Non-transitory computer-readable medium, method and system for foreline deposition diagnostics and control

Abstract

Systems, apparatus, and methods for fore-line diagnostics and control are disclosed. In some embodiments, the fore-line coupled to a chamber venting device is equipped with one or more sensors positioned between the chamber venting device and the purging system. The sensors are positioned to measure pressure in the fore-line as an indicator of conductivity. The sensors are coupled to a trained machine learning model configured to provide signals when the fore-line requires a cleaning cycle or when preventative maintenance should be performed. In some embodiments, the trained machine learning predicts when cleaning or preventative maintenance will be required.

Description

[Technical Field]

[0001] The embodiments of the present invention are generally related to frontline diagnostics and control, and more specifically to predictive frontline diagnostics. [Previous Technology]

[0002] Semiconductor substrates are processed for a wide range of applications, including the manufacture of devices and microdevices. These processes can produce high-quality films on the substrate under certain processing conditions, including temperature, pressure, and precursor flow rates within the processing chamber. Any variation in process parameters can lead to changes in film thickness and profile.

[0003] Although the purpose of film deposition is to deposit most of the material onto the substrate, process chamber components, chamber exhaust devices, and front-end components may also be coated. In order to maintain wafer-to-wafer process conditions, a cleaning formulation is performed to etch away undesirable coatings from the chambers and front-end components. However, even the cleaning process may result in undesirable and sometimes flammable coatings in the front-end.

[0004] In conventional methods, a cleaning process is run after each wafer or after a certain number of wafers without measuring the amount of deposits in the front-end. This may result in cleaning being too frequent, which could be a waste of cleaning resources and time, or insufficient cleaning, leading to unreliable variations in process parameters and potentially dangerous deposits in the front-end.

[0005] What is needed is a system and methods to overcome the shortcomings of conventional methods. [Summary of the Invention]

[0006] This disclosure generally relates to systems, apparatus, and methods for frontline diagnostics and control. In some embodiments, a frontline device coupled to a chamber exhaust system is equipped with one or more sensors positioned between the chamber exhaust system and the elimination system. The sensors are positioned to measure pressure in the frontline as an indicator of conductivity. The sensors are coupled to a trained machine learning model configured to provide signals when the frontline requires a cleaning cycle or when preventative maintenance should be performed. In some embodiments, the trained machine learning predicts when cleaning or preventative maintenance will be required.

[0007] In one embodiment, a system for frontline diagnostics and control is disclosed, the system comprising: a frontline coupled to an exhaust device of a processing chamber; a first sensor positioned to measure deposit accumulation in the frontline; and an accumulation monitor coupled to the first sensor, the accumulation monitor including a trained machine learning (ML) model and configured to generate an output indicating deposit accumulation and trigger a correction action when an indicated deposit accumulation is at or above an accumulation threshold.

[0008] In another embodiment, a computer-readable medium is disclosed, comprising instructions for performing a method for frontline deposition diagnostics. The method includes receiving accumulated data from at least one of a plurality of sensors; providing the accumulated data to a accumulation monitor, the accumulation monitor including a trained machine learning (ML) model configured to produce an output indicating deposition accumulation; and generating a correction action based on the output when the accumulated data is indicated to be at an accumulation threshold.

[0009] In another embodiment, a method for monitoring deposition accumulation in the front end of a processing system is disclosed, comprising receiving sensor data from one or more sensors coupled to the front end of a semiconductor processing system; processing the sensor data with a trained machine learning model to estimate the accumulation; and triggering a correction action corresponding to the estimated accumulation of the trained machine learning model.

Implementation Method

[0016] In the following text, reference is made to embodiments of the present disclosure. However, it should be understood that the present disclosure is not limited to the specific embodiments described. Rather, any combination of the following features and elements, whether or not related to different embodiments, is considered for implementing and practicing the present disclosure. Furthermore, although embodiments of the present disclosure may achieve advantages over other possible solutions and / or over the prior art, whether a particular advantage is achieved by a given embodiment does not limit the present disclosure. Therefore, the following appearances, features, embodiments, and advantages are merely illustrative and are not to be considered elements or limitations of the appended claims unless expressly stated in the claims. Similarly, references to "the present disclosure" should not be construed as a generalization of any inventive subject matter disclosed herein and should not be considered elements or limitations of the appended claims unless expressly stated in the claims.

