Lab on a Chip View Article Online
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
PAPER
Cite this: Lab Chip, 2014, 14, 424
View Journal | View Issue
PDMS lab-on-a-chip fabrication using 3D printed templates Germán Comina, Anke Suska and Daniel Filippini* The fabrication of conventional PDMS on glass lab-on-a-chip (LOC) devices, using templates printed with a commercial (2299 US$) micro-stereo lithography 3D printer, is demonstrated. Printed templates replace clean room and photolithographic fabrication resources and deliver resolutions of 50 μm, and up to 10 μm in localized hindrances, whereas the templates are smooth enough to allow direct transfer and proper sealing to glass substrates. 3D printed templates accommodate multiple thicknesses, from 50 μm up to several mm within the same template, with no additional processing cost or effort. This
Received 19th August 2013, Accepted 16th October 2013
capability is exploited to integrate silicone tubing easily, to improve micromixer performance and to produce multilevel fluidics with simple access to independent functional surfaces, which is illustrated by time-resolved glucose detection. The templates are reusable, can be fabricated in under 20 min, with
DOI: 10.1039/c3lc50956g
an average cost of 0.48 US$, which promotes broader access to established LOC configurations with minimal fabrication requirements, relieves LOC fabrication from design skills and provides a versatile
www.rsc.org/loc
LOC development platform.
Introduction PDMS (polydimethylsiloxane) on glass lab-on-a-chip (LOC) devices constitute a prevalent LOC configuration, and a comprehensive source of proven solutions.1 Although simpler and more affordable than alternatives such as micro electro mechanical systems, it still requires specialized skills and resources,1 which condition access for potential users.1,2 In this work, we simplify PDMS on glass LOC (PDMS-LOC) fabrication, replacing clean room resources by affordable 3D printed templates. Efforts to simplify LOC device fabrication have produced alternative methods, which demand fewer resources than standard photolithographic procedures. Examples of these alternatives are microscope projection techniques that simplify2,3 or even eliminate the need for photolithographic masks.4–6 However, clean room access is still necessary for coating and development procedures, which multiply with design complexity. For instance, the seemingly trivial incorporation of multiple thicknesses in a template entails additional coating steps, masks and alignment procedures.7,8 Direct printing of LOC, implemented by micro stereo lithography (MS4,5), has traditionally required expensive infrastructure, whereas the emergence of affordable 3D printers has the potential to grant broader access to custom Optical Devices Laboratory – Department of Physics, Chemistry and Biology (IFM), Linköping University, Linköping 58183, Sweden. E-mail: [email protected]; Tel: +46 13 28 1282
424 | Lab Chip, 2014, 14, 424–430
made microstructures, by minimizing resources and fabrication skills. Economical 3D printers have been used to fabricate millifluidic and microfluidic devices,10 as well as reactionware for chemical synthesis and analysis.11 In the first case, a thermoplastic extrusion9,10 printer was used, whereas in the second instance a robocasting platform12 that enabled printing with different materials, demonstrated the potential for direct printing of functional substances, and precursors, chemically decorating active reactionware.11 This type of printer connects to the vision of users access to physical products by downloading and locally printing a digital design.11,13 However, these printer technologies have inherent limitations in resolution and surface roughness, which are determinant to fabricate templates compatible with current PDMS fluidics. The recent introduction of affordable 3D printers based on MS aided by digital micro mirror device (DMD) projection,14 offering 50 μm resolution and smooth surface finish, open up new possibilities for highly simplified and versatile PDMS-LOC fabrication. Here we study the use of an affordable MS 3D printer (Miicraft,14 2299 US$) to fabricate templates for classic PDMS-LOC devices.1,15 The resolution limits are investigated and tested on practical structures, such as diffusion and chaotic mixers.16,17 The inherent advantages of 3D printed templates, such as natural fabrication of micrometric channels combined with mm high elements in the same template, are exploited to integrate connection tubing sealed to the PDMS replica, to improve mixer geometry, and to
This journal is © The Royal Society of Chemistry 2014
View Article Online
Lab on a Chip
configure open measurement chambers for straightforward chemical functionalization in multilevel devices, illustrated by time resolved glucose detection. This first demonstration of 3D printed reusable templates for PDMS-LOC, also highlights an attractive fabrication cost of 0.48 US$ per template.
Paper
coating on completely dried out templates does not allow the PDMS to cure.
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Experimental Template fabrication Templates were designed with free computer aided design (CAD) software for Mac OS X 10.8.3 (Autodesk® Inventor® Fusion, Autodesk Inc.) on a MacBook Air computer (13-inch, late 2010, 1.86 GHz Intel Core 2 Duo, 4 GB 1067 MHz). The CADs are exported as .stl mesh files that are converted by the software provided with the 3D printer (Miicraft® Suite). The Miicraft® Suite allows for alignment and scaling of the .stl files, segmentation of the aligned 3D mesh and printer control. Segmentation transforms the 3D mesh in 2D cross sections (bitmap black and white files in .png format at 480 × 768 pixels resolution) in 50 μm steps. The .png files are accessible and may be modified for refinements and minor corrections common to progressive optimization. These corrections do not require CAD skills, and were performed in Photoshop™ CS4, although simpler tools capable of single pixel editing, such as Microsoft Paint, are adequate. Miicraft®14 is an MS 3D printer (2299 US$) with 450 ppi (~56 μm) lateral, and 50 μm vertical resolution. The Miicraft® Suite controlling the printing process allows setting exposure time for each layer, printing speed (2 or 3 cm per hour) and vertical separation between exposed layers (50 and 100 μm); this final setting provides an easy way to duplicate the thickness of a design sliced at 50 μm per layer. After optimization, our templates were printed under the following conditions of exposure time, printing speed and vertical step size: Fig. 1: 12 s, 2 cm per hour, 50 μm; Fig. 2: 11 s, 3 cm per hour, 50 μm and 100 μm (for the 500 μm deep channels); Fig. 3: 10 s, 3 cm per hour, 50 μm. After printing, the templates were sonicated (FinnSonic m15) in industrial grade ethanol for 20 s, and air-dried. The dry templates were post cured in UV light for 600 s, in the post-curing chamber of the printer. The printer uses a proprietary resin (138 US$ per 500 ml), with different proportions of modified acrylate, modified acrylate oligomer, acrylate monomer, epoxy monomer, photo initiator and additives as principal components. The specific formulation is not disclosed. Since PDMS does not cure on the templates, the surfaces must be protected with a PDMS compatible material before casting the polymer. We developed a procedure consisting of 2 min sonication in industrial grade ethanol and subsequent ink airbrushing (Neo CN Iwata, gravity feed dual action airbrush with 0.35 mm nozzle). Effective surface protection requires the ink (Pentel NN60) to be applied in thin layers immediately after the template surface is taken from the ethanol and begins to dry. Ink
This journal is © The Royal Society of Chemistry 2014
Fig. 1 (a) CAD with 50 μm and 2 mm features combined in the same design. (b) 3D printout of the structure in (a). The inset highlights hindrances developed at the 50 μm to 100 μm channel step. (c) Image of the printed pillar guides used to integrate silicone tubing in the PDMS structure. (d) PDMS device assembled on a glass substrate showing no texture at the PDMS–glass interface. (e) Microscope image of tubing connection to a 500 μm channel. (f) Epi-fluorescence microscope image of PDMS structure filled with fluorescein solution.
Fig. 2 (a) Template for diffusion and chaotic mixers. The pillars are guides for silicone tubing integration and the external frame contains the PDMS during curing. (b) Detail of the chaotic mixer design. (c) Detail of the chaotic bitmap edited refinements. (d) Printed template. (e) 100 μm deep diffusion mixer showing the mixture of 1 mM fluorescein with 1 mM rhodamine B at 60 μl min−1. (f) 100 μm deep chaotic mixer imaged in the same conditions as in (e). (g) Idem to (f) operated at 30 μl min−1. (h) Idem to (g) but for a 500 μm deep channel. (i) Detail of the 500 μm deep chaotic mixer 3D structure.
