KR20120047871A - Methods and apparatus to predict etch rate uniformity for qualification of a plasma chamber - Google Patents

Abstract

A method of predicting etch rate uniformity is provided to qualify a healthy state of a processing chamber during substrate processing of substrates. The method includes executing a recipe and receiving processing data from a first set of sensors. The method further includes analyzing the processing data using the subsystem health check prediction model to determine the calculated data, wherein the calculated data includes at least one of etch rate data and uniformity data. The subsystem health confirmation prediction model is constructed by correlating measurement data from a set of film substrates with processing data collected during similar processing of a set of non-film substrates. The method also includes performing a comparison of the calculated data against a set of control limits defined by the subsystem health confirmation prediction model. The method further includes generating a warning if the calculated data is outside the set of control limits.

Description

METHODS AND APPARATUS TO PREDICT ETCH RATE UNIFORMITY FOR QUALIFICATION OF A PLASMA CHAMBER}

Advances in plasma processing have provided growth in the semiconductor industry. In today's competitive market, the ability to minimize waste and produce high quality semiconductor devices gives device manufacturers a competitive edge. As such, tight control of the processing environment is generally required to achieve satisfactory results during substrate processing.

One skilled in the art knows that the conditions of the processing chamber can affect the quality of the semiconductor device being produced. Thus, the ability to correctly qualify a processing chamber may reduce the cost of owning a processing tool and reduce waste. In one example, by correctly qualifying the processing chamber, the recipe may be adjusted to account for chamber conditions. As another example, by correctly qualifying the processing chamber, the processing chamber can be maintained in good working conditions, extending the life of the chamber and reducing the potential for waste. As described herein, the term “qualify the processing chamber” refers to a process that identifies the conditions of the processing chamber and / or the actions required to bring the chamber into compliance.

Metrology methods may be used to qualify processing chambers. In metrology methods, actual metrology tools may be used to perform measurements such as film thickness or critical dimensions (CD) of the substrate. An example of a commercially available instrument capable of making such measurements is the ASET-F5x thin film metrology system from KLA-Tencor. Metrology may be performed before and after the substrate is processed. After the metrology data has been collected, the etch rate and / or CD bias value for the substrate can be determined. From the spatial map of the measured etch rate and / or CD bias values, uniformity can be calculated. As described herein, uniformity may be calculated by taking the standard deviation of the etch rate and / or CD bias value.

Although metrology methods can provide an accurate method of qualifying a processing chamber, metrology methods can be an expensive and time consuming procedure. In one example, the task of measuring the CD bias of only one substrate may take up to an hour. As a result, most of the metrology may be done after the substrate lot has been processed instead of between the substrates. For this reason, the entire substrate lot may be damaged before the problem can be identified.

The invention is illustrated by way of example and not by way of limitation in the figures of the accompanying drawings in which like reference numerals refer to like elements.
1 shows a simple flow diagram illustrating a method of constructing a predictive etch rate model for qualifying a processing chamber.
2 illustrates a general overview of a method of constructing a System Health Check (SSHC) prediction model for qualifying a processing chamber, in an embodiment of the invention.
3 illustrates an implementation of constructing an SSHC prediction model, in an embodiment of the invention.
4A shows a pole scan of a substrate.
4B shows a simple diagram showing etch rate uniformity using concentric circles to partition substrate metrology points, in an embodiment of the invention.
FIG. 5 shows a simple flow diagram illustrating a method of applying an SSHC prediction model to qualify a processing chamber, in an embodiment.

The invention is now described in detail with reference to some embodiments as shown in the accompanying drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to one skilled in the art that the present invention may be practiced without some or all of these specific details. In other instances, well known process steps and / or structures have not been described in detail in order not to unnecessarily obscure the present invention.

Various embodiments are described below, including methods and techniques. The present invention covers an article of manufacture comprising a computer readable medium having stored thereon computer readable instructions for performing embodiments of the inventive technique. Computer-readable media may include, for example, semiconductor, magnetic, magneto-optical, optical, or other forms of computer-readable media that store computer readable code. In addition, the present invention may also cover devices for practicing embodiments of the present invention. Such an apparatus may include dedicated and / or programmable circuits for performing tasks belonging to embodiments of the present invention. Examples of such apparatus include a general purpose computer and / or a dedicated computing device, if properly programmed, and a combination of computer / computing device and dedicated / programmable circuits adapted to various tasks belonging to embodiments of the present invention. You may.

