AI-PC – Controlling Manufacturing with AI

As semiconductor devices shrink and gain complexity, sources of process variance are becoming harder to control. Solving for nano-scale physics often means compensating for complex physics interactions between large numbers of physical properties. Simulating these processes can take hours or days. How can we hope to build high volume manufacturing solutions?

Artificial Intelligence Process Control (AI-PC) is both an enabling technology for the next generation of process tools, and a means to improve yield, cycle time, and cost efficiency of existing process flows.

For semiconductor equipment makers, AI-PC is a critical enabling technology, enabling next generation tools that can make sophisticated optimization adjustments on a per-wafer or even per-millisecond basis, significantly reducing process variance even in the face of increased process complexity.

For semiconductor fab operators, AI-PC will bring an improvement over classic methods for process development and Advanced Process Control (APC). Not only are AI-PC solutions able to discover sources of variance that are too complex for standard SPC charts and associated linear methods, the associated increase in automation means less people managing more equipment.

AI Process Controllers (AI-PC) have three significant advantages when controlling manufacturing processes.

First, they can control and optimize processes that are many orders of magnitude more complex than what is possible today. This can be done at millisecond speed in situations where solving the physics takes hours or is simply impractical to represent in a solvable way.

Second AI-PC controls are adaptive – they can continually update their models in response to feedback, automatically adjusting for measured and even unmeasured process disturbances. This can greatly accelerate process development, and also reduce the variance in process control once in production.

Third, they are predictive – for each control iteration, they can not only pick optimal control values, but they can also predict the outcome – in other words, virtual metrology is built-in to the controller. This virtual metrology can be used to inform control of later processing steps or to detect anomalous outcomes inline.

This talk will introduce AI-PC, its benefits, and provide example use cases of where it could benefit the industry.