Method and apparatus for detecting the end point of an aging process

Abstract

A method for detecting the endpoint of an aging process in a processing chamber, the method comprising the steps of: obtaining aging progress data indicating the progress of the aging process for each of the first plurality of substrates, and collecting historical parameter values ​​from a plurality of sensors disposed in the processing chamber; normalizing the historical parameter values ​​of each of the first plurality of substrates relative to a plurality of parameter values ​​for a particular substrate among the first plurality of substrates; generating an MVA model by applying a set of coefficients to the normalized parameter values ​​of each of the first plurality of substrates, and regressing the set of coefficients based on the aging progress data; and using the MVA model to determine the endpoint of the aging process, the MVA model having a plurality of substantially real-time parameter values ​​measured when aging is performed on each of the second plurality of substrates.

Description

Technical Field

[0001] The present application relates to methods and apparatus for detecting the endpoint of a seasoning process in a processing chamber. Background Art

[0002] Semiconductor device manufacturing includes a series of processes that create circuits in semiconductors (e.g., silicon substrates) according to circuit designs. Many manufacturing processes, such as deposition or etching, are performed in processing chambers, such as chemical vapor deposition (CVD) chambers. Efficient, non-damaging cleaning of processing chambers is important for smooth operation of the process as well as improved device yield and better product performance. Thus, periodic chamber maintenance is required to improve the performance of the chamber to produce high-quality semiconductor devices.

[0003] After such chamber maintenance, a seasoning process is required to restore chamber conditions suitable for substrate manufacturing processing before continuing normal production processing. As processing is performed in the processing chamber, material is gradually deposited on the interior surfaces of the chamber. This can cause a shift in process variables until the entire interior surface is deposited with material. Therefore, a method is needed to detect the end point of the seasoning process to identify when to stop the seasoning process and when production processing can continue in the processing chamber. Summary of the invention

[0004] The present disclosure provides methods and apparatus for detecting the endpoint of a seasoning process in a processing chamber.

[0005] In one embodiment, an apparatus for detecting an endpoint of a seasoning process in a processing chamber may include: a processing chamber; a set of sensors coupled to the processing chamber; at least one memory configured to store seasoning data indicating progress of a past seasoning process for each substrate in a first plurality of substrates; and historical parameter values ​​measured by a plurality of sensors selected from the set of sensors during past seasoning processes performed on each substrate in the first plurality of substrates. The apparatus may further include a controller configured to normalize the historical parameter values ​​for each substrate in the first plurality of substrates relative to a plurality of parameter values ​​for a particular substrate in the first plurality of substrates. The controller may also be configured to generate a multivariate analysis (MVA) model by applying a set of coefficients to the normalized historical parameter values ​​for each substrate in the first plurality of substrates, and determining the set of coefficients by performing a regression operation on output values ​​of a logistic equation. The input to the logistic equation is the output of the MVA model having the normalized historical parameter values ​​for each substrate in the first plurality of substrates. The controller may be further configured to receive a plurality of substantially real-time parameter values ​​from the plurality of sensors while processing the second plurality of substrates, and determine an endpoint of the seasoning process using the MVA model with the plurality of substantially real-time parameter values.

[0006] In another embodiment, a method for detecting an endpoint of a seasoning process in a processing chamber may include the steps of obtaining seasoning progress data indicating a progress of a past seasoning process for each substrate of a first plurality of substrates; obtaining historical parameter values ​​measured by a plurality of sensors selected from a set of sensors coupled to the processing chamber during the past seasoning process performed on each substrate of the first plurality of substrates. The method may also include the steps of normalizing the historical parameter values ​​of each substrate of the first plurality of substrates relative to the plurality of parameter values ​​of a particular substrate of the first plurality of substrates; generating a multivariate analysis (MVA) model by applying a set of coefficients to the normalized historical parameter values ​​of each substrate of the first plurality of substrates; and determining the set of coefficients by performing a regression operation on the output of a logistic equation. The input of the logistic equation is the output of the MVA model with the normalized historical parameter values ​​of each substrate of the first plurality of substrates. The method may further include the steps of receiving a plurality of substantially real-time parameter values ​​from a plurality of sensors when processing each substrate of a second plurality of substrates, and determining an endpoint of the seasoning process using the MVA model with the plurality of substantially real-time parameter values.

