High-precision modular microfluidics by micromilling of interlocking injection-molded blocks

Crystal E. Owens and A. John Hart*
Department of Mechanical Engineering and Laboratory for Manufacturing and Productivity, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA. E-mail: ajhart@mit.edu

Received 4th September 2017 , Accepted 12th December 2017

First published on 12th December 2017


Wider use and adaptation of microfluidics is hindered by the infrastructure, knowledge, and time required to build prototype systems, especially when multiple fluid operations and measurements are required. As a result, 3D printing of microfluidics is attracting interest, yet cannot readily achieve the feature size, smoothness, and optical transparency needed for many standard microfluidic systems. Herein we present a new approach to the design and construction of high-precision modular microfluidics, using standard injection-molded blocks that are modified using micromilling and assembled via elastically averaged contacts. Desktop micromilling achieves channel dimensions as small as 50 μm depth and 150 μm width and adhesive films seal channels to allow internal fluid pressure of >400 kPa. Elastically averaged connections between bricks result in a mechanical locating repeatability of ∼1 μm, enabling fluid to pass between bricks via an O-ring seal with >99.9% reliability. We demonstrated and tested block-based systems for generating droplets at rates above 9000 min−1 and COV <3%, and integrated optical sensors. We also show how blocks can be used to build easily reconfigurable interfaces with glass microfluidic devices and imaging hardware. Microfluidic bricks fabricated by FDM and SLA 3D printing cannot achieve the dimensional quality of molded bricks, yet 3D printing allows customized bricks to be integrated with standard LEGOs. Our approach enables a wide variety of modular microfluidic units to be built using a widely available, cost-effective platform, encouraging use in both research and education.


I. Introduction

Microfluidics leverages the simplified physics and enhanced control of fluid flow at small sizes and low Reynolds number. As a result, microfluidic systems can be configured to perform a wide variety of functions including fundamental investigations of transport, measurements of chemical reaction rates, water quality testing, low-cost medical diagnostics, droplet production for emulsions and synthesis, and more.1–3

However, unlike circuit board electronics, which largely use standard components and leverage digitally-driven design and fabrication tools, microfluidic systems lack a universally established construction paradigm. The process of designing and fabricating a microfluidic system is compounded further by requirements for high precision manufacturing and by practical issues including materials compatibility and sealing. As a result, wafer-based microfabrication such as chemical etching of silicon or glass substrates remains the dominant method of fabricating precision microfluidic systems. Cleanroom infrastructure and operation is expensive, and incurs high material and labor costs and long lead times. Indeed, the commercial viability of many lab-on-a-chip and diagnostic tools has been limited by the high capital cost of manufacturing processes needed to meet the dimensional tolerances required for reliable function.4,5

The adaptation of soft lithography to microfluidics in 1998 catalyzed wider adoption of microfluidics in research because users could design, prototype, and replicate new systems using master templates and polydimethylsiloxane (PDMS). This reduced production time from ∼days to ∼hours once a master mold was formed, and provided new material properties via the soft and flexible elastomers.6–8 However, drawbacks to use of PDMS include its permeability to many liquids, adsorption of small hydrophobic species, the complexity of interconnecting and interfacing with fluid lines and sensors, and once again the necessity of expert technique.

In most cases, microfluidic systems are designed to integrate all necessary functions on a single chip. An alternative approach is to divide the system into modular units that can be assembled to create the final system. Modular systems are especially desirable if they can be made by common machining and laboratory tools, so that any researcher could create or modify unit “building blocks”. In addition, modularity makes it easier for components developed and validated within a research team to be made available to others via digital data (e.g., CAD models), for piecewise integration into new systems, lowering the learning curve and accelerating the collective discovery process.

More recently, researchers have sought to adapt consumer-grade printing technologies to achieve lower-cost and more accessible prototyping of microfluidics. These include inkjet and laser toner printing, wax printing, and polymer 3D printing methods including fused filament fabrication, stereolithography, and jetting.9,10 Many researchers have created and shared two- and three-dimensional (3D) printed designs to produce both monolithic and modular microfluidic systems.3,11–22 These devices were used, for instance, to assemble DNA,23 create emulsions,3 enable fluid flow control using a network,11 and integrate optical and other sensors directly into printed units.24 Engineers continue to develop the underlying technologies, and recently 3D-printed microfluidic devices with planar channels having diameters as small as 13 μm.25 Of particular relevance are two recently developed 3D-printed modular microfluidic systems, one of which uses O-rings and magnets to provide reversible sealing between units,56 and a second that interconnects in a LEGO-like fashion using PDMS blocks.20

However, the use of 2D and 3D printing methods for high-performance microfluidic fabrication is restricted by several limitations, including material compatibility, resolution of the minimum feature size and surface roughness, and long-term dimensional stability, particularly when in contact with a fluid. Modularity places additional stringent requirements on part accuracy, repeatability, and interchangeability – essential criteria to enable rapid construction of systems from a component library – and for maintenance of tight sealing performance between modules, particularly if the system is also reconfigurable.

Contrasting 3D printing, injection molding can produce parts from a wide variety of polymers, with excellent dimensional control (<50 μm tolerances), very smooth surface finish (average roughness <1 μm), and a production speed of ∼seconds per part. However, high tooling cost makes injection molding prohibitive except for in production of a large number of identical units.

Herein we present a new approach to the design and construction of high-precision modular microfluidics, using standard injection-molded blocks that are modified using micromilling and assembled via elastically averaged contacts. We demonstrate this principle using store-bought LEGO bricks. This hybrid approach leverages the materials and dimensional quality of standard injection molded units. Custom modifications using desktop laboratory tools enables deployment of a library of bricks with both fluidic and active (sensing, sorting/filtering) functions. We investigate the dimensional tolerances and mounting repeatability of LEGO bricks, develop machining processes to fabricate the microfluidic modules, establish designs for reversible sealing between blocks, and build demonstration devices including a modular droplet generation system. Finally, we compare geometry and material properties of micromachined LEGO bricks with those of 3D-printed components, and qualify each method for producing modular systems.

II. Experimental methods and analysis

a. Micromilling

Channels were fabricated by micromilling (Roland SRM-20) the sides of as-received LEGO bricks, with dimensions ranging from 150 to 500 μm in width and from 20 to 1000 μm in depth (Performance Micro Tool uncoated carbide endmills, 2 flutes, diameter 0.0040–0.0200′′). Initial milling parameters were guided using software (GWizard, CNCCookbook), and then refined by experiments to reduce surface roughness and tool breakage. Typical cutting parameters for a 0.0190′′ diameter square endmill were 9500 revolutions per minute, feed of 270 mm min−1 (ABS) or 190 mm min−1 (PC), and depth of cut of 0.120 mm. For 0.0070′′ diameter square endmill: 9500 revolutions per minute, feed of 50 mm min−1 (ABS) or 20 mm min−1 (PC) and depth of cut of 0.03 mm. Climb milling was found to give the best surface quality. A complete brick took 1–40 minutes to mill and rotate to each face, depending on total path length and channel dimensions.