Lab Chip, 2014, 14, 424–430 | 425
View Article Online
Paper
Lab on a Chip
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
rendered apparent by a semi-glossy finish. It takes some practice to paint templates consistently with an optimum finish, but if it is inadequate, the template can be rinsed in ethanol and airbrushed again. Templates with insufficient paint do not allow PDMS to cure and are wasted, whereas templates with an excess of paint, characterized by a matte finish, permit PDMS curing but transfer paint to the PDMS replicas. These stained PDMS replicas can be cleaned by a 2 min sonication in ethanol. Surface roughness was measured with a stylus profiler (Dektak 6, Veeco Instrument Inc.) along 1 mm tracks resolved in 6000 points. The average roughness was 182 nm for the finished templates and 540 nm after applying the protective coating (1.08% of the minimum specified thickness). Templates were completed with the manual insertion of silicone tubing (Esska.de GmbH, Hamburg) in the template guides. Designed guide diameters, required for a proper sealing of the tubing within the channels, were: 0.6 mm, 0.9 mm and 1.6 mm for tubing inner diameters of 0.5 mm, 0.8 mm and 1.5 mm respectively. The average weight of each template in this work was 1.8 g, which corresponded to a resin cost of 0.48 US$ per template. Carefully treated templates could be reused. Mechanical failure of delicate tubing guides, which broke off and detached with the cured PDMS, occurred after 4 re-uses in the best case. Templates with more robust features did not show mechanical failure, but were not explicitly tested to gauge the failure limit for the surface coating.
PDMS devices
Fig. 3 (a) Templates used for multilevel fluidics. The blue coloured template is symmetric along the ss′ direction and the PDMS replica from the red coloured template is flipped over for assembly with the blue replica and placed on the glass substrate. The cross-section highlights the use of different thicknesses in the template to configure a measurement chamber, with access to a functionalized substrate. (b) Filling of the multilevel structure with fluorescein solution and fluorescence microscopy detail of the inter-level connections in 1, 2, 3, 4. (c) Idem to (b) 130 ms elapsed after injection. (d) +460 ms. (e) +920 ms. (f) Detail of tubing connection with channel space. (g) Cross section of open measuring chamber connected to 100 μm deep channel. (h) Response to 2.5, 5 and 10 mM glucose concentrations and inset showing the colour change for 10 mM, with respect to the control (ctrl.).
1.5 ml of 2 : 1 ethanol–ink solution is necessary to cover a single template properly (up to 43 × 27 mm2), which is
426 | Lab Chip, 2014, 14, 424–430
To fabricate the PDMS devices, Dow Corning Sylgard 184 base18 and curing agent were mixed in 10 : 1 proportion and stirred in a cup for 2 min. The mixture was degassed in a desiccator connected to a rotary pump for 30 min, and subsequently poured on the coated templates. The PDMS filled templates were cured at 65 °C for 2 h in an environmental oven (Gallem Kamp incubator) and the PDMS replicas, with embedded tubing attached, were separated with tweezers and placed directly on clean glass slides (Menzel-Glaser, Braunschweig, Germany). Only in the case of chaotic mixers, which were tested up to 90 μl min−1 (continuous infusion/Dual NE-1000 syringe pump, from New Era Pump Systems Inc., http://www.syringepump.com), the PDMS and glass were oxygen plasma treated (Diener electronic, Pico model) during 1 min, for greater adherence.
Glucose sensing The assay for glucose sensing utilizes glucose oxidase (GOx), horseradish peroxidase (HRP) and a non-coloured substrate (Ampliflu Red, also known as Amplex Red). GOx catalyzes the oxidation of glucose to glucono delta-lactone and hydrogen peroxide, which reacts with HRP stoichiometrically 1 : 1 to catalyze de-N-acetylation and oxidation of colorless Ampliflu Red (10-acetyl-3,7-dihydroxyphenoxazine) to resorufin.19,20
This journal is © The Royal Society of Chemistry 2014
View Article Online
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Lab on a Chip
Resorufin is a pink-colored, red fluorescent compound, with an absorption maximum of 563 nm. Resorufin is also a substrate for HRP, which further oxidizes it to a colourless/ nonfluorescent compound.19,20 Response time can be widely regulated through the proportions of constituents. The assay in this instance contained: 10 units ml−1 of glucose oxidase enzyme activity, 10 units ml−1 of horseradish peroxidase activity, 250 μM Ampliflu Red and 0.3 M trehalose in phosphate buffered saline (PBS, pH 7.4). All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA). For assay characterization, 1 μl of assay reagent was dropped on top of flat PDMS, and 1 μL of glucose in PBS, pH 7.4 (0; 2.5; 5; 10 mM) was added while video imaged with a Galaxy Note 2 rear camera (see details in the next sub-section). Surface functionalization of the PDMS-LOC microfluidic chamber was accomplished depositing the assay reagent on a flat PDMS surface and allowing it to dry in darkness, at room temperature, for 2 hours. The chamber was sealed to the functional surface, and a 10 mM glucose solution, followed by an air bubble, was pumped into the device at 30 μl min−1 (Dual NE-1000 syringe pump). The pumping was stopped when the solution reached the measuring chamber, while the device was imaged at 30 fps with the rear camera of a Galaxy Note 2 smart phone.
Imaging Template imaging (Fig. 1b and c, Fig. 2d) was implemented with an Olympus SZ60 stereo zoom microscope, fitted with a Canon EOS 500D DSLR camera (15 mPixels, APS-C cmos sensor). Bright field and epi-fluorescence imaging was conducted on a Zeiss Axiovert 40 cfl inverted routine microscope. A Zeiss HBO50 illuminator, housing a mercury vapour short arc lamp HBO 50, provided fluorescence excitation. The Canon EOS 500D camera serviced both microscopes. A Samsung Galaxy Note 2, 8 Mpix rear side camera was used in photo mode for LOC imaging and for full HD video acquisition (30 fps), for which purpose it was positioned facing the LOC devices placed on a glass stage back illuminated by an iPod Touch (4th generation, 960 × 640 pixel screen resolution, with iOS 6) running Led Torch v1.37 app (www.smallte.ch), set to pure white (rgb 255, 255, 255). Glucose response, for three test concentrations and control, was video captured with the Samsung Galaxy Note 2 rear camera and converted to a .jpeg image sequence at 0.5 fps using MPEG Streamclip v 1.9.2. This sequence was processed with a dedicated script, in Matlab R2008b, developed for this purpose. A manual grid was used to select the 4 regions of interest (ROI) (control and 3 glucose concentrations) and extract average intensities in the red, green and blue channels representing the ROIs of the composition (Fig. 3h). The procedure was automatically repeated in all pictures extracted from the original video. The time response data was copied and pasted to Microsoft Excel 2011 for Mac, for final editing. Pictures in
This journal is © The Royal Society of Chemistry 2014
Paper
this communication were contrasted and composed with Apple Keynote 09, v5.3.
