Virtual Twins: Layers Of Challenges

In this interview, David Fried, corporate vice president at Lam Research, discusses the concept of virtual twins in semiconductor manufacturing. He distinguishes virtual twins from digital twins, emphasizing that virtual twins incorporate governing physical constraints and are validated with data. Using the analogy of a city, he explains that a virtual twin consists of multiple layers, each tailored to specific applications (e.g., buildings for drone navigation, roads for driving, subways for public transit). These layers must be connected to a common ground truth to ensure consistency. In semiconductor equipment, similar layers exist: mechanical, process, and sustainability twins. The challenge lies in integrating these independently developed layers, which often have different data structures and governing equations. Fried highlights the need for standards to facilitate integration, drawing parallels to EDA standards like PDKs. He also mentions federated learning and cross-industry examples (aerospace, banking) as models for sharing knowledge across layers. The complexity varies greatly between layers, from simple geometry to computationally intensive plasma physics. Virtual twins enable trade-off analysis across disciplines, such as balancing process performance with sustainability. They also accelerate engineering productivity and are key to achieving autonomous ’lights-out’ fabs. The conversation concludes with the idea that virtual twins continuously evolve, absorbing new data and capabilities.