Data processing device, data processing method, and program

Abstract

The present invention realizes highly versatile data processing for data collected in a manufacturing process. The data processing device comprises: a calculation unit, which collects data groups corresponding to prescribed steps in a process and calculates the effect of the prescribed steps for each data group; a segmentation unit, which segments a feature space so that the distribution of each of the plurality of data groups corresponding to the prescribed steps in the feature space is classified according to each of the calculated effects; and an output unit, which outputs specific data, which specifies each area of ​​the segmented feature space.

Description

Technical Field

[0001] This invention relates to a data processing apparatus, a data processing method, and a program. Background Technology

[0002] Previously, it was known that data processing devices collected or measured data used in various manufacturing processes (e.g., semiconductor manufacturing processes) and performed various analyses. By using such data processing devices to analyze the collected data, it became possible to model the manufacturing process and execute simulation processing, thereby achieving optimization of the manufacturing process and improvement of product quality.

[0003] Patent Document 1: U.S. Patent Application Publication No. 2017 / 0177997

[0004] Patent Document 2: U.S. Patent Application Publication No. 2015 / 0211122

[0005] Patent Document 3: Japanese Patent Application Publication No. 2009-152269

[0006] On the other hand, modeling the manufacturing process is time-consuming and costly. Furthermore, improving simulation accuracy requires building a separate model for each manufacturing device; even for the same manufacturing process, different devices require different models. Therefore, conventional data processing devices for processing data collected during the manufacturing process are both time-consuming and costly, and lack versatility. Summary of the Invention

[0007] One aspect aims to achieve highly versatile data processing of data collected during the manufacturing process.

[0008] According to one form, the data processing device has:

[0009] The calculation unit collects data clusters corresponding to the specified steps within the process and calculates the effects of each specified step based on each data cluster.

[0010] The segmentation unit segments the feature space to establish the distribution of multiple data groups within that feature space according to each calculated effect classification and the aforementioned steps; and

[0011] The output unit outputs specific data, which specifically represents the regions of the aforementioned feature space after segmentation.

[0012] It can perform highly versatile data processing on data collected during the manufacturing process. Attached Figure Description

[0013] Figure 1This is a diagram illustrating an example of the overall structure of a data processing system.

[0014] Figure 2 This is a diagram showing specific examples of data groups processed in each unit.

[0015] Figure 3 It is a diagram used to provide a summary of the parsing result data stored in the parsing result storage section.

[0016] Figure 4 This is a diagram illustrating an example of the hardware configuration of a data processing device.

[0017] Figure 5 This is a diagram illustrating an example of the functional structure of the data parsing unit.

[0018] Figure 6 This is a diagram illustrating a specific example of the processing by the effect calculation unit.

[0019] Figure 7 This is a diagram showing an example of a data set stored in a data storage unit.

[0020] Figure 8 This is a diagram illustrating a specific example of the processing in the classification department.

[0021] Figure 9 This is a diagram showing an example of a Proxel calculated by the Proxel computation unit.

[0022] Figure 10 This is a first flowchart illustrating the Proxel computation process based on the classification department and the Proxel computation department.

[0023] Figure 11 This is the first diagram used to illustrate the advantages of computing Proxel.

[0024] Figure 12 This is the second figure used to illustrate the advantages of calculating Proxel.

[0025] Figure 13 This is Figure 3, used to illustrate the advantages of calculating Proxel.

[0026] Figure 14 This is a second flowchart illustrating the Proxel computation process based on the classification department and the Proxel computation department. Detailed Implementation

[0027] Hereinafter, various embodiments will be described with reference to the accompanying drawings. Furthermore, in this specification and the drawings, constituent elements having substantially the same functional configuration are labeled with the same reference numerals, thereby omitting repeated descriptions.

[0028] [First Implementation Method]

[0029] <Overall Structure of a Data Processing System>

[0030] First, the overall structure of the data processing system will be explained. Figure 1 This is a diagram illustrating an example of the overall structure of a data processing system. (See diagram for example.) Figure 1 As shown, the data processing system 100 includes: a data processing device 110; and terminals 121, 131, and 141 in each unit 120, 130, and 140 (unit name = "Unit A", "Unit B", and "Unit C"). The data processing device 110 and the terminals 121, 131, and 141 in each unit 120, 130, and 140 are connected via a network 150 to enable communication.

[0031] A data parsing program is installed in the data processing device 110. By executing the data parsing program, the data processing device 110 functions as a data parsing unit 111.

