Film formation condition output device, output method, output program, and film formation device

Abstract

The present application provides a film formation condition output device, output method, output program, and film formation device, which can output more appropriate film formation conditions. The film formation condition output device has: a first calculation section that calculates a first film formation condition for achieving a target film formation result using a linear model that models film formation processing by a film formation device; a second calculation section that calculates a second film formation condition for achieving the target film formation result using a nonlinear regression model that models the film formation processing by the film formation device and can calculate a confidence interval of a predicted film formation result; and a selection section that selects and outputs either the first film formation condition or the second film formation condition based on a determination result of whether the confidence interval of the predicted film formation result when the second film formation condition is calculated satisfies a prescribed condition.

Description

TECHNICAL FIELD

[0001] The present application relates to a film formation condition output device, a film formation condition output method, a film formation condition output program, and a film formation device. BACKGROUND

[0002] In a film formation device, in order to explore the most suitable film formation conditions for achieving a target film formation result, various prediction models have been proposed in the past that model the film formation processing by the film formation device using a machine learning technique. As one example, a nonlinear regression model can be cited.

[0003] In the case of a nonlinear regression model, the expressiveness of the model is higher compared to a linear regression model, and even if the output follows a complex regression curve, the prediction accuracy can be locally improved. However, on the other hand, the prediction accuracy in the large can be reduced due to bias in data sampling, lack of the number of data samples, and the like. Therefore, in the case where film formation processing is performed using the film formation conditions that are optimized by using a nonlinear regression model, the actual film formation result can sometimes deviate greatly from the target film formation result.

[0004] <RELATED ART DOCUMENTS>

[0005] <PATENT DOCUMENTS>

[0006] Patent Document 1: Japanese Patent No. 6647473 SUMMARY

[0007] <PROBLEMS TO BE SOLVED BY THE INVENTION>

[0008] The present application provides a film formation condition output device, a film formation condition output method, a film formation condition output program, and a film formation device that can output more appropriate film formation conditions.

[0009] <MEANS FOR SOLVING THE PROBLEMS>

[0010] According to one aspect, a film formation condition output device has:

[0011] a first calculation section that calculates a first film formation condition for achieving a target film formation result using a linear model that models the film formation processing by a film formation device;

[0012] a second calculation section that calculates a second film formation condition for achieving the target film formation result using a nonlinear regression model that models the film formation processing by the film formation device and that can calculate a confidence interval of a predicted film formation result; and

[0013] a selection section that selects and outputs either the first film formation condition or the second film formation condition based on a determination result of whether or not the confidence interval of the predicted film formation result when the second film formation condition is calculated satisfies a prescribed condition.

[0014] <Effects of the Invention>

[0015] According to the present application, a film formation condition output device, a film formation condition output method, a film formation condition output program, and a film formation device capable of outputting more appropriate film formation conditions can be provided. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a view showing an application example of the film formation condition output device.

[0017] Figure 2 is a view showing one example of the hardware configuration of the film formation condition output device.

[0018] Figure 3 is a view showing one example of the learning data set stored in the learning data set storage unit.

[0019] Figure 4 is a view showing one example of the functional configuration of the film formation condition output device in the learning stage.

[0020] Figure 5 is a view showing one example of the functional configuration of the film formation condition output device in the prediction stage.

[0021] Figure 6 is a view showing one example of the flowchart of the flow of the film formation condition output processing.

[0022] Figure 7 is a view showing a flowchart of the flow of the optimization processing using the learned linear model.

[0023] Figure 8 is a view showing a flowchart of the flow of the optimization processing using the learned Gaussian process regression model.

[0024] Figure 9 is a view showing a flowchart of the flow of the selection processing.

[0025] Figure 10 is a view showing a specific example of the selection processing. DETAILED DESCRIPTION

[0026] Hereinafter, each embodiment will be described with reference to the accompanying drawings. Note that in this specification and the accompanying drawings, for constituent elements having substantially the same function configuration, the same reference sign is given and repetitive description is omitted.

[0027] [First Embodiment]

[0028] [Application Example of Film Formation Condition Output Device]

[0029] First, an application example of the film formation condition output device of the first embodiment will be described. Figure 1 is a view showing an application example of the film formation condition output device. In this view, Figure 1 (a) of FIG. 1 shows an application example of the film formation condition output device in a learning phase.

[0030] As shown in (a) of FIG. 1, in the learning phase, the film formation system 100 has a film formation device 120, a film formation condition output device 140, and a film thickness measurement device 150. Figure 1

[0031] The film formation device 120 performs film formation processing on the object (pre-processing wafer 110) to generate the result (post-processing wafer 130). In the present embodiment, the film formation device 120 is described as a single-chip film formation device having a plurality of plasma sources, but the film formation device 120 is not limited thereto. For example, the film formation device 120 can be a single film formation device having a single plasma source, or a batch film formation device that performs film formation processing on a plurality of pre-processing wafers 110. However, in the case where the film formation condition output device 140 described below is processed in each region in the batch film formation device, processing is set to be performed simultaneously in all regions.