Virtual metrology methods may be used to qualify the processing chamber. Current virtual metrology methods may be based on predictive models for specific processing chambers. To construct a predictive model, data collected during processing of a set of film substrates and / or CD may be calculated based on pre-processed and post-processed measurement data for the same set of film substrates. It may be correlated for a set of on-wafer measurements, such as spatial maps of bias data.

To facilitate the discussion, FIG. 1 shows a simple flow diagram describing how to construct a predictive etch rate model that qualify a processing chamber.

In a first step 100, a process of constructing a prediction model is initiated. The predictive model can start at any stage during the wet clean cycle.

In a next step 102, measurement data before processing is obtained for a set of substrates. To construct the predictive model, a set of test substrates is used. A set of test substrates is typically a set of film substrates or SensArray wafers. Typically, the film substrate is an unpatterned substrate with a film layer. Before the substrates process the set, measurement data before processing is obtained for a set of data points on each film substrate. In one example, the thickness of each film substrate is measured.

In a next step 104, a set of film wafers is processed. Instead of using the actual recipe used to process the product wafers, a modified version of the recipe may be used. The modified recipe may be a simpler version of the recipe, and may exhibit the same etching behavior as the recipe. During the process, a set of sensors (such as light emission sensor, pressure measurement sensor, temperature measurement sensor, gas measurement sensor, etc.) are used to capture processing data.

After substrate processing ends, in the next step 106, post-processing measurement data is obtained for the processed test substrates.

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

In a next step 110, a predictive model is constructed. The prediction model may be based on measured spatial maps of etch rates and processing data collected by the sensors. In one example, the calculated average etch rates are set as target etch rate values in the prediction model. The processing data is then correlated to the target etch rate values to form a predictive model. However, even after the prediction model is constructed, the prediction model may still require constant updates. Updates may occur due to changing conditions of the processing chamber during the course of the scheduled maintenance cycle.

In one example, drift may occur due to varying chamber conditions, deposition on sensors, and the like. To account for drift, the prediction model may be normalized based on a predetermined set of known drift values. In one example, after the wet clean, the processing chamber may be in an ideal state where no drift has occurred. However, after several weeks of substrate processing, the gas distribution system may have experienced several percent drift. To account for that drift, the prediction model may be adjusted accordingly.

In another example, as part of the wet cleaning, the chamber walls may have been cleaned and cleaned and the corroded hardware parts may have been replaced. If the predictive model was originally constructed when the processing chamber was "not clean", the predictive model may need to be adjusted to account for the "new" chamber conditions.

Due to varying chamber conditions, a compensating or moving window model may be provided to update the predictive model (step 112. That is, steps 102-108 may be applied to a new set of film substrates). Thereafter, the results from the new test run may be used to update the prediction model.

Although the predictive model may be constructed from virtual metrology methods, there are some limitations to current virtual metrology methods.

First, current virtual metrology methods do not provide an accurate way to qualify a processing chamber because uniformity cannot be determined from a predictive model. Although the prediction model cannot accurately predict the average etch rate and / or CD bias for the substrate, the values provided by the prediction models are still only average values. One skilled in the art knows that actual etch rates and / or CD bias values may vary across the surface of the substrate. For this reason, for example, the average etch rate may not represent actual etch rate values across the surface of the substrate. Therefore, uniformity cannot be determined. As a result, the predictive model may not always qualify the processing chamber correctly.

Another limitation is that the robustness of the predictive model very often depends on the granularity of the processing data collected by the sensor. Most processing tools have sensors that cannot provide the necessary data granularity that may be required to construct robust predictive models. Although sensors are available to provide high fidelity data that may be needed to construct predictive models, most processing tools lack the ability to perform an analysis. As a result, most predictive models provide more errors than desired.

In addition to the limitations described above, the cost of constructing and maintaining a prediction model can be quite expensive. In one example, a typical predictive model may cost about several hundred thousand dollars to construct and maintain. The cost is due in part to the expensive film substrates being used. Even after the prediction model is constructed, additional costs are incurred each time the prediction model is updated. In addition, even if less expensive substrates are available, if the device manufacturer wants to use the predictive model in qualifying the processing chamber, the device manufacturer may be required to continue using the more expensive film substrate in the manufacturing environment.