[0032] Data parsing unit 111 collects data clusters via network 150 through terminals 121, 131, and 141 in each unit 120, 130, and 140. Figure 1 In the example, the data includes initial data, setting data, output data, measurement data, experimental data, and target data. Furthermore, the data parsing unit 111 stores the collected data clusters in the data storage unit 112. However, the method for collecting the data clusters is not limited to this; for example, the administrator of the data processing device 110 can obtain recording media containing the data clusters from each of the units 120, 130, and 140, and collect the data clusters by reading them from the recording media.

[0033] In addition, the data parsing unit 111 parses the data group stored in the data storage unit 112 and stores the parsing result data in the parsing result storage unit 113.

[0034] Unit 120 (unit name = "Unit A") includes a semiconductor manufacturing apparatus that performs a semiconductor manufacturing process. Additionally, unit 120 includes a measuring device for measuring measurement data during the semiconductor manufacturing process, and an experimental value measuring device for measuring experimental data of the resulting product (semiconductor or intermediate product) manufactured in the semiconductor manufacturing process. Furthermore, unit 120 includes a terminal 121 constituting the data processing system 100 and a database for storing data.

[0035] The semiconductor manufacturing apparatus executes a semiconductor manufacturing process based on initial data, setting data, and target data input through terminal 121. Furthermore, the semiconductor manufacturing apparatus establishes a correspondence between the output data obtained from executing the semiconductor manufacturing process and the initial data, setting data, and target data, and stores this correspondence in a database.

[0036] The measuring instrument measures and stores measurement data in a database during the execution of a semiconductor manufacturing process based on a semiconductor manufacturing apparatus. The experimental value measuring instrument measures experimental data on the resulting product (semiconductor or intermediate product) manufactured in the semiconductor manufacturing process and stores it in a database.

[0037] Terminal 121 inputs initial data, setting data, and target data used in the semiconductor manufacturing process to configure the semiconductor manufacturing apparatus. In addition, terminal 121 sends a data set (initial data, setting data, output data, measurement data, experimental data, and target data collected in the semiconductor manufacturing process) stored in the database to data processing device 110.

[0038] In units 130 (unit name = "unit B") and 140 (unit name = "unit C"), the same semiconductor manufacturing process as in unit 120 is performed. Therefore, units 130 and 140 contain the same apparatus as in unit 120. However, in Figure 1 In the example, unit 130 does not include the experimental value measuring device. Additionally, unit 140 does not include either the measuring device or the experimental value measuring device.

[0039] Thus, given the different devices contained in each unit, the information items in the data groups sent from each terminal 121, 131, 141 within each unit 120, 130, 140 to the data processing device 110 will also differ. For example, the data group sent from terminal 131 within unit 130 may not contain experimental data (or a portion thereof). Similarly, the data group sent from terminal 141 within unit 140 may not contain measurement data or experimental data (or a portion thereof).

[0040] <Specific Examples of Data Clusters>

[0041] Next, the data groups processed in units 120, 130, and 140 will be explained. Figure 2 This is a diagram illustrating specific examples of data groups processed in each unit. Here, the data group processed in unit 120 is explained.

[0042] like Figure 2As shown, the semiconductor manufacturing apparatus of unit 120 executes multiple semiconductor manufacturing processes (process names = processes I to M), and each semiconductor manufacturing process includes multiple steps (step names = steps 1 to N). Here, a "step" is defined as the smallest processing unit in the semiconductor manufacturing process that causes a change in state (attributes of the object being processed, the state of the semiconductor manufacturing apparatus, the environment within the semiconductor manufacturing apparatus, etc.). Therefore, in this embodiment, where the state changes over time, steps before and after the passage of time are defined as independent steps.

[0043] exist Figure 2 In this context, data group 201 is a data group corresponding to the following processes and steps, namely:

[0044] • Among the multiple semiconductor manufacturing processes performed by the 120-unit semiconductor manufacturing apparatus, the semiconductor manufacturing process with process name = "Process I" is the semiconductor manufacturing process.

[0045] • A step whose step name is “STEP1” among the multiple steps in the semiconductor manufacturing process.

[0046] like Figure 2 As shown, data group 201 includes items as information such as "initial data (I)", "setup data (R)", "output data (E)", "measurement data (Pl)", "experimental data (Pr)" and "target data (Pf)".

[0047] The "Initial Data (I)" section contains initial data input via terminal 121 within unit 120. In the case of a semiconductor manufacturing process, the initial data includes:

[0048] • Initial CD (critical dimensions)

[0049] Material

[0050] • Thickness

[0051] Aspect ratio

[0052] • Mask coverage, etc.

[0053] The "Setting Data (R)" contains setting data for the semiconductor manufacturing apparatus input via terminal 121 within unit 120. The setting data for the semiconductor manufacturing apparatus is data specific to the characteristics of the semiconductor manufacturing apparatus. In the case of a semiconductor manufacturing process, the setting data includes:

[0054] • Pressure (intracavitary pressure)

[0055] • Power (electricity from a high-frequency power supply)

[0056] • Gas (gas flow rate)

[0057] • Temperature (temperature inside the cavity or temperature on the substrate surface), etc.

