Dressing condition determination device and dressing condition determination method

Abstract

The present invention relates to a technology for determining a dressing condition for dressing a polishing pad that is used for polishing a workpiece such as a wafer, a substrate, or a panel. A dressing condition determination device (60) comprises: a storage device (60a) storing an estimation model (62) constructed by machine learning; and a processor (60b) that executes calculation in accordance with an algorithm of the estimation model (62). The estimation model (62) is configured to output, upon input of a target trajectory image showing a target trajectory of a plurality of abrasive grains, which constitute a dressing surface (51a), on a polishing pad (2), a dressing condition for achieving the target trajectory.

Description

Dressing condition determining device and dressing condition determining method

[0001] The present invention relates to a technique for determining dressing conditions for dressing a polishing pad used in polishing a workpiece such as a wafer, substrate, or panel.

[0002] Chemical mechanical polishing (hereinafter referred to as CMP) is a process in which a workpiece (e.g., a wafer, substrate, or panel) is polished by sliding the workpiece against a polishing pad while supplying a polishing liquid onto the polishing pad. The polishing liquid is typically a slurry containing abrasive particles. A polishing apparatus for CMP includes a polishing table that supports a polishing pad having a polishing surface, and a polishing head that presses the workpiece against the polishing pad.

[0003] A polishing apparatus polishes a workpiece as follows: A polishing liquid (typically a slurry) is supplied to the polishing surface of the polishing pad while the polishing table and polishing pad are rotated together. The polishing head presses the surface of the workpiece against the polishing surface of the polishing pad while rotating the workpiece. The workpiece is brought into sliding contact with the polishing pad in the presence of the polishing liquid. The surface of the workpiece is polished by the chemical action of the polishing liquid and the mechanical action of the abrasive grains contained in the polishing liquid and / or the polishing pad.

[0004] The polishing rate (also called removal rate) of a workpiece can vary depending on various factors. The condition of the polishing surface of the polishing pad is one of the factors that affect the polishing rate. That is, the workpiece is brought into sliding contact with the polishing surface while the polishing liquid is held on the polishing surface of the polishing pad. The polishing pad has many minute protrusions on its polishing surface. These many protrusions contribute to the mechanical polishing of the workpiece. Therefore, the condition of the polishing surface of the polishing pad can affect the polishing rate of the workpiece.

[0005] Therefore, in order to improve and stabilize the polishing rate of the workpiece, the polishing surface of the polishing pad is dressed before the workpiece is polished (hereinafter referred to as pad dressing). Pad dressing is performed using a dresser with a dressing surface made of abrasive grains such as diamond particles. More specifically, while the polishing pad and dresser are rotated, the dressing surface of the dresser is pressed against the polishing surface of the polishing pad to roughen the polishing surface (i.e., numerous minute protrusions are formed on the polishing surface).

[0006] JP 2010-76049 A

[0007] Pad dressing is performed under predetermined dressing conditions. It is desirable for the polishing surface of a polishing pad to have a uniform asperity over its entire surface. Therefore, the dressing conditions are determined so as to achieve such a polishing surface. However, in the past, users determined the dressing conditions by trial and error, which sometimes resulted in failure to achieve the intended dressing results.

[0008] Therefore, the present invention provides a technique that can automatically determine the dressing conditions that can achieve the desired state of the polishing surface.

[0009] In one aspect, a dressing condition determination device is provided that determines dressing conditions for dressing a polishing pad by pressing the dressing surface of a rotating dresser against the rotating polishing pad, the dressing condition determination device comprising: a storage device that stores an estimation model constructed by machine learning; and a processor that performs calculations according to an algorithm of the estimation model, wherein the estimation model is configured to output dressing conditions for realizing a target trajectory image showing a target trajectory on the polishing pad of a plurality of abrasive grains that constitute the dressing surface when the target trajectory image is input.