Bricks were fixtured in the mill using a base made of other LEGO bricks. This allowed the bricks to be securely mounted to other bricks, and repeatably re-positioned. An initial zero point was located on this fixture and then bricks were freely attached and rotated on the fixture for milling.

b. Brick size and gap measurements

The size distribution of LEGO bricks was measured with a digital micrometer (Mitutoyo IP65, resolution 0.001 mm). These values were used to determine the size distribution of the gaps that exist between bricks on a baseplate, by comparing the brick dimension to the average distance between brick posts. This was calculated from measurements of >100 bricks, and always was larger than the brick widths by 100−300 μm. While the posts are not a perfect datum reference, elastic averaging reduces the effect of variation in post position relative to the brick edges.

These values were confirmed by microscope measurements of bricks placed side-by-side on a baseplate, and by the success of later experiments that relied on these values.

c. Repeatability measurements

Repeatability of mounting LEGO bricks on a LEGO baseplate was measured by placing randomly selected pairs of same-size bricks onto a baseplate, and measuring brick position using an optical microscope (Zeiss Z1m). One brick was removed and reattached in the same location, and images were analyzed (using AxioVision software) to determine the repeatability of position of the edge of the moved brick relative to the edge of the stationary brick, averaging the distance along a 2 mm length of a brick edge. Angular variation was measured by comparing the angles of each brick in the same place–replace experimental process. A measurement resolution of ∼0.5 μm was determined from a set of control experiments where the brick assembly was replaced and reimaged but bricks were not disassembled. The minimum resolution depended on image quality, which was affected by brick color, and is indicated by the error bar for each measurement set. All measurements were performed by the same researcher.

d. Surface characterization

The surface roughness of the outer sidewalls of bricks before and after milling was measured with a stylus profilometer (2 μm radius stylus, DektakXT) and a white light interferometer (Zygo white light interferometer NewView 5000).

Surface features also were viewed using scanning electron microscopy (Zeiss Merlin High-resolution SEM, in-lens detector).

e. Sealing and interconnection

To seal the brick faces, the milled channels were covered with a thin film of polyethylene with acrylic adhesive (110 μm thickness, ThermalSeal) underlaid with cyanoacrylate adhesive (Krazy Glue KG925) around the edges to anchor the film. The film was cut with the corner of a standard razor blade at fluid inlet points to make apertures. Milled grooves held O-rings (size 001-1/2, 1/8′′ outer diameter, EPDM rubber, McMaster-Carr). The O-ring groove was milled with a 0.0190′′ diameter endmill, rather than a 1/8′′ or 1/16′′ endmill, to reduce the mean length of machining marks and lower the minimum compression required for sealing.55

The inlet/outlet brick was created by milling 1/8′′ through-holes in line with brick posts, and press-fitting 1/8′′ outer diameter tubing into the hole (Tygon, McMaster-Carr #5103K42). Cyanoacrylate adhesive (Krazy Glue KG925) optionally formed a permanent contact between the tubing and hole. An O-ring seat and small through-hole were milled into the body of the brick to enable sealing with fluidic bricks. All connections between bricks occurred through these holes cut on the smooth sidewalls of the bricks.

f. Contact angle

The contact angle of water on brick surfaces was measured by dispensing a droplet of water onto the side surface and using a contact angle goniometer (Rame-Hart, Model 590).

g. Flow characterization

Pressure was measured directly upstream of the LEGO fluid system using a liquid pressure dial gauge (McMaster-Carr #4026K3), when water was pumped through the system at controlled flow rates using a syringe pump (NE-300, Pump Systems Inc.). The maximum holding pressure was measured by inletting a constant flow of water, plugging the outlet, and recording the pressure at which the system first began to leak.

Flow was imaged with an optical microscope (Zeiss Z1m) or with a smartphone (Samsung Galaxy S6) and macrolens (Apexel 20×) held by a LEGO assembly.

Fluids used include colored water (AmeriColor), ambient air, and silicone oil (Sigma-Aldrich, 10 cSt) with surfactant to stabilize emulsions (0.5 vol% Tween20, Fisher Scientific).

h. Sensors

A light sensor brick was constructed by attaching an infrared light sensor to an L-shaped LEGO brick (940 nm, LiteOn LTR-301), in a sensing configuration similar to ref. 3. A microcontroller (Arduino Uno) powered the circuit and collected data.

i. 3D printing

3D-printed bricks were made from a computer aided design model of a LEGO brick (SolidWorks) using stereolithography (SLA; Clear resin, Form2; Formlabs, layer height 0.25 mm, oblique orientation) and fused deposition modeling (FDM; ABS plastic, Stratasys Mojo; layer height 0.170 mm, vertical orientation with posts facing down because that was determined to give the best features).

III. Results

a. Concept

Modular brick-based microfluidic systems (Fig. 1) were created by modifying as-purchased LEGO bricks to include surface-machined microfluidic pathways. Systems were built by assembling these modified bricks (“fluidic bricks”) on a standard LEGO baseplate.26 By this approach, systems could be designed by machining an entire fluid pathway in a single fluidic brick, or by assembling multiple fluidic bricks with pathway sections, along with inlet/outlet bricks connected to external tubing. The exemplary system shown in Fig. 1 is a modular droplet generator, with a brick-integrated optical sensor enabling real-time monitoring of droplet production, as discussed later. Each brick was designed to perform one or more functions including mixing, droplet generation, sorting, and sensing (Fig. 1d).
image file: c7lc00951h-f1.tif
Fig. 1 A modular microfluidic system built from modified LEGO bricks. The microchannels were machined into the faces of the bricks, and sealed using a transparent adhesive-backed polyethylene film. (a) A computer rendered exploded view of the microfluidic modules, with an inset showing the sub-components of the inlet, and (b) the system snapped together, where blue color represents fluid inside bricks. (c) Photograph of a single fluidic brick component with seated O-ring used for interconnection to adjacent brick. (d) Photographs of bricks designed to input fluid, generate droplets, sense light transmission, mix or split a fluid stream, combine and mix two fluid streams, mix using a 2.5D corner geometry, and provide multiple sequential functions on an extended brick.

