JP2012532462A - Method and apparatus for predicting etch rate uniformity for plasma chamber verification - Google Patents

Abstract

A method for predicting etch rate uniformity is provided to verify the normality of a processing chamber during substrate processing. The method executes a recipe and receives process data from a first sensor group. The method also uses a subsystem normality check prediction model to analyze the processing data to obtain calculated data including at least one of etch rate data and uniformity data. The subsystem normality check prediction model is constructed by correcting the measurement data from the film substrate group with the processing data collected during the same processing of the non-film substrate group. The method further compares the calculated data obtained with a control limit group as defined by the subsystem normality check prediction model. This method also generates a warning if the calculated data is outside the control limit group.
[Selection] Figure 2

Description

  Advances in plasma processing have led to the development of the semiconductor industry. In today's highly competitive market, minimizing waste and producing high-quality semiconductor devices will help enhance device manufacturers' competitiveness. Therefore, in order to obtain a track record in substrate processing, generally, strict control of the processing environment is required.

  As is well known to those skilled in the art, the state of the processing chamber can affect the quality of the semiconductor device produced. Therefore, if the processing chamber can be accurately verified, the cost of ownership of the processing tool (cost of ownership) can be reduced and waste can be suppressed. For example, by accurately verifying the processing chamber, the recipe (processing method) can be adjusted in consideration of the state of the chamber. As another example, by accurately verifying the processing chamber, the processing chamber can be maintained in good operating condition, thereby extending the life of the chamber and reducing the potential for waste. As used herein, the term “verify the processing chamber” refers to the process of identifying the conditions of the processing chamber and / or the steps necessary to adapt the chamber.

  Metrology can be used to adapt the processing chamber. In the measurement method, the thickness or critical dimension (CD) of the substrate film may be measured using an actual measurement tool. As an example of a commercially available instrument capable of performing such measurement, there is an ASET-F5x thin film measuring system commercially available from KLA-Tenker Co., Ltd. The measurement may be performed before and after the substrate processing. After collecting the measurement data, the etching rate and / or CD bias value (critical dimension bias value) for the substrate may be obtained. Uniformity can be calculated from a spatial map of measured etch rate values and / or CD bias values. As described herein, uniformity can be calculated by taking the standard deviation of the etch rate and / or CD bias value.

  Metrology is an accurate way to verify the processing chamber, but is expensive and time consuming. For example, measuring the CD bias of just one substrate can take up to an hour. For this reason, the measurement is often performed after the processing of the entire substrate lot is completed, not during the substrate processing. This can damage the entire substrate lot before the problem is identified.

  The present invention will be described below with reference to the accompanying drawings, which are merely illustrative and do not limit the present invention. In the following description, the same components are denoted by the same reference numerals.

6 is a flowchart outlining a method for building a predictive etch rate model for validating a processing chamber.

FIG. 2 is a diagram illustrating an overview of a method for building a subsystem health check (SSHC) prediction model for validating a processing chamber in an embodiment of the present invention.

The figure which shows the implementation process which builds an SSHC prediction model in embodiment of this invention.

The figure which shows the local scan of a board | substrate.

Schematic which shows the uniformity of an etching rate using the concentric circle which divides | segments a board | substrate measurement point in embodiment of this invention.

6 is a flowchart illustrating an outline of a method for verifying a processing chamber using an SSHC prediction model in an embodiment of the present invention.

  Hereinafter, the present invention will be described in detail with reference to embodiments shown in the accompanying drawings. Numerous details described below are for the purpose of fully understanding the present invention. It will be apparent to those skilled in the art that the present invention may be practiced without some or all of these details. In other instances, well known process steps and / or structures have not been described in detail in order not to unnecessarily obscure the features of the present invention.

  Various embodiments including methods and techniques are described below. In addition, the present invention can be configured as a product including a computer-readable medium storing computer-readable instructions for implementing the technique of the present invention. Computer-readable media include, for example, semiconductor media, magnetic media, magneto-optical media, optical media, and other computer-readable media that can store computer-readable code. Furthermore, the present invention can be configured as an apparatus for executing an embodiment of the present invention. Such an apparatus may comprise dedicated and / or programmable circuitry that performs the tasks associated with embodiments of the present invention. Examples of such devices include general purpose computers and / or dedicated computing devices that are programmed as needed, with dedicated / programmable circuits and computers suitable for various tasks associated with embodiments of the present invention. / A combination of computing devices may be used.