[0007] In yet another embodiment, a method for detecting an endpoint of a seasoning process is provided. The method includes obtaining seasoning progress data indicating the progress of the seasoning process for each substrate in a first plurality of substrates. The method may also include obtaining historical parameter values ​​measured from a subset of a plurality of sensors coupled to a processing chamber, each of the plurality of sensors measuring a parameter of the seasoning process when processing each substrate in the first plurality of substrates. The method may also include normalizing the historical parameter values ​​relative to a subset of historical parameter values ​​associated with a particular substrate in the first plurality of substrates. The method may also include normalizing the seasoning progress data for each substrate in the first plurality of substrates relative to the seasoning progress data associated with completing the seasoning process. The method may also include generating a multivariate analysis (MVA) model by applying coefficients to each normalized historical parameter value in the first plurality of substrates, wherein the coefficients are determined by performing a regression operation on a plurality of outputs of a logistic equation, wherein the inputs of the logistic equation are outputs of the MVA model. The method may also include receiving a plurality of substantially real-time parameter values ​​from a subset of sensors when processing each substrate in a second plurality of substrates. The method may also include using the MVA model with multiple substantially real-time parameter values ​​to determine an endpoint of the aging process. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Therefore, in order to be able to understand in detail the manner in which the above-mentioned characteristic structures of the present disclosure are used, a more specific description of the invention briefly summarized above can be made with reference to various embodiments, some of which are shown in the accompanying drawings. However, it should be noted that the accompanying drawings only illustrate exemplary embodiments of the present disclosure and therefore should not be considered as limiting the scope of the present disclosure, as the present disclosure may admit to other equally effective embodiments.

[0009] Figure 1 is a schematic cross-sectional view of a processing chamber according to aspects of the present disclosure.

[0010] Figure 2A is a schematic block diagram of a process control system for detecting an endpoint of a aging process according to aspects of the present disclosure.

[0011] Figure 2B Graphs show the results of a regression process according to aspects of the present disclosure.

[0012] Figure 3 is a flow chart of a method for detecting an endpoint of an aging process according to aspects of the present disclosure.

[0013] Figure 4 An exemplary configuration of a process control system according to aspects of the present disclosure is illustrated.

[0014] To facilitate understanding, identical reference numerals have been used, where possible, to designate identical elements that are common to the figures. It is contemplated that elements and features of one embodiment may be beneficially incorporated in other embodiments without further recitation. DETAILED DESCRIPTION

[0015] Figure 1 1 is a schematic cross-sectional view of a processing chamber 100 according to one aspect of the present disclosure. The processing chamber 100 includes a chamber body 111, a dielectric top 112, a chemical delivery module 121, a first RF match 122, a second RF match 124, and a plurality of sensors 131-138.

[0016] The dielectric top 112 is adjacent to the showerhead assembly 106. The processing space 109 is at least partially defined by the chamber body 111 and the dielectric top 112. The substrate support 108 is disposed in the processing space 109. The showerhead assembly 106 is disposed in the processing space 109 opposite to the substrate support 108. The showerhead assembly 106 includes a plurality of processing gas channels, including a first gas channel 106A and a second gas channel 106B. One or more processing gases are independently delivered to the processing space 109 via the first gas channel 106A and the second gas channel 106B. The first gas channel 106A and the second gas channel 106B are fluidically connected to the processing space 109 via one or more openings (not shown) formed by the showerhead.

[0017] The chemical delivery module 121 delivers various process gases from one or more gas sources (not shown) to the first gas channel 106A and the second gas channel 106B. The various process gases may include precursor gases, carrier gases, and / or purge gases for deposition processes or etching processes.

[0018] The first RF matching circuit 122 is coupled to a radio frequency (RF) power source 123 of, for example, 12.56 MHz, and supplies RF (or AC) power to the induction coil 110 to excite the process gas in the process space 109 and generate a high-density plasma therein. The second RF matching circuit 124 is coupled to a bias power source 125 of, for example, 13.56 MHz, and supplies RF (or AC) power to a cathode 126 disposed in the substrate support 108. The second RF matching circuit 124 controls the ion bombardment energy delivered to the substrate 107 disposed on the substrate support 108. In some embodiments, the substrate 107 is a silicon substrate for semiconductor device manufacturing.

[0019] In some embodiments, cathode 126 is part of an electrostatic chuck (not shown) embedded within substrate support 108. During the etching process, the electrostatic chuck is thermally controlled using a gas (such as helium) to cool the electrostatic chuck and substrate 107. The RF (or AC) power supplied to induction coil 110 and cathode 126 can vary depending on the particular application.

[0020] The plurality of sensors 131-138 may include one or more of voltage and / or current probes (or VI probes), optical sensors, temperature sensors, pressure sensors, gas sensors, and other sensors located at appropriate locations in the processing chamber 100. The sensors 131-138 measure electrical, mechanical, and / or chemical parameters related to the operation of the processing chamber 100.

[0021] For example, sensor 131 is a VI probe for sensing voltage and / or current at the induction coil 110. Sensors 132 and 133 are flow sensors for measuring flow rates in each of the first gas channel 106A and the second gas channel 106B. Sensor 134 is a spectrometer for detecting multiple spectral bands to measure the temperature of the plasma or substrate 107. Sensor 135 is a pressure sensor for measuring the pressure in the processing chamber 100. Sensor 136 is a temperature sensor for measuring the temperature in the processing chamber 100. Sensor 137 is a gas sensor for measuring the gas concentration (or density) in the processing space 109. In one embodiment that can be combined with one or more of the above embodiments, sensor 137 is an oxygen sensor for measuring the oxygen concentration in the processing chamber 100. Based on the measured concentration of oxygen, the adjustment of oxygen over time can be calculated. Sensor 138 is another VI probe for measuring the voltage and / or current at the cathode 126. Many other sensors can be configured at appropriate locations in the processing chamber 100 to measure parameters required to identify the state of the processing chamber 100 during its operation.

