1. Introduction

Microfluidics has a broad range of applications in point-of-care diagnostics, drug discovery, biomarkers, tissue engineering, and many others. Optofluidics uses fluidic technology and optics in an integrated system [1], and has found many applications including lab-on-a-chip devices, fluid-based and controlled lenses, optical sensors for fluids and for suspended particles, biosensors, imaging tools, etc. [2]. Merging optofluidics technologies with silicon photonics may offer unprecedented sophistication and control in optofluidic systems, but is currently limited by the scaling and fabrication limitations of microfluidic technologies.

Traditional microfluidic device fabrication technology uses polydimethylsiloxane (PDMS), glass, silicon, embossed or injection molded plastics, [3]. These technologies require access to cleanroom equipment and involve steps such as photolithographic microfabrication, molding and release, and careful alignment and bonding of each layer to fabricate a complete device. The achievable device dimensions are far larger than typical silicon photonic elements, and the fabrication steps are often cumbersome and time consuming, requiring days to even weeks to create a successful microfluidic device. In this work, we instead use custom 3D printing technology to interface directly with silicon photonic devices. This eliminates the need to fabricate and bond discrete layers individually, and also enables the utilization of device volume in all 3 dimensions for highly integrated component placement and fluidic routing. Moreover, this technology allows rapid fabrication (less than 15 minutes for most devices), modification, and testing of devices. Unlike commercially available 3D printing technologies, the printer used here demonstrates the ability to fabricate true microfluidic (< 100 $\mu$m) structures. Previous work has demonstrated the capability to print microfluidic devices with channels as small as 18 $\mu$m × 20 $\mu$m [4], valves with only 150 $\mu$m diameter, and highly integrated pumps and mixers [5]. This approaches the scale of many functional silicon photonic devices.

Here we demonstrate the ability to automate the tuning process of silicon photonic microring resonators using 3D-printed optofluidic modules, and build upon our previous work showing manual control of an integrated silicon photonic and microfluidic circuit [6]. This also extends the work of previous researchers, who have demonstrated microfluidic tuning of photonic circuit elements using of preselected fluids [7] and integrated mixing of fluids [8] using traditional microfluidic technologies. In contrast, this work uses a single silicon photonic chip directly interfaced with a 3D-printed microfluidic device, giving the designer the ability to quickly iterate on designs or change the functionality of the optofluidic device by replacing the 3D-printed microfluidics chip with another while reusing the photonic chip.

Here, we employ these advanced printing capabilities to demonstrate custom 3D-printed microfluidic automated control of a microring resonator’s resonance wavelength with on-chip mixing. The 3D-printed microfluidic mixing and delivery device is printed using a custom digital light processor stereolithographic (DLP-SLA) printer with a 365 nm LED light source, a pixel pitch of 7.6 $\mu$m in the plane of the projected image, and capable of depositing layers that are as small as 1 $\mu$m thick [4]. We used a custom resin consisting of 0.38% (w/w) Avobenzone UV absorber and 1% (w/w) phenylbis(2,4,6-trimethylbenzoyl) phosphine oxide (Irgacure 819) photoinitiator in PEGDA [9], [10]. The resin is not proprietary and available to the research community for use in other experiments and applications, many of which are being pursued. We also note that all the resin constituents are available commercially and were used as received.

2. 3D-printed microfluidics mixer design

The design and working principle of the device are illustrated in Fig. 1, which shows two distinct chips: a silicon photonic chip (left) containing an array of microring resonators varying in size from 10 $\mu$m to 80 $\mu$m, and a 3D-printed microfluidic chip (right) that clamps directly on the top of the silicon chip. The microfluidic chip mixes water and saline with reconfigurable concentrations, and then passes the resulting solution over the microring resonators, thereby tuning the resonance. To allow for microfluidic access while also protecting the photonic structures from clamping damage, a 2 $\mu$m thick SiO$_2$ cladding layer with 160 ×140 $\mu$m etched windows over the ring resonators is deposited on top of the silicon guiding layer.