After milling, fluidic bricks are sealed by applying a strip of polyethylene film with adhesive backing, with apertures in the film located at the desired ends of the milled channels (Fig. 1c). The apertures are sized such that capillary pressure retains fluid inside when a brick is detached from a system (Fig. S1). Contact-based sealing between bricks is achieved using miniature O-rings placed in micromilled grooves, with a small overlap between the polyethylene film, cut with a circular aperture, and the outer surface of the O-ring. The dimensional tolerances of LEGO bricks, and the ensuing repeatability and interchangeability of mounting, results in reliable and reversible seals between adjacent bricks, enabling rapid system assembly (Video S1).

b. Micromilling on bricks

By desktop micromilling, surface-machined channels and cavities were made in bricks, with features of 150–500 μm in width, and 50–500 μm in depth, cut into any face of the bricks (Video S2). Among many candidate methods for creating channels in bricks (Table S1) including laser ablation and hot embossing, micromilling was chosen because of its combination of accessibility, versatility, speed, and dimensional quality. In this work, a three-axis desktop micromill (Roland SRM-20, ∼$5000 as of Spring 2016) was employed along with carbide endmills (see methods). Milling parameters were calculated and refined by experiments to reduce tool breakage and surface roughness (see Table S2 for typical speeds and feeds).

In the mill, bricks were held using a LEGO fixture attached to the table of the micromill with double-sided tape; this was possible because of the low cutting forces of micromilling (<5 N). The repeatability of brick mounting made it simple to rotate and reposition blocks to mill on different faces, allowing a 2.5-dimensional structure with fluid path corners that lined up well between different faces (Fig. S2). It typically took ∼10 minutes to mill each fluidic brick, including refixturing for each face and milling of the O-ring seats; for example, large straight channels took 4 minutes, and the 2.5D corner mixing brick in Fig. 1 took 15 minutes.

A scanning electron micrograph shows the rectangular sidewalls and circular corrugations (amplitude <1 μm) on the as-milled channel surface (Fig. 2b) and a laser-interferometry scan shows the sidewall texture (Fig. S3). With this setup we achieved channel widths as small as ≈150 μm (smaller dimensions were compromised by material overheating, Fig. 2d and e), and depths ranging from 50 to 1000 μm (Fig. 2c and d). The corner radius at the base of the channel was typically 5–10 μm, and the angle of the sidewall relative to the unmachined surface was 89–90°. The channel floors had an average surface roughness of 0.90 μm (Fig. 2f, Table 1). Within channels, the mill could also produce freestanding walls as small as 20 μm width, with aspect ratio >5.


image file: c7lc00951h-f2.tif
Fig. 2 Micromilling of channels into the faces of ABS LEGO bricks. (a) Photo and schematic of typical milling setup using LEGO fixture. (b) SEM image of square channel junction used as a droplet generator, showing circular tool marks on channel floor. (c) Side view of channels milled to 10–1000 μm depth and (d) top vide of channels with 150–500 μm width. (e) A comparison of endmill size to channel size shows good agreement for endmills wider than 200 μm, with the dotted line indicating a channel width equal to the endmill diameter (at cutting parameters noted in Table S2); a mill with a higher spindle speed would be expected to achieve improved accuracy at smaller dimensions. (f) Surface roughness versus milling feed rate, measured by stylus profilometry.
Table 1 Properties and demonstrated performance of LEGO-based microfluidic bricks
Category Attribute Value
Geometry Channel dimensions 50–1500 μm (depth) ≥ 150 μm (width)
Channel shape Rounded or rectangular cross-section, any two-dimensional path on a surface
Surface roughness (average, Ra) Native brick: 0.055 μm Machined brick: 0.60–1.20 μm
 
Flow and temperature Pressure limit >400 kPa
Flow rates tested 2 μl min−1 to 5 mL min−1
 
Fabrication and assembly Time to make (average per brick) 10 minutes (milling) 2 minutes (cleaning and sealing with film) 10 seconds (assembly)
Repeatability of brick–brick mounting (remove and remount) <3 μm
Interchangeability (remove and replace with a different brick of same size) <15 μm
O-ring sealing reliability >99%
 
Materials Brick material Polycarbonate (PC) or acrylonitrile butadiene styrene (ABS); suitable for aqueous solutions and suspensions, bases, alcohols
O-ring material EPDM rubber; suitable for aqueous solutions, mild acids, organic solvents
Operating temperature −3 to 70 °C, based on material properties of all components


The ability to quickly and easily machine features in this dimensional range enables the prototyping of planar pathways. This is useful for laminar flow mixing and controlled droplet generation as well as for processes that benefit from larger dimensions and nonplanar geometries including advective mixing, bulk filtration and manipulation of inertial effects to generate secondary flows for particle separation.27 For the micromill used in this work, the runout of the spindle (∼5 μm) limited the smallest tool that could be used without breaking, and the maximum spindle rotation speed (9500 rpm) limited the smallest well-made cut; these attributes are also material dependent. However, high-performance micromilling—using industrial-grade milling machines with higher spindle speeds—is capable of creating features with 10 μm width and below, and much smaller depth, with surface roughness of 10 nm, with the tradeoff of higher machine cost and more frequent endmill breakage when smaller tools are used.28,29 Other techniques such as hot embossing and laser ablation could be used to create even smaller channels and/or complex surface textures on brick surfaces.

c. Brick size and assembly tolerances

The dimensional consistency of LEGO bricks, and the repeatability of their mounting using elastically averaged contacts, is also key to their performance as modular microfluidic units. For brick-based systems with in-plane (i.e., side-to-side) interconnection this can be quantified according to the variation of brick width (σb) and the repeatability of mounting of any single brick (Rb), which is defined here as one standard deviation of the position of a brick edge in one direction as a brick was repeatedly removed and reattached to a baseplate. Assuming the respective variances are independent and normally distributed, the interchangeability of brick mounting (I) is defined as
image file: c7lc00951h-t1.tif

To guide modular system design, we measured the gap spacing and repeatability of LEGO bricks on a baseplate (Fig. 3a). The gap between nominally identical bricks was found to be roughly proportional to the brick size, with 2-post bricks having 172 μm gap width and 8-post bricks having 252 μm gap width, or 177 μm gap width on average for all bricks. When mounted to a baseplate, the standard deviation in gap width was 7 μm for 2-post bricks, 25 μm for 8-post bricks, and 25 μm variation between all bricks (or average single-brick variation of 14.4 μm) (Fig. 3b and S4).


image file: c7lc00951h-f3.tif
Fig. 3 Gap width and mounting repeatability of LEGO bricks and design of the O-ring seat in a milled brick. (a) Gap measurements were made by microscope imaging of the separation between adjacent bricks. (b) Distribution of brick–brick gap spacing between bricks with a range of size. (c) Average mounting repeatability values for each LEGO brick size on a baseplate and with an additional top plate. (d) Components of the O-ring seat that influence sealing capabilities (e) by influencing the percentage compression of the O-ring. The probability curve lies almost entirely between the lower and upper sealing thresholds, with an O-ring seat variation of σ = 45 μm.