[0022] The processing chamber 100 includes a controller 105 for controlling aspects of the processing chamber 100 during plasma processing. The controller 105 includes a central processing unit (CPU) 101, a memory 102, and support circuits 103 for the CPU 101. The controller 105 is coupled to different components and the processing chamber 100 and facilitates the control of the different components and the operation of the processing chamber 100. For example, the controller 105 (as shown in the figure) is coupled to sensors 131 to 138, a chemical delivery module 121, a first RF matching circuit 122 and a second RF matching circuit 124. The controller 105 can be one of any form of general purpose computer that can be used in an industrial environment to control various chambers and sub-processors. The memory 102 stores software (source or object code) that can be executed or called to control the overall operation of the processing chamber 100 in the manner described herein. The controller 105 may also include a memory 104 that stores measurements from different sensors 131 to 138. The stored measurements may include historical parameter values ​​for different processes performed in the processing chamber 100 .

[0023] The controller 105 controls the operation of controllable components in the process chamber 100. For example, the controller 105 may select and activate different groups of the plurality of sensors 131 to 138 for different types of processes, such as a conditioning process and / or a production process, such as deposition or etching, performed in the process chamber 100. The controller 105 may receive signals from the activated sensors to monitor various parameters of the performed processes.

[0024] Examples of the processing chamber 100 may include a chemical vapor deposition (CVD) chamber, such as the AKT 1000 available from Applied Materials, Inc., Santa Clara, California. ® PECVD Chambers, PRODUCER™ Chambers and PRECISION 5000 ® Chamber.

[0025] Figure 2A FIG. 2 is a schematic block diagram of a process control system 200 for detecting the endpoint of a seasoning process according to aspects of the present disclosure. The process control system 200 may be incorporated into Figure 1 The process control system 200 may be implemented in the controller 105 described herein, or may be implemented on a computing system separate from the controller 105. The process control system 200 includes an aging schedule data storage 202, a historical sensory data storage 204, a sensor interface 206, a sensory data normalizer 210, an MVA model generator 212, an MVA model regressor 214, and an MVA analyzer 216.

[0026] During processing of each substrate in the first plurality of substrates, aging progress data {R1, R2, . . . R n The seasoning progress data is stored in the seasoning progress data memory 202. The progress data in the seasoning progress data memory 202 may be progress data of a previously completed seasoning process. The seasoning progress data may include any parameter values ​​indicating the progress of the seasoning operation at different times during processing each substrate of the first plurality of substrates (e.g., 1, 2, 3, ... N). As each substrate is processed, the parameters in the seasoning progress data change as the seasoning process progresses and the processing chamber is aged.

[0027] The seasoning progress data in the seasoning process data storage 202 may be recorded and maintained sequentially or with reference to a timestamp. This enables the progress data to be analyzed sequentially to estimate the end point of the seasoning process at which the processing chamber has been fully seasoned. For example, in the case where the processing chamber performs an etching process, the seasoning progress data for each substrate may include an etching rate indicating the rate at which material is removed from the surface of the substrate. For example, the etching rate may be measured by measuring the amount of material after a period of time has been etched.

[0028] In another example, the aging progress data may also include an etch selectivity indicating the ratio of etching rates between two materials under the same conditions, and / or an undercut factor indicating the lateral extent of etching under the etching mask. For example, the etch selectivity may be measured by etching two overlapping layers and measuring the thickness of the bottom layer after etching. For example, the undercut factor may be measured by placing a resist layer over an etch structure having a length L and a width W, and measuring a signal related to a change in resistance on the resist layer.

[0029] In yet another example, the seasoning progress data may include the concentration of oxygen (O2), or nitrogen (N2), or another gas in the processing chamber during the seasoning process. The seasoning progress data storage 202 may store progress data for many seasoning processes. For example, progress data for more than ten (10) seasoning processes performed within the previous three month period may be stored in the seasoning progress data storage 202.

[0030] In one embodiment that may be combined with one or more of the above embodiments, the aging progress data for each substrate processed during the aging process may be normalized relative to the data obtained after the aging process is completed (i.e., after the aging chamber). For example, in the aging process of the etching chamber, the measured etching rate when processing the third substrate is 3.0 angstroms / minute, and the measured etching rate when the chamber is fully aging (e.g., when processing the twentieth substrate) is 10.0 angstroms / minute. In this case, the normalized aging progress data for the third substrate is 0.3 (=3 / 10).

[0031] In another embodiment that may be combined with one or more of the above embodiments, when the measured oxygen (O2) adjustment in the process chamber is less than a predetermined threshold, the process chamber may be considered fully seasoned. In that case, the measured oxygen (O2) adjustment may be normalized relative to the threshold. For example, when the measured oxygen (O2) adjustment during seasoning of the third substrate is 0.9 sccm and the threshold is 0.5 sccm, the normalized seasoning progress data for the third substrate is 0.8 (= (0.9-0.5) / 0.5).

