Substrate processing apparatus and information processing system

Abstract

The substrate processing apparatus of the present invention comprises: at least one detector for detecting physical quantities of an object during the grinding and / or cleaning and / or drying of the substrate; a transformation unit for transforming the detection values ​​detected by the aforementioned detector during grinding and / or cleaning and / or drying into feature quantities for each processing step of a learned machine learning model; and an inference unit for outputting a predicted value of at least one of the number of defects, defect size, and defect location in the substrate of the object by inputting object data including the aforementioned feature quantities into the learned machine learning model; the aforementioned learned machine learning model is learned using a learning dataset that includes feature quantities containing the detection values ​​detected by the detector during grinding and / or cleaning on the production line of the object or on a production line of the same type as the production line of the object, as input data, and at least one of the number of defects, defect size, and defect location in the substrate as output data.

Description

[Technical Field]

[0001] This invention relates to a substrate processing apparatus and an information processing system. [Previous Technology]

[0002] The cleaning performance of a substrate processing apparatus (e.g., a polishing apparatus) is evaluated by measuring the substrates (specifically, wafers) discharged from the apparatus after processing such as polishing, cleaning, and drying using a dedicated defect inspection device (see, for example, Patent Document 1). Because defect inspection is costly (mainly time-consuming), it is difficult to perform a full inspection after substrate processing (e.g., polishing, cleaning, drying) at the manufacturing site; therefore, sampling inspection is implemented. [Prior Art Documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2002-257533 [Summary of the Invention]

[0004] (The problem that the invention is intended to solve)

[0005] However, uninspected substrates (specifically wafers) may also be defective due to insufficient cleaning, etc.

[0006] The present invention addresses the aforementioned problems by providing a substrate processing apparatus and information processing system that can determine whether a substrate after processing is defective without using a defect inspection device. (Means for solving the problem)

[0007] One embodiment of the substrate processing apparatus of the present invention comprises: at least one detector that detects a physical quantity of an object during the grinding and / or cleaning and / or drying of the substrate; a transformation unit that, for each processing step, transforms the detection values ​​detected by the aforementioned detector during grinding and / or cleaning and / or drying into a feature quantity using a learned machine learning model; and an inference unit that, by inputting object data including the aforementioned feature quantity into the learned machine learning model, outputs a predicted value of at least one of the number of defects, defect size, and defect location in the substrate of the object; the aforementioned learned machine learning model is learned using a learning dataset containing the feature quantity of the detection values ​​detected by the detector during grinding and / or cleaning on the production line of the object or on a production line of the same type as the production line of the object, as input data, and at least one of the number of defects, defect size, and defect location in the substrate as output data.

[0008] When this configuration is adopted, since the defect inspection device is not used for inspection, the predicted value of at least one of the defect number, defect size and defect location of the substrate after substrate processing can be obtained. Therefore, it can be inferred whether the substrate after substrate processing is a defective product without inspection by the defect inspection device.

[0009] Furthermore, in the aforementioned substrate processing apparatus, the input data of the aforementioned machine learning model during learning further includes the dwell time of each unit contained in the substrate processing apparatus. Each unit contained in the substrate processing apparatus further includes a unit dwell time statistics unit, which systematically counts the dwell time of the unit. The aforementioned object data input into the aforementioned machine learning model after learning may also further include the dwell time of each unit counted by the aforementioned unit dwell time statistics unit.

[0010] Furthermore, in the aforementioned substrate processing apparatus, the input data of the aforementioned machine learning model during learning further includes a second feature quantity that changes the position of the component used for grinding or cleaning. The aforementioned transformation unit transforms the position of the component used for grinding or cleaning into the second feature quantity. The aforementioned object data input into the aforementioned machine learning model after learning is completed may also further include the second feature quantity of each component transformed by the aforementioned transformation unit.

[0011] Furthermore, in the aforementioned substrate processing apparatus, the input data of the aforementioned machine learning model during learning further includes processing recipe information, which includes instruction values ​​for the units contained in the substrate processing apparatus. The aforementioned object data input to the aforementioned machine learning model after learning may also further include processing recipe information, which includes further instruction values ​​for the units contained in the substrate processing apparatus.

[0012] Furthermore, the aforementioned substrate processing apparatus includes: a regression analysis unit that outputs correlation parameters representing a correlation between a plurality of detection values ​​and one of the defect number, defect size, and defect location in the substrate, according to a predetermined regression analysis algorithm; a receiving unit that receives at least one detector whose output is based on the feature quantity contained in the input data of the machine learning model; and a learning unit that relearns the aforementioned machine learning model by changing the feature quantity of the detection values ​​received by the aforementioned detector; the aforementioned inference unit can also output the aforementioned predicted value using the machine learning model relearned by the aforementioned learning unit.

[0013] Another aspect of the information processing system of the present invention includes: a transformation unit that transforms the detection values ​​detected by a detector of a substrate processing apparatus during the grinding and / or cleaning and / or drying of the substrate into feature quantities for each processing step of a learned machine learning model; and an inference unit that, by inputting object data including the aforementioned feature quantities, outputs a predicted value of at least one of the number of defects, defect size, and defect location in the substrate; wherein the aforementioned learned machine learning model is learned using a learning dataset that includes feature quantities containing the detection values ​​detected by a detector during grinding and / or cleaning of the object on a production line or a production line of the same type as the object's production line as input data, and at least one of the number of defects, defect size, and defect location in the substrate as output data. (Effects of the Invention)

[0014] When one embodiment of the present invention is adopted, since the defect inspection device is not used for inspection, the predicted value of at least one of the defect number, defect size and defect location of the substrate after substrate processing can be obtained. Therefore, it is possible to infer whether the substrate after substrate processing is a defective product without using a defect inspection device.

