JP7835401B2 - PCB processing control device and system - Google Patents

Description

This invention relates to a substrate processing control device and system for controlling a substrate processing apparatus, and more specifically, to a substrate processing control device and system that uses artificial intelligence to control the substrate processing apparatus in order to focus laser light into the interior of the substrate and form a peelable modified layer.

Traditionally, various single crystals have been used as materials for semiconductor substrates. Among these single crystals, silicon (Si), silicon carbide (SiC), gallium nitride (GaN), gallium phosphide (GaP), and diamond (C) are considered difficult to machine, making it challenging to cut them into substrates from bulk materials such as ingots or blocks using conventional machining methods. Therefore, as an alternative to machining, a processing technique is being researched that involves irradiating the main surface of the substrate with laser light, focusing it to a predetermined depth, and forming a processing mark to create a peelable modified layer.

Furthermore, diamond substrates and gallium nitride substrates are used as bases for heteroepicatial growth of diamond crystals on single-crystal silicon or magnesium oxide (MgO) substrates, and gallium nitride on substrates such as silicon, silicon carbide, and _sapphire ( _Al₂O₃ ), respectively. Laser processing technology, as described above, is being researched as a method for reusing these base substrates and for processing the growth substrates into thin wafers.

Japanese Patent Publication No. 2018-183801

Yohei Yamada, et al., "Precision Laser Slicing Technology for Single-Crystal Silicon," Journal of the Japan Society for Precision Engineering, 2019, Vol. 85, No. 5, pp. 419-425. Taijiro Fujita, et al., "An Attempt at Delamination of the (100) Surface by Laser Slicing of Diamond," Proceedings of the 2021 Annual Conference of the Japan Society for Abrasive Technology, Japan Society for Abrasive Technology, pp. 29-30.

However, because the single-crystal materials mentioned above vary in their solid structure and chemical bonding types, and therefore have different properties, in order to form a modified layer that can be peeled off using laser processing technology, it was necessary to individually consider the processing method suitable for each substrate, taking into account the material, crystal structure, bonding morphology, crystal plane orientation, crystal growth off-angle, and other parameters of the single-crystal substrate to be processed, as well as processing conditions such as the characteristics of the laser beam, the focusing of the laser beam into the substrate, the laser beam scanning method, and the depth and thickness of the modified layer to be formed.

This invention is proposed in view of the above-described circumstances and aims to provide a substrate processing control device and system that controls a substrate processing apparatus to process a crystal substrate using an appropriate processing method according to the crystal substrate to be processed and the processing conditions.

To solve the above-mentioned problems, the substrate processing control device according to this application is a substrate processing device that controls a substrate processing apparatus to irradiate a laser beam toward the main surface of a crystal substrate placed on a stage, and to focus the laser beam to a predetermined depth from the main surface to form a peelable modified layer. The control device includes a machine learning unit that receives explanatory variables, including parameters related to the crystal substrate to be processed and parameters related to the processing conditions of the crystal substrate, and outputs objective variables, including parameters related to the optical system for irradiating the crystal substrate with laser light and parameters related to the stage drive, according to the explanatory variables; and a control unit that controls the substrate processing apparatus based on the objective variables output from the machine learning unit.

The machine learning unit includes a PCB processing model generation unit that generates a PCB processing model by learning from multiple training data sets consisting of pairs of explanatory variables and corresponding target variables; a PCB processing model storage unit that stores the PCB processing model generated by the PCB processing model generation unit; and a PCB processing model application unit that applies the PCB processing model stored in the PCB processing model storage unit to infer the target variable from the explanatory variables. The input explanatory variables are processed in the PCB processing model application unit, and the target variable inferred by the PCB processing model application unit may be output.

The substrate processing model generation unit and the substrate processing model storage unit may constitute a substrate processing learning unit that generates a substrate processing model by learning training data and stores the generated substrate processing model. The substrate processing model application unit and the substrate processing model storage unit may constitute a substrate processing inference unit that stores the generated substrate processing model and infers the target variable from the input explanatory variables by applying the substrate processing model.

The explanatory variables may include, as parameters relating to the crystal substrate, at least one of the crystal material, crystal structure, crystal type, plane orientation, off-angle, cleavage plane, slip plane, Burgers vector, and dislocation plane. The explanatory variables may further include, as parameters relating to the crystal substrate, at least one of the wavelength dependence of transmittance, the coordinates of crystal defects, and the coordinates of surface foreign matter. The explanatory variables may also include, as parameters relating to the processing conditions, at least one of the slicing thickness of the crystal substrate, the depth from the surface of the crystal substrate to the modified layer, and the thickness of the modified layer.

The objective variable may include at least one of the optical system parameters: wavelength, pulse width, brightness, and repetition frequency of the laser light. The objective variable may also include at least one of the stage drive parameters: scanning direction and scanning speed.

It may further include a console for inputting explanatory variables. It may also further include sensors for detecting parameters of the crystal substrate.

The substrate processing system according to this application includes a substrate processing control device, a substrate processing apparatus comprising a substrate support device including a stage on which a crystal substrate is placed and a stage drive device for driving the stage, and an optical system for irradiating laser light toward the main surface of the crystal substrate placed on the stage of the substrate support device, wherein the substrate processing apparatus is controlled by the substrate processing control device.

