Fabricating microfluidic valve master molds in SU-8 photoresist

We fabricated the PR µvalve molds on glass using three different lithographic methods. Figure 1 presents major fabrication steps. We cleaned glass substrates (50 × 50 × 3.2 mm³, McMaster Carr) to improve SU-8 adhesion by gently etching them in NH₄OH/H₂O₂/H₂O (1:1:1) at 75 °C for 1 h to remove any organic residues, rinsing them with DI-water and methanol, and drying them with N₂. We followed the manufacturer's instructions to make layers 3, 15 or 25 µm thick in SU-8 (MicroChem Corp.) and 25 µm thick in SPR-220 (Shipley Company). To make PR-films, we spin coated, soft-baked, UV-irradiated and post-baked the PR layer. To make PR-features we imprinted a 2D pattern onto the PR layer by exposing it to UV through a photo-mask. We removed any excess PR with solvents after post-baking, rinsed solvents from the master, and then dried with N₂ gas. Photo-masks were drawn in AutoCAD 3D 2013 (Autodesk, Inc.) and then printed at 40 640 dpi resolution on high neutral density films (Photoplot Store). To improve micro-features adhesion to the glass substrate we covered the glass substrate with a 15 µm SU-8 base layer.

Zoom In Zoom Out Reset image size

The positive–negative PR method (figure 1(a)) made SPR-220 µvalve ridges on top of a base layer and protected them from SU-8 solvents with a 50 nm gold film (K675XD, Quorum Emitech) followed by a 3 µm polymerized SU-8 film.

The TFW method (figure 1(b)) exploited the capillarity of SU-8 flowing over rectangular cross-sectional features. As the film is spin coated, its adherence to both vertical and horizontal surfaces rounds all features. The radius of curvature decreases with increasing spinning speed and decreasing in viscosity [21]. To make such rounded µvalve ridges we covered a square-profile SU-8 feature with a 15 µm SU-8 2010 blanket spun at 1500 rpm, soft-baked at 115 °C for 5 min, exposed to 365 nm light at 140 mJ cm⁻² and post-baked at 95 °C for 4 min (figure 1(b)). To enhance optical contrast and enable alignment of subsequent PR features, we added 3 µm Al₂O₃ microspheres to the SU-8 (20 mg:1 mL).

The FPL method (figure 1(c)) followed the principles of a fountain pen; a microinjection needle acted as the nib to draw SU-8 from a reservoir and to deposit it onto the master. The setup consisted of four parts: a dissection microscope (Z45-L, Cambridge Instruments), a 4D micromanipulator (5 µm XYZ resolution, 360° YZ motion, MX10-R, Siskiyou Corporation), a custom syringe holder, and a Hamilton syringe fitted with a microinjection needle. The syringe holder was machined in Delrin® Acetal Resin and connected a micrometer head (2.5 µm/div, 150-832, Mitutoyo) to the piston of a blunt-needle 10 µL Hamilton syringe, allowing us to dispense as little as 0.4 nL/div (disregarding capillary forces and gravity), comparable to the volume of a µvalve.

To make microinjection needles, we thermoformed borosilicate capillary tubes (1 mm OD, 0.5 mm ID) that closely fit the syringe needle OD and clipped its end to make a 10–15 µm diameter port. A micropipette puller (P-87, Sutter Instrument Co.) controlled the taper length and port diameter.

We loaded the syringe with SU-8 and inserted its needle into the microinjection needle. A droplet of epoxy resin bonded the two needles together. After the glue hardened for 30 min, we pushed the SU-8 into the microinjection needle. The air bubble trapped between the walls of the two needles acted as a pressure cushion allowing continuous dispensing of SU-8. We tested three PR formulations: SU-8 2002, 2010 and 2015. We achieved the best results with SU-8 2010. Volatiles in the SU-8 made the droplet form a skin that prevented continuous flow. To correct this issue we kept the microinjection needle in a droplet of cyclopentanone (Sigma-Aldrich Co.) when not writing. We also discarded the first droplets of PR dispensed from the needle before writing.

We manually moved the microinjection needle between the ends of the channel molds that needed to be connected by a µvalve. The volume dispensed was controlled by calibrating the micrometer head to the volume that it forced out. The ridge size depended strongly on the SU-8 formulation and the volume dispensed. More dilute SU-8, such as SU-8 2002, had a larger decrease in volume after soft-baking. SU-8 2002 reliably produced ridges less than 10 µm tall. Valve molds with the wrong shape or size were removed with acetone and redrawn after soft-baking and prior to polymerization. The typical height of FPL ridges varied by ±5 µm for similar volumes of dispensed SU-8, microinjection needle diameters, and SU-8 formulations. For applications requiring more precise valve dimensions, the dispensing process could be automated. The final µvalve mold was soft-baked at 100 °C for 2 min, exposed to 155 mJ cm⁻² at 365 nm (Blak-Ray B-100SP UV Lamp, UVP LLC.), and post-baked at 100 °C for 2 min. Each µvalve mold was fully polymerized before the next one was fabricated.