Results and discussion Micro stereo lithography 3D printers, like the Miicraft14 used in this work, utilize a DMD to illuminate a photocurable resin with precise 2D black and white patterns at a resolution of 450 ppi (~56 μm), and to displace successive exposures in the vertical dimension at 50 μm steps, within a maximum printing volume of 43 × 27 × 180 mm3. 3D printing allows the fabrication of microstructures with multiple thicknesses in a single printout, whereas conventional photolithography would require multiple masks and repeated coatings, development and alignment procedures.1,7,8,16 Computer aided designs (CADs) are converted by the printer software into 2D bitmap image slices (.png format), which remain available for retouches and finishing details, without requiring CAD re-editing, thus easing optimization cycles. Since retouching only requires bitmap editing skills, with readily available software, such as Microsoft Paint, the specialized skills for LOC design and CAD composition can be completely separated from its fabrication, and the 3D printer becomes a generic vehicle to transmit the design. Achievable resolution depends on multiple factors, such as proximity between elements and thickness differences. In addition to the layout, the exposure time for each layer, as well as printing speed, determine the final result. Our planar templates are 2 mm thick and demand printing times under 20 min, weighing on average 1.8 g, with a cost of 0.48 US$ per template, which compares favourably with conventional fabrication services. Furthermore, printed templates can be reused several times. The sole material requirements for our technique are: 3D template printing, protective coating to allow PDMS curing, PDMS pouring and assembly. Sealing to external tubing is simplified, because the template can easily incorporate guides for silicone tubing, which become integrated to the PDMS replica. The test structure used to assess the limit of resolution achievable between close elements, and between elements of different thicknesses is depicted in Fig. 1. The structure incorporates features ranging from 50 μm to 2 mm and also illustrates the assembly of connection tubing directly on the template. Fig. 1a shows the 3D CAD with three 50 × 50 μm2 crosssection channels at a close spacing of 50 μm. These channels continue in three different directions and join to 100 × 100 μm2 sections, which in turn connect to 200, 300, and 500 × 500 μm2 cross sections, terminated in cylindrical tubing guides. The zoomed area (Fig. 1b) details the connection between 50 μm and 100 μm deep channels. At this resolution the step between different channel depths induces a repeatable artifact. Although unintended by design, these features are useful to create hindrances beyond the printer resolution, which
Lab Chip, 2014, 14, 424–430 | 427
View Article Online
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Paper
can enable particle separation,21 or trap detection chemistries on beads.22 The insertion of silicone tubing in the printed template guide (Fig. 1c) is implemented with guiding pillars (1.6 mm O.D.), with 1.5 mm I.D. tubing. The small mismatch is absorbed compressively and provides the necessary seal between channel and tube, required to prevent PDMS diffusion and subsequent blockages. Thus, a continuous duct is configured, connecting the micro channels to the tubing lumen (Fig. 1e). Since PDMS does not cure on Miicraft proprietary resin,14 the template surface had to be coated with a PDMS compatible material. Consequently, the templates were airbrushed with permanent ink (Pentel NN60), after which degassed PDMS was poured and cured at 65 °C for 2 hours, before releasing the PDMS replica and coupled tubing. PDMS adheres to silicone tubing solidly, providing a robust interconnection to the micro fluidics. The microscope image of the PDMS device mounted on a glass slide (Fig. 1d), captures the texture transferred to the PDMS replica (540 nm average roughness), which was only noticeable on free surfaces, such as channel ceilings, whereas on the large area in contact with the glass surface, this texture vanished completely. To test the integrity of the structure, the device was filled with a fluorescein solution and imaged in an epi-fluorescence microscope (Fig. 1f). No leakages or blockages were observed, whereas hindrances created on the template were functional and could support printer resolution up to 10 μm, for localized features. To better assess potential practical uses of the 3D printed templates, we investigated reactors incorporating chaotic mixer designs16,17,23,24 and the viability of multi-level PDMS fluidics.8 The laminar character of the flow, which restricts the mixing process to slow molecular diffusion, complicates rapid and efficient mixing in microfluidic devices.17 Passive micromixers introduce geometric features along the micro channels,25 which break the laminar flow inducing chaotic advection to improve mixing efficiency.17,21,24 The ability to fabricate the geometries required to affect the flow regime, with printed templates, is a practical test for this method. Comparative results between a chaotic mixer and a diffusion reactor of the same dimensions are represented in Fig. 2. The printed template (Fig. 2a) incorporates both mixers in an external frame 1 mm high, which is used to confine the poured PDMS to the template, thus making the fabrication procedure cleaner and simpler. The bottom picture corresponds to the template after PDMS curing, when it is ready to be detached. Details of the mixer geometric obstacles, as specified by CAD, and the flow direction are indicated in Fig. 2b, whereas final adjustments to the design, made directly on the .png bitmap images, are shown in Fig. 2c. The resulting printout, for optimized exposure time and printing speed (11 s, 3 cm hour−1), is shown in Fig. 2d. To create a mixer, the texture in the template is advantageous, since the obstacles are not just smooth surfaces but an intricate 3D boundary (Fig. 2i) that contributes to the mixing efficiency.
428 | Lab Chip, 2014, 14, 424–430
Lab on a Chip
The behaviour of 1 mM of fluorescein and rhodamine B aqueous solutions, injected at 60 μl min−1 in the 100 μm deep reactor is shown in Fig. 2e. The solutions do not mix even at the distal end of the structure, which is underscored by a black arrow to indicate the last position where two colours can be distinguished clearly. In contrast, the chaotic design mixed both solutions close to the proximal side of the channel (Fig. 2f), and even at 30 μl min−1, it mixed the substances before the distal side (Fig. 2g). A crucial advantage of printed templates is that changing channel thickness does not imply additional cost or different materials and protocols. Fig. 2h captures the performance of a 500 μm deep chaotic mixer at 30 μl min−1. As expected, mixing efficiency improves, as the Reynolds number scales with channel thickness,17 but the thicker channel also introduces complex 3D topologies (Fig. 2i), hard to achieve in conventional methods, which contribute to turbulence. In addition to well-sealed channels, embedded connectors and practical microfluidic structures, 3D printed templates can be used to create microfluidic structures with more than one level. To connect between levels, through holes could easily be created using printed pillars rising above the PDMS surface; however, this configuration creates meniscuses and uneven PDMS surfaces, which preclude an efficient stacking of multiple levels. The alternative explored in this work entirely eliminates the need for through holes and the effect of meniscuses. The problem was solved using two PDMS replicas with channels, which overlap only at the desired interconnection locations. To create these connections, once the replicas were fabricated, one was flipped over and placed in contact with the second, facing the flat bottom sides. Thus, overlapping channels created interconnections, and the flat contact surfaces provided a robust sealing for the complete circuit. Flipping over one replica of the assembly requires designing one template as the mirror image of the desired result. To simplify design and avoid mirroring, one template has a generic configuration (blue template in Fig. 3a) and is symmetric with respect to the ss′ axis. Under these conditions, the custom layer (created with the red template in Fig. 3a) can be planned as an overlay of the blue template design, thus easily aligning connections between levels. Once the replicas are created, one of them can be flipped over along the ss′ axis without affecting relative alignment between layers. PDMS replicas were detached from the templates after 2 hours, at 65 °C. The custom replica (coloured red in Fig. 3a) was flipped over along the ss′ axis, aligned with the flat side of the generic replica, pressed together and returned to the oven at 65 °C for another 30 min before use. Resulting structures were perfectly sealed and ready to use. To test device integrity, a fluorescein solution was injected at 500 μl min−1 flow (Fig. 3b–e correspond to a 920 ms interval). It held tight with no need of plasma treatment for enhanced adhesion. The flow must traverse 10 inter level connections, to close the circuit, which proved simple to fabricate and reliable in
This journal is © The Royal Society of Chemistry 2014
View Article Online
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Lab on a Chip
operation. Connections were aligned visually, and fluorescence microscopy revealed well-defined channels and proper coupling between levels (Fig. 3 1–4). The continuous connection between tubing lumen and channel space, as well as the solid sealing between tubing and PDMS is apparent in Fig. 3f. Chemical functionalization is central to regular and multilevel fluidics, and can be implemented in various ways. Direct 3D printing of functional regions has been demonstrated in reactionware,11 using a different printing technology. MS 3D printers used in this work cannot change the printing materials easily, but both resolution and surface finish are better suited to make PDMS templates, and can easily combine thick and thin features in the same template. In our case, a thick feature was left above the PDMS level to configure a precisely defined open chamber that enabled communication of the two levels fluidic design with a functionalized surface, which could be prepared separately. The advantage is that surface chemistry could be independently controlled over an unconstrained area and then attached to the fluidics without special positioning demands (Fig. 3a and h). This concept was tested with time resolved glucose detection for concentrations within the diagnostic range. The response to three test concentrations of glucose (and control), within the clinically relevant range, are displayed in Fig. 3h. The sensing surface contains glucose oxidase (GOx), horseradish peroxidase (HRP) and a non-coloured substrate (Ampliflu Red). GOx catalyzes the oxidation of glucose to glucono delta-lactone and hydrogen peroxide, which is then used by HRP to catalyze de-N-acetylation and oxidation of colourless Ampliflu Red (10-acetyl-3,7-dihydroxyphenoxazine) to resorufin. Resorufin is a pink-coloured compound, which is also a substrate for HRP, which further oxidizes it to a colourless/nonfluorescent compound.19,20 The results in Fig. 3h illustrate the faster production of resorufin for higher glucose concentrations, due to higher levels of available hydrogen peroxide produced through GOx activity, and the subsequent further consumption of resorufin by HRP, that renders the indicator colourless. The dynamic characteristic of the response is suitable for quantitative detection independently of the colorimetric readout capabilities, which may vary substantially. In our case, colour is just used to track the minima, which are video acquired at 30 fps (a common feature of any modern cell phone) and can be followed with a resolution of 33 ms. For the response range of 300 s, in our experiment (Fig. 3h), 33 ms corresponds to 0.011% of the range, whereas 1 count in 255 levels of any colour channel used for colour quantification is 0.039% (about 35 times worse). In this work practical PDMS on glass LOC devices have been fabricated using minimal resources and procedure complexity. The capabilities conferred by affordable latest generation MS 3D printers, enable integration of multiple thicknesses in the same template at no extra cost or effort, while retaining all classical PDMS-LOC advantages. The technology offers the possibility to develop LOC devices at low cost, together with great flexibility and simplicity to
This journal is © The Royal Society of Chemistry 2014
Paper
iterate designs, thus supporting fast optimization cycles of multiple prototypes per day. The ability to create sealed connectors and to implement multilevel fluidic designs with robust interconnections renders this technology attractive for numerous development scenarios. Furthermore, the cost of reusable templates below 0.48 US$ per piece makes it suitable not only for development but also for limited production of specialized devices. 3D printed templates inherit the vision of affordable 3D printers delivering a physical version of a remotely conceived digital design, thus separating design skills from fabrication efforts, and greatly simplifying access to PDMS-LOC configurations.