When assembled, a pair of bricks makes contact via regular cylindrical posts on the top of the lower brick, and a rectangular web of the upper brick. Mechanical interference between these features causes frictional and elastic forces that hold the bricks together. Bricks expand elastically by a small amount (<50 μm; Fig. S5b) when mounted due to the stress exerted by posts on the baseplate, but not enough to completely fill the brick–brick gap; in fact, the use of bricks as the milling fixture self-compensates for such deformation during milling.

The mounting repeatability of all tested bricks was <3 μm (Fig. 3c). Repeatability improved with increasing brick size (number of posts) as expected from elastic averaging theory. In addition, when the pairs of bricks were also constrained on the top using a third brick that spanned the gap (essentially doubling the number of engaged posts per brick; see Fig. 3a), repeatability dropped below 1.4 μm, or <1% of our minimum channel width (Fig. 3b). Combining the values for brick dimensions and repeatability, as-received bricks and modified bricks were measured to have an interchangeability I = 7–25 μm, depending on brick size (Fig. S4).

The process of mechanical contact explains the impressive repeatability of LEGO bricks. When multiple posts fit into multiple webs, the compliant posts on the brick surface each deform slightly, causing an elastic averaging of position that (in its ideal form) reduces the error in position; the theory of elastic averaging suggests the brick–brick repeatability scales as image file: c7lc00951h-t2.tif, where c ≥ 1 is a constant, r is the surface roughness, and n is the number of connection points.30 LEGO bricks have multiple (3–4) connection points per post (see Fig. S5d) and surface roughness of <0.1 μm. Therefore, theory estimates the repeatability to be <0.1 μm at the points of contact.

d. Reversible brick-to-brick sealing

To establish consistency and enable prototyping of compact modular systems, for much of the following work we selected 1 × 2 (PC, transparent) and 2 × 2 (ABS, white) brick sizes with the same respective colors. The dimensional variation of brick size and mounting guided our design of the O-ring sealing interface between fluidic bricks (Fig. 3d and e). The sealing performance was characterized using a screw-driven compression fixture, whereby we determined that the nominally designed interface could withstand at least 400 kPa.

To make each fluidic brick, an O-ring was placed in a milled circular groove such that its raised surface protruded into the brick–brick gap with sufficient margin to form a seal upon contact with the adjacent brick. Therefore, the reliability of sealing depends on the space between the surfaces that constrain the O-ring; the total distance between brick faces includes the native gap and milled hole, minus one layer of plastic film, along with the variation in the uncompressed O-ring thickness. Necessarily, the required compression (10% compression, or 100 μm for a 1018 μm-thick O-ring, Fig. S6 and S7) and interface dimensions achieved sealing against the roughness left by surface milling, without squeezing the O-ring out of its seat (which occurs at 50% compression) when bricks were assembled. As a result, we designed the groove depth to give nominal compression of 30% of the O-ring thickness. Based on the gap statistics, we predict that >99.9% of assemblies (±20% compression = 4.5σ) chosen at random from as-received bricks will seal until either the O-ring material fails or it is dislodged by high fluid pressure (Fig. 3e). Calculations based on empirical data from an O-ring handbook suggest the pressure limit to be 5–25 MPa (Fig. S6).31 When bricks were assembled, we never observed leakage at a pressure lower than 400 kPa, at which the sealing film, but not the O-ring, occasionally delaminated from the brick.

Interestingly, the repeatability of fluidic bricks with O-rings and sealing film was measured to be 1.6 μm and 1.0 μm (Fig. 3, without and with a top clamping brick, respectively). This was improved over the unmodified bricks of the same size, suggesting that the O-ring provided a further constraint due to the force exerted against the neighboring brick.

Using this reversible interface, assembling or reconfiguring a brick-based fluidic system is as simple and as quick as working with unmodified LEGO bricks (Video S2). Importantly, the small compression force allows gentle assembly by hand, yet the relatively small area of compression enables sealing to high pressures. Nevertheless, permanent sealing of the bricks together such as by solvent or thermal welding of a plastic cover over the surfaces could give a greater pressure capacity.

e. Flow characterization

A resistor network model (Fig. 4a) can be used to predict the pressure drop of the brick-based systems, and assess how the network design and channel geometry influence the required driving pressure. The pressure-driven flow was assumed primarily viscous and laminar based on the low Reynolds and Bond numbers. The fluid resistance is therefore proportional to R ∼ ΔPh4/QL, for a pressure drop ΔP, channel hydraulic diameter h, length L, and flow rate Q.
image file: c7lc00951h-f4.tif
Fig. 4 Flow resistance and pressure. (a) Resistor model of brick system depicting contributions to flow resistance from each component. (b) Pressure drop for fluidic bricks increases linearly with flow rate; “build” refers to consecutive brick refurbishment, which involves replacing the sealing film and O-ring and cleaning brick surfaces with isopropyl alcohol. It is compared with the pressure drop for two bricks in series, which has a doubled value, as expected. Error bars indicate standard deviation in pressure drop when bricks were disassembled and reassembled without rebuilding between tests, and are within the resolution of the pressure gauge. (c) Based on the resistor model, maximum size (in terms of number of bricks) of a serially connected system. Size is limited by measured pressure capacity of 400 kPa. The highlighted region indicates typical system size (4–8 units) determined from a survey of recent articles in Lab on a Chip (see Fig. S9). Arrows to the right of the 10 mL min−1 line indicate the accessible space for any brick on that line.

Fluidic bricks with a channel cross section of 500 μm × 500 μm had a resistance of 730 Pa s mL−1 for water, which added linearly with each brick in series (Fig. 4b). The resistance did not change (within 25 Pa, set by the pressure sensor resolution) when bricks were removed and remounted, and consistent within 135 Pa s mL−1 when bricks were refurbished. For refurbishing, the tape was removed from a brick, and the brick was cleaned thoroughly with isopropyl alcohol, dried, covered with new sealing film, given a new O-ring, and reinserted into the system.

