Etching Processing System, Etching Quality Prediction Method, and Etching Quality Prediction Program

Abstract

To reduce time to obtain etching quality inspection results.SOLUTION: An etching process system comprises a prediction unit that predicts etching quality of a substrate to be predicted by inputting the image data of the substrate to be predicted into a learned model that has been learned using learning data in which image data of the substrate taken by an imaging device placed on a substrate transfer path and information to predict the etching quality of the substrate have been mapped.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present disclosure relates to an etching processing system, an etching quality prediction method, and an etching quality prediction program.

Background Art

[0002] In the manufacturing process of a substrate, for example, various inspections are performed on the etched substrate. Among these, for inspections of so-called etching quality (also referred to as the outcome of etching), such as CD (Critical Dimension) values and etching amounts, generally, an external inspection apparatus is used. For this reason, it takes a certain amount of time for the etching processing system to obtain the inspection result of the substrate after the etching processing.

[0003] Due to such circumstances, in the current etching processing system, for example, it is difficult to perform control such as feeding back the inspection result of the etching quality to the etching processing of the next substrate.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] The present disclosure shortens the time until the inspection result of the etching quality is obtained.

Means for Solving the Problems

[0006] An etching processing system according to an aspect of the present disclosure has, for example, the following configuration. That is, A prediction unit that predicts the etching quality of a substrate to be predicted is provided. The prediction is made by inputting image data of the substrate to be predicted into a learned model that has been trained using learning data in which image data of the substrate captured by an imaging device arranged on the conveyance path of the substrate is associated with information for predicting the etching quality of the substrate.

Advantages of the Invention

[0007] The time until the inspection result of the etching quality is obtained can be shortened.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Embodiments for Carrying Out the Invention

[0009] Hereinafter, each embodiment will be described with reference to the accompanying drawings. In this specification and the drawings, components having substantially the same functional configuration are denoted by the same reference numerals, and redundant descriptions are omitted.

[0010] [First Embodiment] <System Configuration of the Etching Processing System> First, the system configuration of the etching processing system according to the first embodiment will be described separately for the learning phase and the prediction phase.

[0011] (1) In the case of the learning phase FIG. 1 is a first diagram showing an example of the system configuration of the etching processing system in the learning phase. As shown in FIG. 1, the etching processing system 100 in the learning phase includes a processing chamber 101, an imaging device 102, and a learning device 103.

[0012] The example of FIG. 1 shows that the substrate before processing is conveyed along the conveyance path in the etching processing system 100, etched in the processing chamber 101, then conveyed along the conveyance path as the substrate after processing, photographed by the imaging device 102, and then carried out.

[0013] Among these, the processing chamber 101 etches the accommodated substrate before processing to generate a substrate after processing.

[0014] The imaging device 102 is arranged on the conveyance path in the etching processing system 100. In this embodiment, the imaging device 102 photographs the substrate after processing to generate post-processing image data. Further, the imaging device 102 transmits the generated post-processing image data to the learning device 103. It is assumed that the post-processing image data generated by the imaging device 102 is image data including a plurality of color components (for example, R component, G component, B component).

[0015] The learning device 103 acquires the post-processing image data transmitted from the imaging device 102, and also acquires the etching quality information output when the external inspection device 110 inspects the etching quality of the substrate after processing.

[0016] Further, the learning device 103 generates learning data in which the post-processing image data and the etching quality information are associated, and generates a learned model by performing model learning using the generated learning data. Furthermore, the learning device 103 sets the model parameters of the generated learned model in a prediction device described later.

[0017] The external inspection device 110 inspects the etching quality of the substrate after processing carried out from the etching processing system 100. In the external inspection device 110 in this embodiment, as the inspection of the etching quality, for example, an absolute value indicating the quality of the etching is measured to generate etching quality information. The absolute value indicating the quality of the etching includes, for example, any one of a film thickness value, a CD (Critical Dimension) value, a value related to the cross-sectional shape of the substrate, and the like.

[0018] (2) In the prediction phase FIG. 2 is a first diagram showing an example of the system configuration of the etching processing system in the prediction phase. As shown in FIG. 2, the etching processing system 200 in the prediction phase includes a processing chamber 101, an imaging device 102, and a prediction device 203.

[0019] The example of FIG. 2 shows a state where the substrate before processing is conveyed on the conveyance path in the etching processing system 200, etched in the processing chamber 101, then conveyed on the conveyance path as the substrate after processing, photographed by the imaging device 102, and then carried out.

[0020] Similar to the learning phase, the processing chamber 101 etches the accommodated substrate before processing to generate a substrate after processing.

[0021] In addition, the imaging device 102 is disposed on the conveyance path in the etching processing system 200. In this embodiment, by photographing the substrate after processing, image data after processing is generated. Further, the imaging device 102 transmits the generated image data after processing to the prediction device 203.

[0022] The prediction device 203 predicts the etching quality by inputting the image data after processing transmitted from the imaging device 102 into the learned model, and outputs prediction etching quality information.

[0023] Thus, according to the etching processing system 200, every time one substrate before processing is etched, the etching quality can be predicted and prediction etching quality information can be output. Thereby, compared with the case where an external inspection device inspects the etching quality and outputs the etching quality information, the time until the etching processing system obtains the inspection result of the etching quality can be significantly shortened.

[0024] As a result, according to the etching processing system 200, it becomes possible to perform control such as feeding back the inspection result of the etching quality to the etching process of the substrate before the next process in the processing chamber 101 (refer to the dotted line).

[0025] <Arrangement example of imaging device> Next, an arrangement example of the imaging device 102 in the etching processing system 100 or 200 will be described. FIG. 3 is a diagram showing an arrangement example of the imaging device. Among these, FIG. 3(a) is a diagram showing the module configuration of the entire etching processing system 100 or 200.