[0010] In one aspect, the dressing conditions include at least one of the rotation speed of the dresser, the rotation speed of the polishing pad, the oscillation speed of the dresser on the polishing pad, and a dressing time. In one aspect, the target trajectory image is a simulation image of the trajectories of the multiple abrasive grains generated by performing a simulation of an operation of dressing the polishing pad by pressing the dressing surface of the rotating dresser against the rotating polishing pad. In one aspect, the estimation model is a trained model constructed by machine learning using training data including multiple dressing conditions and multiple trajectory images obtained by dressing simulations performed under the multiple dressing conditions. In one aspect, the estimation model includes an encoder that extracts feature quantities from the target trajectory image, a decoder that restores the target trajectory image from the feature quantities, and a regression analyzer that performs regression analysis on the extracted feature quantities to output the dressing conditions. In one embodiment, the target trajectory is a trajectory in which the trajectories of the abrasive grains are uniformly distributed on the polishing pad and the intersections of the trajectories of the abrasive grains are uniformly distributed on the polishing pad.

[0011] In one aspect, a dressing condition determination method is provided that determines dressing conditions for dressing a polishing pad by pressing the dressing surface of a rotating dresser against the rotating polishing pad, the dressing condition determination method inputting a target trajectory image showing the target trajectory on the polishing pad of a plurality of abrasive grains that make up the dressing surface into an estimation model constructed by machine learning, and outputting dressing conditions for realizing the target trajectory from the estimation model.

[0012] In one aspect, the dressing conditions include at least one of the rotation speed of the dresser, the rotation speed of the polishing pad, the oscillation speed of the dresser on the polishing pad, and a dressing time. In one aspect, the target trajectory image is a simulation image of the trajectories of the multiple abrasive grains generated by performing a simulation of an operation of dressing the polishing pad by pressing the dressing surface of the rotating dresser against the rotating polishing pad. In one aspect, the estimation model is a trained model constructed by machine learning using training data including multiple dressing conditions and multiple trajectory images obtained by dressing simulations performed under the multiple dressing conditions. In one aspect, the estimation model includes an encoder that extracts feature quantities from the target trajectory image, a decoder that restores the target trajectory image from the feature quantities, and a regression analyzer that performs regression analysis on the extracted feature quantities to output the dressing conditions. In one embodiment, the target trajectory is a trajectory in which the trajectories of the abrasive grains are uniformly distributed on the polishing pad and the intersections of the trajectories of the abrasive grains are uniformly distributed on the polishing pad.

[0013] The above-described dressing condition determining apparatus and dressing condition determining method can automatically determine dressing conditions that can realize a desired state of the polishing surface using an estimation model.

[0014] FIG. 1 is a schematic diagram showing an embodiment of a polishing apparatus; FIG. 2 is a schematic diagram showing an example of trajectories of abrasive grains constituting a dressing surface on the polishing surface of a polishing pad; FIG. 3 is a diagram showing an example of a target trajectory image; FIG. 4 is a schematic diagram showing an embodiment of an estimation model; FIG. 5 is a schematic diagram showing an example of training data including a plurality of dressing conditions and a plurality of corresponding trajectory images; FIG. 6 is a schematic diagram for explaining a training phase in which an estimation model is constructed by machine learning using training data; and FIG. 7 is a schematic diagram for explaining a practical phase in which dressing conditions for realizing the trajectories of abrasive grains on the target trajectory image are determined using an estimation model constructed by machine learning.

[0015] Hereinafter, embodiments of the present invention will be described with reference to the drawings. Fig. 1 is a schematic diagram showing one embodiment of a polishing apparatus. The polishing apparatus is an apparatus for chemically and mechanically polishing a wafer W, which is an example of a workpiece used in the manufacture of semiconductor devices. As shown in Fig. 1, this polishing apparatus includes a polishing table 5 that supports a polishing pad 2 having a polishing surface 2a, a polishing head 7 that presses the wafer W against the polishing surface 2a, a polishing liquid supply nozzle 8 that supplies a polishing liquid (e.g., a slurry containing abrasive grains) to the polishing surface 2a, and an operation control unit 30 that controls the operation of the polishing apparatus.