[0017] Discloses systems, equipment and methods for monitoring deposit accumulation in the front line of a processing system using accumulated historical process and cleaning data, as well as real-time data from one or more sensors placed along the front line.

[0018] During the processing of a semiconductor substrate, accumulation can occur on the chamber components and in the front line coupled to the chamber, such as by epitaxial processes, chemical vapor deposition (CVD) processes, plasma-enhanced chemical vapor deposition (PECVD) processes, etching processes, cleaning processes, or other processes performed in the processing chamber as part of the manufacture of semiconductor devices.

[0019] The inventors have discovered that accumulated patterns in the front line can be detected directly and / or indirectly by sensors used with the front line. By collecting sensor data and accumulated patterns from multiple processing cycles, machine learning (ML) algorithms trained with training data can be used to develop thresholds that indicate alarms to the user, the required front line (or chamber) cleaning, or the required preventive maintenance (PM). The trained machine learning model can use the identified patterns to develop thresholds for providing corrective actions (such as stopping operation of the processing chamber, or that cleaning or preventive maintenance (PM) should be performed on the front line), and to predict when such thresholds may be needed.

[0020] By providing thresholds for cleaning and PM monitored by a trained ML model, cleaning / PM cycles can be predictively determined based on the processing parameters of individual and multiple sets of processing chambers, enabling more efficient scheduling of plant operations. Furthermore, inventory of cleaning and PM-related chemicals, replacement parts, and labor resources can be managed more effectively. An exemplary system for frontline deposition diagnostics and control.

[0021] Figure 1 depicts a system 100 for frontline deposition and diagnostic control according to certain embodiments. A processing chamber 103 is coupled to a frontline 106 via an exhaust device (not shown) of the processing chamber 103, thereby coupling the processing chamber 103 to a pump 109, which is in turn coupled to a removal system 112.

[0022] Processing chamber 103 may be an epitaxial chamber, a chemical vapor deposition (CVD) chamber, a plasma-enhanced CVD (PECVD) chamber, an etching chamber, a polishing station, or any processing chamber used to deposit or remove one or more materials from a substrate by any means. As understood by those skilled in the art, when used for deposition or etching, a processing chamber (such as processing chamber 103) is controlled by one or more formulations of gases and materials controlled to be input into the processing chamber. During operation of processing chamber 103, materials not deposited on or etched from the substrate are deposited on chamber components, in frontline 106, and in pump 109 and removal system 112. The systems and methods disclosed herein seek to diagnose deposits in the frontline and indicate, via various algorithms and methods, when frontline cleaning and / or frontline preventative maintenance (PM) should be performed.

[0023] Multiple valves may be positioned along the length of frontline 106 to isolate treatment chamber 103, pump 109, and control pressure within frontline 106. A chamber isolation valve 115 may be located downstream of treatment chamber 103 to isolate treatment chamber 103 during frontline cleaning and PM operations (discussed further below). A pressure control valve 118 may be located downstream of chamber isolation valve 115 and coupled to frontline 106.

[0024] The pressure control valve (PCV) 118 includes a rotary plate and a position sensor. The rotary plate controls the gas pressure in the front line 106 by controlling the conductivity of such gas, and the position sensor is used to sense the rotation angle of the rotary plate. The angle of the PCV 118 is an indication of the angle of the rotary plate, which can be used as an indicator of the pressure and / or conductivity within the front line 106, as described below, and multiple measurement angles can be indicators of changes in pressure and / or conductivity. As will be discussed below, changes in conductivity can be an indicator of deposit accumulation within the front line 106.

[0025] Downstream of PCV 118, front line 106 is coupled to pump isolation valve 121, which isolates pump 109 from front line 106. Downstream of pump isolation valve 121, front line is coupled to pump 109.

