Polishing apparatus and polishing method

Abstract

To provide a polishing device which can obtain a desired film thickness profile.SOLUTION: A polishing device includes: a polishing unit 14a; a film thickness measuring device 8 which measures a film thickness profile of a substrate W; and a control device 30 which controls at least operations of the polishing unit 14a and the film thickness measuring device 8. The control device 30 preliminarily stores a response model prepared by considering a change of a polishing amount between a plurality of monitor regions of the substrate W accompanying a change of pressure in each of pressure chambers 7a to 7h. Furthermore, the control device acquires a film thickness profile before polishing the substrate W using the film thickness measuring device 8 and polishes the substrate W with an optimum polishing recipe prepared on the basis of a difference between the film thickness profile of the substrate W before polishing and a target film thickness of the substrate W and the response model. The next substrate W is polished with a new optimum polishing recipe prepared on the basis of a target polishing amount of the next substrate W and the response model corrected using the optimum polishing recipe and the film thickness profile before and after polishing the previous substrate.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a polishing apparatus and a polishing method for a substrate such as a wafer, and particularly to a polishing apparatus and a polishing method for obtaining a desired film thickness profile by polishing a substrate. The present invention also relates to a polishing method for polishing a substrate using such a polishing apparatus.

Background Art

[0002] In recent years, with the increasing integration and high density of semiconductor devices, circuit wiring has become increasingly finer, and the number of layers of multilayer wiring has also increased. When attempting to achieve multilayer wiring while reducing the circuit size, the step height becomes larger while following the surface unevenness of the lower layer. Therefore, as the number of wiring layers increases, the film covering property (step coverage) for the step shape in thin film formation deteriorates. Therefore, in order to perform multilayer wiring, it is necessary to improve this step coverage and perform a planarization process in the appropriate process. In addition, as the miniaturization of photolithography progresses, the depth of focus becomes shallower, so it is necessary to planarize the surface of the semiconductor device so that the uneven step on the surface of the semiconductor device is within the depth of focus.

[0003] Therefore, in the manufacturing process of semiconductor devices, planarization of the semiconductor device surface has become increasingly important. The most important technology in this surface planarization is chemical mechanical polishing (CMP: Chemical Mechanical Polishing). This chemical mechanical polishing (hereinafter referred to as CMP) is a process in which a polishing liquid (slurry) containing abrasive grains such as silica (SiO 2 ) is supplied onto the polishing surface of a polishing pad, and a substrate such as a wafer is brought into sliding contact with the polishing surface to perform polishing.

[0004] A polishing apparatus for performing CMP includes a polishing table that supports a polishing pad having a polishing surface, and a polishing head (substrate holding device) for holding a substrate. Polishing of the substrate using such a polishing apparatus is performed as follows. While rotating the polishing table together with the polishing pad, slurry is supplied onto the polishing pad. The polishing head presses the substrate against the polishing surface of the polishing pad while rotating the substrate. The substrate is brought into sliding contact with the polishing pad in the presence of the slurry, and the surface of the substrate is planarized by a combination of the chemical action of the slurry and the mechanical action of abrasive grains contained in the slurry.

[0005] During polishing of the substrate, since the surface of the substrate is brought into sliding contact with the rotating polishing pad, a frictional force acts on the substrate. Therefore, in order to prevent the substrate from coming off the polishing head during polishing of the substrate, the polishing head is provided with a retainer ring. This retainer ring is arranged so as to surround the substrate, and during polishing of the substrate, the retainer ring presses the polishing pad outside the substrate while rotating.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0007] In recent years, due to reasons such as coping with various initial film thickness profiles that can vary according to semiconductor devices and CMP processes, and improving yield, there has been an increasing demand to more precisely control the film thickness profile of the substrate (that is, to improve the in-plane uniformity representing the flatness of the surface of the substrate).

[0008] Furthermore, the allowable range with respect to the target film thickness has been narrowing, and in the conventional polishing method of obtaining the film thickness index value of the substrate using a film thickness sensor during polishing and terminating the polishing of the substrate based on the film thickness index value, it has become difficult to keep the film thickness within the required allowable range. For example, the polishing amount during one rotation of the polishing table may be larger than the allowable range. In this case, if the end point of polishing is determined based on the film thickness index value obtained from the film thickness sensor arranged on the polishing table, the film thickness after the completion of polishing may exceed the allowable range with respect to the target film thickness.

[0009] Furthermore, the film thickness profiles before polishing may differ between the substrates to be polished, or the polishing conditions (for example, the state of the polishing surface of the polishing pad) may differ between polishing apparatuses. Due to these combined factors as well, it has become difficult to precisely control the film thickness profile and keep the film thickness profile within the required allowable range.