According to this invention, it is possible to control the processing of a crystal substrate appropriately according to the crystal substrate and processing method, using artificial intelligence based on supervised machine learning.

This is a perspective view showing the general configuration of a substrate processing control system. This is a block diagram showing the schematic configuration of a substrate processing control system. This is a block diagram showing the general structure of the machine learning department for circuit board processing. This is a diagram showing the configuration of the circuit board processing model. This diagram illustrates the operation of the machine learning unit for circuit board processing. This is a perspective view showing the crystalline structure of a zincblende-type structure. A perspective view showing the crystal structure of a diamond-shaped structure. This is a perspective view showing the crystal structure of a wurtzite-type structure. This is a perspective view showing the hexagonal crystal structure of gallium nitride. This is a schematic diagram illustrating the relationship between the basic structure of the tetrahedrons constituting the crystal substrate and the laser scanning direction. This is a schematic diagram showing the scanning direction of the laser beam relative to the crystal substrate. This diagram illustrates the formation of processing marks on a crystalline substrate where the plane orientation of the main surface and the cleavage plane are the same. This diagram illustrates the formation of processing marks on a crystalline substrate where the plane orientation of the main surface and the cleavage plane are different. This diagram illustrates the formation of processing marks on a crystalline substrate where the plane orientation of the main surface and the cleavage plane are different. This is a perspective view showing the crystal structure of a rock salt crystal.

The following describes in detail embodiments of the substrate processing control device and system with reference to the drawings. This embodiment envisions a substrate processing control device that controls a substrate processing device to form a peelable modified layer by placing a crystalline substrate on a stage, irradiating it with laser light toward its main surface, and focusing the laser light to a predetermined depth from the main surface. Specifically, it is assumed that a peelable substrate is created by controlling the modified layer to generate cleavage along the cleavage plane and continuously forming this cleavage plane in the plane direction of the crystalline substrate. Here, the plane direction of the crystalline substrate refers to the direction in which the main surface of the crystalline substrate extends, and a layer formed continuously in the plane direction refers to a layer formed so as to extend parallel to the main surface.

Figure 1 is a perspective view showing a schematic configuration of the substrate processing control system of this embodiment. Figure 2 is a block diagram showing a schematic configuration of the substrate processing system. The substrate processing system includes a substrate processing control device 10 and a substrate processing apparatus 100 controlled by the substrate processing control device 10.

The substrate processing apparatus 100 includes a substrate support device 110 that supports and drives the crystal substrate 200, and an optical system 120 that irradiates laser light B toward the main surface 200a of the crystal substrate 200. The substrate support device 110 includes a stage 112 on which the crystal substrate 200 is placed, a stage drive device 111 that supports the stage 112 so that it can be driven in the horizontal XY direction and the vertical Z direction, and so that it can be rotated around the Z axis, and a fixing device 113 that fixes the crystal substrate 200 to the stage 112. The fixing device 113 can be an adhesive layer, a mechanical chuck, an electrostatic chuck, a vacuum chuck, etc.

The optical system 120 includes a laser light source 121 and a laser focusing unit 127, which includes an objective lens 125 and an aberration adjustment unit 126. The laser light B emitted from the laser light source 121 is directed towards the main surface 200a of the crystal substrate 200 via the laser focusing unit 127. The optical system 120 may further include other optical elements, such as a diffractive optical element (DOE).

The substrate processing control device 10 includes a console 13 with a display, keyboard, etc., and a sensor 14 capable of detecting the material, size, crystal orientation, etc., of the crystal substrate 200 to be processed. The sensor 14 may be configured by combining, for example, a transmittance measuring device, a laser interference size measuring device, an X-ray diffractometer, etc.

As shown in Figure 2, in the substrate processing control device 10, parameters related to the crystal substrate 200, such as the material, size, and crystal orientation of the crystal substrate 200, input from the console 13 or sensor 14, and parameters related to processing conditions, such as the slicing thickness, input from the console 13, are input as explanatory variables to the substrate processing machine learning unit 11. Based on the input explanatory variables, the substrate processing machine learning unit 11, following inference by machine-learned artificial intelligence, outputs parameters related to the optical system 120, including the wavelength and pulse width of the laser light B, and parameters related to the driving of the stage 112 in the substrate support device 110, including the driving direction and driving speed of the stage 112, as target variables.

Furthermore, in the substrate processing control device 10, the target variable output from the substrate processing machine learning unit 11 is sent to the control unit 12, and the control unit 12 controls the substrate processing apparatus 100 according to the target variable. Specifically, the control unit 12 sends a control signal to the optical system 120 of the substrate processing apparatus 100 based on the parameters related to the optical system 120 of the target variable, and sends a control signal to the substrate support device 110 of the substrate processing apparatus 100 based on the parameters related to the drive of the stage 112 of the substrate support device 110, controlling the drive of the stage 112 of the substrate support device 110.

As shown in Figure 1, the machine learning unit 11 and control unit 12 of the PCB processing control device 10 may be configured using a personal computer as the main body 16 of the PCB processing control device 10. In this case, the machine learning unit 11 and control unit 12 may be implemented by executing an appropriate program on the personal computer in the main body 16.

Figure 3 is a block diagram showing the schematic configuration of the PCB processing machine learning unit 11. In the PCB processing machine learning unit 11, the PCB processing training data storage unit 21 stores training data used for supervised machine learning. Training data consists of a large number of pairs of explanatory variables and corresponding target variables. The PCB processing training data storage unit 21 has a large amount of training data pre-collected for use in machine learning. Training data may be created, for example, by computer simulation.

