The chip manufacturing industry is under pressure to continuously increase computing power, shrink chip sizes, and manage power in these packaged circuits. Silicon is widely used because it can be produced in large quantities and still remains pure. However, silicon can only be so thin, and its material properties are limited to three dimensions.
Now a team at the University of Pennsylvania's School of Engineering and Applied Science has developed a technology that uses a two-dimensional semiconductor that is so thin that the material will sit on top of the silicon as a thin film.
The team has developed a technology that deposits indium selenide (Inse) at low temperatures, integrates it with silicon chips, and grows it into full-size, industrial-scale wafers. Wafer-level production is critical to the viability of the material. The more chips in a batch, the lower the cost. The ability of the Inse to carry an electric charge is excellent, but it is tricky to make it into a film large enough.
Postdoc Seunguk Song overcame this shortcoming by applying a growth technique. Seunguk Song and Deep Jariwala, an associate professor in the Department of Electrical and Systems Engineering (ESE) at the University of Pennsylvania, led the study, whose findings were published in the journal Matter.
***vecteezy
The two-dimensional materials they study are a broad class of crystals and compounds that are structurally (and in some cases chemically) stable at very small thicknesses under environmental conditions. This means being able to produce thin films with a thickness of only one to three atoms, which are electronically active and, in many cases, have far more electronic and optical properties than 3D silicon.
The transition from silicon to 2D materials should have little impact on the manufacturing process, and if the impact is too great, the cost-benefit analysis becomes more difficult. If this happens, then such chips will occupy more niches, such as defense, space, etc.
Artificial intelligence is the use of massive amounts of data for large-scale computing. This is where energy efficiency really matters for AI hardware. While silicon hardware is currently improving in terms of energy efficiency, the progress is mainly focused on the architecture level. Advanced technology at the equipment level is exhausted, and the basic limits of calculating energy efficiency are also reached. Then left.
Three options: change the material from silicon to something else, such as a two-dimensional material (which is the strongest contender), or change the physics of the device's operation, or both. For this reason, it is imperative to manufacture chips with new, more energy-efficient materials, which are conducive to artificial intelligence and big data computing. It remains to be seen how much of the business impact and overall energy consumption of AI use will have.