[0026] Here, in explaining the arrangement example of the imaging device 102, first, the module configuration of the entire etching processing system 100 or 200 including the specific module in which the imaging device 102 is arranged will be briefly described.

[0027] As shown in FIG. 3(a), the etching processing system 100 or 200 includes, for example, · Six processing chambers PM (Process Module), · A transfer chamber VTM (Vacuum Transfer Module), · Two load lock chambers LLM (Load Lock Module), · A loader module (Loader Module), · Three load ports LP (Load Port), and has.

[0028] The six processing chambers PM are arranged around the transfer chamber VTM and perform an etching process on the substrate before processing. Note that the processing chamber 101 shown in FIG. 1 or FIG. 2 refers to any one of the six processing chambers PM.

[0029] Inside the transfer chamber VTM, a transfer device VA is arranged, which transfers the substrate before processing from the two load lock chambers LLM to the six processing chambers PM, and transfers the processed substrate after the etching process from the six processing chambers PM to the load lock chambers LLM.

[0030] The load lock chamber LLM is provided between the transfer chamber VTM and the loader module LM and switches between an atmospheric atmosphere and a vacuum atmosphere.

[0031] Inside the loader module LM, a transfer device LA for transferring the substrate before processing or the substrate after processing is arranged. The transfer device LA transfers the substrate before processing accommodated in a FOUP (Front Opening Unified Pod) attached to each load port LP to two load lock chambers LLM. Also, the transfer device LA transfers the substrate after processing from the two load lock chambers LLM to an empty FOUP attached to each load port LP. That is, the load port LP serves as an entrance and exit when loading the substrate before processing and unloading the substrate after processing.

[0032] The loader module LM is provided with an orienter ORT for aligning the position of the substrate before processing. The orienter ORT detects the center position, eccentricity amount, and notch position of the substrate before processing. Correction of the substrate before processing based on the detection result by the orienter ORT is performed using the robot arms AA and AB of the loader module LM.

[0033] FIG. 3(b) is a diagram schematically showing the inside of the orienter ORT where the imaging device 102 is arranged. As shown in FIG. 3(b), the imaging device 102 is arranged, for example, in the vicinity of a detector 301 that detects the notch position of the substrate before processing (reference sign W).

[0034] In this way, by arranging the imaging device 102 on the transfer path in the etching processing system 100 or 200, the imaging device 102 can capture an image of the substrate after etching processing (or the substrate before etching processing).

[0035] <Hardware Configuration of Learning Device and Prediction Device> Next, the hardware configurations of the learning device 103 and the prediction device 203 will be described. In this embodiment, since the learning device 103 and the prediction device 203 have the same hardware configuration, the hardware configuration of the learning device 103 will be described here.

[0036] FIG. 4 is a diagram showing an example of the hardware configuration of the learning device. As shown in FIG. 4, the learning device 103 includes a processor 401, a memory 402, an auxiliary storage device 403, a user interface device 404, a connection device 405, and a communication device 406. Each hardware of the learning device 103 is interconnected via a bus 407.

[0037] The processor 401 includes various arithmetic devices such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 401 reads various programs (for example, a learning program, etc.) onto the memory 402 and executes them.

[0038] The memory 402 includes main storage devices such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The processor 401 and the memory 402 form a so-called computer, and the computer realizes various functions by the processor 401 executing various programs read onto the memory 402.

[0039] The auxiliary storage device 403 stores various programs and various data used when various programs are executed by the processor 401. Note that the learning data storage unit 620 described later is realized in the auxiliary storage device 403.

[0040] The user interface device 404 includes, for example, a keyboard or a touch panel for the user of the learning device 103 to perform input operations of various commands, or a display device for displaying the internal state of the learning device 103 to the user.

[0041] The connection device 405 is a connection device that connects to each part (for example, the imaging device 102, etc.) within the etching processing system 100. The communication device 406 is a communication device for communicating with an external device (not shown) via a network.

[0042] <Specific examples of learning data> Next, specific examples of the learning data generated by the learning device 103 will be described. FIG. 5 is a first diagram showing specific examples of the learning data. As shown in FIG. 5, the learning data 500 includes "input data" and "correct answer data" as items of information.

[0043] In the "input data", post - processed image data generated by photographing the post - processed substrate etched in the processing chamber 101 with the imaging device 102 is stored. The example in FIG. 5 shows a state where post - processed image data with file names such as "post - processed image data 1", "post - processed image data 2", ··· is stored.

[0044] In the "correct answer data", etching quality information output by carrying out the post - processed substrate etched in the processing chamber 101 and performing an etching quality inspection with the external inspection device 110 is stored in association with the post - processed image data. The example in FIG. 5 shows a state where, as etching quality information, file names such as "etching quality information 1", "etching quality information 2", ··· are stored in association with file names such as "post - processed image data 1", "post - processed image data 2", ···

[0045] <Functional configuration of the learning device> Next, the functional configuration of the learning device 103 will be described. FIG. 6 is a first diagram showing an example of the functional configuration of the learning device. As described above, a learning program is installed in the learning device 103, and when the program is executed, the learning device 103 functions as a learning unit 610.

[0046] The learning unit 610 has a model 611 and a comparison / modification unit 612. The learning unit 610 reads the learning data 500 from the learning data storage unit 620 and inputs the processed image data (file names: "processed image data 1", "processed image data 2",...) stored in the "input data" into the model 611. Thereby, the model 611 outputs output data. Also, the learning unit 610 inputs the etching quality information (file names: "etching quality information 1", "etching quality information 2",...) stored in the "correct answer data" into the comparison / modification unit 612.

[0047] The comparison / modification unit 612 updates the model parameters of the model 611 so that the output data output by the model 611 approaches the etching quality information input by the learning unit 610. Thereby, the learning unit 610 generates a learned model.

