Method for identifying wafer area and wafer area identification device for executing method

Identifying the wafer area through optical sensors and neural network models solves the problem of distinguishing chip-free and chip-free areas in the wafer in the prior art, achieving more accurate thickness measurements, and reducing chip defect rate.

CN118936332APending Publication Date: 2024-11-12KC TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311340698.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2023-05-09
Filing Date
2023-10-17
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The prior art is difficult to accurately distinguish the area with chips and areas without chips in the wafer, resulting in inaccurate measurement of wafer thickness, which in turn affects the chip's defect rate.

Method used

Using an area identification model based on optical sensors and neural networks, multiple optical spectral signals are obtained on the wafer through probes, and the modular area, modular area, wafer ring area and air area are identified and distinguished. These signals are used to infer crystals. Round film thickness.

Benefits of technology

It improves the accuracy of wafer area identification and the accuracy of thickness measurement, and reduces the chip defect rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118936332A_ABST
    Figure CN118936332A_ABST
Patent Text Reader

Abstract

The invention discloses a method for identifying a wafer area and a wafer area identification device for executing the method. The wafer area identification device comprises a probe which comprises an optical sensor and is used for acquiring optical spectrum signals of which signal values are changed according to the thickness characteristic of a wafer at a plurality of positions on the wafer; and the processor is used for identifying whether the current optical spectrum signal is obtained from the module area of the chip on the wafer or not by using a neural network-based area identification model which takes the current optical spectrum signal obtained by the probe as input.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to techniques for identifying wafer regions and techniques for inferring wafer thickness. Background Art

[0002] Generally speaking, the Chemical Mechanical Polishing (CMP) process is a standard process for grinding the surface of a wafer by relative rotation between a wafer with a grinding layer and a grinding plate for semiconductor manufacturing. The grinding device for performing the CMP process includes: a grinding plate, on which a grinding pad is attached; a grinding head, on which a wafer to be ground is mounted, which contacts and rotates with the top of the grinding pad; and an adjuster, which presses the surface of the grinding pad with a predetermined pressure while performing fine cutting, so as to adjust so that the micropores formed on the surface of the grinding pad appear on the surface. In order to determine the relevant grinding parameters applied to the wafer grinding process, it is crucial to accurately measure or infer the current thickness of the wafer. Summary of the invention

[0003] According to one embodiment, a wafer region identification device may include: a probe, which includes an optical sensor, which obtains optical spectral signals whose signal values ​​change according to the thickness characteristics of the wafer at multiple positions on the wafer; and a processor, which uses a neural network-based region identification model that uses the current optical spectral signal obtained by the probe as input to identify whether the current optical spectral signal is obtained from a die area where a chip is configured on the wafer.

[0004] The processor may use the region discrimination model to identify whether the current optical spectrum signal is obtained from a non-die region on a wafer where no chips are configured.

[0005] The mold-free area may include at least one of an edge area of ​​the wafer where no chips are arranged and a scribeline area between mold areas.

[0006] The processor may use the region discrimination model to identify whether the current optical spectrum signal is obtained from a wafer ring frame region for fixing the wafer.

[0007] The processor may use the region discrimination model to identify whether the current optical spectrum signal is obtained from an air region outside the wafer ring stand.

[0008] When the processor recognizes that a current optical spectrum signal is obtained from the die area, the processor can infer the film thickness of the wafer based on the current optical spectrum signal.

[0009] The processor may infer the film thickness of the wafer using a neural network-based wafer thickness inference model that takes the current optical spectral signal as input.

[0010] The processor may utilize the plurality of optical spectral signals and the region identification model to distinguish the types of regions corresponding to respective locations of the wafer.

[0011] The optical spectral signal obtained by the probe may include signals obtained by reflecting light of multiple wavelength bands on the wafer.

[0012] The region discrimination model can distinguish a first signal region corresponding to the pattern region and a second signal region corresponding to the pattern-free region from a signal of a specific wavelength extracted from an optical spectrum signal obtained according to the movement trajectory of the probe.

[0013] The region identification model may be a region identification model based on a convolutional neural network.

[0014] According to one embodiment, a method for identifying a wafer area may include the following operations: obtaining, through a probe at multiple positions on the wafer, an optical spectral signal whose signal value changes according to the thickness characteristics of the wafer; and using a neural network-based area identification model that uses the current optical spectral signal obtained by the probe as input to identify whether the current optical spectral signal is obtained from a die area of ​​a chip configured on the wafer.

