Final Industry Thoughts on Full-Nodes, Inter-Nodes, Leading-Nodes and Trailing-Nodes – Part 3
In part two of this blog series, I provided broader industry context about AI and IoT, and introduced “multicolor” as a 3D technique. I showed that 3D techniques have extended the roadmap in NAND as we now define density not in terms of memory cells per mm2, but in cells per mm3. I described that similar thinking is coming to logic, with transistor features moving vertically into successive layers of the chip. In the final piece of this blog series, I’ll discuss an idea called “integrated materials systems” that Applied’s Prabu Raja introduced at the Industry Strategy Symposium (ISS) panel. First, a brief history of the evolution of materials engineering and a little bit about Prabu’s background.
A Brief History: From Unit Processes and Systems to Integrated Processes and Systems
In the early days of Moore’s Law (Era One), wafers moved from one unit process system to the next, with each individual tool duly performing its task (e.g. deposition, etch, implant and anneal), one material at a time. At 90nm (Era Two), materials engineering became so demanding that chambers for two critical steps – pre-clean and selective epitaxy – were combined into the same platform. A wafer could move from one chamber to the next within the same platform, remaining under vacuum and free of atmospheric contaminants.
Some Background on Prabu Raja
Prabu Raja once led Applied’s PVD business, helping the company build one of the most successful product positions in the industry. In 2012, he took over the etch business and led the development of the Applied Centris™ Sym3™ Etch system, Applied’s fastest-ramping product ever. By the time of the ISS panel, Prabu was responsible for the entire semiconductor equipment business at Applied Materials. The company now develops new products in teams that combine experts in chemistry as well as physics, mechanical engineering and software, including AI. As Moore’s Law slows, such an interdisciplinary approach will be needed to successfully identify, integrate and optimize more materials from the periodic table. The future will be about combining multiple materials in entirely new ways, in an approach Prabu calls “integrated materials systems.”
Armed with this background information, I will now summarize what Prabu said at the ISS panel.
From Bulk Materials to Integrated Materials
The easy days of working with individual materials from the periodic table are over. Continued advances in performance, power and area/cost (PPAC) require “integrated materials” that work together to form novel structures with precise electrical properties. Today, we are working with ten different types of silicon dioxide and eight different types of silicon nitride. At the ultra-nanoscale, materials behave differently: surface properties now dominate the bulk properties we grew up with. Properties such as Surface tension, Capillary action, Wall and Grain boundary scattering all come into play. We need to control and manipulate atoms – in how we assemble them, disassemble them and modify them. Crystal orientation and doping uniformity (of adding measured amounts of one material into another) also become important. Finally, when a group of such nanomaterials are stacked together in one place, the interfaces between them will have a big impact on device performance and need to be engineered carefully at the atomic level.
From Integrated Process Systems to Integrated Materials Systems
Assembling these integrated materials at the atomic scale to achieve predictable properties and interactions is an increasingly complex task that requires multiple process steps to be carried out in a precise arrangement, often within the same platform, and sometimes within a single chamber. Engineering materials in vacuum permits precise engineering of the material-to-material interfaces.
Integrated materials systems will allow designers to make devices they never dreamed possible. Selective processing is a new type of capability where we can selectively deposit or selectively remove only those materials that we want and only where we want them. Etching expands from conventional plasma-based to chemistry-based, offering lateral and even non-line-of-sight removals; deposition and etch will be combined in a single platform; implant will be used for shaping; e-beam inspection will create analytics that can influence the process flow to improve outcomes; and AI will be used to identify unseen relationships between process knobs and outcomes.
New Devices, New Materials, New Collaboration
As I wrote in part two of this blog series, data growth is accelerating just as Moore’s Law is slowing. To realize the promise of the Internet of Things, Big Data and AI, the industry needs new system architectures, new devices and new materials and materials engineering. The industry has an opportunity to collaborate at all levels to speed new materials to market. System architects are looking to break the “memory wall” that forms a bottleneck between processors and storage. New memories like MRAM, PCRAM and Intel® 3D XPoint™ technology aim to improve the connection between logic and memory – while neuromorphic approaches seek to merge the two. These new devices are based on complex multi-materials stacks that need to be precisely deposited and removed to maintain interface quality.
Given the complexity, progress depends on a new level of industry collaboration. Customers are asking for earlier and deeper collaboration: the integrated materials systems era is here. At Applied Materials, we look forward to a renaissance in innovation, from new materials to devices, design techniques and systems.
Intel and 3D XPoint are trademarks of Intel Corporation or its subsidiaries in the U.S. and/or other countries.