[0032] Note that the pre-processing wafer 110 refers to a wafer (substrate) before film formation processing in the film formation device, and the post-processing wafer 130 refers to a wafer (substrate) after film formation processing in the film formation device.

[0033] The film thickness measurement device 150 measures the film thickness of each position of the pre-processing wafer 110 and the film thickness of each position of the post-processing wafer 130, respectively. In addition, the film thickness measurement device 150 notifies the film formation condition output device 140 of the film thickness of each position measured with respect to the pre-processing wafer 110 and the post-processing wafer 130, respectively. Note that in the present embodiment, as one example of the film formation result, the film thickness measurement device 150 notifies the film formation condition output device 140 of the film thickness of each position measured, but the film formation result notified to the film formation condition output device 140 is not limited to the film thickness of each position. For example, instead of (or in addition to) the film thickness of each position, the film quality of each position can be notified.

[0034] The film formation condition output device 140 sets the film formation conditions input by the user 160 in the film formation device 120, and the film formation device 120 performs film formation processing based on the set film formation conditions, thereby obtaining the measured film formation result.

[0035] ​In addition, the film-forming condition output device 140 generates multiple learning data sets, including the set film-forming conditions and the obtained film-forming results, and stores them as learning datasets in the learning dataset storage unit 141. Furthermore, the film-forming condition output device 140 reads the stored learning datasets, uses the multiple learning data sets to learn multiple prediction models (linear models, nonlinear regression models), and outputs the learning results (the multiple prediction models that have been learned).

[0036] on the other hand, Figure 1 (b) illustrates an application example of a film-forming condition output device during the prediction stage. For example... Figure 1 As shown in (b), during the prediction phase, the film-forming system 100 has a film-forming device 120 and a film-forming condition output device 140.

[0037] It should be noted that in this embodiment, the film-forming apparatus 120 is described as the same individual as the film-forming apparatus 120 in the learning phase. However, the film-forming apparatus in the learning phase and the film-forming apparatus in the prediction phase can be, for example, different individuals of the same type.

[0038] The film-forming condition output device 140 executes multiple learned prediction models (linear models and nonlinear regression models) based on the initial and target film-forming conditions input by the user 160. Furthermore, the film-forming condition output device 140 selects the optimal film-forming conditions and predicted film-forming results from the execution results of the multiple learned prediction models and outputs them to the user 160. It should be noted that the optimal film-forming conditions are set in the film-forming device 120 after being confirmed by the user 160.

[0039] Thus, by setting the configuration of the optimal film-forming conditions selected from the execution results of multiple prediction models learned by the film-forming apparatus 120, more suitable optimal film-forming conditions can be set.

[0040] Specifically, in cases where the prediction accuracy of the nonlinear regression model is reduced due to data sampling bias or insufficient data sample quantity, the optimal film-forming conditions and predicted film-forming results are selected based on the results of the linear model. Conversely, in cases where the prediction accuracy of the nonlinear regression model is high, the optimal film-forming conditions and predicted film-forming results are selected based on the results of the nonlinear regression model.

[0041] Thus, in the film forming apparatus 120, a processed wafer 130 that achieves the target film forming result can be generated.

[0042] <Hardware Composition of Film Formation Condition Output Device>

[0043] Figure 2This is a diagram illustrating an example of the hardware configuration of the film formation condition output device 140. (See diagram for example.) Figure 2 As shown, the film-forming condition output device 140 includes a processor 201, a memory 202, an auxiliary storage device 203, an I / F (Interface) device 204, a communication device 205, and a drive device 206. It should be noted that the various hardware components of the film-forming condition output device 140 are interconnected via a bus 207.

[0044] The processor 201 has various computing elements such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 201 reads and executes various programs (e.g., the film formation condition output program described later) on the memory 202.

[0045] The memory 202 has main storage elements such as ROM (Read Only Memory) and RAM (Random Access Memory). The processor 201 and the memory 202 form a so-called computer, which performs various functions by executing various programs read from the memory 202 through the processor 201.

[0046] The auxiliary storage device 203 is used to store various programs and various data used by the processor 201 when the programs are executed. For example, the learning dataset storage unit 141 is implemented in the auxiliary storage device 203.

[0047] The I / F device 204 is a connecting element for connecting an external device, namely an operating device 211 and a display device 212, to the film formation condition output device 140. The I / F device 204 receives operations on the film formation condition output device 140 via the operating device 211 (e.g., input of film formation conditions, input of initial film formation conditions, input of target film formation result, etc.). Furthermore, the I / F device 204 outputs the execution results based on the film formation condition output device 140 (e.g., predicted film formation result, optimized film formation conditions, etc.) and displays them on the display device 212.