Implementation Method

[0016] Hereinafter, various embodiments will be described with reference to the drawings. However, detailed descriptions will be omitted where necessary. For example, detailed descriptions of already known matters and repetitive descriptions of substantially the same structures will be omitted. This is to avoid unnecessary length in the following descriptions and to make it easier for those skilled in the art to understand. This embodiment describes a polishing apparatus for flattening the surface of a substrate W by chemical mechanical polishing (CMP), as an example of a substrate processing apparatus.

[0017] FIG1 is a diagram showing an example of the schematic configuration of the substrate processing apparatus of this embodiment. As shown in FIG1, the substrate processing apparatus 1 is a polishing apparatus for flattening the surface of a substrate W such as a silicon wafer (hereinafter also simply referred to as a wafer) by chemical mechanical polishing (CMP).

[0018] As shown in FIG. 1, the substrate processing apparatus 1 includes, for example, a rectangular box-shaped frame 2. The frame 2 is roughly rectangular when viewed from above. A substrate transport path 3 extending in the length direction is provided in the center of the frame 2. A loading / unloading section 10 is provided at one end of the substrate transport path 3 in the length direction. A polishing section 20 is provided on one side of the substrate transport path 3 in the width direction, that is, in the direction orthogonal to the length direction when viewed from above, and a cleaning section 30 is provided on the other side. A substrate transport section 40 for transporting substrates W is provided on the substrate transport path 3. In addition, the substrate processing apparatus 1 includes a control section 50 for uniformly controlling the operation of the loading / unloading section 10, the polishing section 20, the cleaning section 30, and the substrate transport section 40.

[0019] The loading / unloading unit 10 includes a front loading unit 11 for housing the substrate W. A plurality of front loading units 11 are provided on one side of the rack 2 along its length. The plurality of front loading units 11 are arranged along the width of the rack 2. The front loading units 11 may be equipped with, for example, open cassettes, SMIF (Standard Manufacturing Interface) cassettes, or FOUP (Front Opening Unified Pod). Both SMIF and FOUP are cassettes that house the substrate W internally and are sealed containers covered by partition walls, maintaining an environment independent of the external space.

[0020] Furthermore, the loading / unloading unit 10 includes: two transport robots 12 for moving substrate W from and into the front loading unit 11; and a travel mechanism 13 for moving each transport robot 12 along the arrangement of the front loading unit 11. Each transport robot 12 has two arms, one above the other, and these two arms are used separately before and after processing the substrate W. For example, each transport robot 12 uses its upper arm when returning the substrate W to the front loading unit 11, and uses its lower arm when removing the substrate W from the front loading unit 11 before processing.

[0021] The polishing unit 20 includes a plurality of polishing devices 21 (21A, 21B, 21C, 21D) for polishing the substrate W. The plurality of polishing devices 21 are arranged along the length of the substrate transport path 3. Each polishing device 21 includes: a polishing table 23 for rotating a polishing pad 22 having a polishing surface; an upper annular turntable 24 for holding the substrate W and pressing the substrate W against the polishing pad 22 on the polishing table 23 while polishing; a polishing fluid supply nozzle 25 for supplying polishing fluid, dressing fluid, etc. to the polishing pad 22; a dressing device 26 for dressing the polishing surface of the polishing pad 22; and an atomizer 27 for spraying a mixture of liquid such as pure water and gas such as nitrogen, or liquid such as pure water, into a mist onto the polishing surface.

[0022] The polishing apparatus 21 supplies polishing slurry from the polishing slurry supply nozzle 25 to the polishing pad 22, and presses the substrate W onto the polishing pad 22 by means of the upper annular turntable 24. By moving the upper annular turntable 24 relative to the polishing table 23, the substrate W is polished, and the surface of the substrate W is flattened. In addition, the upper annular turntable 24 has a plurality of pressure-reducing zones (e.g., air bladders) arranged in concentric circles. The upper annular turntable 24 adjusts the degree of pressure applied to the substrate W on the polishing pad 22 by adjusting the pressure in these plurality of pressure-reducing zones.

[0023] The dressing device 26 fixes hard particles such as diamond particles or ceramic particles on the rotating part at the front end before contacting the grinding pad 22. By rotating and shaking the rotating part, the entire grinding surface of the grinding pad 22 is evenly dressed to form a flat grinding surface.

[0024] The atomizer 27 uses high-pressure fluid to flush away the grinding debris and grinding particles remaining on the grinding surface of the grinding pad 22, thereby purifying the grinding surface and using the dressing device 26 to dress the grinding surface, that is, to reproduce the grinding surface.

[0025] The cleaning unit 30 includes a plurality of cleaning devices 31 (31A, 31B) for cleaning the substrate W, and a substrate drying device 32 for drying the cleaned substrate W. The plurality of cleaning devices 31 and the substrate drying device 32 are arranged along the length of the substrate transport path 3. A first transport chamber 33 is provided between the cleaning devices 31A and 31B. A transport robot 35 for transporting the substrate W between the substrate transport unit 40, the cleaning devices 31A and 31B is provided in the first transport chamber 33.

[0026] The transfer robot 35 has two arms, one on top and one on the bottom. The two arms are used separately before the substrate W is cleaned in the cleaning device 31A and after the substrate W is cleaned in the cleaning device 31A. For example, the transfer robot 35 uses the lower arm when it takes the substrate W out of the temporary stage (also called the wafer station) 47 (described later) before cleaning and transfers it to the cleaning device 31A, and uses the upper arm when it takes the substrate W out of the cleaning device 31A after cleaning and transfers it to the cleaning device 31A.

[0027] In addition, a second transfer chamber 34 is provided between the cleaning device 31B and the substrate drying device 32. A transfer robot 36 for transferring the substrate W between the cleaning device 31B and the substrate drying device 32 is provided in the second transfer chamber 34.