Conclusions The fabrication of PDMS on glass LOC devices has been demonstrated using affordable MS 3D printers to produce templates. The proposed technique drastically simplifies device fabrication by eliminating the need for clean room facilities and repeated photolithographic steps required for templates with multiple thicknesses. Templates are reusable and can be produced in less than 20 min, at an average cost of 0.48 US$, remarkably accelerating development time, improving flexibility for optimization and cutting costs for multiple iterations. Because PDMS does not cure in contact with the commercial printer resin, a simple procedure to protect the template surface has been developed. Practical fluidic structures, with down to 10 μm hindrances, have been demonstrated, as well as the fabrication of micromixers and multilevel fluidics. 3D printed templates can easily incorporate elements of different thicknesses, which has been exploited in two explicit ways namely, the integration of sealed tubing and the generation of a measurement chamber in a multilevel structure, which can be independently functionalized for chemical sensing, in this case for glucose detection. Affordable last generation MS 3D printers offer the possibility to develop templates suitable for PDMS-LOC configurations of varied complexity at a fraction of cost, time and effort. The printer not only deals with a substantial part of the experimental tasks, which are thus simplified for the end user, but also as originally intended with 3D printer technologies,9,11–13 disentangles fabrication from design skills, rendering the printer a tool to transfer custom PDMS-LOC designs. The concept described here should enable broader access to established LOC solutions, and provide a versatile LOC development platform compatible with fast and affordable optimization.
Acknowledgements G. Comina’s research stay at Linköping University has been financed by the Peruvian National Council for Science and Technology (CONCYTEC). This research has been
Lab Chip, 2014, 14, 424–430 | 429
View Article Online
Paper
supported by grants from Linköping Centre for Life Science Technologies (LIST), Sweden, and from the Swedish Research Council (VR).
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Notes and references 1 Microfluidics and Nanofluidics Handbook, Fabrication, Implementation, and Applications, ed. S. Mitra and S. Chakraborty, CRC Press Taylor & Francis Group, Boca Raton, 2012. 2 J. Love, D. Wolfe, H. Jacobs and G. Whitesides, Microscope Projection Photolithography for Rapid Prototyping of Masters with Micron-Scale Features for Use in Soft Lithography Langmuir, 2001, 17, 6005. 3 J. Behm, K. Lykke, M. Pellin and J. Hemminger, Projection photolithography utilizing a Schwarzschild microscope and self-assembled alkanethiol monolayers as simple photoresists Langmuir, 1996, 12, 2121. 4 C. Sun, N. Fang, D. Wu and X. Zhang, Projection microstereolithography using digital micro-mirror dynamic mask Sens. Actuators, A, 2005, 121, 113. 5 X. Zhang, X. Jiang and C. Sun, Micro-stereolithography of polymeric and ceramic microstructures Sens. Actuators, A, 1999, 77, 149. 6 P. Preechaburana and D. Filippini, Fabrication of monolithic 3D micro-systems Lab Chip, 2011, 11, 288. 7 M. Peterman, P. Huie, D. Bloom and H. Fishman, Building thick photoresist structures from the bottom up J. Micromech. Microeng., 2003, 13, 380. 8 J. Anderson, D. Chiu, R. Jackman, O. Cherniavskaya, J. McDonald, H. Wu, S. Whitesides and G. Whitesides, Fabrication of topologically complex three-dimensional microfluidic systems in PDMS by rapid prototyping Anal. Chem., 2000, 72, 3158. 9 B. Evans, Practical 3D printers: The science and art of 3D printing, Apress, 2012. 10 P. Kitson, M. Rosnes, V. Sans, V. Dragone and L. Cronin, Configurable 3D-printed millifluidic and microfluidic ‘lab on a chip’ reactionware devices Lab Chip, 2012, 12, 3267. 11 M. D. Symes, P. J. Kitson, J. Yan, C. J. Richmond, G. J. T. Cooper, R. W. Bowman, T. Vilbrandt and L. Cronin, Integrated 3D-printed reactionware for chemical synthesis and analysis Nat. Chem., 2012, 4, 349. 12 E. Malone and H. Lipson, Fab@Home: the personal desktop fabricator kit Rapid Prototyping J., 2007, 13, 245.
430 | Lab Chip, 2014, 14, 424–430
Lab on a Chip
13 T. Vilbrandt, A. Pasko and C. Vilbrandt, Fabricating nature Technoetic Arts, 2009, 7, 165. 14 Miicraft® (no date), Miicraft 3D printer, available at: http:// www.miicraft.com/products/ (accessed: 13 August 2013). 15 A. Foudeh, T. Didar, T. Veresa and M. Tabrizian, Microfluidic designs and techniques using lab-on-a-chip devices for pathogen detection for point-of-care diagnostics Lab Chip, 2012, 12, 3249. 16 Y. Suh and S. Kang, A Review on Mixing in Microfluidics Micromachines, 2010, 1, 82. 17 N. Nguyen and Z. Wu, Micromixers – a review J. Micromech. Microeng., 2005, 15, R1–R16. 18 J. Lötters, W. Olthuis, P. Veltink and P. Bergveld, The mechanical properties of the rubber elastic polymer polydimethylsiloxane for sensor applications J. Micromech. Microeng., 1997, 7, 145. 19 V. Towne, M. Will, B. Oswald and Q. Zhao, Complexities in horseradish peroxidase-catalyzed oxidation of dihydroxyphenoxazine derivatives: appropriate ranges for pH values and hydrogen peroxide concentrations in quantitative analysis Anal. Biochem., 2004, 334, 290. 20 M. Zhou, Z. Diwu, N. Panchuk-Voloshina and R. P. Haugland, A stable nonfluorescent derivative of resorufin for the fluorometric determination of trace hydrogen peroxide: applications in detecting the activity of phagocyte NADPH oxidase and other oxidases Anal. Biochem., 1997, 253, 162. 21 I. Dimov, L. Basabe-Desmonts, J. Garcia-Cordero, B. Ross, A. Ricco and L. Lee, Stand-alone self-poweredintegrated microfluidic blood analysis system (SIMBAS) Lab Chip, 2011, 11, 845. 22 N. Lee, Y. Yang, Y. Kim and S. Park, Microfluidic immunoassay platform using antibody immobilized glass beads and its application for the detection of Escherichia coli O157:H7 Bull. Korean Chem. Soc., 2006, 27, 479. 23 S. Kwon, S. Kou, H. Kim, X. Chen, H. Hwang, S. Nam, So Kim, K. Swamy, S. Park and J. Yoon, Sensing cyanide ion via fluorescent change and its application to the microfluidic system Tetrahedron Lett., 2008, 49, 4102. 24 D. Therriault, S. White and J. Lewis, Chaotic mixing in threedimensional microvascular networks fabricated by directwrite assembly Nat. Mater., 2003, 2, 265. 25 V. Mengeaud, J. Josserand and H. Girault, Mixing Processes in a Zigzag Microchannel: Finite Element Simulations and Optical Study Anal. Chem., 2002, 74, 4279.
