METHOD FOR ESTIMATING LIFE OF POLISHING PAD AND POLISHING DEVICE

Abstract

A method for estimating a life of a polishing pad includes: polishing a workpiece by pressing the workpiece against a polishing surface of a polishing pad with an elastic membrane forming a pressure chamber provided in a polishing head; controlling a pressure in the pressure chamber on the basis of a measured value of a film thickness of the workpiece while measuring the film thickness during the polishing of the workpiece; inputting time-series pressure data representing a change in the pressure in the pressure chamber during the polishing of the workpiece into a learned model; and outputting a life index of the polishing pad from the learned model.

Description

CROSS-REFERENCE TO RELATED APPLICATION

This application claims the priority benefit of Japan application serial no. 2022-086935, filed on May 27, 2022. The entirety of the above-mentioned patent application is hereby incorporated by reference herein and made a part of this specification.

BACKGROUND

Technical Field

The disclosure relates to a technique for estimating the life of a polishing pad used in a polishing device for polishing a workpiece such as a wafer, a substrate, or a panel.

Description of Related Art

Chemical mechanical polishing (hereinafter referred to as CMP) is a process of bringing a workpiece (for example, a wafer, a substrate, a panel, or the like) into sliding contact with a polishing pad while supplying a polishing liquid containing abrasive grains such as silica (SiO₂) onto the polishing pad, thereby polishing the workpiece. A polishing device for performing this CMP includes a polishing table that supports a polishing pad having a polishing surface, and a polishing head for pressing a workpiece against the polishing pad.

The polishing device polishes the workpiece as follows. While the polishing table and the polishing pad are rotated together, a polishing liquid (typically slurry) is supplied to the polishing surface of the polishing pad. While rotating the workpiece, the polishing head presses a surface of the workpiece against the polishing surface of the polishing pad. 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 a chemical action of the polishing liquid and a mechanical action of abrasive grains contained in the polishing liquid and/or the polishing pad.

When the workpiece is polished, abrasive grains and polishing chips adhere to the polishing surface of the polishing pad, and polishing performance of the polishing pad deteriorates. Thus, in order to regenerate the polishing surface of the polishing pad, dressing of the polishing pad is performed by a dresser 50. The dresser 50 has hard abrasive grains such as diamond particles fixed to its lower surface, and by slightly scraping off the polishing surface of the polishing pad with the dresser 50, the polishing surface of the polishing pad is regenerated. The polishing pad gradually wears out as the dressing is repeated. When the polishing pad wears out, intended polishing performance cannot be obtained, and thus it is required to periodically replace the polishing pad. Thus, when a usage time of the polishing pad exceeds a predetermined time, or when the number of polished workpieces exceeds a predetermined number, the polishing pad is replaced with a new one.

However, the usage time of the polishing pad and the number of polished workpieces only indirectly indicate wear of the polishing pad and may not accurately reflect the wear of the polishing pad. As a result, a polishing pad that has not yet reached the end of its life may be replaced, or a polishing pad that has been worn beyond its usage limit may continue to be used. In particular, when an excessively worn polishing pad is used, a target film thickness profile of a workpiece may not be achieved.

RELATED ART DOCUMENTS

Patent Documents

• [Patent Document 1] Japanese Patent Application Laid-Open No. 2001-334461
• [Patent Document 2] Japanese Patent Application Laid-Open No. 2000-288915

SUMMARY

In one aspect, a method for estimating the life of a polishing pad is provided, the method including: polishing a workpiece by pressing the workpiece against a polishing surface of a polishing pad with an elastic membrane forming a pressure chamber provided in a polishing head; controlling a pressure in the pressure chamber on the basis of a measured value of a film thickness of the workpiece while measuring the film thickness during the polishing of the workpiece; inputting time-series pressure data representing a change in the pressure during the polishing of the workpiece into a learned model; and outputting a life index of the polishing pad from the learned model.

In one aspect, in addition to the time-series pressure data, polishing data relating to the polishing of the workpiece is input to the learned model. In one aspect, the polishing data includes at least one of information of the workpiece and polishing conditions for the workpiece. In one aspect, in addition to the time-series pressure data, a cut rate of the polishing pad which has been dressed by a dresser is input to the learned model. In one aspect, the steps of inputting the time-series pressure data into the learned model and outputting the life index of the polishing pad from the learned model are performed after the polishing of the workpiece and before polishing of a next workpiece. In one aspect, the pressure chamber is a plurality of pressure chambers, the step of controlling the pressure in the pressure chamber is a step of controlling pressures in the plurality of pressure chambers on the basis of the measured value of the film thickness of the workpiece while measuring the film thickness during the polishing of the workpiece, and the step of inputting the time-series pressure data into the learned model is a step of inputting into the learned model a plurality of pieces of time-series pressure data respectively representing changes in the pressures in the plurality of pressure chambers during the polishing of the workpiece.

In one aspect, a polishing device includes: a polishing head that includes a pressure chamber formed by an elastic membrane and polishes a workpiece by pressing the workpiece against a polishing surface of a polishing pad with the elastic membrane; a film thickness sensor that measures a film thickness of the workpiece; a polishing control part that controls a pressure in the pressure chamber during the polishing of the workpiece on the basis of a measured value of the film thickness; and a pad life calculation part that inputs time-series pressure data representing a change in the pressure during the polishing of the workpiece into a learned model and outputs a life index of the polishing pad from the learned model is provided.

