KR-102963240-B1 - Acceleration of preventive maintenance recovery and recipe optimization using machine learning-based algorithms

KR102963240B1KR 102963240 B1KR102963240 B1KR 102963240B1KR-102963240-B1

Abstract

A method for determining processing chamber conditions using sensor data and a machine learning model is provided. The method includes the step of receiving sensor data by a processing device, the sensor data including chamber data indicating the state of the environment of a processing chamber processing a substrate according to a set of process parameters of a current process. The sensor data further includes spectral data indicating optical emission spectra (OES) measurements of a plasma placed within the processing chamber. The method further includes the step of using the sensor data as input to a machine learning model and the step of obtaining one or more outputs indicating one or more chamber condition metrics. The method further includes the step of determining the recovery state of the processing chamber based on one or more chamber condition metrics. The method further includes the step of causing a modification to the performance of the processing chamber based on the recovery state of the processing chamber.

Inventors

한, 펭유

Assignees

어플라이드 머티어리얼스, 인코포레이티드

Dates

Publication Date: 20260508
Application Date: 20221118
Priority Date: 20211123

Claims (20)

As a method, A step of receiving sensor data by a processing device, comprising: (i) chamber data indicating the state of the environment of a processing chamber processing a substrate according to a set of process parameters of a current process; and (ii) spectrum data indicating optical emission spectra (OES) measurements of a plasma placed within a processing chamber processing a substrate according to a set of process parameters of a current process; A step of using the sensor data as input to a machine learning model by the processing device; A step of obtaining one or more outputs of the machine learning model by the processing device — the one or more outputs represent one or more chamber condition metrics —; A step of determining a recovery state of the processing chamber based on one or more chamber condition metrics by the processing device — said recovery state corresponds to a chamber recovery process performed following a preventive maintenance procedure —; and The processing device comprises the steps of: (a) determining that a first confidence level in which the sensor data accurately indicates the conditions of the processing chamber satisfies a critical condition; and (b) causing a modification to the performance of the processing chamber based on the recovery state of the processing chamber. method.
In Article 1, The method further includes the step of determining an update for at least one process parameter among the set of process parameters to generate an updated set of process parameters based on the one or more chamber condition metrics by the processing device. Modification of the performance of the processing chamber is additionally based on an update to at least one process parameter among the set of process parameters. method.
In Article 2, The step causing a modification to the performance of the above processing chamber is: Making the above substrate or new substrate to be processed according to an updated set of the above process parameters; or Stopping the substrate processing within the processing chamber A step further comprising transmitting a first command that causes at least one of method.
In Article 1, It further includes a step of displaying a notification on a graphical user interface (GUI), The above notice indicates a modification to the performance of the processing chamber, method.
In Article 1, Modifications to the above performance correspond to the chamber seasoning procedure, method.
In Article 1, The above spectrum data is: Further comprising light reflection spectrum measurements corresponding to the reflection pattern of light reflected from the surface of a substrate disposed within the processing chamber, method.
In Article 6, A step of determining one or more spectral features based on combinations of the light emission spectrum measurements and the light reflection spectrum measurements to generate feature data; and A method further comprising the step of using the feature data as input to the machine learning model. method.
In Article 1, Determining that the first confidence level that the sensor data accurately indicates the conditions of the processing chamber satisfies the critical condition is: A step of using the sensor data as input to a statistical model; A step of receiving one or more outputs from the statistical model — said one or more outputs indicate a second confidence level that temporally associated data points of said chamber data and said spectrum data accurately represent the conditions of said processing chamber, and said statistical model is generated using regression between historical chamber data and historical spectrum data —; and A step comprising comparing the second confidence level with one or more critical conditions including the critical condition, method.
In Article 1, One or more outputs of the above machine learning model are based on the spectrum data and the chamber data, and The above machine learning model is trained to predict one or more chamber condition metrics based on the correlation between past OES measurements of plasma and past process result data, method.