[0016] The polishing head 7 is configured to be able to hold a wafer W on its lower surface. The wafer W has a film to be polished. In the following embodiments, a wafer is used as an example of a workpiece, but the workpiece is not limited to a wafer and may be a circular substrate, a rectangular substrate, a panel, or the like used in the manufacture of semiconductor devices.

[0017] The polishing apparatus further includes a support shaft 14, a polishing head swing arm 16 connected to the upper end of the support shaft 14, and a polishing head shaft 18 rotatably supported at the free end of the polishing head swing arm 16. The polishing head 7 is fixed to the lower end of the polishing head shaft 18. A polishing head rotation mechanism (not shown) equipped with an electric motor or the like is disposed within the polishing head swing arm 16. This polishing head rotation mechanism is connected to the polishing head shaft 18 and is configured to rotate the polishing head shaft 18 and the polishing head 7 in the direction indicated by the arrow.

[0018] The polishing head shaft 18 is connected to a polishing head lifting mechanism (including a ball screw mechanism, etc.) not shown. This polishing head lifting mechanism is configured to move the polishing head shaft 18 up and down relative to the polishing head swing arm 16. The up and down movement of the polishing head shaft 18 allows the polishing head 7 to move up and down relative to the polishing head swing arm 16 and the polishing table 5, as shown by the arrows.

[0019] The polishing apparatus further includes a table rotation motor 21 that rotates the polishing pad 2 and the polishing table 5 about their respective axes. The table rotation motor 21 is disposed below the polishing table 5, and the polishing table 5 is connected to the table rotation motor 21 via a table shaft 5a. The polishing table 5 and the polishing pad 2 are rotated by the table rotation motor 21 about the table shaft 5a in the direction indicated by the arrow. The polishing pad 2 is attached to the upper surface of the polishing table 5. The exposed surface of the polishing pad 2 forms a polishing surface 2a that polishes the wafer W.

[0020] The wafer W is polished as follows. The wafer W is held by the polishing head 7 with its surface to be polished facing downward. While the polishing head 7 and the polishing table 5 are rotating, a polishing liquid (e.g., a slurry containing abrasive grains) is supplied onto the polishing surface 2a of the polishing pad 2 from a polishing liquid supply nozzle 8 provided above the polishing table 5. The polishing pad 2 rotates integrally with the polishing table 5 around its central axis. The polishing head 7 is moved to a predetermined height by a polishing head lifting mechanism (not shown). Furthermore, while the polishing head 7 is maintained at the predetermined height, the wafer W is pressed against the polishing surface 2a of the polishing pad 2. The wafer W rotates integrally with the polishing head 7. With the polishing liquid present on the polishing surface 2a of the polishing pad 2, the wafer W is brought into sliding contact with the polishing surface 2a. The surface of the wafer W is polished by a combination of the chemical action of the polishing liquid and the mechanical action of the abrasive grains contained in the polishing liquid and / or the polishing pad 2.

[0021] The operation control unit 30 is composed of at least one computer. The operation control unit 30 includes a storage device 30a in which a program is stored and an arithmetic unit 30b that executes calculations according to instructions included in the program. The storage device 30a includes a main storage device such as a random access memory (RAM) and an auxiliary storage device such as a hard disk drive (HDD) or a solid state drive (SSD). Examples of the arithmetic unit 30b include a CPU (central processing unit) and a GPU (graphics processing unit). However, the specific configuration of the operation control unit 30 is not limited to these examples.

[0022] The polishing apparatus includes a dressing mechanism 50 that dresses the polishing surface 2a of the polishing pad 2. The dressing mechanism 50 includes a dresser 51 that is in sliding contact with the polishing surface 2a of the polishing pad 2, a dresser shaft 52 to which the dresser 51 is connected, and a dresser swing arm 55 that rotatably supports the dresser shaft 52. The lower surface of the dresser 51 forms a dressing surface 51a, which is made of abrasive grains (e.g., diamond particles).

