High density 3D printed microfluidic valves, pumps, and multiplexers

In this paper we demonstrate that 3D printing with a digital light processor stereolithographic (DLP-SLA) 3D printer can be used to create high density microfluidic devices with active components such as valves and pumps. Leveraging our previous work on optical formulation of inexpensive resins (RSC Adv., 2015, 5, 106621), we demonstrate valves with only 10% of the volume of our original 3D printed valves (Biomicrofluidics, 2015, 9, 016501), which were already the smallest that have been reported. Moreover, we show that incorporation of a thermal initiator in the resin formulation along with a post-print bake can dramatically improve the durability of 3D printed valves up to 1 million actuations. Using two valves and a valve-like displacement chamber (DC), we also create compact 3D printed pumps. With 5-phase actuation and a 15 ms phase interval, we obtain pump flow rates as high as 40 μL min(-1). We also characterize maximum pump back pressure (i.e., maximum pressure the pump can work against), maximum flow rate (flow rate when there is zero back pressure), and flow rate as a function of the height of the pump outlet. We further demonstrate combining 5 valves and one DC to create a 3-to-2 multiplexer with integrated pump. In addition to serial multiplexing, we also show that the device can operate as a mixer. Importantly, we illustrate the rapid fabrication and test cycles that 3D printing makes possible by implementing a new multiplexer design to improve mixing, and fabricate and test it within one day.

found experimentally that it is ∼3.5-5.5h a where h a is the inverse of the resin absorption coefficient), while the minimum reliable lateral dimension is 4 times the pixel pitch in the build plane. Nonetheless, some impressive work has been done with commercially available 3D printing tools and materials, including reactionware devices, 9,10 snap-together discrete microfluidic elements, 11 a continuous nitrate-monitoring device for water contamination, 12 mail-order microfluidics using a commercial service bureau, 13 active devices (valves and pumps), 14 and microfluidic circuitry (analogous to electronic circuitry). 15 The primary disadvantages of these approaches to date is the relatively large minimum void size and consequent overall device size.
In this paper we show significant miniaturization of 3D printed microfluidic devices with integrated valves and pumps based on our previous resin formulation work 8 and our demonstration of the first reported 3D printed valves. 1 Specifically, we show how to use a DLP-SLA 3D printer with our inexpensive custom resin formulation to fabricate robust membrane valves 40 pixels in diameter (1.08 mm) with a minimum chamber height of 60 µm. These valves are only 10% the volume of our previous 3D printed valves, 1 and we have improved their durability from 800 actuations to 1 million actuations. To achieve such durability, we modify the resin composition by adding a thermal initiator such that a post-printing 30 minute oven cure drives further polymerization of the material to create a greater degree of cross linking and mechanical toughness. We then demonstrate a simple pump structure consisting of two valves and one displacement chamber (DC), and experimentally characterize its maximum back pressure and maximum flow rate. Finally, we combine 5 valves and one DC into a compact 3-to-2 multiplexer with integrated pump, utilizing the flexibility of 3D printing to densely arrange device elements within the 3D volume of the device. We also show that the multiplexer can be used as a mixer and that its mixing efficiency can be improved by increasing the number of inlets in the DC.

Materials
Our resin formulations consist of monomer, photoinitiator, and absorber, which for this work are poly(ethylene glycol) diacrylate (PEGDA, MW 258), Sudan I, and Irgracure 819, respectively. 1,16,17 We also include a thermal initiator, azobisisobutyronitrile (AIBN), for post-print thermal curing, the details of which are discussed in Sect. 3.1. It is important to note that use of a low molecular weight PEGDA results in excellent water stability for fabricated parts, 16 with no swelling or degradation in mechanical strength. PEGDA, Sudan I, and AIBN are obtained from Sigma-Aldrich (St. Louis, MO), and Irgacure 819 from BASF (Vandalia, Illinois). Each is used as received.
The specific resin formulation employed for the work reported in this paper is the 0.4% Sudan I resin discussed in Ref. 8. It is prepared by mixing 0.4% (w/w) Sudan I, 1% (w/w) Irgacure 819, and 0.01% (w/w) AIBN with PEGDA, and sonicating for 30 minutes. The resin is stored in an amber glass bottle wrapped in aluminum foil to protect it from light exposure.

