Management System and Management Method for Epitaxial Film Deposition Apparatus

Abstract

To provide a technique capable of obtaining stable quality as quality of a SiC epitaxial film on a system, etc., for controlling an epitaxial film deposition apparatus.SOLUTION: A control system 1 for creating a recipe of processing an epitaxial film deposition apparatus 20 includes a processor and a memory. The processing of the epitaxial film deposition apparatus 20 includes processing of depositing an epitaxial film using epitaxial growth on a substrate (a SiC substrate 24), and the processor creates a first processing recipe right after maintenance of the epitaxial film deposition apparatus 20 and a recipe of inputting information on a quality evaluation value of the processed epitaxial film 25 to construct or update a processing model as an initial condition and form a quality evaluation value in an allowable range including a target value as a recipe for the second and following processing right after the maintenance on the basis of the model.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to technologies such as a management system for an epitaxial film formation apparatus.

Background Art

[0002] The most important issues in the realization of a sustainable society are the depletion of energy resources and the excessive emission of greenhouse gases such as CO2. For this reason, power conversion devices with excellent energy efficiency and low CO2 emissions have become important. Many power conversion devices are composed of a power module in which an Insulated Gate Bipolar Transistor (IGBT), which is a switching element, and a PiN diode (PND), which is a rectifying element, are connected in parallel. Therefore, reducing the loss of semiconductor elements directly leads to energy savings in power conversion devices. As a technology for reducing the loss of semiconductor elements, a method of forming elements with 4H-type silicon carbide (4H-SiC: hereinafter also referred to as SiC) has attracted attention. In order to improve the reliability and reduce the cost of SiC elements, high-quality and inexpensive SiC epitaxial substrates (hereinafter also referred to as substrates) are required.

[0003] As one of semiconductor manufacturing devices, there is an epitaxial film formation apparatus. As an epitaxial film formation apparatus for SiC, there is a SiC epitaxial growth apparatus. Epitaxial growth is one of thin-film crystal growth technologies, in which crystal growth is performed on a crystal serving as a base substrate and arranged in alignment with the crystal plane of the substrate. SiC epitaxial growth is a technology for forming SiC on an off-cut SiC substrate. Generally, since the donor concentration of SiC substrates is high, it is necessary to adjust the donor concentration and film thickness according to the application of the withstand voltage used, and epitaxial growth (in other words, epitaxial film formation) is performed for the production of SiC elements. Requirements for epitaxial growth technology cover a wide range, such as the increase in the size of epitaxial growth accompanying the increase in the diameter of the substrate, ensuring the uniformity of the donor concentration, ensuring the uniformity of the epitaxial film thickness, high-speed growth, and reduction of crystal defects.

[0004] As one of the systems for performing evaluation, management, control, etc. related to semiconductor manufacturing, there is a management system for an epitaxial film deposition apparatus. This management system manages film deposition conditions (which may also be called "recipes", etc.), which are conditions for controlling the film deposition process, for setting in the film deposition apparatus. This management system grasps and evaluates the state and results of the film deposition process based on the film deposition conditions by the film deposition apparatus, and generates suitable film deposition conditions.

[0005] As a prior art example related to the above, Japanese Unexamined Patent Application Publication No. 2020-123675 (Patent Document 1) can be cited. Patent Document 1 describes a technique for determining a recipe according to the change over time of a semiconductor manufacturing apparatus as a management system for the semiconductor manufacturing apparatus.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0007] However, in the conventional semiconductor device manufacturing technology, it has been found that the quality of the SiC epitaxial film (which may be abbreviated as "epi film") generated in the film deposition process of the film deposition apparatus varies due to changes over time of the film deposition apparatus, maintenance, etc. Therefore, in order to obtain a stable epi film, there is a problem that the film deposition cost increases.

[0008] In particular, even if the present inventors have once set the film formation conditions suitable for the epitaxial film formation apparatus, it has been found that the quality of the epitaxial film of the film formation result fluctuates every time the film formation apparatus is maintained. Before and after maintenance, although it is generally difficult to quantify, fluctuations in the apparatus state occur. For example, substances associated with film formation accumulate on the wall surface of the vacuum chamber (also called a chamber) of the film formation apparatus. Immediately after maintenance, due to such fluctuations in the apparatus state, the past film formation conditions may no longer be suitable. Therefore, the quality of the epitaxial film of the film formation result immediately after maintenance may fall outside the range of the predetermined quality conditions defined by manufacturing specifications and the like.

[0009] An object of the present invention is to provide a technique capable of obtaining stable quality as the quality of a SiC epitaxial film with respect to techniques such as a management system of the epitaxial film formation apparatus.

Means for Solving the Problems

[0010] A typical embodiment of the present invention has the following configuration. The management system of the epitaxial film formation apparatus according to the embodiment is a management system that generates a recipe for the process of the epitaxial film formation apparatus, includes a processor, and the process of the epitaxial film formation apparatus includes a process of forming an epitaxial film using epitaxial growth on a substrate. The processor inputs information on the recipe of the first process immediately after maintenance of the epitaxial film formation apparatus and the evaluation value of the quality of the epitaxial film as the processing result as initial conditions, constructs or updates the model of the process, and based on the model, generates a recipe for the second and subsequent processes immediately after maintenance such that the evaluation value of the quality is within an allowable range including a target value.

Effects of the Invention

[0011] According to a typical embodiment of the present invention, regarding technologies such as the management system of the epitaxial film deposition apparatus, stable quality can be obtained as the quality of the SiC epitaxial film. Other issues, effects, etc. are shown in [Embodiments for Carrying Out the Invention].

Brief Description of the Drawings

[0012]

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

Embodiments for Carrying Out the Invention

[0013] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same reference numerals are generally assigned to the same parts, and repeated descriptions are omitted. In the drawings, the representation of each component may not represent the actual position, size, shape, and range, etc. for the purpose of facilitating the understanding of the invention, and the present invention is not necessarily limited to the position, size, shape, and range, etc. disclosed in the drawings.

[0014] In the description, when explaining the processing by a program, the program, function, processing unit, etc. may be the main subject of the explanation. However, the main hardware for these is a processor, or a controller, device, computer, system, etc. composed of such a processor. A computer executes processing according to a program read onto a memory while appropriately using resources such as a memory and a communication interface by a processor. As a result, a predetermined function, processing unit, etc. are realized. A processor is composed of, for example, a semiconductor device such as a CPU or a GPU. A processor is composed of a device or circuit capable of performing a predetermined operation. The processing is not limited to software program processing and can also be implemented by a dedicated circuit. A dedicated circuit such as an FPGA or an ASIC can be applied. The program may be pre-installed as data in the target computer, or may be distributed and installed as data from a program source to the target computer. The program source may be a program distribution server on a communication network or a non-transitory computer-readable storage medium. The program may be composed of a plurality of program modules. A computer system may be composed of a plurality of devices. A computer system may be composed of a client-server system, a cloud computing system, an IoT system, etc. In the description, various data and information may be explained in the form of, for example, a table or a list, but are not limited to such a structure and form. Data and information for identifying various elements may be explained in the form of identification information, identifier, ID, name, number, etc., but these expressions are mutually replaceable.

[0015] [Problems, etc.] A supplementary explanation will be given for the above-described problems and the like. An outline of a management system and a management method for a semiconductor manufacturing apparatus as a comparative example with respect to the embodiment will be described below. The management system of the comparative example records and holds the process history information of the film forming apparatus. The process history information is, in other words, time-series data of the process, and includes information such as the recipe used in the process and the evaluation result of the epi quality as the process result. The management system autonomously estimates the change over time of the chamber from the process history information and determines whether maintenance is required. The recipe is the film forming conditions for realizing a film forming result close to the target value using a model. In the technology of this comparative example, in order to estimate the change over time of the apparatus state from the process history information including the recipe and the epi quality, there is no film forming for confirming the change over time.

