Researchers built a switch 1,000 times faster than today’s AI chips, and it barely generates any heat
Scientists have developed a novel electronic switch that operates at speeds 1,000 times faster than current technology, potentially revolutionizing the field of ultra-high-speed electronics and data processing. Collaborators from the University of Regensburg and the University of Michigan have demonstrated a switch functioning in the petahertz frequency range, which equates to a million billion electrical cycles per second. In contrast, existing transistors function in the gigahertz range, corresponding to billions of cycles per second.
Transistors, fundamental components of contemporary electronics, regulate the flow of electricity by switching on and off to encode information as binary data. Enhancing their switching speed can lead to the development of more powerful processors, faster communication systems, and accelerated memory devices. The unprecedented speed was achieved through an innovative technique involving ultrafast laser pulses precisely targeted at a specially designed material composed of layered graphene and tungsten diselenide.
The brief laser bursts triggered electrons within this material to transition between distinct valleys in its electronic band structure, effectively functioning as an on/off mechanism at speeds far exceeding those of conventional semiconductors. According to lead researcher Rupert Huber, this marks the first instance of electrically controlling such valley switching at these ultrahigh speeds, which could pave the way for a new generation of electronics working at optical, petahertz frequencies.
While this experiment serves primarily as a proof of concept, researchers acknowledge that practical application of petahertz electronics remains several years away, necessitating further advancements in materials science and device engineering. Nevertheless, this breakthrough offers an exciting glimpse into a future where computing, communications, and signal processing could be performed on remarkably brief timescales, dramatically surpassing current semiconductor capabilities. The study detailing these findings was recently published in Nature.