[0023] The dresser shaft 52 is connected to a dresser pressing mechanism (including, for example, an air cylinder) (not shown) disposed within the dresser swing arm 55. This dresser pressing mechanism is configured to press the dressing surface 51 a of the dresser 51 against the polishing surface 2 a of the polishing pad 2 via the dresser shaft 52. Furthermore, the dresser shaft 52 is connected to a dresser rotation mechanism (including, for example, an electric motor) (not shown) disposed within the dresser swing arm 55. This dresser rotation mechanism is configured to rotate the dresser 51 via the dresser shaft 52 in the direction indicated by the arrow.

[0024] The polishing surface 2a of the polishing pad 2 is dressed as follows. While the polishing pad 2 is rotated together with the polishing table 5 by the table rotation motor 21, pure water is supplied to the polishing surface 2a from a pure water supply nozzle (not shown). While the dresser 51 is rotated about the dresser shaft 52 by a dresser rotation mechanism (not shown), the dressing surface 51a of the dresser 51 is pressed against the polishing surface 2a by a dresser pressing mechanism (not shown). With pure water present on the polishing surface 2a, the dresser 51 is brought into sliding contact with the polishing surface 2a. While the dresser 51 is rotating, the dresser swing arm 55 is rotated about the support shaft 58 to swing the dresser 51 in the radial direction of the polishing surface 2a. In this way, the polishing pad 2 is slightly scraped away by the dressing surface 51a of the dresser 51, and the polishing surface 2a is dressed (regenerated).

[0025] Dressing of the polishing pad 2 is carried out before the new polishing pad 2 is used for polishing a wafer for the first time, or while polishing a wafer, or after polishing a wafer but before polishing the next wafer.

[0026] The dressing of the polishing pad 2 is performed under predetermined dressing conditions, which include at least one of the rotation speed of the dresser 51, the rotation speed of the polishing pad 2 (polishing table 5), the dressing time, and the oscillation speed of the dresser 51 on the polishing surface 2 a.

[0027] During dressing of the polishing pad 2, the abrasive grains constituting the dressing surface 51a move over the polishing surface 2a of the polishing pad 2, scratching the polishing surface 2a and roughening the polishing surface 2a. Figure 2 is a schematic diagram showing an example of the trajectory of the abrasive grains constituting the dressing surface 51a on the polishing surface 2a of the polishing pad 2. The symbol T shown in Figure 2 represents the trajectory of the abrasive grains on the polishing surface 2a. As shown in Figure 2, as the abrasive grains scratch the polishing surface 2a, a portion of the polishing surface 2a rises to form a protrusion 40. Similarly, at the intersections where multiple trajectories T of the abrasive grains intersect, a portion of the polishing surface 2a rises to form a protrusion 40.

[0028] At the intersections of the loci T, the multiple loci intersect at an intersection angle greater than 0 degrees. If the intersection angle is 0 degrees, the protrusions 40 formed by one abrasive grain may be removed by another abrasive grain. Therefore, the intersection angle of the loci T of the abrasive grains is greater than 0 degrees, for example, in the range of 10 to 170 degrees.

[0029] The numerous protrusions 40 formed by the abrasive grains of the dresser 51 contribute to the mechanical polishing of the wafer. Therefore, the state of the polishing surface 2a of the polishing pad 2 can affect the wafer polishing rate. To improve and stabilize the wafer polishing rate, it is desirable that the trajectories of the abrasive grains are uniformly distributed on the polishing pad 2, and that the intersections of the abrasive grain trajectories are uniformly distributed on the polishing pad 2. The trajectories of the abrasive grains on the polishing surface 2a can change depending on the dressing conditions.

[0030] 1, the polishing apparatus further includes a dressing condition determination device 60 that determines dressing conditions in order to create a desired state of the polishing surface 2a of the polishing pad 2. The dressing condition determination device 60 is connected to the operation control unit 30.

