The semiconductor industry is in the final throes of its most recent cyclical downturn, but clear demand drivers on the horizon, such as 5G and autonomous driving, have created a decidedly upbeat mood at SEMI’s Strategic Materials Conference, held this week in San Jose, California. Increased connectivity in daily lives will not only dramatically boost semiconductor volumes, but the physical challenges of improving chip performance have positioned materials as the key enabling technology of the fourth industrial revolution – creating opportunities for suppliers to capture significant value.
Most speakers were quick to underscore the importance of materials innovation. According to Dave Anderson, president of SEMI Americas, “We are entering the era of the material scientist,” and the role of materials in semiconductor manufacturing “has never been more important.” Carlos Diaz, senior director, corporate research at foundry major TSMC, said that the future “belongs to new materials and processes,” while Bertrand Loy, president and CEO, Entegris, told attendees the world is on the brink of the fourth industrial revolution, where technology will be fusing “physical, digital, and biological worlds and transforming our collective lives.”
Len Jelinek, senior director/semiconductor manufacturing, IHS Markit, noted that 2019 has been a challenging year for semiconductor revenue – expectations are for a 12.5% decline YOY – but said he is not forecasting “doom and gloom” because of positive consumer demand trends beyond 2019. These include the rollout of 5G networks, internet of things (IoT), artificial intelligence (AI), and autonomous vehicles. Jelinek emphasized the foundational impact of 5G in particular. “Don’t think of 5G’s impact only in terms of handsets. It’s an enabling technology that will have broad-based impact” and will be key to creating a sustainable recovery in semiconductor demand in the second half of 2020.
The current semiconductor downturn – the industry’s 10th – was initiated by an imbalance in memory supply and demand, and the lack of resolution of trade issues between China and the US is threatening to amplify volatility. Smartphones, the number-one application for semiconductors, are currently challenged by extended replacement cycles, and total handset shipments are set for its second year of decline. “We, as consumers, are waiting for revolutionary features such as 5G speeds, biometrics, foldable handsets and AI capabilities,” Jelinek says. Recent iterations have been merely evolutionary, and premium handset costs have escalated, he adds.
Automotive electronics, which account for about 10% of global semiconductor demand, will eke out slight growth in 2019, Jelinek says. “Long-term semi component revenue growth within the Auto segment will focus on increasing content within cars supporting advanced safety features.”
During his session, Duncan Meldrum, chief economist and founder of Hilltop Economics, addressed recent threats of a recession. “Underlying economic fundamentals are strong, but we are at that point in the business cycle where it doesn’t take much to knock the economy into recession,” he says. “I am telling people to have a contingency plan in place.”
Nevertheless, Meldrum laid out reasons for optimism. Most economies have plenty of jobs, and consumers have been confident despite negative headlines. “For the average person, a tariff trade war gets to be noise. If they don’t see immediate impact, they tend to eventually discount all the headline noise. The same goes for Washington politics or Brexit.” There are no serious signs of inflation pressures in the US or other major economies, he adds.
Beyond the cycle
Longer-term, explosive growth in connected devices will create a runway for semiconductor volume growth. According to SEMI, over 30 billion devices are currently connected and another 200 million are added daily. By 2020, the number of connected devices will reach 1 trillion.
“The growth profile for industry will be very strong and a multiplicity of drivers will bring more stability to this industry,” Loy adds. “But before this future becomes a reality we have a lot of work to do.” Current chips need to be faster and cheaper. “Physical scaling is not going to get us there, we’ve hit those limits,” Loy adds. “We have to look at new architectures and materials.”
Loy called on the materials sector to need to “up our game” and spend more on R&D. “Customers want us to make our products in very tight process window and ship to control. They want extreme purity for everything. It’s a long list of to-dos and it’s going to cost us a lot,” he adds. Among the needed innovations are photoresist hard masks to hand high aspect ratio, new etch chemistries for better rates and higher selectivity, and new cleaning chemistries for high aspect ratio geometry with high selectivity.
Loy also identified contamination control as a key challenge for material suppliers. “When you think about purity and contaminants, you need to think about size, concentration levels, and classes. To optimize yields and lower wafer defectivity, our customers expect materials to be very pure and exhibit low variability.” The payoff for customers is large; a 1% yield improvement can mean $150 million in annual net profit for a leading-edge logic fab, Loy says. For a 3D NAND fab, that figure can be around $110 million per year. But these requirements are getting exponentially tighter. From 28 to 7 nm, the metal impurity concentration limit became 1,000 times lower, Loy notes.
