Enhancement of microfluidic particle separation using cross-flow filters with hydrodynamic focusing

Abstract

A microfluidic chip is proposed to separate microparticles using cross-flow filtration enhanced with hydrodynamic focusing. By exploiting a buffer flow from the side, the microparticles in the sample flow are pushed on one side of the microchannels, lining up to pass through the filters. Meanwhile a larger pressure gradient in the filters is obtained to enhance separation efficiency. Compared with the traditional cross-flow filtration, our proposed mechanism has the buffer flow to create a moving virtual boundary for the sample flow to actively push all the particles to reach the filters for separation. It further allows higher flow rates. The device only requires soft lithograph fabrication to create microchannels and a novel pressurized bonding technique to make high-aspect-ratio filtration structures. A mixture of polystyrene microparticles with 2.7 μm and 10.6 μm diameters are successfully separated. 96.2 ± 2.8% of the large particle are recovered with a purity of 97.9 ± 0.5%, while 97.5 ± 0.4% of the small particle are depleted with a purity of 99.2 ± 0.4% at a sample throughput of 10 μl/min. The experiment is also conducted to show the feasibility of this mechanism to separate biological cells with the sample solutions of spiked PC3 cells in whole blood. By virtue of its high separation efficiency, our device offers a label-free separation technique and potential integration with other components, thereby serving as a promising tool for continuous cell filtration and analysis applications.

FIG. 1.

Schematic of (a) typical cross-flow filtration and (b) our proposed separation mechanism. (a) The particles from the sample flow spread across the microchannels and only the particles reach to the filters passing through for filtration (b) The particles in the sample flow are concentrated by buffer flow and move toward one side of the channels, then are separated by the microfilters. Those particles, which are smaller than the filter size pass through the filter, move along the waste channel and are collected from the outlet. Meanwhile, other particles, larger than the filter size, continuously move along the main channel and are collected from the target outlet. (c) The flow channels are modeled to three different flow resistances R₁ (for the main channel), R₂ (for the filters), and R₃ (for the waste channel). Key geometrical dimensions in our filtration mechanism: length of the main channel (L), width of the main channel (W_main), width of the sample flow (b₀), and width of filter pore (W_pore).

FIG. 2.

(a) Trajectories of the sample flow from our three-dimensional flow simulation model. The color indicates the pressure gradient. The insets are the close-up views to locally display the trajectories of the sample flow around the filtration regions. The trajectories of the sample flow move toward the filtration regions while the buffer flows are at higher flow rates. The higher density of these flow trajectories indicates a greater portion of the sample flow being pushed through the filters with a larger pressure drop. (b) The relationship between the distance from the beginning of the filter region and the width of the sample flow.

FIG. 3.

(a) Fabrication process of our devices. (b) The silicon mold had a larger etching depth in the microchannel due to a large opening area and a shorter depth in the filter gap. (c) SEM picture for microchannel close to the inlet with microfilters. The picture shows the main channel, waste channel, and an array of filters with pores in rectangular cross section. The main channel is 100 μm wide, 30 μm high; the waste channel is 65 μm wide, 30 μm high; the filters has the quadrangle pore in 20 μm wide, 40μm long with a 60° tilt angle. There were 4 μm gaps between the filters. (d) The PDMS microchannel height was measured by an optic type surface analyser. (e) When the pressurized PDMS bonding technique was used at the pressure of ∼15 kPa, most of the filter areas were bonded onto the substrate.

FIG. 4.

(a) A virtual flow boundary was formed between the sample flow (e.g., DI water) and buffer flow (e.g., blue ink). Both flows were pushed into the microchannel at an equal flow rate of 1$\hspace{0.17em}\mu \mathrm{l}$/min. (a-i) The width of the sample flow (b₀) was 50 μm when two flows initially merged. (a-ii) The sample flow width gradually decreased as the flow proceeded in the microchannel. (a-iii) Most of the sample flow was pushed through the filter into the waste outlet. (b) The images of the particles moving along in the microchannel showed the filtration process in three steps: (b-i) the particles in the sample flow were pushed toward one side of the channel (particle-focusing step); (b-ii) small particles passed through the filters and flowed into the waste channel while large particles remained in the main channel (particle filtration step); and (b-iii) most of the small particles were in the waste channel and proceeded, while the large particles in the main channel, even though the sample flow had mostly flown into the waste channel (particle isolation step).

FIG. 5.

(a)–(c) Representative microscopic images showed: (a) the mixture of the microparticles of 2.7 μm and 10.6 μm before separation, (b) the samples collected from the target outlet after separation, and (c) the samples collected from the waste outlet after separation. (d)–(f) Flow cytometer data indicated: (d) Relative concentration of the samples before separation, (e) relative concentration of the samples collected from the target outlet, and (f) relative concentration of the samples collected from the waste outlet. Note: The numbers near the gated groups represented the percentage of the grouped particle amount to the total particle amount.

FIG. 6.

Separation efficiency of our devices at different flow conditions: (a) Recovery rates of the large particles (in the target outlet) and small particles (in the waste outlet) as the buffer flow rate increased from 2.5 μl/min to 40 μl/min (V_sample = 10 μl/min). (b) The purity of the large particles and small particles as the buffer flow rate increased from 2.5 μl/min to 40 μl/min.

FIG. 7.

(a) PC3 cells were spiked in to whole blood at the concentration of PC3:RBCs = 10³:10⁴ cells/μl for testing. The recovery rates of PC3 cells and RBCs at the concentrations of 10³:10⁴ cells/μl for PC3 cells:RBCs, as the flow rates changed from V_buffer:V_sample = 10:10 (μl/min) to 40:10. (b) PC3 cells were spiked in to whole blood at the two different concentrations (PC3:RBCs = 10²:10⁵ and 10³:10⁴ cells/μl) for testing. The recovery rates of PC3 cells and RBCs at the flow rates of 40:10 (μl/min). (c) The recovery rates of PC3 cells as the flow rates changed from 40:10 (μl/min) to 200:50.