[0016] FIG. 16 shows graphs of examples of changes over time for (a) film forming conditions and (b) epi quality in the comparative example. The horizontal axis is the film forming time (corresponding time point or cycle). For example, in the period from time point t1 to time point t2, the film forming conditions are set so that the epi quality becomes a substantially constant value v1 within the target range V0. The predetermined target range V0 is a range from the lower limit value V2 to the upper limit value V3 centered on the value V1. For example, maintenance of the film forming apparatus was performed at time point tm1. As a result, the apparatus state fluctuates inherently. As a result, in the period from time point t3 after maintenance to time point t4, with the same film forming conditions as before maintenance, the epi quality has deviated from the target range V0 to a value v2. Similarly, maintenance of the film forming apparatus was performed at time point tm2. As a result, in the period from time point t5 after maintenance to time point t6, with the same film forming conditions as before maintenance, the epi quality has deviated from the target range V0 to a value v3.

[0017] As described above, in the comparative example, the inherent fluctuation of the apparatus state before and after maintenance is not considered, and the recipe is determined even immediately after maintenance assuming the same constant apparatus state before and after maintenance. Therefore, as described above, for multiple film formations after immediately after maintenance (for example, each film formation indicated by a plurality of plots from time point t3), the epi quality may deviate from the target range V0.

[0018] Since there are various factors causing fluctuations in the state of the film forming apparatus, it is generally difficult to quantify. In the embodiment, as a mechanism, it is not necessary to quantify the state of the film forming apparatus. Instead, for the recipe as the input of the model, the evaluation value of the epi-quality of the film forming result as the output is used to control the generation and adjustment of a suitable recipe before and after maintenance.

[0019] <Embodiment 1> Using FIGS. 1 to 9, the technology such as the management system of the epitaxial film forming apparatus according to Embodiment 1 of the present invention will be described. The management system of Embodiment 1 is a system having a function of generating and proposing a recipe which is an optimal film forming condition according to the apparatus state of the epitaxial film forming apparatus, and is mainly realized by a computer system. The management method of Embodiment 1 is a method having steps executed by the management system of Embodiment 1. The management system of Embodiment 1 constructs and updates, by machine learning, a model for generating film forming conditions related to epitaxial film formation, using, as an initial condition, the value of the quality of the epitaxial film (which may be described as epi-quality) of the epitaxial film forming result immediately after the maintenance of the epitaxial film forming apparatus. Note that immediately after maintenance, in other words, refers to the time point when the first film forming process is carried out after maintenance.

[0020] [Management System] FIG. 1 shows a schematic configuration of a management system for an epitaxial film deposition apparatus according to Embodiment 1. The management system 1 according to Embodiment 1 is a management system for an epitaxial film deposition apparatus 20 and is implemented by a computer system 10. The computer system 10 includes an integrated management unit 11, an apparatus control unit 12, a recipe search unit 13, an analysis and evaluation unit 14, an input device 15A, and an output device 15B, which are interconnected. Note that the input device 15A and the output device 15B may be externally connected devices. The user U1 operates and uses the management system 1 by the computer system 10 via the input device 15A and the output device 15B. The user U1 is, for example, a person who manages a manufacturing process (particularly a film deposition process). The computer system 10 is connected to a control unit 21 of the epitaxial film deposition apparatus 20 through communication 200 such as wired or wireless communication.

[0021] The epitaxial film deposition apparatus 20 includes a control unit 21 and a chamber 22. The control unit 21 controls driving and the like of each part of the film deposition apparatus. The chamber 22 is configured as, for example, a vacuum chamber and includes a stage 23 inside the vacuum chamber. The stage 23 has its movement in the horizontal direction (e.g., linear movement or rotation) controlled. On the stage 23, a SiC substrate 24 or the like is arranged as a target substrate. The film deposition apparatus 20 forms an epitaxial film 25 by performing an epitaxial film deposition process on the surface of the SiC substrate 24. Along with such an epitaxial film deposition process, a predetermined substance accumulates in the chamber 22. As an example, the deposition location 26 is the side wall surface of the chamber 22. The film deposition apparatus 20 includes a sensor 27 described later. The sensor 27 measures a predetermined physical quantity such as thickness and amount with respect to the deposit at the deposition location 26. The mode of the chamber 22 and the mode of deposition are not limited to this example and there are various modes.

[0022] The integrated management unit 11 is a part that performs input, output, storage, management, etc. of necessary data and information. The device control unit 12 is a part that controls the operation of the film forming apparatus 20 while communicating with the film forming apparatus 20. The recipe search unit 13 is a part that performs a process of searching for an optimal recipe for setting in the film forming apparatus based on machine learning or the like. The analysis and evaluation unit 14 is a part that analyzes and evaluates the result of film formation with the recipe and obtains an evaluation value of the quality of the epitaxial film or the like. The analysis and evaluation unit 14 or the user U1 obtains values such as the film thickness, impurity concentration, crystal defect density, etc. of the epitaxial film as a result of the analysis and evaluation.

[0023] [Management System - Modification Example (1)] Figure 2 shows an implementation example of the management system 1 in the modification example of Embodiment 1. In the implementation example of Figure 2, the same components as in Figure 1 are interconnected through the LAN 201. In the configuration example of Figure 2, each part such as the integrated management unit 11 is implemented by a computer system. That is, it has the computer system CS1 of the integrated management unit 11, the computer system CS2 of the device control unit 12, the computer system CS3 of the recipe search unit 13, and the computer system CS4 of the analysis and evaluation unit 14, and they are interconnected through the LAN 201. Each computer system may be composed of a connection of multiple computers, or for example, a machine with multiple GPUs connected in parallel may be applied. In the configuration example of Figure 2, rapid processing is possible by parallel distributed processing using multiple computer systems.

[0024] [Management System - Modification Example (2)] FIG. 3 shows an implementation example of the management system 1 in another modification of Embodiment 1. In the implementation example of FIG. 3, as part of the same components as in FIG. 1, the recipe search unit 13 is provided in the form of a server or the like on the cloud computing system 202 via a wide area communication network. The computer system CS5 including the integrated management unit 11 and the like is connected via a wide area communication network to the recipe search unit 13 by a server or the like on the cloud computing system 202. Since the processing of the recipe search unit 13 involves machine learning and the like, the load is relatively high, and high performance and a large amount of computing resources may be required. Therefore, in this implementation example, by implementing the recipe search unit 13 on the cloud computing system 202, rapid processing is possible. Also, in this implementation example, it is also possible to share one recipe search unit 13 among a plurality of computer systems related to a plurality of film forming apparatuses.

[0025] [Management System - Details] FIG. 4 shows a detailed configuration example of the management system 1 of Embodiment 1 in FIG. 1. In the configuration of FIG. 4, each unit such as the integrated management unit 11 is composed of a computer or a circuit having a processor 116, a ROM 117, a RAM 118, an interface 115, and a bus or the like that interconnects them. The functions of each unit may be realized by processing by a computer program or may be implemented by a dedicated circuit using an FPGA or the like. Also, in the configuration of FIG. 4, it has a controller 211 connected to the apparatus control unit 12 and a controller 212 connected to the analysis and evaluation unit 14, and these controllers correspond to a control unit or an operation unit for the user U1 to input settings and instructions.

[0026] [Epitaxial Film Forming Apparatus] FIG. 5 shows an implementation configuration example related to the epitaxial film forming apparatus 20 of FIG. 1. (A) of FIG. 5 shows an implementation configuration example in the form of a cluster apparatus. This cluster apparatus includes a substrate analysis chamber 22A, a cleaning chamber 22B, an analysis and evaluation chamber 22C, a processing chamber 22D, a regeneration chamber 22E, and a load lock chamber - transfer chamber 22F.

[0027] (B) of FIG. 5 shows an example of a mounting configuration in the form of a SiC epitaxial device. This SiC epitaxial device is a SiC epitaxial cluster device including an analysis / regeneration / growth chamber, and includes a substrate analysis chamber 22a, an epi analysis chamber 22b, a defect analysis chamber 22c, an epi growth chamber 22d, a CMP chamber 22e, and a load lock chamber / transfer chamber 22f.