In one aspect, the pad life calculation part is configured to input polishing data relating to the polishing of the workpiece into the learned model in addition to the time-series pressure data. In one aspect, the polishing data includes at least one of information of the workpiece and polishing conditions for the workpiece. In one aspect, the polishing device further includes a dresser that dresses the polishing surface of the polishing pad, and the pad life calculation part is configured to input a cut rate of the polishing pad dressed by the dresser into the learned model in addition to the time-series pressure data. In one aspect, the pad life calculation part is configured to input the time-series pressure data into the learned model and output the life index of the polishing pad from the learned model after the polishing of the workpiece and before polishing of a next workpiece. In one aspect, the pressure chamber is a plurality of pressure chambers, the polishing control part is configured to control pressures in the plurality of pressure chambers on the basis of the measured value of the film thickness of the workpiece while measuring the film thickness during the polishing of the workpiece, and the pad life calculation part is configured to input into the learned model a plurality of pieces of time-series pressure data respectively representing changes in the pressures in the plurality of pressure chambers during the polishing of the workpiece.

The change in the pressure in the pressure chamber of the polishing head during the polishing of the workpiece reflects a degree of wear of the polishing pad. Accordingly, the learned model can accurately estimate the life of the polishing pad from the time-series pressure data indicating the change in the pressure in the pressure chamber of the polishing head.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram showing one embodiment of a polishing device.

FIG. 2 is a cross-sectional view showing one embodiment of a polishing head.

FIG. 3 is a graph showing an example of time-series pressure data showing a change over time in pressure in a pressure chamber of the polishing head when a polishing pad is new.

FIG. 4 is a graph showing an example of time-series pressure data showing a change over time in the pressure in the pressure chamber of the polishing head when the polishing pad is significantly worn.

FIG. 5 is a graph showing another example of time-series pressure data showing a change over time in the pressure in the pressure chamber of the polishing head when the polishing pad is significantly worn.

FIG. 6 is a diagram showing an example of a learned model.

FIG. 7 is a diagram representing an example of a relationship between the number of wafers and a usage time of a polishing pad when a plurality of wafers is polished until one polishing pad reaches the end of its life.

FIG. 8 is a diagram representing an example of a relationship between the number of wafers and a usage time of each polishing pad when a plurality of wafers is polished using a plurality of polishing pads until each polishing pad reaches the end of its life.

DESCRIPTION OF THE EMBODIMENTS

Embodiments of the disclosure provide an improved technique that can accurately estimate the life of a polishing pad.

Embodiments of the disclosure will be described below with reference to the drawings. FIG. 1 is a schematic diagram showing one embodiment of a polishing device. The polishing device is a device that chemically and mechanically polishes a wafer W, which is an example of a workpiece used in manufacturing of semiconductor devices. As shown in FIG. 1, this polishing device 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, and a polishing liquid supply nozzle 8 that supplies a polishing liquid (for example, slurry containing abrasive grains) to the polishing surface 2a.

The polishing head 7 is configured such that it can hold the wafer W on its lower surface. The wafer W has a polishing target film. In the following embodiments, a wafer is used as an example of a workpiece, but the workpiece is not limited to the wafer and may be a circular substrate, a rectangular substrate, a panel, or the like used in manufacturing of semiconductor devices.

The polishing device further includes a support shaft 14, a polishing head swing arm 16 connected to an upper end of the support shaft 14, and a polishing head shaft 18 rotatably supported by a free end of the polishing head swing arm 16. The polishing head 7 is fixed to a lower end of the polishing head shaft 18. A polishing head rotating mechanism (not shown) including an electric motor and the like is disposed in the polishing head swing arm 16. This polishing head rotating mechanism is connected to the polishing head shaft 18 and configured to rotate the polishing head shaft 18 and the polishing head 7 in a direction indicated by the arrow.

The polishing head shaft 18 is connected to a polishing head elevating mechanism (not shown) (including a ball screw mechanism and the like). This polishing head elevating mechanism is configured to move the polishing head shaft 18 vertically relative to the polishing head swing arm 16. This vertical movement of the polishing head shaft 18 allows the polishing head 7 to move vertically relative to the polishing head swing arm 16 and the polishing table 5 as indicated by the arrows.

The polishing device further includes a table rotation motor 21 that rotates the polishing pad 2 and the polishing table 5 about their 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 configured to be rotated by the table rotation motor 21 about the table shaft 5a in a direction indicated by the arrow. The polishing pad 2 is attached to an upper surface of the polishing table 5. An exposed surface of the polishing pad 2 constitutes the polishing surface 2a that polishes the wafer W.

