The article discusses the challenges faced by researchers working with MoS2 semiconductors and their innovative approach to optimizing transistor threshold voltages. Unlike silicon, which can be adjusted by doping, MoS2 requires alternative methods such as using different metals for wiring. Researchers employed machine learning to determine the best combinations of materials and wiring, achieving high chip production yields overall. However, they encountered significant yield issues with more complex circuit components, particularly with 64-bit registers. The final chip comprised 5,900 transistors and successfully implemented the RISC-V instruction set.
The researchers investigated a novel approach to transistor threshold voltage adjustment using wiring material rather than traditional doping, overcoming challenges of MoS2 semiconductors.
Utilizing machine learning, the team optimized wiring and materials for over 99.9% chip yield, although yield for complex circuitry like 64-bit registers dropped to only 7%.
Collection
[
|
...
]