In accordance with embodiments of the present invention, a method is provided for generating a subsystem health check (SSHC) prediction model for qualifying a processing chamber. Embodiments of the present invention include methods for constructing an SSHC prediction model that in most cases may be applied to less expensive substrates (such as non-film substrates) that can be reused a finite number of times. Embodiments of the invention also include methods of qualifying a processing chamber based on uniformity. Embodiments of the invention also include methods for implementing an SSHC prediction model in a manufacturing environment.

In this document, various implementations may be discussed using an etch rate as an example. However, the present invention is not limited to the etching rate and may be applied to other process parameters such as, for example, CD bias. Instead, the discussion is meant by way of illustration, and the invention is not limited by the examples provided.

In one embodiment of the present invention, a method is provided in which an SSHC prediction model is built using data from at least two different substrate types. In one example, on-wafer measurements for a film substrate may be correlated to data collected during similar processing of a set of non-film substrates. In one aspect of the invention, the inventors have found that measurement data from one type of substrate may be correlated to sensor data from a second type of substrate if a predetermined relationship is established between the two data sets. .

In one example, there is some relationship between metrology data for a set of film substrates and sensor data for the same set of film substrates. This relationship is the basis for prior art predictive models. If the same modified recipe is applied to a set of non-film substrates (such as less expensive bare silicon substrates), a correlation may be established between the two sets of sensor data because the data is being collected in the same processing environment. have. Through substitution, a correlation may be established between the measurement data of the set of film substrates and the sensor data for the set of non-film substrates. Based on this correlation, an SSHC prediction model that qualify the processing chamber may be constructed from sensor data collected from non-film substrates.

To further remove drift and / or noise in the SSHC prediction model, in one embodiment, data may be collected at different periods within the wet clean cycle. In one example, the SSHC prediction model may be constructed based on data sets collected at the start of the wet clean cycle, in the middle of the wet clean cycle, and at the end of the wet clean cycle. Thus, the SSHC prediction model (unlike the prior art prediction model) does not need to be updated every time the processing chamber undergoes a wet clean, since the SSHC prediction model already describes such a situation. In addition, similar data sets between different chambers with the same hardware configuration can be used to identify and eliminate interchamber variations (such as variations caused by installation and intersensor variations).

As mentioned previously, one of the limitations of the prior art predictive model is that the predictive model is always based on data that may lack granularity. In order to provide the required data needed to construct a robust SSHC prediction model, sensors capable of collecting highly granular data may be used. Examples of sensors include, but are not limited to, for example, a VI probe sensor, an OES sensor, a pressure sensor, and the like.

With more volumes of data with higher granularity, robust data analysis modules may be used to process the data and construct SSHC prediction models. In one embodiment, the robust data analysis module is a high speed processing computing engine that can be configured to handle large volumes of data. The robust data analysis module may also be configured to receive processing data directly from sensors instead of having data relayed through the manufacturing facility host controller or even through the process module controller. Application No. 12 / 555,674, filed September 8, 2009, by Huang et al. Describes an exemplary analysis computer suitable for performing the analysis.

 In one embodiment, the SSHC prediction model may be used to predict uniformity. One skilled in the art knows that the etch rate may not be uniform across the surface of the substrate. Many factors can affect uniformity. For example, the angle at which gas may be dispensed into the processing chamber may affect uniformity. In another example, power distribution within the processing chamber may affect uniformity.

Although the etch rate is not uniform across the surface of the substrate, empirical evidence indicates that certain areas of the substrate may have substantially the same etch rate. In one embodiment of the invention, the substrate may be divided into three concentric circles (in an abstract sense), and the regions within each concentric circle are considered to have experimentally identical uniformity. In one embodiment, uniformity may be calculated from the etch rates of the processed substrate. First, the average etch rate of each concentric circle is determined. Each average etch rate is then multiplied by the number of data points measured in concentric circles (or for non-film substrates without metrology measurements, they are assumed to be present by metrology tools) Multiplied by the number of points (“virtual” points) (see FIG. 4)). Values for all three concentric circles are added and an average etch rate for the substrate may be calculated. Uniformity is then determined by calculating the standard deviation of each concentric mean etch rate relative to the overall average etch rate of the substrate. The overall substrate uniformity is then determined by calculating the standard deviation of all actual or "virtual" etch depths and computing its percentage to average etch rate.