[0028] The cleaning device 31 includes a roller sponge (hereinafter also referred to as a roller) type cleaning module, which is used to clean the substrate W. Alternatively, cleaning devices 31A and 31B can be of the same type or different types of cleaning modules. Furthermore, cleaning devices 31A and 31B can replace the roller sponge type cleaning module with, for example, a pencil sponge (hereinafter also referred to as a pen) type cleaning module or a two-fluid jet type cleaning module. Here, in the case of a two-fluid jet type cleaning module, the two fluids are, for example, a mixture of nitrogen (N2) and pure water.

[0029] The substrate drying apparatus 32 is equipped with a drying module that performs, for example, nitrogen (N2) drying or Rotagoni drying using isopropyl alcohol (IPA). After the substrate W is dried using Rotagoni, the shutter 1a of the partition wall between the substrate drying apparatus 32 and the loading / unloading section 10 is opened, and the substrate is removed from the substrate drying apparatus 32 by the transfer robot 12.

[0030] The substrate transport unit 40 includes: a lift 41, a first linear conveyor 42, a second linear conveyor 43, and a swing conveyor 44. In the substrate transport path 3, starting from the loading / unloading unit 10 side, there are sequentially arranged a first transport position TP1, a second transport position TP2, a third transport position TP3, a fourth transport position TP4, a fifth transport position TP5, a sixth transport position TP6, and a seventh transport position TP7.

[0031] The elevator 41 is a mechanism for moving the substrate W up and down at the first transport position TP1. The elevator 41 receives the substrate W from the transport robot 12 of the loading / unloading unit 10 at the first transport position TP1. Then, the elevator 41 delivers the substrate W received from the transport robot 12 to the first linear conveyor 42. A shutter 1b is provided on the partition wall between the first transport position TP1 and the loading / unloading unit 10. When the substrate W is transported, the shutter 1b is opened, and the substrate W is delivered from the transport robot 12 to the elevator 41.

[0032] The first linear conveyor 42 is a mechanism for conveying substrate W between two of the first conveying position TP1, the second conveying position TP2, the third conveying position TP3, and the fourth conveying position TP4. The first linear conveyor 42 includes: a plurality of conveying arms 45 (45A, 45B, 45C, 45D); and a linear guide mechanism 46 for moving each conveying arm 45 in a plurality of horizontal height directions.

[0033] The conveying arm 45A moves between the first conveying position TP1 and the fourth conveying position TP4 via the linear guide mechanism 46. The conveying arm 45A is used to receive the substrate W from the elevator 41 and deliver the substrate W to the pass hand of the second linear conveyor 43.

[0034] The conveying arm 45B moves between the first conveying position TP1 and the second conveying position TP2 via a linear guide mechanism 46. At the first conveying position TP1, the conveying arm 45B receives the substrate W from the elevator 41, and at the second conveying position TP2, it delivers the substrate W to the grinding apparatus 21A. The conveying arm 45B is equipped with a lifting drive unit, which rises when delivering the substrate W to the upper annular turntable 24 of the grinding apparatus 21A, and descends after delivering the substrate W to the upper annular turntable 24. Similarly, the conveying arms 45C and 45D are also equipped with the same lifting drive unit.

[0035] The conveying arm 45C moves between the first conveying position TP1 and the third conveying position TP3 via the linear guide mechanism 46. At the first conveying position TP1, the conveying arm 45C receives the substrate W from the elevator 41 and delivers the substrate W to the grinding device 21B at the third conveying position TP3. Furthermore, the conveying arm 45C also functions as a retrieval arm, receiving the substrate W from the annular turntable 24 above the grinding device 21A at the second conveying position TP2 and delivering the substrate W to the grinding device 21B at the third conveying position TP3.

[0036] The conveying arm 45D moves between the second conveying position TP2 and the fourth conveying position TP4 via the linear guide mechanism 46. The conveying arm 45D functions as an access arm for receiving the substrate W from the annular turntable 24 above the grinding device 21A or the grinding device 21B at the second conveying position TP2 or the third conveying position TP3, and for delivering the substrate W to the retrieval arm of the swing conveyor 44 at the fourth conveying position TP4.

[0037] The swing conveyor 44 has an arm that can move between the fourth transport position TP4 and the fifth transport position TP5, and delivers the substrate W from the first linear conveyor 42 to the second linear conveyor 43. Furthermore, the swing conveyor 44 delivers the substrate W polished by the polishing unit 20 to the cleaning unit 30. A temporary storage table 47 for the substrate W is provided on the side of the swing conveyor 44. The swing conveyor 44 reverses the substrate W received at the fourth transport position TP4 or the fifth transport position TP5 and loads it onto the temporary storage table 47. The substrate W loaded on the temporary storage table 47 is transported to the first transport chamber 33 by the transport robot 35 of the cleaning unit 30.

[0038] The second linear conveyor 43 is a mechanism for conveying substrate W between two of the fifth conveying position TP5, the sixth conveying position TP6, and the seventh conveying position TP7. The second linear conveyor 43 includes: a plurality of conveying arms 48 (48A, 48B, 48C); and a linear guide mechanism 49 that moves each conveying arm 48 horizontally at a plurality of heights. The conveying arms 48A move between the fifth conveying position TP5 and the sixth conveying position TP6 by means of the linear guide mechanism 49. The conveying arms 45A function as access arms for receiving substrate W from the swing conveyor 44 and delivering the substrate W to the polishing apparatus 21C.

[0039] The transfer arm 48B moves between the sixth transfer position TP6 and the seventh transfer position TP7. The transfer arm 48B functions as an access arm for receiving the substrate W from the polishing device 21C and delivering the substrate W to the polishing device 21D. The transfer arm 48C moves between the seventh transfer position TP7 and the fifth transfer position TP5. The transfer arm 48C functions as a receiving arm for receiving the substrate W from the annular turntable 24 above the polishing device 21C or polishing device 21D at the sixth transfer position TP6 or the seventh transfer position TP7, and delivering the substrate W to the access arm of the swing conveyor 44 at the fifth transfer position TP5. Furthermore, the operation of the transfer arm 48B when delivering the substrate W is the same as the operation of the first linear conveyor 42 described above, and therefore its description is omitted.

