Etching system and etching method

Abstract

To provide an etching system that can produce an etched substrate having a desired surface shape.SOLUTION: An etching system includes: an etching condition calculation unit that calculates etching conditions on the basis of information regarding a crystalline state of a substrate and target values regarding a surface shape of the substrate;and an etching apparatus that etches the substrate on the basis of the calculated etching conditions.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an etching system and an etching method. [Background technology]

[0002] An etching system is known that includes a container containing molten alkali, in which a substrate is immersed in the molten alkali and then gradually pulled up at a constant speed. Such an etching system is disclosed in Patent Document 1. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-100133 Summary of the Invention [Problem to be solved by the invention]

[0004] Conventional etching systems do not take into account the effect that the crystal orientation of the substrate has on the finish. The crystal orientation of the substrate affects the etching rate. Etching systems that do not take crystal orientation into account often fail to produce etched substrates with the desired surface shape.

[0005] The present disclosure solves at least one of the problems of the prior art. It is possible to provide an etching system that can obtain an etched substrate having a desired surface shape. [Means for solving the problem]

[0006] A first etching system of the present disclosure is an etching system including an etching condition calculation unit that calculates etching conditions based on information regarding the crystalline state of a substrate and target values ​​regarding the surface shape of the substrate, and an etching apparatus that etches the substrate based on the calculated etching conditions.

[0007] A second etching system of the present disclosure is the first etching system, wherein the information on the crystalline state is information obtained from the observation results of at least one selected from the group consisting of dislocations, defects, and lattice distortions within the crystal of the substrate.

[0008] A third etching system of the present disclosure is the second etching system, wherein the observation results are obtained by observing the substrate using an X-ray three-dimensional topography method.

[0009] A fourth etching system of the present disclosure is an etching system in which, in the first etching system, the etching condition calculation unit uses training data including the etching conditions, information on the crystalline state, and information on the surface shape of the substrate etched based on the etching conditions to generate a trained model obtained by supervised learning in which the information on the crystalline state and a target value related to the surface shape are used as explanatory variables and the etching conditions are used as target variables.

[0010] A fifth etching system of the present disclosure is an etching system that, in addition to the first etching system, further includes a grinding condition calculation unit that calculates grinding conditions based on target values ​​related to the crystalline state of the substrate, and a grinding device that grinds the substrate based on the calculated grinding conditions.

[0011] A sixth etching system of the present disclosure is the fifth etching system, wherein the grinding condition calculation unit includes a trained model obtained by supervised learning using training data including the grinding conditions and information regarding the crystalline state of the substrate after grinding, with information regarding the crystalline state of the substrate as an explanatory variable and the grinding conditions as a target variable.

[0012] The seventh etching system of the present disclosure is the sixth etching system, further comprising a crystalline state acquisition unit that acquires information regarding the crystalline state, wherein the crystalline state acquisition unit acquires information regarding the crystalline state of the substrate after grinding by the grinding device, and the etching condition calculation unit calculates the etching conditions using the information regarding the crystalline state acquired by the crystalline state acquisition unit.

[0013] A first etching method of the present disclosure is an etching method that includes calculating etching conditions based on information regarding the crystalline state of a substrate and target values ​​regarding the surface shape of the substrate, and etching the substrate based on the calculated etching conditions. [Effects of the Invention]

[0014] The present disclosure may provide an etching system that results in an etched substrate having a desired surface topography. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a hardware configuration diagram of an etching system 100 according to a first embodiment. [Figure 2] FIG. 1 is an explanatory diagram of a grinding device 1. [Figure 3A] 10 is an explanatory diagram showing the observation results of a notch 61 in a substrate 6 by a crystalline state observation device 3. FIG. [Figure 3B] 10 is an explanatory diagram showing the observation results of a notch 61 in a substrate 6 by a crystalline state observation device 3. FIG. [Figure 3C] 10 is an explanatory diagram showing the observation results of a notch 61 in a substrate 6 by a crystalline state observation device 3. FIG. [Figure 4A] 10 is an explanatory diagram of an image processing method for the observation results of the notch 61. FIG. [Figure 4B] 10 is an explanatory diagram of an image processing method for the observation results of the notch 61. FIG. [Figure 5] FIG. 2 is a block diagram of a control device 5. [Figure 6A]FIG. 1 is a flow chart showing the machine learning procedure of the etching system 100. [Figure 6B] FIG. 1 is a flow chart showing the machine learning procedure of the etching system 100. [Figure 7] FIG. 2 is a flow chart showing the operation procedure of the etching system 100. DETAILED DESCRIPTION OF THE INVENTION

