Machine learning model for semiconductor manufacturing processes
Inventors
Huang, Zhiqiang • Tan, Li Ming • LOH, Joanna Kejun • KOENTJORO, Olivia Fatma • Lindley, Roger Alan
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12444655-B2
Publication Date
2025-10-14
Expiration Date
Abstract
The disclosure describes methods and systems for training and deploying a machine learning predictive model for use in a semiconductor manufacturing process. Specifically, the present disclosure provides for training machine learning predictive models for manufacturing components using design data, process parameters, gas flow configurations from a pixelated showerhead, temperature profile across an electrostatic chuck, and measured uniformity profiles of processed wafers. The present disclosure also provides for deploying the machine learning predictive model to effectuate real-time adjustments to a manufacturing process.
Core Innovation
The invention uses a trained machine learning predictive model in a processor-implemented method for processing a semiconductor wafer. The method inputs at least one of semiconductor wafer design data or process parameters, and also inputs a configuration such as a gas flow configuration for a pixelated showerhead, a configuration for an electrostatic chuck, or a configuration for a plurality of RF field generators. The trained machine learning predictive model generates a predicted uniformity profile that is matched to a target uniformity profile.
After directing a controller to process the semiconductor wafer, the method receives a measured uniformity of components on the processed semiconductor wafer. The method determines whether the measured uniformity is within a tolerance limit, and upon determination that the measured uniformity profile is within the tolerance limit, determines that processing of the semiconductor wafer has completed. The measured uniformity includes measured deposition layer thickness or measured critical dimension uniformity.
When the measured uniformity profile is outside the tolerance limit, the method refines at least one of semiconductor wafer design data, the process parameters, or the configuration for the electrostatic chuck or RF field generators. The disclosed approach ties the predictive uniformity profile generation to specific hardware configurations and measurable uniformity profiles.
Claims Coverage
The independent claims recite four inventive-method variants, with five inventive features across the claim sets, that combine a trained machine learning predictive model, specific input configurations, generation of a predicted uniformity profile, comparison to a target, execution of wafer processing, and termination when measured uniformity is within a tolerance limit.
Model-based predicted uniformity profile from wafer inputs and pixelated showerhead gas flow
inputting, into a trained machine learning predictive model, at least one of semiconductor wafer design data or process parameters; inputting, into the trained machine learning predictive model, a gas flow configuration for a pixelated showerhead; receiving a generated predicted uniformity profile from the trained machine learning predictive model; determining that the generated predicted uniformity profile matches a target uniformity profile.
Termination based on measured component uniformity within tolerance limit
directing a controller to process the semiconductor wafer; receiving a measured uniformity of components on the processed semiconductor wafer; determining whether the measured uniformity is within a tolerance limit; upon determination that the measured uniformity profile is within the tolerance limit, determining that processing of the semiconductor wafer has completed.
Electrostatic chuck configuration used for predicted uniformity profile
inputting, into the trained machine learning predictive model, at least one of semiconductor wafer design data or process parameters; inputting, into the trained machine learning predictive model, a configuration for an electrostatic chuck; receiving a generated predicted uniformity profile from the trained machine learning predictive model; determining that the generated predicted uniformity profile matches a target uniformity profile.
RF field generator configuration used for predicted uniformity profile
inputting, into a trained machine learning predictive model, at least one of semiconductor wafer design data or process parameters; inputting, into the trained machine learning predictive model, a configuration for a plurality of RF field generators; receiving a generated predicted uniformity profile from the trained machine learning predictive model; determining that the generated predicted uniformity profile matches a target uniformity profile.
Joint input of RF field generator, electrostatic chuck, and pixelated showerhead configurations
inputting, into a trained machine learning predictive model, at least one of semiconductor wafer design data or process parameters; inputting, into the trained machine learning predictive model, at least two of a configuration for a RF field generator, a configuration for an electrostatic chuck, and a gas flow configuration for a pixelated showerhead; receiving a generated predicted uniformity profile from the trained machine learning predictive model; determining that the generated predicted uniformity profile matches a target uniformity profile.
Overall, the independent claims cover processor-implemented methods that generate and use a predicted uniformity profile from a trained machine learning predictive model, where the predicted profile depends on semiconductor wafer design data or process parameters and one or more specific chamber-related configurations. Processing is performed by a controller and completion is determined when the measured uniformity profile is within a tolerance limit.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?