[0015] The identifying operation may include the following operation: using a region discrimination model to identify whether the current optical spectrum signal is obtained from a non-pattern region on a wafer where no chips are configured.

[0016] The identifying operation may include the following operation: using a region discrimination model to identify whether the current optical spectrum signal is obtained from a wafer ring stand region for fixing the wafer.

[0017] The identifying operation may include the following operation: using a region discrimination model to identify whether the current optical spectrum signal is obtained from an air region outside the wafer ring stand.

[0018] The method may further include the following operation: in a case where it is identified that the current optical spectrum signal is obtained from the die region, inferring a film thickness of the wafer based on the current optical spectrum signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a diagram for providing an overview of a process for identifying wafer regions according to one embodiment.

[0020] Figure 2 FIG. 1 is a diagram showing a configuration of a wafer region identification device according to an embodiment.

[0021] Figure 3 is a diagram for explaining an example of an optical spectrum signal according to a measurement position of a probe according to one embodiment.

[0022] Figure 4 is a diagram illustrating a signal of a specific wavelength band extracted from an optical spectrum signal according to one embodiment.

[0023] Figure 5 and Figure 6 is a diagram for explaining changes in an optical spectrum signal according to a measurement position of a probe according to one embodiment.

[0024] Figure 7a , Figure 7b and Figure 7c is a diagram showing an example of an optical spectroscopy signal obtained by a probe according to one embodiment.

[0025] Figure 8 is a diagram showing the structure of a region discrimination model according to one embodiment.

[0026] Fig. 9 is a flow chart illustrating the operation of a method of identifying wafer regions according to one embodiment.

[0027] Fig.10 is a flow chart illustrating the operation of a learning method for causing a region discrimination model to learn according to one embodiment.

[0028] Fig.11 and Fig.12 is a diagram for explaining learning data used in region discrimination model learning according to one embodiment. DETAILED DESCRIPTION

[0029] Hereinafter, the embodiments are described in detail with reference to the accompanying drawings. However, various changes can be made to the embodiments, so the scope of the patent application is not limited or restricted by these embodiments. It should be understood that all changes, equivalents and even substitutes made to the embodiments are included in the scope of the rights.

[0030] The terms used in the embodiments are used only for the purpose of illustration and should not be interpreted as intended to be limited. If the context does not have a clear different meaning, then the expression of the singular includes the expression of the plural. In this specification, the terms "including" or "having" should be understood to specify the existence of the features, numbers, steps, actions, constituent elements, parts or these combinations recorded in the specification, and do not preclude the existence or additional possibility of one or more other features or numbers, steps, actions, constituent elements, parts or these combinations.

[0031] If there is no other definition, all terms used herein, including technical or scientific terms, have the same meaning as those generally understood by a technician with general knowledge in the technical field to which the embodiment belongs. Terms such as those defined in generally used dictionaries should be interpreted as having the same meaning as that in the context of the relevant technology, and if not clearly defined in this application, they should not be interpreted as ideal or overly formal meanings.

[0032] In addition, when describing with reference to the accompanying drawings, the same components are given the same reference numerals regardless of the figure numbers, and repeated descriptions thereof are omitted. When describing the embodiments, if it is determined that the specific description of the related known technology may unnecessarily confuse the gist of the embodiments, the detailed description thereof is omitted.

[0033] In addition, when describing the constituent elements of the embodiments, terms such as first, second, A, B, (a), (b) and the like may be used. These terms are only used to distinguish the constituent element from other constituent elements, and the nature, order or sequence of the corresponding constituent elements are not limited by the terms. When a constituent element is described as being "connected", "combined" or "joined" to other constituent elements, it should be understood that the constituent element may be directly connected or joined to other constituent elements, but other constituent elements may also be "connected", "combined" or "joined" between the constituent elements.

[0034] The term "module" used in various embodiments herein may include a unit implemented by hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit. A module may be a component that is integrally constructed or the smallest unit or a portion thereof of the above components that performs one or more functions. For example, according to one embodiment, a module may be implemented in the form of an ASIC (application-specific integrated circuit).

[0035] The constituent elements included in one embodiment and the constituent elements having common functions are described in other embodiments using the same names. Unless otherwise stated, the description described in one embodiment may also be applied to other embodiments, and the specific description is omitted within the scope of repetition.