[0016] 1 is a hardware configuration diagram of an etching system 100 according to Example 1. The etching system 100 includes a grinding device 1, a shape measuring device 2, a crystalline state observing device 3, an etching device 4, and a control device 5.

[0017] The grinding device 1 grinds the substrate 6. Specifically, the grinding device 1 grinds the edge portion of the substrate 6. FIG. 2 is an explanatory diagram of the grinding device 1. The grinding device 1 includes a grinding unit 11 and a holding unit 12. The grinding unit 11 includes a spindle 13 and a grinding wheel 14. The grinding wheel 14 has a groove 15. The grinding wheel 14 is rotated by the spindle 13, and the groove 15 is brought into contact with the substrate 6 to grind the outer periphery of the substrate 6.

[0018] The holding unit 12 holds the substrate 6. The holding unit 12 may rotate the substrate 6. The grinding device 1 is equipped with a position adjustment mechanism (not shown). The position adjustment mechanism adjusts the relative position and / or attitude of the grinding unit 11 and the holding unit 12 (i.e., the substrate 6).

[0019] The configuration of the control device 5 will be described later (FIG. 5), but the hardware includes at least a processor 50 and a memory 51. The processor 50 executes a program 52 stored in the memory 51 to control the grinding device 1.

[0020] The grinding wheel 14 has grooves 15 formed along the direction of rotation. The grooves 15 are brought into contact with the portion of the substrate 6 to be machined, thereby grinding the substrate 6. The grinding wheel 14 is composed of abrasive grains made of metal and / or inorganic compounds, etc., and a bond that holds the abrasive grains and forms the grinding wheel 14. The material of the bond is not particularly limited, but it may be metal or resin. The grinding wheel 14 does not necessarily have grooves 15. For example, a grinding wheel that grinds the substrate 6 by bringing a cylindrical or flat grinding wheel surface into contact with the portion to be machined while rotating can also be used. The position and attitude of the grinding wheel 14 relative to the substrate 6 can be controlled by a position adjustment mechanism to machine the outer periphery of the substrate 6 into any desired shape.

[0021] 2 shows how the notch 61 of the substrate 6 is ground. In addition to the notch 61, the grinding device 1 may also grind the outer periphery of the substrate 6 or an orientation flat.

[0022] Typically, the substrate 6 is preferably a semiconductor wafer. The material of the substrate 6 is not particularly limited, but examples thereof include silicon, germanium, gallium arsenide, gallium nitride, silicon carbide, and iridium phosphide.

[0023] The shape measuring device 2 is a device that measures the surface shape of the substrate 6 and acquires information about the surface shape. The surface shape is not particularly limited, but examples include the two-dimensional shape of the substrate 6 (front and side surfaces, etc.) and the surface roughness of the substrate 6. The information about the shape may be the measurement results of the surface shape itself, or a numerical value calculated based on the measurement results. The timing for acquiring the information about the surface shape is not particularly limited, but it is preferably when a predetermined process by the etching system 100 is completed. The predetermined process may include a grinding process of the substrate 6 by the grinding device 1 and an etching process of the substrate 6 by the etching device 4.

[0024] The above is an example of the timing of acquiring information, and the acquisition timing may include, for example, the supply of the substrate 6 to the etching system 100 or the processing of the set. The supply of the substrate 6 to the etching system 100 may be performed by an operator or by a supply device (not shown). The supply device is an automatic control device that can supply an unprocessed substrate 6 to the etching system 100 and collect a processed substrate 6. The information acquired by the shape measuring device 2 is output from the shape measuring device 2, input to the control device 5, and stored in the memory 51.