 

Fig. 1. (Left) Layout of silicon photonics chip. The zoom in box shows a ring resonator with the etched oxide window. (Right) CAD model of the microfluidics chip. The blue outline over the chip corresponds with the blue outline of the same shape on the microfluidics chip which demonstrates the area where the microfluidics chip interfaces with the ring resonators.

Download Full Size | PDF

The microfluidic mixing and delivery mechanism can be described by following the fluids from entry to exit of the microfluidic chip. First, the unmixed fluids are prepared in two 20 cc plastic syringes pumped using a dual drive pump controlled by a computer. The fluids enter the microfluidic device via two separate pinholes (Fig. 2(a)) using 0.5 mm inner diameter flexible PTFE tubing. After entering the device the fluids pass through a pair of pneumatically controlled valves (Fig. 2(b)) with the supply of pneumatic pressure coming from the leftmost pinhole. They then combine at a dilution serpentine channel (Fig. 2(c)) for mixing. At this scale, the small Reynolds number ensures laminar flow. While more efficient methods of microfluidic mixing are possible [11], these simpler devices performed well and more complicated mixing methods were not employed. After mixing, the fluid passes through a channel with dimensions calibrated for optical examination (Fig. 2(d)) to validate the mixing efficacy, which is discussed in further detail below. The mixed fluid then passes through a 40 $\mu$m deep interface channel (Fig. 2(e)) where it comes into contact with the microring resonators. A tight seal with the silicon photonics chip is ensured by using a 20 $\mu$m tall micro-gasket around the channel. Before exiting the device through the rightmost pinhole, the fluid passes through another pneumatically controlled valve which, when used in combination with the other valves, can stop the flow of fluid and create an isolated chip free from contamination with external fluids. In order to ensure precise alignment during clamping, an alignment tab (Fig. 2(f)) is designed so that it sits flush with the photonics chip.

 

Fig. 2. 3D CAD model of the microfluidics mixer. (a) Chip-to-world interface where tubing can be attached. (b) Microfluidic valves actuated by air pressure. (c) Serpentine dilution mixing channel. (d) Optical concentration examination region. (e) Photonics overlay channel where the fluid interacts with the photonic devices. (f) Alignment tab to align the microfluidics chip with the photonics chip properly.

Download Full Size | PDF

Using the clamping mechanism shown in Fig. 3 we seal the micro-gasket on the microfluidics chip to the silicon photonics chip. The base housing of the clamping mechanism was 3D-printed using commercially available 3D-printing techniques. An acrylic slide is fastened to the base with 4 steel screws. The acrylic slide is used so that a microscope can properly visualize aligning a fiber array to the silicon photonics chip. After automated tuning of the cladding refractive index is complete (described below) three (3) pneumatically controlled valves are closed isolate the system and prevent the internal mixed fluids from further interacting with external fluids in the syringe pumps.

 

Fig. 3. Clamped Rendering of the 3D printed device

Download Full Size | PDF

3. Mixer validation for concentration

On-chip measurement and validation of fluid mixing was performed to ensure the on-chip mixing reliability. There are two common approaches to measuring the concentration of a liquid in a microfluidic structure: using a fluorescence-based or an absorption-based analyte [5]. We employ an absorption-based method that only requires a simple imaging system. The mixed fluid is imaged using a microscope, and Beer’s Law is used to relate the general absorbance of the fluid to the concentration [5]:

Figure 4 shows the concentration validation testing setup where the use of a de-embedded microfluidic chip was fabricated and tested to isolate the performance of the mixing. A solution of black dye was used as the absorptive fluid (fluid A) and water was used as the non absorptive fluid (fluid B).

 

Fig. 4. Concentration testing setup. A computer is connected to a microscope camera and a dual-channel syringe pump. On the microscope is placed the microfluidics chip. The microscope is focused on the region of interest and a real image can be seen in the figure labeled ‘Microscope View.’