[0026] As will be understood by those skilled in the art, in some embodiments, one or both of the chamber isolation valve 115 and the pump isolation valve 121 may be a PCV. In this case, the PCV can be used as an isolation valve.

[0027] Along front line 106, sensors 124 to 133 are distributed along front line 106. Sensor 124 is located between processing chamber 103 and chamber isolation valve 115, sensor 127 is located between chamber isolation valve 115 and PCV 118, sensor 130 is located between PCV 118 and pump isolation valve 121, and sensor 133 is located between pump isolation valve 121 and pump 109.

[0028] In some embodiments, one or more of the sensors 124 to 133 are positioned to directly or indirectly measure buildup within the front line. In some embodiments, one or more of the sensors 124 to 133 may be ultrasonic sensors, which include a transducer and a receiver for an ultrasonic signal. The ultrasonic transducer provided as one or more of the sensors 124 to 133 can detect buildup within the front line 106, but is not necessarily located within the front line 106; however, in some embodiments, this type of sensor may be located within the front line 106. When positioned on the outside of the front line 106, the ultrasonic transducer transmits ultrasonic signals through the wall of the front line, which is coupled to the interior space of the front line 106, such as the hollow portion of a conduit constituting the front line 106. When frontline 106 is clean (i.e., without deposits), the ultrasonic signal can travel through frontline 106 at the speed of sound detected by the receiver (possibly slowed down by the gas in the frontline). However, when deposits are present in the frontline, these deposits will slow down the travel rate of the ultrasonic signal, thus prolonging the time it takes from generating the ultrasonic signal at the transducer to receiving it at the receiver. As the deposits continue to accumulate, the ultrasonic signal will take an increasingly longer amount of time to reach the receiver. Furthermore, different materials deposited in frontline 106 will have different effects on the speed of the ultrasonic signal, the only constant being that additional deposit accumulation will gradually slow down the signal.

[0029] In some embodiments, one or more of the sensors 124 to 133 may be piezoelectric sensors, such as quartz crystal microbalance (QCM) sensors. A QCM sensor is a mass-sensitive piezoelectric element based on an oscillating quartz crystal, capable of detecting nanogram mass changes in the quartz component. In an embodiment, as material accumulates on the quartz component, the quartz component will oscillate at a lower rate due to the increased mass. When placed within frontline 106, for example at the location indicated by one or more of the sensors 124 to 133, the signal generated by the sensor will change in relation to the accumulation as deposition occurs on the QCM sensor.

[0030] In some embodiments, one or more of the sensors 124 to 133 may be pressure sensors coupled to frontline 106 to measure pressure within frontline 106. As material accumulates in frontline 106, pressure changes caused by the decreasing volume of gas passing through frontline 106 (due to accumulation) can be detected as an indicator of accumulation.

[0031] Although several embodiments of sensors 124 to 133 have been discussed above, those skilled in the art will understand that one or more of the sensors 124 to 133 may include a variety of sensor types. For example, the sensors may include one or more of pressure sensors, ultrasonic sensors (UT sensors), and / or QCM sensors.

[0032] In embodiments, PCV 118 may additionally be used as a sensor to indicate accumulation in front 106. In these embodiments, as accumulation within front 106 increases, PCV 118 may be required to increase the angle of the plate to maintain the target conductivity of the gas. An increase in angle relative to a previous angle may indicate an increase in accumulation in front 106, while a decrease in angle (e.g., after cleaning or PM) may indicate less or no accumulation.

[0033] One or more of the sensor types discussed above can be used in conjunction with PCV 118 to provide data indicating a continuously increasing or decreasing accumulation. For example, when sensor 127 detects a change in accumulation, PCV 118 can respond by changing its angle. Overall, these signals provide information about how system 100 responds to front-line deposition accumulation.

[0034] Controller 136 is coupled to each of sensors 124 to 133, and PCV 118, and in some embodiments is also coupled to chamber isolation valve 115 and pump isolation valve 121. Controller 136 is additionally coupled to processing chamber 103 to receive information about the processing recipe and to provide the processing chamber with signals about the status of front-end 106.

