.png){kind=icon}
.png){kind=icon}
Return
Chalcogenide Selector-Only Memory
.png){kind=icon}
PRE
AI
DOI：10.1016/j.mattod.2026.103429.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The explosive growth of data-centric artificial intelligence has exposed fundamental limitations of the conventional memory hierarchy, intensifying the demand for storage class memory (SCM) technologies that simultaneously offer low latency, high endurance, and scalable density. Selector-only memory (SOM), also referred to as self-selecting memory, has recently emerged as a disruptive chalcogenide-based paradigm by integrating selector and memory functionalities within a single amorphous material, thereby simplifying the device architecture of conventional one-selector-one-memory (1S1R) designs. In this Review, we present a comprehensive overview of chalcogenide-based SOM, tracing its evolution from ovonic threshold switch (OTS) to SOM chips. We systematically examine the materials landscape of SOM, spanning tellurium-, selenium-, and sulfur-based systems, and critically compare their trade-offs in switching speed, power consumption, memory window, voltage drift, endurance, and thermal stability. Beyond single-layer compositions, recent advances in multi-chalcogenide stacking and heterostructure engineering are highlighted as effective strategies to decouple switching dynamics from structural relaxation, enabling large memory window and ultra-low drift simultaneously. We further discuss the competing physical mechanisms proposed for polarity-dependent threshold modulation, including atomic segregation mechanism, graded band-gap model, anisotropic structural unit model and interfacial band-bending model, and assess their implications for device reliability and scaling. Finally, we outline key challenges and opportunities for SOM in three-dimensional vertical integration, multilevel operation, and emerging system architectures such as compute express link (CXL)-based memory expansion and neuromorphic computing. By bridging materials science, device physics, and system-level considerations, this Review positions SOM as a promising platform for next-generation memory-centric computing.
Keywords:
Selector-only memory
Ovonic threshold switch
Chalcogenide
Phase change memory
Neuromorphic computing
Journal
M
IF:
22
Papers:
371
Citations:
0
View more organizations