This journal is © The Royal Society of Chemistry 2014
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
PAPER
Cite this: Lab Chip, 2014, 14, 424
View Journal | View Issue
PDMS lab-on-a-chip fabrication using 3D printed templates Germán Comina, Anke Suska and Daniel Filippini* The fabrication of conventional PDMS on glass lab-on-a-chip (LOC) devices, using templates printed with a commercial (2299 US$) micro-stereo lithography 3D printer, is demonstrated. Printed templates replace clean room and photolithographic fabrication resources and deliver resolutions of 50 μm, and up to 10 μm in localized hindrances, whereas the templates are smooth enough to allow direct transfer and proper sealing to glass substrates. 3D printed templates accommodate multiple thicknesses, from 50 μm up to several mm within the same template, with no additional processing cost or effort. This
Received 19th August 2013, Accepted 16th October 2013
capability is exploited to integrate silicone tubing easily, to improve micromixer performance and to produce multilevel fluidics with simple access to independent functional surfaces, which is illustrated by time-resolved glucose detection. The templates are reusable, can be fabricated in under 20 min, with
DOI: 10.1039/c3lc50956g
an average cost of 0.48 US$, which promotes broader access to established LOC configurations with minimal fabrication requirements, relieves LOC fabrication from design skills and provides a versatile
www.rsc.org/loc
LOC development platform.
Introduction PDMS (polydimethylsiloxane) on glass lab-on-a-chip (LOC) devices constitute a prevalent LOC configuration, and a comprehensive source of proven solutions.1 Although simpler and more affordable than alternatives such as micro electro mechanical systems, it still requires specialized skills and resources,1 which condition access for potential users.1,2 In this work, we simplify PDMS on glass LOC (PDMS-LOC) fabrication, replacing clean room resources by affordable 3D printed templates. Efforts to simplify LOC device fabrication have produced alternative methods, which demand fewer resources than standard photolithographic procedures. Examples of these alternatives are microscope projection techniques that simplify2,3 or even eliminate the need for photolithographic masks.4–6 However, clean room access is still necessary for coating and development procedures, which multiply with design complexity. For instance, the seemingly trivial incorporation of multiple thicknesses in a template entails additional coating steps, masks and alignment procedures.7,8 Direct printing of LOC, implemented by micro stereo lithography (MS4,5), has traditionally required expensive infrastructure, whereas the emergence of affordable 3D printers has the potential to grant broader access to custom Optical Devices Laboratory – Department of Physics, Chemistry and Biology (IFM), Linköping University, Linköping 58183, Sweden. E-mail: [email protected]; Tel: +46 13 28 1282
424 | Lab Chip, 2014, 14, 424–430
made microstructures, by minimizing resources and fabrication skills. Economical 3D printers have been used to fabricate millifluidic and microfluidic devices,10 as well as reactionware for chemical synthesis and analysis.11 In the first case, a thermoplastic extrusion9,10 printer was used, whereas in the second instance a robocasting platform12 that enabled printing with different materials, demonstrated the potential for direct printing of functional substances, and precursors, chemically decorating active reactionware.11 This type of printer connects to the vision of users access to physical products by downloading and locally printing a digital design.11,13 However, these printer technologies have inherent limitations in resolution and surface roughness, which are determinant to fabricate templates compatible with current PDMS fluidics. The recent introduction of affordable 3D printers based on MS aided by digital micro mirror device (DMD) projection,14 offering 50 μm resolution and smooth surface finish, open up new possibilities for highly simplified and versatile PDMS-LOC fabrication. Here we study the use of an affordable MS 3D printer (Miicraft,14 2299 US$) to fabricate templates for classic PDMS-LOC devices.1,15 The resolution limits are investigated and tested on practical structures, such as diffusion and chaotic mixers.16,17 The inherent advantages of 3D printed templates, such as natural fabrication of micrometric channels combined with mm high elements in the same template, are exploited to integrate connection tubing sealed to the PDMS replica, to improve mixer geometry, and to
This journal is © The Royal Society of Chemistry 2014
View Article Online
Lab on a Chip
configure open measurement chambers for straightforward chemical functionalization in multilevel devices, illustrated by time resolved glucose detection. This first demonstration of 3D printed reusable templates for PDMS-LOC, also highlights an attractive fabrication cost of 0.48 US$ per template.
Paper
coating on completely dried out templates does not allow the PDMS to cure.
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Experimental Template fabrication Templates were designed with free computer aided design (CAD) software for Mac OS X 10.8.3 (Autodesk® Inventor® Fusion, Autodesk Inc.) on a MacBook Air computer (13-inch, late 2010, 1.86 GHz Intel Core 2 Duo, 4 GB 1067 MHz). The CADs are exported as .stl mesh files that are converted by the software provided with the 3D printer (Miicraft® Suite). The Miicraft® Suite allows for alignment and scaling of the .stl files, segmentation of the aligned 3D mesh and printer control. Segmentation transforms the 3D mesh in 2D cross sections (bitmap black and white files in .png format at 480 × 768 pixels resolution) in 50 μm steps. The .png files are accessible and may be modified for refinements and minor corrections common to progressive optimization. These corrections do not require CAD skills, and were performed in Photoshop™ CS4, although simpler tools capable of single pixel editing, such as Microsoft Paint, are adequate. Miicraft®14 is an MS 3D printer (2299 US$) with 450 ppi (~56 μm) lateral, and 50 μm vertical resolution. The Miicraft® Suite controlling the printing process allows setting exposure time for each layer, printing speed (2 or 3 cm per hour) and vertical separation between exposed layers (50 and 100 μm); this final setting provides an easy way to duplicate the thickness of a design sliced at 50 μm per layer. After optimization, our templates were printed under the following conditions of exposure time, printing speed and vertical step size: Fig. 1: 12 s, 2 cm per hour, 50 μm; Fig. 2: 11 s, 3 cm per hour, 50 μm and 100 μm (for the 500 μm deep channels); Fig. 3: 10 s, 3 cm per hour, 50 μm. After printing, the templates were sonicated (FinnSonic m15) in industrial grade ethanol for 20 s, and air-dried. The dry templates were post cured in UV light for 600 s, in the post-curing chamber of the printer. The printer uses a proprietary resin (138 US$ per 500 ml), with different proportions of modified acrylate, modified acrylate oligomer, acrylate monomer, epoxy monomer, photo initiator and additives as principal components. The specific formulation is not disclosed. Since PDMS does not cure on the templates, the surfaces must be protected with a PDMS compatible material before casting the polymer. We developed a procedure consisting of 2 min sonication in industrial grade ethanol and subsequent ink airbrushing (Neo CN Iwata, gravity feed dual action airbrush with 0.35 mm nozzle). Effective surface protection requires the ink (Pentel NN60) to be applied in thin layers immediately after the template surface is taken from the ethanol and begins to dry. Ink
This journal is © The Royal Society of Chemistry 2014
Fig. 1 (a) CAD with 50 μm and 2 mm features combined in the same design. (b) 3D printout of the structure in (a). The inset highlights hindrances developed at the 50 μm to 100 μm channel step. (c) Image of the printed pillar guides used to integrate silicone tubing in the PDMS structure. (d) PDMS device assembled on a glass substrate showing no texture at the PDMS–glass interface. (e) Microscope image of tubing connection to a 500 μm channel. (f) Epi-fluorescence microscope image of PDMS structure filled with fluorescein solution.
Fig. 2 (a) Template for diffusion and chaotic mixers. The pillars are guides for silicone tubing integration and the external frame contains the PDMS during curing. (b) Detail of the chaotic mixer design. (c) Detail of the chaotic bitmap edited refinements. (d) Printed template. (e) 100 μm deep diffusion mixer showing the mixture of 1 mM fluorescein with 1 mM rhodamine B at 60 μl min−1. (f) 100 μm deep chaotic mixer imaged in the same conditions as in (e). (g) Idem to (f) operated at 30 μl min−1. (h) Idem to (g) but for a 500 μm deep channel. (i) Detail of the 500 μm deep chaotic mixer 3D structure.