[0035] Controller 136 includes a trained machine learning (ML) model 144 containing training data 139; and an ML model 141. Training data 139 consists of data related to processing in processing chamber 103 (e.g., features) (e.g., recipe data), and processing chamber sensor data (including time-series data about such data) generated by one or more processing chamber sensors during processing operation. Training data 139 may be stored in a database or other data storage system. Training data further includes data from one or more sensors 124 to 133 (i.e., data about at least one of accumulation for an accumulation sensor and pressure for a pressure sensor), data from PCV 118, data from chamber isolation valve 115, and data from pump isolation valve 121 (e.g., conductivity, angle data, etc.), which are direct or indirect indicators of accumulation in front-end 106, including a time-series representation of the accumulation. In some embodiments, training data 139 may include operational data from pump 109. Training data may be time-series data from each of the aforementioned data sources, thereby indicating differences in operational data from each data source over time. In some embodiments, training data may be historical data, synthetic data, data created by the developer for training purposes, or some combination of such data. In some embodiments, training data may be grouped based on the processing chamber formulation, materials depositable in frontline 106, deposition characteristics of the deposited materials, or any other grouping suitable for training ML model 141. Training data 144 may further include data received in situ from sensors and valves, and train ML model 141 on an ongoing basis, thereby providing a continuously updated trained ML model 144.

[0036] Based on training data 144, one or more parameters of the frontier 106 can be estimated or inferred by the trained ML model 141. For example, the material deposition rate in the frontier 106 can be derived from time-of-flight data of the ultrasonic signal detected by the ultrasonic sensor, the resonant frequency shift rate of the QCM sensor, or the rate of pressure difference change in a segment of the frontier 106 from one or more pressure sensors. The deposition rate within one or more segments of the frontier 106 can be inferred / estimated by the trained ML model 141 and can be considered as a metric for making decisions about previously measured deposition rates, or compared with one or more thresholds. Another exemplary parameter of the frontier 106 that can be estimated or inferred by the trained ML model 141 is the health of the frontier 106 (e.g., conductivity). The trained ML model 141 can estimate / infer the condition of partial blockage compared to the health status based on a combination of the location of PCV 118, the temperature of frontline 106 at PCV 118, and signals from an ultrasonic sensor, QCM sensor, or other sensor. The inferred / estimated conditions can be compared with previously measured conditions of frontline 106, or one or more thresholds, to determine whether frontline 106 is healthy (i.e., suitable for operating the frontline level conduction gas) or whether cleaning or PM is required.

[0037] In some embodiments, the ML model 141 may be trained for a specific processing formulation performed by the processing chamber, for a specific material deposited in the front end, or a combination thereof.

[0038] ML model 141 may be a supervised or unsupervised ML model or a combination of such models. In embodiments where ML model 141 is a supervised model, a support vector machine (SVM), a regression model, or any supervised learning model capable of receiving training data 139 and providing continuous outputs indicating or predicting deposition levels in frontier 106 may be used. In embodiments where ML model is an unsupervised model, a neural network, or any unsupervised learning model capable of receiving training data 139 to train ML model 141, may be used to provide clustered or classified outputs indicating or predicting deposition accumulation in frontier 106.

[0039] During training, the ML model 141 is trained to develop one or more accumulated thresholds in the front-end 106 based on training data 139. These thresholds indicate the need for front-end cleaning, the need for PM, or some other indicator. In this case, the thresholds may be deposition levels in the front-end 106, deposition rates in the front-end 106, blockage / partial blockage versus healthy (sufficiently open) front-end 106, as determined by data provided by one or more sensors in the front-end 106. For example, given a formulation for the processing chamber and the front-end conditions required to properly process the substrate using that formulation, training data from the formulation, processing chamber, and front-end sensors can be compiled over time. This training data is used to train the ML model to identify one or more thresholds for normal operating system operation, the presence or absence of which indicates the need for front-end cleaning or PM. Once trained, the trained ML model can receive "real-time" processing data to provide indicators when front-end conditions require cleaning or PM. In this scenario, indicators can originate from reaching or exceeding threshold regions, signaling one or more categories in the unsupervised model that cleaning is required. In supervised models, cleaning can be indicated when one or more dimensions reach or exceed a threshold. Furthermore, trained ML models can predictively indicate when cleaning or PM is needed, enabling efficient scheduling of resources and plant operations.

