Scholarlink

<p>Resistive random-access memory (RRAM) is a promising beyond-CMOS technology due to its non-volatility, scalability, and high ON/OFF ratio. Furthermore, a single RRAM cell can operate as an analog resistor, meaning that it can be used in more novel computing applications such as tunable synapses in neuromorphic computing. However, issues of poor uniformity and poor yield persist in RRAM. In this work, we performed a systematic study on six different TiO<sub>X</sub>/TiN RRAM samples with varying layer thicknesses. For each sample, we investigate how the layer thicknesses, presence of background oxygen, and testing parameters influence key performance metrics in each sample. These metrics include switching probabilities, failure mechanisms, operating voltage, ON/OFF ratio, device-to-device variation, and tunable resistance capabilities to determine their viability as multi-state memory cells. Our results reveal clear trade-offs between high ON/OFF ratio, high resistance tunability and high device-to-device variations. Furthermore, operating conditions within the same sample that produce large ON/OFF ratios are often not the same as those that produce low variations and high switching probabilities. The results presented in this paper provide a detailed insight into how intrinsic device properties and operating conditions influence RRAM performance and could be used for considerations in designing future RRAM technologies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Investigation of key performance metrics in TiOX/TiN based resistive random-access memory cells

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月3日

查看全文
  • Brandon R. Zink,
  • William A. Borders,
  • Advait Madhavan,
  • Brian D. Hoskins,
  • Jabez McClelland

摘要

Resistive random-access memory (RRAM) is a promising beyond-CMOS technology due to its non-volatility, scalability, and high ON/OFF ratio. Furthermore, a single RRAM cell can operate as an analog resistor, meaning that it can be used in more novel computing applications such as tunable synapses in neuromorphic computing. However, issues of poor uniformity and poor yield persist in RRAM. In this work, we performed a systematic study on six different TiOX/TiN RRAM samples with varying layer thicknesses. For each sample, we investigate how the layer thicknesses, presence of background oxygen, and testing parameters influence key performance metrics in each sample. These metrics include switching probabilities, failure mechanisms, operating voltage, ON/OFF ratio, device-to-device variation, and tunable resistance capabilities to determine their viability as multi-state memory cells. Our results reveal clear trade-offs between high ON/OFF ratio, high resistance tunability and high device-to-device variations. Furthermore, operating conditions within the same sample that produce large ON/OFF ratios are often not the same as those that produce low variations and high switching probabilities. The results presented in this paper provide a detailed insight into how intrinsic device properties and operating conditions influence RRAM performance and could be used for considerations in designing future RRAM technologies.