Abstract
Contact resistance is traditionally regarded as an obstacle to be eliminated in transistors, limiting the charge injection across a wide range of semiconductors. Here we show that controlled variations in contact resistance, induced by asymmetric electrode geometry, can be exploited as a design parameter rather than treated as a drawback. Using ZnON/DNTT heterojunction antiambipolar transistors (AATs), we show that electrode placement defines distinct current pathways and enables multiple Gaussian-like transfer curves within a single device platform. Combining four electrode layouts with dual operating modes yields eight distinct Gaussian-like transfer profiles. This expanded functionality demonstrates that contact engineering enables tunable analog responses directly relevant to neuromorphic computing. The ability to adjust Gaussian amplitude, position, and width provides hardware-efficient implementations of activation functions, continuous weight representations, and probabilistic processing. Based on the obtained Gaussian responses, reinforcement learning tasks such as Duffing oscillator prediction and power consumption forecasting are performed, illustrating the applicability of AATs to nonlinear dynamic systems.