Extreme ultraviolet (EUV) lithography is finally here. How far will it go?
Explore
Extreme ultraviolet (EUV) lithography is finally here. How far will it go?
In late 2017, IBM successfully launched the next era of high-performance cognitive and AI hardware with the deployment of its POWER9 technology in the Summit and Sierra supercomputers, the most powerful supercomputers in the world to date. Not to be outdone, IBM also launched the next era of constant-encryption mainframes, with the popular release of the new z14 enterprise mainframe for the new digital economy.
Continue reading
Joint program accelerates EUV mask development for advanced nodes
Explore
Joint program accelerates EUV mask development for advanced nodes
For more than a decade, IBM has firmly believed that one of the keys to true transistor scaling lies in the adoption of Extreme Ultraviolet (EUV) lithography as a critical patterning technique. This belief culminated in 2015 with IBM’s announcement of the world’s first 7-nanometer-node chip, and again in 2017 with the announcement of a 5nm node chip. In a single exposure step, EUV can create a high-resolution pattern of tens of billions of transistors with one “mask,” which today’s more common semiconductor technology can only attain with patterning processes with multiple masks and prohibitive complexity. Thus, EUV has been critical in enabling scaling for leading-edge logic applications at advanced nodes – like 7nm and 5nm. However, almost every aspect of EUV technology presents new challenges that have to be dealt with in a systematic, collaborative way, and the mask is a critical enabling element.
IBM Research and Photronics, Inc. recently signed a joint research and development agreement to develop manufacturing-grade EUV mask processes for leading-edge logic applications for 7nm, 5nm nodes, and beyond.
Figure 1. A state-of-the-art EUV mask
High-fidelity challenges, high-fidelity results
Masks are designed templates that project or reflect light in a pre-defined pattern onto a silicon wafer, which is used in conjunction with a photo-sensitive film on the wafer to define the circuit designs necessary to build a semiconductor device. EUV operates outside the visible spectrum at 13.5nm wavelength, much shorter than the deep-UV 193nm light of prior technologies. Because of this, much higher fidelity patterns can be achieved, albeit with new mask materials required to capture and create contrast in EUV images. Much like a high-definition television allows you to see many more details in the picture, blemishes and imperfections are also more noticeable with EUV lithography, and they have a disproportionate impact on the ability of an EUV mask to uniformly yield the tens of billions of transistors designed therein.
“Photronics’ industry-proven mask making capability and development team in the US, together with IBM’s advanced research infrastructure and semiconductor technology, form the ideal combination to accelerate industry-enabling solutions for EUV.”
IBM and Photronics aim to address the specific defect challenges of yielding EUV masks, namely achieving the required pattern fidelity and tolerances to yield 7nm and 5nm devices with EUV, and qualify mask repair processes that will correct critical defects captured in the mask making process. Through this collaborative effort, IBM has already been able to demonstrate yielding baselines for its 7nm and 5nm logic processes at its semiconductor research fab in Albany, NY, and continues to enable critical process and device learning at advanced nodes.
There is no doubt that further refinements in EUV mask architecture and processing will be required to achieve the necessary contrast and pattern fidelity for future scaling. With that goal in mind, IBM and Photronics will use the multi-year collaboration to define and demonstrate the roadmap requirements for EUV masks – a critical element to IBM’s semiconductor research for the next decade.
In-memory computing with 1 million devices for applications in AI
Explore
IBM Scientists Demonstrate In-memory Computing with 1 Million Devices for Applications in AI
IBM Research scientists demonstrated that an unsupervised machine-learning algorithm, running on one million phase change memory (PCM) devices, successfully found temporal correlations in unknown data streams. When compared to state-of-the-art classical computers, this prototype technology is expected to yield 200x improvements in both speed and energy efficiency, making it highly suitable for enabling ultra-dense, low-power, and massively-parallel computing systems for applications in AI.
Learn more
IBM Research Alliance builds new transistor for 5 nanometer (nm) technology
Explore
New transistor for 5nm technology
IBM, its Research Alliance partners GLOBALFOUNDRIES and Samsung, and equipment suppliers have developed an industry-first process to build silicon nanosheet transistors that will enable 5 nanometer (nm) chips.
Learn more
Press release
Cognitive chip manufacturing
Explore
Cognitive chip manufacturing
IBM and Tokyo Electron Limited (TEL) recently signed a Joint Development Program to add cognitive capabilities to TEL’s chip manufacturing equipment to track down anomalies – and keep their equipment running perfectly.
Learn more
Carbon nanotubes
Even with the advancement of 10, 7 and 5nm chips, the industry is preparing itself for life after silicon. Carbon nanotubes hold promise as the next material for the next generation of computers. IBM researchers developed a technique that could make the technology more viable.
Learn more
The technologies that will deliver AI
Explore
The era of AI — and the technologies that will deliver it
Mukesh Khare, IBM Research vice president of semiconductor research, explores the answers to the questions -- what does AI and the explosion of data imply for computing devices? And where should AI data be processed – the cloud or the edge?
Learn more
Research empowered by design automation
Explore
Future semiconductor research leadership empowered by design automation
IBM Research and Synopsys have entered into a joint development agreement (JDA) to develop new state-of-the-art Electronic Design Automation (EDA) flows. The goal: use software to holistically automate the method, evaluation, and assessment – the flow – of the elements that go into today’s latest transistors, like our 3D FINFET 7nm chip, and beyond. We’ll be able to know about, and act upon nano-scale variables in design, materials, and processing techniques across a chip’s entire design – preventing costly errors, while optimizing which technology path to pursue. And all of this analysis will be done before a single wafer is made.
Learn more
One big wire change in ‘97 still helping chips achieve tiny scale
Explore
One big wire change in ‘97 still helping chips achieve tiny scale
The IT state-of-the-art 20 years ago was rapidly ending, and we – and I mean everyone in the industry that made a device with a chip inside of it – needed something new to keep up with the demand for ever-faster, better electronics. These were the days of laptops with 233 MHz speeds, and Deep Blue was exploring a mere 200 million possible chess positions per second. Without one element in the eleventh group of the periodic table, Cu, our computers and devices would not have advanced much beyond the speed and power of two decades ago.
Learn more