Download Full Size | PDF

Figure 5 shows the mixing results obtained by switching between two pairs of flow rates $r_A , r_B$: {($r_A =$90 $\mu$L/min, $r_B =$ 10 $\mu$L/min), ($r_A = 10$ $\mu$L/min, $r_B = 90$ $\mu$L/min) }. Other pairs ($r_A, r_B$) were also tested with a total flow rate $r_A + r_B$ kept constant at 100 $\mu$L/min in order to adjust the concentration levels while maintaining the same mixing time withing the chamber. The total volume of the mixing chamber is 60 $\mu$m $\times$ 200 $\mu$m $\times$ 61 mm $=$ 0.73 $\mu$L, corresponding to a mixing time of approximately 0.43 s. This mixing time was found experimentally to be near the lower limit for complete mixing in this proof-of-principle experiment for the highest concentration differences, though scaling the dimensions and employing more advanced mixing techniques may allow for better response times of 10’s of ms as demonstrated in other work [11]. While these tests were performed at several target concentrations, Fig. 5 target concentrations between 90% and 10% is one of the more difficult targets to obtain owing to the large difference in concentrations.

 

Fig. 5. Average concentration vs. time. Target concentrations of 90% and 10%.

Download Full Size | PDF

As seen in Fig. 5, the actual time for the mixing transients to approach equilibrium is on order 10’s of seconds, likely owing to limits of the maximum linear force applied by the syringe pumps (57 lbs), resulting in a non-instantaneous change in flow rates.

4. Automated resonance tuning

An automated control system operates the mixer and the calibration probe laser, and measures the output power of the add or drop ports. To control the mixing we operate a dual channel syringe pump. The two different channels control the flow rate of two syringes filled with different fluids, de-ionized water (fluid A) and 26.3% concentration aqueous solution of NaCl by mass (fluid B) which is the saturation limit of NaCl in water. We initialize the mixer with flow rates of fluid A and B at 50 $\mu$L/min. We then sweep the laser a through a free spectral range ( 2.3 nm) and fit the measured dropped power $P_{\textrm {drop}}$ to a rate-equation model of the optical resonance:

Figure 6 (top) shows the initial and final spectral response resulting from the automatic tuning of one of the microring resonators in the array. It shows the resonance shifting from 1600.8 nm to 1600.0 nm. The resonances are captured at two different times during the experiment as indicated in Fig. 6 (bottom) by vertical red lines. The first vertical red line corresponds to when the program sequence that controls the resonance was initiated. The second vertical red line corresponds to when the program no longer makes further adjustments to the resonance. The target resonant frequency was achieved and maintained with an accuracy of ±50 pm. The system has an effective tuning range of 4 nm, which is 1.7 times the FSR. The tuning range was calculated measuring the change in optical resonant wavelength of the microring resonator after adjusting the concentration of the fluid passing over the resonator from 0% to 26.3% concentration by mass of an aqueous solution of NaCl. This represents the maximum tuning range as 26.3% concentration is approximately the saturation limit of NaCl in water.

 

Fig. 6. Spectral tuning of a microring resonator. (Top) Measured (blue) and theoretical fit (red) of the dropped power as a function of input wavelength. (Bottom) Resonant frequency as a function of time.

Download Full Size | PDF

5. Conclusion

In summary, we have demonstrated the automatic control and tuning of microphotonic elements using 3D-printed optofluidic devices. While microring resonators were used in the present experiment as a proof-of-principle, other photon photonic circuit elements such as Bragg gratings, photonic crystals, waveguides, etc [12] may also benefit from 3D-printed microfluidics. The 3D-printed microfluidics control method demonstrated here may enable large-scale control, tuning, and reconfigurability of photonic circuits. For example, owing to the small feature size and precision of the 3D-printing technique, the overlay geometry of the microfluidics interface can target multiple elements individually. Furthermore, multiple interfacing channels and mixing modules can address multiple photonic elements for reconfigurable circuits.