[0010] Therefore, an object of the present invention is to provide a polishing apparatus capable of obtaining a desired film thickness profile. Another object of the present invention is to provide a polishing method for polishing a substrate using such a polishing apparatus.

Means for Solving the Problems

[0011] In one aspect, at least one polishing unit including a polishing table for supporting a polishing pad and a substrate holding device for pressing a substrate against the polishing pad, a film thickness measuring device for measuring a film thickness profile of the substrate, and a control device for at least controlling operations of the polishing unit and the film thickness measuring device. The substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed to surround the substrate. The control device stores in advance a response model created in consideration of changes in the polishing amount between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber. The control device uses the film thickness measuring device to obtain a film thickness profile of the substrate before polishing, and polishes the substrate with an optimal polishing recipe including at least the pressure of a compressed fluid supplied to the plurality of pressure chambers and a polishing time, created based on a target polishing amount which is a difference between the film thickness profile of the substrate before polishing and a target film thickness of the substrate, and the response model. Then, the next substrate is polished with a new optimal polishing recipe created based on a target polishing amount of the next substrate, the optimal polishing recipe, and a response model corrected using the film thickness profiles of the substrate before and after polishing. A polishing apparatus is provided.

[0012] In one aspect, the optimal polishing recipe is created using an optimization calculation that minimizes an objective function including at least a term of a difference between the target polishing amount and a predicted polishing amount calculated using the response model. In one aspect, the objective function further includes a term of a difference between the compressed fluid pressure of the optimal polishing recipe and a preset reference compressed fluid pressure, and / or a term of a difference between the compressed fluid pressure of the optimal polishing recipe and the compressed fluid pressure of the optimal polishing recipe of a wafer polished previously. In one aspect, the optimization calculation is a quadratic programming method. In one aspect, the number of the plurality of monitor regions is larger than the number of the plurality of pressure chambers. In one aspect, the response model is a response model created by further considering the change in the polishing amount between a plurality of monitor regions of the substrate in accordance with the change in the pressing force of the retainering on the polishing pad, and the optimal polishing recipe further includes the pressing force of the retainering.

[0013] In one aspect, the film thickness measuring device is configured to be able to measure the film thickness at a plurality of measurement points set for each of the plurality of monitor regions. In one aspect, the response model includes a response coefficient representing an increase amount of the polishing rate per unit polishing pressure in each of the plurality of monitor regions. In one aspect, the apparatus further includes a plurality of local load applying devices that apply a local load to a part of the retainering, and the response model is created by further considering the change in the polishing amount between the plurality of monitor regions in accordance with the change in the local load.

[0014] In one aspect, at least one polishing unit including a polishing table for supporting a polishing pad and a substrate holding device for pressing a substrate against the polishing pad, a film thickness measuring device for measuring a film thickness profile of the substrate, and a control device for at least controlling operations of the polishing unit and the film thickness measuring device are provided. The substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate. The control device preliminarily stores a film thickness profile before polishing of a plurality of substrates and a response model when each of the plurality of substrates is polished, the response model being created in consideration of changes in polishing amounts between a plurality of monitor regions of the substrate accompanying changes in pressure in each pressure chamber. The film thickness profiles before polishing of the plurality of substrates are preliminarily classified into a plurality of groups to which similar film thickness profiles belong. The control device acquires the film thickness profile of the substrate before polishing using the film thickness measuring device, determines the group to which the film thickness profile of the substrate before polishing belongs from the plurality of groups, and polishes the substrate with an optimal polishing recipe including at least the pressure of a compressed fluid supplied to the plurality of pressure chambers and a polishing time, the optimal polishing recipe being created based on a target polishing amount which is a difference between the film thickness profile of the substrate before polishing and a target film thickness of the substrate, and the response model associated with the determined group. Then, a polishing apparatus is provided which polishes the next substrate with a new optimal polishing recipe created based on a target polishing amount of the next substrate, and the response model corrected using the optimal polishing recipe and the film thickness profiles of the substrate before and after polishing.

[0015] In one aspect, a polishing apparatus includes a plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing a substrate against the polishing pad, and a control device for at least controlling the operation of the polishing units. The substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate. The substrate is polished by a plurality of polishing steps including a first polishing and a second polishing performed by a polishing unit different from the polishing unit that performed the first polishing. The polishing unit that performs the first polishing has a film thickness sensor capable of measuring the film thickness profile of the substrate. The control device stores in advance a second polishing response model created in consideration of changes in the polishing amount between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber. After performing the first polishing, the control device uses the film thickness sensor to obtain the film thickness profile of the substrate before the second polishing, and based on the target polishing amount, which is the difference between the film thickness profile of the substrate before the second polishing and the target film thickness of the substrate, and the second polishing response model, polishes the substrate with a second polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time. Then, the polishing apparatus polishes the next substrate with a new second polishing optimum recipe created based on the target polishing amount of the next substrate, the second polishing optimum recipe, and the second polishing response model corrected using the film thickness profiles of the substrate before and after the second polishing is provided.