[0048] The communication device 205 is a communication element for communicating with other devices (e.g., film forming device 120) via a network.

[0049] The drive unit 206 is a component used to mount the storage medium 213. The storage medium 213 referred to here includes media that store information optically, electrically, or magnetically, such as CD-ROMs, floppy disks, and optical discs. Alternatively, the storage medium 213 may include semiconductor memories that store information electrically, such as ROMs and flash memory.

[0050] It should be noted that the various programs installed in the auxiliary storage device 203 can be installed, for example, by placing the distributed storage medium 213 on the drive device 206, and installing the various programs stored in the storage medium 213 by the drive device 206. Alternatively, the various programs installed in the auxiliary storage device 203 can be downloaded and installed via a network (not shown).

[0051] <Specific Examples of Learning Using Datasets>

[0052] Next, a specific example of a learning dataset stored in the learning dataset storage unit 141 will be explained. Figure 3 This is a diagram showing an example of a learning dataset stored in the learning dataset storage department.

[0053] like Figure 3 As shown, the learning dataset 300 includes multiple learning datasets such as Learning Data 1, Learning Data 2, Learning Data 3, ... Furthermore, each learning dataset consists of groups of film formation conditions and film formation results. Figure 3 The example shows that “Learning Data 1” consists of a group of film formation conditions 300_1 and film formation results 300_2.

[0054] like Figure 3 As shown, the information items for film formation condition 300_1 include "Type of process data" and "Process data".

[0055] The “Types of Process Data” section stores the types of process data related to film formation. Figure 3 Examples show that the types of process data include "pressure", "plasma gas flow rate", "output of plasma source #1" to "output of plasma source #7", "temperature", and "film formation time".

[0056] The “Process Data” section stores the values ​​of the process data corresponding to the “Type of Process Data”.

[0057] On the other hand, such as Figure 3 As shown, the film formation result 300_2 includes the following information items: "Measurement location" and "Film thickness data".

[0058] The "Measurement Location" stores measurement points on each location above the processed wafer 130, used to determine the film thickness measured by the film thickness measuring device 150 (in this embodiment, the film thickness measuring device 150 is set to measure the film thickness at n locations). Additionally, the "Film Thickness Data" stores the values ​​of the film thickness data measured at each location corresponding to the "Measurement Location".

[0059] <Functional Composition of Film Formation Condition Output Device>

[0060] Next, use Figure 4 as well as Figure 5 The functional configuration of the film-forming condition output device 140 is explained.

[0061] (1) Film-forming condition output device during the learning phase

[0062] Figure 4 This diagram illustrates an example of the functional configuration of a film-forming condition output device during the learning phase. As described above, a film-forming condition output program is installed in the film-forming condition output device 140, and by executing this program during the learning phase, the film-forming condition output device 140 functions as a first learning unit 410 and a second learning unit 420.

[0063] The first learning unit 410 has a linear model 411. The first learning unit 410 updates the model parameters of the linear model 411 using the learning dataset 300 read from the learning dataset storage unit 141. Specifically, the first learning unit 410 updates the model parameters in such a way that the output of the learning dataset contains film-forming conditions (e.g., film-forming condition 300_1) as input is close to the film-forming result (e.g., film-forming result 300_2) contained in each learning dataset. It should be noted that the learned linear model (learning result) with updated model parameters is used in the prediction phase.

[0064] The second learning unit 420 has a Gaussian process regression model 421, which is an example of a nonlinear regression model. The Gaussian process regression model is a non-parametric probabilistic model that can output the variance of the prediction result (in this embodiment, the range of the confidence interval) along with the prediction result (in this embodiment, the predicted film formation result).

[0065] exist Figure 4 In this process, the second learning unit 420 updates the model parameters of the Gaussian process regression model 421 using the learning dataset 300 read from the learning dataset storage unit 141. Specifically, the second learning unit 420 updates the model parameters such that the output of the learning dataset contains film-forming conditions (e.g., film-forming condition 300_1) as input is close to the film-forming results (e.g., film-forming result 300_2) contained in each learning dataset. It should be noted that the learned Gaussian process regression model (learning result) with updated model parameters is used in the prediction phase.

[0066] The Gaussian process regression model used in the prediction phase can be expressed by equation (2) based on the learned Gaussian process regression model generated in the learning phase, i.e., equation (1).

[0067] (Mathematical Formula 1)

[0068] p(y|x)=N(0,∑ (n) (1)

[0069] (Mathematical Formula 2)

[0070]

[0071] It should be noted that in equations (1) and (2) above, X and y represent the film-forming conditions and film-forming results included in the learning data used during the learning phase, respectively. n+1 and y n+1 These represent the film-forming conditions and film-forming results obtained in the optimization calculation, respectively. Additionally, Σ (n) and Σ n+1 This represents the diagonal partitioning of the covariance matrix of a Gaussian process.