Polishing of the wafer W is performed as follows. The wafer W is held by the polishing head 7 with its polishing target surface facing downward. While the polishing head 7 and the polishing table 5 are being rotated, the polishing liquid (for example, slurry containing abrasive grains) is supplied onto the polishing surface 2a of the polishing pad 2 from the polishing liquid supply nozzle 8 provided above the polishing table 5. The polishing pad 2 rotates integrally with the polishing table 5 about its central axis. The polishing head 7 is moved to a predetermined height by the polishing head elevating mechanism (not shown). Further, the polishing head 7 presses the wafer W against the polishing surface 2a of the polishing pad 2 while being maintained at the predetermined height. The wafer W rotates together with the polishing head 7. In a state in which the polishing liquid is on the polishing surface 2a of the polishing pad 2, the wafer W is brought into sliding contact with the polishing surface 2a. A surface of the wafer W is polished by a combination of a chemical action of the polishing liquid and a mechanical action of the abrasive grains contained in the polishing liquid and/or the polishing pad 2.

The polishing device includes a film thickness sensor 24 that measures a film thickness of the wafer W on the polishing surface 2a. The film thickness sensor 24 is configured to produce a measured value of the film thickness that directly or indirectly indicates the film thickness of the wafer W. This measured value of the film thickness varies depending on the film thickness of the wafer W and indicates the film thickness of the wafer W. The measured value of the film thickness may be a value representing the film thickness of the wafer W itself, or may be a physical quantity or a signal value that will be converted into the film thickness.

Examples of the film thickness sensor 24 include an optical film thickness sensor and an eddy current sensor. The optical film thickness sensor is configured to irradiate the surface of the wafer W with light and determine the film thickness of the wafer W from a spectrum of reflected light from the wafer W. The eddy current sensor is configured to induce an eddy current in a conductive film formed on the wafer W and output a signal value that varies in accordance with impedance of an electrical circuit including the conductive film and a coil of the eddy current sensor. Known devices can be used for the optical film thickness sensor and the eddy current sensor.

The film thickness sensor 24 is installed inside the polishing table 5 and rotates together with the polishing table 5. More specifically, the film thickness sensor 24 is configured to measure the film thickness at a plurality of measurement points on the wafer W while traversing the wafer W on the polishing surface 2a each time the polishing table 5 rotates once. In the present embodiment, the film thickness sensor 24 is disposed to measure the film thickness at a plurality of measurement points including a center of the wafer W. Accordingly, the plurality of measurement points are arranged in a radial direction of the wafer W.

The polishing device further includes a polishing control part 30 that controls a pressure in a pressure chamber (which will be described later) of the polishing head 7 during the polishing of the wafer W on the basis of measured values of the film thickness obtained by the film thickness sensor 24, and a pad life calculation part 40 that inputs time-series pressure data representing a change in pressure in the pressure chamber of the polishing head 7 during the polishing of the wafer W into a learned model 42 and outputs a life index of the polishing pad from the learned model 42.

The polishing control part 30 is configured of at least one computer. The polishing control part 30 includes a storage device 30a that stores a program for controlling the pressure in the pressure chamber of the polishing head 7 on the basis of the measured values of the film thickness, and a computing device 30b that performs computing in accordance with 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 solid state drive (SSD). Examples of the computing device 30b include a central processing unit (CPU) and a graphics processing unit (GPU). However, a specific configuration of the polishing control part 30 is not limited to these examples.

The pad life calculation part 40 is configured of at least one computer. The pad life calculation part 40 includes a storage device 40a that stores the learned model 42, a program for calculating the life of the polishing head 7 using the learned model 42, and a program that performs machine learning for constructing the learned model 42, and a computing device 40b that performs computing in accordance with instructions included in these programs. The storage device 40a 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 solid state drive (SSD). Examples of the computing device 40b include a central processing unit (CPU) and a graphics processing unit (GPU). However, a specific configuration of the pad life calculation part 40 is not limited to these examples.

Each of the polishing control part 30 and the pad life calculation part 40 may be configured of a plurality of computers. For example, each of the polishing control part 30 and the pad life calculation part 40 may be configured of a combination of an edge server and a cloud server.

The film thickness sensor 24 is connected to the polishing control part 30. The measured values of the film thickness generated by the film thickness sensor 24 are sent to the polishing control part 30. That is, the measured values of the film thickness at the plurality of measurement points of the wafer W are output from the film thickness sensor 24, sent to the polishing control part 30, and stored in the storage device 30a. The pad life calculation part 40 is connected to the polishing control part 30.

The polishing device includes a dresser 50 that dresses the polishing surface 2a of the polishing pad 2. This dresser 50 includes a dressing disc 51 in sliding contact with the polishing surface 2a of the polishing pad 2, a dresser shaft 52 to which the dressing disc 51 is connected, and a dresser swing arm 55 that rotatably supports the dresser shaft 52. A lower surface of the dressing disc 51 constitutes a dressing surface 51a, and this dressing surface 51a is configured of abrasive grains (for example, diamond grains).