Once the SSHC prediction model is constructed, the SSHC prediction model may be moved to manufacturing. Since the SSHC prediction model is constructed based in part on data collected from non-film substrates, the cost of implementing an SSHC prediction model in a manufacturing environment is significantly less than that of prior art prediction models. One reason for the significant cost reduction is that the SSHC prediction model can be applied to processing data collected from less expensive non-film substrates. In addition, reduced measurement requirements provide savings due to fewer metrology tools required to meet manufacturing needs. Faster time to qualify the chamber may also result in fewer at-risk fabrication wafers being processed because problems in the processor chamber can be detected more quickly. Accordingly, the SSHC prediction model provides an effective model for qualifying the processing chamber while effectively reducing the cost of ownership.

The features and advantages of the present invention may be better understood with reference to the drawings and the following description.

FIG. 2 shows a general overview of a method of constructing a Subsystem Health Check (SSHC) prediction model for qualifying a processing chamber in one embodiment of the invention.

In a first step 202, pre-processing measurements are performed on a set of film substrates. Similar to the prior art, metrology tools (such as KLA-Tencor thin film metrology tools) may be used to make measurements such as measuring the thickness of a substrate. Since the thickness of the substrate may vary over the entire substrate, different data points on the substrate may be measured (such as the data points shown in the 49 point pole scan of the substrate 402 of FIG. 4A).

In a next step 204, the set of film substrates is processed. Similar to the prior art, modified recipes may be used within the test environment. The modified recipe may be a simpler version of the manufacturing recipe and tends to simulate the etching behavior of the manufacturing recipe.

One skilled in the art knows that parameter changes may affect the average etch rate and / or uniformity of the substrate. Thus, in order to qualify the processing chamber correctly, parameter changes may have to be taken into account in generating the SSHC prediction model. To improve the robustness of the model, the system response to potential changes in manufacturing is evaluated. Recipe parameters may change as one or more test substrates are processed. In one example, the first few substrates (eg, three substrates) may be implemented using a basic modified recipe. For the fourth and fifth substrates, the pressure value may be changed to account for pressure level changes that may occur in the processing chamber. In another example, the gas flow distribution is adjusted for the following four substrates so that more percentage of the total gas flows toward the center of the substrate (due to corroded chamber consumables typically found near the etch of the substrate). Offset slightly higher edge etch rates that may occur.

After the set of film substrates are processed, in a next step 206, the set of non-film substrates (such as bare silicon substrates) may be processed using the same modified recipe as applied to the set of film substrates. In one example, when the pressure value is increased for the fourth film substrate, the same pressure value is applied to the fourth non-film substrate. While the non-film substrates are being processed, the sensors may also be collecting processing data. In one embodiment, the order of processing the set of film substrates and the set of non-film substrates does not limit the invention. That is, either step 204 or step 206 may occur first.

As previously mentioned in the prior art, one of the limitations of the prior art methods is due to the granularity of the data collected. In one embodiment, the methods described herein are applied in a processing tool that supports sensors (such as VI probe sensors, OES sensors, pressure senses, etc.) that may collect higher segmentation data. In addition, the high speed processing computing analysis module may be implemented to rapidly process and analyze data. In one embodiment, the high speed processing computing analysis module may be an Advanced Process and Equipment Control System (APECS). The APECS module may be configured to quickly analyze a plurality of data and may provide feedback to a processing module (PM) controller in the processing chamber such that the PM controller is etch rate and / or uniform for the next incoming substrate. It makes it possible to predict sex. Application No. 12 / 555,674, filed September 8, 2009, by Huang et al. Describes an exemplary analysis computer suitable for performing the analysis.

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

In the next step 210, the difference between the preprocessing measurement data and the postprocessing measurement data may be calculated, and an average etch rate and / or uniformity may be calculated for each film substrate. A discussion of uniformity is provided in detail in FIGS. 3 and 4 below.

