Inexpensive and rapid fabrication of PDMS microfluidic devices for biological testing applications using low cost commercially available 3D printers

All microfluidic devices used in this study were designed using the open-source 3D modelling software FreeCAD (www.freecadweb.org/, version 0.19.0) Designs of all microfluidic devices can be found in the supplementary files. All stereolithography (.STL) design files generated in FreeCAD were then loaded onto the 3D printers for device fabrication.

Master-moulds to cast PDMS microfluidic chips were fabricated using one of the following commercial SLA 3D printers: Anycubic Photon D2 DLP (Anycubic, Guangdong, China) or Form 3B Low Force Stereolithography^TM (Formlabs Inc. PKG-F3B-WSVC-MSP-BASIC, Somerville, MA, USA).

For 3D printed moulds using the Form 3B, models were printed with clear resin (Clear V4 Resin, Formlabs Inc., RS-F2-GPCL-04, MA, USA) at a minimum resolution of 25 µm, following the manufacturer recommended print settings using PreForm software version 3.28.0 (Formlabs Inc., Somerville, MA, USA). For 3D printed counter-moulds using the Anycubic Photon D2 DLP, models were printed with black resin (Anycubic, SKU TL4425, Guangdong, China) at a minimum resolution of 51 µm, following the manufacturer recommended print settings using Chitubox software v.1.9.4 (Chitubox Inc, Guangdong, China).

2.3.1. UV and heat treatment of mould.

2.3.2. Sanding and nail polish (NP) coating of mould.

Polydimethylsiloxane (PDMS) (Sylgard 184, Dow Corning, Copley, OH, USA) base and curing agent were mixed in a 10:1 ratio and casted into post-processed 3D printed moulds obtained from both Form 3B and Anycubic printers and allowed to cure at room temperature in the fume hood for 12 h to allow for bubbles to dissipate. The moulds containing PDMS were then incubated at 75 °C for 1.5 h. Finally, they were removed from the mould using a metal scalpel. Inlet and outlet holes were generated in the resultant PDMS device using a hole puncher (UniCore Manual Punchers, QIAGEN WB10028, Germantown, MD, USA). Finally, the PDMS microfluidic device was adhered to a glass microscope slide (VWR 100491–372, Radnor, PA, USA) using medical-grade double-sided tape (ARcare 90106NB, Little Rock, AR, USA). For experiments measuring the bond between PDMS-glass surfaces, PDMS devices and glass microscope slides were bonded using oxygen plasma. The adhered device was baked at 95 °C for 10 min.

Post-processing of the final PDMS devices was performed using a coating solution containing polydopamine (PDA) to increase surface hydrophilicity. Coating solutions were prepared by dissolving dopamine hydrochloride (Sigma-Aldrich H8502, St. Louis, MO, USA) in phosphate-buffered saline pH 8.5. PDMS of dimensions 20 × 20 × 5 mm (length × width × height) were dipped and submerged in 0.1% or 0.01% PDA coating solution for 2, 8, and 24 h with gentle shaking at room temperature. Finally, they were washed with distilled water and air dried in a fume hood.

Following complete curing, the PDMS microfluidic devices were imaged using brightfield microscopy (Zeiss Axiovert 25 CFL, Carl Zeiss Inc., Oberkochen, Baden-Württemberg, Germany) to visualize the device features (microfluidic channels and wells) and characterize the device for potential defects. Images were acquired and processed using AxioVision (Carl Zeiss Inc.) and ImageJ computer software (https://imagej.nih.gov/ij/download.html, National Institutes of Health (NIH), Bethesda, MD, USA).

The optical transmittance of PDMS microfluidic devices casted and released from 3D printed moulds was measured in the visible spectrum between 400 and 700 nm using the Imaging Multimode Plate Reader (Cytation 5, Agilent BioTek Instruments, Winooski, VT, USA). Small PDMS circular cut-outs of approximately 22 mm in diameter with 5 mm thickness were placed in 6-well tissue culture plates (Montreal Biotech Inc. MBI703001C, Dorval, QC, Canada) then scanned in the plate reader. As a positive control, PDMS was casted and cured in a 12-well plate.