3D printer
We use an Asiga Pico Plus 27 DLP-SLA 3D printer as described in Ref. 8, which has an LED peak wavelength of 412 nm and an in-plane resolution (pixel pitch) of 27 µm. Microfluidic devices in an individual print run are fabricated on a glass slide (25 mm x 37.5 mm x 1.2 mm) which is attached to the printer build table with double-sided tape. We experienced no issues with the slide damaging the teflon film comprising the bottom of the resin tray as long as we followed the 3D printer manufacturer's build table alignment procedure. Each slide is prepared by cleaning with acetone and isopropyl alcohol (IPA), followed by immersion in 2% 3-(trimethoxysilyl) propyl methacrylate in toluene for 2 hours. After silane deposition slides are kept in toluene until use.
There are two reasons we use glass slides. The first is that they avoid the need to fabricate the first device layer on the rough (anodized Al) surface of the 3D printer build table, which, especially for resins with high optical absorbance, requires a significantly longer exposure time for the first layer to attach to the build table. Long exposure times deplete the available binding sites on the surface of the layer, making attachment of the next layer problematic. The second reason is that the smooth surfaces of the glass slide offer convenient optical access to observe the interior components of the microfluidic device.

Device fabrication
Our build layer thickness, l, is 10 µm, which results in a normalized layer thickness, ζ = l/h a , of 0.57 for the 0.4% Sudan I resin. This is well within the optimal build thickness range we established in Ref. 8.
The key active component in our devices is a membrane valve, the structure of which is shown in Fig. 1a. 1 The valve consists of a 20 µm thick membrane (i.e., 2 build layers) sandwiched between two cylindrical voids which comprise a control chamber (100 µm high) and a fluid chamber (60 µm high), each 40 pixels (1.08 mm) in diameter. The corresponding dimensions of our original 3D printed valves are 50 µm membrane thickness, with 250 µm control chamber and 250 µm fluid chamber heights, both of which are 2 mm in diameter. 1 The valves in this paper are only 10% of the volume of the valves in our original paper (0.165 mm 3 compared to 1.73 mm 3 ). The valves in our original paper were fabricated with a different DLP-SLA 3D printer (B9 Creator) prior to developing our quantitative approach to resin formulation. 8 When no pressure is applied to the control chamber (as illustrated in Fig. 1b), the valve is open and fluid can flow between the two channels at the bottom of the fluid chamber. A photomicrograph of an open valve is shown in Fig. 1d. The lighting makes it easy to see the pixelation of the bottom surface of the fluid chamber. The measured surface roughness for horizontal surfaces fabricated with 0.4% Sudan I resin is 0.5 µm with a length scale the size of the pixel pitch. As shown in Fig. 1c, when pressure is applied to the control chamber the membrane deflects downward and seals the fluid channels. The central circular region in which the membrane contacts the bottom of the fluid chamber can be clearly seen in Fig. 1e.
In our devices valves are connected with flow channels that 2 | 1-9