[0036] Recently, the size of the chip in the wafer has gradually increased. As the chip size increases, the part of the outer area of ​​the wafer without the chip gradually becomes wider. If the part with the chip and the part without the chip in the wafer area cannot be accurately distinguished, the wafer film thickness measurement based on the optical sensor may be inaccurate, and the inaccurate thickness measurement will lead to an increase in the chip defect rate in the outer area of ​​the wafer. Therefore, it is necessary to distinguish between the area (such as the die area) where the structure (such as the chip) is set in the wafer and the area (non-die area) where the structure is not set. High-level identification technology of the wafer area. If the wafer area and the area with the chip inside the wafer are accurately distinguished, the film thickness of the wafer can be more accurately measured. According to one or more embodiments described in this specification, a technical means is provided that can more accurately measure the wafer area and the area with the chip inside the wafer.

[0037] Figure 1 is a diagram for providing an overview of a process for identifying wafer regions according to one embodiment.

[0038] Reference Figure 1 , the wafer region identification device 100 can identify (distinguish or recognize) the region on the wafer 110 based on one or more optical spectral signals 130 (or spectral data) measured by a probe (not shown). In one embodiment, the wafer region identification device 100 can use a region identification model based on machine learning to determine the structure region on the wafer 110 during the chemical mechanical polishing (CMP) process of the wafer 110. In one embodiment, the wafer region identification device 100 may include a probe. According to an embodiment, the probe may also exist outside the wafer region identification device 100. The wafer region identification device 100 can obtain one or more optical spectral signals 130 in real time during the polishing of the wafer 110 through an optical device such as a probe, and more accurately identify the region in the wafer 110 whose thickness is to be measured based on the one or more optical spectral signals 130 obtained.

[0039] The probe including the optical sensor may obtain the optical spectrum signal 130 while moving along the specific moving track 120. The optical sensor may measure the reflected light returned after the light output from the light source of the probe and the output light is reflected on the medium. The optical spectrum signal may be measured by the probe while the wafer 110 is rotating. The optical spectrum signal 130 obtained by the probe may include an optical spectrum signal whose signal value changes according to the thickness characteristics of the wafer 110. For example, the probe may obtain the optical spectrum signal 130 at a plurality of measurement positions 124 within the interval from the starting point 122 to the end point 126 while moving along the moving track 120. For example, the optical spectrum signal 132 may be obtained at the measurement position 124 by the optical sensor included in the probe, and the optical spectrum signal corresponding to each corresponding sampling position may be obtained at the sampling position reached later. The measurement and collection of the optical spectrum signal is started with the starting point 122 as the reference, and the measurement and collection of the optical spectrum signal is ended when the end point 126 is reached.

[0040] The wafer 110 corresponds to the wafer region, and a wafer ring frame region 115 (or R-ring region) for fixing (or clamping) the wafer may exist around the wafer 110. In one embodiment, the probe can obtain not only optical spectrum signals of the inner region of the wafer 110, but also optical spectrum signals of the wafer ring frame region 115 and the outer region of the wafer ring frame region 115 (e.g., air region).

[0041] One or more optical spectral signals 130 obtained by the probe are transmitted to the wafer region identification device 100, and the wafer region identification device 100 can identify the wafer region based on the received one or more optical spectral signals 130. In one embodiment, the wafer region identification device 100 can use a neural network-based region identification model that uses the current optical spectral signal obtained by the probe as input to identify whether the current optical spectral signal is obtained from the die area of ​​the chip configured on the wafer. The region identification model can be, for example, a region identification model based on a convolutional neural network, but is not limited thereto. The region identification model can be generated by a learning process, and the learning process for the region identification model is performed in Figures 10 to 12 Detailed description in.

[0042] In one embodiment, the wafer region identification device 100 can use the region identification model to identify whether the current optical spectrum signal is obtained from the non-mold area on the wafer where the chip is not configured. The non-mold area may include at least one of the edge area of ​​the wafer where the chip is not configured and the scribe line area between the die areas. In addition, the wafer region identification device 100 can use the region identification model to identify whether the current optical spectrum signal is obtained from the wafer ring frame area 115 for fixing the wafer or the air area outside the wafer ring frame.

