Lithium-Modulated 3-Terminal Artificial Synapse with Long-Term Memory Function Induced by Fluorinated Self-Assembled Monolayer

Neuromorphic computing has been introduced as a revolutionary computing architecture to overcome the limitations of von Neumann architecture. Neuromorphic computing is known for its low operational speed and high power consumption, especially when dealing with large volumes of data in areas such as artificial intelligence and big data. This new architecture conducts calculations and memory operations at the same time within a specific resistor, known as an 'artificial synapse'. These artificial synapses, which are structured in parallel circuits, can be distinguished by their adjustable analog electrical conductance. For use in neuromorphic accelerators, artificial synapses need to have the ability to adjust synaptic weights variably, linearly, and symmetrically with a high dynamic range, and they need to be able to maintain updated synaptic weights.<br/>Electrolyte-gated transistors (EGTs) are seen as a potential option for their low switching voltage thanks to the high capacitance of the electrolyte, their simple production process, and the high flexibility in device configuration. The way EGTs work is by altering the channel conductance through physical or chemical interactions of cations or anions in the electrolyte with the channel, thereby representing changes in synaptic weight. However, due to the natural characteristic of electrolytes to balance charges, an EDL formed at the interface by an immediately applied electric field is likely to be transient. Therefore, considering charge interactions with ions at the interface becomes crucial for achieving adjustable conductance with high linearity and long-term retention.<br/>In this study, we present Heneicosafluorododecylphosphonic acid (F21-DDPA) as an interfacial monolayer in EGT, specifically highlighting its role in the physisorption of ions caused by large dipole moment at the channel/electrolyte interface. The device uses amorphous indium gallium zinc oxide (a-IGZO) as the channel material, which is notable for its low off-current and high field-effect mobility in n-type thin-film transistors, and water-in-bisalt lithium electrolyte as the polymer dielectric, which is remarkable for its high areal capacitance and large electrochemical window. With the introduction of F21-DDPA in EGT, a strong ion-dipole force is observed between the lithium ions and F21-DDPA, leading to the immobilization of lithium ions at the channel/electrolyte interface. As a result, the lithium EGT with F21-DDPA shows exceptional synaptic characteristics, such as switching conductance with near-linearity and high dynamic range (~10), and long-term retention (&gt;30 min), even at a switching voltage below 3 V. Specifically, through chemical analysis and DFT calculations, we speculate that lithium ions become strongly trapped within the van der Waals gap of F21-DDPA. This finding suggests the potential of a specific capacity of lithium ions within F21-DDPA. We propose an EGT-based artificial synapse with multimodal behavior by suggesting a new method of integration for F21-DDPA, indicating its significant potential in neuromorphic accelerators performing both training and inference tasks.