placed on top of the glass slide and the mixture is cured at 70 °C for 75 minutes. The molded PDMS is then released in a DI water bath. Each cast molded bulk piece may have an excess attached thin layer of PDMS. The excess layer of PDMS is about 20 μm in thickness.

2.5 Step 5: bonding and laser ablating to the final footprint

After cleaning and preparing the PDMS surfaces for bonding by sonicating the film and bulk in an isopropyl alcohol bath, the bonding surfaces are activated with 30 seconds of air plasma. To ensure a permanent bond between the bulk and film surfaces, the two pieces were aligned and bonded using a Finetech Lambda flip chip bonder within a few minutes of the surface activation.18 This placement process step determines the critical channel-edge distance. The machine can pick up the glass-backed PDMS film using the head component and place it with sub-micron precision onto the bulk PDMS substrate. The bonded pieces are baked at 80 °C for 30 minutes to strengthen the permanent plasma bond.

The bonded pieces are peeled from the silicon die and placed on an acrylic substrate. The permanent film–bulk PDMS bond should be stronger than the film–silicon bond. The edges of the bulk PDMS are roughly laser ablated with an Epilog Fusion M2 Laser. The laser settings used for the devices in this work are: two passes of CO2 laser rasterization with 5.5 Watts power and 7% speed. This low power laser ablation was optimized to ablate the thin PDMS film extending past the bulk periphery while only roughening the top surface of the bulk component. The acrylic substrate reduces the sidewall roughness generated by the laser process.19 Therefore, the device maintains the original desired footprint with only minor damage to the non-critical top surface. The final device is removed from the acrylic substrate and cleaned in a sonicated isopropyl alcohol bath. The microfluidic device is now ready for coring of the ports and final bonding to an experimental substrate. Fig. 2 shows a completed microfluidic device fabricated with the above mentioned process.

3 Results and discussion

3.1 Channel characterization

The process detailed above can be used to locate critical channel features near the periphery of the microfluidic device, and this can be extended to remove any extraneous PDMS material to refine and reduce the microfluidic chip footprint. Both modifications can aid system integration and packaging efforts. However, the devices created through this process must also maintain functionality, primarily requiring a secure PDMS-substrate bond that allows the microfluidic channel to maintain a sufficient pressure difference for fluid flow. The thin channel wall between the edge of the window cutout and the channel feature may deform and even break prematurely during fluid flow. In fact, Quake valves operate with the deflection of the channel wall membranes, but for this device, such action may be unfavorable due to potential deviations from fully developed laminar flow characteristics at critical sensor locations.20 Additionally, the use of a laser cutter along the channel wall may modify the material behavior, such as the bonding capacity of the PDMS at that place.

The channels are characterized after plasma bonding to glass microscope slides, which have an average surface roughness of Ra = 0.07 μm. The microfluidic channels are first characterized by examining the relationship between channel pressure and flow rate for channels of various widths. Multiple devices of each channel width are fabricated with various channel-edge distances. A 0.5% red fluorescent latex beads in an aqueous solution is injected into the channel by a pressure source. The bead paths are recorded by an Andor Zyla camera connected to a Nikon Eclipse Ti microscope, and the images are analyzed in imageJ to track particle paths.21 The pressure drop across the channel is measured using a Labsmith uPS1800. A schematic of the testing setup is shown in Fig. 3. The same flow rate and pressure experiments were conducted on control devices. The control devices are fabricated with a typical casting process, i.e. pour a thick layer of PDMS mixture on the wafer mold and separate individual devices with a razor blade after a bake cure. These control devices do not have a cut out window, i.e. the device channels are well separated from the device perimeters, and thus can be made with a typical casting process.22 Microfluidic devices with channel widths of 60 μm, 80 μm, 100 μm, and 120 μm were tested. The devices have an average channel height of 13.56 μm and length of 11 mm, as measured by a Keyence 3D Laser Scanning Microscope.


image file: c7ay02177a-f3.tif
Fig. 3 Schematic of the testing setup. A fluorescent latex beads solution is injected into the device under test with a Harvard Syringe Pump. A tared Labsmith pressure sensor measures the pressure drop across the microfluidic channel while the fluorescent beads in the channel are tracked to determine channel flow velocity.