[0043] Through the above operation, the wafer region identification device 100 can more accurately distinguish the internal / external area of ​​the wafer 110 and the internal mold area / non-mold area of ​​the wafer 110. Accordingly, the boundary between the area of ​​the wafer 100 and other areas can be accurately grasped, and the area where the thickness of the wafer 110 can be measured can be expanded. By improving the quality of data used to measure the thickness of the wafer 110, the measurement accuracy of the thickness of the wafer 110 can be improved.

[0044] Figure 2 FIG. 1 is a diagram showing a configuration of a wafer region identification device according to an embodiment.

[0045] Reference Figure 2 The wafer region identification device 100 may include a probe 210, a processor 220, and a memory 230. In some embodiments, at least one of these components (such as the probe 210) may be omitted from the wafer region identification device 100, or one or more other components (such as the wafer thickness estimation device 240, the storage module) may be added.

[0046] The probe 210 includes an optical sensor, and an optical spectrum signal whose signal value changes according to the thickness characteristics of the wafer can be obtained at multiple positions on the wafer. While the probe 210 moves along a specified moving track, the optical sensor can measure the reflected light returned after the light output from the light source is reflected to obtain the optical spectrum signal. The optical spectrum signal obtained by the probe 210 may include signals obtained after light of multiple bands is reflected on the wafer.

[0047] The memory 230 stores information required when the processor 220 performs processing operations. For example, the memory 230 may store commands executable by the processor 220, optical spectrum signals obtained by the probe 210, sensor data of other sensors, etc. The memory 230 may include volatile memories such as RAM, DRAM, SRAM, and / or non-volatile memories such as flash memory that are well known in the art.

[0048] The processor 220 may control the overall operation of the wafer region identification device 100. The processor 220 may be composed of one or more processors. The processor 220 may include, for example, a general-purpose processor such as a CPU (central processing unit), an AP (application processor), a DSP (digital signal processor), or an NPU (neural network processing unit).

[0049] The processor 220 can execute the commands stored in the memory 230 and perform the operations of the wafer region identification device 100 described in this specification. In one embodiment, when the commands executable by the processor 220 stored in the memory 230 are executed by the processor 220, the processor 220 can use multiple optical spectral signals and a region identification model to distinguish the types of regions corresponding to each position of the wafer. The region identification model can be, for example, a region identification model based on a convolutional neural network.

[0050] In one embodiment, the processor 220 may use a neural network-based region discrimination model that takes the current optical spectrum signal obtained by the probe as input to identify whether the current optical spectrum signal is obtained from a die region where a chip is configured on a wafer.

[0051] In one embodiment, the processor 220 may use a regional identification model to identify whether the current optical spectrum signal is obtained from a non-mode area on the wafer where no chips are configured. The non-mode area may include at least one of an edge area of ​​the wafer where no chips are configured and a scribe line area between the die areas. In one embodiment, the regional identification model may distinguish a first signal area corresponding to the die area and a second signal area corresponding to the non-mode area in a signal of a specific wavelength extracted from an optical spectrum signal obtained according to a moving trajectory of the probe.

[0052] In one embodiment, the processor 220 may use a regional identification model to identify whether the current optical spectrum signal is obtained from a wafer ring frame region (a region where the wafer ring frame is located) for fixing the wafer. In addition, the processor 220 may use a regional identification model to identify whether the current optical spectrum signal is obtained from an air region outside the wafer ring frame.

[0053] In one embodiment, the optical spectrum signal obtained from the die area is transmitted to a device for measuring the film thickness of the wafer, that is, the wafer thickness inference device 240, and the wafer thickness inference device 240 can infer the thickness of the wafer based on the received optical spectrum signal. According to the embodiment, the wafer area identification device 100 performs the function of the wafer thickness inference device 240, and the wafer area identification device 100 can also infer the thickness of the wafer based on the optical spectrum signal obtained from the die area. For example, when the processor 220 recognizes that the current optical spectrum signal is obtained from the die area, it can infer the film thickness of the wafer based on the current optical spectrum signal. At this time, the processor 220 can use a wafer thickness inference model based on a neural network that uses the current optical spectrum signal as input to infer the film thickness of the wafer. The film thickness inference of the wafer can be based on the die area of ​​the wafer instead of the die-free area, thereby improving the accuracy of the inference of the film thickness of the wafer. The higher the accuracy of the inference of the film thickness, the more accurately and precisely the speed of cutting the film according to the area of ​​the wafer can be adjusted, and the defect rate of the chip located in the outer part of the wafer can be reduced.