The substrate processing model generation unit 22 creates a substrate processing model by learning the training data supplied from the substrate processing training data storage unit 21. Figure 4 shows an example of the configuration of the substrate processing model. This substrate processing model is composed of a neural network. The neural network includes multiple layers composed of nodes and can be configured to optimize weights and biases through supervised learning, i.e., learning from training data. Note that the substrate processing model is not limited to a neural network and may be implemented using other types of configurations.

The PCB processing model storage unit 23 stores the PCB processing model generated by the PCB processing model generation unit 22. If the PCB processing model is constructed using a neural network, it stores the parameters of the PCB processing model, such as weights and biases, which are set through training with the training data. Even if the PCB processing model is implemented using a configuration other than a neural network, the PCB processing model storage unit 23 still stores the parameters of the PCB processing model.

The PCB processing model application unit 24 reads the parameters of the PCB processing model from the PCB processing model storage unit 23 to reproduce the PCB processing model, applies the PCB processing model to the input explanatory variables, and outputs the target variable. If the PCB processing model is composed of a neural network, the weights and biases of the PCB processing model are read from the PCB processing model storage unit 23 to reproduce the PCB processing model.

In the PCB processing machine learning unit 11, the PCB processing model generation unit 22 and the PCB processing model storage unit 23 constitute a PCB processing learning unit 31 that generates a PCB processing model through supervised learning. Furthermore, the PCB processing model storage unit 23 and the PCB processing model application unit 24 constitute a PCB processing inference unit 32 that infers the target variable from the explanatory variables by applying the PCB processing model generated through supervised learning.

Figure 5 illustrates the operation of the substrate processing machine learning unit 11. The substrate processing machine learning unit 11 receives explanatory variables including parameters related to the crystal substrate 200, such as its material, size, and crystal orientation, and parameters related to processing conditions, such as the slicing thickness and the thickness of the modified layer. The parameters related to the crystal substrate 200, such as its material, size, and crystal orientation, may be input using values measured by the sensor 14, or these parameter values may be input from the console 13. Measurement of the crystal substrate 200 by the sensor 14 is performed before the crystal substrate 200 is placed on the stage 112 and processing is carried out. Parameters related to processing conditions, such as the slicing thickness and the thickness of the modified layer, may also be input from the console 13.

Table 1 shows examples of explanatory variables to be input to the substrate processing machine learning unit 11. The explanatory variables include parameters related to the crystal substrate 200. These parameters include parameters specific to each individual crystal substrate 200, such as the material, size, and crystal orientation of the crystal substrate 200. Parameters specific to the crystal substrate 200 also include the XYZ coordinates of crystal defects and the XY coordinates of surface foreign matter. These parameters specific to the crystal substrate 200 can be detected by measurement using the sensor 14, and are shown in the "Crystal Substrate Detection Data" column in Table 1.

Table 2 shows examples of materials for the crystal substrate 200. Table 2 includes single-crystal silicon, single-crystal diamond, 4H-SiC, 6H-SiC, single-crystal gallium nitride, single-crystal gallium arsenide, single-crystal gallium phosphide, single-crystal sapphire, and single-crystal magnesium oxide as examples. The crystal structure and cleavage planes of each material are also described in Table 2. The substrate processing control device 10 is not limited to these types of materials and can handle other types of materials as well. For example, the material may be polycrystalline, such as polycrystalline silicon, or amorphous, such as glass.

As shown in Table 1, the parameters related to the crystalline substrate 200 include parameters determined by the material of the crystalline substrate 200, such as the crystal structure, chemical bonding, plane orientation, off-angle, cleavage plane orientation, slip plane orientation, slip direction, Burgers vector, and dislocation planes. Referring to Table 2, the crystal structure and cleavage planes are shown as material-dependent parameters. These parameters can be considered attributes of the material of the crystalline substrate 200, and are therefore shown in the "Attributes of Crystalline Substrate 200" column in Table 1.

Table 1 shows the slip plane orientation and slip direction in the material attributes column for the crystal substrate 200. In crystalline materials, slip planes arise when crystals above and below a crystal plane shift in a slip direction, which is a direction specified by the crystalline material. This shift occurs by introducing dislocations, which are line defects, into the crystalline material, and the slip plane is formed by the plastic deformation of the crystalline material. The Burgers vector is defined as a vector representing this shift. In a slip system, the slip plane and slip direction are paired, and most of these planes and directions coincide with the closest-packed plane and closest-packed direction in the atomic arrangement of the crystalline material. On the other hand, cleavage is considered a brittle fracture and is said to occur near the starting point of deformation. In this embodiment, the processing mechanism involves focusing laser light B to a predetermined depth from the main surface 200a of the crystal substrate 200 to generate cleavage in the crystalline material, starting from the dislocation-induced shift.

The explanatory variables also include parameters related to processing conditions. These parameters include the thickness of the crystal substrate 200 sliced from the crystal substrate 200, and the thickness of the modified layer formed inside the crystal substrate 200. These parameters are shown in the "Processing Conditions" column of Table 1.