[0048] <Functional configuration of the prediction device> Next, the functional configuration of the prediction device 203 will be described. FIG. 7 is a first diagram showing an example of the functional configuration of the prediction device. An etching quality prediction program is installed in the prediction device 203, and when the program is executed, the prediction device 203 functions as an etching quality prediction unit 710.

[0049] The etching quality prediction unit 710 is an example of a prediction unit. The etching quality prediction unit 710 has a learned model 711 generated by the learning unit 610, inputs the processed image data of the substrate after processing to be predicted into the learned model 711, predicts the etching quality, and outputs predicted etching quality information.

[0050] <Specific example of predicted etching quality information> Next, a specific example of the predicted etching quality information output from the prediction device 203 will be described while comparing it with the etching quality information output from the external inspection device 110. Here, it is assumed that the external inspection device 110 is a film thickness measurement device, and the etching quality information and the predicted etching quality information are the actually measured film thickness value or the predicted film thickness value.

[0051] FIG. 8 is a first diagram showing the relationship between the etching quality information and the predicted etching quality information. In FIG. 8, the horizontal axis represents the actually measured film thickness value output from the film thickness measuring device, and the vertical axis represents the predicted film thickness value output from the prediction device 203. Each plot represents a set of the actually measured film thickness value and the predicted film thickness value at a plurality of measurement points for each of the plurality of processed substrates. Note that the straight line 801 indicates the position where the actually measured film thickness value and the predicted film thickness value coincide.

[0052] As shown in FIG. 8, each plot position is generally along the straight line 801, and no plot that greatly deviates from the straight line 801 is found. Therefore, it can be said that the predicted film thickness value output by the prediction device 203 can generally reproduce the actually measured film thickness value. Thus, according to the prediction device 203, good prediction accuracy can be achieved.

[0053] <Flow of Etching Quality Prediction Process> Next, the flow of the etching quality prediction process by the etching processing systems 100 and 200 will be described. FIG. 9 is a flowchart showing the flow of the etching quality prediction process.

[0054] First, the processes in the learning phase by the etching processing system 100 (steps S901 to S904) are performed.

[0055] In step S901, the learning device 103 acquires the processed image data generated by photographing the processed substrate after the etching process.

[0056] In step S902, the learning device 103 acquires the etching quality information output by performing the etching quality inspection on the processed substrate after the etching process by the external inspection device 110.

[0057] In step S903, the learning device 103 generates learning data using the acquired post - processing image data as input data and the acquired etching quality information as correct answer data.

[0058] In step S904, the learning device 103 performs learning using the generated learning data and generates a learned model.

[0059] When the processing in the learning phase (steps S901 to S904) by the etching processing system 100 is completed, subsequently, the processing in the prediction phase (steps S905 to S907) by the etching processing system 200 is performed.

[0060] In step S905, the prediction device 203 acquires post - processing image data generated by photographing the post - processed substrate to be predicted that has been etched.

[0061] In step S906, the prediction device 203 inputs the acquired post - processing image data into the learned model, predicts the etching quality, and outputs predicted etching quality information.

[0062] In step S907, the prediction device 203 determines whether to end the etching quality prediction process. If it is determined not to end in step S907 (if No in step S907), the process returns to step S905.

[0063] On the other hand, if it is determined to end in step S907 (if Yes in step S907), the etching quality prediction process ends.

[0064] <Summary> As is clear from the above description, the etching processing system according to the first embodiment · Acquires post - processing image data of a substrate photographed by an imaging device arranged on the conveyance path of the substrate and etching quality information of the substrate output by being inspected by an external inspection device. · A learned model is generated by performing learning using learning data in which post - processing image data of a substrate and etching quality information of the substrate are associated with each other. · By inputting the post - processing image data of the substrate to be predicted into the learned model, the etching quality of the substrate to be predicted is predicted, and predicted etching quality information is output.

[0065] Thus, in the etching processing system according to the first embodiment, an imaging device is arranged inside, and by operating the learned model using the post - processing image data of the substrate, an inspection result of the etching quality is obtained. Thereby, according to the first embodiment, as compared with the case where an external inspection device performs an inspection on the post - processed substrate carried out from the etching processing system, the time until the inspection result of the etching quality is obtained can be significantly shortened.

[0066] [Second Embodiment] In the above - described first embodiment, the case of predicting an absolute value indicating the result of etching by using the post - processing image data generated by photographing the post - processed substrate etched in the processing chamber 101 has been described.

[0067] In contrast, in the second embodiment, · pre - processing image data generated by photographing the pre - processed substrate before the etching process in the processing chamber 101, and · post - processing image data generated by photographing the post - processed substrate after the etching process in the processing chamber 101, The case of predicting a relative value indicating the result of etching as the etching quality by using the difference image data therebetween will be described. The relative value indicating the result of etching includes, for example, any one of an etching amount, an etching rate, a ΔCD value, etc. Hereinafter, the second embodiment will be described centering on the differences from the first embodiment.

[0068] <System Configuration of the Etching Processing System> First, the system configuration of the etching processing system according to the second embodiment will be described separately for the learning phase and the prediction phase.

[0069] (1) In the case of the learning phase FIG. 10 is a second diagram showing an example of the system configuration of the etching processing system in the learning phase. As shown in FIG. 11, the etching processing system 1000 in the learning phase includes a processing chamber 101, an imaging device 102, and a learning device 1003. In the learning phase of the present embodiment, the substrate before processing and the substrate after processing are configured to be inspected for items related to etching quality by an external inspection device 110.

[0070] Specifically, in the learning phase of the present embodiment, the substrate before processing and the substrate after processing are processed according to the following procedure.

[0071] First, the substrate before processing is inspected for items related to etching quality (for example, when the etching quality is "etching amount", the film thickness of the substrate before processing) in the external inspection device 110. After that, the substrate before processing is transported on the transport path in the etching processing system 1000 and photographed by the imaging device 102 arranged on the transport path.