[0032] The normalized seasoning progress data indicates how much processing chambers are seasoned for each substrate during the seasoning operation. The normalized seasoning progress data can be used as reference data to determine a set of coefficients via regression analysis as described below in conjunction with the MVA model regressor 214. In one embodiment that can be combined with one or more of the above embodiments, the seasoning progress data can be normalized before the data is stored in the seasoning progress data memory 202. In another embodiment that can be combined with one or more of the above embodiments, the seasoning progress data can be stored in the seasoning progress data memory 202 without normalization. In that case, the seasoning progress data can be normalized later by the process control system 200.

[0033] During the operation of the processing chamber, the historical sensing data storage 204 stores historical parameter values ​​detected from multiple sensors installed in the processing chamber. To obtain historical parameters of the aging process, parameter values ​​are obtained by multiple sensors during the aging process. The parameter values ​​are recorded and stored in the historical sensing data storage 204 in chronological order.

[0034] In one embodiment that may be combined with one or more of the above embodiments, each of the plurality of sensors may measure a parameter value at a point in time during processing of a particular substrate. For example, each of the plurality of sensors may detect a parameter value at a median time of processing performed on a particular substrate. In another embodiment that may be combined with one or more of the above embodiments, each sensor may detect a parameter value at a plurality of points in time during processing of a particular substrate. In this case, for a particular substrate, a maximum or minimum parameter value, or an average of parameter values ​​over time may be determined.

[0035] The parameter values ​​of the process performed in the processing chamber may include voltage, current, flow rate, spectral data, temperature, pressure, or many other characteristics associated with the aging process. Some parameter values ​​are measured in the processing chamber during the aging process. Other parameter values ​​are measured outside the processing chamber but adjacent to the processing chamber. For example, the voltage, current, or flow rate of the gas can be measured outside the processing chamber.

[0036] In one embodiment that may be combined with one or more of the above embodiments, different subsets of multiple sensors may be selected and enabled for different types of aging processes. For example, for an etching process, the process control system 200 may select and enable flow rate sensors for etching gases such as silicon tetrachloride (SiCl4) gas and chlorine (Cl2). For a deposition process, the process control system 200 may select and enable flow rate sensors for silicon-containing gases such as one or more of silane, halosilanes, or organosilanes.

[0037] During the seasoning process, a first plurality of substrates are sequentially processed in the process chamber. The first plurality of substrates may be used only for seasoning of the chamber, but not for production substrates. Each substrate of the first plurality of substrates is loaded into the process chamber and a predetermined process is performed on the substrate to season the process chamber. The process chamber is considered seasoned when a change in a measured parameter of the predetermined process stabilizes within a threshold or range.

[0038] For example, when executing a predetermined process, a current parameter is measured and compared with a previously measured parameter to determine the variance between the current parameter and the previously measured parameter. The variance of each parameter is compared with a predetermined threshold (range) for the parameter to determine whether the parameter has stabilized. For convenience in calculating the variance, the parameters of interest may be normalized so that the time curves of the parameters may be used simultaneously.

[0039] The sensing data normalizer 210 normalizes a plurality of parameters P of each substrate in the first plurality of substrates. i,j , where j=1, 2, ..., M, is a specific parameter of the substrate, and i=1, 2, ..., N. Thus, the sensing data normalizer 210 obtains a set of normalized parameter values ​​x for each substrate i=1, 2, ..., N in the first plurality of substrates.i,j During the processing of each substrate i=1,2,...,N, a number of parameters P are recorded i,j One substrate in the first plurality of substrates, substrate A (ie, i=A), is selected as the basis for normalizing the parameters of each substrate (i≠A). The parameters P of the selected substrate A,j A parameter used to normalize all other substrates in the first plurality of substrates.

[0040] In one embodiment that may be combined with one or more of the above embodiments, the sensing data normalizer 210 calculates x for each parameter j of other substrates. i,j =[(P i,j -P A,j ) / P A,j ]*100 to normalize the parameters of each of the other substrates (i≠A), where x i,j is the parameter of substrate i relative to the parameter P of the selected substrate A A,j That is, each parameter of the other substrates is normalized relative to the corresponding parameter of the selected substrate A.

[0041] The selected substrate may be any substrate in the first plurality of substrates processed during the seasoning operation. For example, the selected substrate may be the first, second, third, fourth, or fifth substrate in the first plurality of processed substrates. In one embodiment that may be combined with one or more of the above embodiments, the substrate selected as the normalization basis is selected so that all parameter values ​​P for that substrate are A,j is non-zero. For example, assume that five different sensors are selected to be activated to collect sensing data during a certain aging process. That is, for each substrate, five parameter values ​​are recorded. Assume that when processing the second substrate, only four sensors output non-zero values ​​and the fifth sensor outputs a value of substantially zero. Also assume that when processing the third substrate, all five sensors output non-zero values. In that case, the third substrate can be selected as the basis for the normalization process. Selecting a substrate with all non-zero parameter values ​​avoids potential problems of zero division when normalizing the parameter values ​​of other substrates.