[0016] In one aspect, a polishing apparatus includes a plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing a substrate against the polishing pad, a film thickness measuring device for measuring a film thickness profile of the substrate, and a control device for at least controlling operations of the polishing units and the film thickness measuring device. The substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed to surround the substrate. The substrate is a substrate polished by a plurality of polishing steps including a first polishing and a second polishing performed by a polishing unit different from the polishing unit where the first polishing is performed. The polishing unit for performing the second polishing has a film thickness sensor capable of measuring the film thickness profile of the substrate. The control device stores in advance a first polishing response model created in consideration of a change in the polishing amount between a plurality of monitor regions of the substrate accompanying a change in the pressure in each pressure chamber. The control device uses the film thickness measuring device to obtain the film thickness profile of the substrate before the first polishing, and based on a target polishing amount that is a difference between the film thickness profile of the substrate before the first polishing and a target film thickness of the substrate, and the first polishing response model, polishes the substrate with a first polishing optimal recipe including at least the pressure of a compressed fluid supplied to the plurality of pressure chambers and a polishing time. After performing the first polishing, the substrate is transported to a polishing unit having the film thickness sensor, and the film thickness profile of the substrate after the first polishing is obtained using the film thickness sensor. A next substrate is polished with a new first polishing optimal recipe created based on a target polishing amount of the next substrate, the first polishing optimal recipe, and the first polishing response model corrected using the film thickness profiles of the substrate before and after the first polishing.

[0017] In one aspect, a polishing apparatus includes a plurality of polishing units including a polishing table for supporting a polishing pad, a substrate holding device for pressing a substrate against the polishing pad, and a film thickness sensor capable of measuring a film thickness profile of the substrate, and a control device for at least controlling the operation of the polishing units. The substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate. The substrate is a substrate polished by a plurality of polishing steps including a first polishing and a second polishing performed in a polishing unit different from the polishing unit in which the first polishing is performed. The control device stores in advance a first polishing response model and a second polishing response model created in consideration of changes in the polishing amount between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber. The control device transports the substrate to any one of the plurality of polishing units, and uses the film thickness sensor to obtain a film thickness profile of the substrate before the first polishing. Based on the target polishing amount, which is the difference between the film thickness profile of the substrate before the first polishing and the target film thickness of the substrate, and the first polishing response model, the substrate is subjected to the first polishing with a first polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time. Using the film thickness sensor, a film thickness profile of the substrate before the second polishing is obtained. Based on the target polishing amount, which is the difference between the film thickness profile of the substrate before the second polishing and the target film thickness of the substrate, and the second polishing response model, the substrate is subjected to the second polishing with a second polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time. Using the film thickness sensor, a film thickness profile of the substrate after the second polishing is obtained. The next substrate is subjected to the first polishing with a new first polishing optimum recipe created based on the target polishing amount of the next substrate and the first polishing response model corrected using the first polishing optimum recipe and the film thickness profiles of the substrate before and after the first polishing. The next substrate is subjected to the second polishing with a new second polishing optimum recipe created based on the target polishing amount of the next substrate and the second polishing response model corrected using the second polishing optimum recipe and the film thickness profiles of the substrate before and after the second polishing. A polishing apparatus is provided.

[0018] In one aspect, a polishing method for pressing a substrate held by a substrate holding device having an elastic film that forms a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate against a polishing pad supported by a polishing table, wherein a film thickness measuring instrument is used to obtain a film thickness profile of the substrate before polishing, and based on a target polishing amount that is a difference between the film thickness profile of the substrate before polishing and the target film thickness of the substrate, and a response model, the substrate is polished with an optimal polishing recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, and the target polishing amount of the next substrate and a new optimal polishing recipe created based on the optimal polishing recipe and a response model corrected using the film thickness profiles of the substrate before and after polishing are used to polish the next substrate, and the response model is created in consideration of changes in the polishing amounts between a plurality of monitor regions of the substrate accompanying changes in the pressure of each pressure chamber, is provided.

[0019] In one aspect, the optimal polishing recipe is created using an optimization calculation that minimizes an objective function including at least a term of a difference between the target polishing amount and a predicted polishing amount calculated using the response model. In one aspect, the objective function further includes a term of a difference between the compressed fluid pressure of the optimal polishing recipe and a preset reference compressed fluid pressure, and / or a term of a difference between the compressed fluid pressure of the optimal polishing recipe and the compressed fluid pressure of the optimal polishing recipe of a wafer polished previously. In one aspect, the optimization calculation is a quadratic programming method. In one aspect, the number of the plurality of monitor regions is larger than the number of the plurality of pressure chambers. In one aspect, the response model is a response model created in consideration of changes in the polishing amounts between a plurality of monitor regions of the substrate accompanying changes in the pressing force of the retainer ring against the polishing pad, and the optimal polishing recipe further includes the pressing force of the retainer ring.