[0028] The management system 1 of Embodiment 1 manages semiconductor manufacturing equipment including the epitaxial film forming equipment 20. Further, as components, the management system 1 requires an analysis and evaluation device for a substrate or an epi film, or means for acquiring analysis and evaluation result data by the analysis and evaluation device. The analysis and evaluation device or means corresponds to the analysis and evaluation unit 14 in FIG. 1, the analysis and evaluation chamber 22C in (A) of FIG. 5, and the epi analysis chamber 22b in (B) of FIG. 5. The necessary components may be in a form integrated into one device as a cluster device as in the example of (A) of FIG. 5, or may be in the form of a SiC epi device as in (B). Further, in a form provided with the substrate analysis chambers 22A and 22a, as in Embodiment 3 described later, the management system 1 may automatically acquire and use substrate information from the substrate analysis chamber or the like.

[0029] The substrate analysis chambers 22A and 22a are chambers for analyzing and evaluating a substrate (the SiC substrate 24 in FIG. 1) to be film-formed, and as a result, substrate information described later can be obtained. The processing chamber 22D is a chamber for processing the substrate. The epi growth chamber 22d is an example of the processing chamber 22D, and is a chamber for performing a film-forming process of epitaxial growth on the SiC substrate. The analysis and evaluation chamber 22C is a chamber for analyzing and evaluating the substrate processed in the processing chamber 22D. The epi analysis chamber 22b is an example of the analysis and evaluation chamber 22C, and is a chamber for evaluating, for example, the film thickness and impurity concentration of the epi film 25 in FIG. 1. The defect analysis chamber 22c is an example of the analysis and evaluation chamber 22C, and is a chamber for analyzing and evaluating defects in the epi film 25.

[0030] Further, the management system 1 may be configured to include a regeneration chamber 22E and a CMP chamber 22e. The regeneration chamber 22E is a chamber for regenerating the substrate processed in the processing chamber 22D. The CMP chamber 22e is an example of the regeneration chamber 22E and is a chamber for regenerating the substrate by CMP (chemical mechanical polishing). This enables reuse of the substrate (SiC substrate 24 in FIG. 1) by CMP or the like even when a desired epi result cannot be obtained as a result of the process (for example, when the epi quality of the process result with the generated recipe does not meet the allowable range described later).

[0031] The substrate analysis and evaluation apparatus includes, for example, a substrate shape measuring apparatus, and a defect and surface roughness evaluation apparatus. The substrate shape measuring apparatus measures the substrate shape such as the shape of the soris and wafer edge, and the plate thickness. The evaluation apparatus evaluates surface defects, internal defects, and surface roughness using, for example, X-rays, PL light, CL, radiation topography, a laser, or a microscope.

[0032] The epitaxial film formation analysis and evaluation apparatus includes, for example, an epi concentration evaluation apparatus, an epi film thickness evaluation apparatus, and a defect evaluation apparatus. The defect evaluation apparatus evaluates defects using, for example, X-rays, PL light, CL, radiation topography, a laser, or a microscope.

[0033] [Device state change] In the SiC epitaxial film formation apparatus, which is the epitaxial film formation apparatus 20 in FIG. 1, during film formation, as an example of device state change, substances are deposited as follows. During SiC epi growth, by-products derived from the material gas firmly adhere to members other than the SiC substrate 24 (for example, the inner wall of the chamber 22 and the susceptor). This by-product (in other words, deposit) is exposed to high temperatures during epi growth, evaporates, and becomes a factor causing temporal changes in epi quality. In other types of semiconductor manufacturing apparatuses, such as CVD apparatuses, it is relatively easy to remove by-products by gas cleaning or the like. However, in the SiC epitaxial film formation apparatus, currently, effective methods such as gas cleaning have not been established. Therefore, maintenance such as frequently opening the chamber is required to remove by-products, which increases costs.

[0034] In addition, in the SiC epitaxial film forming apparatus, as a result of performing maintenance on the chamber or the like, as described above, the state of the apparatus inherently changes. Even when the same recipe is applied before and after maintenance, the epitaxial quality of the film forming result may change and may not satisfy the allowable range. Therefore, the management system according to Embodiment 1 discovers an optimal recipe for the second and subsequent times based on the input information of the first-time process immediately after maintenance. In other words, the management system updates the model so that the quality of the epitaxial film in the processing result with the estimated recipe becomes an optimal recipe that is as close as possible to the target value within the allowable range.

[0035] [Management System - Function] The main functions and processing outlines of the management system 1 are as follows. The processing outline is also shown in the flow of FIG. 7 described later. The management system 1 inputs the first recipe immediately after the maintenance of the film forming apparatus 20 and the epitaxial quality (using the analysis evaluation result value as the quality), and compares the epitaxial quality estimated using the model of the prior art example or the model before maintenance (f(x) in FIG. 8 described later) with the latest epitaxial quality input above, and calculates a compensation coefficient (A in FIG. 8). The management system 1 updates the model to a new model (F(x) in FIG. 8) using the calculated compensation coefficient. This model is a machine learning model (described later in FIG. 8) for recipe generation and estimation. This model update is, in other words, an update for compensating for the change in the apparatus state before and after maintenance.

[0036] The management system 1 estimates an optimal recipe for the target SiC substrate 24 using the updated new model, and estimates the epitaxial quality of the processing result for each of the estimated candidate recipes. The management system 1 generates an optimal recipe such that the estimated epitaxial quality satisfies the allowable range. The management system causes film formation for the second and subsequent times immediately after maintenance to be performed using the generated recipe, and stores data such as the recipe and epitaxial quality during the implementation as history (in other words, processing history information) in a database (DB).

[0037] In addition, when the epi quality of the result estimated from the above optimal recipe does not meet the allowable range, the management system 1 of Embodiment 1 may be controlled as follows. That is, the management system 1 notifies the user U1 of the information on the optimal recipe calculated at that time through a graphical user interface (GUI), and conveys to the user U1 that the allowable range cannot be satisfied with that recipe. The user U1 checks this notification, determines whether to allow the film formation to be performed with that recipe, and makes an input according to the determination. Even when the allowable range cannot be satisfied with that recipe, if an input allowing the implementation is made, the film formation with that recipe is allowed. If an input not allowing the implementation is made, it is possible to take measures such as reviewing the set value of the allowable range. This function corresponds to the parts of steps S105 and S106 in the flow of FIG. 7 described later.

[0038] In addition, in each film formation, even when the epi quality of the film formation result in the recipe does not meet the allowable range, the management system ends the flow according to the notification to the user U1 on the GUI and the input of the determination by the user U1. This function corresponds to the parts of steps S111 and 112 in the flow of FIG. 7 described later.

[0039] [Management System - Functional Block Configuration] FIG. 6 shows an example of the functional block configuration of the management system 1 of Embodiment 1 in FIG. 1. The computer system 10 in FIG. 6 includes a central processing unit 104, a DB 105, an input device 15A, an output device 15B, a model configuration unit 107, a recipe estimation unit 108, a device control unit 110, a process processing unit 111, an analysis and evaluation unit 112, and a convergence determination unit 113.

[0040] As input information 101 through the input device 15A, it has a target value, an allowable range, an analysis and evaluation result, the first recipe after maintenance, epi quality, etc. The epi quality is an evaluation value such as concentration. The target value is the target value of the epi quality. The allowable range is a range determined according to specifications, etc., configured using the target value. The analysis and evaluation result is information that can be obtained from the analysis and evaluation device. The information on the first recipe and epi quality after maintenance is the recipe applied in the first film formation after that maintenance and the evaluation value of the epi quality of the film formation result.

[0041] The user U1 can operate the input device 15A (e.g., keyboard or mouse) to input each piece of information and make it part of the input information 101. Alternatively, the integrated management unit 11 of the management system 1 may automatically acquire and input the input information that can be automatically acquired.

[0042] As output information 102 through the output device 15B, it has an optimal recipe, an analysis and evaluation result, etc. The optimal recipe is set in the epitaxial film formation device 20 (also described as the film formation device 20) as the recipe to be applied after the second and subsequent times after maintenance. In other words, the recipe set in the film formation device 20 is updated to the new recipe.

[0043] The user U1 can check the content of the output information 102 on the display screen of the display device, which is one of the output devices 15B. The user U1 can give instructions, make settings, input the input information 101, etc. according to the GUI on the display screen.