[0042] In the following description, for illustrative purposes, it is assumed that the second substrate is used as a basis for normalization. That is, the sensing data normalizer 210 is described as normalizing the parameter value of each substrate in the first plurality of substrates using the parameter value measured when processing the second substrate. However, as described above, other substrates may be selected as a basis for normalization.

[0043] In one embodiment that may be combined with one or more of the above embodiments, the parameter value is normalized by calculating a deviation of the parameter value obtained when processing each substrate of the first plurality of substrates from the parameter value obtained when processing the selected substrate (i.e., the second substrate). For example, if the pressure in the processing chamber measured when processing the selected (i.e., the second) substrate is 40 Torr and the pressure measured when processing the fifth substrate is 100 Torr, the normalized chamber pressure of the fifth substrate is 2.5 (=100 / 40).

[0044] The MVA model generator 212 generates a multivariate analysis (MVA) model using the normalized parameter values. In one embodiment that may be combined with one or more of the above embodiments, the MVA model may be generated by converting a coefficient set {c k,j}, k=0,1,...,N is applied to the normalized parameter values ​​of each substrate to obtain the polynomial equation. The polynomial equation for the i-th substrate is as follows: Therefore, the value y i is a linear combination of the parameter values ​​recorded for the ith substrate. In some cases, y i is the weighted average of the parameter values ​​recorded for substrate i.

[0045] The MVA model generator 212 uses the output y of the polynomial equation i To calculate the following logistic equation:

[0046]

[0047] Where S i is the aging index of the ith substrate. The aging index indicates the progress of the aging operation. Therefore, the MVA model generator 212 will output a respective aging index for each substrate in the first plurality of substrates, wherein each aging index is a function of a set of coefficients and a set of normalized parameter values.

[0048] Given a set of aging data for a processing chamber, regression may be used to calculate the coefficients of the MVA model as described above from the aging data. Figure 2B Graph 220 illustrates the results of an exemplary regression process according to aspects of the present disclosure. Figure 2B , the vertical axis indicates an aging index representing the progress of the aging operation, and the horizontal axis indicates the sequence of substrates processed during the aging operation. For example, about 47 data points 222 are shown in graph 220. During the aging operation, each data point 222 represents normalized aging progress data {R1, R2, ... R i}.

[0049] Curve 224 is a graph showing the normalized parameter value x for each substrate in the first plurality of substrates.i,j A graph of the logistic equation based on the regression of the MVA model. The coefficient set c k,j This can be determined by performing different regression methods on the logistic equation. i The model of may be used to minimize the squared error of data point 222 relative to curve 224 .

[0050] In one embodiment that may be combined with one or more of the above embodiments, a least squares regression method may be used to fit the curve 224 to the normalized aging progress data {R1, R2, . . . R i}. The least squares regression method is used to minimize the logistic equation {S1, S2,... S i Each output of} is related to the aging progress data {R1, R2,... R i The sum of the squares of the errors between each of}.

[0051] The substantially horizontal dashed line 226 indicates the threshold aging index at which the chamber is considered fully aged. As illustrated, the dashed line 226 is depicted at an aging index of approximately 0.95. However, other indices such as 0.85, 0.9, or 0.97 may be used as the threshold aging index.

[0052] Return to reference Figure 2A , the MVA model regressor 214 performs regression analysis of the coefficient set to fit the aging progress data to a logistic equation. For each substrate number i, the MVA model regressor 214 converts each aging index S i Fitting to the corresponding aging progress data R i , to determine the coefficient set {c k,j}.

[0053] In one embodiment that may be combined with one or more of the above embodiments, the MVA model regressor 214 uses a support vector machine (SVM) model, which is a machine learning algorithm that analyzes data for classification and regression analysis. The SVM model can determine a set of coefficients that minimize the error between outputs by adjusting the position of the hyperplane to have the maximum margin from the vector of aging indices. In another embodiment that may be combined with one or more of the above embodiments, the MVA model regressor 214 uses a least squares fitting method to determine the coefficients that minimize the individual aging indices S for each substrate i. i Corresponding aging progress data R i The coefficient set of the least squares error between {c k,j In another embodiment that may be combined with one or more of the above embodiments, the MVA model regressor 214 uses another suitable regression model, such as linear regression or polynomial regression, to determine a set of coefficients that minimizes the error.

[0054] Once the coefficient set {c k,j}, the MVA analyzer 216 stores the polynomial equation , and the aging index S for each substrate is generated by applying the output of the polynomial equation to the logistic equation i In one embodiment that may be combined with one or more of the above embodiments, the MVA analyzer 216 may be configured to analyze the aging index S based on the aging index S. i Determines the endpoint of the aging operation.

[0055] When processing each substrate in the second plurality of substrates, the sensor interface 206 receives a plurality of substantially real-time parameter values ​​from the plurality of sensors. Depending on the seasoning process being performed, the substantially real-time parameter values ​​may be obtained from different sensors of the plurality of sensors. To compare the sensed data (i.e., parameter values) of the second plurality of substrates with the historical sensed data, as described above, the substantially real-time parameter values ​​include sensed parameter values ​​of the same type as the historical parameter values.