Inlet and outlet bricks had flow resistance of <1% of the fluidic bricks, and both the corner bend and O-ring seat contributed negligibly (<5%) to the overall pressure drop; this was determined by measuring the flow resistance of a system with only secondary components, and a system with paths cut with varying numbers of 90° corners (Fig. S8).

Using the resistor network model, we estimated the total pressure drop of a brick network as a function of channel hydraulic diameter, number of bricks, and flow rate (Fig. 4c). For example, at a moderate flow rate of 0.1 mL min−1, networks of >100 bricks in series could be built, without exceeding the measured pressure limits of the reversible O-ring seals. In comparison, microfluidic systems described recently in Lab on a Chip typically have ∼4–8 elements such as inlets and outlets, testing and incubation chambers, heaters, sensors, and cameras (green shaded region in Fig. 4c and S9).32–51 Therefore, we can conclude that the brick system presented here could accommodate the complexity of a typical microfluidic system by allowing one function per module, and be scaled up to build much larger systems. While a LEGO system could be bulkier when built this way, after the first steps it can be redesigned to fit into a smaller space (for example, Fig. S14).

f. Modular brick-based systems

Optical elements can be integrated in separate bricks and attached to the common baseplate, enabling construction of modular systems with integrated sensing. To demonstrate this, we show two systems for optical imaging during fundamental microfluidic flow processes: generation of water-in-oil emulsions, and controlled mixing of liquids in laminar flow (Fig. 5).
image file: c7lc00951h-f5.tif
Fig. 5 Modular brick-based fluidics with integrated optical sensing. (a) A LEGO-mounted light source and light sensor were mounted on opposite sides of a transparent fluidic brick, shown here with the sensor present and removed (inset). The brick behind the sensor is interchanged between a laminar mixing brick and a droplet generating brick. (b) When mixing two colored fluids for a series of concentrations, the sensor measures the mixing ratio; two interchangeable sensors give the same linear slope within experimental error (<1%). A darkfield micrograph (inset) shows laminar flow at the indicated point. (c) Configuring the system as a droplet generator, the sensor signal marks the passage of droplets data that marked the passage of droplets, shown at higher resolution in (d). The data is processed (e) to measure the average size, production rate, and distribution of both for a variety of flow rates.

An exemplary sensing brick was built by mounting a photodiode sensor and an emitting fiber optic cable each on “L”-shaped LEGO bricks that could be mounted in close proximity to transparent (polycarbonate) fluidic bricks (Fig. 5), with the fluid channel in line with the light path of the sensor pair. The LEGO mounts ensure consistent alignment of the emitter and sensor bricks relative to each other and to the fluidic brick when repositioned to any point along the fluid path. The polycarbonate brick and polyethylene film together allowed 80% transmission of light in the visible range (400–700 nm), whereas transparent SLA 3D printed bricks allowed 70%.

A modular fluid mixing system was built by assembling the required fluidic blocks with the emitter-detector on a common baseplate. Combining two miscible solutions of colored water at flow rates between 0.02 and 0.10 mL min−1, we constructed a curve of light transmission (measured as the output voltage, see methods) versus concentration (Fig. 5b). Measurements were taken with two LEGO-based light sensors used in turn at the same location, each having sensitivity of 2% concentration, or 0.5 volt/100% (Fig. 5b). The two sensors provided the same linear relation within 1% and individual data points within 8% (experimental limit, 0.01 V). This technique could be used to observe the extent of a precipitation or synthesis reaction based on color or transmittance, and could be extended for more sophisticated imaging and spectroscopy.

Microfluidic droplet generators are used widely for applications including material synthesis and templating, cell encapsulation and manipulation, and reaction control.52,53 Challenges remain in passively generating vast quantities of monodisperse droplets with good size control. In particular, the ability to rapidly prototype and test droplet generators is important; the complexity of the fluid interface and dependence on channel geometry means that experimental characterization is often more accurate than analysis- or simulation-based design of droplet generation devices.52

Utilizing the modular construction, we interchanged the mixing reaction brick with a droplet generating brick, adding one inlet for silicone oil, enabling immediate inline characterization of water-in-oil droplet production. A T-junction was milled into a transparent brick to form a droplet generator block (as in Fig. 1d and 5a).54 Silicone oil was used as the continuous phase and colored water as the dispersed phase. The voltage recorded by the sensor indicated the passage of droplets. This data (Fig. 5c) was first processed to calculate production rate, and the width of the spikes was used along with input flow rates to measure the size and dispersity of generated droplets (Fig. 5d and e) at a maximum tested frequency of approximately 9000 min−1. We show droplet size controlled by the flow rate ratio, image file: c7lc00951h-t3.tif, and size C.O.V. of 1–10% within this range, which is comparable to droplet generators with the same geometry made by conventional microfluidic fabrication such as PDMS replica molding.53 The use of the brick sensor data to estimate droplet size was validated by video analysis, using a smartphone camera (Samsung Galaxy S6) or high-speed camera (Phantom 660), depending on the droplet production rate.

Moreover, the brick-based approach enables integration of wafer-based microfluidics; in Fig. S10 we show the mounting of a commercial glass droplet generation chip (Micronit). Bricks were assembled to align with the inlet and outlet points of the chip, and in a fashion similar to the above, bricks served to position the chip, feed in fluid, and characterize the production of droplets in a reconfigurable manner. This allowed us to generate droplets at smaller sizes (reaching 1 nL, governed by channel geometry) than we could in untreated micromilled Lego bricks, and shows an alternate utility of the modular approach to more easily integrate and interconnect conventionally microfabricated components.

g. 3D printed bricks and attachments

The dimensional qualities of molded bricks and micromilled channels exceed the performance of current desktop 3D printing techniques; however, 3D printing enables rapid fabrication of complementary pieces for these systems, especially when they require larger channels, large extended features, or internal geometries that cannot be micromilled.

We fabricated (see methods) brick-like structures using fused deposition modeling (FDM; ABS with Stratasys Mojo, layer height 0.17 mm, Fig. 6c) and stereolithography (SLA; Formlabs Clear resin with Form2, layer height 0.025 mm, Fig. 6b). Bricks were printed at multiple orientations and a best orientation was chosen for each process that maximized width accuracy and planar isotropy in repeatability (Fig. S11). The orientation influenced the surface quality. When the bricks were printed at an angled orientation, the layer-wise corrugations resulted in higher surface roughness on the brick sidewalls and within the channels.


image file: c7lc00951h-f6.tif
Fig. 6 Evaluation of 3D printed bricks, made by SLA and FDM techniques. (a) Comparison of cost and performance attributes of 3D printed bricks compared to stock LEGO bricks with milled channels (normalized by the value for milling). Error bars indicate general range of values. Side-by-side images of orange injection-molded LEGO bricks with (b) SLA bricks and (c) FDM bricks printed from CAD models of LEGO bricks. (d) A view of an SLA brick right after printing, showing the support structures. (e) An SLA 3D-printed inlet, which maintained a smooth profile through the brick, increasing efficiency and cleanliness. (f and g) 3D printed camera mount imaging a system and (h) image take with that mount.