  Virtual metrology can be used to verify the processing chamber. Current virtual metrology is based on, for example, a predictive model for a given processing chamber. In constructing the predictive model, the data collected during processing of the film substrate group is used to calculate the etching rate and / or CD bias (critical dimension) that can be calculated based on pre-processing measurement data and post-processing measurement data for the same film substrate group. Bias) data may be correlated to a group of on-wafer measurement values such as a spatial map.

  For ease of discussion, the flowchart of FIG. 1 outlines a method for building a predictive etch rate model for validating a processing chamber.

  In the first step 100, the process of building a prediction model is started. The predictive model can be initiated at any stage of the wet cleaning cycle.

  In the next step 102, preprocessing measurement data relating to the substrate group is acquired. A test board group is used to construct a prediction model. Usually, a film substrate group, for example, a SensArray wafer group is used as the test substrate group. A film substrate is typically an unpatterned substrate with a film layer. Prior to processing a group of substrates, pre-processing measurement data is acquired for a group of data points on each film substrate. For example, the thickness of each film substrate is measured.

  In the next step 104, the film wafer group is processed. Instead of processing the product wafer using an actual recipe (processing method), a modified version of the recipe may be used. The modified recipe is a simplified version of the production recipe and may give the same etching behavior as the production recipe. During processing, processing data is collected using a sensor group (such as an optical emission sensor, a pressure measurement sensor, a temperature measurement sensor, and a gas measurement sensor).

  After the substrate processing is completed, in the next step 106, post-processing measurement data is acquired for the processed test substrate.

  In the next step 108, the difference between the pre-process measurement data and the post-process measurement data (etch depth) for each data point may be calculated to determine the average etch rate for each film substrate.

  In the next step 110, a prediction model is constructed. The predictive model may be based on etch rate spatial map measurements and process data collected by sensors. For example, the calculated value of the average etching rate is set as the target etching rate value of the prediction model. Next, a prediction model is constructed by correlating the processing data with the target etch rate value. However, there are cases where a continuous update (update) is required even after the prediction model is constructed. Updates may occur due to changes in the state of the processing chamber during the regular maintenance cycle.

  For example, drift may occur due to variations in chamber conditions, deposition on sensors, and the like. In order to take into account drift, the prediction model may be normalized based on a predetermined set of drift values. For example, assume that after wet cleaning, the processing chamber is in an ideal state where no drift occurs. However, several weeks of drift may occur in the gas distribution subsystem after weeks of substrate processing. In order to take into account such drift, the prediction model may be adjusted accordingly.

  As another example, assume that as part of a wet clean, the chamber walls were cleaned, rubbed, and corroded hardware parts were replaced. If the prediction model is first built with the process chamber being “not clean”, the prediction model needs to be adjusted to take into account the “new” chamber state.

  Depending on chamber state variations, a compensation or moving window model may be prepared (step 112) to update the prediction model. That is, steps 102 to 108 may be repeated for a new film substrate group. The prediction model may be updated using the result of the new test operation (test run).

  Although a prediction model can be constructed by a virtual measurement method, there are several problems with the current virtual measurement method.

  First, current virtual metrology methods do not provide an accurate method for validating a processing chamber because uniformity cannot be determined from a predictive model. For example, even though the predictive model can accurately predict the average etch rate and / or CD bias of the substrate, the number given by the predictive model is only an average number. As will be apparent to those skilled in the art, the actual etch rate and / or CD bias value can vary across the substrate surface. Thus, for example, the average etch rate may not reflect the actual etch rate across the substrate surface. Therefore, uniformity cannot be determined, and as a result, the predictive model cannot always accurately verify the processing chamber.

  Another problem is that often the robustness of the prediction model depends on the granularity of the processing data collected by the sensor. Most of the processing tools include sensors that cannot provide the necessary data granularity that is considered necessary to build a robust prediction model. Even though the sensor can provide the high fidelity data needed to build a predictive model, most processing tools do not have analytical capabilities. As a result, most prediction models give an error that is greater than an acceptable level.