We used MSL in PDMS (Sylgard 184, Dow Corning Corp.) to fabricate thin membranes by superposing a flow channel and a control channel. As seen in figure 1 of Melin et al [2], we used two layers of PDMS to create a push-down microfluidic valve. PDMS was cast at 1:10 ratio (curing agent/base elastomer) for all layers. We poured a 5 mm thick control layer over the control master and baked it for 1 h at 100 °C. Holes were punched at inlets for external connections to operate the µvalves. To make a 40 µm flow layer membrane, PDMS was spin coated over the flow master at 1270 rpm and baked for 10 min at 100 °C. Irreversible bonding of PDMS layers and glass relied on surface oxidation to activate the surface of the cross-linked PDMS [22]. Both PDMS layers were oxidized for 30 s in air plasma (400 mTorr) and we used a custom made XY-stage for quick and fine alignment to bond the layers together with the control layer properly superimposed over the valve features so that a pressurized control line would deform and seal the valve. We baked the assembly for 1 h at 85 °C to promote further PDMS bonding and then punched flow layer inlets. Cleaned microscope slides (see above) were plasma treated for 5 min to remove surface hydrocarbons. Then a PDMS assembly and the slide were plasma oxidized for another 45 s and brought into contact binding them to each other irreversibly, thus making a MSL chip.

We measured the PR feature heights and recorded surface profiles with a 20 µg force stylus profiler (Dektak 6M, Veeco Instruments Inc.). Profiler data was exported to Matlab to plot and measure PR-feature dimensions. SU-8 masters were imaged with a scanning electron microscope (SEM, Quanta 600F, FEI Company) and the chip in bright-field with a CMOS camera (1.3 MP B/W, MCE-B013-U, Mightex) connected to an inverted microscope (IMT2, Olympus). To test valve operation we first filled the control channels with degased H₂O and used 5–20 psi air pressure to close the µvalve. We then visualized µvalve operation with fluorescein (20 mM in PBS) flowing through the channels under 488 nm light from a 100 W mercury–xenon arc lamp and by tracking flowing 15 µm polystyrene spheres. We used 1–4 psi hydrostatic pressure to maintain liquid flow.

TFW and FPL successfully created robust masters with µvalve ridges adjacent to other SU-8 micro-features. We tested the µvalves in MSL PDMS devices made with one flow layer which included a µvalve, a 25 µm T-channel and a 3 µm weir filter (figure 3), and one control layer which allowed us to actuate the PDMS membrane to open/close the µvalve. Figure 3 shows how the masters' components of figure 2 relate to a working PDMS chip (figure 4).

Zoom In Zoom Out Reset image size

The two methods produced distinct rounding patterns. The TFW µvalves had wide, shallow-gradient wings (figures 2(b), (e), (h)), which worked perfectly with filtered reagents or live cells, but not with hard particles (e.g. 15 µm polystyrene spheres), since particles were trapped under the wings during operation. PR surface tension forces rounded the FPL µvalves. They have near-parabolic profile with dimensions similar to those of the flow channels. Figure 2 shows the large differences in aspect ratios (height-to-width) and the SEM images the overall shape. TFW µvalve aspect ratios varied from sample-to-sample from 1:20 (figure 2(c)) to 1:40 (figure 2(b)) and were dependent on the pressure applied to PDMS when closing the chip with a slide. FPL µvalves had more consistent aspect ratios, ≈1:7 and operational µvalves as narrow as 15 µm were easily drawn (data not shown).

As a test, we constructed a channel crossing with one valved input and three outputs, where the middle output includes a weir-filter. Contrast analysis of epi-fluorescence images showed that closed µvalves completely stopped the fluid flow (dark band across the input channel in figures 4(a) and (b)). Valve closing pressure depended on the dimensions of the valve and the fluid pressure in the flow line. Higher flow line pressures required higher control line pressures to completely seal the valve. 15 µm bead solutions were used to show that the valves could properly stop beads with a quick response time and that the weir filter would allow fluid flow but not beads (figure 4). During later experiments, both types of µvalves closed without PDMS layer delamination and required actuating pressures similar to those chips cast with reflowed SPR-220 masters. [25] TFW and FPL worked on opaque substrates unlike Futai et al [14] and did not require any additional micromolding steps like Kim et al [15].