In Fig. 6a we present a comparison between LEGO bricks with micromilled channels 3D printed bricks that include channels. This shows micromilling gives 10× smaller features, 10–100× smoother surfaces and 10× more consistent dimensions than 3D printing, as well as a much higher proportion of pieces that meet tolerances to snap together, and improved long-term stability in fluids. Nevertheless, the mounting repeatability FDM and SLA bricks (2.5 and 8 μm, respectively) was nearly equivalent to LEGO bricks (also see Fig. S12). The repeatability of SLA-printed bricks improved after the first 5–10 cycles, indicating that the contact surfaces were wearing. In addition, SLA bricks were prone to fracture after 50–100 mounting cycles (Fig. S12b). The width variation of SLA and FDM brick widths was σ ∼ 70 and 30 μm, respectively, leading to a total interchangeability of 70 and 30 μm, respectively. The dimensions of SLA bricks increased after contact with water, indicating absorption and reducing long-term interchangeability.

One advantageous use of 3D printing for modular systems is fabrication of inlet and outlet bricks with smooth, wide internal channels that change orientation within the brick (Fig. 6e). Compared to the LEGO-based inlet assembly (Fig. 1d), the 3D printed geometry also reduces dead space, increasing efficiency and cleanliness of the fluid inside, particularly if the brick was to be regularly flushed and reused with multiple fluids. However, we have found no commercial polymers that can be 3D printed that are as transparent as injection-molded polycarbonate, dimensionally stable in contact with water, and stiff enough to be precisely located via the elastic averaging approach.

The repeatability of 3D-printed LEGO interfaces allows fabrication of other peripheral components, such as a smartphone camera holder (Fig. 6f–h), which was used to image several experiments described in this article (images in Fig. 1, 3 and 5, and Video S1). When the phone and mount were removed and reattached, measurements of images taken with the camera had <0.25 mm standard deviation of position for a focus 80 mm from the camera. We also used a $10 macrolens phone attachment to magnify the image 20× to take images and video of fluid flow with resolution up to 20 μm per pixel.

IV. Discussion

A summary of the performance of the modified LEGO brick-based microfluidic system is presented in Table 1. All in all, micromilling of LEGO bricks provides an accessible route to prototyping microfluidic devices with ∼10 μm–1 mm critical dimensions, and ∼μm-scale repeatability which allows easy reconfiguration and integration of external imaging and sensing hardware. The cost of each unit is significantly lower than lithographic fabrication, comparing ∼$0.15 for a brick to ∼$50 for a lithographic microdevice, and ∼$5000 for a desktop micromilling machine to much more for photolithography and cleanroom infrastructure. In addition, while 3D-printing is continuing to advance, it likely will be many years before it can match injection-molded LEGOs in optical clarity, surface smoothness, and tight tolerance of dimensions.

However, the LEGO-based modular microfluidic system presented here has several opportunities for improved functionality. The ∼1 μm roughness of the milled channels, while normal for milled plastics, is rough compared to typical glass and PDMS surfaces. Also, the minimum channel dimensions achieved here, 50 μm × 150 μm, are too large for many important operations including single-cell analysis and generation of very small and precise droplets via flow constrictions. In addition, the modularity itself can introduce complications, as each sealed point between blocks is a potential failure point. However, once a system is designed, and in cases where saving space is a priority, a series of blocks can be integrated into a single chip (for example, Fig. 1d). Similar to how a breadboard is useful for prototyping electronic circuits, an expanded kit of microfluidic bricks can be used to prototype a fluidic system before the final design is determined and fabricated as a single chip. Because all four vertical faces of a block can be used for microfluidic pathways, and blocks can stand face-to-face, a modular brick-based system can fit into a tight area.

Further, while milling does enable accurate control of channel depth, it does not allow freeform channel pathways such as embedded helices, which would be possible with 3D printing. With these features and limitations in mind, the brick-based microfluidic system can find application in biological analysis and treatment, especially in general sample preparation, incubation, and cell culture, as well as in chemical titrations, controlled mixing of fluids, and droplet generation.

The uncoated thermoplastic material of LEGO bricks is not compatible with most organic solvents; however, we have additionally explored methods to improve the chemical compatibility by applying a thin layer of a resistant material, such as Parylene-C (Fig. S13). This coating successfully protected bricks from a variety of organic solvents that damaged uncoated bricks, including acetonitrile, dimethyl sulfoxide, tetrahydrofuran, toluene, dichloromethane, N,N-diisopropylethylamine, hexanes, dimethylformamide, and acetone. In this case, the Parylene-C would coat the entire internal channel, including the underside of the film. In addition, the O-ring may be selected from a variety of materials, for example Viton, which may be used with most fluids except esters and ketones, and Kalrez, which may be used with acids, acetone, and other organic solvents. In this case, the Parylene-C could be scratched off with force but was not removed by the presence of water or other mentioned solvents, or by assembling the LEGO system. In future work, the adhesion could be more quantifiably determined by scratch tests before and after a specified duration of exposure to fluids. Another alternative to improve chemical compatibility would be to micromachine entire bricks out of a material like PTFE or PE, or use injection molding to manufacture standard units, followed by micromilling for customization. If the manufacturing tolerances are sufficiently tight, fabricated bricks with the same mechanical structure should demonstrate interchangeability similar to current LEGO bricks. The polyethylene sealing film is the most limiting in chemical compatibility, due to its adhesive and the possibility of trapping debris. The film may be heat-welded on, removing the adhesive from the system.

In addition, beyond the examples of sensor bricks here, a much larger variety of functional bricks are possible, including those performing filtration, heating, sorting/separations, and pressure sensing. The ability to mount parts on 3D printed stands facilitates easy interfacing of peripheral hardware, such as syringe pumps, actuators, and valves. Milling and assembly of microfluidic bricks can also create non-planar and three-dimensional architectures, which can reduce the device footprint, enhance advection, and enable new structural designs such as the brick for nonplanar advective mixing in Fig. 1d, and the concept for three-dimensional fluid splitting in Fig. S14. The brick-based designs and their fabrication instructions are also ideal for dissemination using standard CAD/CAM file formats (see ESI design files) as a basis for others to modify the designs for specific needs, and to contribute new designs to the community.