  In addition to the problems described above, there is also a problem that the cost of constructing and maintaining a prediction model is very high. For example, for a typical predictive model, the cost of construction and maintenance is about hundreds of thousands of dollars. Part of this cost is due to the use of expensive film substrates. Even after the prediction model is constructed, an additional cost is incurred each time the prediction model needs to be updated. Although cheaper substrates can be used, if the device manufacturer chooses to validate the processing chamber using a predictive model, it is necessary to continue to use more expensive film substrates in the production environment.

  In an embodiment of the present invention, a method for generating a subsystem health check (SSHC) prediction model for validating a processing chamber is provided. Embodiments of the present invention provide a method for building an SSHC prediction model that is applicable to cheaper substrates that are often reusable a finite number of times (such as non-film substrates). Embodiments of the present invention also provide a method for validating a processing chamber based on uniformity. Furthermore, embodiments of the present invention provide a method for executing an SSHC prediction model in a production environment.

  Although various embodiments using etch rates are described herein, the present invention is not limited to etch rates and is applicable to other process parameters such as CD bias. The following description is merely an example, and the present invention is not limited to these examples.

  In an embodiment of the present invention, a method for building an SSHC prediction model using data on at least two different types of substrates is provided. For example, on-wafer measurements for film substrates may be correlated with data collected during similar processing for non-film substrate groups. In one aspect of the present invention, if a relationship is established between two data groups, measurement data obtained from one type of substrate can be correlated with sensor data obtained from a second type of substrate. The inventors have found that this is possible.

  For example, it is assumed that there is a relationship between measurement data relating to a film substrate group and sensor data relating to the same film substrate group. In the prior art, this relationship is used as the basis of the prediction model. If the same modified recipe is applied to a non-film substrate group (eg, a cheaper bare silicon substrate), the data is collected in the same processing environment, so a correlation can be established between the two sensor data groups. By performing the replacement, a correlation can be established between the measurement data of the film substrate group and the sensor data of the non-film substrate group. Based on this correlation, an SSHC prediction model for validating the processing chamber can be constructed from sensor data collected for non-film substrates.

  In addition, data may be collected during different periods of the wet cleaning cycle to eliminate drift and / or noise in the SSHC prediction model. For example, the SSHC prediction model may be constructed based on data collected at the start of the wet cleaning cycle, during the wet cleaning cycle, and just before the end of the wet cleaning cycle. Therefore, the SSHC prediction model (unlike the conventional prediction model) is constructed taking into account wet cleaning of the processing chamber, so there is no need to update each time the processing chamber is wet cleaned. Furthermore, using similar data groups obtained between different chambers with the same hardware configuration, variations between chambers (due to installation, variations between sensors, etc.) can be identified and eliminated.

  As described above, one of the problems with the conventional prediction model is that the prediction model is generally based on data that may have insufficient granularity. In order to provide data necessary for constructing a robust SSHC prediction model, a sensor capable of collecting highly granular data may be used. Examples of such sensors include, but are not limited to, VI probe sensors, OES sensors, pressure sensors, and the like.

  The SSHC prediction model may be constructed by performing data processing using a robust data analysis module with a higher granularity and a larger amount of data. In one embodiment, the robust data analysis module is a fast processing computing engine that can be configured to handle large amounts of data. In addition, a robust data analysis module is configured to receive process data directly from the sensor instead of relaying data via the production equipment host controller or relaying data via the process module controller. May be. Application No. 12 / 555,674 filed September 8, 2009 by Huang et al. Describes an example of an analysis computer suitable for performing the analysis.

  In one embodiment, uniformity may be predicted using an SSHC prediction model. Those skilled in the art will appreciate that the etch rate is not always uniform across the entire substrate surface. Many factors affect uniformity. For example, the angle at which the gas is distributed within the processing chamber can affect uniformity. As another example, energization within the processing chamber can affect uniformity.

  Although the etch rate may not be uniform across the substrate surface, empirical evidence has shown that the etch rate is approximately the same for a given area of the substrate. In one embodiment of the present invention, it is assumed that (in an abstract sense) the substrate is divided into three concentric circles, and the regions within each concentric circle are empirically provided with the same uniformity. In one embodiment, the uniformity can be calculated from the etch rate of the processed substrate. First, the average etching rate in each concentric circle is obtained. Multiply each average etch rate by the number of data points measured in concentric circles (in the case of non-film substrates that do not have measurement measurements, the measurement points assumed by the metrology tool ("virtual" points)) Multiply the numbers (see Figure 4)). The values for all three concentric circles can be added to calculate the average etch rate of the substrate. Next, the uniformity is obtained by calculating the standard deviation of the average etching rate of each concentric circle with respect to the total average etching rate of the substrate. Next, the uniformity of the entire substrate is determined by calculating the standard deviation of all actual or “virtual” etch depths and calculating the ratio to the average etch rate.

