Microfluidic resonators with two parallel channels for independent sample loading and effective density tuning

This paper reports doubly clamped microchannel embedded resonators with two independent and parallel channels integrated for effective sample density tuning for the first time. With the aid of such a unique design, each fluidic channel can be independently accessed thus different liquid samples can be loaded simultaneously. The proposed fluidic resonators are batch fabricated by depositing silicon nitride, polysilicon, and silicon nitride sequentially on top of a set of 4-inch silicon wafers and sacrificing the middle polysilicon layer with potassium hydroxide (KOH). The sacrificial process defines two parallel channels and releases doubly clamped beam resonators simultaneously. In addition, an off-chip vacuum clamp with optical and fluidic access is custom-made to operate each resonator with enhanced quality factor. The microfluidic resonators mounted on the custom vacuum clamp are thoroughly characterized with a laser Doppler vibrometer and used to measure the effective sample density ranging from 395 to 998 kg/m.


Background
Microfluidic resonators including suspended microchannel resonators (SMRs) [1] and hollow microtube resonators (HMRs) [2,3] have been applied for measurements of liquids, synthetic particles, and biological matters including cells, bacteria, virus, and exosomes with the outstanding performance. However, the fabrication process of such promising microfluidic resonators is yet laborious thus expensive since the majority of related research effort is focused on further improvement of device performance, mostly resolution, by making smaller resonators or achieving high quality factor via wafer-level on-chip vacuum packaging. In addition, only a single channel or bifurcated channels are embedded within resonators to date which limit the independent injection and characterization of multiple samples simultaneously. This may particularly be useful in an industrial application where a calibration standard needs to be present in the sensor all the time while samples of interest are tested. Additionally, such an arrangement also helps in achieving higher accuracy through the concept of differential measurements of density. Therefore, multiple samples of interest must be tested sequentially one at a time and this, in turn, increases the overall analysis time, risk of contamination, and inconvenience in calibration thus results in higher operation and maintenance costs. Moreover, mass densities of liquids are a few orders of magnitude larger than those of gases, the intermediate range of sample density is not available with microfluidic resonators with a single embedded channel [4][5][6]. In this paper, we propose simple fabrication of microfluidic resonators with two parallel channels and demonstrate independent sample loading and effective sample density tuning. Lee et al. Micro and Nano Syst Lett (2020) 8:16 Experimental details Fabrication Figure 1 shows design and fabrication processes of the proposed doubly clamped microfluidic resonators with two embedded parallel channels. Proposed microfluidic resonators are batch fabricated with a set of 4-inch silicon wafers by deposition, etching, and sacrificial process [7,8]. Silicon nitride and polycrystalline silicon are sequentially deposited by low pressure chemical vapor deposition (LPCVD). Next, regions for two parallel channels of microfluidic resonators are patterned by photolithography and reactive ion etching on the polysilicon layer. Then, another LPCVD silicon nitride layer is deposited on the patterned polysilicon and previously deposited silicon nitride layer. The secondly deposited silicon nitride layer defines channel sidewalls and top covers of microfluidic resonators. Finally, patterned polysilicon regions are sacrificially removed by potassium hydroxide (KOH) etching and microfluidic resonators are released at the same time. Two parallel microchannels (length of 500 µm, width of 16 µm, and height of 3 µm) are integrated on top of a plain doubly clamped beam (length of 500 µm, width of 56 µm, and height of 0.4 µm). Fundamental resonance frequency of fabricated resonators with two parallel microchannels is calculated to be 218.8 kHz. Larger sample delivery and waste collection channels are connected with two channels embedded in the microfluidic resonator and four liquid ports that are configured during the polysilicon sacrificial etching. Figure 2 shows optical and scanning electron micrographs of a fabricated doubly clamped microfluidic resonator with two embedded channels. The scanning electron micrograph in Fig. 2d shows an intentionally damaged resonator near its clamping region that confirms well-defined parallel channel structures.

Device mount
In order to measure the mechanical resonance frequency (with high quality factor) of the microfluidic resonators, a custom off-chip vacuum clamp is designed and made using a precision CNC milling machine. The plug-n-play style vacuum clamp made it highly practical to exchange the sensors after each measurement.

Measurements
To test the custom-made vacuum clamp, amplitude and phase responses for a microfluidic resonator (with air present in the both channels) are collected with a laser doppler vibrometer (MSA-500, Polytec) at partial vacuum as well as atmospheric pressure. Figures 4a and b show data for 0.062 mbar where the fundamental resonance frequency and corresponding quality factor are 218.8 kHz and 10,940, respectively. Figure 4c shows the quality factor as a function of the vacuum level and Fig. 4d shows the maximum amplitude at the fundamental resonance frequency as a function of the vacuum level. At 1 atm and 6.1 × 10 −5 mbar, resonance frequencies are 209.2 and 218.9 kHz, respectively and quality factors are 186 and 15,094, respectively. Both the quality factor and the maximum amplitude are significantly increased under vacuum condition.
Next, the microfluidic resonator is characterized while its two channels are filled with different liquid samples. Figure 5a shows the resonance frequency shift when two parallel channels are partially or fully filled with ethanol or water with referenced to the case when both channels are empty (filled with air). For a same pair of samples, four scenarios are possible upon sample loading considering duplication. When injection channels for two different samples are switched, there are slight mismatches in the resonance frequency shift observed at around -24, -29, and -46 kHz. Those discrepancies are attributed to the fabrication tolerance. Figure 5b represents the resonance frequency shift as a linear function of the average density, (ρ A + ρ B )/2. With ethanol and air, the effective sample density as low as 395 kg/m 3 is easily configured. As expected, further down tuning of the effective sample density is enabled by adding more embedded channels.

Conclusions
This paper reports microfabrication of doubly clamped beam resonators with two parallel embedded channels that can be injected with identical or different liquid samples or selectively emptied. Especially, the effective sample density can be significantly lowered by injecting a liquid sample into only one channel. The custom offchip vacuum clamp introduced herein simply realizes