[0058] The "Output Data (E)" includes the output data from the semiconductor manufacturing apparatus of unit 120 during the execution of a semiconductor manufacturing process with process name = "Process I" and step name = "STEP1" based on the semiconductor manufacturing apparatus of unit 120. The output data output from the semiconductor manufacturing apparatus is data that depends on the characteristics of the semiconductor manufacturing apparatus. In the case of a semiconductor manufacturing process, the output data includes:

[0059] • Vpp (potential difference)

[0060] • Vdc (DC self-bias)

[0061] •OES (Observation Spectrometry-Based Emission Intensity)

[0062] • Reflect (electrical power of reflected waves)

[0063] • Top DCS current (based on Doppler current meter readings), etc.

[0064] The "Measurement Data (Pl)" includes measurement data measured by the measuring instrument of unit 120 during the execution of a semiconductor manufacturing process with process name = "Process I" and step name = "STEP1" based on the semiconductor manufacturing apparatus of unit 120. The measurement data measured by the measuring instrument is data independent of the characteristics of the semiconductor manufacturing apparatus. In the case of a semiconductor manufacturing process, the measurement data includes:

[0065] Plasma density

[0066] • Ion energy

[0067] • Ion flux, etc.

[0068] The "Experimental Data (Pr)" section contains experimental data obtained by measuring the results generated from a semiconductor manufacturing process with process name "Process I" and step name "STEP1" executed by a semiconductor manufacturing apparatus of unit 120 using an experimental value measuring device. The experimental data measured by the experimental value measuring device is independent of the characteristics of the semiconductor manufacturing apparatus. In the case of a semiconductor manufacturing process, the experimental data includes:

[0069] • Etching rate

[0070] • Deposition rate

[0071] • XY position (XY coordinates)

[0072] • Film type

[0073] ·Vertical / Lateral (vertical / horizontal distinction), etc.

[0074] The "Target Data (Pf)" contains the target data input via terminal 121 within unit 120. Target data refers to the attributes that the resulting object should achieve when the semiconductor manufacturing apparatus of unit 120 executes the entire semiconductor manufacturing process with process name "Process I". In the case of a semiconductor manufacturing process, the target data includes:

[0075] •CD (critical dimension)

[0076] •Depth

[0077] Taper (cone angle)

[0078] • Tilting (angle of inclination)

[0079] •Bowing (warpage), etc.

[0080] also, Figure 2 The data set shown is an example, and the types of data included in each information item are not limited to those shown in the diagram. Furthermore, it is assumed that the data set contains different information items and different types of data for each unit, each process, and each step.

[0081] <Summary of the parsed data>

[0082] Next, a summary of the parsing result data stored in the parsing result storage unit 113 after the data parsing unit 111 of the data processing device 110 parses the data groups collected from each unit 120, 130, and 140 will be described. Figure 3It is a diagram used to provide a summary of the parsing result data stored in the parsing result storage section.

[0083] exist Figure 3 In this context, data group 301 is a data group corresponding to the semiconductor manufacturing process with process name = "Process I" and step name = "STEP1", and includes multiple data groups collected from units 120, 130, and 140.

[0084] Specifically, in data group 301, in addition to data group 201 collected from unit 120, there are also data groups for units 130 and 140 respectively, which are established with semiconductor manufacturing process with process name = "process I" and step name = "STEP1".

[0085] In the data processing apparatus 110, multiple data groups that undergo the same process and the same steps are analyzed, and data groups that produce the same effect are grouped together. In semiconductor manufacturing apparatuses, even when performing the same process and the same steps, there are cases where different results are obtained due to differences in the data contained in the data groups. Therefore, by grouping data groups that produce the same effect together and calculating specific data for each specific group, it is possible to calculate the range of data contained in each data group that is allowed to produce the same effect.

[0086] exist Figure 3 In this context, groups 311 to 314 are groups obtained by grouping data groups in data group 301 that achieve the same effect into one group. The specific data (ranges of each data point) specific to a group that achieves the same effect in the same process and step can be considered the smallest data unit in semiconductor manufacturing processes that imparts the same change to a "state". That is, the specific data (ranges of each data point) specific to this group can be considered the smallest data unit in the micro-processing of semiconductor manufacturing processes.

[0087] Thus, the smallest data unit (process element) in the micro-processing of semiconductor manufacturing is called "proxel" in the first embodiment. This is the same terminology used to refer to the smallest unit of an image (picture element) as "pixel" and the smallest unit of a volume (volume element) as "voxel".