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

Dressing of the polishing surface 2a of the polishing pad 2 is performed as follows. While the polishing pad 2 is rotated by the table rotation motor 21 together with the polishing table pure water is supplied to the polishing surface 2a from a pure water supply nozzle (not shown). While the dressing disc 51 is rotated about the dresser shaft 52 by the disc rotating mechanism (not shown), the dressing surface 51a of the dressing disc 51 is pressed against the polishing surface 2a by the disc pressing mechanism (not shown). The dressing disc 51 is brought into sliding contact with the polishing surface 2a with the pure water on the polishing surface 2a. While the dressing disc 51 is rotating, the dresser swing arm 55 is swung around a support shaft 58 to swing the dressing disc 51 in a radial direction of the polishing surface 2a. In this manner, the polishing pad 2 is slightly scraped off by the dressing disc 51, and the polishing surface 2a is dressed (regenerated). The dressing of the polishing surface 2a of the polishing pad 2 is performed during the polishing of the wafer W or after the polishing of the wafer W.

Next, the polishing head 7 will be explained. FIG. 2 is a cross-sectional view showing one embodiment of the polishing head 7. The polishing head 7 includes a head body 61 fixed to an end portion of the polishing head shaft 18, an elastic membrane 64 attached to a lower portion of the head body 61, and a retainer ring 62 disposed below the head body 61. The retainer ring 62 is disposed around the elastic membrane 64. This retainer ring 62 is an annular structure that holds the wafer W in order to prevent the wafer W from jumping out of the polishing head 7 during the polishing of the wafer W.

Four pressure chambers C1, C2, C3, and C4 are provided between the elastic membrane 64 and the head body 61. The pressure chambers C1, C2, C3, and C4 are formed by the elastic membrane 64 and the head body 61. The central pressure chamber C1 is circular and the other pressure chambers C2, C3, and C4 are annular. These pressure chambers C1, C2, C3, and C4 are arranged concentrically.

Gas transfer lines F1, F2, F3, and F4 are connected to the pressure chambers C1, C2, C3, and C4. One ends of the gas transfer lines F1, F2, F3, and F4 are connected to a compressed gas supply (not shown) serving as a utility provided in a factory in which the polishing device is installed. Compressed gas such as compressed air is supplied to the pressure chambers C1, C2, C3, and C4 through the gas transfer lines F1, F2, F3, and F4. The compressed gas in the pressure chambers C1, C2, C3, and C4 presses the wafer W against the polishing surface 2a of the polishing pad 2 via the elastic membrane 64.

The gas transfer line F3 communicating with the pressure chamber C3 is connected to a vacuum line (not shown) so that a vacuum can be formed in the pressure chamber C3. When a vacuum is formed in the pressure chamber C3, the elastic membrane 64 constituting the pressure chamber C3 is depressed upward. As a result, the polishing head 7 can hold the wafer W with the deformed elastic membrane 64 like a suction cup. In one embodiment, an opening may be formed at a portion of the elastic membrane 64 forming the pressure chamber C3, and the wafer W may be sucked and held by the polishing head 7 by forming a vacuum in the pressure chamber C3.

An annular elastic membrane 66 is disposed between the head body 61 and the retainer ring 62, and a pressure chamber C5 is formed inside the elastic membrane 66. The pressure chamber C5 is connected to the compressed gas supply via a gas transfer line F5. The compressed gas is supplied into the pressure chamber C5 through the gas transfer line F5, and the compressed gas in the pressure chamber C5 presses the retainer ring 62 against the polishing pad 2.

The gas transfer lines F1, F2, F3, F4, and F5 extend through a rotary joint 70 attached to the polishing head shaft 18. The gas transfer lines F1, F2, F3, F4, and F5 communicating with the pressure chambers C1, C2, C3, C4, and C5 are provided with pressure regulators R1, R2, R3, R4, and R5. The compressed gas from the compressed gas supply is supplied independently into the pressure chambers C1 to C5 through the pressure regulators R1 to R5. The pressure regulators R1 to R5 are configured to regulate pressures of the compressed gas in the pressure chambers C1 to C5.

The pressure regulators R1 to R5 can change internal pressures of the pressure chambers C1 to C5 independently from each other, and thus, pressing pressures against four corresponding regions of the wafer W, that is, a central portion, an inner intermediate portion, an outer intermediate portion, and an edge portion thereof, and a pressing pressure of the retainer ring 62 against the polishing pad 2 can be adjusted independently. The gas transfer lines F1, F2, F3, F4, and F5 are also connected to air release valves (not shown) so that the pressure chambers C1 to C5 can be opened to the atmosphere. In the present embodiment, the elastic membrane 64 forms the four pressure chambers C1 to C4, but in one embodiment, the elastic membrane 64 may form three or less, or five or more pressure chambers. Only a single pressure chamber may be provided.

The pressure regulators R1 to R5 are connected to the polishing control part 30. The polishing control part 30 is configured to control the pressures in the pressure chambers C1 to C5 via the pressure regulators R1 to R5 on the basis of the measured values of the film thickness of the wafer W. More specifically, the polishing control part 30 receives a plurality of measured values of the film thickness of the wafer W from the film thickness sensor 24 (see FIG. 1), determines target pressure values for each of the pressure chambers C1 to C5 to achieve a target film thickness on the basis of the plurality of measured values of the film thickness, and sends the target pressure values to the pressure regulators R1 to R5. The pressure regulators R1 to R5 operate to maintain the pressures in the pressure chambers C1 to C5 at corresponding target pressure values. Accordingly, the target pressure values for the pressure chambers C1 to C5 determined by the polishing control part 30 correspond to actual pressures in the pressure chambers C1 to C5. In one embodiment, the pressures in the pressure chambers C1 to C5 may be measured by pressure sensors.