[0040] In some embodiments, as a result of training the ML model 141 with training data, the trained ML model 144 can predict a given deposition accumulation in the front-end 106 at a given time based on a given formulation and associated process parameters. In predictive embodiments, the controller can predict when to perform cleaning or PM on the front-end 106, for example, to adapt to plant scheduling, process chamber 103 maintenance scheduling, and notify resource purchasing and delivery decisions.

[0041] In some embodiments, controller 136 is an open-loop controller that includes thresholds for sensor and valve data, which indicate when to perform cleaning, PM, or other actions. In other embodiments, controller 136 may be a combination of an open-loop control method and a trained machine learning model 144.

[0042] In operation, sensors 124 to 133, valves 115 to 121, and one or more of the processing chamber 103 in some embodiments provide processing data to a trained ML model 144. The trained ML model 144 receives these inputs, processes them according to a trained machine learning model, and provides one or more outputs, generating indicators to perform frontline cleaning or PM when the outputs reach or exceed one or more thresholds. In some embodiments, the controller 136 may isolate the processing chamber from the frontline 106 and provide signals to the cleaning system to perform frontline cleaning (discussed further below), or provide indicators to the user of the system 100 to perform cleaning. Exemplary system for frontline deposition diagnostics and control using the RPS frontline cleaning system.

[0043] Figure 2 depicts a system 200 for frontline deposition and diagnostic control according to certain embodiments, the system including a frontline cleaning system. System 200 is similar to system 100, with the addition of a frontline cleaning system 210, a frontline cleaning isolation valve 215, and improvements discussed below. In some embodiments, the frontline cleaning 210 is similar to the frontline cleaning system disclosed in U.S. Patent 10,500,614 to Hilkene et al., entitled "TEMPERATURE CONTROLLED REMOTE PLASMA CLEAN FOR EXHAUST DEPOSIT REMOVAL," which was issued on December 10, 2019, to Applied Materials of Santa Clara, CA, the entire disclosure of which is incorporated herein by reference.

[0044] In embodiments including frontline cleaning 210 for frontline deposition diagnostics and control, one or more of sensors 124 to 133 may include thermal sensors. In some embodiments, frontline cleaning 210 includes a remote plasma source (RPS) for generating plasma and supplying it to frontline 106 to remove accumulations. In some embodiments, the RPS generates plasma from a mixture of NF3 and argon. Since the reaction of plasma with accumulations in frontline 106 can result in an exothermic reaction, the heat generated during the cleaning process can be an indicator of the amount of remaining accumulations in frontline 106. Subsequently, as accumulations are removed due to reaction with plasma, the amount of heat generated decreases until a negligible amount of heat is generated, indicating that the cleaning cycle is complete.

[0045] In the embodiment of Figure 2, the trained ML model 144 may further include parameters and thresholds for monitoring accumulations in the front line during a cleaning cycle performed by front line cleaning 210. Training data 139 for developing the parameters and thresholds includes data from a thermal sensor included in one or more sensors 124 to 133. This training data 139 is provided to train the ML model 141 to produce the trained ML model 144. Exemplary method for front line diagnostics and control

[0046] Figure 3 depicts a method 300 for frontline diagnostics and control according to certain embodiments.

[0047] At 305, method 300 receives sensor data from one or more sensors coupled to a front-end of a semiconductor processing system, and at 310, method 300 processes the sensor data with a trained machine learning model to estimate accumulations in the front-end. In an embodiment, the trained machine learning model is trained via: receiving sensor training data from a database including sensor data from previous operations of one or more semiconductor processing systems; classifying the sensor training data to distinguish between one of the clean surfaces of the front-end and one of multiple deposition thicknesses of material deposited on the front-end; and generating model parameters based on the classification for the trained machine learning model. In some embodiments, the trained machine learning model is further trained to detect temperature changes in the front-end. It should be understood that the classification in this case may relate to supervised ML or unsupervised ML.