[0072] Furthermore, the following mathematical formula 3 is a non-diagonal block of the covariance matrix of the Gaussian process.

[0073] (Mathematical Formula 3)

[0074]

[0075] From the above equation (2), it can be derived that the predicted value of the film formation result follows the probability distribution shown in equation (3).

[0076] (Mathematical Expression 4)

[0077]

[0078] Using equation (3) above, when using the Gaussian process regression model generated during the learning phase, x n+1 The film formation result under the film formation conditions is given by the average value of the Gaussian distribution on the right side of the above equation (3), i.e., the following mathematical formula 5.

[0079] (Mathematical Formula 5)

[0080]

[0081] Furthermore, in the optimization of film formation conditions in the Gaussian process regression model, the average value of the Gaussian distribution, i.e., the following mathematical formula 6, applies to film formation conditions x that approximate the target film formation result. n+1 Perform the calculation.

[0082] (Mathematical Expression 6)

[0083]

[0084] In addition, the confidence interval for predicting film formation results using the Gaussian process regression model 421 is calculated based on the variance of the Gaussian distribution on the right side of the above equation (1), i.e., the following mathematical formula 7.

[0085] (Mathematical Expression 7)

[0086]

[0087] (2) Film-forming condition output device in the prediction stage

[0088] Figure 5 This diagram illustrates an example of the functional configuration of a film-forming condition output device during the prediction phase. As described above, a film-forming condition output program is installed in the film-forming condition output device 140, and by executing this program during the prediction phase, the film-forming condition output device 140 functions as the following components.

[0089] ·Optimization Department 510

[0090] First Forecasting Department 520

[0091] ·Optimization Department 530

[0092] ·Second Forecasting Department 540

[0093] Selection Department 550

[0094] The optimization unit 510 receives initial film-forming conditions and a target film-forming result input by the user 160. Furthermore, the optimization unit 510 obtains a first predicted film-forming result (first film-forming result) output by the first prediction unit 520 by inputting the received initial film-forming conditions into the first prediction unit 520, and adjusts the film-forming conditions by comparing them with the target film-forming result. Additionally, the optimization unit 510 obtains a first predicted film-forming result output by the first prediction unit 520 by inputting the adjusted film-forming conditions into the first prediction unit 520, and further adjusts the film-forming conditions by comparing them with the target film-forming result.

[0095] It should be noted that in the optimization unit 510, the film formation conditions are optimized by repeatedly comparing the first predicted film formation result with the target film formation result and adjusting the film formation conditions based on the comparison results. For example, the optimization unit 510 optimizes the film formation conditions while successively changing multiple types of process data included in the film formation conditions.

[0096] In addition, in the optimization unit 510, if the difference between the first predicted film formation result and the target film formation result meets the specified conditions, the optimization process of the film formation conditions is ended, and the film formation conditions at the end time are notified to the selection unit 550 as the first optimized film formation conditions (first film formation conditions).

[0097] It should be noted that in the optimization unit 510, as a predetermined condition, such as when the following conditions are met, it is determined that the difference between the first predicted film formation result and the target film formation result meets the predetermined conditions.

[0098] The difference between the average of the first predicted film formation results at n measurement points and the average of the target film formation results at n measurement points is below a specified threshold.

[0099] • The deviation of the first predicted film formation result at n measurement points is below the specified threshold.

[0100] The first prediction unit 520 is an example of a first calculation unit, which has a learned linear model 521 generated by the first learning unit 410 during the learning phase. The first prediction unit 520 obtains initial or adjusted film-forming conditions from the optimization unit 510 and inputs them into the learned linear model 521. In addition, the first prediction unit 520 notifies the optimization unit 510 of the first predicted film-forming result output by the learned linear model 521 by inputting the initial or adjusted film-forming conditions. Furthermore, the first prediction unit 520 notifies the selection unit 550 of the first predicted film-forming result that meets the prescribed conditions determined in the optimization unit 510.

[0101] It should be noted that, in Figure 5 In the example, the case where the optimization unit 510 repeatedly adjusts the film-forming conditions until the specified conditions are met is explained, thereby optimizing the film-forming conditions. However, based on the learned linear model, there are sometimes situations where it is not necessary to repeatedly adjust the film-forming conditions.

[0102] In this case, the optimization unit 510 does not determine whether the prescribed conditions are met (or sets the difference between the first predicted film formation result and the target film formation result to always meet the prescribed conditions), and notifies the selection unit 550 of the first optimized film formation conditions.