[0054] Figure 3 is a diagram for explaining an example of an optical spectrum signal according to a measurement position of a probe according to one embodiment.

[0055] Reference Figure 3 , the internal area of ​​the wafer 310 can be divided into a die area 312 where chips are configured and a non-die area where chips are not configured. The non-die area may include an edge area 314 in the wafer 310 where chips are not configured and a scribe line area 316 between the die areas 312. The die area 312 may include, for example, a plurality of chips arranged in rows and columns. The die areas 312 are separated from each other by the scribe line areas 316, and individual chips in the wafer 310 may include circuits. The dies have a chip shape before being cut from the wafer 310, and can be separated from each other by cutting.

[0056] The optical spectrum signal can be obtained by the optical sensor while the probe moves along the moving trajectory 320 that crosses the area on the wafer 310. For example, an optical spectrum signal 332 can be obtained from a measurement position 322 located on the moving trajectory 320, and an optical spectrum signal 334 can be obtained from a measurement position 324. The optical spectrum signal 332 is measured in a mold-free area of ​​the wafer 310, and the optical spectrum signal 334 is measured in a mold area of ​​the wafer 310. In the optical spectrum signal 332 and the optical spectrum signal 334, the x-axis represents the wavelength length of the signal, and the y-axis represents the intensity of the signal. The optical spectrum signal measured in the mold-free area (for example: the outer area of ​​the wafer without a chip and the scribe groove area) may act as unnecessary noise when measuring the thickness of the wafer 310.

[0057] Figure 4 is a diagram illustrating a signal of a specific wavelength band extracted from an optical spectrum signal according to one embodiment.

[0058] Reference Figure 4 , shows an example of an optical spectrum signal 410 obtained by a probe. The intensity of light transmitted to the optical sensor included in the probe is correlated with the thickness of the wafer and the material reflected by the light. For example, when the material of the wafer film changes, the optical properties of the reflected light according to the thickness may change accordingly. The optical spectrum signal 410 may have different spectral signal forms depending on the characteristics of the area where the light is reflected. Graph 422 shows the signal component of the first wavelength length (e.g., 551.51 nm) in the optical spectrum signal 410, and graph 424 shows the signal component of the second wavelength length (e.g., 454.49 nm) in the optical spectrum signal 410. In graphs 422 and 424, the x-axis represents time and the y-axis represents the intensity of the signal.

[0059] Figure 5 and Figure 6 is a diagram for explaining changes in an optical spectrum signal according to a measurement position of a probe according to one embodiment.

[0060] Reference Figure 5 , the probe 520 can obtain an optical spectrum signal through the optical sensor in the interval between the starting point 542 and the end point 544 while moving along the moving track 530. For example, the probe 520 can obtain an optical spectrum signal at each sampling position at a certain time interval (or a certain distance interval) in the interval between the starting point 542 and the end point 544. The probe 520 rotates multiple times along the moving track 530 to obtain the optical spectrum signal.

[0061] In one embodiment, the starting point 542 on the moving trajectory 530 may start from the air region. The sampling position is performed according to the air region -> wafer ring frame region 515 -> wafer region 510 -> wafer ring frame region 515 -> air region as time passes, and when reaching the end point 544, the acquisition of the optical spectrum signal may be stopped.

[0062] In chart 550, the intensity variation of the optical spectrum signal measured at the sampling position over time from the starting point 542 to the end point 544 is shown. It can be seen from chart 550 that the intensity of the optical spectrum signal in the air region is constant, and after the wafer ring frame region 515 intensity increases, the intensity in the wafer region 510 further becomes larger. In addition, it can be seen that between the wafer ring frame region 515 and the wafer region 510, the signal intensity of the optical spectrum signal changes sharply. Based on each regional characteristic of such optical spectrum signal, the region inside / outside the wafer can be effectively distinguished. The regional discrimination model described in this manual is a model for learning each regional characteristic of this optical spectrum signal. The reflectivity or absorptivity of light are different according to the surface state of the wafer, and therefore, the shape or form of the optical spectrum signal on the wafer surface measured by the probe can be compared, and the region inside / outside the wafer and the die area / no-die area of ​​the wafer can be effectively distinguished.