[0048] At 315, method 300 triggers a corrective action corresponding to the accumulated estimates of the trained machine learning model. This corrective action may include one or more of the following: stopping operation of the processing system, initiating cleaning operations in the semiconductor processing system or front-end, or warning the user to perform preventative maintenance on the semiconductor processing system or front-end. A signal may be generated in response to a threshold indicated by the trained machine learning model being reached or exceeded. An exemplary processing system for front-end diagnostics and control.

[0049] Figure 4 depicts a processing system 400 for frontline diagnostics and control according to certain embodiments, which can perform the methods described herein, such as the method for frontline diagnostics and control of Figure 3; and the system described with respect to Figures 1 and 2 capable of performing such methods.

[0050] The processing apparatus 401 of the processing system 400 includes a central processing unit (CPU) 402 connected to a data bus 416. The CPU 402 is configured to process, for example, computer-executable instructions stored in memory 408 or storage device 410, and to cause the processing system 400 to perform the methods described herein, for example, with respect to Figure 3. The CPU 402 is included to represent a single CPU, multiple CPUs, a single CPU having multiple processing cores, and other forms of processing architecture capable of executing computer-executable instructions.

[0051] The processing system 400 further includes an input / output (I / O) device 412 and an interface 404, which allows the processing device 401 to connect to the input / output device 412. The input / output device may be a semiconductor processing system and associated support system, a keyboard, a display, a mouse, a pen input device, or other devices that allow interaction with the processing system 400. It should be noted that the processing system 400 may connect to external I / O devices (e.g., external display devices) via physical and wireless connections.

[0052] The processing device 401 further includes a network interface 406 that provides the processing system 401 with access to an external network 414 and thereby to an external computing device.

[0053] The processing device 401 further includes a memory 408, which in this example includes a receiving module 418, a processing module 420, a triggering module 422, a training module 424, and a classification module 426 for performing the operations described in Figure 3, which can be performed on the system depicted in Figure 1 and / or Figure 2. Although not depicted in Figure 4, the memory 408 may include other configurations.

[0054] It should be noted that although for simplicity it is illustrated as a single memory 408 in Figure 4, the various states stored in memory 408 may be stored in different physical memories, including memories far from the processing system 400, but may be accessed by the CPU 402 via internal data connections (such as bus 416).

[0055] Storage device 410 includes accumulated data 428, sensor data 430, ML model data 432, trained ML learning model data 434, model parameter data 436, sensor training data 438, and correction action data 440, as described in conjunction with Figures 1 to 3. Although not depicted in Figure 4, storage device 410 may include other formats.

[0056] Like memory 408, a single storage device 410 is depicted in Figure 4 for simplicity, but the various states stored in the storage device 410 may be stored in different physical storage devices, but such different physical storage devices are CPU 402 via internal data connections (such as bus 416) or externally (such as network interface 406). A person familiar with this artifice will understand that one or more elements of the processing system 400 may be remotely located and accessed via the network 414 . Other Notes

[0057] The preceding description is provided to enable anyone familiar with the art to practice the various embodiments described herein. The examples discussed herein do not limit the scope, applicability, or embodiments articulated in the scope of the applied patent. Various modifications to such embodiments will be obvious to those skilled in the art, and the general principles defined herein can be applied to other embodiments. For example, changes may be made to the function and arrangement of the components discussed without departing from the rubric of the present disclosure case. Various instances may be appropriately omitted, replaced, or added to various procedures or parts. For example, the described method may be performed differently from the order of the order described, and various steps may be added, omitted, or combined. Furthermore, features about some instance descriptions can be combined in some other instances. For example, the equipment may be implemented using any number of states described herein, or the method may be practiced using any number of states described herein. Furthermore, the rubric of this disclosure is intended to cover such equipment or methods practiced using other structures, functionalities, or structures and functions in addition to or in addition to the various states of this disclosure articulated herein. It should be understood that any state of the present disclosure case revealed herein may be manifested by one or more elements of the claim.