[0056] The substantially real-time parameter values ​​are provided to the sensed data normalizer 210 for normalization. The sensed data normalizer 210 normalizes the plurality of substantially real-time parameter values ​​as described above. The normalized real-time sensed data for each substrate in the second plurality of substrates is provided to the MVA analyzer 216. The MVA analyzer 216 applies the normalized real-time sensed data to the polynomial equation , to determine the output y i The MVA analyzer 216 places the output of the polynomial equation into a logistic equation to generate an aging index for each substrate in the second plurality of substrates.

[0057] In one embodiment that may be combined with one or more of the above embodiments, the MVA analyzer 216 detects the end point of the aging process based on whether the aging index of the second plurality of substrates meets the completion criteria of the aging process. The completion criteria of the aging process may be any number of substrates having an aging index greater than any threshold value. For example, the arbitrary number of substrates may be 3, and the arbitrary threshold aging index may be 0.95. That is, at least three substrates need to have an aging index higher than 0.95 to indicate that the aging process has been completed.

[0058] For example, suppose Figure 2BThe seasoning operation is shown to be performed on the second plurality of substrates, and the criteria for completion of the seasoning process have been determined. Data point 227 representing the seasoning index of the thirty-third substrate of the second plurality of substrates processed during the seasoning operation is the third substrate having a seasoning index above the threshold value 226 of 0.95. Thus, data point 227 indicates that the seasoning process of the chamber is completed when the step of processing the thirty-third substrate is completed. Thus, the MVA analyzer 216 can detect the end point of the seasoning process when the step of processing the thirty-third substrate is completed. Of course, the completion criteria of the seasoning process can be any number of substrates having a seasoning index greater than any threshold value.

[0059] Figure 3 A method 300 for detecting an endpoint of a seasoning process is illustrated according to aspects of the present disclosure. The method begins at operation 302, where a process control system for controlling a seasoning operation obtains seasoning progress data for each substrate in a first plurality of substrates. The seasoning progress data may indicate the progress of a previously performed seasoning process. The seasoning progress data may include a sensed value obtained from a sensor in a process chamber or a calculated value obtained based on the sensed value.

[0060] The seasoning progress data may be normalized relative to seasoning data measured while the process chamber was operating in a fully seasoned state prior to purging the process chamber.The process control system may be directly or remotely coupled to the process chamber in which the seasoning process is performed.

[0061] At operation 304, the process control system obtains historical parameter values ​​measured using a plurality of sensors installed in the process chamber. The parameter values ​​used to determine the progress of the seasoning operation of a particular process chamber may depend on the process performed in that chamber. That is, if there are many parameter values ​​for a particular process chamber, a subset of those parameter values ​​may correspond to a particular seasoning process performed in that process chamber. Another subset of parameter values ​​may correspond to a different seasoning process for that chamber. That is, the parameter values ​​selected to determine the seasoning state of a process chamber may be based on the particular process performed therein.

[0062] The process control system selects and activates a plurality of sensors in the process chamber for monitoring the seasoning operation. The plurality of sensors selected for monitoring may depend on the type of seasoning process performed. The process control system derives a plurality of parameter values ​​for each substrate in the first plurality of substrates processed during the seasoning process from the sensed data captured by the selected plurality of sensors. Subsequently, the process control system normalizes the parameter values ​​in the plurality of parameter values ​​for each substrate relative to the corresponding parameter values ​​associated with a particular substrate in the first plurality of substrates, such as the second substrate. In one embodiment that may be combined with one or more of the above embodiments, the particular substrate may be one of the first, third, fourth, or higher order substrates processed in the process chamber.

[0063] At operation 306, the process control system generates a multivariate analysis (MVA) model using the normalized parameter values. In one embodiment that may be combined with one or more of the above embodiments, the MVA model may be a polynomial equation obtained by applying a set of coefficients to the normalized parameter values ​​for each substrate in the first plurality of substrates. The output of the MVA model generator may be the output of the polynomial equation for each substrate in the first plurality of substrates. The output of the MVA model generator is used as an input to a logic equation that generates an output of an aging index. The aging index indicates the progress of the aging process. Each aging index is a function of the set of coefficients.

[0064] At operation 308, the MVA model is regressed with respect to the normalized aging progress data. Regression may be applied using various regression algorithms so as to minimize the sum of the errors between the output of the logistic equation and the aging progress data. In one embodiment that may be combined with one or more of the above embodiments, a support vector machine (SVM) model may be used to determine a set of coefficients that minimizes the error by adjusting the position of the hyperplane to have the maximum margin from the vector of the reference data. In another embodiment that may be combined with one or more of the above embodiments, the MVA model may be regressed using other appropriate regression models (such as linear regression or polynomial regression) to determine a set of coefficients that minimizes the error.