3.2 Fluid flow

The fluid flow in the channels can be describe as Poiseuille flow. For a fully developed laminar flow, where the width of a channel is much larger than the height, the fluid flow can be approximated as a two-dimensional laminar flow, and the spatial distribution of the flow velocity in the channel is
image file: c7ay02177a-t1.tif
where u is the flow velocity, Δp is the pressure drop along the channel, μ is the viscosity of the fluid, L is the length of the channel, h is the height of the channel, and y is the vertical distance between image file: c7ay02177a-t2.tif and image file: c7ay02177a-t3.tif.23 This equation is used to determine the theoretical maximum velocity of the fluorescent particles and hence the flow velocity in the channel. The flow velocities of the channels with different channel widths and different channel-edge distances are compared with theoretical velocities to determine whether the fabrication process introduces any non-idealities.

For each channel width, multiple devices are tested with various channel-edge distances. The reference device has no cut out window and has a channel-edge distance of at least 5 mm. It was fabricated with a traditional, monolithic, cast-molded process and hand-diced into individual chips. The other devices fabricated have approximate channel-edge distances of 250 μm, 150 μm, and 50 μm. Fabrication variations result in channel-edge distances that fluctuate along the edge of each device, and the edge distances notated in this work are the smallest distances between the channel edge and the location where the PDMS terminates (see Fig. 1j). Devices of various channel-edge distances were fabricated to examine critical channel-edge dimensions that may cause non-idealities in flow behavior.

As shown in Fig. 4, the relationship between the maximum flow velocity and the pressure difference closely follows the theoretical relationship for all permutations of channel-edge distances and channel widths. The maximum flow velocity varies linearly with the pressure difference, and the deviation of the results from the theoretical results are likely due to variations in the channel lengths and heights introduced in the fabrication process and the approximation of the flow as a two-dimensional laminar flow. Additionally, there exists the entry effect, which can cause deviation of the fluid flow behavior from the assumed fully developed laminar flow. The pressures measured correspond to volumetric flow rates up to 6 μL min−1. While it is possible that higher applied pressures can induce some deformation of the channel and hence introduce fluid–structure interactions, under those conditions, the reference channel behavior may also no longer follow the ideal relationship of a fully developed laminar flow. The experimental results demonstrate that a device fabricated with the above mentioned process is an appropriate alternative to a conventional microfluidic device, especially when size and footprint may limit the practicability of system integration. In future studies, it may be interesting to determine the pressures at which non-idealities appear, as a function of the channel-edge distance.


image file: c7ay02177a-f4.tif
Fig. 4 Maximum flow velocity versus pressure difference for various channel sizes. Each channel size is tested with several channel-edge distances, with a non-windowed device acting as a reference device. The theoretical velocity is calculated using the Poiseuille velocity profile.

Depending on the application and required pressures, an appropriate microfluidic device can be fabricated to have channel features located near the periphery. The performance and functionality may be limited by fabrication yield and alignment precision rather than feasibility of making a device with a maximum channel-edge, area, or footprint geometry. We will discuss yield in a later section.

3.3 Critical pressure

Here, we examine the failure modes of the fabricated microfluidic devices and the associated critical pressures at which the PDMS device delaminates or leaks. A small channel-edge distance means there is a small amount of PDMS material available to sustain the channel pressure. This potential failure point limits the functionality and maximum flow rate allowed within the channel. Using the same experimental setup described in Fig. 3, the volumetric flow rate is gradually increased until there is a change in the measured pressure difference that indicates a leak. When the leak is due to the delamination at the thin membrane, there is an immediate and acute loss of pressure, coupled with turbulent movement of the fluorescent beads. Leakage at the inlets is another failure mode and is characterized by a sustained measured pressure, even with increased volumetric flow. In this work, we consider both modes of failure. It is worth pointing out that both failure modes are indicative of the bonding strength between the PDMS membrane and the glass substrate.

Fig. 5 shows the correlation of the critical pressure, the pressure difference at which local leakage and/or delamination initiates, with the channel-edge distance. The critical pressure is independent of the channel width, and the devices with smaller channel-edge distances ruptured at pressures comparable to those with larger distances. In fact, the range of critical pressures (pressure differences) shown in Fig. 5 are in accord with those measured by Eddings et al. in the determination of the optimal PDMS–PDMS bonding technique.24 The variation observed in the maximum sustainable pressure difference can be attributed to variations of the bonding strength associated with intrinsic variations of the plasma bonding process. Hence, the developed fabrication process does not accelerate the delamination or leakage failure of microfluidic devices, for the channel-edge distances reported in this work.


image file: c7ay02177a-f5.tif
Fig. 5 Critical pressure as a function of the channel-edge distance for all channel widths.