Metal 20-gauge blunt tip needles (HaBeuniver HB13510ML921DZT, Fuzhou, FuJian, China) were connected to FEP tubing (Masterflex^® Transfer Tubing, VWR MFLX06406-60, Radnor, PA, USA) and then connected to the PDMS microfluidic devices at inlet and outlet positions. We created a PDMS device as a 'test' case that contains three 5 cm long rectangular-shaped channels (1 mm wide and 0.5 mm deep) in parallel (CAD file available upon request) to perform fluid flow experiments. For all experimentation with fluid flow, a commercial syringe pump (PHD ULTRA, Harvard Apparatus, Holliston, MA, USA) was used to infuse the microfluidic chip with distilled water. Food colouring dyes (red, blue, and yellow) were added to the distilled water to create coloured solutions for easier visualization of fluid flow through the entirely sealed microfluidic system. The syringe pump was programmed to generate flow rates ranging from 0.5 µl min⁻¹ to 100 µl min⁻¹, for a period of 24 h.

High shear stresses can negatively impact cellular proliferation and viability. Low and non-lethal shear stresses must be achieved in order to use microfluidic devices for cell-based assays. To determine the theoretical minimum and maximum shear stresses in our system, the flow rates within the microfluidic device was converted to shear stress values using an online tool provided by Darwin Microfluidics, Microfluidic Flow Rate and Shear Stress Calculator (https://darwin-microfluidics.com/blogs/tools/microfluidic-flowrate-and-shear-stress-calculator, Darwin Microfluidics, Paris, Île-de-France, France).

PDMS is hydrophobic naturally. To evaluate the uniformity of the fabricated PDMS chip surface, the contact angle of five positions of each PDMS device were measured using an Optical Contact Analyzer (DataPhysics Instruments GmbH, Filderstadt, Germany) to determine the contact angle, using a sessile drop technique. In this method, 5 µl of high-performance liquid chromatography grade water was deposited on the PDMS surface at a rate of 2 µl s⁻¹ and imaged once a droplet settled and formed. The WCAs of the resulting images were analysed using SCA 20 software (version 2.04, Build 4). For each replicate, a minimum of five separate droplets were quantified across the surface of the PDMS device, and reported values are an average of these independent measurements.

PDMS-glass bonded devices were anchored to a spring balance (Ajax Scientific ME495-2000, Toronto, Canada) and a force was applied to the PDMS device, separating the PDMS from the glass slide. The force (in Newtons) at which the PDMS peeled from the glass surface was recorded.

We set out to utilize commercial 3D printing, an emerging technique used in the milli- and micro-fluidic field, as a tool to design PDMS microfluidic devices and characterize critical processing steps and printing considerations that result in optically clear PDMS microfluidic chips. Positive master moulds were printed using FormLabs 3B+ and AnyCubic Photon D2 and were subjected to various post-processing steps to generate a smooth 3D printed part (figure 1(a)). These processing steps involved high heat treatment, physical sanding, and applying a chemical coat. The resulting moulds were used to cure and create optically clear PDMS (figure 1(b)). In our test case involving three separate channels (dimensions: 100 µm height × 1 mm width × 3 cm length; volume of 3 µl), we were able to flow dyed solutions with flow rates ranging from 0.5 µl min⁻¹ to 100 µl min⁻¹ (figure 1(b)). This corresponds to a speed range of 0.083 mm s⁻¹ to 16.7 mm s⁻¹ and a shear stresses range of 4.5–891 Pa. Therefore, despite opting to use an adhesive rather than the conventional and stronger plasma bonding approach for sealing the device, our system can withstand high pressures. Furthermore, by selecting the proper media flow rates, we can achieve low shear stresses that have been demonstrated to be non-lethal to many mammalian cell lines [27]. Nonetheless, it is likely this can be improved through improved channel dimensions.