As shown in Table 3 below, the parameters for the optical system 120 include parameters related to the laser light B, such as the wavelength, pulse width, brightness, repetition frequency, and beam profile of the laser light B. These parameters are shown in the "Laser Light B" column of the optical system 120 in Table 3. The parameters for the optical system 120 also include parameters related to the laser focusing unit 127, such as the numerical aperture, magnification, focal length, working distance, depth of focus, and, if necessary, diffractive optical elements for splitting the laser light B into multiple beams. The parameters for the diffractive optical elements may include focusing conditions such as the shape, number, and arrangement of the laser transmitted light that splits the laser light B using the diffractive optical elements. These parameters are shown in the "Laser Focusing Unit 127" column of the optical system 120 in Table 3. The parameters related to the driving of the stage 112 include the driving direction, driving speed, and rotational speed. Here, the rotational speed applies when the stage 112 is rotationally driven. These parameters are shown in the "Stage Driving" column of Table 3. Furthermore, since the driving of stage 112 is relatively equivalent to scanning laser beam B, the stage driving can be interpreted as scanning laser beam B.

The substrate processing machine learning unit 11 applies a substrate processing model generated by supervised machine learning to the input explanatory variables. It then outputs parameters related to the optical system 120, such as laser wavelength and laser output, and parameters of the substrate support device 110, such as the driving direction and driving speed of the stage 112, as target variables.

Referring again to Figure 2, the target variable output from the substrate processing machine learning unit 11 of the substrate processing control device 10 is supplied to the control unit 12. The control unit 12 controls the substrate processing apparatus 100 based on the target variable supplied from the substrate processing machine learning unit 11. For example, it sets the optical system 120 of the substrate processing apparatus 100 based on parameters related to the optical system 120, such as the wavelength and repetition frequency of the laser light B supplied from the substrate processing machine learning unit 11. Furthermore, it tracks and controls the position and speed of the stage 112 of the substrate support device 110 in real time based on parameters related to the driving of the stage 112, such as the driving direction and driving speed of the stage 112 supplied from the substrate processing machine learning unit 11.

Thus, according to this embodiment, the substrate processing apparatus can be controlled to process the crystal substrate 200 using an appropriate processing method according to the crystal substrate 200 to be processed and the processing conditions. Therefore, there is no need to individually consider the substrate processing apparatus 100 according to the material of the crystal substrate 200 and the processing conditions of the substrate, and the workload for substrate processing can be reduced. Furthermore, because it is controlled by artificial intelligence machine learning, substrate processing can be performed accurately and reliably.

Thus, according to this embodiment, by using machine learning of artificial intelligence, in addition to parameters such as the material, crystal structure, bonding mode, crystal plane orientation, and crystal growth off-angle of the crystal substrate 200 made of the single crystal material to be processed, it is also possible to consider processing condition parameters such as the characteristics of the laser beam B, the focusing of the laser beam B into the crystal substrate 200, the scanning method of the laser beam B, and the depth and thickness of the modified layer to be formed. This allows for the formation of a peelable modified layer using appropriate laser beam B processing techniques for single crystal materials having various solid structures and types of chemical bonding.

In this embodiment, it was assumed that a removable modified layer would be formed by irradiating the main surface 200a of the crystal substrate 200 with laser light B, creating a processing mark to a predetermined depth from the main surface 200a. However, the edges of the crystal substrate 200 may also be appropriately processed. For example, processing marks may not be formed at a predetermined distance from the outer edge of the crystal substrate 200, and instead, cleavage may spontaneously occur from the processing marks along the edge toward the edge, forming the modified layer. Such processing of the edges of the crystal substrate 200 can also be taught to the substrate processing model through supervised learning.

Furthermore, the substrate processing state of the crystal substrate 200 processed by controlling the substrate processing apparatus 100 with the substrate processing control device 10 of this embodiment may be evaluated. For example, the modified layer formed on the crystal substrate 200 may be evaluated using image data obtained from sensing results of images obtained from an infrared microscope or optical microscope, such as the cleavage direction, the appearance on the side irradiated by the laser beam B (crack formation, etc.), and the depth of crystal disorder. The evaluation results may be displayed, for example, on the display of the console 13.

The following describes an example of forming a modified layer on a specific crystalline substrate 200 by irradiating it with laser light B using the substrate processing control device 10 and substrate processing control system of this embodiment. In this embodiment, parameters such as the material, size, and crystal orientation of the crystalline substrate 200, as shown in Table 1, may be detected by measuring the crystalline substrate 200 with a sensor 14. These parameters may also be input from the console 13. Furthermore, the sensor 14 may also detect the XYZ coordinates of crystal defects and the XY coordinates of surface foreign matter on the crystalline substrate 200. Since parameters such as the crystal structure, chemical bonding, and surface orientation of the crystalline substrate 200 are specified according to the material of the crystalline substrate 200, they may be read from a database or, if necessary, input from the console 13. Parameters related to processing conditions, such as the slicing thickness and the thickness of the modified layer, may also be input from the console 13.

As shown in Figure 5, these parameters are input as explanatory variables to the substrate processing machine learning unit 11, and the substrate processing machine learning unit 11 outputs the corresponding objective variables using artificial intelligence. As shown in Table 3, the objective variables include the wavelength, pulse width, and repetition frequency of the laser light B, the driving direction of the stage 112, and the driving speed. The substrate processing control device 10 controls the substrate processing apparatus 100 based on these objective variables to appropriately process the crystal substrate 200. Therefore, processing by the substrate processing apparatus 100 does not need to be adjusted even if the crystal substrate 200 to be processed is changed, and an appropriate processing method according to the crystal substrate 200 to be processed is selected under the control of artificial intelligence.