[0020] In one aspect, the film thickness measuring device measures the film thickness at a plurality of measurement points set for each of the plurality of monitor regions. In one aspect, the response model includes a response coefficient representing an increase in the polishing rate per unit polishing pressure in each of the plurality of monitor regions. In one aspect, the response model is created by further considering the change in the polishing amount among the plurality of monitor regions with the change in the local load applied to a part of the retainer ring by a plurality of local load applying devices.

[0021] In one aspect, there is a polishing method of pressing a substrate held by a substrate holding device having an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate, against a polishing pad supported by a polishing table for polishing. A film thickness profile before polishing of a plurality of substrates and a response model when each of the plurality of substrates is polished are preliminarily stored. The film thickness profile before polishing of the plurality of substrates is preliminarily classified into a plurality of groups to which similar film thickness profiles belong. The film thickness profile before polishing of the substrate is obtained, and the group to which the film thickness profile belongs is determined from the plurality of groups. Based on the target polishing amount, which is the difference between the film thickness profile of the substrate before polishing and the target film thickness of the substrate, and the response model associated with the determined group, at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time are included. The substrate is polished with an optimal polishing recipe, and the next substrate is polished with a new optimal polishing recipe created based on the target polishing amount of the next substrate and the response model corrected using the optimal polishing recipe and the film thickness profiles of the substrate before and after polishing. A polishing method is provided.

[0022] In one aspect, a method of polishing a substrate by a plurality of polishing steps performed in a plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing the substrate against the polishing pad, wherein the substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate, the plurality of polishing steps including a first polishing and a second polishing performed in a polishing unit different from the polishing unit in which the first polishing is performed, the polishing unit performing the first polishing having a film thickness sensor capable of measuring a film thickness profile of the substrate, preparing in advance a second polishing response model created in consideration of changes in the polishing amount between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber, after performing the first polishing, using the film thickness sensor to obtain the film thickness profile of the substrate before the second polishing, and based on the target polishing amount, which is the difference between the film thickness profile of the substrate before the second polishing and the target film thickness of the substrate, and the second polishing response model, performing the second polishing of the substrate with a second polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, and performing the second polishing of the next substrate with a new second polishing optimal recipe created based on the target polishing amount of the next substrate, the second polishing optimal recipe, and the second polishing response model corrected using the film thickness profiles of the substrate before and after the second polishing, is provided.

[0023] In one aspect, a method of polishing a substrate is performed by a plurality of polishing steps in a plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing the substrate against the polishing pad. The substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate. The plurality of polishing steps include a first polishing and a second polishing performed in a polishing unit different from the polishing unit in which the first polishing is performed. The polishing unit performing the second polishing has a film thickness sensor capable of measuring a film thickness profile of the substrate. A first polishing response model prepared in advance in consideration of changes in the polishing amount between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber is prepared. Using a film thickness measuring device, a film thickness profile of the substrate before the first polishing is obtained. Based on the target polishing amount, which is the difference between the film thickness profile of the substrate before the first polishing and the target film thickness of the substrate, and the first polishing response model, the substrate is polished in a first polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time. After performing the first polishing, the substrate is transported to a polishing unit having the film thickness sensor, and using the film thickness sensor, a film thickness profile of the substrate after the first polishing is obtained. A new first polishing optimum recipe is created based on the target polishing amount of the next substrate, the first polishing optimum recipe, and the first polishing response model corrected using the film thickness profiles of the substrate before and after the first polishing, and the next substrate is polished in the new first polishing optimum recipe. A polishing method is provided.