[0044] The management system 1 includes, as each part realized based on program processing by the central processing unit 104, a model configuration unit 107, a recipe estimation unit 108, an apparatus control unit 110, a process processing unit 111, an analysis and evaluation unit 112, and a convergence determination unit 113. The central processing unit 104 is composed of a processor or the like in FIG. 4, and performs processing while appropriately using resources such as a memory and a communication interface. A control program or the like is stored in a ROM or a secondary storage device (not shown). The DB 105 can be composed of a memory, a secondary storage device, or a DB server or the like. The central processing unit 104 performs processing while reading and writing various data and information stored in a memory, the DB 105, or the like. Various data and information such as input information 101, output information 102, related information, and setting information are sorted and stored in the DB 105. Data and information being processed are stored in a memory such as a RAM.

[0045] The model configuration unit 107 configures and manages a machine learning model. The recipe estimation unit 108 estimates a recipe based on the model. These (107, 108) correspond to the recipe search unit 13 in FIG. 1. The apparatus control unit 110 controls the film forming apparatus 20. The process processing unit 111 performs processing related to the film forming process in the film forming apparatus 20. These (110, 111) correspond to the apparatus control unit 12 in FIG. 1. The analysis and evaluation unit 112 corresponds to the analysis and evaluation unit 14 in FIG. 1.

[0046] [Processing Flow (1)] FIG. 7 shows the main processing flow by the management system 1 (particularly the computer system 10, the central processing unit 104, and a processor or the like) of the first embodiment. This flow has steps S101 to S112. In step S101, the management system 1 inputs, as input information 101, information such as the first recipe immediately after maintenance and the epi quality (corresponding evaluation value) of the film forming result with that recipe.

[0047] In step S102, the management system 1 calculates, by the recipe estimation unit 108, the compensation coefficient of the recipe generation model (FIG. 8 described later) in the model configuration unit 107.

[0048] In step S103, the management system 1 inputs information on a desired target value and an allowable range as part of the input information 101. For example, the user U1 sets the target value and the allowable range through the input device 15A and the output device 15B.

[0049] In step S104, the management system 1 uses the model compensated using the compensation coefficient in step S102 to generate an optimal recipe for realizing an epi-quality value as close as possible to the target value within the allowable range, and estimates the epi-quality value of the result of epi film formation based on this recipe. This estimation of the epi quality can be realized, for example, as a simulation.

[0050] In step S105, the management system 1 determines and checks whether the epi-quality value that is the result of the above estimation satisfies the allowable range. If it is satisfied (Y), it proceeds to step S107, and if it is not satisfied (N), it proceeds to step S106.

[0051] In step S106, the management system 1 notifies the user U1 through the GUI of the output device 15B that the recipe generated in step S104 does not satisfy the allowable range (the one specified in step S103). Then, the management system 1 accepts an input based on the confirmation and determination of the user U1 as to whether to allow the film formation with that recipe. If an input allowing the film formation with that recipe is given (Y), it proceeds to step S107, and if an input not allowing it is given (N), it returns to step S103. In step S103, the setting of the allowable range is reviewed.

[0052] In step S107, the management system 1 causes the film formation apparatus 20 to execute an epitaxial film formation process according to the above optimal recipe by the process processing unit 111. Then, the management system 1 inputs the information on the recipe used at that time to the DB105 and saves it as part of the history.

[0053] In step S108, the management system 1 performs an analysis and evaluation of the epi quality of the film formation processing result in the above recipe by the analysis and evaluation unit 112, and creates analysis and evaluation result information (in other words, processing result information) including the epi quality.

[0054] In step S109, the management system 1 receives and acquires the above processing result information by the central processing unit 104. Alternatively, the user U1 may input the processing result information through the input device 15A. The central processing unit 104 associates the processing result information with the information of the recipe used for the processing and stores it in the DB105 as a part of the history.

[0055] In step S110, the management system 1 updates the model for recipe generation using the information of the updated DB105 in the model configuration unit 107.

[0056] In step S111, the management system 1 determines and checks whether the epi quality of the film formation result with the above optimal recipe satisfies the allowable range. If it is satisfied (Y), the flow ends. If it is not satisfied (N), it proceeds to step S112.

[0057] In step S112, the management system 1 notifies the user U1 through the GUI of the output device 15B that the allowable range was not satisfied with the recipe for which the film formation process was performed this time. Then, the management system 1 accepts an input based on the confirmation and determination of the user U1 as to whether to end with the film formation result. If an input to end with the film formation result is made (Y), the flow ends. If an input not to end is made (N), it returns to, for example, step S104. In step S104, the recipe is reviewed. The same processing is repeated thereafter.

[0058] Based on the above flow, in each time after the second time immediately after maintenance, epitaxial film formation is performed with an optimal or preferred recipe, and optimal or preferred epi quality can be stably obtained. Note that in the example of the above flow, the management system 1 can collectively perform estimation and proposal of the optimal recipe as described above in a plurality of times after the second time after maintenance. Not limited to this, the management system 1 may perform estimation and proposal of the optimal recipe as described above each time in each time after the second time after maintenance. The management system 1 may perform estimation and proposal of the above recipe in units of each set number of times.

[0059] [Model] FIG. 8 is an explanatory diagram of a method of updating a model using machine learning and the like in Embodiment 1. Hereinafter, a method of configuring a model for generating / estimating a recipe and the like will be described. (a) shows an outline of a model for a recipe. In a machine learning model, input information includes a recipe and epi quality for each film formation, and output information includes estimation results of an optimal recipe and epi quality. In Embodiment 1, in particular, the input of the model includes the recipe and epi quality of the first time immediately after maintenance, and the output of the model includes the estimated optimal recipe and epi quality to be applied after the second time and later immediately after maintenance. Optimal means that the estimated epi quality satisfies within an allowable range and is as close as possible to a target value.

[0060] (b) shows an outline of model update. Using the expression of a function, the conventional or previous model is represented by f(x), and the updated new model is represented by F(x). The new model F(x) is derived as F(x)=A*f(x)+W0 as shown in the figure. A is a compensation coefficient. W0 is the device state immediately after maintenance.

[0061] (c) shows a model update method using machine learning. In this method, considering the device state (W0) immediately after maintenance of the film forming apparatus 20, an optimal recipe is generated based on the model. In this method, as one of the epi qualities (corresponding evaluation values) of the film formation results until maintenance, the concentration distribution C tis used. Let the latent variable representing the device state be X t and the subscript t represents time.

[0062] The concentration distribution C, which is the output according to the device state, t is described as shown in the figure. Equation 1 is C t = g(X t , θ) + δ. θ is a value according to the recipe, substrate information (described later), etc. δ is the variation. Equation 2 is X t = f 1 (X t-1 , Z t-1 ) + ε. Z t-1 is a value according to the presence or absence of maintenance, the film thickness of the deposit, the elapsed time, etc. ε is the variation. g(X t , θ) and f 1 (X t-1 , Z t-1 ) are functions. The functional forms of these functions are unknown. Therefore, the management system 1 derives the functional forms of these functions by performing regression analysis.

[0063] Excluding the first time immediately after the maintenance of the film forming apparatus 20, the concentration distribution C at the second time and later after the maintenance is described by the following equation. Equation 3 is C = A * C t + C 0 . Equation 4 is A = f 2 (C 0 , C -1 ). C 0 is the measured value of the concentration distribution at the first time immediately after the maintenance. C -1 is the measured value of the concentration distribution before the maintenance. The compensation coefficient A can be derived from the measured values of the concentration distribution (C 0 , C -1 ) as described above. Therefore, the concentration distribution C immediately after the maintenance can be derived using the concentration distribution C t .

[0064] Regarding the details of the recipe generation or the machine learning method, known methods such as neural networks and support vector machines can be applied. For example, methods such as RNN (Recurrent Neural Network) and LSTM (Long-Short Term Memory) can be applied.