[0058] As used herein, the word "illustrative" means "to be used as an example, instance, or illustration." Any state described herein as “illustrative” is not necessarily construed as superior to or advantageous over another.

[0059] As used herein, phrases referring to "at least one" in a list of items refer to any combination of such items, including a single member. As an instance, “at least one of a, b, or c” is intended to cover a, b, c, ab, ac, bc, and abc, and any combination of multiple identical elements (e.g., aa, aaa, aab, aac, abb, acc, bb, bbb, bbc, cc, and ccc, or any other order of a, b, and c).

[0060] As used herein, the term "determine" encompasses a variety of actions. For example, "determine" may include calculation, processing, derivation, investigation, searching (e.g., searching in a table, database, or other data structure), ascertainment, etc. Furthermore, "determine" may include receiving (e.g., receiving information), accessing (e.g., accessing data in memory), etc. In addition, "determine" may include parsing, selecting, picking, building, etc.

[0061] The methods disclosed herein include one or more steps or actions for implementing such methods. The method steps and / or actions may be interchanged without departing from the scope of the claims. In other words, unless a specific order of steps or actions is specified, the order and / or use of specific steps and / or actions may be modified without departing from the scope of the claims. Furthermore, the various operations of the above methods can be performed by any suitable means capable of performing the corresponding functions. Such means may include various hardware and / or software components and / or modules, including but not limited to circuits, application-specific integrated circuits (ASICs), or processors. Typically, in the presence of operations illustrated in the figures, such operations may have corresponding paired means plus functional components with similar numbering.

[0062] The following claims are not intended to be limited to the embodiments shown herein, but rather to encompass the entire range consistent with the language of the claims. In claims, unless specifically stated otherwise, references to singular elements do not mean "one and only one," but rather "one or more." Unless otherwise specifically stated, the term "some" means one or more. No claim element shall be construed under patent law unless the element is explicitly stated using the term "component for," or, in the case of a method claim, the element is stated using the term "step for." All structural and functional equivalents of elements throughout the various forms described in this disclosure, known to or to be known by those skilled in the art, are expressly incorporated herein by reference and are intended to be covered by the claims. Furthermore, the disclosure herein is not intended to be offered to the public, whether or not such disclosure is expressly stated in the claims.

[0063] Although the embodiments described above are related to this disclosure, other and further embodiments of this disclosure may be designed without departing from the basic scope of this disclosure, and the scope of this disclosure is determined by the appended claims. [Simplified Explanation of the Diagram]

[0010] In order to understand the above features of this disclosure in detail, the present disclosure, which has been briefly summarized above, can be described in more detail with reference to the embodiments, some of which are illustrated in the accompanying drawings. However, it should be noted that the drawings only illustrate exemplary embodiments and should therefore not be regarded as limiting its scope, and other equally effective embodiments may be permitted.

[0011] Figure 1 depicts a system for frontline deposition and diagnostic control according to certain embodiments.

[0012] Figure 2 depicts a system for frontline deposition and diagnostic control according to certain embodiments, the system including a frontline cleaning system.

[0013] Figure 3 depicts a method for frontline diagnostics and control according to certain embodiments.

[0014] Figure 4 is a processing system according to certain embodiments for performing methods for frontline diagnostics and control.

[0015] To facilitate understanding, the same reference numerals are used to denote elements common in the drawings where possible. It is contemplated that elements and features of one embodiment can be advantageously incorporated into other embodiments without further description. [Biomaterial Storage]

[0065] Domestic storage information (please note in order of storage institution, date, and number): None. International storage information (please note in order of storage country, institution, date, and number): None.

Claims

1. A system for upstream pipeline diagnosis and control, comprising: A pre-line is coupled to an exhaust device in a processing chamber; a sensor is positioned to measure deposit accumulation in the pre-line. The system also includes an accumulation monitor coupled to the first sensor, the accumulation monitor comprising a trained machine learning (ML) model and configured to generate an output indicating deposition accumulation, and to trigger a correction action when the indicated deposition accumulation is at or above an accumulation threshold, wherein the trained machine learning model is trained via a procedure including: receiving sensor training data from a database including sensor data from previous operations of a semiconductor processing chamber; classifying the sensor training data to distinguish between a clean surface of the front-end pipeline and a deposition thickness of a material deposited on the front-end pipeline; and generating model parameters for the trained machine learning model based on the classification.