[0024] In one aspect, a method of polishing a substrate by a plurality of polishing steps performed in a plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing the substrate against the polishing pad, wherein the substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate, the plurality of polishing steps include a first polishing and a second polishing performed in a polishing unit different from the polishing unit in which the first polishing is performed, the polishing unit performing the first polishing and the polishing unit performing the second polishing each have a film thickness sensor capable of measuring a film thickness profile of the substrate, a first polishing response model and a second polishing response model prepared in advance in consideration of a change in the polishing amount between a plurality of monitor regions of the substrate accompanying a change in the pressure in each pressure chamber are prepared, the substrate is transported to the polishing unit performing the first polishing, and the film thickness profile of the substrate before the first polishing is obtained using the film thickness sensor, and the substrate is first polished with a first polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount, which is the difference between the film thickness profile of the substrate before the first polishing and the target film thickness of the substrate, and the first polishing response model, the film thickness profile of the substrate before the second polishing is obtained using the film thickness sensor, and the substrate is second polished with a second polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount, which is the difference between the film thickness profile of the substrate before the second polishing and the target film thickness of the substrate, and the second polishing response model, the film thickness profile of the substrate after the second polishing is obtained using the film thickness sensor, and the next substrate is first polished with a new first polishing optimum recipe created based on the target polishing amount of the next substrate, the first polishing response model corrected using the first polishing optimum recipe and the film thickness profiles of the substrate before and after the first polishing, and the next substrate is second polished with a new second polishing optimum recipe created based on the target polishing amount of the next substrate, the second polishing response model corrected using the second polishing optimum recipe and the film thickness profiles of the substrate before and after the second polishing.A polishing method is provided.

Advantages of the Invention

[0025] According to the present invention, the optimal polishing recipe obtained using the response model takes into account the change in the polishing amount between regions of the substrate accompanying the change in the pressure of each pressure chamber. Therefore, the film thickness profile of the substrate can be controlled more precisely.

Brief Description of the Drawings

[0026]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

[0027] Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a plan view showing the overall configuration of a polishing apparatus according to an embodiment. The polishing apparatus shown in FIG. 1 is a CMP apparatus that executes a series of polishing processes for polishing the surface of a wafer, which is an example of a substrate, cleaning the polished wafer, and drying the cleaned wafer.

[0028] As shown in FIG. 1, the polishing apparatus includes a substantially rectangular housing 10 and a load port 12 on which a substrate cassette for accommodating a large number of wafers (substrates) is placed. The load port 12 is disposed adjacent to the housing 10. The load port 12 can be mounted with an open cassette, a SMIF (Standard Manufacturing Interface) pod, or a FOUP (Front Opening Unified Pod). SMIF and FOUP are sealed containers that can house a substrate cassette inside and maintain an environment independent of the external space by covering it with a partition wall.

[0029] Inside the housing 10, a plurality (four in this embodiment) of polishing units 14a to 14d for polishing wafers, a first cleaning unit 16 and a second cleaning unit 18 for cleaning the polished wafers, and a drying unit 20 for drying the cleaned wafers are accommodated. The polishing units 14a to 14d are arranged along the longitudinal direction of the substrate processing apparatus, and the cleaning units 16, 18 and the drying unit 20 are also arranged along the longitudinal direction of the substrate processing apparatus. Further, the polishing apparatus has a control device 30 located inside the housing 10 for controlling the operation of each unit.

[0030] A first substrate transfer robot 22 is disposed in the region surrounded by the load port 12, the polishing unit 14a, and the drying unit 20, and a substrate transfer device 24 is disposed in parallel with the polishing units 14a to 14d. The first substrate transfer robot 22 receives the wafer before polishing from the load port 12 and delivers it to the substrate transfer device 24, and receives the dried wafer from the drying unit 20 and returns it to the load port 12. The substrate transfer device 24 transports the substrate received from the first substrate transfer robot 22 and exchanges the substrate with each of the polishing units 14a to 14d.

[0031] A second substrate transfer robot 26 is disposed between the first cleaning unit 16 and the second cleaning unit 18 and transfers wafers between these cleaning units 16, 18 and the substrate transfer device 24. A third substrate transfer robot 28 is disposed between the second cleaning unit 18 and the drying unit 20 and transfers wafers between these units 18, 20. The first substrate transfer robot 22, the substrate transfer device 24, the second substrate transfer robot 26, and the third substrate transfer robot 28 constitute a substrate transfer unit for transferring wafers between the load port 12, the polishing units 14a to 14d, the cleaning units 16, 18, and the drying unit 20.

[0032] In the present embodiment, as the first cleaning unit 16, a substrate cleaning device that scrubs and cleans the substrate by rubbing roll sponges against both the front and back surfaces of the wafer in the presence of a chemical solution is used. As the second cleaning unit 18, a substrate cleaning device using a pen-type sponge (pen sponge) is used. In one embodiment, as the second cleaning unit 18, a substrate cleaning device that scrubs and cleans the wafer by rubbing roll sponges against both the front and back surfaces of the wafer in the presence of a chemical solution may be used. Further, as the drying unit 20, a spin drying device that holds the wafer, ejects IPA vapor from a moving nozzle to dry the wafer, and further dries the wafer by rotating it at a high speed is used.

[0033] The wafer is polished by at least one of the polishing units 14a to 14d. The polished wafer is cleaned by the first cleaning unit 16 and the second cleaning unit 18, and the further cleaned substrate is dried by the drying unit 20. In one embodiment, the polished substrate may be cleaned by either the first cleaning unit 16 or the second cleaning unit 18.

