Machine learning models for semiconductor manufacturing processes

Abstract

This disclosure describes methods and systems for training and deploying machine learning predictive models for use in semiconductor manufacturing processes. Specifically, this disclosure provides training a machine learning predictive model for manufacturing a component using design data, process parameters, gas flow configuration from a pixelated showerhead, a temperature profile across an electrostatic chuck, and a measured uniformity profile of a processed wafer. This disclosure also provides deploying a machine learning predictive model to achieve real-time adjustments to the manufacturing process.

Claims

1. A processor implementation for processing semiconductor wafers using a trained machine learning predictive model, The steps include inputting at least one of semiconductor wafer design data or process parameters into a trained machine learning predictive model, The steps include inputting the gas flow configuration of the pixelated showerhead into the aforementioned trained machine learning prediction model, The steps include receiving the predicted uniformity profile generated from the aforementioned trained machine learning prediction model, The steps include determining whether the generated predicted uniformity profile matches the target uniformity profile, The steps include instructing the controller to process the semiconductor wafer, The steps include receiving the measured uniformity profile of the components on the processed semiconductor wafer, The steps include determining whether the measured uniformity profile is within acceptable limits, The step of determining that the processing of the semiconductor wafer is complete when it is determined that the measured uniformity profile is within the tolerance limit, Processor implementation method including

2. The method according to claim 1, wherein the process parameters include one or more of the following: the temperature of the electrostatic chuck, the processing time, the type of gas, the power, the RF bias, the pressure, or the temperature inside the processing chamber of the semiconductor wafer.

3. The method according to claim 1, wherein the gas flow configuration of the pixelated showerhead includes a distribution of gas flow in one or more zones of the pixelated showerhead.

4. The method according to claim 1, wherein the measured uniformity profile is the measured thickness of the deposited layer on the processed semiconductor wafer.

5. The method according to claim 1, wherein the measured uniformity profile is the measured limit dimensional uniformity on the processed semiconductor wafer.

6. The method according to claim 1, wherein the measured uniformity profile is received from one or more measuring tools.

7. The method according to claim 1, further comprising the step of refining at least one of the semiconductor wafer design data, the process parameters, or the gas flow configuration of the pixelated showerhead if the measured uniformity profile is determined to be outside the acceptable limits.

8. If the measured uniformity profile is determined to be outside the acceptable limit, The steps include using the trained machine learning prediction model to predict adjustments to the flow rate of at least one process gas from the pixelated showerhead, The steps include determining whether the predicted adjustment to the flow rate of the at least one process gas is feasible using the trained machine learning prediction model, If it is determined that the predicted adjustment is feasible, the steps include: reconstructing the flow rate of the at least one process gas from the pixelated showerhead; The method according to claim 1, further comprising:

9. The method according to claim 8, wherein the step of reconstructing the flow rate of the at least one process gas from the pixelated showerhead includes the step of reconstructing the flow rate of the at least one process gas from at least one zone of the pixelated showerhead.

10. If it is determined that the aforementioned predicted adjustment is not feasible, the step of generating a maintenance warning, The method according to claim 8, further comprising:

11. A processor implementation for processing semiconductor wafers using a trained machine learning predictive model, The steps include inputting at least one of semiconductor wafer design data or process parameters into a trained machine learning predictive model, The steps include inputting the configuration of the electrostatic chuck into the trained machine learning prediction model, The steps include receiving the predicted uniformity profile generated from the aforementioned trained machine learning prediction model, The steps include determining whether the generated predicted uniformity profile matches the target uniformity profile, The steps include instructing the controller to process the semiconductor wafer, The steps include receiving the measured uniformity profile of the components on the processed semiconductor wafer, The steps include determining whether the measured uniformity profile is within acceptable limits, The step of determining that the processing of the semiconductor wafer is complete when it is determined that the measured uniformity profile is within the tolerance limit, Processor implementation method including

12. The method according to claim 11, wherein the process parameters include one or more of the temperature of the electrostatic chuck, processing time, type of gas, power, RF bias, pressure, or temperature in the processing chamber of the semiconductor wafer.

13. The method according to claim 11, wherein the configuration of the electrostatic chuck includes a temperature distribution in one or more zones of the electrostatic chuck.

14. The method according to claim 11, wherein the measured uniformity profile is the measured thickness of the deposited layer on the processed semiconductor wafer.

15. The method according to claim 11, wherein the measured uniformity profile is the measured limit dimensional uniformity on the processed semiconductor wafer.

16. The method according to claim 11, wherein the measured uniformity profile is received from one or more measuring tools.

17. The method according to claim 11, further comprising the step of refining at least one of the semiconductor wafer design data, the process parameters, or the configuration of the electrostatic chuck if the measured uniformity profile is determined to be outside the acceptable limits.

18. A processor implementation for processing semiconductor wafers using a trained machine learning predictive model, The steps include inputting at least one of semiconductor wafer design data or process parameters into a trained machine learning predictive model, The steps include inputting the configurations of multiple RF field generators into the aforementioned trained machine learning prediction model, The steps include receiving the predicted uniformity profile generated from the aforementioned trained machine learning prediction model, The steps include determining whether the generated predicted uniformity profile matches the target uniformity profile, The steps include instructing the controller to process the semiconductor wafer, The steps include receiving the measured uniformity profile of the components on the processed semiconductor wafer, The steps include determining whether the measured uniformity profile is within acceptable limits, If it is determined that the measured uniformity profile is within the tolerance limit, the process of the semiconductor wafer is determined to be complete. Processor implementation method including

19. The method according to claim 18, wherein the process parameters include one or more of the following: the temperature of the electrostatic chuck, the processing time, the type of process gas, the flow rate of the process gas, the power, the RF bias, the pressure, or the temperature inside the processing chamber of the semiconductor wafer.

20. The method according to claim 18, wherein the configuration of the plurality of RF field generators includes a distribution of RF bias generated in one or more of the plurality of RF field generators.

21. The method according to claim 18, wherein the measured uniformity profile is the measured thickness of the deposited layer on the processed semiconductor wafer.

22. The method according to claim 18, wherein the measured uniformity profile is the measured limit dimensional uniformity on the processed semiconductor wafer.

23. The method according to claim 18, wherein the measured uniformity profile is received from one or more measuring tools.

24. The method of claim 18, further comprising the step of refining at least one of the semiconductor wafer design data, the process parameters, or the configuration of the plurality of RF field generators if the measured uniformity profile is determined to be outside the tolerance limit.

25. A processor implementation for processing semiconductor wafers using a trained machine learning predictive model, The steps include inputting at least one of semiconductor wafer design data or process parameters into a trained machine learning predictive model, The steps include inputting at least two of the following into the trained machine learning predictive model: the configuration of the RF field generator, the configuration of the electrostatic chuck, and the gas flow configuration of the pixelated showerhead. The steps include receiving the predicted uniformity profile generated from the aforementioned trained machine learning prediction model, The steps include determining whether the generated predicted uniformity profile matches the target uniformity profile, The steps include instructing the controller to process the semiconductor wafer, The steps include receiving the measured uniformity profile of the components on the processed semiconductor wafer, The steps include determining whether the measured uniformity profile is within acceptable limits, The step of determining that the processing of the semiconductor wafer is complete when it is determined that the measured uniformity profile is within the tolerance limit, Processor implementation method including

Images