Each groove has staggered lengths of 865 μm and 1,000 μm. The grooves were designed to be at an angle of 45° to the channel

wall and were spaced with an interval of 840 μm (center to center) along the length of the channel. The electrodes were then fabricated on the Si wafer with grooves using a lift-off technique [17]. A 10-nm-thick Cr layer and a 40-nm-thick Au layer were deposited sequentially on a predefined photoresist layer on the Si wafer to form the electrode patterns. After defining the electrodes, the wafer was diced into smaller substrates (15 mm × 20 mm). The graphene monolayer was then transferred onto the Si wafer and placed between the electrodes. The resistance of the graphene was about 1 kΩ. Finally, the Si wafer with grooves, electrodes, and graphene was bonded to a polydimethylsiloxane (PDMS) layer, which had a fluidic channel of 100 μm in height, 1.5 mm in ABT-263 cost width, and 20 mm in length defined by replica molding. The PDMS layer was LCL161 cell line sealed to the Si surface by oxygen plasma treatment. Four types of samples were prepared in Figure 1f: Type 1: the electrodes aligned parallel to the flow in the absence of grooves Type 2: the electrodes aligned perpendicular to the flow in the absence of grooves Type 3: the electrodes aligned parallel to the flow in the presence of grooves Type 4: the electrodes aligned perpendicular to the flow in the

presence of grooves A syringe pump (Legato 180; KD Scientific, Holliston, MA, USA) was used to inject fluid through the PDMS microchannel. The flow-induced voltage over the graphene was measured using a digital multimeter (DM 2002; Keithley Instruments, Cleveland, OH, USA). All experiments were carried out at room temperature (25°C). Results and discussion Prior to measuring flow-induced voltage, we investigated the mixing performance of the herringbone grooves. Figure 2a,b shows the simulation results of mixing between pure water and dyed water without and with herringbone grooves, respectively. A 3-D numerical

simulation was performed using COMSOL Multiphysics (ver. 4.3a). The simulation geometry was identical to the actual microchannel device. Figure 2c,d shows the actual experimental data. Two streams Dipeptidyl peptidase of liquid (pure water and red dyed water) were injected into the microchannel via two inlets using a syringe pump. In the absence of herringbone grooves, only a minimal amount of mixing due to thermal diffusion was observed at the outlet of the channel in both simulated and experimental data. On the other hand, significantly more mixing was observed in the device with herringbone grooves. Mixing performance was also evaluated from the coefficient of variation (CV) [18], which is a normalized measure of dispersion of a probability distribution. The CV of concentration is considered a good measure of mixing quality. A positive value (approximately 1.0) indicates no mixing, and a value of 0 indicates complete mixing. As mixing progressed, the CV decayed exponentially from 1 to 0.