[0072] Note that the pre-processing quality information (for example, the film thickness value of the substrate before processing) output by inspecting items related to etching quality in the external inspection device 110 is transmitted to the learning device 1003. Also, the pre-processing image data generated by photographing with the imaging device 102 is transmitted to the learning device 1003.

[0073] The pre - processed substrate photographed by the imaging device 102 is housed in the processing chamber 101 and undergoes etching processing in the processing chamber 101. The post - processed substrate generated by the etching processing is conveyed along the conveyance path and photographed by the imaging device 102 arranged on the conveyance path. Further, the post - processed substrate photographed by the imaging device 102 is carried out and inspected in the external inspection device 110 for items related to etching quality (for example, when the etching quality is "etching amount", the film thickness of the post - processed substrate).

[0074] Note that the post - processed image data generated by photographing the post - processed substrate with the imaging device 102 is transmitted to the learning device 1003. Also, the post - processed quality information (for example, the film thickness value of the post - processed substrate) output when the post - processed substrate is inspected for items related to etching quality in the external inspection device 110 is transmitted to the learning device 1003.

[0075] The learning device 1003 acquires the pre - processed image data and the post - processed image data transmitted from the imaging device 102, and also acquires the pre - processed quality information and the post - processed quality information transmitted from the external inspection device 110.

[0076] Also, the learning device 1003 calculates the difference between the pre - processed image data and the post - processed image data and generates difference image data. Also, the learning device 1003 calculates the difference between the pre - processed quality information and the post - processed quality information to calculate the etching quality information. Also, the learning device 1003 generates learning data in which the calculated etching quality information is associated with the difference image data. Also, the learning device 103 generates a learned model by performing model learning using the generated learning data. Further, the learning device 1003 sets the model parameters of the generated learned model in a prediction device described later.

[0077] (2) In the case of the prediction phase FIG. 11 is a second diagram showing an example of the system configuration of the etching processing system in the prediction phase. As shown in FIG. 11, the etching processing system 1100 in the prediction phase includes a processing chamber 101, an imaging device 102, and a prediction device 1103.

[0078] In the prediction phase of the present embodiment, the substrate before processing and the substrate after processing are processed according to the following procedure.

[0079] First, the substrate before processing is conveyed on the conveyance path in the etching processing system 1100 and photographed by the imaging device 102 disposed on the conveyance path. Note that the pre-process image data generated by being photographed by the imaging device 102 is transmitted to the prediction device 1103.

[0080] The substrate before processing photographed by the imaging device 102 is accommodated in the processing chamber 101 and etched in the processing chamber 101. The substrate after processing generated by the etching process is conveyed on the conveyance path, photographed by the imaging device 102 disposed on the conveyance path, and then carried out.

[0081] Note that the post-process image data generated by the substrate after processing being photographed by the imaging device 102 is transmitted to the prediction device 1103.

[0082] The prediction device 1103 calculates the difference between the pre-process image data and the post-process image data transmitted from the imaging device 102, and generates difference image data. Further, the prediction device 1103 inputs the generated difference image data into the learned model to predict the etching quality and outputs prediction etching quality information.

[0083] Thus, according to the etching processing system 1100, every time one pre-processing substrate is etched, the etching quality can be predicted and the predicted etching quality information can be output. As a result, compared with the case where an external inspection device inspects items related to the etching quality and outputs the pre-processing quality information and the post-processing quality information, the time until the etching processing system obtains the inspection result of the etching quality can be significantly shortened.

[0084] As a result, according to the etching processing system 1100, it becomes possible to perform control such as feeding back the inspection result of the etching quality to the etching processing of the next pre-processing substrate in the processing chamber 101 (see the dotted line).

[0085] <Specific Example of Learning Data> Next, a specific example of the learning data generated by the learning device 1003 will be described. FIG. 12 is a second diagram showing a specific example of the learning data. As shown in FIG. 12, the learning data 1200 includes "acquired image data", "input data", "acquired quality information", and "correct answer data" as information items.

[0086] In the "acquired image data", the pre-processing image data and the post-processing image data generated by photographing the pre-processing substrate before being etched in the processing chamber 101 and the post-processing substrate after being etched by the imaging device 102 are stored respectively. The example in FIG. 12 shows a state where file names such as "pre-processing image data 1" and "pre-processing image data 2" are stored as the pre-processing image data, and file names such as "post-processing image data 1" and "post-processing image data 2" are stored as the post-processing image data.

[0087] In the "input data", the difference image data generated by calculating the difference between the pre-processing image data stored in the "acquired image data" and the corresponding post-processing image data is stored. The example in FIG. 12 shows a state where file names such as "difference image data 1" and "difference image data 2" are stored as the difference image data.

[0088] In "acquired quality information", pre - processing quality information and post - processing quality information, which are output when the pre - processing substrate and the post - processing substrate are each inspected for items related to etching quality in the external inspection device 110, are stored. The example in FIG. 12 shows that as pre - processing quality information, file names such as "pre - processing quality information 1", "pre - processing quality information 2",... are stored, and as post - processing quality information, file names such as "post - processing quality information 1", "post - processing quality information 2",... are stored.

[0089] In "correct data", etching quality information generated by calculating the difference between the pre - processing quality information stored in "acquired quality information" and the corresponding post - processing quality information is stored. The example in FIG. 12 shows that as etching quality information, file names such as "etching quality information 1", "etching quality information 2",... are stored.

[0090] <Functional configuration of the learning device> Next, the functional configuration of the learning device 1003 will be described. FIG. 13 is a second figure showing an example of the functional configuration of the learning device. A learning program is installed in the learning device 1003, and when the program is executed, the learning device 1003 functions as a difference calculation unit 1310 and a learning unit 1320.

[0091] The difference calculation unit 1310 acquires pre - processing image data and post - processing image data generated by photographing the pre - processing substrate and the post - processing substrate respectively by the imaging device 102.