[0065] [Recipe adjustment] FIG. 9 is an explanatory diagram showing a specific example of fluctuations in epi quality and recipe adjustment according to the implementation of each epitaxial film formation and appropriate maintenance in the film formation apparatus 20 in the management system 1 of Embodiment 1. The graph in FIG. 9 shows the time change of the film formation conditions of the recipe (a) and the epi quality of (b) with the horizontal axis being the film formation time (corresponding time point or number of times). In other words, this content of FIG. 9 shows an example of adjusting to an optimal recipe immediately after maintenance according to the fluctuations in the apparatus state and epi quality associated with maintenance. The recipe of (a) shows one parameter value (for example, but not limited to, the gas flow rate) among the film formation conditions applied to the film formation process of SiC epitaxial growth. The epi quality of (b) is an evaluation value of the quality of the SiC epitaxial film (epi film 25 in FIG. 1) resulting from the film formation process of the recipe of (a), for example, the concentration distribution C in FIG. 8 t is the value of.

[0066] The allowable range 900 is a range having a lower limit value V2 and an upper limit value V3 centered on the target value V1 of the epi quality. Also, time points tm1, etc. show examples of the dates and times when the film formation apparatus 20 was maintained. Examples of maintenance include adjusting the state of the chamber 22 by removing deposits in the chamber 22 as described above.

[0067] For example, at time point t1 (the first time in cumulative terms), when film formation is performed with the recipe value set to p1, the epi quality value becomes v1, which satisfies the tolerance range of 900. At time point t2 (the second time in cumulative terms), when film formation is performed after adjusting the recipe value to p2, the epi quality value becomes v2, which is close to the target value V1 within the tolerance range of 900. During the period from time point t2 to time point t10 (the tenth time in cumulative terms), while appropriately fine-tuning the recipe value, film formation is performed each time, and as a result, each epi quality is maintained at a value close to the target value V1 within the tolerance range of 900, such as the value v2.

[0068] Here, after time point t10 (the tenth time in cumulative terms), at time point tm1, maintenance of the film forming apparatus 20 is being carried out. Immediately after this maintenance, time point t11 corresponds to the first time after maintenance (the eleventh time in cumulative terms), time point t12 corresponds to the second time after maintenance (the twelfth time in cumulative terms), and so on for each film formation.

[0069] In the first film formation after maintenance at time point t11, the recipe value p3 is the same as the recipe value at time point t10 immediately before maintenance. As a result of this film formation, the epi quality value has fluctuated to the value v3, as indicated by the black circle. This value v3 has fallen outside the tolerance range of 900 by falling below the lower limit value V2.

[0070] Therefore, based on the recipe value p3 and the epi quality value v3 at time point t11 (the first time after maintenance), the management system 1 performs the process of FIG. 7 described above, and at the next time point t12 (the second time after maintenance), proposes to change the recipe value p3 to the optimal value p4 estimated based on the new model and performs adjustment 901. As a result of this adjustment 901, the epi quality value of the film formation result at the recipe value p4 at time point t12 has become a value v2 close to the target value V1 within the tolerance range of 900.

[0071] Similarly, in each cycle after time point t13, the recipe is adjusted. For example, at time point t12, the recipes for each cycle after time point t13 are also proposed together, and they are adjusted so that the values increase gradually for each cycle. During the period from time point t12 to time point t20, as a result of film formation in each cycle, the value of the epi quality is maintained at a value v2 within the allowable range 900.

[0072] Also, after time point t20, at time point tm2, maintenance of the film forming apparatus 20 is performed. As a result, at the first time point t21 (the 21st time in total) immediately after the maintenance, the epi quality of the film forming result at the recipe value p5 is the value v4. The value v4 is out of the allowable range 900 so as to exceed the upper limit value V3. Therefore, the management system 1 performs an adjustment 902 in the same manner as the previous adjustment 901. In this adjustment 902, it is changed from the recipe value p5 at the first time point t21 immediately after the maintenance to the recipe value p6 at the second time point t22 immediately after the maintenance. After time point t22, along with the fine adjustment of the recipe, the epi quality of the film forming result in each cycle is maintained at the value v2. Thereafter, similarly, the recipe is adjusted according to the epi quality immediately after the maintenance is performed. In the example of FIG. 9, an example of adjustment for one parameter constituting the recipe and one evaluation value of the epi quality is shown, but the present invention is not limited to this, and it is similarly applicable to a plurality of other parameters.

[0073] [Effect etc. (1)] As described above, according to the management system of the epitaxial film forming apparatus of Embodiment 1, etc., stable quality can be obtained as the quality of the SiC epitaxial film. According to Embodiment 1, the labor of the work of creating and setting the recipe by the user can be reduced, and the semiconductor device manufacturing process can be made more efficient. According to Embodiment 1, as in the example of FIG. 9, according to the epi quality of the film forming result of the first cycle immediately after the maintenance, the recipe to be applied to the second and subsequent cycles is adjusted to be the optimal recipe. As a result, in each film formation of the second and subsequent cycles except for the first cycle after the maintenance, the epi quality can be quickly converged and maintained at a suitable value within the allowable range.

[0074] [Embodiment 2] Using FIGS. 10 to 11, a management system and the like of the epitaxial film deposition apparatus according to Embodiment 2 will be described. The basic configuration of Embodiment 2 and the like is the same as that of Embodiment 1. Hereinafter, the components different from those of Embodiment 1 in Embodiment 2 and the like will be mainly described. The management system of Embodiment 2 has a function of issuing a maintenance notice when the epitaxial quality in a recipe estimated for application to a film deposition process immediately after maintenance does not satisfy the allowable range.

[0075] [Management System (2)] FIG. 10 shows a functional block configuration of the computer system 10 in the management system 1 according to Embodiment 2. In addition to the components in Embodiment 1 shown in FIG. 6, this configuration includes a maintenance determination unit 109 and a maintenance effect evaluation unit 106 as processing units related to the maintenance notice function. The output information 102 includes, in addition to the aforementioned information, a maintenance notice and a maintenance evaluation result.

[0076] When the epitaxial quality in the recipe estimated by the recipe estimation unit 108 for application to the film deposition process immediately after maintenance does not satisfy the allowable range, the management system 1 inputs information indicating "maintenance required" from the maintenance determination unit 109 to the central processing unit 104. This is information indicating that it is necessary to perform maintenance on the film deposition apparatus 20 in order to satisfy the allowable range. Upon receiving this input, the central processing unit 104 proceeds to the maintenance notice step (FIG. 11 described later). The central processing unit 104 causes the recipe estimation unit 108 to start generating an optimal maintenance method necessary for generating a suitable recipe. The recipe estimation unit 108 generates an optimal maintenance method based on the model and returns the information to the central processing unit 104. The central processing unit 104 notifies the user U1 of the optimal maintenance method and the like through the GUI.

[0077] There are multiple types of maintenance methods for the film forming apparatus 20. For example, assume there are three types of methods: Method A, Method B, and Method C. The recipe estimation unit 108 selects the optimal maintenance method necessary for generating a suitable recipe (i.e., a recipe whose epi quality meets the allowable range) from such a plurality of candidate maintenance methods.

[0078] The generation of the maintenance method can be realized as follows, for example. The management system 1 manages to include maintenance history information including information such as the date and time and method of maintenance implementation in the processing history information of the DB105. When it is determined by the maintenance determination unit 109 that maintenance is required, based on the model including the maintenance history information, the management system 1 determines and derives which method among the plurality of candidate maintenance methods is the most optimal for maintenance, and detailed parameter values in the maintenance method, etc.

[0079] Also, in Embodiment 2, the effect of such maintenance when the maintenance of the film forming apparatus 20 is carried out is quantitatively evaluated. The maintenance effect evaluation unit 106 performs a quantitative evaluation of the effect of the maintenance and inputs the maintenance effect evaluation result to the central processing unit 104. Examples of the evaluation of the maintenance effect include measuring and evaluating, for example, the deposition amount or thickness of by-products or deposits on the wall in the chamber 22 of FIG. 1. For example, the sensor 27 in FIG. 1 measures a predetermined physical quantity such as the deposition amount or thickness at the deposition location 26. The management system 1 acquires this measurement value and uses it as an evaluation value of the maintenance effect. Examples of the sensor 27 may include a film thickness sensor or an optical film thickness evaluation device, etc. The result of this measurement may be input by the user U1 to the management system 1. The management system 1 may automatically acquire and input the signal of the sensor 27. The management system 1 may also monitor and record the signal of the sensor 27 over time.