3.4 Process reliability

Locating a critical channel feature near the device's periphery can weaken PDMS-substrate bonds and cause premature device failure due to particulates and oil transferred along the edges during sample handling. The contamination of the PDMS edge causes a lack of a permanent plasma bond of visually passable devices and has been the largest contributor to premature device failures. Insufficient cleaning after laser cutting along the device edge introduces residue to the critical bonding edge. Other modes of failure include ripping off chunks of the PDMS edge during a cleaning step using scotch tape. This damage during cleaning only significantly impacts device yield for less than 100 μm edge distances.

The edge distances notated in this work are the smallest distances between the channel edge and the location that the PDMS terminates. For failures due to local delamination, this is often the site of failure. However, for each device, the edge distance along the channel may vary. This is due to variations in the fabrication process for the bulk PDMS molds. The laser cutting process of acrylic may introduce uneven cutting lines and some variations due to the melting of the acrylic. The combination of material, material thickness, and laser cutter will also determine the angle of inclination of the laser kerf, which directly determines the sidewall angles of the PDMS bulk pieces. In addition, the acrylic cementing process to form the bulk PDMS wells may leave the well edges rounded, giving a more sloped PDMS sidewall, which acts as a guide for the final laser ablation and determines the edge distance.

Although the developed process increases the process time required to make a device, this method is robust and only requires technical finesse when extracting the thin films of PDMS, a process step common in multilayer and three dimensional microfluidic systems. Use of the Finetech Flip Chip Bonder as an alignment instrument eliminates the need for imprecise hand alignment on a lubricating methanol layer. In this study, 32 devices were fabricated with a yield of 75%, with most of the yield loss occurring for <100 μm edge distance devices.

3.5 Integrated microfluidic-electronic system

Here, we demonstrate how a microfluidic device fabricated with the developed process may be integrated with an electronic sensing subsystem. In this work, a microfluidic channel is bonded to the surface of a radio frequency (RF) electronic sensor. This is a significant step towards the development of the wireless, intracellular detection system as described by Hu et al.11 To create a flat substrate for the microfluidic channel to be bonded onto, an integrated circuits chip is embedded and planarized in a PDMS film. This ensures a continuous surface extending between the chip's device surface and the supporting PDMS, and this can be done with no leakage of PDMS onto the contact pads of the chip. Using the Finetech flip chip bonder, the microfluidic channel and the integrated circuits chip are precisely aligned and bonded. The final system is shown in Fig. 6. The PDMS device is aligned to arrange the microfluidic channel on sensor devices while exposing contact pads. This system successfully maintains a seal during water flow along the surface of the chip, allowing for electrical measurements during sample transport and fluid flow. Current work is focused on integrating biological samples with intracellular electronic devices that will detected while flowing through the microchannel across the sensor integrated circuit. This novel fabrication process opens up the possibility of a variety of microfluidic device outlines and enables more diverse lab on chip systems, such as the intracellular RF sensor system described above.
image file: c7ay02177a-f6.tif
Fig. 6 Schematic and close up view of chip and microfluidic integrated system. The integrated circuits chip is fabricated with the TSMC (Taiwan Semiconductor Manufacturing Company, Ltd) 40 nm CMOS general process and is semi-exposed, allowing for fluid flow on covered, sensor portions of the chip and probing on the exposed portions of the chip25 The exposed portion of the chip consists of shielding patterns that form a checkerboard design and probe pads (not shown).

4 Conclusions

Lab on a chip systems with microfluidic integration are incredibly powerful and versatile, and they have already been proven to be appropriate for a variety of biological and chemical applications. With the results presented above, we enable further development into systems of high density and functionality in small packages. We have demonstrated that it is possible to fabricate a microfluidic device with critical channel features near the periphery of the device without changing the flow characteristics in the channel. This can be extended to minimize the microfluidic chip footprint, which is advantageous for packaging purposes. In addition, the thin channel walls produced through the previously detailed process do not prematurely break under high pressure, ensuring a robust system operation. Packaging and system integration are important parts of designing lab on chips, and this process only expands the applications and relevance of microfluidic systems.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-114747. The authors also acknowledge the support of Stanford Graduate Fellowships, the Bio-X IIP Award, and the Air Force Office of Scientific Research (MURI: Atomically Thin Systems That Unfold, Interact and Communicate at the Cellular Scale). Work was performed in part in the nano@Stanford labs, which are supported by the National Science Foundation as part of the National Nanotechnology Coordinated Infrastructure under award ECCS-1542152. We thank the Stanford Genome Technology Center for providing the microscope and imaging measurement equipment.