[0063] Reference Figure 6 , a graph 610 is shown, which shows a signal component of a specific wavelength length (e.g., 551.51 nm) in an optical spectrum signal obtained by a probe. The signal component may have a periodicity according to a movement period (or rotation period) of the probe. Observing a graph of a specific area 620 in which a corresponding signal component is enlarged, the signal component may include a signal component 632 measured in an air area, a signal component 634 measured in a wafer ring frame area, a signal component 636 measured in a mold-free area in an inner area of ​​the wafer, and a signal component 638 measured in a mold area in an inner area of ​​the wafer.

[0064] Figure 7a , Figure 7b and Figure 7c is a diagram showing an example of an optical spectroscopy signal obtained by a probe according to one embodiment.

[0065] Figure 7a Examples of optical spectral signals measured in the air region or the wafer ring stand region are shown. Figure 7b shows an example of an optical spectral signal measured in a mode-free region in the inner region of a wafer, Figure 7c An example of an optical spectral signal measured in a die area in the inner region of a wafer is shown. The forms of the optical spectral signals measured in the air region (or wafer ring frame region), the die-free region, and the die area are clearly distinguished from each other, so the inner / outer regions of the wafer and the die area / die-free region can be distinguished based on the forms of the optical spectral signals.

[0066] Figure 8 is a diagram showing the structure of a region discrimination model according to one embodiment.

[0067] Reference Figure 8According to one embodiment, the regional identification model 820 can be based on a convolutional neural network including a convolutional layer 822 and a hidden layer and an output layer 824. The optical spectrum signal 810 obtained by the probe is input to the regional identification model 820, and the convolutional layer 22 can extract features from the input optical spectrum signal 810. The hidden layer and the output layer 824 can identify the region corresponding to the optical spectrum signal based on the features extracted by the convolutional layer 822. The hidden layer can be, for example, a fully connected layer. The hidden layer is connected to the output layer, and the convolutional layer 822 and the hidden layer and the output layer 824 can respectively include multiple artificial neurons (or nodes). The structure of the regional identification model 820 can be variously deformed. In other embodiments, multiple optical spectrum signals obtained by the probe can also be input into the regional identification model 820 at one time.

[0068] The regional identification model 820 can output the probability value and score of the currently input optical spectrum signal corresponding to the optical spectrum signal measured in the die area. As another example, the regional identification model 820 can also output the probability value and score of the currently input optical spectrum signal corresponding to the air area, the wafer ring frame area, the mold-free area and the die area respectively.

[0069] Fig. 9 1 is a flow chart showing the operation of a method for identifying a wafer region according to one embodiment. The method according to one embodiment may be performed by a wafer region identification device (e.g., Figure 1 and Figure 2 The wafer area identification device 100 is used to perform.

[0070] Reference Fig. 9 In operation 910, the wafer region identification device can obtain an optical spectrum signal whose signal value changes according to the thickness characteristics of the wafer by using a probe at multiple positions on the wafer. The probe measures the reflected light returned after the light emitted by the light source is reflected on a medium (e.g., air, a wafer ring frame, or a wafer), thereby obtaining the optical spectrum signal.

[0071] In operation 920, the wafer region identification device may use a neural network-based region identification model that uses the current optical spectrum signal obtained by the probe as an input to identify the wafer region. In one embodiment, the wafer region identification device may use the region identification model to identify whether the current optical spectrum signal is obtained from the mold area where the chip is configured on the wafer. In addition, the wafer region identification device may use the region identification model to identify whether the current optical spectrum signal is obtained from the moldless area where the chip is not configured on the wafer. The wafer region identification device may use the region identification model to identify whether the current optical spectrum signal is obtained from the wafer ring frame area for fixing the wafer. In addition, the wafer region identification device may use the region identification model to identify whether the current optical spectrum signal is obtained from the air area outside the wafer ring frame.