[0080] Note that in Embodiment 2 and the like, similar to Embodiment 1, even when the allowable range cannot be satisfied with the optimal recipe, there is a function (step S106 in FIG. 7) etc. that allows the processing to be carried out at the discretion of the user U1.

[0081] [Processing Flow (2)] FIG. 11 shows the processing flow of the management system 1 in Embodiment 2. The flow in FIG. 11 has an additional part of step S200 related to the maintenance notification function compared to the flow in FIG. 7. In the aforementioned step S105, when the estimated epi quality by the estimated recipe does not satisfy the allowable range (N), the process transitions to step S200. Step S200 has steps S201 to S204.

[0082] In step S201, the management system 1 generates a maintenance method (in other words, maintenance content, etc.) necessary for generating a recipe to achieve an epi quality value close to the target value using the model in the model configuration unit 107 by the recipe estimation unit 108.

[0083] In step S202, the management system 1 notifies the user U1 through the GUI to perform maintenance using the above maintenance method, and the user U1 performs maintenance on the film forming apparatus 20. The maintenance effect evaluation unit 106 evaluates the maintenance effect of the maintenance implementation result as maintenance effect evaluation result information. Alternatively, the user U1 evaluates the maintenance effect and inputs the maintenance effect evaluation result information.

[0084] In step S203, the central processing unit 104 acquires the maintenance effect evaluation result information, associates the maintenance implementation content (including date, time, method, etc.) with the maintenance effect evaluation result, and stores it in the DB105 as part of the maintenance history information.

[0085] In step S204, the management system 1 updates the model using the information of the updated DB105 by the model configuration unit 107. After step S204, the flow ends.

[0086] [Effects, etc. (2)] As described above, according to Embodiment 2, when the allowable range cannot be satisfied by the recipe without maintenance, a suitable maintenance method is notified and implemented so as to obtain a suitable recipe that can satisfy the allowable range. As a result, in the first film formation immediately after the maintenance is carried out, good epi quality that satisfies the allowable range can be obtained as a result of the film formation using the suitable recipe. According to Embodiment 2, in response to the maintenance notification, the efficiency of the maintenance operation can also be improved.

[0087] The effects will be supplemented with reference to FIG. 9 described above. In the case of Embodiment 1, for example, it is assumed that time points tm1, tm2, etc. are regular maintenance. On the other hand, in the case of Embodiment 2, as an operation, such regular maintenance is not performed, or additional maintenance is appropriately performed even during regular maintenance. For example, after the time point t10, if the management system 1 assumes that there is no maintenance as it is, it determines that the epi quality does not satisfy the allowable range in the film formation at the next time point t11. Therefore, the management system 1 generates a suitable recipe and maintenance method in consideration of the above-described maintenance implementation and its implementation result, and notifies the user U1. As a result, maintenance is performed, for example, between the time points t10 and t11. As a result, in the first film formation at the time point t11 immediately after the maintenance, the epi quality can be made within the allowable range.

[0088] <Embodiment 3> The management system of the epitaxial film formation apparatus according to Embodiment 3 will be described with reference to FIG. 12. Embodiment 3 is a form in which functions are added to Embodiment 2. The management system of Embodiment 3 inputs information (referred to as substrate information) of a substrate (SiC substrate 24 in FIG. 1) that is the target of the film formation process, and performs optimal recipe generation and the like in consideration of the substrate information. After installing a substrate (for example, a semiconductor wafer) on the stage 23 in the chamber 22, the management system 1 first evaluates the substrate to obtain substrate evaluation result information.

[0089] [Substrate] Generally, compared with Si substrates, SiC substrates have more defects such as dislocations. For example, the dislocation density affects the defect density after epitaxial film formation. Therefore, when using the defect density as the output (the output of the aforementioned model, the evaluation value of epitaxial quality), it is necessary to consider the defect information of the substrate. Also, warpage of the substrate etc. also affects epitaxial quality. Thus, in Embodiment 3, by taking into account the substrate information including such information, an optimal recipe is generated with high accuracy. In Embodiment 3, substrate information is included in the input information of the training data of machine learning. The management system 1 generates an optimal recipe so that the epitaxial quality of the film formation result using the target substrate is as close as possible to the target value within the allowable range.

[0090] When generating that recipe, there is a need for a constraint that one of the inputs, the substrate information, satisfies a pre-specified substrate specification (in other words, substrate quality). For example, if the target substrate has many defects, there is a possibility that the estimated optimal recipe may not satisfy the allowable range. Therefore, the management system 1 calculates and determines to what extent of the number of defects (in other words, defect density) on the target substrate an optimal recipe can be generated. The management system 1, for example, gradually reduces the number of defects and calculates whether a recipe that satisfies the allowable range can be generated under the conditions of each number of defects. Then, the management system 1 notifies, through the GUI, the conditions of the number of defects and defect density (in other words, the conditions of the substrate specification) for which the recipe can be generated. The management system 1 notifies whether the installed target substrate satisfies or does not satisfy the conditions. After that, when the target substrate does not satisfy the conditions, the management system 1 asks the user U1 through the GUI to confirm whether to change the target substrate to a substrate that satisfies the conditions and requests a judgment. The user U1 makes a judgment and inputs, accordingly, whether to change the substrate or not change the substrate. Also, when it is determined not to change the substrate, the management system 1 transitions to the maintenance execution flow.

[0091] [Management System (3)] The functional block configuration of the management system 1 in Embodiment 3 is the same as that of Embodiment 2 in FIG. 10. The difference is that it includes substrate information as one of the input information 101, and includes information related to substrate specifications such as the above defect density and substrate change notifications as one of the output information 102.

[0092] [Processing Flow (3)] FIG. 12 shows the processing flow of the management system 1 in Embodiment 3. In the flow of FIG. 12, steps related to substrate information processing are added to the flows of FIGS. 7 and 11. Between step S103 and step S104 in FIG. 7, there is step S301. In step S301, the management system 1 inputs and acquires substrate information regarding the target substrate (SiC substrate 24) on stage 23 and registers it in DB105. At this time, the user U1 may input substrate information through the GUI, or the management system 1 may acquire substrate information from another device.

[0093] The content of the process in step S104 is partially different from the above (referred to as S104c). In step S104c, the management system 1 uses the above model and the substrate information to generate an optimal recipe and estimates the corresponding epi quality.

[0094] Also, the content of the process in step S107 is partially different from the above (referred to as S107c). In step S107c, the management system 1 associates the recipe and the substrate information and inputs them into BD105.

[0095] In Embodiment 3, when the allowable range is not satisfied in step S105 (N), the process proceeds to step S302. In step S302, the management system 1 uses the model of the model configuration unit 107 by the recipe estimation unit 108 to derive the substrate specifications required for generating a recipe that realizes an epi quality value close to the target value within the allowable range, for example, conditions such as the above defect density.

[0096] Next, in step S303, the management system 1 determines whether the target substrate on the stage 23 meets the substrate specification conditions based on the substrate specification conditions and the substrate information. If it does not meet the conditions, the management system 1 determines whether to change the target substrate on the stage 23 to another substrate that meets the substrate specification conditions. At this time, the management system 1 notifies the user U1 through the GUI of information such as the defect density of the target substrate and the substrate specification conditions, and accepts a determination / input on whether to change to another substrate when the target substrate does not meet the conditions. The user U1 checks the notification and makes a determination / input on whether to change the substrate. If the substrate is to be changed (Y), the process returns to step S301. In that case, in step S301, the substrate information regarding the changed substrate is input. If the substrate is not changed (N), the process transitions to step S200 related to the aforementioned maintenance notification, and the same processing is performed.

[0097] [Substrate and Model] In Embodiment 3, as described above, a plurality of substrates are treated as the target substrate, and a suitable recipe is generated in response to the differences between individual substrates. In Embodiment 3, the aforementioned model for recipe generation is a model assuming one standard SiC substrate as the target substrate. The management system 1 can generate a suitable recipe according to the substrate specification conditions of each individual substrate based on this model and the substrate information of each individual substrate.