[0088] In the first embodiment, the data parsing unit 111 calculates "Proxel" by parsing the collected data group and stores it as parsing result data in the parsing result storage unit 113.

[0089] <Hardware Composition of Data Processing Device>

[0090] Next, the hardware configuration of the data processing device 110 will be described. Figure 4 This is a diagram illustrating an example of the hardware configuration of a data processing device.

[0091] like Figure 4 As shown, the data processing device 110 includes a CPU (Central Processing Unit) 401, a ROM (Read-Only Memory) 402, and a RAM (Random Access Memory) 403. The CPU 401, ROM 402, and RAM 403 form what is called a computer. Additionally, the data processing device 110 includes an auxiliary storage device 404, an operation device 405, a display device 406, an I / F (Interface) device 407, and a driver device 408. All the hardware components of the data processing device 110 are interconnected via a bus 409.

[0092] CPU 401 executes various programs (e.g., data parsing programs, etc.) installed in auxiliary storage device 404.

[0093] ROM 402 is a non-volatile memory that functions as the main storage device. ROM 402 stores various programs and data required for CPU 401 to execute various programs installed on auxiliary storage device 404. Specifically, ROM 402 stores boot programs such as BIOS (Basic Input / Output System) and EFI (Extensible Firmware Interface).

[0094] RAM403 is a volatile memory such as DRAM (Dynamic Random Access Memory) or SRAM (Static Random Access Memory), which functions as the main storage device. RAM403 provides a working area for various programs installed on the auxiliary storage device 404 to be executed by the CPU 401.

[0095] The auxiliary storage device 404 stores various programs, data sets collected by the CPU 401 through the execution of these programs, and the calculated analysis results. The auxiliary storage device 404 implements a data storage unit 112 and an analysis result storage unit 113.

[0096] The operating device 405 is an input device used by the administrator of the data processing device 110 to input various instructions to the data processing device 110. The display device 406 is a display device that displays internal information of the data processing device 110.

[0097] I / F device 407 is a connection device connected to network 150 for communicating with terminals 121, 131, and 141 in units 120, 130, and 140.

[0098] The drive device 408 is a device for setting the recording medium 410. The recording medium 410 referred to herein includes media that record information optically, electrically, or magnetically, such as CD-ROMs, floppy disks, and optical discs. Alternatively, the recording medium 410 may include semiconductor memory that electrically records information, such as ROM or flash memory.

[0099] Furthermore, various programs installed on the auxiliary storage device 404 may be provided on the drive device 408 via the allocated recording medium 410, and the various programs recorded on the recording medium 410 may be read and installed by the drive device 408. Alternatively, various programs installed on the auxiliary storage device 404 may be installed by downloading via the network 150.

[0100] <Functional Composition of the Data Analysis Unit of the Data Processing Device>

[0101] Next, the functional configuration of the data parsing unit 111 of the data processing device 110 will be explained. Figure 5 This is a diagram illustrating an example of the functional structure of a data parsing unit. For example... Figure 5 As shown, the data parsing unit 111 includes a collection unit 510, an effect calculation unit 520, a classification unit 530, and a Proxel calculation unit 540.

[0102] The collection unit 510 collects data groups 301 (e.g., data group 201, etc.) from terminals 121, 131, and 141 within each unit 120, 130, and 140 via network 150.

[0103] The effect calculation unit 520 is an example of a calculation unit that calculates the effect for each collected data set. The effect calculation unit 520 is configured to acquire, for each collected data set, state data showing the state before execution of the corresponding process and corresponding step, and state data showing the state after execution. It then uses this data to calculate the change in state before and after execution as the effect. Furthermore, the effect calculation unit 520 stores the calculated effect, along with setting data, output data, measurement data, and experimental data, as a data set in the data storage unit 112.

[0104] The classification unit 530 is an example of a segmentation unit that reads out multiple data groups stored in the data storage unit 112 and analyzes their distribution in the feature space. When the data contained in each data group has K categories, the classification unit 530 analyzes the distribution of the data groups in the K-dimensional feature space.

[0105] Specifically, the classification unit 530 groups data groups that produce the same result from multiple read data groups together. Furthermore, the classification unit 530 segments the K-dimensional feature space so that data groups distributed within the feature space are classified into groups.

[0106] The Proxel calculation unit 540 is an example of an output unit. The Proxel calculation unit 540 calculates the Proxel by calculating the range of each data of each of the K categories in each region of the K-dimensional feature space after being segmented by the classification unit 530 (by group-specific specific data), and stores it as parsing result data in the parsing result storage unit 113.