[0034] As shown in FIG. 1, the polishing apparatus according to the present embodiment includes a film thickness measuring device (ITM: In-line Thickness Monitor) 8 for detecting (measuring) the film thickness profile of the surface (polished surface) of the wafer. The type of the film thickness measuring device 8 is arbitrary as long as it can acquire the film thickness profile of the wafer. For example, the film thickness measuring device 8 may be a measuring machine that uses a non-contact detection method such as an eddy current type or an optical type, or may be a measuring device that detects the film thickness profile by a detection head that scans non-contactingly above the wafer. Alternatively, the film thickness measuring device 8 may be a measuring machine that scans a probe in contact with the surface of the wafer and monitors the up and down movement of the probe to detect the distribution of unevenness on the surface of the wafer. In both the contact type and non-contact type detection methods, the detected output is a signal corresponding to the film thickness or the film thickness. When detecting the film thickness profile of the wafer, the film thickness profile may be associated not only with the position in the radial direction but also with the position in the circumferential direction by using the position of the notch or the position of the orientation flat of the wafer as a reference.

[0035] The film thickness measuring device 8 is connected to the control device 30, and the control device 30 is configured to be able to control the operation of the film thickness measuring device 8. Further, the film thickness measuring device 8 transmits its measured value to the control device 30, and the control device 30 can acquire the film thickness profile of the wafer from the measured value transmitted from the film thickness measuring device 8.

[0036] FIG. 2 is a perspective view schematically showing an example of the polishing unit 14a shown in FIG. 1. Since the polishing units 14a to 14d of the polishing apparatus shown in FIG. 1 have the same configuration as each other, the polishing unit 14a will be described below.

[0037] The polishing unit 14a shown in Fig. 2 includes a polishing table 35 to which a polishing pad 33 having a polishing surface 33a is attached, a polishing head 37 that holds the wafer W and presses the wafer W against the polishing pad 33 on the polishing table 35, a polishing liquid supply nozzle 38 for supplying a polishing liquid or a dressing liquid (e.g., pure water) to the polishing pad 33, and a dressing device 40 having a dresser 41 for dressing the polishing surface 33a of the polishing pad 33.

[0038] The polishing table 35 is connected to a table motor 31 disposed below it via a table shaft 35a, and the table motor 31 rotates the polishing table 35 in the direction indicated by the arrow. The polishing pad 33 is attached to the upper surface of the polishing table 35, and the upper surface of the polishing pad 33 constitutes a polishing surface 33a for polishing the wafer W. The polishing head 37 is connected to the lower end of a head shaft 36. The polishing head 37 is configured to hold the wafer W on its lower surface by vacuum suction. The head shaft 36 is configured to move up and down by an up-and-down movement mechanism (not shown).

[0039] The head shaft 36 is rotatably supported by a head arm 42, and the head arm 42 is driven by a head turning motor 54 and configured to turn about a head turning shaft 43. By driving the head turning motor 54, the polishing head 37 moves between a polishing position above the polishing pad 33 and a standby position beside the polishing pad 33.

[0040] The dressing device 40 includes a dresser 41 that is in sliding contact with the polishing pad 33, a dresser shaft 45 to which the dresser 41 is connected, an air cylinder 47 provided at the upper end of the dresser shaft 45, and a dresser arm 48 that rotatably supports the dresser shaft 45. The lower surface of the dresser 41 constitutes a dressing surface 41a, and this dressing surface 41a is composed of abrasive grains (for example, diamond particles). The air cylinder 47 is disposed on a support base 50 supported by a plurality of support columns 51, and these support columns 51 are fixed to the dresser arm 48.

[0041] The dresser arm 48 is driven by a dresser turning motor 55 and is configured to turn about a dresser turning shaft 49. The dresser shaft 45 rotates by the drive of a motor (not shown), and due to the rotation of this dresser shaft 45, the dresser 41 rotates in the direction indicated by the arrow about the dresser shaft 45. The air cylinder 47 functions as an actuator that moves the dresser 41 up and down via the dresser shaft 45 and presses the dresser 41 against the polishing surface (surface) 33a of the polishing pad 33 with a predetermined pressing force.

[0042] Next, with reference to FIG. 3, an example of the polishing head 37 provided in the polishing unit 14a will be described. FIG. 3 is a cross-sectional view schematically showing an example of the polishing head. As shown in FIG. 3, the polishing head 37 basically includes a head body 2 fixed to the lower end of a head shaft 36, a retainer ring 3 that directly presses against the polishing surface 33a (see FIG. 2), and an elastic membrane (membrane) 5 that presses the wafer W against the polishing surface 33a. The retainer ring 3 is disposed so as to surround the wafer W and is connected to the head body 2. The elastic membrane 5 is attached to the head body 2 so as to cover the lower surface of the head body 2.

