Challenges in implementation of silicon-based DNA hybridization sensors

Silicon-based DNA sensors have shown to be able to provide detectable signals to DNA hybridization to tethered probes on their surfaces. Since then, there has been considerable research on methods of integrating these biosensors (named BioFETs) with mainstream silicon processing via post-processing exposure of the gate oxides of large area FETs using etching processes. This method promises lower cost biosensors, but also introduces possible sources of error in the measurements. In addition, due to the localization of the charges of the macromolecules and their polymorphism, it becomes especially challenging to predict a BioFET response signal. Complications arise from possible cross-hybridization with foreign species and with neighboring probes. The purity of PCR samples that are introduced into the BioFET can especially cause unwanted signals and jeopardize the specificity of the sensor. In this communication, we attempt to identify and describe several of these factors that can significantly affect the performance of the BioFET. We show that simulations can greatly aid experiments in describing the sources of the uncertainties and errors. Some of the sources are related to the nature of the chemical interface and the DNA molecules, while others are related to the construction of the device and its operation mode. We also discuss some techniques that might be used in the construction, as well as in the operation of the device, which can help reduce this variability and provide more accurate readings. These techniques can manifest in the biochemical realm, like proper processing of target DNA molecules or selection of proper buffer solutions. In other cases, the construction of the device itself, such as the use of insulators or the method of post-processing silicon, can affect the yield of this sensor. In addition, the operation of the device and thermodynamic conditions can also affect the total device signal-to-noise ratio. We show that all of these processes ha- e very similar effects and must be included, in one form or another, when building working models for a DNA BioFET.