Figure 6 is a perspective view showing the crystal structure of a zincblende-type structure. The zincblende-type structure shown in Figure 6 is found in Group III-V compounds such as gallium arsenide (GaAs) and gallium nitride (GaN). In Figure 6, white circles represent Group III atoms such as gallium (Ga), black circles represent Group V atoms such as arsenic (As) and nitrogen (N), and the lines connecting the white or black circles indicate covalent bonds. The zincblende-type structure is based on a regular tetrahedron formed with four atoms surrounding one atom as its vertices. Figure 6 also shows the (011) plane, which is a cleavage plane.

Figure 7 is a perspective view showing the crystal structure of a diamond-type structure. When the zincblende-type structure shown in Figure 6 is composed of the same type of elements, it results in the diamond-type structure shown in Figure 7. Diamond-type structures are found, for example, in elements belonging to Group IV, such as carbon and silicon. In Figure 6, the white circles represent atoms of carbon, silicon, etc., and the lines connecting the white circles represent covalent bonds. The diamond-type structure is constructed with a regular tetrahedron as its basic structure, formed with four other atoms surrounding one atom as its vertices. Figure 7 also shows the (001) plane, which is the cleavage plane, and the (001) plane, which is sometimes considered the main surface 200a of the crystal substrate 200.

Figure 8 is a perspective view showing the wurtzite structure of gallium nitride. Gallium nitride, belonging to Group III-V compounds, can also adopt the zincblende structure shown in Figure 6, but under constant temperature and pressure, the wurtzite structure shown in Figure 8 is the stable form. In Figure 8, white circles represent gallium (Ga) atoms, black circles represent nitrogen (N) atoms, and the lines connecting the white or black circles indicate covalent bonds.

Figure 9 is a perspective view showing the hexagonal crystal structure of gallium nitride. The unit cell of wurtzite-type gallium nitride is hexagonal, as shown in Figure 9. In Figure 9, which shows the hexagonal structure, as in Figure 8, white circles represent gallium atoms and black circles represent nitrogen atoms. The unit cell of hexagonal gallium nitride shown in Figure 9 has different cleavage planes depending on its polarity. The gallium atomic layer and nitrogen atomic layer are stacked alternately in the c-axis direction, and polarization occurs due to the difference in electronegativity between gallium and nitrogen atoms. The c-plane is a polar plane, the a-plane and m-plane, which are perpendicular to the c-plane, are nonpolar planes, and the r-plane is a semipolar plane. Here, the cleavage planes are the a-plane {11-20} plane and the m-plane {1-100} plane. Note that due to restrictions on the characters that can be used in patent application documents, a negative sign "-" is used instead of an overline for Miller indices. Such hexagonal gallium nitride can be grown along the c-axis on a base substrate. In a hexagonal crystal structure, the base and apex faces are {0001} planes.

Figure 10 is a schematic diagram showing the relationship between the basic structure of the tetrahedrons constituting the crystal substrate 200 and the laser scanning direction. Single crystals having crystal structures such as the zincblende type structure shown in Figure 6 and the diamond type structure shown in Figure 7, and crystal substrates 200 composed of such single crystals, are constructed with tetrahedrons as their basic structure. In crystal substrates 200 where the main surface 200a has a plane orientation such as (100) or (111), one face of the tetrahedron of the basic structure is arranged parallel to the main surface. In this case, the laser beam B can be scanned in a specific direction, taking into account the arrangement of the tetrahedrons of the basic structure, so that cleavage can be formed from dislocations in the crystal material due to laser irradiation. The figure illustrates the arrangement of the tetrahedrons of the basic structure relative to the main surface 200a of the crystal substrate 200, and the direction in which the laser beam B is scanned toward the main surface 200a of the crystal substrate 200, using the tetrahedrons of the basic structure as a reference.

Referring to Figure 1, the direction in which the laser beam B is scanned toward the main surface 200a with respect to the regular tetrahedron corresponds to the direction in which the laser beam B is scanned from the optical system 120, including the laser light source 121, along the main surface 200a of the crystal substrate 200. In the substrate processing machine learning unit 11 shown in Figure 3, the direction in which the laser is scanned is provided to the substrate processing learning unit 31 from the substrate processing training data storage unit 21 as training data for the type of crystal substrate 200 to be processed, based on the occurrence and direction of cleavage originating from displacements due to dislocations in the crystal material for the surface orientation of the crystal substrate 200, such as (100) and (111). This data may then be stored as parameters in the substrate processing model storage unit 23. The substrate processing inference unit 32 can output an appropriate laser scanning direction based on the parameters read from the substrate processing model storage unit 23.

Figure 10(a) shows a first embodiment of the scanning direction of the laser beam B. The dashed lines in the figure indicate the scanning lines of the laser beam B. The laser beam B is scanned along the lines at regular intervals (dot pitch) across the crystal substrate 200, scanning across the entire main surface 200a of the crystal substrate 200 so that adjacent lines are at regular intervals (line pitch). The thick arrows in the figure indicate the direction in which the lines are moved. In the first embodiment, the direction of the scanning lines of the laser beam B is perpendicular to one of the edges of an equilateral triangle that constitutes one face of a regular tetrahedron of the basic figure, which is parallel to the main surface 200a. By scanning in this direction, it is possible to break specific covalent bonds extending from the atom located at the center of the tetrahedron of the basic figure toward each face of the tetrahedron, thereby generating cleavage.