  After the SSHC prediction model is constructed, the SSHC prediction model may be moved to the production process. Since the SSHC prediction model is constructed based in part on data collected from non-film substrates, the cost of implementing the SSHC prediction model in a production environment is significantly lower than conventional prediction models. One reason for significant cost reduction is that the SSHC prediction model can be applied to process data collected from cheaper non-film substrates. Furthermore, the reduction in the number of measurement requests due to the small number of measurement tools required to satisfy the production conditions also leads to cost reduction. Since the chamber verification response time (turnaround time) is short, problems occurring in the processing chamber can be detected more quickly, and the number of production wafers that may be problematic can be reduced. Therefore, the SSHC prediction model provides an effective model for verifying the processing chamber and can effectively suppress the cost of ownership (cost of ownership).

  The features and advantages of the present invention will be further understood from the following description with reference to the drawings.

  FIG. 2 outlines a method for building a subsystem health check (SSHC) prediction model for validating a processing chamber in an embodiment of the present invention.

  In the first step 202, a pre-processing measurement is performed on the film substrate group. Similar to the prior art, a measurement tool (such as a KLA-Tenker thin film measurement tool) may be used to perform measurements such as measuring the thickness of the substrate. Since the thickness of the substrate may vary across the substrate, measurements may be taken at different data points on the substrate (eg, a magnetic pole scan is performed at 49 data points on the substrate 402 shown in FIG. 4A). Do).

  In the next step 204, the film substrate group is processed. Similar to the prior art, a modified recipe may be used in the test environment. The modified recipe is a simplified version of the production recipe and simulates the etching behavior of the production recipe.

  As is well known to those skilled in the art, parameter variations can affect the average etch rate and / or uniformity of the substrate. Therefore, in order to accurately verify the processing chamber, it is necessary to take into account parameter variations when constructing the SSHC prediction model. In order to improve the robustness of the model, the system response to changes in the production environment is evaluated. Recipe parameters may be varied when processing one or more test substrates. For example, the first few substrates (e.g., three substrates) may be run with a basic modification recipe, and the fourth and fifth substrates may change pressure values and occur in the processing chamber. Take into account some pressure level fluctuations. In another example, the gas flow rate toward the center of the substrate can be adjusted to offset the slight increase in edge etch rate that can be caused by corrosion of chamber consumables (as seen near the edge of the substrate). The gas flow distribution for the next four substrates may be adjusted to increase.

  After processing the film substrate group, the next step 206 may process the non-film substrate group (eg, bare silicon substrate group) using the same modified recipe as applied to the film substrate group. For example, when the pressure value rises with respect to the fourth film substrate, the same pressure value is applied to the fourth non-film substrate. A sensor may collect processing data while processing a non-film substrate group. In one embodiment, the order of processing of the film substrate group and the non-film substrate group is not particularly limited. In other words, either step 204 or step 206 may be performed first.

  As mentioned in the prior art, one of the problems with conventional methods has been due to the granularity of the data collected. In one embodiment, the methods described herein are applied to processing tools that hold sensors (eg, VI probe sensors, OES sensors, pressure sensors, etc.) that can collect highly granular data. Furthermore, a high-speed processing calculation analysis module may be mounted to process and analyze data at high speed. In one embodiment, the high-speed processing calculation analysis module may be an advanced process and equipment control system (APECS). The APECS module may be configured to analyze multiple data at high speed (in situ). Also, the APECS module provides feedback to the processing module (PM) controller of the processing chamber so that the PM controller can predict the etch rate and / or uniformity for the next substrate to be loaded. May be. Application No. 12 / 555,674 filed September 8, 2009 by Huang et al. Describes an example of an analysis computer suitable for performing the analysis.

  After completion of the substrate processing, post-processing measurement data is acquired in the next step 208. For example, post-processing measurement data may be collected at the same set of data points (eg, as shown in FIG. 4A) for each processed film substrate.

  In the next step 210, the difference between the pre-processing measurement data and the post-processing measurement data may be calculated so that the average etch rate and / or uniformity is calculated for each film substrate. Details regarding the uniformity will be described later with reference to FIGS.