[0043] The elastic film 5 has a plurality of (eight in the illustrated example) concentrically arranged annular peripheral walls 5a, 5b, 5c, 5d, 5e, 5f, 5g, 5h. Between the upper surface of the elastic film 5 and the lower surface of the head body 2, a circular central pressure chamber 7a located at the center, an annular edge pressure chamber 7h located at the outermost periphery, and six annular intermediate pressure chambers (first to sixth intermediate pressure chambers) 7b, 7c, 7d, 7e, 7f, 7g located between the central pressure chamber 7a and the edge pressure chamber 7h are formed by these plurality of peripheral walls 5a to 5h. In the present embodiment, the number of pressure chambers formed in the elastic film 5 is eight, but the number of pressure chambers is not limited to the present embodiment. The number of pressure chambers may be increased or decreased according to the structure of the elastic film 5.

[0044] In the head body 2, flow paths 4a, 4b, 4c, 4d, 4e, 4f, 4g, 4h communicating with the pressure chambers 7a to 7h are respectively formed. And the flow paths 4a, 4b, 4c, 4d, 4e, 4f, 4g, 4h are respectively connected to a pressure adjusting device 65 via fluid lines 6a, 6b, 6c, 6d, 6e, 6f, 6g, 6h. The pressure adjusting device 65 is connected to a control device 30, and the control device 30 is configured to be able to control the operation of the pressure adjusting device 65.

[0045] A retainer chamber 34 is formed directly above the retainer ring 3, and the retainer chamber 34 is connected to a pressure adjusting device 65 via a flow path 4i and a fluid line 6i formed in the head body 2.

[0046] According to the polishing head 37 configured as shown in FIG. 3, while the wafer W is held by the polishing head 37, by controlling the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h, the wafer W can be pressed with different pressures for each of a plurality of regions (areas) on the elastic film 5 along the radial direction of the wafer W. Thus, in the polishing head 37, by adjusting the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h formed between the head body 2 and the elastic film 5, the pressing force applied to the wafer W can be adjusted for each region of the wafer W. At the same time, by controlling the pressure of the compressed fluid supplied to the retainer chamber 34, the pressing force with which the retainer ring 3 presses the polishing pad 33 (see FIG. 2) can be adjusted.

[0047] When the retainer ring 3 presses the polishing pad 33, the shape of the polishing pad 33 changes according to the pressing force. Therefore, the pressing force of the retainer ring 3 on the polishing pad 33 also becomes a factor affecting the film thickness profile of the polished wafer W.

[0048] The retainer ring 3 is formed of a resin such as engineering plastic (e.g., PEEK), and the elastic film 5 is formed of a rubber material having excellent strength and durability such as ethylene propylene rubber (EPDM), polyurethane rubber, or silicone rubber.

[0049] Next, with reference to FIG. 4, the configuration of the pressure adjustment device 65 shown in FIG. 3 will be described. FIG. 4 is a schematic diagram showing the pressure adjustment device 65 shown in FIG. 3. As shown in FIG. 4, on the fluid lines 6a to 6i, opening / closing valves V1, V2, V3, V4, V5, V6, V7, V8, V9 and pressure regulators R1, R2, R3, R4, R5, R6, R7, R8, R9 are respectively connected. The flow paths 4a to 4i are respectively connected to the fluid supply source 32 via the fluid lines 6a to 6i.

[0050] Furthermore, atmospheric release lines 151 to 159 are connected to the fluid lines 6a to 6i. Atmospheric release valves L1 to L9 are respectively attached to these atmospheric release lines 151 to 159.

[0051] The pressure regulators R1 to R9 each have a pressure regulating function of regulating the pressure of the compressed fluid supplied from the fluid supply source 32 to the pressure chambers 7a to 7h and the retainer chamber 34. The pressure regulators R1 to R9, the on-off valves V1 to V9, and the atmosphere release valves L1 to L9 are connected to the control device 30, and their operations are controlled by the control device 30. When the atmosphere release valves L1 to L9 are actuated, the chambers 7a to 7h and 34 are released to the atmosphere and become atmospheric pressure states.

[0052] Although not shown, a plurality of vacuum lines are respectively connected to the fluid lines 6a to 6i, and a negative pressure can be formed in the chambers 7a to 7h and 34 through these plurality of vacuum lines. In this way, each of the pressure chambers 7a to 7h and 34 is adjusted to any one of a pressurized state, a negative pressure state, and an atmospheric pressure state by the pressure adjusting device 65.