[0025] The crystalline state observation device 3 is a device that observes the crystalline state of the substrate 6 and acquires information about the crystalline state. The crystalline state is not particularly limited, but examples include crystal dislocations in the substrate 6 and lattice strain of the crystals in the substrate 6. The crystalline state observation device 3 includes an observation unit for observing the crystalline state. The observation unit can be selected appropriately depending on the required observation results. The crystalline state observation device 3 may be equipped with one or more observation units. The information about the crystalline state may be the observation results of the crystalline state themselves, or a numerical value calculated based on the observation results.

[0026] The measurement principle employed by the observation unit is not particularly limited, but is preferably non-destructive. Examples of measurement principles include X-ray three-dimensional topography and Raman spectroscopy. The observation unit is composed of various sensors and the like that perform observations employing these measurement principles. Such observation units are well known and can be incorporated into the crystalline state observation device 3.

[0027] The timing for acquiring the information on the crystalline state is not particularly limited, but is preferably the timing when a predetermined process by the etching system 100 is completed. The predetermined process preferably includes at least the grinding process of the substrate 6 by the grinding apparatus 1. In addition to the above, the predetermined process may also include, for example, the supply or setting process of the substrate 6 to the etching system 100, and the etching process of the substrate 6 by the etching apparatus 4. The information acquired by the crystalline state observation apparatus 3 is output from the crystalline state observation apparatus 3, input to the control device 5, and stored in the memory 51.

[0028] 3A, 3B, and 3C are explanatory diagrams of the observation results of the notch 61 of the substrate 6 using the crystalline state observation device 3. FIG. 3A is a plan view of the notch 61 to be observed. FIG. 3B shows the observation results of the linear portion A-1 of the notch 61, and is a schematic representation of an X-ray topography image acquired at regular intervals along the direction of the arrow Ar1 in FIG. 3A. FIG. 3C shows the observation results of the linear portion A-2 of the notch 61, and is a schematic representation of an X-ray topography image acquired at regular intervals along the direction of the arrow Ar2 in FIG. 3A. All of the observation results were obtained using the crystalline state observation device 3, which employs three-dimensional X-ray topography as its measurement principle.

[0029] X-ray topography is a well-known technique for observing, as an image, changes in X-ray diffraction intensity resulting from dislocations, defects, lattice distortion, etc. (hereinafter also referred to as "defects, etc.") within a crystal. The image obtained can typically be a mapping display of X-ray intensity converted into color density. For example, around a crystal defect, distortion of the crystal lattice increases the diffracted X-ray intensity, resulting in a higher color density (darker color). From this shading, the shape and distribution of defects, etc. can be observed. X-ray three-dimensional topography is a three-dimensional representation of an X-ray topography image. That is, it is a well-known technique for acquiring X-ray topography images by scanning an object (here, a substrate 6) three-dimensionally and then synthesizing these images to generate a tomographic image.

[0030] Observation is performed along the directions of arrows Ar1 and Ar2 in FIG. 3A, and the results are composed of multiple X-ray topography images 62, in which the substrate 6 is likened to thin slices, as shown in FIGS. 3B and 3C. By processing and combining these images, a tomographic image of a certain portion of the substrate 6 is obtained. Note that although FIGS. 3B and 3C show the linear portions A-1 and A-2 of the notch 61, the observation location and observation direction (scanning direction) are not limited to those described above. The observation location (in other words, the observation range) and observation direction (scanning direction) may be specified by an operator or may be automatically controlled by a program 52 stored in the memory 51 of the control device 5.

[0031] The observation results obtained in this way may include the crystalline state inside the substrate 6 as well as the crystalline state on the surface of the substrate 6.