[0072] In operation 930, the wafer region identification device may also perform an operation of inferring the film thickness of the wafer based on the optical spectrum signal. When the wafer region identification device recognizes that the current optical spectrum signal is obtained from the die area, the wafer region identification device may infer the film thickness of the wafer based on the current optical spectrum signal. In one embodiment, the wafer region identification device may infer the film thickness using a wafer thickness inference model based on a neural network, wherein the wafer thickness inference model based on the neural network may be learned by inputting an optical spectrum signal and outputting an inferred value of the film thickness of the wafer. The wafer thickness inference model may be generated through a learning process. In the corresponding learning process, the following process may be repeated: inputting an optical spectrum signal corresponding to a specific film thickness into the wafer thickness inference model, adjusting the parameters constituting the wafer thickness inference model (e.g., connection weighting value, deviation) so that the film thickness inference value output by the wafer thickness inference model is close to the actual film thickness value. By performing the learning process with a plurality of film thicknesses of the wafer as the object, a wafer thickness inference model may be generated.

[0073] Fig.10 1 is a flow chart showing the operation of a learning method for learning a region identification model according to one embodiment. The learning method may be performed by a learning device including a processor and a memory.

[0074] Reference Fig.10 In operation 1010, the learning device may select a wavelength having a large characteristic (eg, intensity difference) of a wafer region from among the light of multiple wavelength bands.

[0075] In operation 1020, the learning device may generate learning data based on the selected wavelength. In one embodiment, the learning data may be generated by a probability density function and / or a differential function. The learning device may apply the selected wavelength to the probability density function or the differential function to generate the learning data. The optical spectrum data of the region where the thickness of the wafer is measured at an internal position of the wafer region, i.e., the die region, may be generated as learning data. Based on the waveform of the generated learning data, the function parameters of the probability density function and / or the differential function for generating the learning data may vary.

[0076] The core of the algorithm for generating learning data is to use data that exceeds the specified limit range. According to an embodiment, the algorithm for generating learning data may include: 1) a technique for tracking the trajectory of the probe using hardware parameters (e.g., the angular velocity of the probe, the installation position of the probe, etc.) to generate learning data; 2) a technique for installing a separate data measurement device under the wafer to obtain learning data; 3) a technique for generating learning data using previously acquired optical data while the wafer is in a waiting state; 4) a technique for generating learning data using a function with light intensity differences as a variable; 5) a technique for generating learning data using a probability density function, etc.

[0077] In operation 1030, the learning device may cause the region identification model to learn based on the generated learning data. The learning device inputs the learning data to the region identification model, defines the difference between the output of the region identification model and the actual desired output through a loss function, and updates the parameters of the region identification model (e.g., connection weights and biases between artificial neurons) in a form that reduces the loss value of the loss function. In the process of updating these parameters, an error backpropagation technique may be used. By continuing to perform such a process for each learning data, the learning device may gradually generate a region identification model that is updated in the ideal direction.

[0078] Fig.11 and Fig.12 is a diagram for explaining learning data used in region discrimination model learning according to one embodiment.

[0079] Reference Fig.11 , showing the die area 1115 of the configuration chip in the wafer 1110 and the outer area 1130 of the wafer 1110. The outer area 1130 may include, for example, a wafer ring frame area and / or an air area. According to the optical spectrum data measured by the moving trajectory 1120 of the probe, the area 1140 where the die area 1115 exists and the outer area 1130 of the wafer 1110 may be the area for learning the regional discrimination model. The data corresponding to the corresponding areas 1130 and 1140 may be used as learning data for learning the regional discrimination model.

[0080] Reference Fig.12 , showing the intensity of the signal 1210 obtained according to the moving trajectory of the probe changing with time. Among the obtained signals, the signal 1224 corresponding to the die area of ​​the wafer and the signals 1222 and 1226 corresponding to the outer area of ​​the wafer can be used as learning data for learning the area discrimination model.

[0081] The method according to the embodiment can be implemented as a program command form that can be executed by a variety of computer devices, and recorded in a computer-readable recording medium. The computer-readable recording medium may include program commands, data files, data structures, etc. alone or in combination. The program commands recorded in the medium may also be specially designed and constructed for the embodiment or known and usable by computer software practitioners. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices such as ROMs, RAMs, and flash memories that are constructed in the form of storing and executing program commands. Examples of program commands include not only machine language codes created by compilers, but also high-level language codes that can be executed by computers using interpreters, etc. The hardware device may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0082] Software may include a computer program, code, instructions, or a combination of more than one of them, and may constitute a processing device in a form that operates as required or independently or collectively instructs a processing device. Software and / or data may be interpreted by a processing device, or may be permanently or temporarily embodied in any type of machine, component, physical device, virtual device, computer storage medium or device to provide instructions or data to a processing device. Software may be distributed among networked computer systems and stored or run in a decentralized manner. Software and data may be stored in one or more computer-readable recording media.