Zoom In Zoom Out Reset image size

Next, we sought to determine the accuracy of the commercial SLA printers in printing the CAD-designed moulds, and how well this translated to the final PDMS device. SEM imaging of positive moulds and PDMS devices generated by FormLabs and AnyCubic printers confirmed that the designed channel features could be resolved (figures 1(c) and (d)). Notably, both commercial printers could create channel heights as low as 100 µm, with the FormLabs capable of going as low as 50 µm (figure 1(c)). However, at <100 µm the channel features were less resolved, and both the 3D print and resultant PDMS device did not form sharp, distinct walls and instead were slanted (figures 1(c) and (d)). To quantify this, we measured the angle between the channel wall and a theoretical plane perpendicular to the printed surface, which would otherwise be a perfect wall if the printer replicated the CAD design perfectly (figure 1(e)). Indeed, we found that only the designed channel heights >200 µm resulted in more accurately printed channel walls (<10 degrees deviation) (figure 1(e)). In the XY plane, we found that the AnyCubic printer was able to accurately fabricate the desired channel widths. Surprisingly, the FormLabs printer consistently created channel widths that were ∼100 µm greater than the CAD design (figure 1(f)), a factor that must be considered in 3D design. Nevertheless, with proper 3D design and an understanding of the limitations of each printing method, we showed that commercial 3D printing can certainly be used in the design and fabrication of PDMS microfluidic devices.

We observed a significant amount of PDMS curing inhibition at the interface between the PDMS surface and the 3D printed surface, which has been previously reported in applications involving 3D printed moulds (figure 2). The resultant PDMS did not fully cure, leaving behind a sticky texture with many channel features failing to form, or being formed at very low resolution. To relieve this inhibition, we treated our printed moulds with high heat at 100 °C for up to four hours. This resulted in a completely cured and translucent PDMS surface, through manual observation and brightfield microscopy (figure 2). Notably, the surface of the device was smooth, and the device features were properly formed. We found that longer heat treatments resulted in increased PDMS curation, as evident by an improved PDMS surface texture and perfect channel edges. Importantly, at extremely high temperatures (∼120 °C) or prolonged heating (>8 h), the 3D printed moulds warped and/or cracked, decreasing the likelihood of generating PDMS microfluidic devices with accurate channel dimensions (data not shown). Overall, 3D prints acquired from SLA printers require high heat treatment to allow complete PDMS curing.

Zoom In Zoom Out Reset image size

Despite the successful fabrication of a PDMS microfluidic device, the resultant PDMS devices were optically translucent, likely due to the surface roughness of the 3D printed moulds. The PDMS surface was textured and not clear microscopically (figure 2). Furthermore, when visualizing the surface of the PDMS devices obtained using these moulds by brightfield microscopy at higher magnification (63X magnification), it was immediately clear that the surface was uneven and rough (figure 3(a)). The surface contained various imperfections, such as streaks and ridges. Therefore, we sought to design an inexpensive and simple post-processing method that would dramatically reduce the surface roughness of our 3D prints, ideally translating to the PDMS device. Ultimately, our goal was to reduce surface roughness and imperfections on the PDMS device itself, as this is a critical requirement for optical clarity. We opted to sand the features of the 3D moulds with increasing grit sandpaper, from 1500 to 15 000 grit, to remove large debris and imperfections in the 3D print, which resulted in a visually smoother surface but left streaks and scratches (figure 3(a)). Therefore, with sanding alone, the resulting PDMS devices were translucent and conformed to the scratches visible on the moulds. Polishing and applying chemical-based coatings, such as spray painting and lacquer coatings, as a final surface finish have been described [28]. We explored several chemical coating methods, including silicon-based sprays (mould release spray), polyvinyl alcohol, and polyurethane, all of which inhibited PDMS curation. However, we also assessed the ability of NP, an extremely common coating agent that is readily available, to smoothen the surface of the moulds. We noted that heat-treated and NP coated 3D printed moulds can reduce the surface roughness and significantly reduce the visible surface roughness when combined with the initial sanding of the 3D print (figure 3(a)). Both post-processing steps reduced the surface roughness of the 3D print by almost two-fold (Rq from 119.16 to 63.23; figure 3(b)) and were suitable to cure PDMS.