[0103] The optimization unit 530 receives initial film-forming conditions and a target film-forming result input by the user 160. Furthermore, the optimization unit 530 obtains a second predicted film-forming result (second film-forming result) output by the second prediction unit 540 by inputting the received initial film-forming conditions into the second prediction unit 540, and adjusts the film-forming conditions by comparing them with the target film-forming result. Additionally, the optimization unit 530 obtains a second predicted film-forming result output by the second prediction unit 540 by inputting the adjusted film-forming conditions into the second prediction unit 540, and further adjusts the film-forming conditions by comparing them with the target film-forming result.

[0104] It should be noted that in the optimization unit 530, the film formation conditions are optimized by repeatedly comparing the second predicted film formation result with the target film formation result and adjusting the film formation conditions based on the comparison results. For example, the optimization unit 530 optimizes the film formation conditions while successively changing multiple types of process data included in the film formation conditions.

[0105] Furthermore, in the optimization unit 530, if the difference between the second predicted film formation result and the target film formation result meets the prescribed conditions, the optimization process of the film formation conditions is terminated, and the film formation conditions at the end time are notified to the selection unit 550 as the second optimized film formation conditions (second film formation conditions).

[0106] It should be noted that in the optimization unit 530, as a predetermined condition, such as when the following conditions are met, it is determined that the difference between the second predicted film formation result and the target film formation result meets the predetermined conditions.

[0107] The difference between the average of the second predicted film formation results at n measurement points and the average of the target film formation results at n measurement points is below a specified threshold.

[0108] • The deviation of the second predicted film formation results at n measurement points is below a specified threshold.

[0109] • The number of repetitions must be more than the specified number.

[0110] The second prediction unit 540 is an example of a second calculation unit, which has a learned Gaussian process regression model 541 generated by the second learning unit 420 during the learning phase. The second prediction unit 540 obtains initial or adjusted film-forming conditions from the optimization unit 530 and inputs them into the learned Gaussian process regression model 541. In addition, the second prediction unit 540 notifies the optimization unit 530 of the second predicted film-forming result output by its learned Gaussian process regression model 541 by inputting the initial or adjusted film-forming conditions. Furthermore, the second prediction unit 540 notifies the selection unit 550 of the second predicted film-forming result that is determined to meet the prescribed conditions in the optimization unit 530, and the range of the confidence interval (the range of the confidence interval corresponding to n measurement points) representing the variance of the second predicted film-forming result.

[0111] The selection unit 550 outputs the optimal film formation conditions and the predicted film formation results. Specifically, the selection unit 550 selects and outputs any of the following data.

[0112] • The first optimized film formation conditions notified by the optimization unit 510, and the first predicted film formation results notified by the first prediction unit 520.

[0113] • The second optimal film formation conditions notified by the optimization unit 530, and the second predicted film formation results notified by the second prediction unit 540.

[0114] It should be noted that in the selection section 550, when selecting any option, please refer to the following content.

[0115] • The range of the confidence interval (corresponding to n measurement points) notified by the second prediction unit 540, and / or

[0116] • Comparison results of the absolute value of the difference between the first predicted film formation result notified by the first prediction unit 520 and the target film formation result with the absolute value of the difference between the second predicted film formation result notified by the second prediction unit 540 and the target film formation result (corresponding to n measurement points).

[0117] Thus, in the selection section 550, in order to determine whether it is an example of reduced prediction accuracy of the nonlinear regression model, the range of the confidence interval and the absolute value of the difference are referenced.

[0118] <Film Formation Conditions Output Processing>

[0119] Next, the process of film formation condition output processing based on film formation condition output device 140 will be explained. Figure 6 This is an example of a flowchart illustrating the process of outputting film-forming conditions. It should be noted that steps S601 to S603 are performed during the learning phase.

[0120] In step S601, the first learning unit 410 and the second learning unit 420 read the learning dataset from the self-learning dataset storage unit 141.

[0121] In step S602, the first learning unit 410 uses the read learning dataset to learn the linear model.

[0122] In step S603, the second learning unit 420 uses the read learning dataset to learn the Gaussian process regression model. After the learning process is completed, the process moves to the prediction stage.

[0123] In step S604, the optimization unit 510 and the optimization unit 530 receive the initial film formation conditions and the target film formation conditions input by the user 160.

[0124] In step S605, the optimization unit 510 and the first prediction unit 520 perform optimization processing to explore the most suitable film-forming conditions (first optimal film-forming conditions) using a learned linear model. Details of the optimization processing using the learned linear model will be described later.

[0125] In step S606, the optimization unit 530 and the second prediction unit 540 perform optimization processing to explore the most suitable film-forming conditions (second optimal film-forming conditions) using a learned Gaussian process regression model. Details of the optimization processing using the learned Gaussian process regression model will be described later.

[0126] In step S607, the selection unit 550 performs a selection process to select the optimal film-forming conditions and predict the film-forming results by referring to the range of the confidence interval calculated when the second prediction unit 540 explores the second optimal film-forming conditions. Details of the selection process will be described later.

