Trench field-effect transistors integrated in a microfluidic channel and design considerations for charge detection

Abstract

Field-effect transistors (FETs) combined with a microfluidic system allow for the electrical detection of charged materials moving in a microfluidic channel. Here, we demonstrate trench-shaped silicon FETs with the combination of a microfluidic channel that can be used for simultaneous electrical and optical detection of charged fluorescent beads. The n-channel silicon trench FETs have a maximum transconductance of 1.83 × 10⁻⁵ S at near-zero gate bias voltage, which is beneficial for the high sensitivity of electrical detection. The optical transparency and physical robustness of the integrated microfluidic channel are achieved by a polydimethylsiloxane (PDMS)/glass hybrid cover combining the good sealing characteristics of PDMS, and the thin and flat properties of glass. Device evaluation methodologies and measurement approaches are also presented demonstrating a synchronized time-lapse imaging and electronic detection of bead transport. The proposed device and design consideration could advance the promise of electronic sensing to measure potential differences induced by charged analytes.

The detection of charged materials using field-effect transistors (FETs) finds applications in proteomics and DNA technologies.^1–3 Furthermore, the detection of flowing, charged analytes in a microfluidic channel allows real-time sensing of viral particles and cells.^4–7 Hence, combining FETs and microfluidics is a promising approach for not only flowing charge detection but also practical biomedical devices. To date, various types of FET-microfluidic combinations have been reported, which include ion-sensitive FETs (ISFETs) with direct ink writing (DIW),⁸ silicon nanowire FETs with polydimethylsiloxane (PDMS),⁹ high-electron mobility transistor (HEMT) with PDMS,¹⁰ and graphene FETs with SU-8/PDMS.⁴ Most previous studies reported thick polymer-based covers for microfluidic channels, which could limit the resolution of optical imaging. Although these approaches can work for some applications, high-resolution imaging with surface contact microscopy requires thin cover glass for a short distance between a lens and an object. In terms of FET performance, silicon (Si) is the most successful semiconductor material offering stable and high electrical performance with decent manufacturing cost. Therefore, a device and measurement system that can perform synchronized electrical and optical monitoring of discrete flowing charged analytes could expand the utility of Si FETs as detectors for small, discrete particles.

Direct integration of FETs for the detection and imaging of charged particles within a microfluidic channel requires specialized design and fabrication considerations. They hinge on the effective integration of live epifluorescence imaging with electronic detection and fluidic manipulation of analytes within small, microchannel environments.^11–13 Accordingly, we embedded silicon-based trench FETs within an etched silicon microfluidic channel, topped with roofs made from coverslips coated with PDMS (“hybrid-cover”), to ensure liquid-tight seals, which also allowed imaging by high numerical aperture objectives (Fig. 1). These integrative features enabled synchronized optical and electrical measurements and tracking of charged particles undergoing transport, which were fostered by our detailed design/fabrication processes that placed functioning FETs within a microfluidic channel. Such device and measurement system developments were validated by simultaneous imaging and potential measurements of flowing charged fluorescent beads whose results portend an integrative approach to sensing discrete particles.

FIG. 1.

(a) Schematic of silicon-based trench field-effect transistors (FETs) integrated into a microfluidic channel. (b) Image of a fabricated device showing the PDMS coated coverslip (“hybrid-cover”) bonded to the silicon FETs. (c) Scanning electron microscopy image of a 1-μm depth silicon trench. (d) Simulation result of doping concentration showing the cross-sectional device structure. 10 nm of SiO₂ gate dielectric was thermally grown on the Si trench, and ion implantation was performed based on the simulation. (e) Image of four trench FETs and a microfluidic channel. (f) Magnified image of the device.

To co-fabricate field-effect transistors within a microfluidic channel, we started by embedding trench FETs on a silicon substrate. Four trench FETs were fabricated on the center of the p-type silicon substrate, and a microfluidic channel was formed by covering the trench with the hybrid-cover [Figs. 1(a) and 1(b)]. The input and output ports were made on the silicon substrate using a deep silicon etch to deliver charged targets from the back side of the device. The through-etched input/output ports are shown in Fig. 1(b). The 1 μm deep trench on the central area and the through-etched two ports were connected, and the entire area, except the pads of the transistors, was sealed by the hybrid cover.