Zoom In Zoom Out Reset image size

Next, we used atomic force microscopy (AFM) to obtain higher resolution images, and quantified the root mean square roughness (Rq) of the 3D printed moulds and resulting PDMS surfaces. Interestingly, PDMS devices fabricated from moulds that had not undergone post-processing exhibited lower surface roughness than its mould counterpart (FormLabs: Rq from ∼119 to 99 nm; AnyCubic: Rq from ∼259 to 171 nm). However, with post-processed moulds, there was an extreme reduction in surface roughness of the resulting PDMS device in all cases when compared to the non-processed PDMS device (FormLabs: Rq from ∼63 to 18 nm; AnyCubic: PDMS device with Rq ∼ 32.42 nm). We note that in all technical replicates of the AnyCubic 3D printed mould, we could not get reliable estimates of the surface roughness. Nevertheless, the resulting PDMS surface obtained from the moulds from both commercial printers was >5-fold smoother than PDMS obtained from an untreated mould.

The resultant PDMS surfaces obtained through post-processing exhibited greater clarity, visibly, compared to the non-processed counterpart (figure 3(c)). This was further validated when we assessed the optical transmittance (a measure of clarity) of various post-processed PDMS samples, within the 400–700 nm range, compared to a control PDMS surface cured in a glass mould. We found that PDMS surfaces originating from moulds that were post-processed with both mechanical sanding and NP coating resulted in the highest level of transmission (78%–83%) across scanned wavelengths compared to the non-processed moulds and approached the transmission of the control PDMS surface, which exhibited transmission levels of 85%–89% (figure 3(c)). Notably, the clarity of PDMS surfaces from moulds treated with only a NP coating was still quite high (75%–81%). This suggests that this coating is the main factor for PDMS surface smoothness and clarity, with mechanical sanding responsible for increasing clarity by ∼2%–3%, likely due to the removal of larger surface debris and imperfections. It should be noted that we observed that physical sanding can cause up to 50% removal of resin material that form our microfluidic features, which is approximately ∼40 µm change in height in our devices (figure 3(d)). Additionally, the application of NP was observed to add ∼11 µm of height in our devices (figure 3(d)). These two factors must be a consideration when performing these post-processing steps. Future work will expand on these methods to limit the amount of 3D printed material that is lost from sanding and optimizing the NP coating to minimize the additional material added to our master mould. Nevertheless, our data outline the importance of the post-processing of 3D printing moulds to generate optically clear PDMS devices.

One other potential issue with rough PDMS surfaces is its ability to be bonded to glass surfaces using oxygen plasma. A smoother surface could facilitate greater PDMS-glass bonding. To investigate this, we oxygen plasma bonded PDMS (1 cm × 1 cm × 1 cm cut outs) from moulds produced using the FormLabs printer, with and without NP treatment (figure 3(e)). Surprisingly, we find that our PDMS substrates, regardless of NP treatment, is capable of bonding to glass tightly. When a tensile stress was applied, commonly, the PDMS substrate would break internally before its bond with glass was broken (figure 3(f)). The tensile force required to cause the PDMS or PDMS-glass bond was on average 13.20 and 12.12 Newtons for substrates generated from untreated and NP-treated moulds, respectively (figure 3(g)). This is equates to 0.1320 and 0.1212 Megapascals, which is at least within an order of magnitude and corroborates other reported values [29, 30]. Therefore, PDMS surfaces derived from FormLabs-based master moulds can be oxygen plasma bonded, though a smoother surface may facilitate more even bonding through the microfluidic device.

Despite being highly biocompatible, PDMS is also highly hydrophobic, which can pose significant complications for microfluidic devices intended for cell-based applications. Therefore, we applied a chemical surface modification (previously reported to increase surface hydrophilicity) using PDA [31, 32]. PDMS substrates were coated with various concentrations of PDA for 2, 8, or 24 h, and the WCA was measured as a proxy for surface hydrophilicity. PDA-coated devices coated exhibited significantly reduced WCA compared to non-coated PDMS surfaces (figure 4). Notably, this effect was concentration dependent as the surfaces coated with 0.1% PDA had a smaller WCA than the surfaces coated with 0.01% PDA. Therefore, the resultant PDMS microfluidic devices fabricated using our method can be modified similarly to those reported in the literature to produce hydrophilic surfaces that are suitable for cell culture applications.