[0127] In step S608, the selection unit 550 outputs the selected optimal film-forming conditions and the predicted film-forming result to the user 160, and sets the selected optimal film-forming conditions in the film-forming apparatus 120.

[0128] <Detailed Processing of Film-Forming Conditions Output>

[0129] Next, the details of each step in the film formation condition output processing (optimization processing using a learned linear model (step S605), optimization processing using a learned Gaussian process regression model (step S606), and selection processing (step S607)) will be explained.

[0130] (1) Detailed explanation of the optimization process using the learned linear model (step S605)

[0131] Figure 7 This is a flowchart illustrating the process of optimization using a learned linear model.

[0132] In step S701, the first prediction unit 520 inputs the initial film formation conditions or adjusted film formation conditions notified by the optimization unit 510 into the learned linear model 521.

[0133] In step S702, the first prediction unit 520 notifies the optimization unit 510 of the first predicted film formation result output by the learned linear model 521.

[0134] In step S703, the optimization unit 510 compares the target film formation result with the first predicted film formation result.

[0135] In step S704, the optimization unit 510 determines whether the comparison result meets the specified conditions. If it is determined in step S704 that the specified conditions are not met (if it is No in step S704), the process proceeds to step S705.

[0136] In step S705, the optimization unit 510 adjusts the film formation conditions, notifies the first prediction unit 520 of the adjusted film formation conditions, and then returns to step S701.

[0137] On the other hand, if it is determined in step S704 that the specified conditions are met (if it is Yes in step S704), proceed to step S706. It should be noted that, as described above, if the learned linear model does not need to repeatedly perform the adjustment of film formation conditions, the processing in step S704 is not performed (or, it is set to always meet the specified conditions), and proceed to step S706.

[0138] In step S706, the optimization unit 510 notifies the selection unit 550 of the first optimized film formation conditions, and the first prediction unit 520 notifies the selection unit 550 of the first predicted film formation result. Then, the process returns to... Figure 6 Step S606.

[0139] (2) Detailed explanation of the optimization process using the learned Gaussian process regression model (step S606) Figure 8 This is a flowchart illustrating the process of optimization using a learned Gaussian process regression model.

[0140] In step S801, the second prediction unit 540 inputs the initial film formation conditions or adjusted film formation conditions notified by the optimization unit 530 into the learned Gaussian process regression model 541.

[0141] In step S802, the second prediction unit 540 notifies the optimization unit 530 of the second predicted film formation result output by the learned Gaussian process regression model 541.

[0142] In step S803, the optimization unit 530 compares the target film formation result with the second predicted film formation result.

[0143] In step S804, the optimization unit 530 determines whether the comparison result meets the specified conditions. If it is determined in step S804 that the specified conditions are not met (if it is No in step S804), the process proceeds to step S805.

[0144] In step S805, the optimization unit 530 adjusts the film formation conditions, notifies the second prediction unit 540 of the adjusted film formation conditions, and then returns to step S801.

[0145] On the other hand, if it is determined in step S804 that the specified conditions are met (if it is Yes in step S804), then proceed to step S806.

[0146] In step S806, the optimization unit 530 notifies the selection unit 550 of the second optimized film formation conditions, and the second prediction unit 540 notifies the selection unit 550 of the second predicted film formation result and the range of the confidence interval. Then, it returns to the previous step. Figure 6 Step S607.

[0147] (3) Select the detailed processing (step S607)

[0148] Figure 9 This is a flowchart illustrating the selection process. In step S901, the selection unit 550 obtains the range of the confidence interval (corresponding to n measurement points) notified by the second prediction unit 540.

[0149] In step S902, the selection unit 550 determines whether the range of the obtained confidence interval meets a predetermined condition (whether it is below a predetermined threshold). If, in step S902, the range of the confidence interval does not meet the predetermined condition (is greater than the predetermined threshold) (in the case of No in step S902), the process proceeds to step S905. It should be noted that the selection unit 550 proceeds to step S905 under the following circumstances.

[0150] • Determine if any of the confidence intervals for the corresponding n measurement points is greater than a specified threshold, or

[0151] • When the range of a predetermined number of confidence intervals corresponding to n measurement points is larger than a specified threshold.

[0152] On the other hand, in step S902, if the range of the confidence interval meets the specified conditions (below the specified threshold), then proceed to step S903.

[0153] In step S903, the selection unit 550 calculates the absolute value (first absolute value of difference) of the difference between the first predicted film formation result notified by the first prediction unit 520 and the target film formation result. For example, the selection unit 550 calculates the first absolute value of difference as the sum of the squares of the differences between the first predicted film formation result and the target film formation result at n measurement points.