[0031] The dressing condition determination device 60 is composed of at least one computer. The dressing condition determination device 60 includes a storage device 60a that stores a program and an estimation model 62, and a processor 60b that executes calculations in accordance with instructions included in the program and the algorithm of the estimation model 62. The storage device 60a includes a main storage device such as a random access memory (RAM) and an auxiliary storage device such as a hard disk drive (HDD) or a solid state drive (SSD). Examples of the processor 60b include a CPU (central processing unit) and a GPU (graphics processing unit). However, the specific configuration of the dressing condition determination device 60 is not limited to these examples.

[0032] In one embodiment, the operation control unit 30 and the dressing condition determination device 60 may be integrally formed. For example, the operation control unit 30 and the dressing condition determination device 60 may be configured as a single computer.

[0033] The estimation model 62 is configured to output, when a target trajectory image showing a target trajectory on the polishing pad 2 of the multiple abrasive grains that make up the dressing surface 51a is input, dressing conditions for realizing the target trajectory. The estimation model 62 is a trained model constructed by machine learning. An example of the estimation model 62 is a neural network.

[0034] The target trajectory image is a simulation image generated by executing a simulation of an operation of dressing the polishing pad 2 by pressing the dressing surface 51 a of the rotating dresser 51 against the rotating polishing pad 2. The simulation may be executed by the dressing condition determination device 60 or another simulation device.

[0035] 3 is a diagram showing an example of a target trajectory image. In this example, the trajectories of the target abrasive grains appearing in the target trajectory image are uniformly distributed on the polishing pad 2, and the intersections of the trajectories of the abrasive grains are uniformly distributed on the polishing pad 2. In other words, during dressing of the polishing surface 2a of the polishing pad 2, the multiple abrasive grains constituting the dressing surface 51a of the dresser 51 move uniformly on the polishing surface 2a, forming trajectories with a geometric pattern. The multiple abrasive grains move on the polishing surface 2a with uniformly distributed travel distances.

[0036] The target trajectory image is input to a dressing condition determination device 60. The dressing condition determination device 60 inputs the target trajectory image to an estimation model 62 and causes the estimation model 62 to output dressing conditions for realizing a target trajectory, which is the trajectory of abrasive grains that appears in the target trajectory image. The dressing conditions include at least one of the rotation speed of the dresser 51, the rotation speed of the polishing pad 2, the oscillation speed of the dresser 51 on the polishing pad 2, and the dressing time.

[0037] FIG. 4 is a schematic diagram showing one embodiment of the estimation model 62. The estimation model 62 includes an encoder 65 that extracts features of a target trajectory image, a decoder 67 that restores the target trajectory image from the features, and a regression analyzer 69 that performs regression analysis on the extracted features and outputs dressing conditions. The encoder 65 has an input layer 65a to which the target trajectory image is input. The decoder 67 has an output layer 67a that outputs an image substantially identical to the target trajectory image input to the encoder 65. The encoder 65 is configured to compress the features of the input target trajectory image and extract the features of the target trajectory image. The decoder 67 is configured to restore the target trajectory image from the features extracted by the encoder 65. The combination of the encoder 65 and the decoder 67 constitutes an autoencoder.

[0038] The regression analyzer 69 is configured to perform regression analysis on the feature quantities extracted (compressed) by the encoder 65 and output dressing conditions that can realize the target trajectory, which is the trajectory of the abrasive grains on the target trajectory image input to the encoder 65. The dressing conditions output from the regression analyzer 69 include at least one of the rotation speed of the dresser 51, the rotation speed of the polishing pad 2, the oscillation speed of the dresser 51 on the polishing pad 2, and the dressing time.

[0039] The estimation model 62 is a trained model constructed by machine learning. More specifically, the estimation model 62 is a trained model constructed by machine learning using training data including a plurality of dressing conditions and a plurality of trajectory images obtained by dressing simulations performed under the plurality of dressing conditions. The dressing simulation is a simulation of the operation of dressing the polishing pad 2 with the dressing surface 51 a of the dresser 51.