In one embodiment, steps 202-210 of constructing the SSHC prediction model may be performed at least twice to further remove noise and / or drift. In one embodiment, the steps may be performed at the start of the wet clean cycle (ie, after maintenance has been performed to the processing chamber) and to the end of the wet clean (ie, in a period before the next maintenance is performed). have. In addition, data may be collected during the middle of the wet clean cycle.

Once the data has been collected, in the next step 212, sensor processing data and metrology data may be correlated and an SSHC prediction model may be constructed to qualify the processing chamber. In one embodiment, the SSHC prediction model may be based on a partial least squares model. The partial least squares model is a technique for discovering relationships between two sets of data. The partial least-squares model may have a similar purpose to the least-squares linear fit, but is typically used when there are a large number of independent variables (in input matrix X) and a large number of dependent variables (in input matrix Y). do. In the partial least squares model, the Y variables are continuous but instead consist of a set of independent discrete values or classes. The analysis aims to find linear combinations of X variables that can be used to classify the input data into one of these discrete classes.

As can be seen from FIG. 2, a method is provided for constructing an SSHC prediction model that may take into account the conditions of a processing chamber at different periods during a wet clean cycle. Although the SSHC prediction model as described in FIG. 2 requires additional time to be constructed (relative to the method described in FIG. 1), once configured, the SSHC prediction model is constantly (as required by the prior art methods). It does not need to be updated. Thus, the resources required to construct the SSHC prediction model are typically one-time costs instead of continuing costs (as required by prior art methods). In addition, since the SSHC prediction model may be applied to data collected from non-film substrates, device manufacturers do not have to continue using more expensive film substrates in the manufacturing environment to effectively apply the SSHC prediction model. The cost is significantly reduced.

3 illustrates one implementation of constructing an SSHC prediction model, in one embodiment of the invention.

In a first step 302, a first set of data is collected. In one embodiment, the first set of data is collected at the start of the wet clean cycle (after maintenance is performed on the processing chamber). The first set of data may include processing data collected by the sensors for the set of non-film substrates and metrology data collected for the first set of film substrates. That is, steps 202-210 of FIG. 2 are performed on the set of non-film substrates and the first set of film substrates. In one embodiment, the first set of film substrates and the set of non-film substrates may include the same number of substrates.

In a next step 304, a second set of data is collected. In one embodiment, step 304 is optional. The data collected during step 304 may be used as valid data. Step 304 is typically performed during the middle of a wet clean cycle to account for potential drift that may occur. In one example, some of the hardware components (eg, gas distribution system) may drift after several manufacturing runs.

Unlike film substrates, non-film substrates may be processed multiple times. Those skilled in the art know that at least non-film substrates (such as bare silicon substrates) may be processed up to about 10-15 times before they may no longer be able to handle subsequent processing. Thus, the same set of non-film substrates previously etched in step 302 may be processed again in step 304 and later in step 306.

In a next step 306, a third set of data is collected. In one embodiment, the third set of data is collected at the end of the wet clean cycle. The third set of data is collected for the third set of film substrates and processing data collected by the sensors for the set of non-film substrates (which may be the same set of non-film substrates from step 304). Included metrology data.

By collecting data from different periods during the wet clean cycle, the SSHC prediction model may capture the behavior of the processing chamber through the normal wet clean cycle.

In step 308, the system checks to determine if sufficient model set data has been collected. As discussed herein, model set data refers to the data collected in steps 302 and 306.

If there is insufficient model set data, in step 310, more data is collected before analysis can begin. For example, if not enough film substrates have been processed, insufficient data may be collected. In another example, some of the processed data collected by the sensors may be unacceptable and cannot be used to construct an SSHC prediction model.

However, if there is a sufficient amount of model set data, in the next step 312, the system checks whether a control limit has been specified. In one embodiment, the control limit is an error range that may be acceptable to the user. Since the SSHC prediction model is constructed based on the highly granular data, the control limit may be set as low as 2-3 percent.

If a control limit has not yet been set, then at step 314, the system may require that a control limit be provided.

However, if control limits are specified, then at step 316, the system may begin the process of constructing the SSHC prediction model. The SSHC prediction model includes the average etch rate as the target etch rate correlated to sensor processing data.

