Dr. LAURA MATZ, CEO of Athinia®and CSTO of Merck KGaA, Darmstadt, Germany
As technologies such as IoT, autonomous driving and generative AI accelerate, and with more fabs being built around the world to accommodate new demand for semiconductor chips, the industry is at a critical juncture, grappling with unprecedented opportunities and challenges. Collaboration across the global ecosystem is essential for advancing the development and manufacturing of next-generation technologies. Digital twins, already successful in industries such as healthcare for drug discovery, are now being increasingly adopted in semiconductor design and manufacturing. In fact, the CHIPS and Science act funding to establish the CHIPS Manufacturing USA Institute focused on digital twins is now spurring new innovation and standards in this area.
Digital twins can gain deeper insights into performance, anticipate potential issues before they occur and improve decision-making and collaboration across the industry. By mirroring the real world in a virtual space, digital twins enable optimization of operations, predictive maintenance and innovation in product design and development.
In semiconductors, digital twins can be used to introduce a new material into a technology node or utilized to replicate that material, and device performance from one fab to another. This is becoming increasingly important when bringing fabs from Asia to the US, or even replicating at the same fab from one part of the US to another. Digital twins can transfer learnings in a digital model rather than relying on human translation before a chip is even designed. Adopting an end-to-end supply chain ecosystem digital twin could revolutionize the semiconductor industry by bolstering supply chain resilience, optimizing resource allocation, adopting more sustainable materials or practices, and empowering the semiconductor value chain to anticipate and adapt to upcoming megatrend inflections.
The challenge for the semiconductor industry is building trust to integrate data from multiple companies together or multiple process steps so that the industry can build greater knowledge, and predictability in the digital twin, for advancing a process node or bringing in a new material into a device integration. One of the most valuable aspects of the digital twin is having continuous feedback, and reinforcement to build on greater predictability of the model. The continuous integration of data feeding into that model is really powerful to make it more predictive, but also to highlight when things change, when some parameter isn’t performing the same, and when it’s not delivering the initial specifications set for the process. This continuous feedback loop is powerful to gain the most value out of the digital twin and build the most predictive model. Ultimately, the greater predictability you have, the more people will trust the model, the faster companies can accelerate timelines, and typically the closer companies can get to the optimum versus what human intuition can bring.
To move adoption forward, the industry needs a standard that can be trusted to connect the data sets that are predictive enough to enable new advancements without giving away any proprietary information. Once the model is created, it can be utilized for other processes, and for other companies.
The following are important steps semiconductor companies should embrace when adopting digital twins to realize their full benefits.
- Decide what data is important and critical to incorporate into a digital twin in order to make it predictive. The data is really the foundation because it will not only build the initial model but also reinforce and improve the prediction over time.
- Starting small and obtaining confidence around the predictive nature of a digital twin is important. Begin with a single process step and identify all the key factors that could contribute to the major output variables of that process. Integrating all the various process data including equipment, materials, process conditions, incoming materials and metrology can build an initial model which then can be improved with AI approaches.
- Integrate generative AI, and other AI algorithms such as machine learning into models to make it more predictive which will create the maximum value from the digital twin. There’s an initial inertia that it takes to get going, but once that happens, the predictive capability of a digital twin or of an algorithm, and a predictive model can not only trigger new insights, but also accelerate a company’s development timelines or transfer timelines, depending on the problem.
- Finally, be open to sharing data across the ecosystem. Participating in Athinia®’s secure data analytics platform can help by sharing, aggregating, and analyzing data to unlock efficiencies while improving quality, supply chain, and sustainability – all without taking ownership of data.
By building the confidence and taking it one step at a time while being clear on the problem to be solved, semiconductor companies can help build a predictive model that will help solve problems faster and predict in silico what would be tested in a physical environment. Achieving this will far outweigh any risks in terms of data collaboration with other partners.
AI is set to revolutionize digital twin capabilities, and it’s time for our industry to lead. We must embrace it as the driving force for transformative change, learning from the proven successes in other sectors. For example, around 25% of drugs in the FDA pipeline are now AI-recommended molecules, showcasing AI’s transformative power. We have the opportunity to deliver equally important results.
The true impact of adopting digital twins will be how we solve problems collectively and create digital twins with many partners across the ecosystem. With the pressure to bring capacity as fast as possible and with the acceleration of AI and other use cases, by 2030, it is an opportune time to enable the semiconductor ecosystem to embrace digital twins in full force and become the way the industry will operate, solving the most pressing and complex problems that individual companies cannot address as effectively on their own.
About the author
Dr. Laura Matz is CEO of Athinia®, which focuses on enabling secure data sharing within the semiconductor ecosystem. She is also the Chief Science and Technology Officer for Merck KGaA, Darmstadt, Germany, driving innovation and digitalization across the 3 business sectors, Life Sciences, Healthcare, and Electronics. Serving as an Executive Vice President within Merck KGaA, Darmstadt, Germany. She is responsible for the corporate innovation teams, including the digital office and new digital business models. Laura has over 20 years of experience in semiconductor manufacturing and a decade of experience in running semiconductor materials businesses.