V. Conclusions

We presented a new approach to rapidly construct self-aligning modular microfluidic systems by modification and assembly of interlocking injection-molded blocks. This was demonstrated using micromilling of store-bought LEGO® bricks to create surface fluidic pathways on bricks, and includes procedures for sealing and interconnecting bricks to form modular, reconfigurable, and self-aligning microfluidic systems. By adding bricks with different functionalities in series, in a few minutes we were able to create a variety of microfluidic systems with coupled analysis. To our knowledge, this is the first demonstration of a fully reconfigurable modular microfluidic system that both features reversible, non-permanent seals, and that has precise channel dimensions below 250 μm. The standard interface among all LEGOs will maintain the consistency and interchangeability of the platform, and enable the convergence of mechanical, fluidic, and other elements making a “lab on a brick” a new and viable platform for advancing prototyping of new microfluidic systems.

Conflicts of interest

CEO and AJH have a pending patent application on injection-molded modular interconnecting microfluidic blocks. The authors have no other conflicts of interests to declare.

Acknowledgements

C. E. O was supported by the National Science Foundation Graduate Research Fellowship (Grant # 1122374), the MIT Mechanical Engineering Department Ascher H. Shapiro Fellowship, and the Department of Defense (DoD) through the National Defense Science & Engineering Graduate Fellowship (NDSEG) Program. Materials and components were purchased using funds from the MIT Lincoln Laboratory Advanced Concepts Committee, a Faculty Award to A. J. H. from 3M Corporation, and the National Science Foundation EAGER/Cybermanufacturing Program (Grant # 1547154). We thank Felice Frankel for advice and assistance with photography and image design. We thank Abhinav Rao, Dale Thomas, Justin Beroz, Ron Rosenberg, Shannon Miller, Brian Solomon, Stuart Baker, Ryan Oliver, Nick Dee, and Hangbo Zhao for helpful discussions. This work made use of the MRSEC Shared Experimental Facilities at MIT, supported by the National Science Foundation under award number DMR-14-19807.

