Engineers face big challenges delivering foldable smartphones, rollable displays and next generation DRAMs. However, they also have opportunities to deliver new classes of health care devices and 3D chip stacks.
HALF MOON BAY, Calif — Engineers face big challenges delivering foldable smartphones, rollable displays and next generation DRAMs. However, they also have opportunities to deliver new classes of health care devices and 3D chip stacks.
Those were some of the highlights of a day of technology talks at the annual Industry Strategy Symposium hosted by the Semi trade group here earlier this month.
Jo de Boeck, chief strategy officer at the Imec, kicked off the day with one of the research institute’s signature whirlwind tours of semiconductor technology and applications. It ranged from an update on extreme ultraviolet lithography to work on superconducting qubits.
One of the more mind-bending application areas was a lab-on-a-chip geared to “reprogram” human tissues for medical research or therapy. He described a multi-level array chip geared to capture individual cells. Targeted cells could then be hit with an electric charge so they could ingest a chemical through microfluidic channels.
A lab-on-a-chip array with microfluidics could “reprogram” human tissues. (Source: Imec)
Foldable smartphones require unfolding fat wallets
In displays, foldable phones and rollable OLED TVs are the buzz of the consumer electronics crowd, but don’t expect affordable products anytime soon, said Brian Shieh, general manager of display technology at Applied Materials.
In 2019, foldable handsets such as those promised by Samsung last year will cost as much as $1,800. Consumers may feel cheated by their polymer surfaces that feel cheap and will be subject to scratches and cracks. Researchers are exploring materials that are tougher yet still flexible, but no solutions are on the near horizon, said Shieh.
Meanwhile OLED displays face in a five-year development effort for versions that could meet the cost ceiling of $1,000 for a typical high-end TV. Today’s mobile OLED displays are made in vacuum chambers the size of football fields that handle a dozen deposition layers with overlay accuracy measured in microns, he said.
Given their high costs, OLED TVs will remain an expensive product niche for years. The good news is OLEDs are now used in about 40% of smartphones and will come to dominate handsets over the next few years.
“It will take a lot of work to make OLEDs cost competitive with LCDs for TVs,” he said, describing early work using inkjet printing and other alternative techniques.
Micro LED displays promise to outperform OLEDs with lower energy, longer life, faster refresh and broader color gamut. But they are even more costly, starting their commercial life in large-scale digital signs before reaching TV sizes and long term, perhaps, smartphones.
Pros and cons of current and future display technologies. Click to enlarge. (Source: Yole Development)
Preventing mainstream DRAMs from falling over
Mainstream DRAM technology is approaching scaling limits in several ways, saidAki Sekiguchi, deputy general manager of the corporate innovation division of Tokyo Electron Ltd. (TEL).
Tall capacitor structures already require atomic deposition processes. Drying these thin columns after a wet etch can cause them to collapse from uneven surface tensions. TEL is already putting into production fabs a so-called supercritical drying approach that removes surface tensions.
Looking ahead, chip features are “getting down to tens of atoms, and controlling them to 1% tolerances is hard…so process technology is getting stressed and increasingly different to scale,” Sekiguchi said.
It’s not clear if engineers can find a kind of 3D design for DRAM that has helped NAND flash and logic cells scale. However, he noted many kinds of memories with different speed and capacity characteristics continue to emerge.
“We are very close to inflection points in the next few years,” said Buddy Nicoson, manager of global front-end operations for Micron in a separate talk. Micron is doing “a lot of work in new architectures" for the DRAM first invented in 1968, he said.
Technical challenges facing the 50-year old DRAM. (Source: TEL)
DRAMs gearing up for 12-high stacks
One of the ways DRAMs are going 3D is with chip-level stacks. To date, the High Bandwidth Memory (HBM) approach has been limited to four chips in a stack, but a new Jedec standard will open the door to towers of up to a dozen chips.
An interconnect specialist from Xperi described its approach that in the lab enabled a 15-chip DRAM stack. Xperi’s wafer-to-wafer approach makes strong connections at sizes approaching a single micron.
The so-called Hybrid Bond approach has been used in chips stacks for CMOS imagers for a decade. Besides HBM DRAMs it is also being explored for use in 3D NAND stacks and to create global shutters for more accurate lidar, said Paul Enquist, a vice president of 3D and R&D at Xperi.
A concept for a 15-layer HBM stack of DRAM chips. (Source: Xperi)
Quantum researchers need help in the fridge
Researchers are still three to five years away from identifying useful applications for quantum computers, according to a representative of Google’s effort. The search giant is now testing its second-generation device, called Bristlecone, which contains a qubit chip linked to a control chip that reads out the qubit’s results.
“We are looking for where it does better than classical computers in time, power or other dimensions,” what’s known as quantum supremacy, said Eric Ostby, a lead hardware developer in the search giant’s Santa Barbara lab. The group also wants to find ways to use quantum computers to “drastically accelerate computations for AI,” but that will take a long time, he said.
Google currently uses the Pleiades supercomputer at NASA’s Ames lab to check Bristlecone’s work. In the near future, it will get time on the newer Summit system, however even a traditional supercomputer can be overtaxed simulating a system with more than 50 qubits, he said.
A near term goal is to try to apply traditional error corrections schemes to quantum systems because their results are inherently error-prone. Ultimately quantum systems are only expected to be good for certain kinds of computations, so they are likely to act as specialized accelerators inside traditional data centers.
One early expected applications is for quantum simulations, the focus of 30% of the work of today’s supercomputers. Quantum systems also could speed up discovery of new drugs.
Ostby called on today’s capital equipment suppliers to help in the effort. “Each qubit uses a number of readout wires. We’d like to reduce these…Going to a thousand qubits-plus needs better fabrication. Lots of chips and other devices are needed to handle quantum signals and we would like more of them to be in the fridge,” he said.
The quantum roadmap starts with a search for supremacy, then moves to a hunt for applications.(Source: Google)
— Rick Merritt, Silicon Valley Bureau Chief, EE Times Circle me on Google+