Lab Chip, 2014, 14, 424–430 | 425
View Article Online
Paper
Lab on a Chip
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
rendered apparent by a semi-glossy finish. It takes some practice to paint templates consistently with an optimum finish, but if it is inadequate, the template can be rinsed in ethanol and airbrushed again. Templates with insufficient paint do not allow PDMS to cure and are wasted, whereas templates with an excess of paint, characterized by a matte finish, permit PDMS curing but transfer paint to the PDMS replicas. These stained PDMS replicas can be cleaned by a 2 min sonication in ethanol. Surface roughness was measured with a stylus profiler (Dektak 6, Veeco Instrument Inc.) along 1 mm tracks resolved in 6000 points. The average roughness was 182 nm for the finished templates and 540 nm after applying the protective coating (1.08% of the minimum specified thickness). Templates were completed with the manual insertion of silicone tubing (Esska.de GmbH, Hamburg) in the template guides. Designed guide diameters, required for a proper sealing of the tubing within the channels, were: 0.6 mm, 0.9 mm and 1.6 mm for tubing inner diameters of 0.5 mm, 0.8 mm and 1.5 mm respectively. The average weight of each template in this work was 1.8 g, which corresponded to a resin cost of 0.48 US$ per template. Carefully treated templates could be reused. Mechanical failure of delicate tubing guides, which broke off and detached with the cured PDMS, occurred after 4 re-uses in the best case. Templates with more robust features did not show mechanical failure, but were not explicitly tested to gauge the failure limit for the surface coating.
PDMS devices
Fig. 3 (a) Templates used for multilevel fluidics. The blue coloured template is symmetric along the ss′ direction and the PDMS replica from the red coloured template is flipped over for assembly with the blue replica and placed on the glass substrate. The cross-section highlights the use of different thicknesses in the template to configure a measurement chamber, with access to a functionalized substrate. (b) Filling of the multilevel structure with fluorescein solution and fluorescence microscopy detail of the inter-level connections in 1, 2, 3, 4. (c) Idem to (b) 130 ms elapsed after injection. (d) +460 ms. (e) +920 ms. (f) Detail of tubing connection with channel space. (g) Cross section of open measuring chamber connected to 100 μm deep channel. (h) Response to 2.5, 5 and 10 mM glucose concentrations and inset showing the colour change for 10 mM, with respect to the control (ctrl.).
1.5 ml of 2 : 1 ethanol–ink solution is necessary to cover a single template properly (up to 43 × 27 mm2), which is
426 | Lab Chip, 2014, 14, 424–430
To fabricate the PDMS devices, Dow Corning Sylgard 184 base18 and curing agent were mixed in 10 : 1 proportion and stirred in a cup for 2 min. The mixture was degassed in a desiccator connected to a rotary pump for 30 min, and subsequently poured on the coated templates. The PDMS filled templates were cured at 65 °C for 2 h in an environmental oven (Gallem Kamp incubator) and the PDMS replicas, with embedded tubing attached, were separated with tweezers and placed directly on clean glass slides (Menzel-Glaser, Braunschweig, Germany). Only in the case of chaotic mixers, which were tested up to 90 μl min−1 (continuous infusion/Dual NE-1000 syringe pump, from New Era Pump Systems Inc., http://www.syringepump.com), the PDMS and glass were oxygen plasma treated (Diener electronic, Pico model) during 1 min, for greater adherence.
Glucose sensing The assay for glucose sensing utilizes glucose oxidase (GOx), horseradish peroxidase (HRP) and a non-coloured substrate (Ampliflu Red, also known as Amplex Red). GOx catalyzes the oxidation of glucose to glucono delta-lactone and hydrogen peroxide, which reacts with HRP stoichiometrically 1 : 1 to catalyze de-N-acetylation and oxidation of colorless Ampliflu Red (10-acetyl-3,7-dihydroxyphenoxazine) to resorufin.19,20
This journal is © The Royal Society of Chemistry 2014
View Article Online
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Lab on a Chip
Resorufin is a pink-colored, red fluorescent compound, with an absorption maximum of 563 nm. Resorufin is also a substrate for HRP, which further oxidizes it to a colourless/ nonfluorescent compound.19,20 Response time can be widely regulated through the proportions of constituents. The assay in this instance contained: 10 units ml−1 of glucose oxidase enzyme activity, 10 units ml−1 of horseradish peroxidase activity, 250 μM Ampliflu Red and 0.3 M trehalose in phosphate buffered saline (PBS, pH 7.4). All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA). For assay characterization, 1 μl of assay reagent was dropped on top of flat PDMS, and 1 μL of glucose in PBS, pH 7.4 (0; 2.5; 5; 10 mM) was added while video imaged with a Galaxy Note 2 rear camera (see details in the next sub-section). Surface functionalization of the PDMS-LOC microfluidic chamber was accomplished depositing the assay reagent on a flat PDMS surface and allowing it to dry in darkness, at room temperature, for 2 hours. The chamber was sealed to the functional surface, and a 10 mM glucose solution, followed by an air bubble, was pumped into the device at 30 μl min−1 (Dual NE-1000 syringe pump). The pumping was stopped when the solution reached the measuring chamber, while the device was imaged at 30 fps with the rear camera of a Galaxy Note 2 smart phone.
Imaging Template imaging (Fig. 1b and c, Fig. 2d) was implemented with an Olympus SZ60 stereo zoom microscope, fitted with a Canon EOS 500D DSLR camera (15 mPixels, APS-C cmos sensor). Bright field and epi-fluorescence imaging was conducted on a Zeiss Axiovert 40 cfl inverted routine microscope. A Zeiss HBO50 illuminator, housing a mercury vapour short arc lamp HBO 50, provided fluorescence excitation. The Canon EOS 500D camera serviced both microscopes. A Samsung Galaxy Note 2, 8 Mpix rear side camera was used in photo mode for LOC imaging and for full HD video acquisition (30 fps), for which purpose it was positioned facing the LOC devices placed on a glass stage back illuminated by an iPod Touch (4th generation, 960 × 640 pixel screen resolution, with iOS 6) running Led Torch v1.37 app (www.smallte.ch), set to pure white (rgb 255, 255, 255). Glucose response, for three test concentrations and control, was video captured with the Samsung Galaxy Note 2 rear camera and converted to a .jpeg image sequence at 0.5 fps using MPEG Streamclip v 1.9.2. This sequence was processed with a dedicated script, in Matlab R2008b, developed for this purpose. A manual grid was used to select the 4 regions of interest (ROI) (control and 3 glucose concentrations) and extract average intensities in the red, green and blue channels representing the ROIs of the composition (Fig. 3h). The procedure was automatically repeated in all pictures extracted from the original video. The time response data was copied and pasted to Microsoft Excel 2011 for Mac, for final editing. Pictures in
This journal is © The Royal Society of Chemistry 2014
Paper
this communication were contrasted and composed with Apple Keynote 09, v5.3.