2. The system as claimed in claim 1, wherein the sensor comprises: an ultrasonic transducer; a piezoelectric sensor; a pressure sensor; or a position sensor for a pressure control valve.

3. The system as described in claim 1, wherein the cumulative monitor includes an open-loop control.

4. The system as described in claim 1, wherein the cumulative monitor is coupled to a plurality of sensors.

5. The system as claimed in claim 1, wherein the calibration action includes one or more of the following: generating an alarm to a user, stopping the operation of the semiconductor processing chamber, or initiating a cleaning operation of the semiconductor processing chamber.

6. The system as described in claim 5, wherein the trained machine learning model is further configured to monitor a temperature change in the upstream piping during the cleaning operation.

7. The system as claimed in claim 1, wherein the semiconductor processing chamber is the processing chamber.

8. A non-transitory computer-readable medium including instructions for performing a method for front-end pipeline deposition diagnostics, the method comprising the steps of: receiving sensor data from at least one of a plurality of sensors; providing the sensor data to an accumulation monitor including a trained machine learning (ML) model configured to generate an output indicating deposition accumulation, wherein the trained ML model is trained via a procedure including: receiving sensor training data from a database including sensor data from previous operations of a semiconductor processing chamber; classifying the sensor training data to distinguish between a clean surface of the front-end pipeline and a deposition thickness of a material deposited on the front-end pipeline; generating model parameters for the trained machine learning model based on the classification; and generating a correction action based on the output when the sensor data is indicated to be at or above a accumulation threshold.

9. The non-transitory computer-readable medium as described in claim 8, wherein the plurality of sensors includes one or more of an ultrasonic transducer or a piezoelectric sensor.

10. The non-transitory computer-readable medium as described in claim 8, wherein the plurality of sensors includes a pressure sensor and a pressure control valve.

11. The non-transitory computer-readable medium as described in claim 10, wherein the pressure sensor is located adjacent to the pressure control valve.

12. The non-transitory computer-readable medium as described in claim 8, wherein the plurality of sensors are positioned between a pump and a chamber exhaust device.

13. The non-transitory computer-readable medium as described in claim 12, wherein the trained machine learning (ML) model includes one or more of a supervised learning model or an unsupervised learning model.

14. The non-transitory computer-readable medium as described in claim 8, wherein the calibration action is generated at least in part by an open-loop control configured to monitor the sensor data.

15. The non-transitory computer-readable medium as described in claim 8, wherein the trained machine learning (ML) model includes one or more of a supervised model or an unsupervised model.

16. A method for monitoring deposition accumulation in a front-end pipeline of a processing system, comprising the steps of: receiving sensor data from one or more sensors coupled to a front-end pipeline of a processing system; processing the sensor data with a trained machine learning model to estimate an accumulation, wherein the trained ML model is trained via a procedure comprising: receiving sensor training data from previous operation of one or more processing systems; classifying the sensor training data to distinguish between a clean surface of the front-end pipeline and a deposition thickness of a material deposited on the front-end pipeline; generating model parameters for the trained machine learning model based on the classification; and triggering a correction action corresponding to the accumulation estimated by the trained machine learning model.

17. The method as described in claim 16, wherein the trained machine learning model is further trained to detect a temperature change in the upstream piping.

18. The method as described in claim 16, wherein the one or more sensors include one or more of the following: an ultrasonic transducer; a piezoelectric sensor; a pressure sensor; or a position sensor for a pressure control valve.

19. The method as described in claim 16, wherein the corrective action includes one or more of the following: stopping the operation of the processing system, initiating a cleaning operation, or warning a user.

20. The method as described in claim 16, wherein the one or more sensors are positioned between a pump and a chamber exhaust device.

Images