[0040] The substrate processing apparatus 1 includes at least one detector (not shown) for detecting physical quantities of an object during grinding and / or cleaning and / or drying of a substrate (e.g., a wafer). The physical quantities of the object, for example during grinding, are as follows: • Rotation speed and / or torque of the grinding table 23 • Rotation speed and / or torque of the upper annular turntable 24 • Airbag pressure of the upper annular turntable 24 • Rotation speed and / or load of the dressing device 26 • Slurry / water flow rate • Flow rate of the atomizer 27 • Flow rate of nitrogen (N2) in the atomizer 27

[0041] The physical quantities of the object, for example during cleaning, are as follows: • Drum rotation speed and / or torque and / or load • Pen rotation speed and / or torque and / or load • Flow rate of the solution / water • Wafer rotation speed • Flow rate of nitrogen (N2)

[0042] The physical quantities of the object, for example during drying, are as follows: ‧Flow rate of nitrogen (N2) ‧Flow rate of isopropanol (IPA)

[0043] The substrate processing apparatus 1 may also include, as one example of the aforementioned detectors, a detector for detecting the rotation speed and / or torque of the worktable; a detector for detecting the rotation speed and / or torque of the upper annular turntable; a detector for detecting the pressure of the air bladder on the upper annular turntable; a detector for detecting the rotation speed and / or load of the trimmer; and a detector for detecting the flow rate of the slurry / water. Furthermore, the substrate processing apparatus 1 may also include: a detector for detecting the rotation speed and / or torque and / or load of the drum of the cleaning device 31; a detector for detecting the rotation speed and / or torque and / or load of the pen of the cleaning device 31; a detector for detecting the flow rate of the cleaning solution / water of the cleaning device 31; and a detector for detecting the rotation speed of the wafer of the cleaning device 31. Furthermore, the substrate processing apparatus 1 may also include: a detector for detecting the flow rate of nitrogen (N2); and a detector for detecting the flow rate of isopropanol (IPA).

[0044] Hereinafter, the components constituting the loading / unloading section 10, the grinding section 20, the cleaning section 30 and the substrate conveying section 40 will be referred to as units.

[0045] FIG2 is a block diagram showing an example of the schematic configuration of the control unit of this embodiment. As shown in FIG2, the control unit 50 includes: a unit control unit 51, a processor 6, and a memory unit 7. The unit control unit 51 uniformly controls the operation of each unit of the loading / unloading unit 10, the polishing unit 20, the cleaning unit 30, and the substrate conveying unit 40.

[0046] The memory unit 7 stores a machine learning model 71 that has been learned. This machine learning model 71 is learned using a learning dataset that takes as input the characteristic quantities of the detection values ​​detected by detectors during grinding and / or cleaning, including each processing step on the production line of the object or on a production line similar to the object's production line, and takes at least one of the following as output data: the number of defects, defect size, and defect location in the substrate. In this embodiment, an example of the output data of the learning dataset is the number of defects in the substrate. Here, the characteristic quantities are the average, maximum, minimum, total, median, standard deviation, dispersion, steepness, or skewness of the time series values ​​of the detector, or the average, maximum, minimum, total, median, standard deviation, dispersion, steepness, or skewness of the time series data of the differential values ​​of the detection values.

[0047] The processor 6 executes the specified program by reading it from the memory unit 7, thereby performing the functions of the unit dwell time statistics unit 61, transformation unit 62, learning unit 63, inference unit 64, regression analysis unit 65, and processing unit 66.

[0048] The transformation unit 62 transforms the detection values ​​detected by the aforementioned detector during grinding and / or cleaning and / or drying into feature quantities for each processing step of the machine learning model that has been learned.

[0049] The learning unit 63 uses a set of learning data containing the characteristic values ​​detected by the detector during grinding and / or cleaning, which include each processing step on the production line of the object or on the same type of production line as the object, as input data, and at least one of the number of defects, defect size, and defect location in the substrate as output data, so that the machine learning model 71 learns.

[0050] The inference unit 64 outputs a predicted value of at least one of the number of defects, defect size, and defect location in the object substrate by inputting object data containing the aforementioned feature quantities into the machine learning model that has been learned.

[0051] The regression analysis unit 65, according to a predetermined regression analysis algorithm, outputs correlation parameters (e.g., correlation coefficients) for each of the plurality of detection values, representing the correlation between these parameters and one of the defect count, defect size, or defect location in the substrate. Here, the regression analysis algorithm can also be LASSO (least absolute shrinkage and selection operator: LASSO) regression, Ridge regression, SVR (Support Vector Regression: SVR), RFR (random forest Regression: RFR), or Light GBM. The regression analysis unit 65 can also display these correlation parameters (e.g., correlation coefficients) on the display device 8. This allows the operator or user of the substrate processing apparatus 1 to confirm the correlation parameters (e.g., correlation coefficients).

[0052] The receiving unit 66, for example, receives at least one detector whose output is based on the feature quantity contained in the input data of the machine learning model, provided that the operator or user has confirmed the aforementioned output correlation parameters (e.g., correlation coefficients). At this time, the operator or user, for example, inputs or selects the detection value that is based on the feature quantity of the correlation coefficient. Then, the learning unit 63 uses the feature quantity of the detection value of the aforementioned received detector to relearn the aforementioned machine learning model 71. The inference unit 64 outputs the aforementioned prediction value using the machine learning model relearned by the learning unit 63. With this configuration, by having the receiving unit 66 receive the detector whose output is based on the feature quantity of the correlation coefficient, for example, provided that the operator or user outputs it, the prediction accuracy of the machine learning model 71 after relearning can be improved.