As a method for training a machine learning model to determine the state of a processing chamber in a chamber recovery procedure, The above processing chamber processes the current substrate according to the current process, and The above method is: The step of generating training data for the above machine learning model ― The step of generating the above training data is: i) identifying a first training input having past sensor data including past chamber data indicating the state of the environment of a second processing chamber processing a previous substrate according to a previous process, and ii) past spectrum data indicating light emission spectrum (OES) measurements of a previous plasma placed within the second processing chamber processing the previous substrate according to the previous process; The method includes the step of identifying a first target output for the first training input, wherein the first target output comprises past process result data having process result measurements of the previous substrate processed using the second processing chamber according to the previous process —; and (i) providing training data to train the machine learning model for a set of training inputs including the first training input and (ii) a set of target outputs including the first target output, and The above-mentioned trained machine learning model is intended to receive a new input having new sensor data including i) new chamber data indicating a new state of a new environment of a new processing chamber processing a new substrate according to a new process, and ii) new spectrum data indicating optical emission spectrum (OES) measurements of a new plasma placed within the new processing chamber processing the new substrate according to the new process, and to generate a new output based on said new input. The above new output displays chamber condition metrics corresponding to the recovery status associated with the chamber recovery process performed following preventive maintenance procedures, Method for training a machine learning model.
In Article 10, The above past spectrum data is: Further comprising light reflection spectrum measurements corresponding to the reflection pattern of light reflected from the surface of the previous substrate disposed within the second processing chamber, Method for training a machine learning model.
In Article 11, The method further includes the step of determining one or more spectral features based on combinations of light emission spectrum measurements and light reflection spectrum measurements to generate feature data, The above training data further includes the above feature data, Method for training a machine learning model.
In Article 10, A method further comprising the step of performing a data extrapolation procedure with the past spectrum data to generate OES estimates corresponding to time instances occurring before or after the above OES (optical emission spectra) measurements. Method for training a machine learning model.
In Article 10, A method further comprising the step of performing a data interpolation procedure with the past spectrum data to generate OES estimates for one or more time instances occurring between pairs of the above OES (optical emission spectra) measurements. Method for training a machine learning model.
In Article 10, Each training input within the above set of training inputs is mapped to a target output within the above set of target outputs, Method for training a machine learning model.
In Article 10, The above-mentioned trained machine learning model comprises at least one of a logistic regression type algorithm, a multilayer perception algorithm, or a recurrent neural network (RNN). Method for training a machine learning model.
As a non-transient computer-readable medium containing instructions, When the above commands are executed by the processing device, the processing device: (i) receiving sensor data including chamber data indicating the state of the environment of a processing chamber processing a substrate according to a set of process parameters of a current process, and (ii) spectrum data indicating optical emission spectrum (OES) measurements of a plasma placed within a processing chamber processing a substrate according to a set of process parameters of the current process; Using the sensor data as input to a machine learning model; To obtain one or more outputs of the above machine learning model — said one or more outputs represent one or more chamber condition metrics —; Determining the recovery state of the processing chamber based on one or more of the above chamber condition metrics — said recovery state corresponds to a chamber recovery process performed following a preventive maintenance procedure —; and (a) determining that the confidence level that the sensor data accurately indicates the conditions of the processing chamber satisfies a critical condition, and (b) causing a modification to the performance of the processing chamber based on the recovery state of the processing chamber, Non-transient computer-readable media.
In Article 17, When the above commands are executed by the processing device, the processing device additionally: To generate an updated set of process parameters based on the above one or more chamber condition metrics, an update to at least one process parameter among the above set of process parameters is determined, and Modification of the performance of the processing chamber is additionally based on an update to at least one process parameter among the set of process parameters. Non-transient computer-readable media.
In Article 17, One or more outputs of the above machine learning model are based on the spectrum data and the chamber data, and The above machine learning model is trained to predict one or more chamber condition metrics based on the correlation between past OES measurements of plasma and past process result data, Non-transient computer-readable media.
In Article 17, Determining that the confidence level that the sensor data accurately indicates the conditions of the processing chamber satisfies the critical condition is: Using the above sensor data as input to a statistical model; Receiving one or more outputs from the above statistical model — said one or more outputs indicate a confidence level that temporally associated data points of said chamber data and said spectrum data accurately represent the conditions of said processing chamber, and said statistical model is generated using regression between past chamber data and past spectrum data —; and Comparing the above confidence level with one or more critical conditions including the above critical condition, Non-transient computer-readable media.