References

  1. G. M. Whitesides, The origins and the future of microfluidics, Nature, 2006, 442, 368–373 CrossRef CAS PubMed.
  2. E. K. Sackmann, A. L. Fulton and D. J. Beebe, The present and future role of microfluidics in biomedical research, Nature, 2014, 507, 181–189 CrossRef CAS PubMed.
  3. K. C. Bhargava, B. Thompson and N. Malmstadt, Discrete elements for 3D microfluidics, Proc. Natl. Acad. Sci. U. S. A., 2014, 111, 15013–15018 CrossRef CAS PubMed.
  4. S. Yadav, Analysis of value creation and value capture in microfluidics market, MS thesis, Massachusetts Institute of Technology, 2010 Search PubMed.
  5. G. Salgado and S. F. Pinheiro, Barriers to the Diffusion of Microfluidics from Research to Market, MS thesis, Catolica Lisbon, 2016 Search PubMed.
  6. Y. Xia and G. M. Whitesides, Soft Lithography, Annu. Rev. Mater. Sci., 1998, 28, 153–184 CrossRef CAS.
  7. P. Garstecki and G. M. Whitesides, Formation of bubbles and droplets in microfluidic systems, 2005, vol. 53 Search PubMed.
  8. D. C. Duffy, J. C. McDonald, O. J. A. Schueller and G. M. Whitesides, Rapid Prototyping of Microfluidic Systems in Poly(dimethylsiloxane), Anal. Chem., 1998, 70, 4974–4984 CrossRef CAS PubMed.
  9. K. Iwai, et al., Finger-powered microfluidic systems using multilayer soft lithography and injection molding processes, Lab Chip, 2014, 14, 3790–3799 RSC.
  10. B. Weigl, G. Domingo, P. Labarre and J. Gerlach, Towards non- and minimally instrumented, microfluidics-based diagnostic devices, Lab Chip, 2008, 8, 1999–2014 RSC.
  11. R. D. Sochol, et al., 3D Printed Microfluidic Circuitry via Multijet-Based Additive Manufacturing, Lab Chip, 2015, 16, 668–678 RSC.
  12. A. K. Au, N. Bhattacharjee, L. F. Horowitz, T. C. Chang and A. Folch, 3D-printed microfluidic automation, Lab Chip, 2015, 15, 1934–1941 RSC.
  13. S. Mueller, T. Mohr, K. Guenther, J. Frohnhofen and P. Baudisch faBrickation: Fast 3D Printing of Functional Objects by Integrating Construction Kit Building Blocks, 2014, pp. 3827–3834 Search PubMed.
  14. A. Konda, J. M. Taylor, M. A. Stoller and S. A. Morin, Reconfigurable microfluidic systems with reversible seals compatible with 2D and 3D surfaces of arbitrary chemical composition, Lab Chip, 2015, 15, 2009–2017 RSC.
  15. S. Tsuda, et al., Customizable 3D Printed ‘Plug and Play’ Millifluidic Devices for Programmable Fluidics, PLoS One, 2015, 10, e0141640 Search PubMed.
  16. K. G. Lee, et al., 3D printed modules for integrated microfluidic devices, RSC Adv., 2014, 4, 32876 RSC.
  17. P. F. O'Neill, et al., Advances in three-dimensional rapid prototyping of microfluidic devices for biological applications, Biomicrofluidics, 2014, 8, 52112 CrossRef PubMed.
  18. A. K. Au, W. Lee and A. Folch, Mail-order microfluidics: evaluation of stereolithography for the production of microfluidic devices, Lab Chip, 2014, 14, 1294–1301 RSC.
  19. A. J. L. Morgan, et al., Simple and Versatile 3D Printed Microfluidics Using Fused Filament Fabrication, PLoS One, 2016, 1–17 CAS.
  20. K. Vittayarukskul and A. P. Lee, A truly Lego ®-like modular microfluidics platform, J. Micromech. Microeng., 2017, 27, 35004 CrossRef.
  21. N. P. Macdonald, et al., Comparing Microfluidic Performance of Three-Dimensional (3D) Printing Platforms, Anal. Chem., 2017, 89, 3858–3866 CrossRef CAS PubMed.
  22. A. J. L. Morgan, et al., Simple and versatile 3D printed microfluidics using fused filament fabrication, PLoS One, 2016, 11, 1–17 Search PubMed.
  23. W. G. Patrick, et al., DNA Assembly in 3D Printed Fluidics, PLoS One, 2015, 10, e0143636 Search PubMed.
  24. P. K. Yuen, Embedding objects during 3D printing to add new functionalities, Biomicrofluidics, 2016, 10, 44104 CrossRef PubMed.
  25. H. Gong, B. P. Bickham, A. T. Woolley and G. P. Nordin, Custom 3D printer and resin for 18 μm × 20 μm microfluidic flow channels, Lab Chip, 2017, 17, 2899–2909 RSC.
  26. C. E. Owens and A. J. Hart, Lego Brick Microfluidics, in MicroTAS 2016 Conference Papers, 2016, pp. 934–935 Search PubMed.
  27. N. Nivedita and I. Papautsky, Continuous separation of blood cells in spiral microfluidic devices, Biomicrofluidics, 2013, 7, 1–14 CrossRef PubMed.
  28. G. Byrne, D. Dornfeld and B. Denkena, Advancing Cutting Technology, CIRP Ann., 2003, 52, 483–507 CrossRef.
  29. V. Venkatesh, N. Swain, G. Srinivas, P. Kumar and H. C. Barshilia, Review on the Machining Characteristics and Research Prospects of Conventional Micro-Scale Machining Operations, Mater. Manuf. Processes, 2016, 32, 235–262 CrossRef.
  30. A. H. Slocum, Precision Machine Design, Society of Manufacturing Engineers, 1992 Search PubMed.
  31. Parker O-ring Handbook, 2001, p. 56 Search PubMed.
  32. B. Zhuang, et al., A fully integrated and automated microsystem for rapid pharmacogenetic typing of multiple warfarin-related single-nucleotide polymorphisms, Lab Chip, 2016, 16, 86–95 RSC.
  33. J. Zhang, et al., Fundamentals and Applications of Inertial Microfluidics: A Review, Lab Chip, 2016, 16, 10–34 RSC.
  34. K. K. Zeming, et al., Real-time modulated nanoparticle separation with an ultra-large dynamic range, Lab Chip, 2016, 16, 75–85 RSC.
  35. O. Wagner, et al., Biocompatible fluorinated polyglycerols for droplet microfluidics as an alternative to PEG-based copolymer surfactants, Lab Chip, 2016, 16, 65–69 RSC.
  36. K. Tsougeni, G. Papadakis, M. Gianneli, A. Grammoustianou, V. Constantoudis, B. Dupuy, P. Petrou, S. Kakabakos and A. Tserepi, E. G. and E. G. Plasma nanotextured polymeric lab-on-a-chip for highly efficient bacteria capture and lysis, Lab Chip, 2015, 1, 120–131 Search PubMed.
  37. S. Tang, Y. Fang, Z. Liu, L. Zhou and Y. Mei, Tubular optical microcavities of indefinite medium for sensitive liquid refractometers, Lab Chip, 2016, 16, 182–187 RSC.
  38. F. Stumpf, et al., LabDisk with complete reagent prestorage for sample-to-answer nucleic acid based detection of respiratory pathogens verified with influenza A H3N2 virus, Lab Chip, 2016, 16, 199–207 RSC.
  39. A. Spencer, et al., A high-throughput mechanofluidic screening platform for investigating tumor cell adhesion during metastasis, Lab Chip, 2016, 16, 142–152 RSC.
  40. S. U. Senveli, et al., A surface acoustic wave biosensor for interrogation of single tumour cells in microcavities, Lab Chip, 2016, 16, 163–171 RSC.
  41. F. Schuler, et al., Digital Droplet PCR on Disk, Lab Chip, 2015, 16, 208–216 RSC.
  42. C. Martino, C. Statzer, D. Vigolo and A. J. deMello, Controllable generation and encapsulation of alginate fibers using droplet-based microfluidics, Lab Chip, 2016, 16, 59–64 RSC.
  43. K. Han, Y.-J. Yoon, Y. Shin and M. K. Park, Self-powered switch-controlled nucleic acid extraction system, Lab Chip, 2016, 16, 132–141 RSC.
  44. E. M. Hamad, et al., Inkjet printing of UV-curable adhesive and dielectric inks for microfluidic devices, Lab Chip, 2016, 16, 70–74 RSC.
  45. A. C. Glavan, A. Ainla, M. M. Hamedi, M. T. Fernández-Abedul and G. M. Whitesides, Electroanalytical Devices with Pins and Thread, Lab Chip, 2015, 16, 112–119 RSC.
  46. W. A. Cox-Muranami, E. L. Nelson, G. P. Li and M. Bachman, Large area magnetic micropallet arrays for cell colony sorting, Lab Chip, 2016, 16, 172–181 RSC.
  47. Z. Z. Chong, et al., Active droplet generation in microfluidics, Lab Chip, 2016, 16, 35–58 RSC.
  48. X. Zhang, Q. Chen, Z. Li and A. Jian, Optofluidic tunable lenses using laser-induced thermal gradient, Lab Chip, 2015, 16, 104–111 Search PubMed.
  49. Y. B. Bae, et al., Microfluidic assessment of mechanical cell damage by extensional stress, Lab Chip, 2016, 16, 96–103 RSC.
  50. H. Babahosseini, et al., The impact of sphingosine kinase inhibitor-loaded nanoparticles on bioelectrical and biomechanical properties of cancer cells, Lab Chip, 2015, 16, 188–198 RSC.
  51. A. Aung, et al., 3D cardiac μtissues within a microfluidic device with real-time contractile stress readout, Lab Chip, 2016, 16, 153–162 RSC.
  52. S. L. Anna, Droplets and Bubbles in Microfluidic Devices, 2016,  DOI:10.1146/annurev-fluid-122414-034425.
  53. P. Zhu and L. Wang, Lab on a Chip Passive and active droplet generation with microfluidics: a review, Lab Chip, 2017, 17, 34–75 RSC.
  54. P. Garstecki, M. J. Fuerstman, H. A. Stone and G. M. Whitesides, Formation of droplets and bubbles in a microfluidic T-junction-scaling and mechanism of break-up, Lab Chip, 2006, 6, 437–446 RSC.
  55. B. N. J. Persson and C. Yang, Theory of the leak-rate of seals, J. Phys.: Condens. Matter, 2008, 20, 11 Search PubMed.
  56. P. K. Yuen, A reconfigurable stick-n-play modular microfluidic system using magnetic interconnects, Lab Chip, 2016, 16, 3700–3707 RSC.

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c7lc00951h

This journal is © The Royal Society of Chemistry 2018