[0053] In one embodiment, the input data for the aforementioned machine learning model further includes the dwell time of each unit contained in the substrate processing device within that unit. Based on this, the unit dwell time statistics unit 61 counts the dwell time of each unit contained in the substrate processing device 1 within that unit. In another embodiment, the object data input to the aforementioned machine learning model after learning further includes the dwell time of each unit counted by the aforementioned unit dwell time statistics unit within that unit.

[0054] In one example of this embodiment, the input data during the learning of the aforementioned machine learning model further includes a second feature quantity that transforms the position of the component used for grinding or cleaning. Based on this premise, the transformation unit 62 transforms the position of the component used for grinding or cleaning into the second feature quantity. At this time, the aforementioned object data input into the machine learning model after the learning is completed further includes the second feature quantity of each component transformed by the transformation unit 62.

[0055] In one example of this embodiment, the input data during the learning of the aforementioned machine learning model further includes processing scheme information containing instruction values ​​for the units included in the substrate processing apparatus. These instruction values ​​are setting values ​​for the physical quantities of the object (e.g., the number of revolutions of the grinding table 23), and each physical quantity setting value is one value for each processing step. Based on this premise, the object data input into the machine learning model after the learning process is completed further includes processing scheme information containing instruction values ​​for the units included in the substrate processing apparatus.

[0056] Continuing with the explanation of the residence time of the substrate in each unit using FIG3. FIG3 is a graph showing an example of the residence time in each unit of this embodiment. In the graph of FIG3, the vertical axis represents the processing step number, and the horizontal axis represents time. FIG3 shows the residence time of each unit in the following process. That is, the transfer robot 12 delivers the substrate to the elevator 41, and then the transfer arm 45A of the first linear conveyor 42 takes the substrate out from the elevator 41. The transfer arm 45A delivers the substrate W to the grinding device 21A at the second transfer position TP2. After being ground by the grinding device 21A, the transfer arm 45D of the first linear conveyor 42 receives the substrate from the grinding device 21A and delivers the substrate W to the swing conveyor 44 at the fourth transfer position TP4. The swing conveyor 44 reverses the received substrate and loads it onto the temporary stage 47. Then, the lower arm of the transfer robot 35 receives the substrate W and transports it to the cleaning device 31A. After being cleaned by cleaning device 31A, the upper arm of transfer robot 35 removes substrate W from cleaning device 31A and transfers it to cleaning device 31B. After being cleaned by cleaning device 31B, the arm of transfer robot 36 removes substrate W from cleaning device 31B and transfers it to substrate drying device 32. After being dried by substrate drying device 32, transfer robot 12 removes the substrate from substrate drying device 32.

[0057] In Figure 3, TRBDs represent the dwell time of the substrate in the arm of the transfer robot 12 before the grinding process, TLFT represents the dwell time of the substrate in the elevator 41, TLTP1 represents the dwell time of the substrate in the transfer arm 45A of the first linear conveyor 42, and TPoliA represents the dwell time of the substrate in the grinding device 21A. The dwell time TPoliA of the substrate in the grinding device 21A includes waiting time in addition to the grinding process time. This waiting time can also be used as input data for the learning and inference of the machine learning model. The dwell time of the TLTP3 series substrate in the conveying arm 45D of the first linear conveyor 42, the dwell time of the TSTP series substrate in the swing conveyor 44, the dwell time of the TWS1 series substrate in the temporary stage, the dwell time of the TRB1L series substrate in the lower arm of the conveying robot 35, the dwell time of the TCL1A series substrate in the cleaning device 31A, the dwell time of the TRB1LU series substrate in the upper arm of the conveying robot 35, the dwell time of the TCL3A series substrate in the cleaning device 31B, the dwell time of the TRB3 series substrate in the arm of the conveying robot 36, the dwell time of the TCL4A series substrate in the substrate drying device 32, and the dwell time of the TRBDe series dried substrate in the arm of the conveying robot 12.

[0058] As shown in Figure 3, the grinding process has a plurality of processing steps, the first cleaning process also has a plurality of processing steps, the second cleaning process also has a plurality of processing steps, and the drying process also has a plurality of processing steps. These processing steps are assigned numbers in a manner that makes them identifiable. Here, these processing steps are designated as numbers 1 to N (i.e., processing steps are numbered 1 to N, where N is a natural number) and explained below.

[0059] Continuing, an example of the data construction of the feature quantities contained in the input data of the input machine learning model will be explained using Figures 4 and 5. Here, it is explained that there are M detectors (M is a natural number) from detector D1 to detector DM in the substrate processing device 1. Figure 4 shows an example of the data construction of the feature quantities contained in the input data of the input machine learning model. As shown in Figure 4, the feature quantity of the value of detector D1 is represented by an arrangement (or vector) of the feature quantities of processing steps 1 to N, respectively, as elements. Similarly, the feature quantity of the value of detector D2 is represented by an arrangement (or vector) of the feature quantities of processing steps 1 to N, respectively, as elements, and the feature quantity of the value of detector DM is represented by an arrangement (or vector) of the feature quantities of processing steps 1 to N, respectively, as elements. Thus, the feature quantity of the value of detector Di (i is an integer from 1 to M) contained in the input data of the input machine learning model is represented by an arrangement (or vector) of the feature quantities of processing steps 1 to N, respectively, as elements.