The target etch rate may be calculated using model set metrology data. In one example, the substrate location may have a pre-processing measurement of about 500 nm. Once the substrate has been processed, the thickness of the substrate at the same location is now 375 nm. If the etch depth is the difference between the pre-processing measurement and the post-processing measurement, the etch depth (eg, 0 degree orientation, 115 mm radius) at a given data point is 125 nm. If the process time for the substrate is 2 minutes, the etch rate is 62.5 nm per minute for that data point (eg, substrate position). Once the etch rate is determined, each of the data points on the substrate may now be associated with the etch rate.

Unlike the prior art, the average etch rate of the substrate is not calculated by adding up all 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 concept that the etch rates are not the same across the substrate. Instead, the experimental test splits (in an abstract manner) into three concentric circles (as shown by concentric circles 450, 440 and 430 in FIG. 4B) with each data point in the concentric circles having a similar etch rate. It can be.

To facilitate the discussion, the average etch rates are 62.5 nm / min, 72.5 nm / min and 82.5 nm for concentric circles 450 (ER1), concentric circles 440 (ER2) and concentric circles 430 (ER3), respectively. Assume After three average etch rates have been determined, each average etch rate may be normalized. In one example, concentric circles 450 have nine data points. Accordingly, the average etching rate (62.5 nm / minute) is multiplied by nine. After the average etch rate is normalized, the overall average etch rate is combined by three normalized average etch rates, and dividing the total by the number of data points (in this example, measurements were made at 49 data points). Is calculated. In this example, the overall average etch rate for the substrate is 75.6 nm / minute.

Once the average etch rate is calculated, then uniformity may be calculated. As described above, the etch rate values of the substrate may be divided into three concentric circles. That is, the etch rates at a given data point in the concentric circles may be essentially the same. Thus, the substrate is substantially uniform in concentric circles. Accordingly, uniformity for concentric circles may be determined by calculating the standard deviation of the average etch rate for the concentric circles relative to the overall average etch rate of the substrate.

In a next step 318, the model set prediction etch rate and uniformity may be calculated using the collected sensor processing data. Etch rate and uniformity for non-film substrates may be predicted using previously established correlations (ie, partial least squares modeling) between sensor data and metrology information.

In the next step 320. The system checks to determine if valid data has been collected. As described above, valid data may be collected during optional step 304. If valid data is present, then at step 322, the average etch rate and uniformity based on the valid data may be calculated.

In the next step 324, SSHC model predictive etch rate and uniformity for the valid data set are compared against actual metrological measurements. This step may be used to verify the SSHC prediction model.

Once the comparison is performed, the system may proceed to the next step 326. Similarly, if no valid data exists, the system may proceed to the next step 326.

In the next step 326, the system may compare the predictive error rate with a control limit. In one example, when the control limit is 3 percent, the prediction error rate should be 3 percent or less.

If the prediction error is unacceptable, then at step 328, a warning is issued to notify the user that the SSHC prediction model may need to be adjusted.

However, if the prediction error rate is acceptable, then in step 330, the SSHC prediction model may be moved to manufacturing and made available to qualify the processing chamber.

5 shows a simple flow diagram illustrating a method of applying an SSHC prediction model to qualify a processing chamber in one embodiment. Subsystem health verification tests may be performed before manufacturing is performed. That is, an SSHC prediction model may be applied to the data collected from the test substrate to determine the health state of the processing chamber.

In a first step 504, the user may activate the SSHC prediction model. User specifications (eg recipes, file names, etc.) may be entered. Such information may also be taken from library 502. Library 502 may include recipe specific parameters (such as upper control limit, lower control limit, target etch rate, uniformity, etc.).

In the next step 506, the system may perform a preemptive assessment. Proactive assessment may be performed to determine readiness of the processing chamber.

If the preemptive assessment failed (eg, outside the predefined threshold range), at the next step 526, the system may determine the cause of the alarm (failure). In one embodiment, the cause of the alarm may result from connectivity, a change in frequency, and / or a change in temperature (528).

If the cause of the alarm is due to connectivity, one or more sensors may not be properly connected. If the sensor is connected incorrectly, the sensor cannot capture processing data.