The polishing head 7 can apply separate pressures to a plurality of regions of the wafer W. For example, the polishing head 7 can press different regions of the surface of the wafer W against the polishing surface 2a of the polishing pad 2 with different pressures. Accordingly, the polishing head 7 can control a film thickness profile of the wafer W to achieve a target film thickness profile.

In one embodiment, the polishing control part 30 calculates a plurality of average polishing rates of a plurality of regions on the wafer W corresponding to the pressure chambers C1 to C4 from the measured values of the film thickness of the wafer W being polished, calculates a difference between the average polishing rate of each region and a predetermined target polishing rate, and determines the pressure values in the pressure chambers C1 to C4 to reduce the difference. For example, in a case in which the average polishing rate of the region corresponding to the pressure chamber C1 is below the target polishing rate, the pressure of the pressure chamber C1 is increased.

In another embodiment, the polishing control part 30 calculates a plurality of average values of the film thickness of a plurality of regions on the wafer W corresponding to the pressure chambers C1 to C4 from the measured values of the film thickness of the wafer W being polished, calculates a difference between the average value of the film thickness of each region and a predetermined target film thickness value that changes with the progress of polishing, and determines target pressure values in the pressure chambers C1 to C4 required to eliminate the difference.

The wafer W is polished under preset polishing conditions. The polishing conditions include a rotation speed of the polishing table 5, a rotation speed of the polishing head 7, initial pressures in the pressure chambers C1 to C4 of the polishing head 7, a type and a flow rate of the polishing liquid supplied to the polishing surface 2a of the polishing pad 2, and so on. If the polishing conditions are constant, polishing rates (also called removal rates) of wafers of the same type are generally the same. However, when the polishing pad 2 wears down, grooves formed on the polishing surface 2a become shallower, making it difficult for the polishing liquid to be held on the polishing surface 2a. As a result, polishing rates of wafers are lowered even though the polishing conditions are the same. Accordingly, when wear of the polishing pad 2 progresses, it is required to replace the polishing pad 2 with a new one.

The pad life calculation part 40 shown in FIG. 1 is configured to estimate the life of the polishing pad 2 using the learned model 42. The time-series pressure data representing changes in pressure in the pressure chambers of the polishing head 7 acquired during the polishing of the wafer W is input to the learned model 42. The pad life calculation part 40 performs calculation in accordance with an algorithm defined by the learned model 42 to determine the life index of the polishing pad 2.

The pressures in the time-series pressure data are pressures in at least one of the pressure chambers C1 to C4 described above. In the embodiments described below, the time-series pressure data indicating a change in pressure in the pressure chamber C1 is used, but in other embodiments, time-series pressure data indicating a change in pressure in the pressure chamber C2, C3, or C4 other than the pressure chamber C1 may be input to the learned model 42. In yet other embodiments, a plurality of pieces of time-series pressure data representing changes in pressure in a plurality of pressure chambers of the pressure chambers C1 to C4 or all pressure chambers C1 to C4 may be input to the learned model 42.

The time-series pressure data input to the learned model 42 is data indicating a change in pressure in the pressure chamber C1 of the polishing head 7 with the lapse of polishing time when one wafer is being polished. More specifically, the time-series pressure data is data indicating a change in pressure in the pressure chamber C1 from the start of polishing of one wafer to the end of polishing. This time-series pressure data may change depending on a state of wear of the polishing pad 2. That is, when the polishing pad 2 is new and there are grooves on the polishing surface 2a, the polishing liquid (for example, slurry) supplied onto the polishing surface 2a is held in the grooves. Accordingly, during the polishing of the wafer W, a sufficient amount of polishing liquid is between the wafer W and the polishing surface 2a of the polishing pad 2. The polishing head 7 can polish the wafer W at a high polishing rate while the pressure in the pressure chamber C1 are maintained within an appropriate range.

FIG. 3 is a graph showing an example of the time-series pressure data showing a change over time in pressure in the pressure chamber C1 of the polishing head 7 during the polishing of the wafer W when the polishing pad 2 is new. During the polishing of the wafer W, the pressures in the plurality of pressure chambers C1 to C4 (see FIG. 2) of the polishing head 7 change to eliminate variations in film thickness in the surface of the wafer W and change such that the film thickness of the entire wafer W reaches the target film thickness. In accordance with such pressure control, as shown in FIG. 3, the pressure in the pressure chamber C1 fluctuates somewhat, but generally changes within the appropriate range.

FIG. 4 is a graph showing an example of the time-series pressure data showing a change over time in pressure in the pressure chamber C1 of the polishing head 7 during the polishing of the wafer W when the polishing pad 2 is worn to some extent. When wear of the polishing pad 2 progresses and the grooves on the polishing surface 2a become shallower, the polishing liquid is less likely to enter a gap between the wafer W and the polishing surface 2a of the polishing pad 2. As a result, the polishing rate of the wafer W is lowered. When the polishing pad 2 wears further, as shown in FIG. 5, the pressure in the pressure chamber C1 further increases in order to increase the polishing rate.