[0032] The notch 61 is broadly composed of straight line portions A-1 and A-2 and curved line portions B-1, B-2, and B-3. The curved line portion B-1 is a curved line portion that continues from the outer periphery of the substrate 6. The curved line portion B-1 is sandwiched between the outer periphery and the straight line portion A-1. The outer periphery can be locally considered to be a straight line. The curved line portion B-1 can be said to be a "single-R" portion with a rounded (rounded) portion on one side. Similarly, the curved line portion B-2 is a single-R portion sandwiched between the straight line portion A-2 and the outer periphery. On the other hand, the curved line portion B-3 is surrounded by the straight line portions A-1 and A-2, which are oriented in different directions, and can be said to be a "double-R" portion with rounded portions on both sides.

[0033] The inventors have found that the straight line sections A-1 and A-2 are the sections where changes in the crystalline state are most likely to occur. The crystalline state affects the etching rate. Therefore, it is preferable that the observation targets of the crystalline state observation device 3 include at least the straight line sections A-1 and A-2.

[0034] 4A and 4B are explanatory diagrams of an image processing method for the observation results of the notch 61. FIG. 4A is an X-ray topography image 62 acquired by a crystalline state observation device 3 employing X-ray three-dimensional topography. The X-ray topography image 62 is a single image in which the substrate 6 is likened to a thin slice. FIG. 4B is a processed image 63 obtained by binarizing the X-ray topography image 62 of FIG. 4A. The crystalline state observation device 3 binarizes the X-ray topography image 62 to obtain the processed image 63, which can then be combined to obtain a tomographic image. This tomographic image, or a numerical value calculated based on this tomographic image, can be used as information regarding the crystalline state of a predetermined range of the notch 61.

[0035] There is no particular limitation on the method for obtaining information about the crystalline state from the tomographic image obtained from the processed image 63. For example, there is a method of calculating the ratio of basic units of a specific color (black or white) to the number of basic units (pixels, picture elements) that make up the image in the processed image 63. This calculation may be performed for all images that make up the tomographic image. In this way, information (numerical values) about the crystalline state at a given position in the notch 61 can be obtained. As already explained, in an X-ray topography image, the color density increases in areas where there are more defects, etc. Therefore, by calculating the ratio of black pixels, information about the distribution of defects, etc. can be obtained. Furthermore, by calculating the ratio of white pixels, information about the distribution of areas where there are no defects, etc. can be obtained.

[0036] Note that the meanings of black and white areas may differ depending on the setting of thresholds and the like in the binarization process. For example, the thresholds in the binarization process may be appropriately selected depending on the target values ​​for the surface shape of the substrate 6 required after the etching process. Furthermore, the binarization process is just one example, and any processing method may be adopted as long as it can digitize the distribution of defects and the like. However, binarization is advantageous in that it is simple, can uniquely identify the distribution, and can speed up the process.

[0037] The information about the crystalline state may include information about the penetration depth of defects, etc. The penetration depth of defects, etc. can be obtained from the distribution of defects, etc. in the processed image 63. Specifically, one method is to detect convex defects, etc. extending from the surface of the substrate 6 toward its interior, and calculate their penetration depth. An example of the procedure is as follows.

[0038] First, a distribution of defects, etc. is obtained from the multiple processed images 63 constituting the tomographic image. Next, convex defects are detected from the distribution of defects, etc. A convex defect may refer to, for example, a location where defects, etc. identified for each pixel, are adjacently clustered and extend from the surface of the substrate 6 to its interior. Next, the penetration depths of convex defects present at approximately the same location in multiple processed images 63 are compared. The penetration depth may be defined, for example, as the maximum depth (distance) of the convex defect extending from the surface of the substrate 6 to its interior. Convex defects may occur with a three-dimensional spread from the surface of the substrate 6 to its interior. Through the above process, it is possible to capture an overall image of a convex defect that is originally continuous and extends three-dimensionally from the multiple processed images 63 constituting the tomographic image. Next, the largest value (deepest value) obtained as a result of the comparison may be determined as the penetration depth of the convex defect.

