Artificial intelligence (AI) has begun to impact our way of life, with new applications in medicine, sales, finance and more hitting the market and leading companies such as Intel, AMD, IBM, Microsoft, Meta and Google trying to find success producing, enabling or using this new technology. AI requires powerful chips with stringent specifications that allow faster model training and reduced power consumption. Heterogeneous integration is one of the key ways to meet the strict requirements that are needed for AI products, and fan-out panel-level packaging (FOPLP) is one of the best options for meeting heterogeneous integration requirements. FOPLP also offers lower cost advantages.
While FOPLP has many advantages, it faces significant process challenges, such as die placement errors and substrate warpage control. One of the key challenges is the trade-off between overlay yield and throughput during the lithography processing steps. A user exposes multiple dies per exposure shot to increase throughput, but this can result in lower overlay yield because of pick-and-place die placement errors. To overcome this low-yield issue, each die needs to be properly aligned; unfortunately, this impacts throughput. So, a compromise is required. Finding the right balance between throughput and overlay is one of the biggest challenges for FOPLP.
In this paper we will address the tradeoff between lithography throughput and overlay yield and demonstrate an integrated feedforward adaptive shot solution. This feedforward approach uses a third-party metrology system to measure reconstituted panel die location data and then sends the data to the stepper via a network. With feedforward algorithm technology, the stepper uses smart adaptive shot technology to generate an optimized variable shot size layout. This layout helps ensure that overlay yield is within specification and does so with the minimum number of exposure steps. With feedforward adaptive shot technology, the user can maximize the throughput of the stepper and allow for better overlay yield at the same time.