The microfluidic channel is gradually tapered to a 2 μm-width channel right above the trench FETs for promoting a smooth flow of target materials mediated by hydrostatic pressure differences created in the bounding, fluidic ports [Fig. 1(c)]. N-type ion implantation was performed (phosphorus, 5 × 10¹⁵ cm⁻², 25 keV) to form source and drain on the p-type trench channel (boron, 2 × 10¹⁵ cm⁻³). The simulation result of doping concentration shows that the channel can be formed in the trench channel [Fig. 1(d)]. The SiO₂ was grown (10 nm) as a gate dielectric using a furnace. After etching the source/drain contact holes, source/drain metal was formed using a liftoff process. The designed trench FETs utilize the potential of charged targets as a gate potential. Therefore, no gate metal on the trench FETs is required. However, to characterize the performance of the trench FETs, testing trench FETs with metal gates were fabricated on the corner of the device. The trench FET has an electron channel width of 2 or 4 μm and length of 2 μm.

To fabricate the hybrid cover, PDMS base and crosslinker were mixed at a 5:1 ratio and degassed for 30–60 min under vacuum. The degassed PDMS was spin-coated on an oxygen (O₂) plasma-treated coverslip (Fisherfinest^TM, 18 × 18 mm) followed by a curing process (65 °C for 4 h). To remove the edge bead and flatten the hybrid cover surface, all the edge area was cut using a diamond scriber. Finally, the hybrid cover was bonded to the trench FET substrate after O₂ plasma treatment for irreversible bonding. Figures 1(e) and 1(f) show the microfluidic channel formed on the four FETs.

Figure 2(a) shows the fabricated device bonded to a printed circuit board (PCB). The device and PCB were wire-bonded to electrically connect each other. Figure 2(b) shows the measurement setup, which enables the electrical measurement and optical measurement at the same time. The PCB with silicon device was inverted and laid on a microscope objective. The source and drain of the transistor were connected to a low noise current pre-amplifier (SR 570, Stanford Research Systems) to apply drain voltage (V_D) and detect the drain current (I_D) generated by the gate potential change in the microfluidic channel. The preamplifier can detect drain current changes as small as 1 pA. To test the effect of gate potential through the liquid, two platinum (Pt) electrodes were placed in each port delivering input signal generated by a function generator.

FIG. 2.

(a) Fabricated device bonded to a printed circuit board (PCB) and electrically connected using wire-bonding. (b) Measurement setup enabling simultaneous optical and electrical measurements and tracking of charged materials flowing through the device. (c) Epifluorescence micrograph of charged beads.

To test microfluidic functionalities, charged beads (100 nm, FluoSpheres^® carboxylate-modified microspheres, Life technologies) were successfully transported with the full device. Figure 2(c) shows the charged beads within the device, while a movie shows controlled transport mediated by electrophoretic forces (supplementary material movie 1).

The characterization and quantitation of the sensing abilities of the trench FETs is necessary for the estimation of the overall sensitivity of the device to detect flowing, charged particles. The metal-gated trench FETs were first characterized to extract basic parameters, and the “liquid-gated” trench FETs were assessed to show the viability of the potential change detection.

Figure 3(a) shows the transfer curve of the transistor for both a low drain voltage (V_D = 0.1 V) and a high drain voltage (V_D = 1.1 V). The curve shows typical n-channel metal-oxide-semiconductor (NMOS) characteristics with the maximum transconductance (g_m) of 1.82 × 10⁻⁵ S. In FETs, g_m is the change in the drain current (I_d) divided by the small change in the gate voltage (V_g) with a constant drain voltage. Although the g_m can be used to get the absolute value of drain current change (i.e., output signal) in a certain gate voltage change (i.e., input signal), it does not reflect the relative sensing ability at a specific drain current level. Therefore, we have introduced a relative sensitivity value r, showing the sensing ability in a drain current level. The value r is expressed as follows, the g_m divided by I_d:

$$r=\frac{g_m}{I_d}=\frac{d{I}_d}{d{V}_g\times I_d},$$ (1)

where g_m is the transconductance, I_d is the drain current, and V_g is the gate voltage.¹⁴

FIG. 3.

(a) I_D-V_G characteristic and g_m/ID (i.e., relative sensitivity value r) plot of a trench FET. (b) I_D-V_D characteristic. (c) Schematic of a biasing condition for the detection of potential change in a microfluidic channel. (d) Drain current change measured with FET 1–3 depending on the potential change delivered to the liquid gate from a platinum electrode.