[0060] Here, the components included in the aforementioned substrate processing apparatus 1 are referred to as components U1 to UL (L being a natural number). The processor 6, for example, stores the positions of components 1 to L included in the substrate processing apparatus 1 in a time sequence in the memory unit 7. The positions of components U1 to UL can be calculated from the instruction signals of the unit control unit 51 indicating the operation of components U1 to UL using a predetermined conversion formula, or they can be determined from the correspondence between the instruction signals and known positions, or they can be the positions detected by the detector. As shown in FIG4, the feature quantity of the position of component U1 is represented by an arrangement (or vector) of the feature quantities of processing steps 1 to N, respectively, as elements. Similarly, the feature quantity of the position of component U2 is represented by an arrangement (or vector) of the feature quantities of processing steps 1 to N, respectively, as elements. Similarly, the feature quantity of the position of component UN is represented by an arrangement (or vector) of the feature quantities of processing steps 1 to N, respectively, as elements. Thus, the feature quantity of the position of component Uj (j is an integer from 1 to L) contained in the input data of the input machine learning model is represented by the arrangement (or vector) of the feature quantities of processing steps 1 to N respectively as elements.

[0061] Figure 5 shows an example of the data structure of the residence time of each unit in the input data of the input machine learning model. As shown in Figure 5, the residence time of each unit in the input data of the input machine learning model is represented by an arrangement (or vector) of the residence time of the substrate in each of the above units as elements.

[0062] Continuing, an example of the learning process and inference process of the machine learning model 71 will be explained using Figures 6A and 6B. Figure 6A is a schematic diagram illustrating an example of the learning process of the machine learning model. Figure 6B is a schematic diagram illustrating an example of the inference process of the machine learning model. As shown in Figure 6A, the learning process involves the machine learning model 71 learning using a set of learning data, which includes the arrangement of at least one feature quantity of each processing step described in Figure 4, processing scheme information, and the arrangement of the dwell time of each unit as input data, and the number of defects on the substrate as output data. As shown in Figure 6B, the inference process involves inputting the arrangement of at least one feature quantity of each processing step, processing scheme information, and the arrangement of the dwell time of each unit into the machine learning model 71 for the target substrate, and outputting the number of defects on the target substrate. Here, the feature quantity in the inference process is the same as the feature quantity in the learning process.

[0063] In addition, the arrangement of one or more feature quantities of each processing step, the processing scheme information, and the arrangement of the dwell time of each unit are used as input data for machine learning. However, it can be any one of them or any combination of two.

[0064] Figure 7 is an example of a graph comparing the measured value of the number of defects with the predicted value of the number of defects. In Figure 7, the vertical axis represents the predicted value of the number of defects, and the horizontal axis represents the measured value of the number of defects. The graphs for each grinding device 21A, 21B, 21C, and 21D are plotted by changing the density of the plot. The closer the distribution of the plotted groups is to the dashed line L1, the higher the prediction accuracy of the number of defects. As shown in Figure 7, the distribution of the plotted groups for any grinding device is close to the dashed line L1, and the prediction accuracy of the number of defects for any grinding device is high.

[0065] As described above, the substrate processing apparatus 1 of this embodiment includes: at least one detector that detects physical quantities of an object during the grinding and / or cleaning and / or drying of the substrate; a transformation unit 62 that transforms the detection values ​​detected by the aforementioned detector during grinding and / or cleaning and / or drying into feature quantities for each processing step of a learned machine learning model; and an inference unit 64 that outputs a predicted value of the number of defects in the substrate by inputting object data including the aforementioned feature quantities into the learned machine learning model 71. The learned machine learning model 71 is learned using a learning dataset that includes feature quantities containing the detection values ​​detected by the detector during grinding and / or cleaning on the production line of the object or on a production line of the same type as the production line of the object, as input data, and the number of defects in the substrate as output data.

[0066] When this configuration is adopted, since inspection is not performed using a defect inspection device, a predicted value for at least one of the defect number, defect size, and defect location of the substrate after substrate processing can be obtained. Therefore, it is possible to infer whether the substrate after substrate processing is a defective product without inspection using a defect inspection device. <Modified Example 1 of Machine Learning Model>

[0067] Continuing with Figures 8A and 8B, let's illustrate a modified example 1 of the learning and inference processes of the machine learning model 71. Figure 8A is a schematic diagram illustrating the modified example 1 of the learning process of the machine learning model. Figure 8B is a schematic diagram illustrating the modified example 1 of the inference process of the machine learning model. As shown in Figure 8A, the learning process involves the machine learning model 71 learning from a set of learning data, using the arrangement of at least one feature quantity of each processing step described in Figure 4, processing scheme information, and the arrangement of dwell time for each unit as input data, and the number of defects on the substrate and the defect size of the substrate as output data. As shown in Figure 8B, the inference process involves inputting the arrangement of at least one feature quantity of each processing step, processing scheme information, and the arrangement of dwell time for each unit into the machine learning model 71 for the target substrate, and outputting the number of defects on the target substrate and the defect size of the target substrate. The feature quantities in this inference process are the same as those in the learning process. <Modified Example 2 of the Machine Learning Model>

[0068] Continuing with Figures 9A and 9B, let's illustrate a modified example 2 of the learning and inference processes of the machine learning model 71. Figure 9A is a schematic diagram illustrating the modified example 2 of the learning process of the machine learning model. Figure 9B is a schematic diagram illustrating the modified example 2 of the inference process of the machine learning model. As shown in Figure 9A, the learning process involves the machine learning model 71 learning from a set of learning data, using the arrangement of at least one feature quantity of each processing step described in Figure 4, processing scheme information, and the arrangement of dwell time for each unit as input data, and using the number of defects on the substrate and the location of defects on the substrate as output data. As shown in Figure 9B, the inference process involves inputting the arrangement of at least one feature quantity of each processing step, processing scheme information, and the arrangement of dwell time for each unit into the machine learning model 71 for the target substrate, and outputting the number of defects on the target substrate and the location of defects on the target substrate. Here, the feature quantity in the inference process is the same as the feature quantity in the learning process. <Modified Example 3 of the Machine Learning Model>