[0039] The X-ray three-dimensional topography method is advantageous in that it can easily and non-destructively visualize the distribution of defects, etc. The inventors have discovered a relationship between defects, etc. and the surface shape of the substrate 6 after etching. The etching system 100 calculates information about the crystalline state from the observation results using the X-ray three-dimensional topography method, and calculates etching conditions based on this information. This configuration allows for the production of an etched substrate with a better surface shape that is closer to the target value.

[0040] The above is information about the crystalline state obtained by the crystalline state observation device 3 including an observation unit that employs X-ray three-dimensional topography as its measurement principle. Similarly, when using the crystalline state observation device 3 that includes an observation unit that employs another measurement principle (for example, Raman spectroscopy), information about the crystalline state can be calculated based on an image that quantitatively shows changes in the crystalline state in a predetermined area of ​​the substrate 6.

[0041] 1, the etching device 4 performs an etching process on the ground substrate 6. Examples of the etching process include an alkaline etching process using an alkali and an acid etching process using an acid.

[0042] 5 is a block diagram of the control device 5. The control device 5 is a computer including a processor 50 and a memory 51. A program 52 is stored in the memory 51. The processor 50 executes the program 52 stored in the memory 51 to control each part of the etching system 100.

[0043] More specifically, the control device 5 controls each unit of the etching system 100 by having the processor 50 execute a program 52 stored in the memory 51. This program 52 may include an instruction set for controlling the grinding device 1 in accordance with grinding conditions. The grinding conditions preferably typically include setting parameters for the grinding device 1. The setting parameters for the grinding device 1 may include, for example, the material of the substrate 6, the grinding speed, the depth of cut, and the type of grinding wheel. The grinding conditions may be input into the control device 5 by an operator or may be stored in advance in the memory 51.

[0044] The program 52 may also include an instruction set for controlling the etching according to the etching conditions of the etching apparatus 4. The etching conditions typically preferably include setting parameters of the etching apparatus 4. The setting parameters of the etching apparatus 4 may include the material of the substrate 6, the type of etching solution, the processing time, etc.

[0045] The program 52 includes a grinding condition calculation unit 521, a grinding control unit 522, a crystalline state acquisition unit 523, an etching condition calculation unit 524, an etching control unit 525, and a surface shape acquisition unit 526 as a command set for controlling each unit of the etching system 100. The grinding condition calculation unit 521 also includes a first target acquisition unit 521A and a trained model 521B as submodules. The etching condition calculation unit 524 also includes a second target acquisition unit 524A and a second trained model 524B as submodules.

[0046] The grinding condition calculation unit 521 includes a command set for acquiring a grinding target set for the substrate 6, and a command set for predicting information on the crystalline state of the substrate 6 after grinding based on the grinding target. The grinding target includes at least information on the crystalline state of the substrate 6 after grinding.

[0047] The first target acquisition unit 521A, which is a sub-module of the grinding condition calculation unit 521, includes a command set for receiving input of the grinding target Pm1 and storing the acquired grinding target Pm1. The first target acquisition unit 521A may further include a command set for creating an input screen (GUI: Graphical User Interface) to assist the operator in inputting the grinding target Pm1. The grinding target Pm1 includes a target value Pm2 related to the crystalline state after grinding. The target value Pm2 related to the surface state is preferably a value that can be acquired by a method similar to the information related to the crystalline state acquired by the crystalline state observation device 3. The grinding target Pm1 may also include a shape measurement result of the substrate 6 before grinding, and information related to the material of the substrate 6, etc.

[0048] The first trained model 521B is a trained model that receives input of a target value Pm2 related to the crystalline state included in the grinding target Pm1 and outputs the grinding conditions Pm3 required to obtain that target value. The first trained model 521B is obtained by supervised learning using a learning dataset (training dataset) including the grinding conditions and information related to the crystalline state of the substrate 6 processed under those grinding conditions, with information related to the crystalline state as the explanatory variable and the grinding conditions as the target variable. Note that the training dataset may further include other data included in the grinding target Pm1, such as the material of the substrate 6. These other data can be used as explanatory variables.