[0092] Also, the difference calculation unit 1310 calculates the difference between the acquired pre - processing image data and post - processing image data, and generates difference image data. Further, the difference calculation unit 1310 stores the generated difference image data in the "input data" and "acquired image data" of the learning data 1200 in the learning data storage unit 1330 in association with the pre - processing image data and the post - processing image data respectively.

[0093] Further, the difference calculation unit 1310 acquires the pre - processing quality information and the post - processing quality information, which are output when the external inspection device 110 inspects items related to the etching quality for the pre - processing substrate and the post - processing substrate, respectively.

[0094] Also, the difference calculation unit 1310 calculates the difference between the acquired pre - processing quality information and the post - processing quality information, and generates etching quality information. Further, the difference calculation unit 1310 stores the generated etching quality information in the "correct answer data" and "acquired quality information" of the learning data 1200 in the learning data storage unit 1330 in association with the pre - processing quality information and the post - processing quality information, respectively.

[0095] The learning unit 1320 includes a model 1321 and a comparison / modification unit 1322. The learning unit 1320 reads the learning data 1200 from the learning data storage unit 1330 and inputs the difference image data (file names: "difference image data 1", "difference image data 2", ···, etc.) stored in the "input data" to the model 1321. Thereby, the model 1321 outputs output data. Also, the learning unit 1320 inputs the etching quality information (file names: "etching quality information 1", "etching quality information 2", ···, etc.) stored in the "correct answer data" to the comparison / modification unit 1322.

[0096] The comparison / modification unit 1322 updates the model parameters of the model 1321 so that the output data output by the model 1321 approaches the etching quality information input by the learning unit 1320. Thereby, the learning unit 1320 generates a learned model.

[0097] <Relationship between difference image data and etching quality information> Next, the relationship between the difference image data generated by the difference calculation unit 1310 and the etching quality information will be described. Here, as an example, the relationship between the difference image data and the etching amount will be described.

[0098] FIG. 14 is a diagram showing the relationship between differential image data and etching amount. In FIG. 14, the horizontal axis represents the luminance value of the differential image data (the difference in luminance value between the pre-process image data and the post-process image data), and the vertical axis represents the etching amount. Each plot is performed according to the following procedure. · For the substrate before processing, measure the film thickness at a plurality of measurement points, generate film thickness value information including a plurality of film thickness values, and output the same. · Photograph the substrate before processing, and extract the luminance values at the plurality of measurement points from the pre-process image data. · Generate a substrate after processing by subjecting the substrate before processing to an etching process. · Photograph the substrate after processing, and extract the luminance values at the plurality of measurement points from the post-process image data. · For the substrate after processing, measure the film thickness at the plurality of measurement points, generate film thickness value information including a plurality of film thickness values, and output the same. · At each of the plurality of measurement points, generate a pair of the difference in luminance value and the difference in film thickness value (i.e., etching amount), and plot the pair on a graph.

[0099] In FIG. 14, reference numeral 1410 represents an approximate straight line calculated based on the pair of the difference in luminance value and the etching amount. As shown in FIG. 14, there is a certain degree of correlation between the differential image data and the etching amount.

[0100] <Functional Configuration of Prediction Device> Next, the functional configuration of the prediction device 1103 will be described. FIG. 15 is a second diagram showing an example of the functional configuration of the prediction device. An etching quality prediction program is installed in the prediction device 1103, and when the program is executed, the prediction device 1103 functions as a differential image generation unit 1510 and an etching quality prediction unit 1520.

[0101] The differential image generation unit 1510 acquires the pre-process image data and the post-process image data of the substrate before processing and the substrate after processing to be predicted. Further, the differential image generation unit 1510 calculates the difference between the pre-process image data and the post-process image data, and generates differential image data.

[0102] The etching quality prediction unit 1520 is another example of the prediction unit and has a learned model 1521 generated by the learning unit 1320. The etching quality prediction unit 1520 predicts the etching quality by inputting the differential image data generated by the differential image generation unit 1510 into the learned model 1521, and outputs predicted etching quality information.

[0103] <Specific example of predicted etching quality information> Next, a specific example of the predicted etching quality information output from the prediction device 1103 will be described while comparing it with the etching quality information generated by calculating the difference between the pre - processing quality information and the post - processing quality information output from the external inspection device 110. FIG. 16 is a second diagram showing the relationship between the etching quality information and the predicted etching quality information. Specifically, the example in FIG. 16 is when each pre - processing substrate with different initial film thicknesses and film types is subjected to an etching process. · The etching amount output from the prediction device 1103 (vertical axis: predicted etching amount), and · The etching amount (horizontal axis: measured etching amount) generated by calculating the difference between the film thickness value of the pre - processing substrate and the film thickness value of the post - processing substrate output from the external inspection device 110, are compared.

[0104] In FIG. 16, reference numeral 1610 indicates the relationship between the predicted etching amount and the measured etching amount when the pre - processing substrate with an initial film thickness of 1.0 μm and a film type of silicon oxide film (Ox) is subjected to an etching process.

[0105] Also, reference numeral 1620 indicates the relationship between the predicted etching amount and the measured etching amount when the pre - processing substrate with an initial film thickness of 0.1 μm and a film type of silicon oxide film (Ox) is subjected to an etching process.

[0106] Also, reference numeral 1630 indicates the relationship between the predicted etching amount and the measured etching amount when the pre - processing substrate with an initial film thickness of 1.0 μm and a film type of silicon oxide film (Ox) is subjected to an etching process.

[0107] Reference numeral 1640 shows the relationship between the predicted etching amount and the measured etching amount when the substrate before processing with an initial film thickness of 1.5 μm and a film type of photoresist (PR) is etched.