Zoom In Zoom Out Reset image size

3D printing technologies have made great strides in biomedical and biomolecular research [5, 6, 17, 38, 39]. Though 3D printing has been used in the fabrication of PDMS microfluidic devices, its use remains limited. We aimed to provide a detailed guideline for the fabrication of PDMS devices using 3D printing technology. Our focus was to construct a simple and versatile method to generate PDMS microfluidic devices, from start to finish, with a product suitable for mammalian cell applications. We pushed the limits of the available FormLabs and AnyCubic 3D printers and showed that we could generate channel features in the <100 µm range, though larger feature sizes result in a more accurate final product. We were able to generate devices with small features (>25 µm channel height) that were still capable of fluid flow. This is, to the best of our knowledge, the lowest reported PDMS structure generated with 3D printing. Future developments in the field of 3D printing will maximize the resolution limit while maintaining the accuracy of the final printed part.

The potential applications for PDMS microfluidics are vast, spanning from diagnostic systems to sample testing and drug discovery [3, 8, 40, 41]. To date, 3D printing has not yet produced PDMS microfluidic devices with the same level of resolution and clarity when compared to devices obtained from more traditional methods. This is largely due to the limitations of the printer, PDMS curing inhibition that may result from different types of photocurable commercial resins and the smoothness of the resulting moulds that are generated by 3D printing. Our described post-processing steps can be applied in a wide range of laboratory settings, producing clear PDMS devices that approach those created via photolithographic techniques.

The first critical component of our post-processing step is the heat treatment of the 3D print. Contingent on the type of resin used, most photocurable resins are comprised of acrylate monomers, crosslinkers, colouring agents/photo absorbers, plasticizers, and most importantly, the photo initiator. Phosphine-oxide-based photo initiator fragments can leach from the moulds and inhibit the PDMS catalyst [42, 43]. In contrast, unreacted monomers strongly bind and attach PDMS to the 3D-printed surface. Both photo initiator and unreacted monomers can remain even after the 3D printing process, and their combined actions result in PDMS curing inhibition. UV treatment maximizes the polymerization of the resin and high-temperature treatment of the moulds allows any present photo initiator to be vapourized, both processes reducing the presence of residual photo initiator and resin monomer, enabling proper PDMS curing [44]. In line with these studies, we found that high heat treatment enables complete and proper PDMS curation. Notably, moulds are susceptible to cracking and breaking at prolonged levels of dry heating at >100 °C, so we propose the use of shorter heat treatments or high heat treatment with humidity (such as the autoclave), which does not visibly affect the structural integrity of the mould.

The second important post-processing step involves physical and chemical surface modifications to the 3D print to obtain a smooth surface. Several spray paint or lacquer coatings have previously been used to conceal the layer markings of 3D printed objects [28]. Similarly, chemical coats with SU-8 and fluorosilanes have been used to improve PDMS curing and clarity [14]. Furthermore, Parylene-C coatings have also been used to aid in PDMS release in addition to protecting master moulds, extending the number of cycles of moulding with various polymers [45, 46]. However, these methods require specialized pre-treatments, equipment, and/or deposition methods. We provide an alternative, low-cost coating solution involving NP, which significantly increases the surface smoothness of the 3D printed moulds, resulting in extremely clear PDMS surfaces (Rq < 33 nm). In comparison, the surface roughness of Pyrex glass has been estimated to be Rq = ∼1 nm [47]. Although our fabrication method does not produce PDMS surfaces rivalling that of pure glass substrates, we outline a novel post-processing method that approaches the surface smoothness of glass and outperforms other reported methods that use 3D printing to generate PDMS devices [48, 49].

The final part of our post-processing outline involves the application of a coating to the PDMS surface to increase hydrophilicity. PDMS is a highly hydrophobic substrate. This is particularly problematic for cell biological applications since adherent mammalian cells will not adhere to a hydrophobic surface [1, 10, 50, 51]. Additionally, air bubbles can readily form within a hydrophobic microfluidic device as hydrophilic media or solutions are flowed through the device [52]. Therefore, increasing the hydrophilicity of the PDMS surface allows for biomedical applications and promotes proper operation of the microfluidic device.