[0098] [Effect, etc. (3)] As described above, according to Embodiment 3, in addition to the same effects as in Embodiments 1 and 2, as a result of film formation with a suitable recipe according to the characteristics of each individual substrate applied in each film formation, stable epi quality can be obtained. For example, in the example of FIG. 9, each film formation at each time point is a substrate with individual characteristics as the target substrate, and suitable epi quality can be obtained each time, including immediately after maintenance.

[0099] <Embodiment 4> Using FIG. 13, the management system of the epitaxial film deposition apparatus according to Embodiment 4 and the like will be described. Embodiment 4 is a form in which functions are added to Embodiment 3. The management system of Embodiment 4 further has a function of generating an optimal recipe and the like in consideration of the characteristics of individual film deposition apparatuses and chambers in a plurality of film deposition apparatuses 20 and a plurality of chambers 22. In Embodiment 4, as candidates for processing a substrate, there are a plurality of epitaxial film deposition apparatuses and a plurality of chambers in the epitaxial apparatus. The processor inputs, as one of the input information of the model, information on the epitaxial film deposition apparatus and the chamber that are the targets for processing the substrate among the candidates, and based on the model, generates a recipe for the processing in the chamber of the target epitaxial film deposition apparatus.

[0100] The configuration of the management system 1 according to Embodiment 4 is the same as that of FIG. 10 described above. The difference is that, as one of the input information 101, it has a device number and a chamber number. The processing flow in Embodiment 4 is different from that in FIG. 12, for example, in step S301, the device number and the chamber number are further input and registered in the DB105 in association with other information.

[0101] The device number is an ID for identifying an individual film deposition apparatus 20, and the chamber number is an ID for identifying an individual chamber 22. The user U1 inputs, through the GUI, the device number of the film deposition apparatus 20 to be used for film deposition and the chamber number of the chamber 22 to be used in the film deposition apparatus 20 (for example, the epitaxial growth chamber 22d in FIG. 5(B)). Alternatively, the management system 1 may automatically grasp or acquire the target device number and chamber number from other devices or the like. If the device number and the chamber number are not input or specified, the management system 1 uses the default setting values of the device number and the chamber number. This default setting value can also be set by the user U1 through the GUI.

[0102] The management system 1 performs processes such as recipe generation in step S104 using a model corresponding to the film forming apparatus 20 and the chamber 22 identified by the input device number and chamber number. In step S107, the management system 1 associates information such as substrate information, device number, chamber number, model, and recipe, and stores the information in the DB105.

[0103] [Film Forming Apparatus and Chamber] When different film forming apparatuses 20 and chambers 22 are used for film formation, even if the same recipe is applied, there may be differences in the epi quality of the respective film formation results. Therefore, the management system 1 of Embodiment 4 grasps the individual film forming apparatuses 20 and their chambers 22 used for film formation based on the device number and chamber number, and generates a suitable recipe corresponding to the characteristics of the film forming apparatuses 20 and the chambers 22. Specifically, in the machine learning model, the characteristics of the individual film forming apparatuses 20 and chambers 22 are reflected. Also, a model for each individual film forming apparatus 20 and chamber 22 may be configured and used, or a technique such as transfer learning may be used. In transfer learning, one model assumed with a certain film forming apparatus 20 and chamber 22 fixed is configured, and based on that model, for other film forming apparatuses 20 and chambers 22, a corresponding model is configured by transfer learning. Thereby, based on a small amount of fixed information, a high-precision model corresponding to each film forming apparatus 20 and chamber 22 can be configured.

[0104] FIG. 13 shows, as a supplement, an example of the relationship between a plurality of film forming apparatuses 20, a plurality of chambers 22, and models to be applied in Embodiment 4. The table in (a) shows, as a first method, an example of management information when individual models are configured for each individual film forming apparatus 20 and chamber 22. The table in (a) has, as columns, an apparatus number, a chamber number, and a model. For example, in the chamber with chamber number = 11 in the film forming apparatus 20 with apparatus number = 1, model M111 is applied, and in another chamber with chamber number = 12, another model M112 is applied. In another film forming apparatus 20 with apparatus number = 2, another model (M221, M222) is applied. The table in (b) shows, as a second method, an example of management information when a common one model is configured for a plurality of film forming apparatuses 20 and a plurality of chambers 22. For example, in two film forming apparatuses 20 with apparatus numbers = 1 and 2, the same model M100 is used. In one film forming apparatus 20 with apparatus number = 3, the same model M300 is used.

[0105] [Effects, etc. (4)] As described above, according to Embodiment 4, in addition to the same effects as in Embodiment 3, as a result of film formation with a suitable recipe according to the characteristics of each individual film forming apparatus 20 and chamber 22 applied in each film formation, stable epi quality can be obtained. For example, the state including the influence of maintenance can be different for each chamber 22 of the film forming apparatus 20. Even in that case, an optimal recipe can be generated based on a model in which the characteristics of each individual chamber 22 are considered, and the epi quality of the film formation immediately after maintenance can be stabilized.

[0106] [Modification Examples of Embodiments 1 to 4] As a modification of Embodiments 1 to 4, the following is also possible. The configuration of the management system 1 of the modification is the same as that of FIG. 10, for example, and the difference is that, as one of the input information 101, it has hyperparameters. These hyperparameters refer to parameters set for known machine learning algorithms, in other words, setting information for machine learning algorithms / models. The processing flow in this modification is different from the flow of FIG. 12 in Embodiment 3, for example, in that in step S301, hyperparameters are input as one of the input information. The user U1 inputs and sets hyperparameters through the GUI.

[0107] In machine learning, when using a huge amount of data, the amount of calculation, time, and load increase when searching for the optimal recipe. Although it is possible to shorten the time required for the search by devising such as increasing the search calculation step (such as the width of changing the value), the accuracy decreases. Therefore, in Embodiment 4, by inputting and setting appropriate hyperparameters for the model for recipe generation, both high accuracy and short-time calculation are achieved. For example, the management system 1 sets, as one of the hyperparameters, the start value of the recipe search to a value of a recipe close to the optimal solution. The optimal solution is, for example, a recipe (corresponding film formation conditions) such that the epi quality becomes the target value V1 in FIG. 9. According to Embodiment 4, therefore, even when the search calculation step of the recipe search is small, the time required for the search may be shortened. That is, it is possible to improve the efficiency of processing and work.

[0108] [GUI] In Embodiments 1 to 4, or their modifications, the following example of the GUI can be applied. In this example, a display example of a screen including the GUI for the user U1 in the modification having the function of the above hyperparameters is shown.

[0109] Figures 14 and 15 show GUI examples. In Figure 14, column g1 is a recipe search function column where the enable / disable state for the functions described in Embodiment 1 and the like can be set. Column g2 is a maintenance notification message column where the aforementioned maintenance notification messages are displayed. Column g3 is a device / chamber information input column where the aforementioned device number and chamber number can be input and confirmed. Column g4 is a substrate information input column where the aforementioned substrate information can be input and confirmed. For example, values such as the substrate ID and thickness of the substrate for each slot can be input.

[0110] Column g5 is a target value / tolerance range input column where the aforementioned target value, upper limit value, and lower limit value can be input and confirmed for each parameter representing epi quality. Column g6 is a hyperparameter input column where the values of the aforementioned hyperparameters can be input and confirmed. Column g7 is a data set input column immediately after maintenance where the recipe and epi quality can be input and confirmed as the data set for the first film deposition immediately after the aforementioned maintenance. Column g8 is a maintenance evaluation result input column where the evaluation results of the maintenance effect (such as the maintenance method of the maintenance content and the evaluation values of each maintenance item) when the aforementioned maintenance is performed (for example, at time point tm1 in Figure 9) can be input and confirmed. Each column in Figure 14 mainly corresponds to an input column for setting information.

[0111] Next, in FIG. 15, each column mainly corresponds to an output column for results. Column g9 is an optimal recipe output column, and for example, the content of the generated optimal recipe to be applied after the second and subsequent times immediately after maintenance in Embodiment 1, and the estimated result of the epi quality in that optimal recipe are displayed. Column g10 is an epi quality output column immediately after maintenance, and for example, the evaluation value (measured value and description) of the epi quality in the film formation result after the second and subsequent times immediately after maintenance in Embodiment 1, and the corresponding estimated value (the value estimated by the management system 1 corresponding to the optimal recipe in column g9) are displayed. Column g11 is a history information output column, and the history information stored in DB105 is displayed. Column g11 includes, for example, a process information column, a recipe content column, and an epi quality evaluation result column. The process information column has a process number, a process date and time, a substrate ID, etc. The process information column may also have other information such as a device number and a chamber number.