Another cause of the alarm may be due to large frequency changes. In one example, the frequency reported by the electrical probe (such as the VI probe sensor) and the frequency reported by the generator are compared with each other. If the difference between the two frequencies is above the threshold (eg too large), then a problem may exist for example in the VI probe sensor. In one example, the VI probe sensor may be too active. In another example, the VI probe sensor may be disabled.

Another cause of the alarm may be due to a large temperature change between the current temperature in the processing chamber and the desired set point temperature (ie, recipe temperature). In one example, when the processing tool is first turned on, the temperature in the processing chamber may require several minutes to reach the desired set point temperature. The system checks to ensure that the temperature in the processing chamber is within the desired threshold before processing can begin.

Once the system passes the preemptive assessment, in the next step 508, the system checks to determine if the SSHC prediction model exists.

If the SSHC prediction model does not exist, then in step 522, the SSHC prediction model is constructed (as described in FIGS. 2 and / or 3).

However, if an SSHC prediction model exists, the substrate is processed and processing data is collected. In one embodiment, an SSHC prediction model is applied to the processing data to predict the etch rate and / or uniformity (calculated data) for the current substrate (step 510).

In the next step 512, the system checks to determine if the etch rate and uniformity are within control limits. If the etch rate and / or uniformity are not within the control limits, then at step 514, a failure notification may be issued. Failure notification may provide details about potential problems that may exist within the processing chamber. In one example, the failure notification may indicate that the recipe may need to be adjusted to account for drift in the processing chamber. In another example, the failure notification may indicate that there was a sudden temperature increase during substrate processing, which may have been caused by insufficient coolant to adjust the chamber temperature.

However, if the etch rate and uniformity are within control limits, then at step 516, the system may check to determine if the degree of SSHC verification has been satisfied. In one example, the user may have set a limit indicating that the processing chamber is in good operating condition if the etch rate and uniformity for three consecutive substrates are within control limits.

If the degree of SSHC verification is not met, at the next step 518, the system may continue (return to step 510) to the next substrate. Steps 510-518 are repetitive. In one embodiment, steps 510-518 may be performed until the degree of SSHC verification is satisfied. In yet another embodiment, a failure notification may be issued to a user who is unable to meet the degree of SSHC verification within a predetermined number of substrates, thereby informing the user about a potential problem with the system.

However, if the degree of SSHC verification is satisfied, then at step 520, the processing chamber has passed the subsystem health check and the manufacturing run may begin.

As will be appreciated from the foregoing, methods are provided for qualifying a processing chamber. By extrapolating data from two or more test runs during the wet clean cycle, an SSHC prediction model may be constructed that takes into account changes that may occur during the wet clean cycle. In the case of a robust SSHC prediction model, the SSHC prediction model provides an effective model to qualify the processing chamber while effectively reducing the cost of ownership.

Although the invention has been described by several preferred embodiments, there are variations, substitutions, and equivalents that fall within the scope of the invention. While various examples are provided herein, these examples are illustrative and are not limiting of the invention.

In addition, the name and summary of the invention are provided herein for convenience and should not be used to interpret the scope of the claims. In addition, the abstract is written in a highly condensed form and is provided herein for convenience and, therefore, should not be used to interpret or limit the entire invention as expressed in the claims. When the term “set” is used herein, such term is intended to have a commonly understood mathematical meaning covering a member that is zero, one member, or two or more members. In addition, there are many alternative ways of implementing the methods and apparatus of the present invention. Accordingly, the following appended claims are intended to be construed to include all such alterations, substitutions, and equivalents that fall within the true spirit and scope of the present invention.

Claims (20)