[0049] The machine learning algorithm is not particularly limited, and any known algorithm can be applied. Examples include linear regression, decision tree, random forest, support vector machine, and neural network. Note that the information on the crystalline state of the substrate included in the training dataset can be obtained by the crystalline state observation device 3. The training method will be described in more detail later.

[0050] The etching system 100 equipped with the first trained model 521B can automate the difficult task of setting grinding conditions. As discovered by the present invention, information about the crystalline state of the substrate 6 affects subsequent etching processes. Therefore, setting grinding conditions so that the ground substrate 6 meets the target value for the crystalline state results in an etched substrate with an excellent surface shape. However, setting grinding conditions generally requires skilled techniques. Controlling the parameters of the grinding device to achieve a predetermined crystalline state can sometimes take some time to find the grinding conditions. On the other hand, the etching system 100 equipped with the first trained model 521B allows for faster grinding condition search, resulting in a faster overall process.

[0051] The grinding control unit 522 includes a command set for controlling the grinding device 1 and grinding the substrate 6 based on the grinding conditions Pm3 created by the first trained model 521B or any grinding conditions.

[0052] The crystalline state acquisition unit 523 includes a command set for controlling the crystalline state observation device 3 to acquire information Pm4 relating to the crystalline state of the substrate 6 after grinding.

[0053] The second target acquisition unit 524A, which is a sub-module of the etching condition calculation unit 524, includes a command set for receiving an input of the etching target Pm5 and storing the acquired etching target Pm5. The second target acquisition unit 524A may further include a command set for creating an input screen (GUI) for assisting the operator in inputting the etching target Pm5. The etching target includes at least a target value Pm6 (target value) related to the surface shape of the substrate 6 after etching.

[0054] The second trained model 524B is a trained model that receives input of a target value Pm6 regarding the surface shape of the substrate 6 after etching, which is included in the etching target Pm5, and input of information Pm4 regarding the crystalline state of the substrate 6 after grinding, and outputs etching conditions Pm7 required to perform processing in accordance with the etching target Pm5. The second trained model 524B is obtained by supervised learning using a training dataset containing information regarding the crystalline state of the substrate 6 before etching, etching conditions, and information regarding the surface shape of the substrate 6 processed under the etching conditions, with the information regarding the crystalline state and the information regarding the surface shape as explanatory variables and the etching conditions as target variables. Note that the information regarding the surface shape of the substrate 6 included in the training dataset can be obtained by the shape measurement device 2. Examples of machine learning algorithms include those similar to those that can be used for the machine learning of the first trained model 521B.

[0055] The etching system 100 equipped with the second trained model 524B can automate the setting of difficult etching conditions. As discovered by the present invention, information about the crystalline state of the substrate 6 affects the surface shape of the substrate 6 after etching. Therefore, setting etching conditions based on information about the crystalline state and a target value for the surface shape results in an etched substrate with an excellent surface shape. However, setting etching conditions generally requires skilled techniques. On the other hand, the etching system 100 equipped with the second trained model 524B allows for faster search for etching conditions, resulting in faster overall processing.

[0056] The etching control unit 525 includes a command set for controlling the etching device 4 and etching the substrate 6 based on the etching condition Pm7 created by the second trained model 524B or any etching condition.

[0057] Next, the operation of the etching system 100 will be described. First, the machine learning procedure of the etching system 100 will be described. FIGS. 6A and 6B are flow diagrams showing the machine learning procedure of the etching system 100. The machine learning procedure is, in other words, the training flow of the first trained model 521B and the second trained model 524B.

[0058] 6A shows a training flow for obtaining the first trained model 521B. First, in step S101, grinding conditions are input from the outside to the control device 5. The grinding conditions input to the control device 5 are stored in the memory 51.

[0059] Next, in step S102, the substrate 6 is ground based on the grinding conditions input to the control device 5. The grinding control unit 522 controls the grinding device 1 based on the input grinding conditions, and grinds the substrate 6.