[0112] The recipe content column has each parameter value of the recipe applied to the process. The evaluation result column has each evaluation value of the epi quality. Column g12 is a maintenance evaluation result output column, and for example, the evaluation result of the maintenance effect when maintenance is performed with "maintenance required" in the above-described Embodiment 2 is displayed.

[0113] The management system 1 holds data such as a table corresponding to each column of the above GUI example in DB105 or memory. Based on the data, the management system 1 generates and displays screen data (for example, it may be a web page) for display on the display screen of the output device 15B.

[0114] As another GUI, the management system 1 may display a graph or the like of the result of detecting and monitoring the change over time of the physical quantity of the deposit by the sensor 27 in FIG. 1 in association with other information.

[0115] [Appendix] As described above, the present invention has been specifically described based on the embodiments. However, the present invention is not limited to the above-described embodiments and can be variously modified without departing from the gist. Each component may be singular or plural unless otherwise particularly limited. A form by combining the embodiments is also possible. Except for the essential elements, addition, deletion, replacement, etc. of the components of the embodiments are possible. In the embodiments, regarding the generation and adjustment of the recipe, the case of applying it to SiC epitaxial film formation has been described. However, it can be similarly applied to other processes and apparatuses, for example, lithography (exposure, electron beam lithography, X-ray lithography, etc.), other film formation (CVD, PVD, evaporation, sputtering, thermal oxidation, etc.), pattern processing (etching, electron beam, laser, etc.), ion implantation (plasma, etc.), cleaning (liquid, ultrasonic, etc.), and the like.

Explanation of Signs

[0116] 1…Management system, 10…Computer system, 20…Epitaxial film formation apparatus, 11…Integrated management unit, 12…Apparatus control unit, 13…Recipe search unit, 14…Analysis and evaluation unit, 15A…Input device, 15B…Output device, U1…User, 21…Control unit, 22…Chamber, 23…Stage, 24…SiC substrate, 25…Epitaxial film, 26…Deposition location, 27…Sensor.

Claims

1. A management system for generating a processing recipe of an epitaxial film deposition apparatus, comprising: a processor, wherein the processing of the epitaxial film deposition apparatus includes a process of forming an epitaxial film by using epitaxial growth on a substrate, and the processor inputs information on the recipe of the first processing immediately after maintenance of the epitaxial film deposition apparatus and the evaluation value of the quality of the epitaxial film as the processing result, and constructs or updates the model of the processing as initial conditions; based on the model, generates a recipe for the second and subsequent processing immediately after maintenance, such that the evaluation value of the quality is within an allowable range including the target value; the recipe includes a gas flow rate as a parameter value; the evaluation value of the quality includes a concentration distribution; the maintenance includes maintenance related to deposits in a chamber where the film forming process is performed; the processor sets the recipe of the first processing immediately after maintenance to be the same as the value of the recipe of the processing immediately before maintenance, executes the first processing immediately after the maintenance, obtains the evaluation value of the quality of the processing result immediately after the maintenance, updates the model with this information as the initial conditions, and based on the updated model, generates a recipe for the second and subsequent processing immediately after maintenance, such that the evaluation value of the quality is within the allowable range including the target value; the model is a machine learning model, and is a model that handles fluctuations in the state of the epitaxial film deposition apparatus before and after maintenance using a compensation coefficient; the compensation coefficient is calculated using the evaluation value of the quality of the processing result immediately before maintenance and the evaluation value of the quality of the first processing result immediately after maintenance; A management system for an epitaxial film deposition apparatus.

2. In the management system for an epitaxial film deposition apparatus according to Claim 1, if there is no recipe in which the estimated evaluation value of the quality satisfies the allowable range in the recipe generated based on the model, the processor determines that maintenance is necessary and outputs a message indicating that maintenance is necessary. A management system for an epitaxial film deposition apparatus.

3. In the management system for an epitaxial film deposition apparatus according to Claim 2, When the processor determines that maintenance is necessary, it uses the model to generate a maintenance method required for generating a recipe that satisfies the allowable range, outputs the maintenance method together with the fact that maintenance is necessary, and updates the model using an evaluation value of the maintenance effect as a result of performing the maintenance according to the maintenance method. A management system for an epitaxial film deposition apparatus.

4. In the management system for an epitaxial film deposition apparatus according to claim 3, The processor uses, as the evaluation value of the maintenance effect, a detected value of the physical quantity of by-products in the chamber of the epitaxial film deposition apparatus immediately after the maintenance. A management system for an epitaxial film deposition apparatus.

5. In the management system for an epitaxial film deposition apparatus according to claim 1, The processor inputs substrate information including an evaluation value of the quality of the substrate as one of the input information of the model, generates conditions for the quality of the substrate required for generating a recipe that satisfies the allowable range when there is no recipe that satisfies the allowable range based on the model and the substrate information, outputs whether to change to a substrate that satisfies the conditions, and when it is input that the substrate is not to be changed, determines that maintenance is necessary and outputs the fact that maintenance is necessary. A management system for an epitaxial film deposition apparatus.

6. In the management system for an epitaxial film deposition apparatus according to claim 1, As candidates for performing the processing on the substrate, it has a plurality of epitaxial film deposition apparatuses and a plurality of chambers in the epitaxial film deposition apparatus, The processor inputs, as one of the input information of the model, information on the epitaxial film deposition apparatus and the chamber that are to be the targets for performing the processing on the substrate among the candidates, and generates a recipe for the processing in the chamber of the target epitaxial film deposition apparatus based on the model. A management system for an epitaxial film deposition apparatus.

7. In the management system for an epitaxial film deposition apparatus according to claim 1, If the evaluation value of the quality in the recipe estimated as the recipe for the second and subsequent processes immediately after the maintenance does not satisfy the allowable range, the processor outputs whether to perform the process although the allowable range is not satisfied in the recipe, and if it is input that the process is to be performed, causes the process in the recipe to be performed. A management system for an epitaxial film forming apparatus.

8. In the management system for an epitaxial film forming apparatus according to claim 1, the epitaxial film forming apparatus includes an analysis chamber for analyzing and evaluating the quality of the epitaxial film based on measured values, the management system acquires the analysis and evaluation results in the analysis chamber. A management system for an epitaxial film forming apparatus.

9. A management method in a management system for generating a recipe for processing an epitaxial film forming apparatus, the management system includes a processor, the processing of the epitaxial film forming apparatus includes a process of forming an epitaxial film by using epitaxial growth on a substrate, the processor inputs information on the recipe for the first process immediately after the maintenance of the epitaxial film forming apparatus and the evaluation value of the quality of the processing result of the epitaxial film as initial conditions, and constructs or updates a model of the process; generates, as a recipe for the second and subsequent processes immediately after the maintenance, a recipe in which the evaluation value of the quality is within an allowable range including a target value, based on the model; and has the recipe includes a gas flow rate as a parameter value, the evaluation value of the quality includes a concentration distribution, the maintenance includes maintenance related to deposits in the chamber where the film forming process is performed, the processor sets the recipe for the first process immediately after the maintenance to be the same as the value of the recipe for the process immediately before the maintenance, causes the first process immediately after the maintenance to be executed, acquires the evaluation value of the quality of the processing result of the first process immediately after the maintenance, updates the model with this information as the initial conditions, and based on the updated model, generates a recipe in which the evaluation value of the quality is within the allowable range including the target value, as a recipe for the second and subsequent processes immediately after the maintenance; the model is a machine learning model and is a model that handles fluctuations in the state of the epitaxial film forming apparatus before and after maintenance using a compensation coefficient. The compensation coefficient is calculated using an evaluation value of the quality of the processing result immediately before the maintenance and an evaluation value of the quality of the processing result of the first time immediately after the maintenance. A method for managing an epitaxial film forming apparatus.

Images