[0154] In addition, the selection unit 550 calculates the absolute value (second absolute value of difference) of the difference between the second predicted film formation result notified by the second prediction unit 540 and the target film formation result. For example, the selection unit 550 calculates the second absolute value of difference as the sum of the squares of the differences between the second predicted film formation result and the target film formation result at n measurement points.

[0155] Furthermore, the selection unit 550 determines whether the absolute value of the first difference is greater than or equal to the absolute value of the second difference.

[0156] In step S903, if the absolute value of the first difference is determined to be smaller than the absolute value of the second difference (if it is No in step S903), proceed to step S905.

[0157] In step S905, the selection unit 550 selects the first predicted film formation result and the first optimized film formation conditions notified by the first prediction unit 520.

[0158] On the other hand, in step S903, if it is determined that the absolute value of the first difference is the same as or greater than the absolute value of the second difference (if it is Yes in step S903), then proceed to step S904.

[0159] In step S904, the selection unit 550 selects the second predicted film formation result and the second optimized film formation conditions notified by the second prediction unit 540.

[0160] <Specific examples of selection processing>

[0161] Next, we will explain a specific example of the selected processing method. Figure 10 This is a diagram illustrating a specific example of the selection process. Wherein, Figure 10 (a-1) to (a-3) show the first predicted film formation results when the first optimal film formation conditions are calculated using the learned linear model 521. Figure 10 In (a-1) to (a-3), reference numeral 1010a is a schematic diagram showing the processed wafer 130, and the difference in density indicates the difference in film thickness. In addition, graph 1020a shows the film thickness in radius 1011a, the dotted line 1021a represents the target film formation result, and the solid line 1022a represents the first predicted film formation result.

[0162] on the other hand, Figure 10 (b-1) to (b-3) show the second predicted film formation results when the second optimal film formation conditions are calculated using the learned Gaussian process regression model 541. Figure 10 In (b-1) to (b-3), reference numeral 1010b is a schematic diagram showing the processed wafer 130, and the difference in density indicates the difference in film thickness. In addition, graph 1020b shows the film thickness within a radius of 1011b, the dotted line 1021b represents the target film deposition result, the solid line 1022b represents the second predicted film deposition result, and the shaded area 1023b represents the range of the confidence interval.

[0163] Figure 10 (a-1) and Figure 10 (b-1) illustrates the selection process when the same initial film-forming conditions and the same target film-forming conditions are input during the prediction stage. In this case, the selection unit 550 determines the following and selects the second predicted film-forming result (Figure 1002b).

[0164] • In step S902, the range of the confidence interval is below the threshold.

[0165] In step S903, |first predicted film formation result - target film formation result| ≥ |second predicted film formation result - target film formation result|

[0166] in addition, Figure 10 (a-2) and Figure 10 (b-2) shows that the same initial film-forming conditions and the same target film-forming conditions are input during the prediction phase (however, with...). Figure 10 In cases where (a-1) and (b-1) are different, the selection unit 550 makes the following determination and selects the first predicted film formation result (Figure 1020a).

[0167] • In step S902, the range of the confidence interval is below the threshold.

[0168] In step S903, |first predicted film formation result - target film formation result| < |second predicted film formation result - target film formation result|

[0169] in addition, Figure 10 (a-3) and Figure 10 (b-3) shows that the same initial film-forming conditions and the same target film-forming conditions are input during the prediction phase (however, with...). Figure 10 In cases where (a-1), (a-2), (b-1), and (b-2) are different, the selection unit 550 makes the following determination and selects the first predicted film formation result (Figure 1020a).

[0170] • In step S902, the range of the confidence interval is larger than the threshold.

[0171] Thus, when using the learned Gaussian process regression model 541, given that the prediction accuracy may be lower than that of the learned linear model 521 based on the initial film formation conditions and the target film formation result, the selection unit 550 refers to the range of the confidence interval and the absolute value of the difference. Therefore, the selection unit 550 can output more suitable optimized film formation conditions and predicted film formation results.

[0172] Summary

[0173] As can be clearly seen from the above explanation, the film-forming condition output device of the first embodiment performs the following calculations.

[0174] • Using a linear model that models the film-forming process based on the film-forming device, the first optimal film-forming conditions for achieving the target film-forming result are calculated.

[0175] • A Gaussian process regression model that models the film formation process based on the film formation device is used to calculate the second optimal film formation condition for achieving the target film formation result.

[0176] Based on the determination result of whether the range of the confidence interval of the predicted film formation result when the second optimal film formation condition is calculated meets the specified conditions, either the first optimal film formation condition or the second optimal film formation condition is selected for output.

[0177] Therefore, according to the first embodiment, a film-forming condition output device or the like can be provided that can output more suitable film-forming conditions.