[0108] Reference numeral 1650 shows the relationship between the predicted etching amount and the measured etching amount when the substrate before processing with an initial film thickness of 0.25 μm and a film type of silicon nitride film (SiN) is etched.

[0109] As shown in the example of FIG. 16, in the case of the prediction device 1103, a higher prediction accuracy can be achieved for a hard film (Ox, SiN) than for an organic film. Also, in the case of the prediction device 1103, a higher prediction accuracy can be achieved when the initial film thickness is smaller.

[0110] <Summary> As is clear from the above description, the etching processing system according to the second embodiment · Acquires the pre-processing image data of the substrate before processing and the post-processing image data of the substrate after processing photographed by the imaging device arranged on the conveyance path of the substrate, and the pre-processing quality information of the substrate before processing and the post-processing quality information of the substrate after processing output by being inspected by an external inspection device. · A learned model is generated by performing learning using the learning data in which the differential image data obtained by differentiating the pre-processing image data and the post-processing image data and the etching quality information obtained by differentiating the pre-processing quality information and the post-processing quality information are associated. · By inputting the differential image data obtained by differentiating the pre-processing image data of the substrate before processing to be predicted and the post-processing image data of the substrate after processing to be predicted into the learned model, the predicted etching quality information of the substrate before processing and the substrate after processing to be predicted is output.

[0111] As described above, in the etching processing system according to the second embodiment, an imaging device is disposed inside, and a learned model is operated using the pre-processing image data of the substrate before processing and the post-processing image data of the substrate after processing, thereby obtaining an inspection result of the etching quality. According to the second embodiment, the time required to obtain the inspection result of the etching quality can be significantly shortened as compared with the case where an external inspection device performs inspection on the substrate before processing carried into the etching processing system and the substrate after processing carried out.

[0112] [Third Embodiment] In the second embodiment described above, when predicting the etching quality, the prediction unit inputs the differential image data obtained by differentiating the pre-processing image data and the post-processing image data into the learned model, and directly predicts the etching quality.

[0113] In contrast, in the third embodiment, two learned models are arranged. One learned model predicts the pre-processing quality from the pre-processing image data, and the other learned model predicts the post-processing quality from the post-processing image data. Then, the etching quality is predicted based on the predicted pre-processing quality and post-processing quality. Hereinafter, the third embodiment will be described centering on the differences from the first and second embodiments.

[0114] [Specific Example of Learning Data] First, a specific example of the learning data generated in the third embodiment will be described. FIG. 17 is a third diagram showing a specific example of the learning data. As shown in FIG. 17, the learning data 1710 includes "input data" and "correct answer data" as information items.

[0115] The "input data" stores the pre-processing image data generated by photographing the substrate before being etched in the processing chamber 101 by the imaging device 102. The example in FIG. 17 shows a state in which file names = "pre-processing image data 1", "pre-processing image data 2",... etc. are stored as the pre-processing image data.

[0116] In the "correct data", the pre - processing quality information output by the external inspection device 110 through inspecting items related to the etching quality of the substrate before processing is stored. The example in FIG. 17 shows a state where file names such as "pre - processing quality information 1", "pre - processing quality information 2",... are stored as the pre - processing quality information.

[0117] Similarly, the learning data 1720 includes "input data" and "correct data" as information items.

[0118] In the "input data", the post - processing image data generated by the imaging device 102 photographing the substrate after etching processing in the processing chamber 101 is stored. The example in FIG. 17 shows a state where file names such as "post - processing image data 1", "post - processing image data 2",... are stored as the post - processing image data.

[0119] In the "correct data", the post - processing quality information output by the external inspection device 110 through inspecting items related to the etching quality of the substrate after processing is stored. The example in FIG. 17 shows a state where file names such as "post - processing quality information 1", "post - processing quality information 2",... are stored as the post - processing quality information.

[0120] <Functional Configuration of the Learning Device> Next, the functional configuration of the learning device will be described. FIG. 18 is a third figure showing an example of the functional configuration of the learning device. A learning program is installed in the learning device 1800, and when the program is executed, the learning device 1800 functions as a learning unit 1810 and a learning unit 1820.

[0121] The learning unit 1810 has a model 1811 and a comparison / modification unit 1812. The learning unit 1810 reads the learning data 1710 from the learning data storage unit 1830 and inputs the pre - processing image data (file names: "pre - processing image data 1", "pre - processing image data 2", ···, etc.) stored in the "input data" into the model 1811. Thereby, the model 1811 outputs output data. Also, the learning unit 1810 inputs the pre - processing quality information (file names: "pre - processing quality information 1", "pre - processing quality information 2", ···, etc.) stored in the "correct answer data" into the comparison / modification unit 1812.

[0122] The comparison / modification unit 1812 updates the model parameters of the model 1811 so that the output data output by the model 1811 approaches the pre - processing quality information input by the learning unit 1810. Thereby, the learning unit 1810 generates a learned model.

[0123] Similarly, the learning unit 1820 has a model 1821 and a comparison / modification unit 1822. The learning unit 1820 reads the learning data 1720 from the learning data storage unit 1830 and inputs the post - processing image data (file names: "post - processing image data 1", "post - processing image data 2", ···, etc.) stored in the "input data" into the model 1821. Thereby, the model 1821 outputs output data. Also, the learning unit 1820 inputs the post - processing quality information (file names: "post - processing quality information 1", "post - processing quality information 2", ···, etc.) stored in the "correct answer data" into the comparison / modification unit 1822.

[0124] The comparison / modification unit 1822 updates the model parameters of the model 1821 so that the output data output by the model 1821 approaches the post - processing quality information input by the learning unit 1820. Thereby, the learning unit 1820 generates a learned model.

[0125] <Functional configuration of the prediction device> Next, the functional configuration of the prediction device in the third embodiment will be described. FIG. 19 is a third diagram showing an example of the functional configuration of the prediction device. An etching quality prediction program is installed in the prediction device 1900, and when the program is executed, the prediction device 1900 functions as a prediction unit 1910, a prediction unit 1920, and an etching quality calculation unit 1930.