Figure 10(b) shows a second embodiment of the scanning direction of the laser beam B. In the second embodiment, the dashed lines in the figure indicate the scanning lines of the laser beam B. The laser beam B is scanned along the lines at regular intervals (dot pitch) across the crystal substrate 200, scanning across the entire main surface 200a of the crystal substrate 200 so that adjacent lines are at regular intervals (line pitch). The thick arrows in the figure indicate the direction in which the lines are moved. In the second embodiment, the direction of the scanning lines of the laser beam B is parallel to one of the sides of an equilateral triangle that constitutes one face of a regular tetrahedron of the basic figure parallel to the main surface. By scanning in this direction, it is possible to break specific covalent bonds extending from the atom located at the center of the regular tetrahedron of the basic figure toward each face of the tetrahedron, thereby generating cleavage.

Figure 11 is a schematic diagram showing the scanning direction of laser beam B on the crystal substrate 200. Laser beam B is scanned in an appropriate direction to cause cleavage in the crystal according to the (100), (111), and other plane orientations of the crystal substrate 200. The scanning direction of laser beam B is determined on the main surface 200a of the crystal substrate 200, taking into consideration the arrangement of the tetrahedra, the basic structure constituting the crystal substrate 200, as explained in Figure 10. On the main surface 200a of the crystal substrate 200, laser beam B is scanned along lines at regular intervals (dot pitch), scanning across the entire main surface 200a of the crystal substrate 200 so that adjacent lines are at regular intervals (line pitch). These scanning directions are set in combination with parameters such as laser beam parameters, dot pitch, and line pitch for the material to be processed and the crystal orientation. Processing conditions are selected that form a peelable processed layer due to the generated cleavage, and that the crystal substrate 200 is peeled in the plane direction at the cleavage plane to create a new substrate.

Figure 11(a) is a schematic diagram showing a first embodiment of a laser beam B scanning the main surface 200a of the crystal substrate 200. In this first embodiment, the laser beam B may scan along the substantially rectangular main surface 200a of the crystal substrate 200 in one direction, such as from right to left or from left to right in the figure. Alternatively, the scanning direction of the laser beam B may be switched between scanning from right to left and from left to right for each adjacent line. The line scanned by the laser beam B may move along the main surface 200a from bottom to top or from top to bottom in the figure.

Figure 11(b) is a schematic diagram showing a second embodiment of the laser beam B scanning the main surface 200a of the crystal substrate 200. The second embodiment is obtained by rotating the scanning direction of the laser beam in the first embodiment shown in Figure 11(a) by 90° within the main surface 200a. In the second embodiment, the laser beam may scan along the main surface 200a of the substantially rectangular crystal substrate 200 in one direction, such as from bottom to top or from top to bottom in the figure. Furthermore, the scanning direction of the laser beam B may be switched between bottom to top and top to bottom for each adjacent line. The line scanned by the laser beam B may move along the main surface 200a from right to left or from left to right in the figure.

Figure 11(c) is a schematic diagram showing a third embodiment of the laser beam B scanning the main surface 200a of the crystal substrate 200. The third embodiment is obtained by rotating the scanning direction of the laser beam B in the first embodiment shown in Figure 11(a) over a predetermined angle within the main surface. In the third embodiment, the laser beam B may scan the main surface 200a of the crystal substrate 200 in one direction, either in a predetermined direction or in the opposite direction. The scanning direction of the laser beam B may be switched for each adjacent line, in a predetermined direction and then in the opposite direction. The line scanning the laser beam B may move in any direction as long as it can scan the entire main surface 200a.

Figure 12 illustrates the formation of processing marks 212 on a crystal substrate 210 where the plane orientation of the main surface 210a and the cleavage plane are the same. Figure 12(a) is a top view of the crystal substrate 210 on which processing marks 212 have been formed, and Figures 12(b) and 12(c) are cross-sectional views of the crystal substrate 210 on which processing marks 212 have been formed. Examples of crystal substrates 210 where the plane orientation of the main surface 210a and the cleavage plane are the same include single-crystal silicon or single-crystal diamond crystal substrates 210 with a (111) plane main surface 210a, and rock salt type crystal substrates 210, which will be described later, with a (100) plane main surface 210a.

As shown in Figure 12(b), the laser beam B is focused from the main surface 210a to a predetermined depth in the crystal substrate 210, forming a processing mark 212. Cracks 214 extend from the processing mark 212 along cleavage planes with the same orientation as the main surface 210a. As shown in Figure 12(a), the laser beam B is irradiated so as to scan the entire main surface 210a. Because the laser beam B is irradiated in a pulsed manner at a predetermined repetition frequency, the processing marks 212 are formed along the scanning line of the laser beam B with a predetermined dot pitch. The thin arrows in Figures 12(a) and 12(b) indicate the direction of scanning of the laser beam B along the line, and the thick arrow in Figure 12(a) indicates the direction of movement along the scanning line of the laser beam B. The laser beam B is scanned so that it has a predetermined line pitch with respect to adjacent lines.