  In one embodiment, the steps of building the SSHC prediction model (202-210) may be repeated at least twice to further eliminate noise and / or drift. In one embodiment, these steps are performed at the beginning of the wet cleaning cycle (ie, after performing maintenance on the processing chamber) and just before the end of the wet cleaning cycle (ie, before the next maintenance is performed). You may make it do. Furthermore, data may be collected even during the wet cleaning cycle.

  After collecting the data, the next step 212 may correlate the sensor process data with the measurement data to build an SSHC prediction model for validating the process chamber. In one embodiment, the SSHC prediction model may be based on a partial least square model. The partial least square model is a technique for finding a relationship between two data groups. The partial least square model is used for the same purpose as the least square linear fitting method, but is usually used when there are a plurality of independent variables (input matrix X) and possibly a plurality of dependent variables (input matrix Y). It is done. In the partial least square model, the Y variable is not continuous but is composed of independent discrete values or discrete classes. An analysis is performed to find a linear combination of X variables used to classify the input data into one of the discrete classes.

  As can be seen from FIG. 2, a method is provided for building an SSHC prediction model taking into account the state of the processing chamber at different periods during the wet cleaning cycle. As described in FIG. 2, even if the time taken to construct the SSHC prediction model is long (compared to the method shown in FIG. 1), in the case of the SSHC prediction model, after the construction (which was necessary in the conventional method) There is no need for frequent updates. Thus, the resources required to build an SSHC prediction model are usually temporary costs and do not require ongoing expenses (which were necessary with conventional methods). In addition, since SSHC prediction models can be applied to data collected from non-film substrates, device manufacturers do not have to continually use more expensive film substrates in production environments to effectively apply SSHC prediction models. The cost of ownership (cost of ownership) can be greatly reduced.

  FIG. 3 shows an implementation process for constructing an SSHC prediction model in the embodiment of the present invention.

  In a first step 302, a first data group is collected. In one embodiment, a first set of data is collected at the beginning of a wet cleaning cycle (after performing processing chamber maintenance). The first data group may include processing data collected by the sensor with respect to the non-film substrate group and measurement data collected with respect to the first film substrate group. That is, steps 202 to 210 in FIG. 2 are executed for the non-film substrate group and the first film substrate group. In one embodiment, the first film substrate group and the non-film substrate group may have the same number of substrates.

  In the next step 304, a second data group is collected. In one embodiment, step 304 is performed as needed. The data collected in step 304 may be used as verification data. In order to account for possible drift, step 304 is typically performed in the middle of a wet cleaning cycle. For example, some hardware components (such as a gas distribution system) may experience drift after multiple production runs (production runs).

  Unlike film substrates, non-film substrates may be processed multiple times. As is well known to those skilled in the art, in the case of a non-film substrate (such as a bare silicon substrate), it can be processed at least about 10-15 times until it cannot withstand further processing. Thus, the same non-film substrate group that was etched in step 302 can be processed in step 304 and further in step 306.

  In the next step 306, a third data group is collected. In one embodiment, the third set of data is collected just before the end of the wet cleaning cycle. The third data group includes processing data collected by the sensor with respect to the non-film substrate group (which may be the same as the non-film substrate group in step 304) and measurement data collected with respect to the third film substrate group. Including.

  By collecting data at different periods between wet cleaning cycles, the SSHC predictive model makes it possible to capture the behavior of the processing chamber through a normal wet cleaning cycle.

  In the next step 308, the system performs a check to determine if sufficient model setting data has been collected. As described above, the model setting data means the data collected in step 302 and step 306.

  If sufficient model setting data has not been collected, in the next step 310, further data collection is performed before starting the analysis. For example, not enough data is collected if the number of processed film substrates is insufficient. As another example, it may be determined that some of the processing data collected by the sensor is unacceptable and cannot be used to build an SSHC prediction model.

  On the other hand, if sufficient model setting data has been collected, in the next step 312, the system may check whether a control limit is defined. In one embodiment, the control limit is a user-acceptable error range. Since the SSHC prediction model is constructed based on high granularity data, the control limit is set to a low value of 2 to 3%, for example.

  If the control limit has not been set, the system asks to define the control limit in the next step 314.