[0107] <Specific examples of processing by each part of the data parsing department>

[0108] Next, specific examples of the processing of the effect calculation unit 520, classification unit 530, and Proxel calculation unit 540 in each part of the data parsing unit 111 (collection unit 510, effect calculation unit 520, classification unit 530, and Proxel calculation unit 540) will be explained.

[0109] (1) Specific examples of the processing in the effect calculation department

[0110] First, a specific example of the processing of the effect calculation unit 520 will be explained. Figure 6 This is a diagram illustrating a specific example of the processing by the effect calculation unit.

[0111] like Figure 6 As shown, the relationship between the specified steps of the semiconductor manufacturing process (process name = "process I", step name = "STEP1") and the data group can be schematically represented as shown by dashed line 600.

[0112] That is, if a semiconductor manufacturing apparatus with predefined settings executes a specified step of a semiconductor manufacturing process, the state before execution (any one of the properties of the object being processed before execution, the state of the semiconductor manufacturing apparatus, or the environment within the semiconductor manufacturing apparatus) will change after execution. Moreover, the execution status of the semiconductor manufacturing process at this time can be specified through output data, measurement data, and experimental data.

[0113] In other words, under specific execution conditions involving setting data, outputting data, measuring data, and experimental data, the effect of specifying the prescribed steps in the semiconductor manufacturing process can be achieved through:

[0114] • Data representing the state before execution (first data) and

[0115] • It is represented by the difference in the data (second data) indicating the state after execution.

[0116] Therefore, in the effect calculation unit 520, data representing the state before execution and data representing the state after execution, corresponding to the data groups for each process and each step, are obtained. For example, the data groups included in output data, measurement data, and experimental data are divided into data that changes due to the execution of the prescribed steps and other data. The changed data is obtained as data representing the state before execution and data representing the state after execution. In addition, the other data is obtained as data for a specific execution condition.

[0117] Furthermore, in the effect calculation unit 520, the effect of each execution state in the process and step is calculated by calculating the difference between the two (data representing the state before execution and data representing the state after execution). In addition, the effect calculation unit 520 establishes a correspondence between the calculated effect and the setting data, output data, measurement data, and experimental data, and stores them as a data set in the data storage unit 112.

[0118] Figure 7 This diagram represents an example of a data group stored in the data storage unit. It is an example of a data group stored in the data storage unit 112 by the effect calculation unit 520 for a semiconductor manufacturing process with process name = "Process I" and step name = "STEP1".

[0119] like Figure 7 As shown, the data group stored in the data storage unit 112 by the effect calculation unit 520 includes items such as "data group identifier", "setting data (R)", "output data (E)", "measurement data (Pl)", "experimental data (Pr)" and "effect" as information.

[0120] A "data cluster identifier" is an identifier used to identify individual data clusters. Figure 7 In this context, a data cluster identifier = "data001" is, for example, a data cluster containing data and effects collected through unit 120 (unit name = "unit A"). Similarly, a data cluster identifier = "data002" is, for example, a data cluster containing data and effects collected through unit 130 (unit name = "unit B").

[0121] In the information items from "Setup Data (R)" to "Experimental Data (Pr)," the data sets collected by each unit 120, 130, and 140 are stored. Figure 2 The data group in the reference (excluding the initial data (I) and target data (Pf)).

[0122] The effects calculated by the effects calculation unit 520 are stored in "Effects". According to... Figure 7For example, in a semiconductor manufacturing process with process name = "Process I" and step name = "STEP1", under specific execution conditions such as establishing corresponding setting data with data group identifier = "data 001", the following effect is obtained: <1> Similarly, in the case of a semiconductor manufacturing process with process name = "Process I" and a step name = "STEP1", under specific execution conditions such as establishing corresponding setting data with data group identifier = "Data 002", the following effect is obtained: <2> ".

[0123] (2) Specific examples of classification processing

[0124] Next, a specific example of the processing in the classification section 530 will be explained. Figure 8 This is a diagram illustrating a specific example of the processing in the classification department.

[0125] like Figure 8 As shown, the classification unit 530 reads out multiple data groups stored in the data storage unit 112 according to each process and each step, and marks them in the feature space 800. Figure 8 In the diagram, a solid circle containing numbers indicates one of the multiple data groups read out, and the numbers recorded inside the solid circle represent the data group identifier for that data group.

[0126] In addition, Figure 8 In the example, to simplify the explanation, the feature space 800 is set to two dimensions (that is, to show the case of two types of data (data type p, data type q) contained in the data group).

[0127] exist Figure 8 In the diagram, the dashed circles surrounding the solid circles indicate that data groups that will achieve the same effect are grouped together. That is, the data groups identified by the data group identifiers recorded within the solid circles contained within the dashed circles are all data groups that have the same effect in the semiconductor process with process name = "Process I" and step name = "STEP1".