Contamination control is even more vital when the potential impacts of latent defects – which are difficult to detect in a fab and during electrical testing – are considered, particularly in emerging applications like autonomous driving, Loy says. “The cost of yield loss is expensive, but failure in a critical optical sensor of a car could be significantly greater, in terms of recalls or even human loss of life.”
To meet tightening purity requirements, Loy recommends throwing out traditional thinking about contamination control. “In the past, we could get away with simple filtrations,” he says. “That’s no longer going to work. We need to collectively, up and down the supply chain, migrate to better filtration and purification and also rethink chemical delivery systems and packaging solutions to preserve the integrity of our products.”
Metrology will also be key, but analytical capability is lagging. “We all like to believe that we cannot control what we cannot see, but that is exactly what we have to do.”
The need for innovation is also being felt at the wafer level. Kevin Light, director, Applications Technology Americas at Siltronic Corp., said that as semiconductor markets become more diversified, silicon suppliers must recognize the distinct challenges each segment faces. Better wafer properties are required for next-generation chips, he adds. “Excessive wafer geometry can cause errors during lithography, especially when printing even smaller linewidths,” he says. The end result can be defocus and placement errors. When dealing with “More than Moore” architectures, wafer requirements are driven by other factors than defects. “More than Moore applications do not benefit from scaling, but instead drive capabilities of separate silicon parameters,” Light says. “In some cases you need high doping, in others the doping needs to be precise.”
Czochralski crystal growth is suitable for high dopant levels, but the concentrations vary at the top and bottom of the ingot. Float Zone crystals avoid oxygen incorporation and provide consistent doping. These variations make Czochralski process suitable for PowerMOS, and Float Zone appropriate for IGBT. Compound semiconductor layers, such as GaN-on-Si, offer potential advantages owing to higher switching speeds and critical breakdown fields, he adds.
“Silicon wafer requirements are diversifying as the devices themselves find increasing use outside of traditional logic,” Light adds. “Moore’s law is alive and next-gen computing will continue to push the limits of flatness and cleanliness. Meanwhile, demands of energy efficiency, electrification, IoT, and 5G drive wafer requirements other than scaling, including extremely high doped or ultra-low oxygen growing techniques, high lifetimes, and substrates engineered for compounds semiconductors.”
Autonomous driving was a frequent discussion topic at SMC. Although IHS Markit does not see it really rolling out until past 2025, the disruption to the auto industry’s status quo is very much being felt now.
Dragos Maciuca, executive technical director, Palo Alto Research and Innovation Center at Ford Motor Company, says cars of the future will be autonomous, connected, electrified, and shared. “The biggest transformation will be the shift from mechanical hardware to software,” he says. “Currently [a car] is a mechanical thing that has some electronics. Going forward, it will be a software-driven system that happens to control some mechanical elements.”
The transition is already way under way, so much so that autonomous technology developed for the automotive industry is already being spun off into other sectors, such as mining and agriculture, and the auto industry’s competitive landscape is already seeing changes. OEMs and carmakers are entering the market from the traditional auto industry side, while companies such as Google are participating from the software side. “Others, like Uber and Lyft, are coming in from the business plan point of view to eliminate drivers and improve margins,” Maciuca adds.
Autonomous driving will require numerous innovations, many of which will require new electronic materials and production processes. “We need weight savings, space savings, and advanced architecture,” Maciuca says. “We also need customization to print circuits as the vehicle comes down the line.” The tech community is proving up to the task. For LIDAR, there were just two technologies available a few years ago, he adds.
The impact on chipmakers is also already being felt. “The automotive industry used to buy older chips,” Maciuca says. “Now we are moving to a stage where we need the very first chips at the most advanced node. And we are using them for safety-critical operations. If an AI chip that is supposed to detect a human fails, the consequences can be very severe.”
Rebecca Coons is a senior editor at Chemical Week.
Republished with permission from Chemical Week.
The SEMI Electronic Materials Group (SEMI EMG) is the backbone of the Strategic Materials Conference. EMG is a technology community representing SEMI member companies that provide substrates, polymers, metals, organic and inorganic materials, chemicals, and gases that are developed or in use for the manufacturing of electronics. The group is open to SEMI Members involved in materials manufacture, distribution, and services throughout the microelectronics industry. For more details, please visit the website.