As shown in Figure 12(b), cracks 214 extending from the processing marks 212 extend in the scanning direction of the laser beam B along the direction of the cleavage plane, which has the same surface orientation as the main surface 210a, around the processing marks 212. As shown in Figure 12(c), the cracks 214 extending from each processing mark 212 connect with the cracks 214 extending from adjacent processing marks 212, thereby forming a modified layer inside the crystal substrate 210 in which processing marks 212 and cracks 214 extending parallel to the main surface 210a are connected. The connection of cracks 214 extending from adjacent processing marks 212 is achieved by appropriately setting parameters related to the optical system 120, as shown in Table 3.

The modified layer separates the crystalline substrate 210 into the crystalline substrate from the main surface 210a to the modified layer and the remaining crystalline substrate deeper than the modified layer. Since the modified layer is formed by the interconnection of processing marks 212 and cracks 214, the crystalline substrate from the main surface 210a to the modified layer and the remaining crystalline substrate deeper than the modified layer can be separated by cleavage at the modified layer. The crystalline substrate from the main surface 210a to the modified layer is called a sliced substrate because, similar to slicing, it is a thin crystalline substrate cut from the original crystalline substrate 210. The thickness of the sliced substrate is referred to as the slicing thickness. By separating the sliced substrate at the modified layer, a new crystalline substrate is obtained from the crystalline substrate 210.

Figures 13 and 14 illustrate the formation of processing marks 222 on a crystal substrate 220 where the plane orientation of the main surface 220a and the cleavage plane are different. Figures 13(a) and 13(b) are top and cross-sectional views of the crystal substrate 220 on which processing marks 222 have been formed by the first scanning of laser beam B, respectively. Figures 14(a) to 14(c) are top and cross-sectional views of the crystal substrate 220 on which processing marks 222 have been formed by the second scanning of laser beam B. Examples of crystal substrates 220 where the plane orientation of the main surface 220a and the cleavage plane are different include single-crystal silicon or single-crystal diamond crystal substrates 220 with a (001) plane main surface 220a, and rock salt type crystal substrates 220, which will be described later, with a (111) plane main surface 220a.

As shown in Figure 13(b), during the first scan of the laser beam B, the laser beam B is focused from the main surface 220a to a predetermined depth in the crystal substrate 220, forming a processing mark 222. Following the first scan of the laser beam B, a crack 224 extends from the processing mark 222 along a first cleavage plane with a different surface orientation from the main surface 220a, in the scanning direction of the laser beam B. As shown in Figure 13(a), the laser beam B is irradiated so as to scan the entire main surface 220a. Since the laser beam B is irradiated in a pulsed manner at a predetermined repetition frequency, the processing marks 222 are formed along the scanning line of the laser beam B with a predetermined dot pitch. The thin arrows in Figures 13(a) and 13(b) indicate the direction of scanning of the laser beam B along the line, and the thick arrow in Figure 13(a) indicates the direction of movement along the scanning line of the laser beam B. The laser beam B is scanned such that it has a predetermined line pitch with respect to adjacent lines. The same applies to Figures 14(a) through 14(c).

As shown in Figure 14(b), in the second scan, the laser beam B is focused from the main surface 220a to a predetermined depth in the crystal substrate 220, forming the processing marks 222. As indicated by the thin arrows in Figures 14(a) and 14(b), the second scan of the laser beam B is performed by scanning along the same line as the first scan of the laser beam B, but in the opposite direction. As indicated by the thick arrow in Figure 14(a), the direction in which the line scanning the laser beam B moves is also the opposite direction to that of the first scan of the laser beam B. The spacing of the laser beam B scans along the line (dot pitch) and the spacing in the direction of movement of the line scanning the laser beam B (line pitch) are the same as those of the first scan of the laser beam B.

As shown in Figure 14(b), in the second scanning of the laser beam B, the laser beam B is focused from the main surface 220a to a processing mark 222 of a predetermined depth in the crystal substrate 220. As a result of the second scanning of the laser beam B, a crack 224 extends from the processing mark 222 along a second cleavage plane, which has a different surface orientation from both the main surface 220a and the first cleavage plane, in the scanning direction of the laser beam B. Here, since the scanning direction of the laser beam B in the second scanning is reversed compared to the first scanning of the laser beam B, the profile of the laser beam B with respect to the line is reversed in the second scanning of the laser beam B compared to the first scanning of the laser beam B. As shown in Figure 14(b), when the first and second scans of laser beam B are performed together, the processing marks 222 formed by the irradiation of laser beam B, and the cracks 224 extending from the processing marks 222 along the first and second cleavage planes, are formed symmetrically with respect to the line.

In Figure 14(c), cracks 224 are formed symmetrically from the processing marks 222 with respect to the line, and the cracks 224 extending from each processing mark 222 are interconnected. The connection of the cracks 224 that are not parallel to the main surface 220a, i.e., the first and second cleavage planes, creates a modified layer parallel to the main surface 220a within the crystal substrate 220. The connection of cracks 224 extending from adjacent processing marks 222 is achieved by appropriately setting parameters related to the optical system 120, as shown in Table 3.