[0065] At operation 310, the process control system receives a plurality of substantially real-time parameter values ​​from a plurality of sensors and normalizes the plurality of substantially real-time parameter values ​​while processing each substrate in the second plurality of substrates. The normalized sensory data from operation 304 is provided to the MVA model. The output of the MVA model is input to a logic equation that generates an aging index for each substrate in the second plurality of substrates. The system detects an end point of the aging process based on a completion criterion for the aging process. The criterion may be a number of substrates of the second plurality of substrates having an aging index above a threshold.

[0066] Figure 4 The diagram is configured to implement the process control system 400 described herein, such as the method described above with respect to Figure 3 A method 300 of detecting the endpoint of an aging process is discussed.

[0067] As shown, the process control system 400 includes: a central processing unit (CPU) 402 for executing programmed instructions; a sensor interface 206 that can allow connection of different sensors; a network interface 405, which can include, for example, a transceiver for transmitting and receiving data from an external network; a memory 408, such as a volatile random access memory; a storage 410, such as a non-volatile disk drive, a RAID array, etc.; and an interconnect 412, such as a data bus. In some examples, some or all of the storage 410 can be remote from the process control system 400 and can instead be accessed via the network interface 405.

[0068] CPU 402 retrieves and executes executable instructions stored in memory 408 via interconnect 412. As shown in the figure, memory 408 includes MVA model generator 212, MVA model regressor 214, MVA analyzer 216 and sensor data normalizer 210. These components correspond to the Figure 2A The components of process control system 200 are described. Memory 408 includes program code for implementing these components.

[0069] The CPU 402 also retrieves data from the memory 410 and processes the data. The memory 410 includes the aging progress data memory 202 and the historical sensing data memory 204, as described in relation to Figure 2A describe.

[0070] The previous description is provided so that any person skilled in the art can practice the various embodiments described herein. It is obvious to those skilled in the art that various modifications can be easily made to these embodiments, and the general principles defined herein can be applied to other embodiments. For example, the functions and arrangements of the discussion elements can be changed without departing from the scope of the present disclosure. Various examples can ignore, replace, or add various programs or components as appropriate. At the same time, the features described in some embodiments or examples can be combined with some other embodiments or examples. For example, the numerous aspects described herein can be used to implement devices or practice methods. In addition, the scope of the present disclosure is intended to cover such devices or methods that use other structures, functions, or structures and functional practices other than the various aspects of the disclosure described herein. It should be understood that any aspect of the disclosure disclosed herein can be realized by one or more elements of the claims.

[0071] As used herein, the word “exemplary” means “serving as an example, instance, or illustration.” Any aspect described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects.

[0072] As used herein, the term "determine" or "determining" includes various actions. For example, "determining" may include calculating, computing, processing, deriving, investigating, searching (e.g., searching in a table, a database, or another data structure), deciding, etc. At the same time, "determining" may include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory), etc. At the same time, "determining" or "determining" may include solving, selecting, choosing, establishing, etc.

[0073] The methods disclosed herein include one or more operations or actions for implementing the methods. The operations and / or actions may be interchangeable with each other without departing from the claims. In other words, unless a specified order of operations or actions is specified, the order and / or use of specific operations and / or actions may be changed without departing from the claims. In addition, the various operations described herein may be performed by any suitable means capable of performing the corresponding functions. The 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.

[0074] The following claims are not intended to be limited to the embodiments shown herein, but to the full extent consistent with the language of the claims. Within the claims, unless otherwise stated, reference to a singular element does not mean "one and only one", but means "one or more". Unless otherwise explicitly stated, the term "some" refers to one or more. Unless the phrase "means for..." is used to expressly describe an element, the claim element is not interpreted according to the provisions of 35 USC §112 (f) of the United States Patent Act. All structural and functional equivalents of the elements throughout the various aspects of the present disclosure that are known or will later be known to those skilled in the art will be expressly incorporated herein by reference and are intended to be included in the claims.

[0075] While the foregoing is directed to embodiments of the present disclosure, other and further embodiments of the disclosure may be devised without departing from the basic scope thereof, and the scope of the disclosure is to be determined by the claims that follow.

Claims

1. A device for detecting the end point of an aging treatment, the device comprising: a processing chamber; a set of sensors coupled to the processing chamber; At least one memory configured to store: seasoning progress data indicating a progress of at least one seasoning process for each substrate in the first plurality of substrates; and historical parameter values ​​measured by a plurality of sensors selected from the set of sensors while processing each of the first plurality of substrates during the aging process; and A controller configured to: normalizing the historical parameter value for each substrate in the first plurality of substrates relative to a plurality of historical parameter values ​​for one substrate in the first plurality of substrates; generating a multivariate analysis model by applying a set of coefficients to the normalized historical parameter values ​​for each substrate of the first plurality of substrates; determining the set of coefficients by performing a regression operation on output values ​​of a logistic equation, wherein an input to the logistic equation is an output of the multivariate analysis model having normalized historical parameter values ​​for each of the first plurality of substrates; receiving a plurality of substantially real-time parameter values ​​from the plurality of sensors while processing a second plurality of substrates; and An endpoint of the aging process is determined using the multivariate analysis model having the plurality of substantially real-time parameter values.