[0060] Next, in step S103, information regarding the crystalline state of the substrate 6 is acquired. Specifically, the crystalline state acquisition unit 523 controls the crystalline state observation device 3 to observe the crystalline state of the substrate 6, and acquires information regarding the crystalline state through predetermined image processing, etc. The information regarding the crystalline state of the substrate 6 is stored in the memory 51.

[0061] The above flow yields one entry that corresponds to the grinding conditions and information about the crystalline state of the substrate 6 after grinding. The above procedure is repeated for multiple grinding conditions to obtain a training data set consisting of multiple entries (step S104). Note that the above entry may also include other information, such as the material of the corresponding substrate 6. The other information may be used as one of the explanatory variables in the learning step.

[0062] Next, in step S105, the first trained model 521B is obtained by supervised learning using the training data set and with at least information on the crystalline state as an explanatory variable and the grinding conditions as a target variable.

[0063] 6B shows a training flow for obtaining second trained model 524B. First, in step S201, etching conditions are input from the outside to control device 5. The etching conditions input to control device 5 are stored in memory 51.

[0064] Next, in step S202, information regarding the crystalline state of the substrate 6 before etching is acquired. This information may be acquired by the crystalline state acquisition unit 523. The order of steps S201 and S202 may be reversed. Alternatively, steps S201 and S202 may be performed simultaneously. That is, the etching conditions and information regarding the crystalline state may be input simultaneously.

[0065] Next, in step S203, etching is performed based on the etching conditions input to the control device 5. In the grinding process, the etching control unit 525 controls the etching device 4 based on the input etching conditions to etch the substrate 6.

[0066] Next, in step S204, information about the surface shape of the substrate 6 is acquired. Specifically, the surface shape acquisition unit 526 controls the shape measurement device 2 to measure the surface shape of the substrate 6 and acquires information about the surface shape of the substrate 6 after etching. The information about the surface shape of the substrate 6 is stored in the memory 51.

[0067] The above flow produces one entry that corresponds information about the crystalline state before etching, the etching conditions, and information about the surface shape of the substrate 6 after etching. The above procedure is repeated for multiple etching conditions to produce a training data set consisting of multiple entries (step S205). Note that the above entry may also include other information, such as the material of the corresponding substrate 6. The other information can be used as one of the explanatory variables in the learning step.

[0068] Next, in step S206, the second trained model 524B is obtained by supervised learning using the training data set, with at least information about the crystal state and information about the surface shape as explanatory variables and the etching conditions as objective variables. This completes the machine learning procedure in the etching system 100.

[0069] Next, the operation procedure of the etching system 100 including the first trained model 521B and the second trained model 524B will be described.

[0070] 7 is a flow diagram showing the operation procedure of the etching system 100. In other words, FIG. 7 is a flow diagram of a method for etching the substrate 6 using the etching system 100.

[0071] First, in step S301, the first target acquisition unit 521A obtains a grinding target Pm1 through input by an operator or the like. This grinding target Pm1 includes a target value Pm2 related to the crystalline state. Note that the target value Pm2 related to the crystalline state can be determined more simply by using the training data of the first trained model 521B as a grinding process database or the training data of the second trained model 524B as an etching process database.

[0072] Next, in step S302, the first trained model 521B uses the target value Pm2 related to the crystalline state, which is included in the grinding target Pm1, as input data, and predicts and generates the grinding condition Pm3.

[0073] Next, in step S303, the grinding control unit 522 controls the grinding device 1 in accordance with the grinding conditions Pm3 to grind the substrate 6. This results in a substrate 6 having a crystalline state similar to the target value Pm2 for the crystalline state.

[0074] Next, in step S304, the crystalline state acquisition unit 523 controls the crystalline state observation device 3 to acquire information Pm4 regarding the crystalline state of the substrate 6 after grinding. This is not a predicted value, but a value acquired from an actual measurement value. Note that if only grinding of the substrate 6 is performed using the etching system 100, the steps after this may be unnecessary.