In Fig. 3(a), the maximum r value of the transistor was extracted to 8.67 S/A at V_d = 1.1 V and 5.35 S/A at V_d = 0.1 V. At zero gate voltage, the r value was 4.5 S/A. Because the peak r value is observed near zero gate bias, the FET sensor needs no reference gate bias or little reference bias to utilize the maximum sensing region. Normally, the maximum r value is seen at the subthreshold region, where the FET is not fully turned-on (i.e., below the threshold voltage of the transistor). In this region, the drain current is exponentially increased, and the subthreshold slope S [Eq. (2)] has the minimum value (i.e., the steepest slope),

$$S={\left(\frac{d\left({\mathit{log}}_{10}{I}_d\right)}{d{V}_g}\right)}^{-1}=2.3\frac{kT}{q}\left(1+\frac{C_{dm}}{C_{ox}}\right),$$ (2)

where k is the Boltzmann constant, T is the temperature, q is the electronic charge, C_dm is the depletion-layer capacitance, and C_ox is the gate dielectric oxide capacitance.¹⁵

The maximum r value is obtained when the subthreshold slope is the minimum. According to Eq. (2), the increased r value can be accomplished by reducing C_dm and increasing C_ox. Therefore, decreasing the thickness of gate oxide and reducing the trap density in the gate oxide and/or the interface of the silicon and the gate oxide will further improve sensitivity. Figure 3(b) describes the output characteristics of the trench transistor showing good saturation of the drain current. The saturation drain current indicates that the electron channel of the trench transistor is well-formed having the pinch-off condition.

After evaluating the trench transistor with a metal gate, the “liquid-gated” trench transistors were integrated within a microfluidic channel and then tested. Figure 3(c) describes the measurement setup applying a 5 V square wave input through phosphate buffered saline (0.05× PBS, pH 8.0) in the microfluidic channel.

The source electrodes and the Pt electrode in the other port were grounded, and the three drain electrodes were biased at 1 V. Figure 3(d) shows the measured drain current vs time for the three transistors. Each drain current was responded to the input voltage delivered to the liquid gate verifying the ability to detect the charge differences. It is worth noting that the initial response time and the amplitude of the drain current could vary in different locations in the microfluidic channel due to process variation of each transistor.

Negatively charged fluorescent beads (carboxylated) were used to test the charge detection ability of the device and the measurement system (fluorescence microscopy with 520 nm green filter). Figure 4 shows the customized measurement system showing the movement of the beads both optically and electrically. Figure 4(a) depicts the optically captured image while the charged fluorescent beads flow through the FET integrated microfluidic channel. We can also generate the movie using the periodically captured images while measuring the electric potential changes. Understandably, epi-illumination conditions generated photoelectric currents in the transistors, causing 10–100 times larger drain currents and increased noise. We, therefore, maximized electrical sensing by shuttering off the illumination during any electrical measurement, which produced the periodic potential cycles seen in Fig. 4(b). In each cycle, bead epifluorescence images were first acquired [i.e., low and short potential in Fig. 4(b)], followed by electrical measurements without illumination [i.e., high and long potential in Fig. 4(b)]. Our data showed that the relative potential was increased when charged beads were present compared to no charged beads, meaning that trench FETs fabricated within microfluidic channels will detect streaming charged targets. Note that the sensitivity of the charge detection would be influenced by the relative size difference between target materials and transistor's channel width (i.e., sensing area) because the relative potential change would be affected by both factors.

FIG. 4.

Optical imaging and electrical detection of charged fluorescent beads. Optical imaging and electrical measurement were staggered to avoid the large photoelectric current generated by the incident light. The relative potential increases with charged beads compared to that of without charged beads.

In this study, we demonstrate that trench FETs integrated within a microfluidic channel will detect charged materials undergoing fluidic transport. The Si-based trench FETs with a PDMS/glass hybrid-cover is a promising approach for combining fluidic components with in situ transistors in ways that support analyte manipulations, while melding electrical measurements and optical imaging capabilities. Although future optimization efforts will improve the sensitivity of charge detection, the device and measurement approach presented here will be of great utility for electrical and biological applications requiring detection of charge materials within microfluidic flows. For future studies, high-k dielectric materials such as Al₂O₃ or HfO₂ could be applied to offer higher sensitivity of the FET sensor.

See the supplementary material movie 1: flowing charged beads controlled by electrophoretic forces in the microfluidic channel.

AUTHOR DECLARATIONS

Conflict of Interest

The authors have no conflicts to disclose.

DATA AVAILABILITY

The data that support the findings of this study are available within the article and its supplementary material.