[0083] Although the embodiments are described above by limited drawings, a person with general knowledge in the technical field can apply various technical modifications and variations based on the above contents. For example, appropriate results can be achieved by executing the described techniques in a different order from the described method, and / or combining or combining the described system, structure, device, circuit and other components in a different form from the described method, or by replacing or replacing with other components or equivalents.

[0084] Therefore, other implementations, other embodiments, and matters equivalent to the claims are also within the scope of the claims.

Claims

1. A wafer area identification device, It is characterized in that include: a probe including an optical sensor that obtains optical spectral signals at a plurality of locations on the wafer whose signal values ​​vary according to thickness characteristics of the wafer; as well as A processor uses a neural network-based region discrimination model that takes the current optical spectrum signal obtained by the probe as input to identify whether the current optical spectrum signal is obtained from a die area where a chip is configured on a wafer.

2. The wafer region identification device according to claim 1, It is characterized in that The processor uses the region discrimination model to identify whether the current optical spectrum signal is obtained from a non-patterned region of the wafer where no chips are configured.

3. The wafer region identification device according to claim 2, It is characterized in that The mold-free area includes at least one of an edge area of ​​the wafer where no chips are arranged and a scribe line area between mold areas.

4. The wafer region identification device according to claim 1, It is characterized in that The processor uses the region discrimination model to identify whether the current optical spectrum signal is obtained from the wafer ring rack region for holding the wafer.

5. The wafer region identification device according to claim 4, It is characterized in that The processor uses a region discrimination model to identify whether the current optical spectrum signal is obtained from an air region outside the wafer ring stand.

6. The wafer region identification device according to claim 1, It is characterized in that When the processor recognizes that a current optical spectrum signal is obtained from the die region, the processor infers the film thickness of the wafer based on the current optical spectrum signal.

7. The wafer region identification device according to claim 6, It is characterized in that The processor infers the film thickness of the wafer using a neural network-based wafer thickness inference model that takes the current optical spectral signal as input.

8. The wafer region identification device according to claim 1, It is characterized in that The processor uses the plurality of optical spectral signals and the region identification model to distinguish the types of regions corresponding to respective locations of the wafer.

9. The wafer region identification device according to claim 1, It is characterized in that The optical spectrum signal obtained by the probe includes signals obtained by reflecting light of multiple wavelength bands on the wafer.

10. The wafer region identification device according to claim 1, It is characterized in that The region discrimination model distinguishes a first signal region corresponding to the pattern region and a second signal region corresponding to the pattern-free region from a signal of a specific wavelength extracted from an optical spectrum signal obtained according to a movement trajectory of the probe.

11. The wafer region identification device according to claim 1, It is characterized in that The region identification model is a region identification model based on convolutional neural network.

12. A method, as a method of identifying a region of a wafer, It is characterized in that The following operations are included: Obtaining optical spectral signals whose signal values ​​change according to the thickness characteristics of the wafer through the probe at multiple positions on the wafer; as well as A neural network-based regional discrimination model is used to identify whether the current optical spectrum signal is obtained from a die area where a chip is configured on a wafer by using the current optical spectrum signal obtained by the probe as input.

13. The method according to claim 12, It is characterized in that The recognition operation includes the following operations: The region discrimination model is used to identify whether the current optical spectrum signal is obtained from a non-patterned region on the wafer where no chips are configured.

14. The method according to claim 12, It is characterized in that The recognition operation includes the following operations: A region discrimination model is used to identify whether the current optical spectrum signal is obtained from the wafer ring rack region for fixing the wafer.

15. The method according to claim 12, It is characterized in that The recognition operation includes the following operations: The region discrimination model is used to identify whether the current optical spectrum signal is obtained from the air region outside the wafer ring stand.

16. The method according to claim 12, It is characterized in that The following operations are also included: In the case of identifying that the current optical spectrum signal is obtained from the die region, the film thickness of the wafer is inferred based on the current optical spectrum signal.

17. The method according to claim 12, It is characterized in that The region discrimination model distinguishes a first signal region corresponding to the pattern region and a second signal region corresponding to the pattern-free region from a signal of a specific wavelength extracted from an optical spectrum signal obtained according to a movement trajectory of the probe.