In this way, the wear of the polishing pad 2 is reflected in the time-series pressure data that indicates a change over time in pressure in the pressure chamber C1 of the polishing head 7. Thus, the pad life calculation part 40 inputs the time-series pressure data acquired during the polishing of the wafer W to the learned model 42 (see FIG. 1) and outputs the life index of the polishing pad 2 from the learned model 42.

The life index of the polishing pad 2 is an index indicating a degree of wear of the polishing pad 2. An operator can determine whether or not to replace the polishing pad 2 on the basis of the life index of the polishing pad 2 calculated by the pad life calculation part 40. For example, the life index of the polishing pad 2 is the number of wafers (workpieces) that can be polished using the polishing pad 2. In another example, the life index of the polishing pad 2 is an index that indicates a ratio of a current usage time of the polishing pad 2 to the life of the polishing pad 2. In yet another example, the life index of the polishing pad 2 is a numerical value (for example, a numerical value from 1 to 10) that directly or indirectly represents the life of the polishing pad 2.

The learned model 42 is constructed through machine learning performed by the pad life calculation part 40. Training data used for the machine learning includes time-series pressure data for training acquired when many wafers were polished in the past using other polishing pads. The training data further includes a life index of a polishing pad that is a correct label. Examples of the machine learning include a support vector regression method (SVR method), the partial least squares method (PLS method), a deep learning method (deep structured learning), a random forest method, a decision tree method, and so on. In one example, the learned model 42 is configured of a neural network constructed by a deep learning method.

FIG. 6 is a schematic diagram showing an example of the learned model 42 constructed using deep learning. The learned model 42 has an input layer 101, a plurality of hidden layers (also called intermediate layers) 102, and an output layer 103. However, the learned model 42 is not limited to the example shown in FIG. 6.

Construction of the learned model 42 using deep learning is performed as follows. The time-series pressure data for training is input to the input layer 101 shown in FIG. 6. When the input layer 101 receives the time-series pressure data for training, the model 42 outputs the life index of the polishing pad from the output layer 103. The pad life calculation part 40 adjusts parameters (a weight, a threshold, and the like) of each node (neuron) to minimize a difference between the life index of the polishing pad output from the output layer 103 and the life index of the polishing pad which is the correct label. Thus, the model is learned to output an appropriate life index of the polishing pad from the output layer 103 on the basis of the data input to the input layer 101.

The learned model 42 is constructed by repeating the above machine learning using a plurality of pieces of time-series pressure data for training obtained when a plurality of wafers is polished using a plurality of polishing pads. By performing the machine learning using a large number of polishing pads, accuracy of the life index of the polishing pad 2 output from the learned model 42 can be improved. Accordingly, the learned model 42 can accurately estimate the life of the polishing pad 2 from the time-series pressure data indicating a change in pressure in the pressure chamber C1 of the polishing head 7. The learned model 42 is stored in the storage device 40a of the pad life calculation part 40.

The life index of the polishing pad serving as the correct label included in the training data is acquired as follows. At least one polishing pad is used to polish a plurality of wafers until the polishing pad reaches the end of its life. FIG. 7 is a diagram showing an example of a relationship between the number of wafers and a usage time of a polishing pad when a plurality of wafers has been polished until one polishing pad reaches the end of its life. In the example shown in FIG. 7, the polishing pad has reached the end of its life when 1000 wafers have been polished. The life index of the polishing pad serving as the correct label can be expressed by a ratio of the number of polished wafers at each time point from the start of use of the polishing pad to the end of the life of the polishing pad to 1000 wafers, the number of wafers obtained by subtracting the number of wafers at each time point from 1000 wafers (the number of polishable wafers), or a numerical value based thereon. The life index at each time point is associated with corresponding time-series pressure data for training acquired at that time point and added to the training data. The training data created in this manner is stored in the storage device 40a of the pad life calculation part 40.

FIG. 8 is a diagram showing an example of a relationship between the number of wafers and a usage time of each polishing pad when a plurality of wafers has been polished using a plurality of polishing pads until each polishing pad reaches the end of its service life. In the example shown in FIG. 8, a plurality of polishing pads are used to acquire training data for machine learning. Although these polishing pads have the same structure and the same material, external factors such as dressing the polishing pads can cause the polishing pads to have different life spans. Thus, in the example shown in FIG. 8, a plurality of polishing pads are used for polishing until the end of their life spans. Polishing pads A, B, and C have the same structure and the same material, the polishing pad A has reached the end of its life after polishing 1000 wafers, the polishing pad B has reached the end of its life after polishing 900 wafers, and the polishing pad C has reached the end of its life after polishing 1100 wafers. The life index of the polishing pad, which is the correct label using each polishing pad, is acquired in the same manner as in the example of FIG. 7.