[0069] Continuing with Figures 10A and 10B, let's illustrate a modified example 3 of the learning process and inference process of the machine learning model 71. Figure 10A is a schematic diagram illustrating the modified example 3 of the learning process of the machine learning model. Figure 10B is a schematic diagram illustrating the modified example 3 of the inference process of the machine learning model. As shown in Figure 10A, the learning process involves the machine learning model 71 learning from a set of learning data, which includes the arrangement of at least one feature quantity of each processing step described in Figure 4, processing scheme information, and the arrangement of the dwell time of each unit as input data, and the number of defects on the substrate, the defect size of the substrate, and the defect location of the substrate as output data. As shown in Figure 10B, the inference process involves inputting the arrangement of at least one feature quantity of each processing step, processing scheme information, and the arrangement of the dwell time of each unit into the machine learning model 71 for the target substrate, and outputting the number of defects on the target substrate, the defect size of the target substrate, and the defect location of the target substrate. Here, the characteristic quantity of the inferred process is the same as the characteristic quantity of the learned process.

[0070] Furthermore, not limited to the combination of output data in Modification Examples 1 to 3, the inference unit 64 can also output one or any two of the defect count, defect size, and defect location in the target substrate. Thus, by inputting the object data containing the aforementioned feature quantities into the machine learning model that has been learned, the inference unit 64 can also output a predicted value of at least one of the defect count, defect size, and defect location in the target substrate.

[0071] The substrate processing apparatus 1 of this embodiment includes a processor 6 and a memory unit 7, but is not limited thereto. <Modification 1 of this embodiment> FIG11A is a schematic configuration diagram of Modification 1 of this embodiment. As shown in FIG11A, it includes an information processing system S1 that can exchange information with the substrate processing apparatus 1 and is connected thereto. The information processing system S1 includes a processor 6 and a memory unit 7 that stores a machine learning model 71. The processor 6 executes the program by reading it from the memory unit 7, and can also perform the functions of a unit dwell time statistics unit 61, a transformation unit 62, a learning unit 63, an inference unit 64, a regression analysis unit 65, and a processing unit 66. <Modification 2 of this embodiment>

[0072] Figure 11B is a schematic diagram of Modification 2 of this embodiment. As shown in Figure 11B, an information processing system S1 is provided, which can exchange information with the board processing device 1 via a communication circuit network CN. The information processing system S1 includes: a processor 6 and a memory unit 7 storing a machine learning model 71. The processor 6 executes a specified program by reading it from the memory unit 7, and can also perform the functions of a unit dwell time statistics unit 61, a transformation unit 62, a learning unit 63, an inference unit 64, a regression analysis unit 65, and a processing unit 66. <Modification 3 of this embodiment>

[0073] Figure 12A is a schematic diagram of Modification Example 3 of this embodiment. As shown in Figure 12A, a server 70 is provided that can exchange information with the substrate processing device 1 via a communication circuit network CN. The substrate processing device 1 has a processor 6 and a memory unit 7a storing a specified program. The server 70 may also have a memory unit 7b storing a machine learning model 71. At this time, the processor 6 of the substrate processing device 1 executes the specified program by reading it from the memory unit 7a, and performs the functions of the unit dwell time statistics unit 61, the transformation unit 62, the learning unit 63, the inference unit 64, the regression analysis unit 65, and the processing unit 66. The inference unit 64 can output a predicted value of at least one of the defect number, defect size, and defect location in the object substrate by inputting object data containing the aforementioned feature quantities into the machine learning model 71 that has been learned by the server 70. <Modification Example 4 of this embodiment>

[0074] Figure 12B is a schematic diagram of a modified example 4 of this embodiment. As shown in Figure 12B, it may also be configured to include a substrate processing device 1 and an information processing system S3. The information processing system S3 may also include: an information processing device 60 that can exchange information with the substrate processing device 1 and a server 70 that can exchange information with the information processing device 60 via a communication circuit network CN. In this case, the information processing device 60 has: a processor 6 and a memory unit 7a that stores a specified program, and the server 70 may also have a memory unit 7b that stores a machine learning model 71. Here, the processor 6 of the information processing device 60 executes a specified program read from the memory unit 7a, thereby performing the functions of the unit dwell time statistics unit 61, transformation unit 62, learning unit 63, inference unit 64, regression analysis unit 65, and processing unit 66. The inference unit 64 can also output a predicted value of at least one of the defect number, defect size, and defect location in the object substrate by inputting object data containing the aforementioned feature quantities into the machine learning model 71 that has been learned by the server 70.

[0075] Thus, the information processing systems S1, S2, and S3 may also include: a transformation unit 62 that transforms the detection values ​​of the substrate during grinding and / or cleaning and / or drying detected by the detectors provided in the substrate processing apparatus 1 into feature quantities for each processing step of the learned machine learning model; and an inference unit 64 that outputs a predicted value of at least one of the number of defects, defect size, and defect location in the substrate by inputting object data containing the aforementioned feature quantities. Here, the learned machine learning model is learned using a learning dataset that includes feature quantities of the detection values ​​during grinding and / or cleaning detected by the detectors on the production line of the object or on a production line of the same type as the production line of the object as input data, and at least one of the number of defects, defect size, and defect location in the substrate as output data.

[0076] When this configuration is adopted, since the defect inspection device is not used for inspection, the predicted value of at least one of the defect number, defect size and defect location of the substrate after substrate processing can be obtained. Therefore, it can be inferred whether the substrate after substrate processing is a defective product without inspection by the defect inspection device.

[0077] Furthermore, at least a portion of the functions of the information processing systems S1, S2, S3, or processor 6 described in the above embodiments can be implemented by hardware or software. In the case of software implementation, the program implementing at least a portion of the functions of the information processing systems S1, S2, S3, or processor 6 can be stored in a computer-readable recording medium for the computer to read and execute. The recording medium is not limited to removable media such as magnetic disks or optical discs, but can also be a fixed recording medium such as a hard disk drive or memory.