[0128] For example, the dashed circle 801 contains data group identifiers: “data 001”, “data 004”, and “data 010”. Although the solid circles containing these data group identifiers are located relatively close to each other in the feature space 800, they do not completely overlap. That is, the data groups identified by each data group identifier are similar, but not entirely identical.

[0129] On the other hand, these data sets were effective in both the semiconductor manufacturing process with process name = "Process I" and the step with step name = "STEP1". <1> The data group. In other words, for multiple data groups grouped together by dashed circles 801 in feature space 800, the effect of executing process I and STEP1 is obtained under any data group. <1> The data set.

[0130] In addition, Figure 8 In the diagram, dashed circle 802 contains data group identifiers "data 005", "data 006", and "data 007", while dashed circle 803 contains data group identifier "data 002". The data groups contained in dashed circle 802 and the data groups contained in dashed circle 803 are located in mutually separate positions.

[0131] On the other hand, the data groups identified by the data group identifiers recorded within the solid circles of the dashed circle 802 all achieve the desired effect when performing process I and STEP1 under each data group. <2> The data group. Similarly, the data group identified by the data group identifier recorded within the solid circle contained in the dashed circle 803 is the data group under which process I and STEP1 are executed, and the effect is obtained. <2> The data set.

[0132] Here, if dashed circles 802 and 803 are grouped together by a dashed circle, they overlap with other dashed circles 801. Therefore, in the classification unit 530, even if the same effect is obtained, they are grouped separately (that is, the classification unit 530 groups the data when segmenting the feature space to prevent the data groups corresponding to different effects from being mixed in the same region).

[0133] Furthermore, the classification unit 530 segments the feature space so that the data groups distributed in the feature space are classified into groups. In addition, the classification unit 530 segments the feature space, for example, by clustering the data groups distributed in the K-dimensional feature space using "effect" as an index.

[0134] (3) Specific examples of processing by the Proxel computing unit

[0135] Next, a specific example of the processing by the Proxel calculation unit 540 will be described. As described above, the Proxel calculation unit 540 calculates the Proxel by calculating the range of each data in each region of the feature space segmented by the classification unit 530 (by using specific data in a specific group). Figure 9 This is a diagram showing an example of a Proxel calculated by the Proxel computation unit.

[0136] like Figure 9As shown, the Proxel calculation unit 540 calculates the range of each data in each region of the feature space by calculating the minimum value and the maximum value of each data included in the data groups grouped into the same group by the classification unit 530.

[0137] Figure 9 An example of shows that the data groups that achieve the same effect as the effect <1> are grouped by the classification unit 530 into a group with the group name = "Group Gr1". Additionally, Figure 9 An example of shows that for "Pressure" of the set data in each data of the data group grouped into the group with the group name = "Group Gr1",

[0138] · Minimum value = "Pressure_1"

[0139] · Maximum value = "Pressure_4".

[0140] The range of each data in the region of the feature space where the data group grouped into the group with the group name = "Group Gr1" is distributed can be specifically represented by the dashed line 900. In addition, the range of each data represented by the dashed line 900 is only the Proxel (specific data specified by the group 311) described in Figure 3 .

[0141] Figure 10 is the first flowchart showing the process flow of Proxel calculation based on the classification unit and the Proxel calculation unit.

[0142] In step S1001, the classification unit 530 reads out the data group corresponding to the process and step of the processing target from the data storage unit 11​​​​​​​​​​​​​​​​

[0148] One advantage of the Proxel Computing Unit 540 is the improved ease of processing multiple data sets collected from units 120, 130, and 140.

[0149] Figure 11 This is the first diagram used to illustrate the advantages of calculating Proxel. Figure 11 In this example, multiple data clusters 1100 are shown, each representing a data cluster collected from units 120, 130, and 140 respectively, all yielding the same result. Furthermore, in... Figure 11 For the sake of simplicity, the data included in each data group is categorized into 5 types.

[0150] The fact that a portion of “Measured Data” = “Ion energy” and a portion of “Experimental Data” = “Etching rate” in multiple data groups 1100 are empty is because the unit does not have the measuring instrument or experimental value measuring instrument to measure the data.

[0151] On the other hand, Figure 11 In this context, Proxel1110 is an example of a Proxel calculated by the Proxel computation unit 540 based on multiple data groups 1100.

[0152] By calculating Proxel1110, the same effect can be obtained ("effect") <10> Multiple data groups are treated as a single data group. In this way, by calculating Proxel1110, incomplete data groups containing empty columns can be imputed and treated as a single, highly versatile data group without empty columns. That is, by calculating Proxel, highly versatile data processing can be achieved.

[0153] (2) To homogenize the density of the data cluster.

