ABSTRACT
Electrolyte-gated silicon field-effect transistors (FETs) capable of detecting single molecules could enable high-throughput molecular sensing chips to advance, for example, genomics or proteomics. For solid-gated silicon FETs it is well-known that nano-scaled devices become sensitive to single elementary charges near the silicon-oxide interface. However, in electrolyte-gated FETs, electrolyte screening strongly reduces sensitivity to charges near the gate oxide. The question arises whether nano-scaling electrolyte-gated FETs can entail a sufficiently large signal-to-noise ratio (SNR) for the detection of single molecules. We enhanced a technology computer-aided design tool with electrolyte screening models to calculate the impact of the FET geometry on the single-molecule signal and FET noise. Our continuum FET model shows that a sufficiently large single-molecule SNR is only obtained when nano-scaling all FET channel dimensions. Moreover, we show that the expected scaling trend of the single-molecule SNR breaks down and no longer results in improvements for geometries approaching the decananometer size. This is the characteristic size of the FET channel region modulated by a typical molecule. For gate lengths below 50 nm, the overlap of the modulated region with the highly conductive junctions leads to saturation of the SNR. For cross-sections below 10-30 nm, SNR degrades due to the overlap of the modulated region with the convex FET corners where a larger local gate capacitance reduces charge sensitivity. In our study, assuming a commercial solid-state FET noise amplitude, we find that a suspended nanowire FET architecture with 35 nm length and 5 × 10 nm2 cross-section results in the highest SNR of about 10 for a 15-base DNA oligo in a 15 mM electrolyte. In contrast with typical silicon nanowire FET sensors which possess micron-scale gate lengths, we find it to be key that all channel dimensions are scaled down to the decananometer range.
ABSTRACT
Electrolyte screening is well known for its detrimental impact on the sensitivity of liquid-gated field-effect transistor (FET) molecular sensors and is mostly described by the linearized Debye-Hückel model. However, charged and pH-sensitive FET sensing surfaces can limit the FET molecular sensitivity beyond the Debye-Hückel screening formalism. Pre-existing surface charges can lead to the breakdown of Debye-Hückel screening and induce enhanced nonlinear Poisson-Boltzmann screening. Moreover, the charging of the pH-sensitive surface groups interferes with biomolecule sensing resulting in a pH interference mechanism. With analytical equations and TCAD simulations, we highlight that the Debye-Hückel approximation can underestimate screening and overestimate FET molecular sensitivity by more than an order of magnitude. Screening strengthens significantly beyond Debye-Hückel in the proximity of even moderately charged surfaces and biomolecule charge densities (≥1 × 1012 q/cm2). We experimentally show the strong impact of both nonlinear screening and the pH interference effect on charge-based biomolecular sensing using a model system based on the covalent binding of single-stranded DNA on silicon FET sensors. The DNA signal increases from 24 mV at pH 7 to 96 mV at pH 3 in 1.5 mM PBS for a DNA density of 7 × 1012 DNA/cm2. Our model quantitatively explains the signal's pH dependence with roughly equal nonlinear screening and pH interference contributions. This work shows the importance of reducing the net charge and the pH sensitivity of the sensing surface to improve molecular sensing. Therefore, tailoring the gate dielectric and functional layer of FET sensors is a promising route to strong silicon FET molecular sensitivity boosts.