Description

Acceleration of preventive maintenance recovery and recipe optimization using machine learning-based algorithms [0001] The embodiments of the present disclosure generally relate to predicting chamber conditions of manufacturing systems. Specifically, the present disclosure relates to determining chamber conditions to identify a chamber recovery state and/or update process parameters. [0002] Substrate processing may include a series of processes for producing electrical circuits on semiconductors, such as silicon wafers, according to circuit design. These processes may be executed in a series of chambers. The successful operation of modern semiconductor fabrication facilities may aim to enable a steady stream of wafers to move from one chamber to another during the process of forming electrical circuits on wafers. In a process that performs multiple substrate processes, the processing conditions of the chambers may change, which may result in the processed substrates failing to meet the desired conditions and results. [0003] One such substrate process may include plasma etching, which is a process of transferring a pattern of a mask material layer to another layer under the mask, such as a conductive or dielectric material layer, by removing a layered material from the wafer surface. Depending on the layered material and the etching chemicals, such a process inevitably generates different kinds of etching byproducts, such as silicon oxide and organic polymers. Some of these byproducts are deposited on the inner surfaces of the chamber where the plasma etching process is performed. The deposition of byproducts can affect the etching performance by, for example, by depositing particles (e.g., flakes) on the substrate or by reacting with the plasma to affect the process results. [0004] To mitigate the effects of etching byproducts, preventive maintenance, such as chamber cleaning, may be used to periodically remove deposits from the chamber walls. Examples of preventive maintenance may include stopping production of the chamber and introducing a cleaning plasma, such as CF4 + O2 plasma, into the chamber to clean silicon oxide deposited during silicon etching. This plasma reacts with the deposited material, and the products of this reaction are pumped out of the chamber. However, it has been observed that after such chamber cleaning, the cleaned chamber walls render the chamber unsuitable for immediate production wafer etching. Chamber seasoning is a procedure of etching a series of substrates (e.g., blank silicon wafers) to restore chamber conditions suitable for production substrate processing. After chamber seasoning, a thin layer of silicon oxide may cover the chamber walls. Subsequently, the chamber is returned to production wafer etching for the next round of chamber cleaning and seasoning. Preventive maintenance may also include removing dust and/or deposits by physical methods (e.g., wiping one or more surfaces of the process chamber). [0005] The following is a brief summary of the present disclosure to provide a basic understanding of some aspects of the present disclosure. This summary is not an extensive overview of the present disclosure. It is not intended to identify any core or important elements of the present disclosure, nor to describe any scope of any specific implementations or claims of the present disclosure. Its sole purpose is to present some concepts of the present disclosure in a brief form as an introduction to the more detailed description that follows. [0006] In an exemplary embodiment, the method comprises the step of a processing device receiving sensor data including chamber data that indicates the state of the environment of a processing chamber processing a substrate according to a set of process parameters of a current process. The chamber data further comprises spectral data indicating an optical emission spectra (OES) measurement of a plasma placed within the process chamber. The processing chamber processes a substrate according to a set of process parameters of a current process. The method further comprises the step of using the sensor data as input to a machine learning model. The method further comprises the step of obtaining one or more outputs of the machine learning model. One or more outputs indicate one or more chamber condition metrics. The method further comprises the step of determining the recovery state of the processing chamber based on one or more chamber condition metrics, wherein the recovery state is associated with a chamber recovery process (e.g., a chamber seasoning procedure) performed following a preventive maintenance procedure. The method further comprises the step of causing a modification to the performance of the processing chamber based on the recovery state of the processing chamber. [0007] In an exemplary embodiment, a method is provided for training a machine learning model to determine the state of a processing chamber in a chamber recovery