[0178] [Second Implementation]

[0179] In the first embodiment described above, the case of using a Gaussian process regression model as a nonlinear regression model was explained. However, the model used as a nonlinear regression model is not limited to the Gaussian process regression model; any nonlinear regression model capable of calculating the prediction result and the confidence interval of the prediction result is acceptable, and other nonlinear regression models may also be used.

[0180] Furthermore, in the first embodiment described above, the case where the film-forming condition output program is installed in the film-forming condition output device 140 was explained (that is, the film-forming condition output device 140 was described as being configured separately from the film-forming device 120). However, it is also possible to configure a storage unit containing the film-forming condition output program within the film-forming device 120, and execute the program through the film-forming device 120, thereby enabling the film-forming device 120 to function as a film-forming condition output device.

[0181] It should be noted that other elements may be combined with the configurations described in the above embodiments, and the present invention is not limited to the configurations shown herein. In this regard, changes can be made without departing from the spirit of the invention, and appropriate decisions can be made depending on the application method.

Claims

1. A film-forming condition output device, comprising: The first calculation unit uses a linear model that models the film formation process based on the film formation apparatus to calculate the first film formation conditions for achieving the target film formation result. The second calculation unit uses a nonlinear regression model that models the film formation process based on the above-mentioned film formation apparatus and is capable of calculating the confidence interval of the predicted film formation result to calculate the second film formation condition for achieving the above-mentioned target film formation result. as well as The selection unit selects and outputs either the first film-forming condition or the second film-forming condition based on a comparison between the range of the confidence interval of the predicted film-forming result when the second film-forming condition is calculated and a threshold. When the range of the aforementioned confidence interval is below the aforementioned threshold, the selection unit compares the first difference and the second difference, and based on the comparison result, selects and outputs either the first film-forming condition or the second film-forming condition. The first difference is the absolute value of the difference between the first film-forming result when the first film-forming condition is calculated and the target film-forming result, and the second difference is the absolute value of the difference between the second film-forming result when the second film-forming condition is calculated and the target film-forming result.

2. The film-forming condition output device according to claim 1, wherein, If the range of the aforementioned confidence interval is larger than the aforementioned threshold, the aforementioned selection unit selects and outputs the aforementioned first film-forming condition.

3. The film-forming condition output device according to claim 1, wherein, When the first difference is smaller than the second difference, the selection unit selects and outputs the first film-forming condition; when the first difference is the same as or larger than the second difference, the selection unit selects and outputs the second film-forming condition.

4. The film-forming condition output device according to claim 1, wherein, The linear model and the nonlinear regression model described above are learned models whose parameters are updated in a way that makes the predicted film formation result under the input film formation conditions close to the corresponding film formation result.

5. A method for outputting film-forming conditions, comprising: The first calculation step uses a linear model that models the film-forming process based on the film-forming device to calculate the first film-forming conditions for achieving the target film-forming result. The second calculation step uses a nonlinear regression model that models the film formation process based on the above-mentioned film formation apparatus and is capable of calculating the confidence interval of the predicted film formation result to calculate the second film formation condition for achieving the above-mentioned target film formation result. as well as The selection process involves comparing the range of the confidence interval of the predicted film formation result when the second film formation condition is calculated with a threshold value, and then selecting and outputting either the first or the second film formation condition. When the range of the aforementioned confidence interval is below the aforementioned threshold, in the aforementioned selection process, the first difference and the second difference are compared, and based on the comparison result, either the aforementioned first film-forming condition or the aforementioned second film-forming condition is selected and output. The first difference is the absolute value of the difference between the first film-forming result when the aforementioned first film-forming condition is calculated and the target film-forming result, and the second difference is the absolute value of the difference between the second film-forming result when the aforementioned second film-forming condition is calculated and the target film-forming result.

6. A computer-readable storage medium storing a film-forming condition output program thereon, which, when executed by a processor, performs the following steps: The first calculation step uses a linear model that models the film-forming process based on the film-forming device to calculate the first film-forming conditions for achieving the target film-forming result. The second calculation step uses a nonlinear regression model that models the film formation process based on the above-mentioned film formation apparatus and is capable of calculating the confidence interval of the predicted film formation result to calculate the second film formation condition for achieving the above-mentioned target film formation result. as well as The selection process involves comparing the range of the confidence interval of the predicted film formation result when the second film formation condition is calculated with a threshold value, and then selecting and outputting either the first or the second film formation condition. When the range of the aforementioned confidence interval is below the aforementioned threshold, in the aforementioned selection process, the first difference and the second difference are compared, and based on the comparison result, either the aforementioned first film-forming condition or the aforementioned second film-forming condition is selected and output. The first difference is the absolute value of the difference between the first film-forming result when the aforementioned first film-forming condition is calculated and the target film-forming result, and the second difference is the absolute value of the difference between the second film-forming result when the aforementioned second film-forming condition is calculated and the target film-forming result.

7. A film-forming apparatus having the computer-readable storage medium of claim 6.

Images