In polishing for obtaining the training data, whether or not the polishing pad has reached the end of its life may be determined, for example, by a skilled operator visually observing grooves formed on the polishing surface of the polishing pad. In another example, the pad life calculation part 40 or the operator may determine that a polishing pad has reached the end of its life on the basis of a difference between a film thickness profile of a wafer polished using the polishing pad and a target film thickness profile. For example, when the difference exceeds a reference value, the pad life calculation part 40 or the operator can determine that the polishing pad has reached the end of its life. The film thickness profile may be replaced with uniformity of the film thickness in the wafer surface. In yet another example, the pad life calculation part 40 or the operator may determine that a polishing pad has reached the end of its life when a polishing rate of a wafer polished using the polishing pad falls below a predetermined threshold.

In one embodiment, the pad life calculation part 40 calculates the life index of the polishing pad 2 using the learned model 42 during a wafer polishing process. That is, after polishing the wafer W and before polishing the next wafer, the pad life calculation part 40 inputs time-series pressure data regarding the polished wafer W into the learned model 42 and outputs the life index of the polishing pad 2 from the learned model 42. If the obtained life index of the polishing pad 2 suggests that it is time to replace the polishing pad 2, the polishing pad 2 is replaced with a new polishing pad before the next wafer is polished.

The change in pressure in the pressure chamber C1 of the polishing head 7 during the polishing of the wafer W may be affected by a state of the wafer W being polished and/or polishing conditions for the wafer W. Thus, in order to calculate a more accurate life index of the polishing pad 2, in one embodiment, the pad life calculation part 40 may input polishing data relating to polishing of the wafer W into the learned model 42 in addition to the time-series pressure data.

The polishing data includes at least one of information of the wafer W and polishing conditions for the wafer W. Examples of the information of the wafer W include a polishing rate of the wafer W, a polishing time of the wafer W required to reach the target film thickness, an initial film thickness and a target film thickness of the wafer W, a film thickness profile (a film thickness distribution) of the wafer W after polishing and a polishing amount profile of the wafer W after polishing. Examples of the polishing conditions for the wafer W include rotation speeds of the polishing table 5 and polishing pad 2, a rotation speed of the polishing head 7, and a flow rate of the polishing liquid supplied to the polishing pad 2.

In one embodiment, the polishing control part 30 may perform control of the pressure in each pressure chamber based on film thickness measurement data during the polishing (for example, executed by comparison with a target polishing rate or a target film thickness value) only for a partial period during the polishing. For example, in a fixed polishing condition period from the start of polishing to a predetermined time, the wafer is polished while a predetermined constant pressure is maintained in the pressure chamber, and a polishing amount profile (polishing amount distribution), an average polishing amount, an average residual film amount, or a residual film profile (residual film distribution) acquired during the fixed polishing condition period may be included in the polishing data.

The training data used for the machine learning to construct the learned model 42 includes the same types of pieces of polishing data that are input to the learned model 42. That is, the training data includes, in addition to the time-series pressure data for training, at least one of wafer information and wafer polishing conditions when the time-series pressure data for training was acquired.

In one embodiment, the pad life calculation part 40 is configured to input a cut rate of the polishing pad 2 dressed by the dresser 50 into the learned model 42 in addition to the time-series pressure data. The cut rate of the polishing pad 2 is a thickness (amount) of the polishing pad 2 scraped off by the dresser 50 per unit time. The thickness (amount) of the polishing pad 2 scraped off by the dresser 50 is a difference between a height of the polishing surface 2a of the polishing pad 2 before dressing and a height of the polishing surface 2a of the polishing pad 2 after dressing.

The dressing disc 51 of the dresser 50 has hard abrasive grains (for example, diamond abrasive grains) on its dressing surface 51a. The abrasive grains are gradually worn away as the polishing pad 2 is repeatedly dressed. Accordingly, the cut rate of the polishing pad 2 also changes gradually. When the cut rate of the polishing pad 2 changes, a state of the polishing surface 2a of the dressed polishing pad 2 is assumed to also change, which may affect the time-series pressure data.

Thus, in order to calculate the life index of the polishing pad 2 more accurately, the pad life calculation part 40 inputs the cut rate of the polishing pad 2 into the learned model 42 in addition to the time-series pressure data. The training data used for the machine learning to construct the learned model 42 includes, in addition to the time-series pressure data for training, the cut rate of the polishing pad when the time-series pressure data for training was acquired.

Further, in one embodiment, the pad life calculation part 40 may be configured to input the polishing data about the wafer W and the cut rate of the polishing pad 2 into the learned model 42 in addition to the time-series pressure data. The training data used for the machine learning to construct the learned model 42 includes, in addition to the time-series pressure data for training, the same type of polishing data as the polishing data input to the learned model 42, and the cut rate of the polishing pad when the time-series pressure data for training was acquired. According to the present embodiment, the pad life calculation part 40 is expected to calculate a more accurate life index of the polishing pad 2.

In the above-described embodiments, the polishing head has the plurality of pressure chambers C1 to C4, but it is also possible to calculate the life index of the polishing pad 2 through machine learning using time-series pressure data of a polishing head having only a single pressure chamber.