[0078] In addition, the program that implements at least a portion of the functions of the information processing system S1, S2, S3 or the processor 6 may also be distributed via communication lines such as the Internet (including wireless communication). Furthermore, the program may also be encrypted or modulated and distributed in a compressed state via wired or wireless lines such as the Internet, or stored on a recording medium.

[0079] Furthermore, the information processing systems S1, S2, and S3 can also function using one or more information devices. When using multiple information devices, one of them can be used as a computer, and by executing a specified program through that computer, it can also function as at least one means of the information processing systems S1, S2, and S3.

[0080] As described above, the present invention is not limited to the embodiments described above. The invention can be further embodied by modifying the constituent elements without departing from its essence. Furthermore, various inventions can be formed by appropriately combining the plurality of constituent elements disclosed in the above embodiments. For example, some constituent elements can be deleted from all the constituent elements shown in the embodiments. Moreover, constituent elements from different embodiments can be appropriately combined. [Simplified Explanation of the Diagram]

[0015] Figure 1 is a diagram showing an example of the schematic configuration of the substrate processing apparatus of this embodiment. Figure 2 is a block diagram showing an example of the schematic configuration of the control unit of this embodiment. Figure 3 is a graph showing an example of the dwell time in each unit of this embodiment. Figure 4 is an example of the data structure of the feature quantities contained in the input data of the machine learning model. Figure 5 is an example of the data structure of the dwell time of each unit contained in the input data of the machine learning model. Figure 6A is a schematic diagram illustrating an example of the learning process of the machine learning model. Figure 6B is a schematic diagram illustrating an example of the inference process of the machine learning model. Figure 7 is an example of a graph comparing the measured value of the number of defects with the predicted value of the number of defects. Figure 8A is a schematic diagram illustrating a modified example 1 of the learning process of the machine learning model. Figure 8B is a schematic diagram illustrating a modified example 1 of the inference process of the machine learning model. Figure 9A is a schematic diagram illustrating a modified example 2 of the learning process of the machine learning model. Figure 9B is a schematic diagram illustrating Modified Example 2 of the inference process of the machine learning model. Figure 10A is a schematic diagram illustrating Modified Example 3 of the learning process of the machine learning model. Figure 10B is a schematic diagram illustrating Modified Example 3 of the inference process of the machine learning model. Figure 11A is a schematic diagram of Modified Example 1 of this embodiment. Figure 11B is a schematic diagram of Modified Example 2 of this embodiment. Figure 12A is a schematic diagram of Modified Example 3 of this embodiment. Figure 12B is a schematic diagram of Modified Example 4 of this embodiment.

Claims

1. A substrate processing apparatus comprising: at least one detector for detecting physical quantities of an object during grinding and / or cleaning and / or drying of the substrate; a transformation unit for transforming the detection values ​​during grinding and / or cleaning and / or drying detected by the detector into feature quantities for each processing step using a learned machine learning model; and an inference unit for inputting object data including the aforementioned feature quantities into the learned machine learning model and outputting a predicted value of at least one of the number of defects, defect size, and defect location in the substrate of the object; wherein the learned machine learning model is learned using a learning dataset containing feature quantities including the detection values ​​during grinding and / or cleaning detected by the detector on the production line of the object or on a production line of the same type as the production line of the object, as input data, and at least one of the number of defects, defect size, and defect location in the substrate as output data.

2. The substrate processing apparatus of claim 1, wherein the input data of the aforementioned machine learning model during learning further includes the dwell time of each unit contained in the substrate processing apparatus, and each unit contained in the substrate processing apparatus further includes a unit dwell time statistics unit that systematically counts the dwell time in the unit, and the aforementioned object data input into the aforementioned machine learning model after learning is completed further includes the dwell time of each unit counted by the aforementioned unit dwell time statistics unit.

3. The substrate processing apparatus of claim 1 or 2, wherein the input data of the aforementioned machine learning model during learning further includes a second feature quantity that transforms the position of the component used for grinding or cleaning, the aforementioned transformation unit transforms the position of the component used for grinding or cleaning into the second feature quantity, and the aforementioned object data input into the aforementioned machine learning model after learning further includes the second feature quantity of each component transformed by the aforementioned transformation unit.

4. The substrate processing apparatus of claim 1 or 2, wherein the input data of the aforementioned machine learning model during learning further includes processing recipe information, which includes instruction values ​​for the units contained in the substrate processing apparatus, and the aforementioned object data input to the aforementioned machine learning model after learning further includes processing recipe information, which includes further instruction values ​​for the units contained in the substrate processing apparatus.

5. The substrate processing apparatus of claim 1 or 2, comprising: a regression analysis unit that outputs correlation parameters relating to one of the following: the number of defects, the size of defects, and the location of defects in the substrate, respectively, according to a predetermined regression analysis algorithm; a receiving unit that receives at least one detector whose output is based on the feature quantity contained in the input data of the machine learning model; and a learning unit that relearns the machine learning model by changing the feature quantity of the detection value of the received detector; and an inference unit that outputs the predicted value using the machine learning model relearned by the learning unit.

6. An information processing system comprising: a transformation unit that transforms, for each processing step, the detection values ​​detected by a detector of a substrate processing apparatus during polishing and / or cleaning and / or drying of the substrate into feature quantities; and an inference unit that, by inputting object data including the aforementioned feature quantities, outputs a predicted value of at least one of the number of defects, defect size, and defect location in the substrate; wherein the aforementioned learned machine learning model is learned using a learning dataset containing feature quantities of the detection values ​​detected by a detector during polishing and / or cleaning of the substrate during each processing step on a production line of the object or a production line of the same type as the object, as input data, and at least one of the number of defects, defect size, and defect location in the substrate as output data.

Images