This year’s top semiconductor stories were mostly about the long and twisting trips a technology takes from idea (or even raw material) to commercial deployment. I’ve been at IEEE Spectrum long enough to have seen some of the early days of things that became commercial only this year.
In chip-making that includes the production of the next evolution of transistor design—nanosheet transistors—and the arrival of nanoimprint lithography. In optoelectronics, it was the commercialization of optical fiber links that go directly into the processor package.
Of course there were also great new technologies recently born, like growing diamond inside ICs to cool them. But there were also, unfortunately, developments that are getting in the way of moving technologies from the laboratory to the semiconductor fab.
Still, if anything, the year’s best semiconductor stories showed that technology is full of fascinating tales.
Peter Crowther
It seems one of our readers’ favorite things was this cool idea. Perhaps you read it while chilling out with a print copy of Spectrum or maybe while on your phone and icing a sore knee. (Okay. I’ll stop.) Stanford professor Srabanti Chowdhury explained how her team has come up with a way to grow diamonds inside ICs, mere nanometers from heat generating transistors. The result was radio devices that were more than 50 degrees Celsius cooler, and a pathway to integrate the highly heat-conductive material in 3D chips. The article was part of a special report on the problem of heat in computing that includes an article on cooling chips with lasers and other great reads.
Left: Stefan Ziegenbalg; Right: ASML
This one had a little bit of everything. It’s the story of how ASML figured out a key unknown in the development of one of the most crucial (and craziest) contraptions in technology today, the light source for extreme ultraviolet lithography. But it’s also a sweet story of a man and his grandfather—but with supernovas, atomic bomb blasts, high-powered lasers, and a cameo by computer pioneer John von Neumann.
Mingrui Ao, Xiucheng Zhou et al.
In past years, we’ve reported plenty about advances in making individual 2D transistors work well. But in April we delivered a story of some 2D semiconductor integration heroics. Researchers in China managed to integrate nearly 6,000 molybdenum disulfide devices to make a RISC-V processor. Amazingly, despite using just laboratory-level manufacturing, the chip’s creators achieve a 99.7 percent yield of good transistors.
Our Japan correspondent, John Boyd, described an exciting potential competitor to EUV lithography. Canon announced that it had sold the first nanoimprint lithography system for chip making. Instead of carrying the chip’s features as a pattern of light, this machine literally stamps them onto the silicon. It’s a technology that’s been decades in the making. In fact, one of my first reporting trips for IEEE Spectrum was to visit a startup using nanoimprint lithography to make specialized optics. I got in a minor car accident on my way there and never got to see the tech in person. But if you want a look, there’s one in Austin, Texas.
IEEE Spectrum; Source image: Natcast
The U.S. CHIPS and Science Act promised to be transformational—not just for chip manufacturing, but for providing R&D and infrastructure that would help close the dreaded lab-to-fab gap that captures and kills so many interesting ideas. The main vehicle for that R&D and infrastructure was the National Semiconductor Technology Center, a legally mandated, US $7.4 billion program to be administered by a public-private partnership. But the Commerce Department ended the latter entity, called Natcast, in late Summer. The vitriol with which it was done shocked many chip experts. Now Commerce has killed another CHIPS Act center, the SMART USA Institute, which was dedicated to digital twins for chip manufacturing.
The idea of bringing speedy, low-power optical interconnects all the way to the processor has fired the imagination of engineers for years. But high cost, low-reliability, and serious engineering issues have kept it from happening. This year we saw the first hint that it was really coming. Broadcom and Nvidia—separately—developed optical transceivers integrated in the same package as network switch chips, which sling data from server rack to server rack inside data centers.
IEEE Spectrum
TSMC and Intel have begun manufacturing new types of transistors, called nanosheets or gate-all-around. We got the first look at what this means for shrinking the next generation of logic chips when both companies reported details of SRAM memory for such new chips. Amazingly, both companies produced memory cells exactly as small as each other right down to the nanometer. Even more amazingly, Synopsys designed a cell using the previous generation of transistors that hit that density as well, but they didn’t perform nearly as well.
Optics Lab
My personal favorite of the year was a story I did myself as part of The Scale Issue, our October special report exploring all kinds of scale in technology. I was assigned an article with a truly global scale—tracing the 30,000 kilometer journey from quartz mine through silicon ingot to smart phone.
From Your Site Articles
Related Articles Around the Web