Results and discussion Micro stereo lithography 3D printers, like the Miicraft14 used in this work, utilize a DMD to illuminate a photocurable resin with precise 2D black and white patterns at a resolution of 450 ppi (~56 μm), and to displace successive exposures in the vertical dimension at 50 μm steps, within a maximum printing volume of 43 × 27 × 180 mm3. 3D printing allows the fabrication of microstructures with multiple thicknesses in a single printout, whereas conventional photolithography would require multiple masks and repeated coatings, development and alignment procedures.1,7,8,16 Computer aided designs (CADs) are converted by the printer software into 2D bitmap image slices (.png format), which remain available for retouches and finishing details, without requiring CAD re-editing, thus easing optimization cycles. Since retouching only requires bitmap editing skills, with readily available software, such as Microsoft Paint, the specialized skills for LOC design and CAD composition can be completely separated from its fabrication, and the 3D printer becomes a generic vehicle to transmit the design. Achievable resolution depends on multiple factors, such as proximity between elements and thickness differences. In addition to the layout, the exposure time for each layer, as well as printing speed, determine the final result. Our planar templates are 2 mm thick and demand printing times under 20 min, weighing on average 1.8 g, with a cost of 0.48 US$ per template, which compares favourably with conventional fabrication services. Furthermore, printed templates can be reused several times. The sole material requirements for our technique are: 3D template printing, protective coating to allow PDMS curing, PDMS pouring and assembly. Sealing to external tubing is simplified, because the template can easily incorporate guides for silicone tubing, which become integrated to the PDMS replica. The test structure used to assess the limit of resolution achievable between close elements, and between elements of different thicknesses is depicted in Fig. 1. The structure incorporates features ranging from 50 μm to 2 mm and also illustrates the assembly of connection tubing directly on the template. Fig. 1a shows the 3D CAD with three 50 × 50 μm2 crosssection channels at a close spacing of 50 μm. These channels continue in three different directions and join to 100 × 100 μm2 sections, which in turn connect to 200, 300, and 500 × 500 μm2 cross sections, terminated in cylindrical tubing guides. The zoomed area (Fig. 1b) details the connection between 50 μm and 100 μm deep channels. At this resolution the step between different channel depths induces a repeatable artifact. Although unintended by design, these features are useful to create hindrances beyond the printer resolution, which
Lab Chip, 2014, 14, 424–430 | 427
View Article Online
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Paper
can enable particle separation,21 or trap detection chemistries on beads.22 The insertion of silicone tubing in the printed template guide (Fig. 1c) is implemented with guiding pillars (1.6 mm O.D.), with 1.5 mm I.D. tubing. The small mismatch is absorbed compressively and provides the necessary seal between channel and tube, required to prevent PDMS diffusion and subsequent blockages. Thus, a continuous duct is configured, connecting the micro channels to the tubing lumen (Fig. 1e). Since PDMS does not cure on Miicraft proprietary resin,14 the template surface had to be coated with a PDMS compatible material. Consequently, the templates were airbrushed with permanent ink (Pentel NN60), after which degassed PDMS was poured and cured at 65 °C for 2 hours, before releasing the PDMS replica and coupled tubing. PDMS adheres to silicone tubing solidly, providing a robust interconnection to the micro fluidics. The microscope image of the PDMS device mounted on a glass slide (Fig. 1d), captures the texture transferred to the PDMS replica (540 nm average roughness), which was only noticeable on free surfaces, such as channel ceilings, whereas on the large area in contact with the glass surface, this texture vanished completely. To test the integrity of the structure, the device was filled with a fluorescein solution and imaged in an epi-fluorescence microscope (Fig. 1f). No leakages or blockages were observed, whereas hindrances created on the template were functional and could support printer resolution up to 10 μm, for localized features. To better assess potential practical uses of the 3D printed templates, we investigated reactors incorporating chaotic mixer designs16,17,23,24 and the viability of multi-level PDMS fluidics.8 The laminar character of the flow, which restricts the mixing process to slow molecular diffusion, complicates rapid and efficient mixing in microfluidic devices.17 Passive micromixers introduce geometric features along the micro channels,25 which break the laminar flow inducing chaotic advection to improve mixing efficiency.17,21,24 The ability to fabricate the geometries required to affect the flow regime, with printed templates, is a practical test for this method. Comparative results between a chaotic mixer and a diffusion reactor of the same dimensions are represented in Fig. 2. The printed template (Fig. 2a) incorporates both mixers in an external frame 1 mm high, which is used to confine the poured PDMS to the template, thus making the fabrication procedure cleaner and simpler. The bottom picture corresponds to the template after PDMS curing, when it is ready to be detached. Details of the mixer geometric obstacles, as specified by CAD, and the flow direction are indicated in Fig. 2b, whereas final adjustments to the design, made directly on the .png bitmap images, are shown in Fig. 2c. The resulting printout, for optimized exposure time and printing speed (11 s, 3 cm hour−1), is shown in Fig. 2d. To create a mixer, the texture in the template is advantageous, since the obstacles are not just smooth surfaces but an intricate 3D boundary (Fig. 2i) that contributes to the mixing efficiency.
428 | Lab Chip, 2014, 14, 424–430
Lab on a Chip
The behaviour of 1 mM of fluorescein and rhodamine B aqueous solutions, injected at 60 μl min−1 in the 100 μm deep reactor is shown in Fig. 2e. The solutions do not mix even at the distal end of the structure, which is underscored by a black arrow to indicate the last position where two colours can be distinguished clearly. In contrast, the chaotic design mixed both solutions close to the proximal side of the channel (Fig. 2f), and even at 30 μl min−1, it mixed the substances before the distal side (Fig. 2g). A crucial advantage of printed templates is that changing channel thickness does not imply additional cost or different materials and protocols. Fig. 2h captures the performance of a 500 μm deep chaotic mixer at 30 μl min−1. As expected, mixing efficiency improves, as the Reynolds number scales with channel thickness,17 but the thicker channel also introduces complex 3D topologies (Fig. 2i), hard to achieve in conventional methods, which contribute to turbulence. In addition to well-sealed channels, embedded connectors and practical microfluidic structures, 3D printed templates can be used to create microfluidic structures with more than one level. To connect between levels, through holes could easily be created using printed pillars rising above the PDMS surface; however, this configuration creates meniscuses and uneven PDMS surfaces, which preclude an efficient stacking of multiple levels. The alternative explored in this work entirely eliminates the need for through holes and the effect of meniscuses. The problem was solved using two PDMS replicas with channels, which overlap only at the desired interconnection locations. To create these connections, once the replicas were fabricated, one was flipped over and placed in contact with the second, facing the flat bottom sides. Thus, overlapping channels created interconnections, and the flat contact surfaces provided a robust sealing for the complete circuit. Flipping over one replica of the assembly requires designing one template as the mirror image of the desired result. To simplify design and avoid mirroring, one template has a generic configuration (blue template in Fig. 3a) and is symmetric with respect to the ss′ axis. Under these conditions, the custom layer (created with the red template in Fig. 3a) can be planned as an overlay of the blue template design, thus easily aligning connections between levels. Once the replicas are created, one of them can be flipped over along the ss′ axis without affecting relative alignment between layers. PDMS replicas were detached from the templates after 2 hours, at 65 °C. The custom replica (coloured red in Fig. 3a) was flipped over along the ss′ axis, aligned with the flat side of the generic replica, pressed together and returned to the oven at 65 °C for another 30 min before use. Resulting structures were perfectly sealed and ready to use. To test device integrity, a fluorescein solution was injected at 500 μl min−1 flow (Fig. 3b–e correspond to a 920 ms interval). It held tight with no need of plasma treatment for enhanced adhesion. The flow must traverse 10 inter level connections, to close the circuit, which proved simple to fabricate and reliable in
This journal is © The Royal Society of Chemistry 2014
View Article Online
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Lab on a Chip
operation. Connections were aligned visually, and fluorescence microscopy revealed well-defined channels and proper coupling between levels (Fig. 3 1–4). The continuous connection between tubing lumen and channel space, as well as the solid sealing between tubing and PDMS is apparent in Fig. 3f. Chemical functionalization is central to regular and multilevel fluidics, and can be implemented in various ways. Direct 3D printing of functional regions has been demonstrated in reactionware,11 using a different printing technology. MS 3D printers used in this work cannot change the printing materials easily, but both resolution and surface finish are better suited to make PDMS templates, and can easily combine thick and thin features in the same template. In our case, a thick feature was left above the PDMS level to configure a precisely defined open chamber that enabled communication of the two levels fluidic design with a functionalized surface, which could be prepared separately. The advantage is that surface chemistry could be independently controlled over an unconstrained area and then attached to the fluidics without special positioning demands (Fig. 3a and h). This concept was tested with time resolved glucose detection for concentrations within the diagnostic range. The response to three test concentrations of glucose (and control), within the clinically relevant range, are displayed in Fig. 3h. The sensing surface contains glucose oxidase (GOx), horseradish peroxidase (HRP) and a non-coloured substrate (Ampliflu Red). GOx catalyzes the oxidation of glucose to glucono delta-lactone and hydrogen peroxide, which is then used by HRP to catalyze de-N-acetylation and oxidation of colourless Ampliflu Red (10-acetyl-3,7-dihydroxyphenoxazine) to resorufin. Resorufin is a pink-coloured compound, which is also a substrate for HRP, which further oxidizes it to a colourless/nonfluorescent compound.19,20 The results in Fig. 3h illustrate the faster production of resorufin for higher glucose concentrations, due to higher levels of available hydrogen peroxide produced through GOx activity, and the subsequent further consumption of resorufin by HRP, that renders the indicator colourless. The dynamic characteristic of the response is suitable for quantitative detection independently of the colorimetric readout capabilities, which may vary substantially. In our case, colour is just used to track the minima, which are video acquired at 30 fps (a common feature of any modern cell phone) and can be followed with a resolution of 33 ms. For the response range of 300 s, in our experiment (Fig. 3h), 33 ms corresponds to 0.011% of the range, whereas 1 count in 255 levels of any colour channel used for colour quantification is 0.039% (about 35 times worse). In this work practical PDMS on glass LOC devices have been fabricated using minimal resources and procedure complexity. The capabilities conferred by affordable latest generation MS 3D printers, enable integration of multiple thicknesses in the same template at no extra cost or effort, while retaining all classical PDMS-LOC advantages. The technology offers the possibility to develop LOC devices at low cost, together with great flexibility and simplicity to
This journal is © The Royal Society of Chemistry 2014
Paper
iterate designs, thus supporting fast optimization cycles of multiple prototypes per day. The ability to create sealed connectors and to implement multilevel fluidic designs with robust interconnections renders this technology attractive for numerous development scenarios. Furthermore, the cost of reusable templates below 0.48 US$ per piece makes it suitable not only for development but also for limited production of specialized devices. 3D printed templates inherit the vision of affordable 3D printers delivering a physical version of a remotely conceived digital design, thus separating design skills from fabrication efforts, and greatly simplifying access to PDMS-LOC configurations.