A method of predicting etch rate uniformity to qualify a healthy state of a processing chamber during substrate processing of a set of substrates, the method comprising:
Executing a recipe on a first of the set of substrates;
Receiving processing data from a first set of sensors during the execution of the recipe;
Analyzing the processing data using a subsystem health check prediction model to determine calculated data, the calculated data including at least one of etch rate data and uniformity data, and the subsystem health check prediction model Is constructed by correlating a first set of data with a second set of data, the first set of data comprising measurement data from a set of film substrates, the second set of data Determining the calculated data comprising processing data collected during similar processing of film substrates;
Performing a comparison of the calculated data of the first substrate to a set of control limits defined by the subsystem health confirmation prediction model; And
If the calculated data is outside the set of control limits, generating a warning. The method of claim 1,
And performing a preemptive assessment to determine readiness of the processing chamber. The method of claim 2,
If the preemptive assessment is outside a predefined threshold, further comprising determining the cause of the problem. The method of claim 1,
And fetching data from a library to support the subsystem health verification prediction model. The method of claim 1,
And wherein said set of control limits are user-configurable. The method of claim 1,
If the calculated data is within the set of control limits, a verification degree is determined to determine whether a predefined number of substrates have passed the comparison between the calculated data and the set of control limits. and verifying the degree). The method of claim 1,
The subsystem health confirmation prediction model is constructed from data collected from a time period,
Wherein the time period is one of at the beginning of the wet cycle, during the wet cycle, and at the end of the wet cycle. The method of claim 7, wherein
And the subsystem health confirmation prediction model is constructed from data collected from one time period. The method of claim 7, wherein
And the subsystem health confirmation prediction model is constructed from data collected from at least two time periods. A processing chamber health verification device for qualifying a processing chamber of a plasma processing system,
A pre-evaluation module for determining readiness of the processing chamber;
A library for storing at least one of recipes and recipe parameters;
Includes a subsystem health check prediction model,
The subsystem health confirmation prediction model,
Receiving processing data from the first set of sensors during substrate processing;
Analyzing the processing data to determine a set of calculated data, wherein the set of calculated data includes at least one of etch rate data and uniformity data. ;
Comparing the set of calculated data with a set of predefined control limits; And
And generate a warning if the set of calculated data is outside the set of predefined control limits. The method of claim 10,
The subsystem health confirmation prediction model, if the calculated data is within the set of control limits, a predefined number of substrates pass the comparison between the calculated data and the set of control limits. And verifying the degree of verification to determine if the processing chamber has been performed. The method of claim 10,
The subsystem health confirmation prediction model,
Acquiring a first set of data from a first set of substrates, wherein the first set of substrates is a set of film substrates, the first set of data comprising measurement data before processing and measurement data after processing Acquiring the first set of data, wherein the post-processing measurement data is collected after executing a recipe on the first set of substrates;
Collecting a second set of data from a second set of substrates, wherein the second set of data is collected by a set of sensors during the execution of a similar execution of the recipe. To do; And
And correlating the second set of data with the first set of data to construct the subsystem health verification prediction model. The method of claim 10,
The subsystem health confirmation prediction model is constructed from data collected from a time period,
Wherein the time period is one of at the start of a wet cycle, during the wet cycle, and at the end of the wet cycle. The method of claim 13,
And the subsystem health verification prediction model is constructed from data collected from one time period. The method of claim 13,
And the subsystem health verification prediction model is constructed from data collected from at least two time periods. An article of manufacture comprising a program storage medium comprising computer readable code, the article of manufacture comprising:
The computer readable code is
Configured to predict etch rate uniformity to qualify the health state of the processing chamber during substrate processing of a set of substrates,
Code for executing a recipe on a first of the set of substrates;
Code for receiving processing data from a first set of sensors during the execution of the recipe;
Code for analyzing the processing data using a subsystem health check prediction model to determine calculated data, the calculated data including at least one of etch rate data and uniformity data, and the subsystem health check prediction model Is constructed by correlating a first set of data with a second set of data, the first set of data comprising measurement data from a set of film substrates, the second set of data Code for determining the calculated data, the processing data collected during similar processing of film substrates;
Code for comparing the calculated data of the first substrate against a set of control limits defined by the subsystem health confirmation prediction model; And
A program storage medium comprising code for generating a warning if the calculated data is outside the set of control limits. 17. The method of claim 16,
Further comprising code for performing a preemptive assessment to determine readiness of the processing chamber. 17. The method of claim 16,
Further comprising code for fetching data from a library to support the subsystem health verification prediction model. 17. The method of claim 16,
If the calculated data is within the set of control limits, verify the degree of verification to determine if a predefined number of substrates have passed the comparison between the calculated data and the set of control limits. The article of manufacture further comprising a program storage medium. 17. The method of claim 16,
The subsystem health confirmation prediction model is constructed from data collected from a time period,
Wherein the time period is one of at the beginning of a wet cycle, during the wet cycle, and at the end of the wet cycle.

Images