IC supply crunch? Engineering simulation helps ease the pain

Increasing the semiconductor supply is a recurring theme as the world struggles to reset itself after two years battling the COVID-19 pandemic. But as viral variants continue to emerge, lockdowns persist in major Chinese manufacturing centers, and the supply chain struggles to right itself, Silicon Semiconductor asked engineering simulation leader Ansys whether multiphysics simulation can speed IC development. Ansys Director of Business Strategy Christophe Bianchi shares his insights.

BY MARK ANDREWS, TECHNICAL EDITOR, SILICON SEMICONDUCTOR

THE CHIPS for America Act. The European Chips Act. Initiatives by GlobalFoundries, Intel, Samsung, TSMC, and more – All are efforts by governments and manufacturers to grow semiconductor supplies when a lack of ICs of every type has dogged a global recovery following the calamitous years spent fighting COVID-19.

How are we doing? According to most measures, the semiconductor shortages faced by automotive, consumer electronics and commercial manufacturers are likely to continue through 2022 and into the first quarters of 2023. Shortages persist despite the fact that SEMI, the SIA and other global trade groups report that there are 60 semiconductor fabs under construction and that manufacturers have already increased production.

The what’s and why’s behind a persistent shortage of semiconductors have already been explored in this business magazine and many others. Essentially, getting more devices into the hands of global OEMs and contract manufacturers is going to take time. Even though the extra capacity coming online later in 2022 through 2024 will no doubt ease the crunch, can anything be done now that has not already been tried?

One place Silicon Semiconductor sought answers was with the engineering simulation experts at Ansys, a leader in the field that is constantly expanding its portfolio of simulation services that embraces not just the ICs that fabs manufacture, but also the in-plant processes of multiple industries.

Engineering simulation has come a long ways since Ansys helped create the field 50 years ago. Today, engineering simulation offers the chance to test components in a virtual environment across thousands of scenarios that will materially affect device functionality, lifetime and adherence to specifications, amongst many parameters. Ansys reports that H3C Semiconductor, for example, uses the company’s multiphysics platform to engineer an advanced processor chip for cybersecurity, AI and 5G backhaul applications, effectively improving product sign-off efficiency and accelerating product development. Through simulation, it is possible to ensure that, once in service, the chips meet the reliability, performance, and longevity requirements of their respective applications, Ansys stated. According to Ansys Director of Business Strategy, Christophe Bianchi, engineering simulation is already at work to improve quality and speed production by guiding designers away from likely-to-fail scenarios while speeding design by automatically applying years’ worth of computational experience to test and verify myriad design options, materials parameters and other key metrics.

MA: Is highly accurate engineering simulation best suited for the design phase of a new product, or can it benefit other key aspects of IC manufacturing to help close the gap between supply and demand?

CB: Engineering simulation is pervasive across the entire engineering practice. Of course, the design phase is where trade-offs are made based on complex simulation, for example, thermal, mechanical, electromagnetic or Infrared. But simulation is also used to develop and optimise manufacturing processes, such as the Chemical-Mechanical Polishing (CMP) process, or tuning the performance of wafer handling in the fab. What we see becoming a critical use of simulation, combined with AI/ML technology, is the advent of digital twins of various manufacturing equipment (such as UV lithography tools). Here, the twin, running in the cloud or on the edge, provides invaluable information to pilot predictive maintenance and therefore optimise both performance and yield of these complex manufacturing processes.

MA: How do today’s most advanced engineering simulation programmes differ from previous generations?

CB: Although performance and capacity remain two critical development threads in engineering simulation, for which the advent of distributed computing has provided a boost in productivity, there are two major paradigm shifts that new generations of engineering simulation impact design methodologies and product developments: Firstly, the multiphysics aspect of simulation becomes even more critical at advanced semiconductor nodes. With design margins shrinking even further, and complexity of products reaching new heights, it is now impossible to treat each phenomenon independently. The combined and interdependent effects of thermal, electromagnetic, radiation, mechanical stress, and power distribution variability must be assessed jointly in a true multiphysics way.

Ansys multiphysics simulation can speed the design process across device types while enabling faster review and approval cycles between team members

Another major shift is the advent of AI/ML methods assisting the designers and developers throughout their engineering tasks. With the increased complexity of today’s most advanced engineering simulations, designers must explore much more complex solution spaces, and the help provided by AI/ML facilitates those efforts by assessing parameter sensibility, and design space optimisation using automatically trained reduced-order models for faster convergence.

MA: Is it possible to estimate, if looking at a hypothetical product, how much time engineering simulation could save compared to developing said product without it or with outmoded tools?

CB: I don’t think it is a question of saving time when the alternative to simulation is prototyping. It is not just the exorbitant cost of prototyping (reaching above $1 million in advanced semiconductor nodes), but the sheer fact that the chances of silicon success without simulation is close to zero, and that is what makes engineering simulation an essential part of the design cycle. I cannot think of a situation where a design team would risk developing a product without simulation for the sake of development time.

MA: One challenge semiconductor manufacturers faced during 2020-21 that persists is addressing demand in a more agile fashion, such as rapidly changing what is being manufactured. Can engineering simulation help manufacturers rapidly retool?

CB: The IC supply crisis the semiconductor industry is currently facing is more a capacity than an agility problem. As the most affected industry by this problem, the automotive sector did not foresee the impact of “just in time” inventory practices when facing fierce competition from computing and communication for the same manufacturing capacity. These industries rebounded faster during the pandemic and secured the wafer supplies, leaving no room for peak demand from auto Tier 1s and OEMs. Running at (or slightly above) full capacity, further agility would not really solve the IC shortage crisis.

On the other hand, leveraging engineering simulation to tune and improve manufacturing yield has the potential to increase production throughout, without the delay and investment required for additional capacity. This is an area where several initiatives have been launched since the beginning of the crisis, ranging from modelling and simulation of process variability to digital twins of complete manufacturing toolchains.

Ansys multiphysics simulation can speed the design process across device types while enabling faster review and approval cycles between team members

MA: If better engineering simulation can indeed shorten lead times in some circumstances, are there other factors manufacturers ought to consider to address rapidly changing customer needs?

CB: In the automotive sector, the shortage of devices resulting from a combination of ultrafast demand change and a “just in time” inventory strategy have been exaggerated by the extremely large number of different components required to equip today’s cars. Calls for more standardisation and a move towards software-defined systems (using less and more standardised programmable devices) are two of the ways we are seeing improvements. But developing such multi-purpose devices brings an order of magnitude in the design complexity and requires significantly more simulation in the development and design phase.

MA: Engineering simulation has grown more sophisticated and capable. Can you please describe key breakthroughs or developmental milestones that enabled this progress?

CB: Ansys has been refining the art of engineering simulation for the past 50 years and is not planning on stopping. Our history of acquisitions demonstrates the will to combine more and more laws of nature in our multi-physics solutions: for instance, we recently added photonics and optical simulation.

The main research and development axes that have increased the capabilities and sophistication of our solutions range from continuous research in numerical methods (For example, iterative solver methods, explicit/implicit/hybrid/bayesian computational model) to AI/ML improvement of the parametrisation of our solvers, high performance computing (task-based, shared memory, message passing, fine grain (GPU) and the future use of exascale and quantum computing) as well as platforms, workflows and data management - all targeting a more pervasive use of the cloud.