References

  1. M. Astolfi, B. Peant, M. A. Lateef, N. Rousset, J. Kendall-Dupont, E. Carmona, F. Monet, F. Saad, D. Provencher, A.-M. Mes-Masson and T. Gervais, Lab Chip, 2016, 16, 312–325 RSC.
  2. G. Luka, A. Ahmadi, H. Najjaran, E. Alocilja, M. DeRosa, K. Wolthers, A. Malki, H. Aziz, A. Althani and M. Hoorfar, Sensors, 2015, 15, 30011–30031 CrossRef CAS PubMed.
  3. W.-L. Chou, P.-Y. Lee, C.-L. Yang, W.-Y. Huang and Y.-S. Lin, Micromachines, 2015, 6, 1249–1271 CrossRef.
  4. P. Abgrall and A. Gue, J. Micromech. Microeng., 2007, 17, R15 CrossRef.
  5. S. Haeberle and R. Zengerle, Lab Chip, 2007, 7, 1094–1110 RSC.
  6. I. Burdallo, C. Jimenez-Jorquera, C. Fernández-Sánchez and A. Baldi, J. Micromech. Microeng., 2012, 22, 105022 CrossRef.
  7. A. Wu, L. Wang, E. Jensen, R. Mathies and B. Boser, Lab Chip, 2010, 10, 519–521 RSC.
  8. R. Wimberger-Friedl, T. Nellissen, W. Weekamp, J. van Delft, W. Ansems, M. Prins, M. Megens, W. Dittmer, C. de Witz and B. van Iersel, J. Micromech. Microeng., 2009, 19, 015015 CrossRef.
  9. D. Welch and J. B. Christen, J. Micromech. Microeng., 2013, 23, 035009 CrossRef.
  10. M. Muluneh and D. Issadore, Lab Chip, 2014, 14, 4552–4558 RSC.
  11. X. Hu, K. Aggarwal, M. X. Yang, K. B. Parizi, X. Xu, D. Akin, A. S. Y. Poon and H.-S. P. Wong, Phys. Rev. Appl., 2017, 8, 014031 CrossRef.
  12. G. Duan, X. Zhao, H. R. Seren, C. Chen, A. Li and X. Zhang, J. Micromech. Microeng., 2016, 26, 124002 CrossRef.
  13. J. Zhao, A. Hashmi, J. Xu and W. Xue, Appl. Phys. Lett., 2012, 100, 243109 CrossRef.
  14. J. I. Padovani, S. S. Jeffrey and R. T. Howe, Technology, 2016, 04, 110–119 CrossRef PubMed.
  15. R. Sivakumarasamy, K. Nishiguchi, A. Fujiwara, D. Vuillaume and N. Clement, Anal. Methods, 2014, 6, 97–101 RSC.
  16. A. Fatona, Y. Chen, M. Reid, M. A. Brook and J. M. Moran-Mirabal, Lab Chip, 2015, 15, 4322–4330 RSC.
  17. J. M. Karlsson, T. Haraldsson, C. F. Carlborg, J. Hansson, A. Russom and W. van der Wijngaart, J. Micromech. Microeng., 2012, 22, 085009 CrossRef.
  18. S. Bhattacharya, A. Datta, J. M. Berg and S. Gangopadhyay, J. Microelectromech. Syst., 2005, 14, 590–597 CrossRef CAS.
  19. H.-B. Liu and H.-Q. Gong, J. Micromech. Microeng., 2009, 19, 037002 CrossRef.
  20. M. A. Unger, H.-P. Chou, T. Thorsen, A. Scherer and S. R. Quake, Science, 2000, 288(5463), 113–116 CrossRef CAS.
  21. J. Schindelin, C. T. Rueden, M. C. Hiner and K. W. Eliceiri, Mol. Reprod. Dev., 2015, 82, 518–529 CrossRef CAS PubMed.
  22. J. Friend and L. Yeo, Biomicrofluidics, 2010, 4, 026502 Search PubMed.
  23. H. A. Stone, in Introduction to Fluid Dynamics for Microfluidic Flows, ed. H. Lee, R. M. Westervelt and D. Ham, Springer, US, Boston, MA, 2007, pp. 5–30 Search PubMed.
  24. M. A. Eddings, M. A. Johnson and B. K. Gale, J. Micromech. Microeng., 2008, 18, 067001 CrossRef.
  25. TSMC, http://tsmc.com/english/dedicatedFoundry/index.htm.

This journal is © The Royal Society of Chemistry 2017