[0154] One advantage of the Proxel computing unit 540 in computing Proxel is that it becomes less susceptible to the influence of density deviations in multiple data clusters collected from each of the units 120, 130, and 140. That is, it can homogenize the density of data clusters in the feature space.

[0155] Figure 12 This is the second figure used to illustrate the advantages of calculating Proxel. Figure 12 In the diagram, the horizontal axis represents the data type P (here, "HF power"), and the vertical axis represents the data type Q (here, "LF power").

[0156] exist Figure 12In the feature space 1200 shown, white circles represent the distribution of each data group, and regular hexagons represent Proxels. For example... Figure 12 As shown, there is a bias in the density of the distribution of multiple data clusters collected from units 120, 130, and 140 in the feature space 1200. In contrast, in the case of Proxel, it is possible to arrange them uniformly in the feature space 1200.

[0157] In this way, by calculating Proxel, data clusters from various regions of the 1200 feature space can be processed equally. Therefore, for example, when using Proxel for machine learning, the influence of bias in the data clusters can be suppressed. In other words, by calculating Proxel, highly versatile data processing can be achieved.

[0158] (3) Capable of forming representative models

[0159] One advantage of Proxel's computing unit 540 is its ability to generate representative models and estimate the results of corresponding steps using Proxel.

[0160] Figure 13 This is Figure 3, used to illustrate the advantages of calculating Proxel. (See Figure 3.) Figure 13 As shown, since the effect of this step is associated with Proxel (specific data specific to group 311), the result of this step 1311 can be estimated when the initial data 1301 is input. Similarly, the result of this step 1312 can be estimated when the initial data 1302 is input.

[0161] In this way, by calculating Proxel, the result can be estimated regardless of the characteristics of the semiconductor manufacturing device. In other words, highly versatile data processing can be achieved by calculating Proxel.

[0162] Summary

[0163] As clearly stated from the above description, in the data processing apparatus 110 according to the first embodiment,

[0164] • Collect data clusters for each process and each step, and calculate the effect for each collected data cluster.

[0165] • Divide the feature space so that the data groups that achieve the same degree of effect in their respective distributions in the feature space are classified into the same groups.

[0166] • The Proxel is calculated by measuring the range of each data point in each region of the segmented feature space and stored as the parsing result data.

[0167] Thus, according to the first embodiment, the following advantages can be obtained: the ease of processing the collected data group is improved, the density of the distribution of the collected data group in the feature space can be made uniform, and a model can be formed to estimate the result of this step, etc.

[0168] That is, according to the first embodiment, it is possible to perform data processing with high versatility on the data group collected in the manufacturing process.

[0169] [Second Embodiment]

[0170] In the above first embodiment, it was described that each data included in the data group is evenly processed. However, in the data group, there are data with high contribution degrees to the effect and data with low contribution degrees. Therefore, in the second embodiment, clustering processing is performed on the basis of weighting the data according to the contribution degree to the effect.

[0171] In addition, in the above first embodiment, it was described that the feature space is divided by clustering processing to generate each region, and Proxel is calculated. In contrast, in the second embodiment, after the feature space is divided by clustering processing to generate each region, for the regions with low contribution degrees to the effect, Proxel is calculated by integrating with other regions.

[0172] Hereinafter, the second embodiment will be described centering on the differences from the above first embodiment.

[0173] <Flow of Proxel Calculation Processing>

[0174] Figure 14 It is the second flowchart showing the flow of Proxel calculation processing based on the classification unit and the Proxel calculation unit. The differences from the Proxel calculation processing described in the first embodiment are steps S1401 to S1404. Figure 10 In step S1401, the classification unit 530 determines the contribution degree to the effect for each data included in the data group stored in the data storage unit 112. <​​​​​​​In step S1403, the Proxel calculation unit 540 integrates regions with low contribution to the effect in each region of the feature space segmented by the classification unit 530 with other regions.

[0178] In step S1404, the Proxel calculation unit 540 calculates the Proxel by calculating the range of each data in each integrated region, and stores it as parsing result data in the parsing result storage unit 113.

[0179] Summary

[0180] As clearly stated from the above description, in the data processing apparatus 110 according to the second embodiment, in addition to the content of the first embodiment described above, the following processing is also performed:

[0181] • When segmenting the feature space, further consider the contribution of each data point contained in the data set to the effect.

[0182] • Integrate regions with low contribution to the effect from the segmented feature space with other regions.

[0183] Therefore, according to the data processing apparatus 110 of the second embodiment, the number of Proxels can be reduced, and Proxels corresponding to the contribution level can be calculated.

[0184] [Other implementation methods]

[0185] In the first and second embodiments described above, Proxel is calculated by calculating the range of each data point in each region of the segmented feature space. However, the method for calculating Proxel is not limited to this.

