We demonstrated a proton-based 3-terminal synapse device which shows symmetric conductance change characteristics. Using the optimized device, we successfully confirmed the improved classification accuracy of neural networks for on-chip training.
A wake-up free Hf0.5Zr0.5O2 (HZO) ferroelectric film with the highest remnant polarization (Pr) value to-date was achieved through tuning of the ozone pulse duration, the annealing process, and the metal/insulator interface. The ozone dosage during the atomic layer deposition of HZO films appears to be a crucial parameter in suppressing the mechanisms driving the wake-up effect. A tungsten capping electrode with a relatively low thermal expansion coefficient enables the induction of an in-plane tensile strain, which increases the formation of the orthorhombic phase while decreasing the formation of the monoclinic phase during the cooling step of the annealing process. Therefore, increasing the annealing temperature TA followed by rapid cooling to room temperature resulted in a substantial increase in the 2Pr value (≈ 64 µC/cm 2 ). However, the leakage current increased considerably, which can affect the performance of metal-insulator-metal (MIM) devices. To reduce the leakage current while maintaining the mechanical stress during thermal annealing, a 10 nm Pt layer was inserted between the W/HZO bottom interface. This resulted in a ~ 20-fold decrease in the leakage current while the 2Pr value remained almost constant (~ 60 µC/cm 2 ). The increase in barrier height at the Pt/HZO interface compared to that of the W/HZO interface coupled with the suppression of the formation of interfacial oxides (WOx) by the introduction of a Pt/HZO interface serves to decrease the leakage current.
Ferroelectric materials are promising candidates for synaptic weight elements in neural network hardware because of their nonvolatile multilevel memory effect. This feature is crucial for their use in mobile applications such as inference when vector matrix multiplication is performed during portable artificial intelligence service. In addition, the adaptive learning effect in ferroelectric polarization has gained considerable research attention for reducing the CMOS circuit overhead of an integrator and amplifier with an activation function. In spite of their potential for a weight and a neuron, material issues have been pointed out for commercialization in conjunction with CMOS processing and device structures. Herein, we review ferroelectric synaptic weights and neurons from the viewpoint of materials in relation to device operation, along with discussions and suggestions for improvement. Moreover, we discuss the reliability of HfO2 as an emerging material and suggest methods to overcome the scaling issue of ferroelectrics.
Wake‐up effect is still an obstacle in the commercialization of hafnia‐based ferroelectric thin films. Herein, the effect of defects, controlled by ozone dosage, on the field cycling behavior of the atomic layer deposited Hf0.5Zr0.5O2 (HZO) films is investigated. A nearly wake‐up free device is achieved after reduction of carbon contamination and oxygen defects by increasing the ozone dosage. The sample which is grown at 30 s ozone pulse duration shows about 97% of the woken‐up Pr at the pristine state whereas that grown below 5 s ozone pulse time shows a pinched hysteresis loop, that underwent a large wake‐up effect. This behavior is attributed to the increase in oxygen vacancy and carbon concentration in the films deposited at insufficient O3 dosage, which is confirmed by X‐ray photoelectron spectroscopy (XPS). The X‐ray diffraction (XRD) scan shows that the increase in ozone pulse time yields the reduction of tetragonal phase; therefore, the dielectric constant reduces. The I–V measurements reveal the increase in current density as the ozone dosage decreases, which might be due to the generation of oxygen vacancies in the deposited film. Finally, the dynamics of wake‐up effect is investigated, and it appears to be explained well by the Johnson–Mehl–Avrami–Kolmogoroff model, which is based on structural phase transformation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.