The modified layer separates the crystalline substrate 220 into the crystalline substrate 220 from the main surface 220a to the modified layer and the remaining crystalline substrate 220 deeper than the modified layer. This connected modified layer promotes cleavage across the entire crystalline substrate 220, allowing the crystalline substrate from the main surface 220a to the modified layer to be separated from the remaining crystalline substrate deeper than the modified layer. The crystalline substrate from the main surface 220a to the modified layer is called a sliced substrate because, similar to slicing, it is a thin substrate crystal cut from the original crystalline substrate 220. The thickness of the sliced substrate is referred to as the slicing thickness. By peeling off the sliced layer with the modified layer, a new crystalline substrate is obtained from the crystalline substrate 220.

In this embodiment, the material, crystal structure, bonding mode, crystal plane orientation, crystal growth off-angle, and other parameters of the crystal substrate 200 to be processed are confirmed. Furthermore, processing conditions such as the characteristics of the laser beam, the focusing of the laser beam into the substrate, the laser beam scanning method, and the depth and thickness of the modified layer to be formed are also considered. A processing method suitable for the crystal substrate 200 is then implemented under artificial intelligence control. This reduces the burden on the operator slicing the crystal substrate 200 and guarantees that the crystal substrate is sliced under the desired conditions.

Figure 15 is a perspective view showing the crystal structure of rock salt (NaCl). In Example 2, single-crystal gallium nitride (GaN), single-crystal gallium arsenide (GaAs), and single-crystal gallium phosphide (GaP) are used as examples of materials for the crystal substrate 200. These single-crystal gallium nitride, single-crystal gallium arsenide, or single-crystal gallium phosphide chlorides have a sodium-type unit cell. In Figure 14, white circles represent gallium (Ga) atoms, black circles represent nitrogen (N), arsenic (As), or phosphorus (P) atoms, and lines connecting the white or black circles indicate covalent bonds.

Similarly to Example 1, a modified layer can be formed inside the crystal substrate 200, which is made of single-crystal gallium nitride, single-crystal gallium arsenide, or single-crystal gallium phosphide, by irradiating it with laser light B based on the target variable output from the substrate processing machine learning unit 11. By cleaving the crystal substrate 200 with this modified layer, a crystal substrate thinner than the original crystal substrate 200 can be obtained, similar to Example 1.

Claims (12)

A substrate processing control device controls a substrate processing apparatus to irradiate a laser beam toward the main surface of a crystal substrate placed on a stage, and to focus the laser beam to a predetermined depth from the main surface to form a peelable modified layer,
A machine learning unit receives explanatory variables, including parameters related to the crystal substrate to be processed and parameters related to the processing conditions of the crystal substrate, and outputs a target variable, including parameters related to the optical system for irradiating the crystal substrate with laser light and parameters related to the driving of the stage, according to the explanatory variables.
A substrate processing control device including a control unit that controls the substrate processing apparatus based on an objective variable output from the machine learning unit. The aforementioned machine learning unit,
A circuit board processing model generation unit generates a circuit board processing model by learning multiple training data sets, with respect to training data consisting of pairs of explanatory variables and corresponding objective variables.
A substrate processing model storage unit for storing the substrate processing model generated by the substrate processing model generation unit,
The board processing model application unit applies the board processing model stored in the board processing model storage unit to infer the objective variable from the explanatory variables,
The substrate processing control device according to claim 1, wherein the input explanatory variables are processed in the substrate processing model application unit, and the target variable inferred by the substrate processing model application unit is output. The substrate processing control device according to claim 2, wherein the substrate processing model generation unit and the substrate processing model storage unit constitute a substrate processing learning unit that learns the training data to generate a substrate processing model and stores the generated substrate processing model. The substrate processing control device according to claim 2, wherein the substrate processing model application unit and the substrate processing model storage unit store the generated substrate processing model and constitute a substrate processing inference unit that applies the substrate processing model to infer the target variable from the input explanatory variables. The substrate processing control device according to claim 1, wherein the explanatory variable includes, as a parameter relating to the crystal substrate, at least one of the crystal material, crystal structure, crystal type, plane orientation, off-angle, cleavage plane, slip plane, Burgers vector, and dislocation plane. The substrate processing control device according to claim 5, further comprising, as parameters relating to the crystal substrate, at least one of the wavelength dependence of transmittance, the coordinates of crystal defects, and the coordinates of surface foreign matter, as explanatory variables. The substrate processing control device according to claim 1, wherein the explanatory variable includes, as a parameter relating to the processing conditions, at least one of the thickness to which the crystal substrate is sliced, the depth from the surface of the crystal substrate to the modified layer, and the thickness of the modified layer. The substrate processing control device according to claim 1, wherein the objective variable includes at least one of the parameters relating to the optical system: wavelength, pulse width, brightness, and repetition frequency of the laser light. The substrate processing control device according to claim 1, wherein the objective variable includes at least one of the scanning direction and scanning speed as parameters related to the driving of the stage. The substrate processing control device according to claim 1, further comprising a console on which the aforementioned explanatory variables can be input. The substrate processing control device according to claim 1, further comprising a sensor for detecting the parameters of the crystal substrate. A substrate processing control device according to any one of claims 1 to 11,
The aforementioned substrate processing apparatus,
A substrate support apparatus including a stage on which the crystal substrate is placed and a stage drive device for driving the stage,
The substrate processing apparatus includes an optical system that irradiates laser light toward the main surface of a crystal substrate placed on a stage of the substrate support apparatus,
The substrate processing apparatus is a substrate processing system controlled by the substrate processing control device.