2. The apparatus of claim 1, wherein when the processing chamber is configured to perform an etching process, the aging progress data comprises at least one of an etching rate, an etching selectivity, or an undercut factor.

3. The apparatus of claim 1, wherein the controller is further configured to select the plurality of sensors from the set of sensors to collect parameter values ​​depending on a process performed in the process chamber.

4. The apparatus of claim 1, wherein the logic equation is: ， where y i is the output of the multivariate analysis model for the ith substrate, and S i is an aging index indicating the progress of the aging process when processing the i-th substrate.

5. The apparatus of claim 1, wherein the regression operation is a least square regression that minimizes the sum of squares of errors between an output of the logistic equation and each of the aging progress data.

6. The apparatus of claim 1, wherein the regression operation is a support vector machine (SVM) regression that adjusts a position of a hyperplane to have a maximum margin from the vector of the aging progress data.

7. The apparatus of claim 4, wherein the controller is further configured to detect the endpoint of the aging process when the aging index of a particular number of substrates in the second plurality of substrates satisfies a threshold value.

8. A method for detecting an end point of an aging process in a process chamber, the method comprising the steps of: obtaining seasoning progress data indicating a progress of a seasoning process for each substrate in the first plurality of substrates; obtaining historical parameter values ​​measured by a plurality of sensors selected from a set of sensors coupled to a processing chamber while processing each substrate of the first plurality of substrates; normalizing the historical parameter value for each substrate in the first plurality of substrates relative to a plurality of historical parameter values ​​for one substrate in the first plurality of substrates; generating a multivariate analysis model by applying a set of coefficients to the normalized historical parameter values ​​for each substrate of the first plurality of substrates; determining the set of coefficients by performing a regression operation on an output of a logistic equation, wherein an input to the logistic equation is an output of the multivariate analysis model having normalized historical parameter values ​​for each of the first plurality of substrates; receiving a plurality of substantially real-time parameter values ​​from the plurality of sensors while processing each substrate of the second plurality of substrates; and An endpoint of the aging process is determined using the multivariate analysis model having the plurality of substantially real-time parameter values.

9. The method of claim 8, wherein when the processing chamber is configured to perform an etching process, the aging progress data comprises at least one of an etching rate, an etching selectivity, or an undercut factor.

10. The method of claim 8, further comprising: The plurality of sensors are selected from the set of sensors to collect depending on the type of process performed by the process chamber.

11. The method of claim 8, wherein the regression operation is a least squares regression that minimizes the sum of squares of errors between each output of the logistic equation and each of the aging progress data, respectively.

12. The method of claim 8, wherein the regression operation is a support vector machine (SVM) regression that adjusts the position of a hyperplane to have a maximum margin from the vector of the aging progress data.

13. The method of claim 8, wherein the logic equation is: ， where y i is the output of the multivariate analysis model for the ith substrate, and S i is an aging index indicating the progress of the aging process when processing the i-th substrate.

14. The method of claim 13, further comprising: The endpoint of the aging process is detected when the aging index of a specific number of substrates in the second plurality of substrates meets a threshold value.

15. A method for detecting the end point of an aging process, the method comprising the following steps: obtaining seasoning progress data indicating a progress of a seasoning process for each substrate in the first plurality of substrates; obtaining historical parameter values ​​measured by a subset of the plurality of sensors coupled to a processing chamber, each sensor of the plurality of sensors measuring a parameter of the seasoning process while processing each substrate of the first plurality of substrates; normalizing the historical parameter values ​​relative to a subset of the historical parameter values ​​associated with one substrate in the first plurality of substrates; normalizing the aging progress data for each substrate of the first plurality of substrates relative to aging progress data associated with completing an aging process; generating a multivariate analysis model by applying a coefficient to each normalized historical parameter value of the first plurality of substrates, wherein the coefficient is determined by performing a regression operation on a plurality of outputs of a logistic equation, wherein inputs to the logistic equation are outputs of the multivariate analysis model; receiving a plurality of substantially real-time parameter values ​​from the subset of sensors while processing each substrate of the second plurality of substrates; and An endpoint of the aging process is determined using the multivariate analysis model having the plurality of substantially real-time parameter values.

16. The method of claim 15, wherein the subset of the plurality of sensors is selected based on a type of process performed in the processing chamber.

17. The method of claim 15, wherein the regression operation is a support vector machine (SVM) regression that adjusts the position of a hyperplane to have a maximum margin from the vector of the aging progress data.

18. The method of claim 15, wherein the logic equation is: ， where y i is the output of the multivariate analysis model for the ith substrate, and S i is an aging index indicating the progress of the aging process when processing the i-th substrate.

19. The method of claim 18, wherein the regression operation is a least squares regression that minimizes a sum of squares of errors between the aging index of each substrate in the second plurality of substrates and each of the aging progress data.

20. The method of claim 18, further comprising: The endpoint of the aging process is detected when the aging index of a specific number of substrates in the second plurality of substrates meets a threshold value.

Images