[0186] For example, if the regions in the segmented feature space are significantly separated within the feature space, the Proxel can be calculated by further calculating the range of data for each of the transformed regions, based on the process of deforming each region to make them adjacent to each other. This reduces the number of blank areas (regions where the Proxel is not specified) in the feature space.

[0187] Furthermore, in the first and second embodiments described above, the Proxel is calculated by calculating the range of each data in each region of the segmented feature space, and the Proxel is stored as parsing result data in the parsing result storage unit 113. However, the parsing result data stored in the parsing result storage unit 113 is not limited to the Proxel. For example, representative data (specific data for specific groups) representing each region of the segmented feature space may also be stored as parsing result data.

[0188] Furthermore, in the first and second embodiments described above, it was explained that a data parsing program was installed in the data processing device 110, and the data parsing unit 111 was implemented in the data processing device 110. However, the data parsing program may also be installed, for example, in the terminals 121, 131, and 141 within each of the units 120, 130, and 140, and the data parsing unit 111 may be implemented in the terminals 121, 131, and 141 within each of the units 120, 130, and 140.

[0189] Furthermore, while the first and second embodiments described above explained the calculation of the Proxel based on data collected during the semiconductor manufacturing process, the data for calculating the Proxel is not limited to data collected during the semiconductor manufacturing process. Even manufacturing processes other than semiconductor manufacturing processes, such as those involving devices using plasma, generally involve more complex data sets. Therefore, the advantages described above can be obtained when calculating the Proxel based on data collected during the manufacturing process included in the device.

[0190] Furthermore, the present invention is not limited to the configurations described in the above embodiments, combinations with other elements, or the configurations shown herein. Modifications can be made in these aspects without departing from the spirit of the invention, and can be appropriately determined depending on the application.

[0191] This application asserts priority based on Japanese Patent Application No. 2017-253694, filed on December 28, 2017, the entire contents of which are incorporated herein by reference.

[0192] Explanation of reference numerals in the attached figures:

[0193] 100...Data processing system; 110...Data processing device; 111...Data parsing unit; 201...Data group; 311-314...Group; 510...Collection unit; 520...Effect calculation unit; 530...Classification unit; 540...Proxel calculation unit; 800...Feature space; 900...Proxel.

Claims

1. A data processing device, comprising: A calculation unit collects data groups corresponding to predetermined steps in the process and calculates the effect of the predetermined step with respect to each data group; a segmentation unit that segments the feature space so as to establish distributions of respective data groups in the feature space according to each of the calculated effect classifications and the prescribed steps; and An output unit outputs specific data, wherein the specific data specifically identifies each region of the feature space after segmentation. The calculation unit calculates the effect for each data group based on a difference between first data indicating a state before executing a predetermined step in the process and second data indicating a state after executing the predetermined step in the process.

2. The data processing device according to claim 1, wherein: The division unit divides the feature space so that data groups having the calculated effects at the same level are classified into the same group.

3. The data processing device according to claim 2, wherein: The segmentation unit weights each data included in the data group based on a contribution of each data included in the data group to the effect, and segments the feature space by using the data group including the weighted data.

4. The data processing device according to claim 1, wherein: The segmentation unit deforms each segmented region of the feature space so that the regions do not overlap with each other.

5. The data processing device according to claim 1, wherein: The above process is a semiconductor manufacturing process. The step is a minimum processing unit for changing a state in the semiconductor manufacturing process, and the state includes at least a property of an object to be processed in the semiconductor manufacturing process.

6. A data processing method, comprising: A calculation process collects data groups corresponding to predetermined steps in the process and calculates the effect of the predetermined step for each data group; a segmentation step of segmenting the feature space so as to establish distributions of the plurality of data groups corresponding to each of the effect classifications calculated and the prescribed steps in the feature space; and The output step outputs specific data, wherein the specific data specifies each region of the feature space after segmentation. In the calculation step, the effect is calculated for each data group based on a difference between first data indicating a state before executing a predetermined step in the process and second data indicating a state after executing the predetermined step in the process.

7. A recording medium storing a program for causing a computer to execute the following steps: A calculation process collects data groups corresponding to predetermined steps in the process and calculates the effect of the predetermined step for each data group; a segmentation step of segmenting the feature space so as to establish distributions of the plurality of data groups corresponding to each of the effect classifications calculated and the prescribed steps in the feature space; and The output step outputs specific data, wherein the specific data specifies each region of the feature space after segmentation. In the calculation step, the effect is calculated for each data group based on a difference between first data indicating a state before executing a predetermined step in the process and second data indicating a state after executing the predetermined step in the process.

Images