[0040] Each trajectory image included in the training data is an image showing the trajectory of abrasive grains constituting the dressing surface 51a in the dressing simulation. The plurality of dressing conditions included in the training data are objective variables (i.e., correct labels), and the plurality of trajectory images corresponding to the plurality of dressing conditions are explanatory variables. The dressing simulation may be performed by the dressing condition determination device 60 or another simulation device.

[0041] 5 is a schematic diagram showing an example of training data including a plurality of dressing conditions and a plurality of corresponding trajectory images. A dressing simulation is performed under each dressing condition, and abrasive grain trajectory images corresponding to each dressing condition are obtained. The plurality of dressing conditions A to F shown in FIG. 5 are different from one another, and as a result, the plurality of trajectory images A to F obtained by the dressing simulation are also different from one another. The plurality of dressing conditions A to F included in the training data correspond one-to-one to the plurality of trajectory images A to F included in the training data.

[0042] The dressing condition determination device 60 executes machine learning using the training data to construct an estimation model 62. An example of the machine learning is deep learning, and an example of the estimation model 62 is a neural network. However, the specific details of the machine learning and the estimation model 62 are not particularly limited as long as the intended purpose can be achieved.

[0043] 6 is a schematic diagram illustrating the training phase in which an estimation model 62 is constructed by machine learning using training data. When one of multiple trajectory images included in the training data is input to the input layer 65a of the encoder 65, a trajectory image is output from the output layer 67a of the decoder 67, and dressing conditions are output from the regression analyzer 69. Machine learning adjusts parameters (weights, thresholds, etc.) of each node (neuron) of the encoder 65, the decoder 67, and the regression analyzer 69 so that the trajectory image output from the decoder 67 matches the trajectory image input to the encoder 65, and the dressing conditions output from the regression analyzer 69 match the dressing conditions corresponding to the trajectory image input to the encoder 65. When a trajectory image (e.g., the target trajectory image) not included in the training data is input, the estimation model 62 constructed by machine learning can output dressing conditions for realizing the trajectory of the abrasive grains on the input trajectory image.

[0044] 7 is a schematic diagram illustrating the practical phase in which an estimation model 62 constructed by machine learning is used to determine dressing conditions for realizing the trajectory of abrasive grains on a target trajectory image. A target trajectory image created in advance by simulation is input to an input layer 65a of an encoder 65 of the estimation model 62. A regression analyzer 69 of the estimation model 62 outputs dressing conditions that can realize the trajectory of abrasive grains on the target trajectory image (i.e., the target trajectory).

[0045] The dressing conditions output from the estimation model 62 are transmitted to the operation control unit 30 shown in Fig. 1. The operation control unit 30 controls the operation of the polishing apparatus including the dressing mechanism 50 in accordance with the dressing conditions, causing the dressing mechanism 50 to dress the polishing surface 2a of the polishing pad 2. The dressing operation in accordance with the dressing conditions determined by the dressing condition determination device 60 can form the trajectory of abrasive grains on the polishing surface 2a of the polishing pad 2, as shown in the target trajectory image in Fig. 7.

[0046] The dressing condition determining device 60 and the dressing condition determining method described above can automatically determine dressing conditions that can realize a desired state of the polishing surface 2 a using the estimation model 62 .

[0047] Dressing of the polishing pad 2 is performed at several times. For example, a new polishing pad 2 is dressed by the dresser 51 before it is used for polishing. This dressing operation is also called break-in dressing. In another example, the polishing pad 2 is dressed by the dresser 51 after a wafer has been polished and before the next wafer is polished. In yet another example, the polishing pad 2 is dressed by the dresser 51 while the wafer is being polished.

[0048] The above-described embodiments have been described for the purpose of enabling a person of ordinary skill in the art to practice the present invention. Various modifications of the above-described embodiments would be obvious to a person skilled in the art, and the technical concept of the present invention may be applied to other embodiments. Therefore, the present invention is not limited to the described embodiments, but is to be interpreted in the broadest scope in accordance with the technical concept defined by the claims.

[0049] The present invention can be used in a technique for determining dressing conditions for dressing a polishing pad used in polishing a workpiece such as a wafer, a substrate, or a panel.