  On the other hand, if the control limit is defined, in the next step 316, the system starts the SSHC prediction model construction process. The SSHC prediction model may include an average etch rate as a target etch rate correlated to sensor process data.

  The target etching rate may be calculated using model setting measurement data. For example, assume that the pretreatment measurement at a substrate position was about 500 nm. After processing the substrate, the thickness of the substrate at the same substrate position is 375 nm. If the etching depth is defined as the difference between the pre-processing measurement value and the post-processing measurement value, the etching depth at a predetermined data point (for example, a radius of 115 mm in the direction of 0 degree) is 125 nm. . If the processing time of the substrate is 2 minutes, the etching rate at the data point (substrate position) is 62.5 nm per minute. After determining the etch rate, each data point on the substrate can be associated with an etch rate.

  Unlike the prior art, the average etch rate of the substrate is not determined by summing all the etch rate values and dividing the total etch rate value by the number of data points. Instead, the average etch rate of the substrate is based on the idea that the etch rate is not the same across the substrate. Empirical testing suggests that when the substrate is divided into three concentric circles (as shown in concentric circles 450, 440 and 430 in FIG. 4B), the etch rate at each data point within a concentric circle is the same. it can.

  For ease of discussion, the average etch rates in concentric circle 450 (ER1), concentric circle 440 (ER2) and concentric circle 430 (ER3) are 62.5 nm / min, 72.5 nm / min and 82.5 nm / min, respectively. Assume that After obtaining the three average etching rates, each average etching rate may be normalized. For example, there are nine data points in the concentric circle 450. Therefore, multiply the average etch rate (62.5 nm / min) by 9. After normalizing the average etch rate, the three normalized average etch rates are summed together and the total sum is divided by the number of data points (in this example, 49 data points were taken). Calculate the etch rate. In this example, the overall average etch rate of the substrate is 75.6 nm / min.

  After calculating the average etching rate, the uniformity may be calculated. As described above, the etching rate value of the substrate can be divided into three concentric circles. That is, the etch rate at data points within one concentric circle is substantially the same. For this reason, the substrate is substantially uniform within one concentric circle. Accordingly, the uniformity of the concentric circles can be determined by calculating the standard deviation of the average etching rate of one concentric circle with respect to the total average etching rate of the substrate.

  In the next step 318, the model predicted etch rate and uniformity may be calculated using the collected sensor processing data. The correlation established previously between sensor data and measurement information (ie, partial least square modeling) can be used to predict etch rate and uniformity for non-film substrates.

  In the next step 320, the system checks to determine if verification data has been collected. As described above, the verification data may be collected at step 304. If verification data is collected, the next step 322 calculates an average etch rate and uniformity based on the verification data.

  In the next step 324, the etching rate and uniformity predicted by the SSHC model for the validation data set are compared with the actual measured measurements. This step can be used to verify the validity of the SSHC prediction model.

  After making the comparison, the system proceeds to the next step 326. If the verification data does not exist, the system proceeds to the next step 326.

  In the next step 326, the system compares the predicted error rate to the control limit. For example, if the control limit is 3%, the prediction error rate must be 3% or less.

  If the prediction error is outside the allowable range, in the next step 328, a warning is issued to inform the user that the SSHC prediction model needs to be adjusted.

  On the other hand, if the prediction error is within an allowable range, in the next step 330, the SSHC prediction model may be moved to the production process and used for verification of the processing chamber.

  The flowchart of FIG. 5 outlines a method for validating a processing chamber by applying an SSHC prediction model in one embodiment. A subsystem normality check test may be performed before the production process (production run). That is, the SSHC prediction model may be applied to data collected from the test substrate to determine the normal state of the processing chamber.

  In an initial step 504, the user activates the SSHC prediction model. User specifications (recipe, file name, etc.) may be input. This information may be extracted from the library 502. The library 502 may include recipe-specific parameters (control upper limit, control lower limit, target etching rate, uniformity, etc.).

  In the next step 506, the system may perform a preliminary assessment. The preparation state of the processing chamber may be determined by preliminary evaluation.

  If the preliminary evaluation fails (eg, out of a predetermined threshold range), the system may determine the cause of the warning (failure) at the next step 526. For example, the cause of the warning may be connectivity, frequency variation, and / or temperature change (step 528).

  If the cause of the warning is connectivity, one or more sensors may not be connected correctly. If the sensor is not properly connected, the sensor cannot collect processing data.

