Effects of Channel Layer Properties on Weight-Update Characteristics of Low-Power Hf-Doped ZnO Thin-Film Transistors

We present a study on weight-update characteristics dependent on channel layer properties (i.e., geometrical and material properties) of low-power Hf–ZnO synaptic thin-film transistors (Syn-TFTs). Here, the channel layer properties can affect deep state capacitance ( ${C}_{\text {deep}}\text {)}$ , which leads to a change in the average subthreshold slope ( $\text{SS}_{\text {avg}}\text {)}$ . This varied SS_avg can also contribute to the sensitivity during synaptic processes (i.e., synaptic depression and potentiation), thus different weight-update characteristics (e.g., dynamic ratio, $\text{dr}_{w}\text {)}$ . To prove these, the static and pulsed characteristics of the Syn-TFTs with different channel layer properties (i.e., channel thickness and composition ratio of HfO₂/ZnO) are monitored. From experimental results, the Syn-TFT with a thinner channel is found to have a larger dr_w compared to the Syn-TFT with a thicker channel. This is mainly because the channel thickness is proportional to ${C}_{\text {deep}}$ . In terms of material properties, it is observed that the Syn-TFT with a higher composition ratio of HfO₂/ZnO has a larger dr_w in comparison with the Syn-TFT based on a lower composition ratio of HfO₂/ZnO. This can be attributed to the strong oxygen bonding energy of Hf ions, which can inhibit the increase of oxygen vacancies in the channel layer, reducing ${C}_{\text {deep}}$ . In this vein, the Syn-TFT with both a thinner thickness of the channel layer and a higher composition ratio of HfO₂/ZnO can achieve the higher dr_w (i.e., 60.9), resulting in the average recognition rate (RR) of 93.4%.