In another embodiment, the polishing control part 30 calculates an average value of the film thickness of each of a plurality of regions on the wafer W corresponding to the pressure chambers C1 to C4 from the measured values of the film thickness of the wafer W being polished (that is, the current film thickness of the wafer W), calculates a difference between the average value of the film thickness of each region and an average value of the film thickness of the entire wafer W, and determines pressure values in the pressure chambers C1 to C4 required to eliminate the difference. For example, when the average film thickness of the region corresponding to the pressure chamber C1 is lower than the average film thickness of the entire wafer W, the polishing control part 30 reduces the pressure value of the pressure chamber C1. In another example, in a case in which the average value of the film thickness in the region corresponding to the pressure chamber C4 exceeds the average value of the film thickness of the entire wafer W, the polishing control part 30 increases the pressure value of the pressure chamber C4.

Time-series pressure data of two different pressure chambers may be input to the input layer of the model during machine learning and estimation of the life index. Time-series pressure data of two pressure chambers in which changes in a pad state such as a groove depth of a polishing pad are different from each other may show different behaviors depending on the changes in the pad state such as the groove depth of the polishing pad. For example, when the groove of the polishing pad becomes shallow, supply of slurry to a center of the wafer is particularly insufficient, and thus a polishing rate of the center of the wafer tends to decrease. Thus, a pressure in a pressure chamber at a central part of the wafer rises prior to a pressure chamber at an outer circumferential part of the wafer during polishing, and polishing uniformity is maintained. By inputting the time-series pressure data of two different pressure chambers into the learned model in this way, it becomes possible to output a more accurate life index.

The above-described embodiments are described for the purpose of enabling a person having ordinary knowledge in the technical field to which the disclosure belongs to implement the disclosure. Various modifications of the above embodiments can be made by those skilled in the art, and the technical idea of the disclosure can be applied to other embodiments. Accordingly, the disclosure is not limited to the described embodiments and is to be construed in its broadest scope in accordance with the technical idea defined by the claims.

Claims

1. A method for estimating a life of a polishing pad, comprising:

polishing a workpiece by pressing the workpiece against a polishing surface of a polishing pad with an elastic membrane forming a pressure chamber provided in a polishing head;

controlling a pressure in the pressure chamber on the basis of a measured value of a film thickness of the workpiece while measuring the film thickness during the polishing of the workpiece;

inputting time-series pressure data representing a change in the pressure during the polishing of the workpiece into a learned model; and

outputting a life index of the polishing pad from the learned model.

2. The method for estimating a life of a polishing pad according to claim 1, wherein, in addition to the time-series pressure data, polishing data relating to the polishing of the workpiece is input to the learned model.

3. The method for estimating a life of a polishing pad according to claim 2, wherein the polishing data includes at least one of information of the workpiece and polishing conditions for the workpiece.

4. The method for estimating a life of a polishing pad according to claim 1, wherein, in addition to the time-series pressure data, a cut rate of the polishing pad which has been dressed by a dresser is input to the learned model.

5. The method for estimating a life of a polishing pad according to claim 1, wherein the steps of inputting the time-series pressure data into the learned model and outputting the life index of the polishing pad from the learned model are performed after the polishing of the workpiece and before polishing of a next workpiece.

6. The method for estimating a life of a polishing pad according to claim 1,

wherein the pressure chamber is a plurality of pressure chambers,

the step of controlling the pressure in the pressure chamber is a step of controlling pressures in the plurality of pressure chambers on the basis of the measured value of the film thickness of the workpiece while measuring the film thickness during the polishing of the workpiece, and

the step of inputting the time-series pressure data into the learned model is a step of inputting into the learned model a plurality of pieces of time-series pressure data respectively representing changes in the pressures in the plurality of pressure chambers during the polishing of the workpiece.

7. A polishing device comprising:

a polishing head that includes a pressure chamber formed by an elastic membrane and polishes a workpiece by pressing the workpiece against a polishing surface of a polishing pad with the elastic membrane;

a film thickness sensor that measures a film thickness of the workpiece;

a polishing control part that controls a pressure in the pressure chamber during the polishing of the workpiece on the basis of a measured value of the film thickness; and

a pad life calculation part that inputs time-series pressure data representing a change in the pressure during the polishing of the workpiece into a learned model and outputs a life index of the polishing pad from the learned model.

8. The polishing device according to claim 7, wherein the pad life calculation part is configured to input polishing data relating to the polishing of the workpiece into the learned model in addition to the time-series pressure data.

9. The polishing device according to claim 8, wherein the polishing data includes at least one of information of the workpiece and polishing conditions for the workpiece.

10. The polishing device according to claim 7, further comprising a dresser that dresses the polishing surface of the polishing pad, wherein

the pad life calculation part is configured to input a cut rate of the polishing pad dressed by the dresser into the learned model in addition to the time-series pressure data.

11. The polishing device according to claim 7, wherein the pad life calculation part is configured to input the time-series pressure data into the learned model and output the life index of the polishing pad from the learned model after the polishing of the workpiece and before polishing of a next workpiece.

12. The polishing device according to claim 7,

wherein the pressure chamber is a plurality of pressure chambers,

the polishing control part is configured to control pressures in the plurality of pressure chambers on the basis of the measured value of the film thickness of the workpiece while measuring the film thickness during the polishing of the workpiece, and

the pad life calculation part is configured to input into the learned model a plurality of pieces of time-series pressure data respectively representing changes in the pressures in the plurality of pressure chambers during the polishing of the workpiece.