  Another cause of warning is a large variation in frequency. For example, the frequency recorded by an electrical probe (such as a VI probe sensor) is compared with the frequency recorded by a generator. If the difference between the two frequencies is greater than or equal to a threshold (eg, very large), there may be a problem with the VI probe sensor. For example, the temperature of the VI probe sensor is too high, or the VI probe sensor is not functioning.

  Another source of warning is a large temperature change between the current temperature in the processing chamber and the desired set temperature (ie, recipe temperature). For example, when the processing tool is first switched on, it may take several minutes for the temperature in the processing chamber to reach the desired set temperature. The system performs a check to ensure that the temperature in the processing chamber is within a desired threshold before processing begins.

  If the preliminary evaluation is passed, the system checks at next step 508 whether a SSHC prediction model exists.

  If no SSHC prediction model exists, the next step 522 builds the SSHC prediction model (as described in FIGS. 2 and / or 3).

  On the other hand, when the SSHC prediction model exists, the substrate is processed and processing data is collected. For example, an SSHC prediction model is applied to the process data to predict etch rate and / or uniformity (calculated data) for the current substrate (step 510).

  In the next step 512, the system checks to see if the etch rate and uniformity are within control limits. If the etch rate and / or uniformity is not within the control limits, a failure notification is issued at the next step 514. A failure notification may provide details regarding problems that may exist in the processing chamber. For example, the failure notice may indicate that the recipe needs to be adjusted to account for drift in the processing chamber. As another example, a failure notice may indicate that the temperature has suddenly increased during substrate processing, such as because there is insufficient coolant to control the chamber temperature.

  If the etch rate and uniformity are within control limits, in the next step 516, the system checks to see if the SSHC verification degree has been met. For example, the user may set a limit that indicates that the processing chamber is in good working condition if the etch rate and uniformity for three consecutive substrates are within control limits.

  If the SSHC verification degree is not met, at the next step 518, the system continues processing with the next substrate (returns to step 510). That is, steps 510 to 518 are repeated. In one embodiment, steps 510-518 may be repeated until the SSHC verification degree is met. In another embodiment, if the SSHC verification degree cannot be met with a predetermined number of boards, a failure notification is issued to the user to inform the user about a problem that may exist in the system. Also good.

  On the other hand, if the SSHC verification degree is satisfied, in the next step 520, the processing chamber passes the subsystem normality check and can start the production process (production run).

  As can be appreciated from the above description, the processing chamber can be verified in various ways. By extrapolating data from two or more test runs (test runs) during a wet cleaning cycle, an SSHC prediction model can be constructed that takes into account the possible variations that may occur in the wet cleaning cycle . The robust SSHC prediction model becomes an effective model for verifying the processing chamber and can reduce the cost of ownership (cost of ownership).

  Although the present invention has been described according to some preferred embodiments, various changes, modifications, substitutions, and the like are possible within the scope of the gist of the present invention. The various embodiments and examples described above are merely examples and do not limit the present invention.

  The names and summary of the invention are for convenience only and are not intended to interpret the claims. Further, the summary of the invention describes a very simplified form and is for convenience, and does not interpret or limit the entire invention described in the claims. As used herein, the term “group” is used in its commonly understood mathematical sense and includes 0, 1 or 2 or more. The method and apparatus of the present invention can be implemented with various modifications. The scope of the claims set forth below includes various modifications, substitutions, and equivalents within the scope of the present invention.

Claims (20)