[0126] The prediction unit 1910 has a learned model 1911 generated by the learning unit 1810, and inputs the pre-process image data generated by photographing the substrate before processing to be predicted into the learned model 1911, thereby predicting the pre-process quality and outputting pre-process quality information.

[0127] The prediction unit 1920 has a learned model 1921 generated by the learning unit 1820, and inputs the post-process image data generated by photographing the substrate after processing to be predicted into the learned model 1921, thereby predicting the post-process quality and outputting post-process quality information.

[0128] The etching quality calculation unit 1930 predicts the etching quality by calculating the difference between the pre-process quality information output from the prediction unit 1910 and the post-process quality information output from the prediction unit 1920, and outputs predicted etching quality information.

[0129] <Specific example of predicted etching quality information> Next, a specific example of the predicted etching quality information output from the prediction device 1900 will be described while comparing it with the etching quality information generated by calculating the difference between the pre-process quality information and the post-process quality information output from the external inspection device 110. FIG. 20 is a third diagram showing the relationship between the etching quality information and the predicted etching quality information. Specifically, the example in FIG. 20 shows the case where each substrate before processing with different initial film thicknesses and film types is subjected to etching processing. · The etching amount (vertical axis: predicted etching amount) output from the prediction device 1900, and · By calculating the difference between the film thickness value of the substrate before processing and the film thickness value of the substrate after processing output from the external inspection device 110, the etching amount (horizontal axis: measured etching amount) generated, and is a comparison of them.

[0130] In FIG. 20, reference numeral 2010 shows the relationship between the predicted etching amount and the measured etching amount when etching a substrate before processing with an initial film thickness of 0.1 μm and a film type of silicon oxide film (Ox).

[0131] Also, reference numeral 2020 shows the relationship between the predicted etching amount and the measured etching amount when etching a substrate before processing with an initial film thickness of 0.25 μm and a film type of silicon nitride film (SiN).

[0132] As shown in the example of FIG. 20, in the case of the prediction device 1900, even for the same hard film, a higher prediction accuracy can be achieved for the silicon oxide film (Ox) than for the silicon nitride film (SiN).

[0133] <Summary> As is clear from the above description, the etching processing system according to the third embodiment · Obtains the pre-processing image data of the substrate before processing and the post-processing image data of the substrate after processing photographed by the imaging device arranged on the conveyance path of the substrate, and the pre-processing quality information of the substrate before processing and the post-processing quality information of the substrate after processing output by being inspected by the external inspection device. · Generates a first learned model by performing learning using learning data in which the pre-processing image data and the pre-processing quality information are associated. · Generates a second learned model by performing learning using learning data in which the post-processing image data and the post-processing quality information are associated. · Inputs the pre-processing image data of the substrate before processing to be predicted into the first learned model to output the pre-processing quality information of the substrate after processing to be predicted. Also, inputs the post-processing image data of the substrate after processing to be predicted into the second learned model to output the post-processing quality information of the substrate after processing to be predicted. ·Calculate the difference between the pre - processing quality information and the post - processing quality information, and output the predicted etching quality information of the substrate before processing and the substrate after processing to be predicted.

[0134] In this way, in the etching processing system according to the third embodiment, an imaging device is arranged inside, and by operating each learned model using the pre - processing image data of the substrate before processing and the post - processing image data of the substrate after processing, the inspection result of the etching quality is obtained. Thus, according to the third embodiment, compared with the case where an external inspection device performs an inspection on the substrate before processing carried into the etching processing system and the substrate after processing carried out, the time until the inspection result of the etching quality is obtained can be significantly shortened.

[0135] [Fourth Embodiment] In the above - described first to third embodiments, the prediction device has been described as having a function of outputting predicted etching quality information. However, the functions of the prediction device are not limited to this. For example, it may have a control unit that feeds back the predicted etching quality and controls the processing content of the substrate to be etched after the substrate to be predicted (for example, corrects the control amount of the control knob).

[0136] In this case, for example, a learned process control model generated by pre - learning about the relationship between the control amount of the control knob and the etching quality is arranged in the control unit, and the control unit operates the control knob with the control amount derived using the model. Specifically, when the predicted etching quality information deviates from the target value, the control unit derives a control amount using the model so as to correct the deviation amount, and operates the control knob with the derived control amount.

[0137] In this way, by adopting a configuration that controls the processing content of the etching process using the predicted etching quality information, according to the etching processing system 200, etching quality conforming to the target value can be realized.

[0138] [Fifth Embodiment] In the above-described first to fourth embodiments, the learning unit has been described as performing learning in the learning phase to generate a learned model. However, the functions of the learning unit are not limited to this. For example, in the prediction phase, when a predetermined condition is satisfied, the learning unit may have a function of performing re-learning on the generated learned model.

[0139] The case of the predetermined condition here refers to, for example, generating learning data by continuously performing the inspection of the etching quality by the external inspection device 110 even in the prediction phase, and the case where a predetermined amount of learning data is accumulated. Alternatively, it refers to the case where the difference between the etching quality information and the predicted etching quality information is monitored by continuously performing the inspection of the etching quality by the external inspection device 110 even in the prediction phase, and the difference between the two becomes equal to or greater than a predetermined threshold value.

[0140] Note that the method of re-learning by the learning unit is arbitrary. For example, a new learned model may be generated by performing learning from 1 using the learning data accumulated in the prediction phase. Alternatively, the learned model during operation in the prediction phase may be updated by performing additional learning using the learning data accumulated in the prediction phase.

[0141] Thus, since the learning unit has a function of performing re-learning, according to the etching processing system 100, the prediction accuracy can be improved after shifting to the prediction phase.