[0075] Next, in step S305, the second target obtaining unit 524A obtains an etching target Pm5 through input by an operator or the like. This etching target Pm5 includes a target value Pm6 related to the surface shape. Note that the target value Pm6 related to the surface shape can be determined more simply by using the grinding process database and / or the etching process database described above.

[0076] Next, in step S306, the second trained model 524B uses, as input data, information Pm4 regarding the crystalline state acquired by the crystalline state acquisition unit 523 and target values ​​Pm6 regarding the surface shape acquired by the second target acquisition unit 524A, and predicts and generates the required etching conditions Pm7.

[0077] Next, in step S307, the etching control unit 525 controls the etching device 4 in accordance with the etching condition Pm7 to etch the ground substrate 6. As a result, a substrate 6 having a surface shape similar to the target value Pm6 for the surface shape is obtained.

[0078] Next, in step S308, the surface shape acquisition unit 526 controls the shape measurement device 2 to acquire information about the surface shape of the etched substrate 6. In principle, the information about the surface shape acquired in this step is similar to the target value Pm6 for the surface shape. However, there are cases in which sufficient similarity cannot be obtained in relation to the prediction accuracy of the first trained model 521B and the second trained model 524B. In this case, the finally obtained information about the surface shape may be used to adjust each unit, particularly the first trained model 521B and the second trained model 524B.

[0079] Although the embodiments of the etching system have been described above, the technical concept of the present disclosure is not limited to the above embodiments. Various modifications and alterations can be made within the scope of achieving the desired effects. [Explanation of symbols]

[0080] 1 Grinding equipment 2 Shape measuring device 3 Crystalline state observation device 4. Etching equipment 5. Control device 6 PCB 11 Grinding section 12 Holding part 13 Spindle 14 Grinding wheels 15 groove 50 processors 51 memory 52 Programs 61 notches 62 X-ray topography images 63 processed images 100 Etching System 521 Grinding condition calculation section 521A 1st target acquisition part 521B First trained model 522 Grinding control section 523 Crystal State Acquisition Unit 524 Etching condition calculation section 524A 2nd target acquisition part 524B Second trained model 525 Etching control unit 526 Surface shape acquisition part

Claims

1. an etching condition calculation unit that calculates etching conditions based on information about the crystalline state of the substrate and a target value about the surface shape of the substrate; an etching apparatus that etches the substrate based on the calculated etching conditions.

2. 2. The etching system according to claim 1, wherein the information on the crystalline state is information obtained from the observation results of at least one selected from the group consisting of dislocations, defects, and lattice distortions within the crystal of the substrate.

3. 3. The etching system according to claim 2, wherein the observation results are obtained by observing the substrate using an X-ray three-dimensional topography method.

4. 2. The etching system according to claim 1, wherein the etching condition calculation unit includes a trained model obtained by supervised learning using training data including the etching conditions, information about the crystalline state, and information about a surface shape of the substrate etched based on the etching conditions, with the information about the crystalline state and a target value related to the surface shape as explanatory variables and the etching conditions as objective variables.

5. a grinding condition calculation unit that calculates grinding conditions based on a target value related to the crystalline state of the substrate; The etching system according to claim 1 , further comprising: a grinding device that grinds the substrate based on the calculated grinding conditions.

6. 6. The etching system according to claim 5, wherein the grinding condition calculation unit includes a trained model obtained by supervised learning using training data including the grinding conditions and information about the crystalline state of the substrate after grinding, with information about the crystalline state of the substrate as an explanatory variable and the grinding conditions as a target variable.

7. further comprising a crystalline state acquisition unit that acquires information about the crystalline state; the crystalline state acquisition unit acquires information about the crystalline state of the substrate after grinding by the grinding device; 7. The etching system according to claim 6, wherein said etching condition calculation section calculates said etching conditions using information on said crystalline state acquired by said crystalline state acquisition section.

8. calculating etching conditions based on information about the crystalline state of the substrate and a target value about the surface shape of the substrate; and etching the substrate based on the calculated etching conditions.

Images