[0050] W Wafer T Abrasive grain trajectory 2 Polishing pad 2a Polishing surface 5 Polishing table 5a Table shaft 7 Polishing head 8 Polishing liquid supply nozzle 14 Support shaft 16 Polishing head swing arm 18 Polishing head shaft 21 Table rotation motor 30 Operation control unit 30a Storage device 30b Arithmetic unit 40 Protrusion 50 Dressing mechanism 51 Dresser 51a Dressing surface 52 Dresser shaft 55 Dresser swing arm 60 Dressing condition determination device 60a Storage device 60b Processor 62 Estimation model 65 Encoder 67 Decoder 69 Regression analyzer

Claims

1. A dressing condition determination device that determines dressing conditions for dressing a polishing pad by pressing the dressing surface of a rotating dresser against the rotating polishing pad, the dressing condition determination device comprising: a storage device that stores an estimation model constructed by machine learning; and a processor that executes calculations according to an algorithm of the estimation model, wherein the estimation model is configured to output dressing conditions for realizing a target trajectory image showing a target trajectory on the polishing pad of a plurality of abrasive grains that constitute the dressing surface when the target trajectory image is input.

2. The dressing condition determining apparatus according to claim 1, wherein the dressing conditions include at least one of a rotation speed of the dresser, a rotation speed of the polishing pad, a swing speed of the dresser on the polishing pad, and a dressing time.

3. The dressing condition determination device of claim 1, wherein the target trajectory image is a simulation image of the trajectories of the multiple abrasive grains generated by performing a simulation of the operation of dressing the polishing pad by pressing the dressing surface of the rotating dresser against the rotating polishing pad.

4. The dressing condition determination device of claim 1, wherein the estimation model is a trained model constructed by machine learning using training data including a plurality of dressing conditions and a plurality of trajectory images obtained by a dressing simulation performed under the plurality of dressing conditions.

5. The dressing condition determination device of claim 1, wherein the estimation model comprises an encoder for extracting features of the target trajectory image, a decoder for restoring the target trajectory image from the features, and a regression analyzer for performing regression analysis on the extracted features to output the dressing conditions.

6. The dressing condition determination device according to claim 1, wherein the target trajectory is a trajectory in which the trajectories of the abrasive grains are uniformly distributed on the polishing pad and the intersection points of the trajectories of the abrasive grains are uniformly distributed on the polishing pad.

7. A dressing condition determination method for determining dressing conditions for dressing a polishing pad by pressing a dressing surface of a rotating dresser against a rotating polishing pad, the dressing condition determination method comprising: inputting a target trajectory image showing a target trajectory on the polishing pad of a plurality of abrasive grains constituting the dressing surface into an estimation model constructed by machine learning; and outputting dressing conditions for realizing the target trajectory from the estimation model.

8. The method of determining a dressing condition according to claim 7, wherein the dressing condition includes at least one of a rotation speed of the dresser, a rotation speed of the polishing pad, a swing speed of the dresser on the polishing pad, and a dressing time.

9. The dressing condition determination method of claim 7, wherein the target trajectory image is a simulation image of the trajectories of the multiple abrasive grains generated by performing a simulation of an operation of dressing the polishing pad by pressing the dressing surface of the rotating dresser against the rotating polishing pad.

10. The dressing condition determination method described in claim 7, wherein the estimation model is a trained model constructed by machine learning using training data including a plurality of dressing conditions and a plurality of trajectory images obtained by a dressing simulation performed under the plurality of dressing conditions.

11. The dressing condition determination method of claim 7, wherein the estimation model comprises an encoder for extracting features of the target trajectory image, a decoder for restoring the target trajectory image from the features, and a regression analyzer for performing regression analysis on the extracted features to output the dressing conditions.

12. A method for determining dressing conditions according to claim 7, wherein the target trajectory is a trajectory in which the trajectories of the abrasive grains are uniformly distributed on the polishing pad and the intersection points of the trajectories of the abrasive grains are uniformly distributed on the polishing pad.

Images