sections.
A fabricated device is shown in Fig. 4c, looking from below through the glass slide substrate. The valves V1, V2, and V3 are occluded (indicated by dashed white lines) behind V4, DC, and V5 (indicated by solid white lines). PTFE tubing is epoxied in the inlets on the left, and the outlet flow channels are on the right. The three inlet tubes contain buffer (Buffer), diluted red dye in water (Red), and diluted black dye in water (Black), respectively. Both Red and Black have previously been pumped through the device, followed by Buffer. This is the reason the flow channels from the Red and Black inlets to the DC are filled with Red and Black, respectively. Figures 4d-4i show an example set of operations conducted with the multiplexer that exercise the various combinations of inlets to outlets. It begins with Red being pumped to Outlet 1 (Fig. 4d), followed by Black to Outlet 2 (Fig. 4e), Buffer to Outlet 2 (Fig. 4f), Buffer to Outlet 1 (Fig. 4g), Red to Outlet 2 (Fig. 4h), and Black to Outlet 1 (Fig. 4i). Its dynamic operation is shown in ESI † Movie S1. During each inlet/outlet combination, the pump is typically run for 50 periods to more than fully flush the previous fluid in the large (500 µm × 500 µm × 2.5 mm) outlet channels. The large outlet channel size is chosen solely to make it easy to see the colored fluids. As a further note, it takes approximately 3 pump periods to flush fluid from the DC when switching from one fluid to another.
The multiplexer can also be used as a mixer by, for example, operating two of the inlet valves simultaneously during pumping, in which case the fluids from the two inlets will be drawn together through the pump and expelled into an outlet. Prior to initiating pump action, we first opened V2, V3, DC, and V5 while raising the reservoirs from which red and black fluid are drawn about 15 cm above the microfluidic device. All of the other valves are closed. Fig. 5a illustrates the gravity-induced flow of Red and Black through the device. The upper right inset shows Black entering the DC from below and Red from above, corresponding to the physical locations of their inlets into the DC. The upper left inset shows the segregated Red/Black flow stream through the DC outlet channel, which maintains its segregation through V5 and Outlet 2 (lower right inset image). Clearly, the only mixing that occurs is due to diffusion across the boundary between the two fluids. Now consider simultaneous pumping from Red and Black into Outlet 2 according to the timing logic in Table 2. The results are shown in Fig. 5d in which each image shows the device state for the corresponding timing logic in Table 2 (note that t 5 is the same state as t 0 ). Prior to taking these images, the device was operated long enough such that it had reached a steady-state condition. At t 1 fluid is draw into the DC through open valves V2 and V3, both of which are closed at t 2 . The inset for t 2 shows the spatial segregation of fresh Red and Black just drawn into the DC. At t 3 the valve to Outlet 2, V5, is opened, following which fluid is expelled from the DC through V5 into Outlet 2 at t 4 . The inset at t 4 shows similar Red/Black segregation in the DC outlet channel, but by the time it makes it through V5 and into Outlet 2 there is much more mixing than in Fig. 5a. However, there is still a discernible red streak near the middle of Outlet 2 (see inset at t 5 ).
As soon as we got this result we realized that mixing could be improved by increasing the degree to which Red and Black are interleaved in the DC, which is easily accomplished with a change in geometry. Consider for example the bottom view of the DC in Fig. 5b in which Red is introduced into the DC through flow channel R1, and Black through B1. By splitting each Red and Black inlet into two inlets and interleaving them as shown in Fig. 5c (labeled as R1, R2, B1, and B2), additional mixing can be created in the DC. The mixing properties of the resultant device are shown in Fig. 5e using the same sequence of steps as Fig. 5d. The inset image for t 2 shows red and black regions localized around their respective DC inlets, while the inset at t 4 shows more Red/Black streams in the DC outlet channel, resulting in better mixing in Outlet 2 as seen in the inset at t 5 . The rapid iteration time enabled by 3D printing allowed us to redesign, fabricate, and test this new DC inlet design within a day.
As a final comment, the 3-to-2 multiplexer in Fig. 4 can be readily scaled to larger numbers of inlets and outlets. At this point it is unclear what the practical scaling limit is, but it will likely be determined by the fabrication yield of the valves, in which case our fabrication techniques would need to be further refined to increase the valve yield.

Summary
In this paper we have demonstrated the potential of 3D printing to enable both rapid fabrication iteration and high density integration of microfluidic components. We have reported the smallest yet 3D printed valves and characterized valve performance and durability. Incorporation of a thermal initiator in the resin together with a post-print bake dramatically improves durability. Fifty two out of 54 valves were successfully tested up to 10,000 actuations, at which point we stopped the tests because of how long they took. One valve was tested to 1 million actuations, after which it still performed well. We have used these valves to create compact pumps and characterized their maximum back pressure and maximum flow rate. Flow rates as high as 40 µL/min have been demonstrated. We have also demonstrated a 3-to-2 multiplexer with integrated pump, and shown that it can also be used as a mixer. Moreover, we have shown the ability to implement and test a new idea to improve mixing within only a day, thereby illustrating the power of 3D printing to enable a "fail fast and often" iterative device development strategy.