A method for predicting the uniformity of an etching rate in order to verify the normality of a processing chamber when processing a substrate group.
Performing a recipe on a first substrate of the substrate group;
Receiving the processing data from the first sensor group when executing the recipe;
Analyzing the processing data using a subsystem normality check prediction model to obtain calculation data, wherein the calculation data includes at least one of etching rate data and uniformity data, and the subsystem normality check prediction The model is constructed by correlating a first data group and a second data group, the first data group including measurement data from a film substrate group, and the second data group is a non-film substrate group. Including process data collected when processing
Performing a comparison between the calculated data relating to the first substrate and a control limit group defined by the subsystem normality check prediction model;
Generating a warning if the calculated data is outside the range of the control limit group;
A method comprising: The method of claim 1, further comprising:
Performing a preliminary assessment to determine a readiness state of the processing chamber. The method of claim 2, further comprising:
A method comprising determining a cause of a problem when the preliminary evaluation is outside a predetermined threshold range. The method of claim 1, further comprising:
Extracting data from the library to support the subsystem health check prediction model. The method of claim 1, comprising:
The method, wherein the control limit group is user-configurable. The method of claim 1, further comprising:
When the calculated data is within the range of the control limit group, the validity of the verification degree is confirmed, and whether a predetermined number of boards have passed the comparison between the calculated data and the control limit group. A method comprising the step of determining whether or not. The method of claim 1, comprising:
The subsystem health check prediction model is constructed from data collected over a period of time, which is determined at the beginning of the wet cycle, during the wet cycle, and at the end of the wet cycle. A method that is any one. The method of claim 7, comprising:
The method, wherein the subsystem health check prediction model is constructed from data collected over a period of time. The method of claim 7, comprising:
The method, wherein the subsystem health check prediction model is constructed from data collected over at least two time periods. A processing chamber health check device for verifying a processing chamber of a plasma processing system, comprising:
A preliminary evaluation module for determining a preparation state of the processing chamber;
A library for storing at least one of recipes and recipe parameters;
A subsystem health check prediction model,
Receive processing data from the first sensor group during substrate processing,
Analyzing the processing data to obtain a calculation data group including at least one of etching rate data and uniformity data;
Comparing the calculated data group with a predetermined control limit group;
A subsystem health check prediction model configured to generate a warning if the calculated data group is outside the range of the predetermined control limit group;
A processing chamber normality check apparatus comprising: The processing chamber normality check apparatus according to claim 10,
In the subsystem normality check prediction model, if the calculation data is within the range of the control limit group, the validity of the degree of verification is confirmed, and a predetermined number of boards are connected to the calculation data and the calculation data. A processing chamber normality check device configured to determine whether the comparison with a control limit group has been passed. The processing chamber normality check apparatus according to claim 10,
The subsystem health check prediction model is
Obtaining a first data group from a first substrate group;
Collect the second data group from the second substrate group,
Constructed by correlating the first data group with the second data group;
The first substrate group is a film substrate group, the first data group includes pre-processing measurement data and post-processing measurement data, and the post-processing measurement data is a recipe for the first substrate group. Collected after execution,
The processing chamber normality check apparatus, wherein the second data group is collected from a sensor group when the same data as the recipe is executed. The processing chamber normality check apparatus according to claim 10,
The subsystem health check prediction model is constructed from data collected over a period of time, which can be any of the start of the wet cycle, during the wet cycle, and at the end of the wet cycle. One is a processing chamber normality check apparatus. The processing chamber normality checking apparatus according to claim 13,
The processing chamber health check device, wherein the subsystem health check prediction model is constructed from data collected in one period. The processing chamber normality checking apparatus according to claim 13,
The processing chamber health check apparatus, wherein the subsystem health check prediction model is constructed from data collected in at least two periods. A product comprising a program storage medium storing a computer-readable code,
The computer readable code is configured to predict etch rate uniformity to verify the normality of a processing chamber when processing substrates;
Code for executing a recipe on the first substrate of the substrate group;
A code for receiving processing data from the first sensor group when executing the recipe;
Analyzing the processing data using a subsystem normality check prediction model to obtain calculation data, wherein the calculation data includes at least one of etching rate data and uniformity data, and the subsystem normality check prediction The model is constructed by correlating a first data group and a second data group, the first data group including measurement data from a film substrate group, and the second data group is a non-film substrate group. Code containing processing data collected when processing
A code for performing a comparison between the calculation data relating to the first substrate and a control limit group defined by the subsystem normality check prediction model;
A code for generating a warning when the calculated data is outside the range of the control limit group. The product of claim 16, further comprising:
A product comprising code for performing a preliminary assessment to determine the readiness of the processing chamber. The product of claim 16, further comprising:
A product comprising code that pulls data from a library to support the subsystem health check prediction model. The product of claim 16, further comprising:
When the calculated data is within the range of the control limit group, the validity of the verification degree is confirmed, and whether a predetermined number of boards have passed the comparison between the calculated data and the control limit group. A product with code to determine whether The product of claim 16, comprising:
The subsystem health check prediction model is constructed from data collected over a period of time, which is determined at the beginning of the wet cycle, during the wet cycle, and at the end of the wet cycle. A product that is one of them.

Images