[0142] [Other Embodiments] In each of the above embodiments, the case where the learning device and the prediction device are configured as separate bodies has been described. However, the learning device and the prediction device may be configured as an integrated device.

[0143] In addition, in each of the above embodiments, the learning device and the prediction device have been described as being integrally configured with a plurality of modules that constitute the etching processing system 100 or 200. However, the learning device and the prediction device may be configured to be connected separately from the plurality of modules that constitute the etching processing system 100 or 200, for example, via a network.

[0144] In addition, in each of the above embodiments, the imaging device 102 has been described as being disposed within the orienter ORT that constitutes the etching processing system 100 or 200. However, the arrangement of the imaging device 102 is not limited to within the orienter ORT. As long as it is an arrangement capable of photographing both the pre-process substrate and the post-process substrate, it can be arranged at any position on the conveyance path. Also, the number of imaging devices to be installed is not limited to one, and a plurality of them may be used.

[0145] In addition, in each of the above embodiments, the film thickness value or the etching amount, which is the difference in the film thickness value, has been described as being used as the etching quality information or the predicted etching quality information. However, the etching quality information or the predicted etching quality information is not limited to these.

[0146] In addition, in the above second embodiment, the method for calculating the difference image data was not mentioned. However, the difference image data may be calculated separately for each color component, for example, and the difference image data of a specific color component having a higher correlation with items related to the etching quality may be selected. Alternatively, the difference image data may be calculated using the pre-process image data and the post-process image data of a specific color component. Alternatively, instead of selecting the difference image data of a specific color component, the difference image data may be processed so as to have a higher correlation with items related to the etching quality.

[0147] Note that the configurations and the like described in the above embodiments are not limited to the configurations shown here, such as combinations with other elements. Regarding these points, it is possible to make changes without departing from the spirit of the present invention, and they can be appropriately determined according to the application form.

Explanation of Reference Numerals

[0148] 100: Etching processing system 101: Processing chamber 102: Imaging device 103: Learning device 110: External inspection device 200: Etching processing system 203: Prediction device 500: Learning data 610: Learning section 710: Etching quality prediction section 1000: Etching processing system 1003: Learning device 1103: Prediction device 1200: Learning data 1310: Difference calculation section 1320: Learning section 1510: Difference image generation section 1520: Etching quality prediction section 1710: Learning data 1720: Learning data 1800: Learning device 1810: Learning section 1820: Learning section 1900: Prediction device 1910: Prediction section 1920: Prediction section 1930: Etching quality calculation section

Claims

1. A prediction unit that predicts the etching quality of a substrate to be predicted by inputting image data of the substrate captured by an imaging device disposed on the conveyance path of the substrate into a learned model that has been learned using learning data in which the image data of the substrate and information for predicting the etching quality of the substrate are associated with each other. An etching processing system having the same.

2. The learned model is a model that has been learned using learning data in which post-processed image data generated by capturing the substrate after etching processing is used as input data and etching quality information output by inspecting the etching quality of the substrate after etching processing is used as correct answer data. The etching processing system according to Claim 1.

3. The learned model uses, as input data, difference image data obtained by taking the difference between pre-processed image data generated by capturing the substrate before etching processing and post-processed image data generated by capturing the substrate after etching processing, and uses, as correct answer data, etching quality information obtained by taking the difference between pre-processed quality information output by inspecting items related to the etching quality of the substrate before etching processing and post-processed quality information output by inspecting items related to the etching quality of the substrate after etching processing. The etching processing system according to Claim 1, which is a model that has been learned using learning data.

4. The learned model includes: A first learned model that has been learned using learning data in which pre-processed image data generated by capturing the substrate before etching processing is used as input data and pre-processed quality information output by inspecting items related to the etching quality of the substrate before etching processing is used as correct answer data; and A second learned model that has been learned using learning data in which post-processed image data generated by capturing the substrate after etching processing is used as input data and pre-processed quality information output by inspecting items related to the etching quality of the substrate after etching processing is used as correct answer data. The prediction unit includes: ​ The etching quality of the substrate to be predicted is predicted based on the pre - processing quality information of the substrate to be predicted, which is output by inputting the pre - processing image data generated by photographing the substrate to be predicted before the etching process into the first learned model, and the post - processing quality information, which is output by inputting the post - processing image data generated by photographing the substrate to be predicted after the etching process into the second learned model. The etching processing system according to claim 1.

5. The prediction unit predicts, as the etching quality, any one of the film thickness value, CD value, and value related to the cross - sectional shape of the substrate to be predicted after the etching process. The etching processing system according to claim 2.

6. The items related to the etching quality include any one of the film thickness value, CD value, and value related to the cross - sectional shape. The prediction unit predicts, as the etching quality, any one of the etching amount, etching rate, and ΔCD value of the substrate to be predicted after the etching process. The etching processing system according to claim 3 or 4.

7. The pre - processing image data and the post - processing image data are image data of a specific color component. The etching processing system according to claim 6.

8. The etching processing system according to claim 1 further includes a control unit that controls the processing content of the substrate to be etched after the substrate to be predicted based on the predicted etching quality of the substrate to be predicted.

9. The etching processing system according to claim 1 further includes a learning unit that generates the learned model and, when a predetermined condition is satisfied, performs re - learning on the generated learned model.

10. A prediction step of predicting the etching quality of the substrate to be predicted by inputting the image data of the substrate, which is photographed by an imaging device arranged on the conveyance path of the substrate, into a learned model that has been learned using learning data in which the image data of the substrate and the information for predicting the etching quality of the substrate are associated. An etching quality prediction method having the above.

11. A prediction step of predicting the etching quality of a substrate to be predicted by inputting image data of the substrate to a learned model that has been learned using learning data in which the image data of the substrate photographed by an imaging device disposed on the conveyance path of the substrate is associated with information for predicting the etching quality of the substrate. An etching quality prediction program for causing a computer to execute the above.

Images