Conclusions The fabrication of PDMS on glass LOC devices has been demonstrated using affordable MS 3D printers to produce templates. The proposed technique drastically simplifies device fabrication by eliminating the need for clean room facilities and repeated photolithographic steps required for templates with multiple thicknesses. Templates are reusable and can be produced in less than 20 min, at an average cost of 0.48 US$, remarkably accelerating development time, improving flexibility for optimization and cutting costs for multiple iterations. Because PDMS does not cure in contact with the commercial printer resin, a simple procedure to protect the template surface has been developed. Practical fluidic structures, with down to 10 μm hindrances, have been demonstrated, as well as the fabrication of micromixers and multilevel fluidics. 3D printed templates can easily incorporate elements of different thicknesses, which has been exploited in two explicit ways namely, the integration of sealed tubing and the generation of a measurement chamber in a multilevel structure, which can be independently functionalized for chemical sensing, in this case for glucose detection. Affordable last generation MS 3D printers offer the possibility to develop templates suitable for PDMS-LOC configurations of varied complexity at a fraction of cost, time and effort. The printer not only deals with a substantial part of the experimental tasks, which are thus simplified for the end user, but also as originally intended with 3D printer technologies,9,11–13 disentangles fabrication from design skills, rendering the printer a tool to transfer custom PDMS-LOC designs. The concept described here should enable broader access to established LOC solutions, and provide a versatile LOC development platform compatible with fast and affordable optimization.
Acknowledgements G. Comina’s research stay at Linköping University has been financed by the Peruvian National Council for Science and Technology (CONCYTEC). This research has been
Lab Chip, 2014, 14, 424–430 | 429
View Article Online
Paper
supported by grants from Linköping Centre for Life Science Technologies (LIST), Sweden, and from the Swedish Research Council (VR).
Published on 18 October 2013. Downloaded by University of Waterloo on 13/12/2013 15:31:28.
Notes and references 1 Microfluidics and Nanofluidics Handbook, Fabrication, Implementation, and Applications, ed. S. Mitra and S. Chakraborty, CRC Press Taylor & Francis Group, Boca Raton, 2012. 2 J. Love, D. Wolfe, H. Jacobs and G. Whitesides, Microscope Projection Photolithography for Rapid Prototyping of Masters with Micron-Scale Features for Use in Soft Lithography Langmuir, 2001, 17, 6005. 3 J. Behm, K. Lykke, M. Pellin and J. Hemminger, Projection photolithography utilizing a Schwarzschild microscope and self-assembled alkanethiol monolayers as simple photoresists Langmuir, 1996, 12, 2121. 4 C. Sun, N. Fang, D. Wu and X. Zhang, Projection microstereolithography using digital micro-mirror dynamic mask Sens. Actuators, A, 2005, 121, 113. 5 X. Zhang, X. Jiang and C. Sun, Micro-stereolithography of polymeric and ceramic microstructures Sens. Actuators, A, 1999, 77, 149. 6 P. Preechaburana and D. Filippini, Fabrication of monolithic 3D micro-systems Lab Chip, 2011, 11, 288. 7 M. Peterman, P. Huie, D. Bloom and H. Fishman, Building thick photoresist structures from the bottom up J. Micromech. Microeng., 2003, 13, 380. 8 J. Anderson, D. Chiu, R. Jackman, O. Cherniavskaya, J. McDonald, H. Wu, S. Whitesides and G. Whitesides, Fabrication of topologically complex three-dimensional microfluidic systems in PDMS by rapid prototyping Anal. Chem., 2000, 72, 3158. 9 B. Evans, Practical 3D printers: The science and art of 3D printing, Apress, 2012. 10 P. Kitson, M. Rosnes, V. Sans, V. Dragone and L. Cronin, Configurable 3D-printed millifluidic and microfluidic ‘lab on a chip’ reactionware devices Lab Chip, 2012, 12, 3267. 11 M. D. Symes, P. J. Kitson, J. Yan, C. J. Richmond, G. J. T. Cooper, R. W. Bowman, T. Vilbrandt and L. Cronin, Integrated 3D-printed reactionware for chemical synthesis and analysis Nat. Chem., 2012, 4, 349. 12 E. Malone and H. Lipson, Fab@Home: the personal desktop fabricator kit Rapid Prototyping J., 2007, 13, 245.
430 | Lab Chip, 2014, 14, 424–430
Lab on a Chip
13 T. Vilbrandt, A. Pasko and C. Vilbrandt, Fabricating nature Technoetic Arts, 2009, 7, 165. 14 Miicraft® (no date), Miicraft 3D printer, available at: http:// www.miicraft.com/products/ (accessed: 13 August 2013). 15 A. Foudeh, T. Didar, T. Veresa and M. Tabrizian, Microfluidic designs and techniques using lab-on-a-chip devices for pathogen detection for point-of-care diagnostics Lab Chip, 2012, 12, 3249. 16 Y. Suh and S. Kang, A Review on Mixing in Microfluidics Micromachines, 2010, 1, 82. 17 N. Nguyen and Z. Wu, Micromixers – a review J. Micromech. Microeng., 2005, 15, R1–R16. 18 J. Lötters, W. Olthuis, P. Veltink and P. Bergveld, The mechanical properties of the rubber elastic polymer polydimethylsiloxane for sensor applications J. Micromech. Microeng., 1997, 7, 145. 19 V. Towne, M. Will, B. Oswald and Q. Zhao, Complexities in horseradish peroxidase-catalyzed oxidation of dihydroxyphenoxazine derivatives: appropriate ranges for pH values and hydrogen peroxide concentrations in quantitative analysis Anal. Biochem., 2004, 334, 290. 20 M. Zhou, Z. Diwu, N. Panchuk-Voloshina and R. P. Haugland, A stable nonfluorescent derivative of resorufin for the fluorometric determination of trace hydrogen peroxide: applications in detecting the activity of phagocyte NADPH oxidase and other oxidases Anal. Biochem., 1997, 253, 162. 21 I. Dimov, L. Basabe-Desmonts, J. Garcia-Cordero, B. Ross, A. Ricco and L. Lee, Stand-alone self-poweredintegrated microfluidic blood analysis system (SIMBAS) Lab Chip, 2011, 11, 845. 22 N. Lee, Y. Yang, Y. Kim and S. Park, Microfluidic immunoassay platform using antibody immobilized glass beads and its application for the detection of Escherichia coli O157:H7 Bull. Korean Chem. Soc., 2006, 27, 479. 23 S. Kwon, S. Kou, H. Kim, X. Chen, H. Hwang, S. Nam, So Kim, K. Swamy, S. Park and J. Yoon, Sensing cyanide ion via fluorescent change and its application to the microfluidic system Tetrahedron Lett., 2008, 49, 4102. 24 D. Therriault, S. White and J. Lewis, Chaotic mixing in threedimensional microvascular networks fabricated by directwrite assembly Nat. Mater., 2003, 2, 265. 25 V. Mengeaud, J. Josserand and H. Girault, Mixing Processes in a Zigzag Microchannel: Finite Element Simulations and Optical Study Anal. Chem., 2002, 74, 4279.
This journal is © The Royal Society of Chemistry 2014