[0053] When a vacuum is formed in any one of the intermediate pressure chambers 7b to 7g (for example, the intermediate pressure chamber 7d) with the wafer W in contact with the lower surface of the elastic membrane 5, the wafer W is held by the polishing head 37 by vacuum suction. Further, when compressed fluid is supplied to any one of the intermediate pressure chambers 7b to 7g (for example, the intermediate pressure chamber 7d) with the wafer W separated from the polishing pad 33, the wafer W is released from the polishing head 37.

[0054] Next, a polishing method for polishing the wafer W using this polishing apparatus will be described. FIG. 5 is a flowchart showing a polishing method according to an embodiment. As shown in FIG. 5, the control device 30 of the polishing apparatus takes out the wafer W from the substrate cassette placed on the load port 12 (see FIG. 1), transports it to the film thickness measuring device 8, and acquires the film thickness profile of the wafer W before polishing (see step 1 in FIG. 5).

[0055] Next, the control device 30 creates an optimal polishing recipe based on the film thickness profile of the wafer W before polishing and the response model (see step 2 in FIG. 5). The response model is stored in the control device 30 in advance. The method for creating this response model will be described later.

[0056] Hereinafter, a method for creating an optimal polishing recipe according to an embodiment will be described.

[0057] First, the control device 30 calculates the initial film thickness for each monitor region of the wafer W corresponding to each pressure chamber 7a to 7h of the elastic film 5 from the film thickness profile of the wafer W before polishing acquired by the film thickness measuring device 8. Hereinafter, for convenience of explanation, the monitor region of the wafer W corresponding to the pressure chamber 7a is represented as Da, and the initial film thickness of the monitor region Da is represented as Ta. Similarly, the monitor region of the wafer W corresponding to the pressure chamber 7b is represented as Db, the initial film thickness of the monitor region Db is represented as Tb, the monitor region of the wafer W corresponding to the pressure chamber 7c is represented as Dc, the initial film thickness of the monitor region Dc is represented as Tc, the monitor region of the wafer W corresponding to the pressure chamber 7d is represented as Dd, the initial film thickness of the monitor region Dd is represented as Td, the monitor region of the wafer W corresponding to the pressure chamber 7e is represented as De, the initial film thickness of the monitor region De is represented as Te, the monitor region of the wafer W corresponding to the pressure chamber 7f is represented as Df, the initial film thickness of the monitor region Df is represented as Tf, the monitor region of the wafer W corresponding to the pressure chamber 7g is represented as Dg, the initial film thickness of the monitor region Dg is represented as Tg, the monitor region of the wafer W corresponding to the pressure chamber 7h is represented as Dh, and the initial film thickness of the monitor region Dh is represented as Th.

[0058] The film thickness measuring device 8 measures the film thickness at a plurality of measurement points in each of the monitor regions Da to Dh and transmits those measurement values to the control device 30. The control device 30 determines the representative values of the plurality of film thickness measurement values in each monitor region Da to Dh as the initial film thicknesses Ta to Th. The representative value is, for example, the average value of the plurality of film thickness measurement values. Next, the control device 30 calculates the difference between the initial film thicknesses Ta to Th and the target film thickness Tt respectively, and calculates the target polishing amounts Ra' to Rh' for each monitor region Da to Dh.

[0059] Next, the control device 30 calculates predicted polishing amounts Ra to Rh in each monitor area Da to Dh using a response model stored in advance in the control device 30. Further, the control device 30 calculates, by optimization calculation, at least the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h and the polishing time so that the predicted polishing amounts Ra to Rh become values close to the target polishing amounts Ra' to Rh'.

[0060] In this optimization calculation, for example, the objective function shown in the following formula (1) is used. Specifically, in the optimization calculation, the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h and the retainer chamber 34, and the polishing time, at which the objective function including the difference between the predicted polishing amounts Ra to Rh and the target polishing amounts Ra' to Rh' becomes minimum, are calculated.

[0061] Objective function = Σ|Predicted polishing amount - Target polishing amount| 2 ···(1)

[0062] In the objective function shown in formula (1), only the polishing amount is considered, but the present embodiment is not limited to this example. For example, as shown in the following formula (2), it is also preferable that the objective function includes terms of the difference between the calculated optimal compressed fluid pressure and a preset reference compressed fluid pressure, and the difference between the calculated optimal compressed fluid pressure and the optimal compressed fluid pressure at the time of polishing the previous wafer.

[0063] Objective function = Σ|Predicted polishing amount - Target polishing amount| 2 + λΣ|Calculated optimal compressed fluid pressure - Reference compressed fluid pressure| 2 + γΣ|Calculated optimal compressed fluid pressure - Optimal compressed fluid pressure of the previous wafer| 2 ···(2)

[0064] Here, λ and γ are weighting coefficients that determine the weights of each term, and any real number of 0 or more can be set. By adding these terms, it is possible to suppress a large change in the calculated optimal compressed fluid pressure for each wafer to be processed and obtain stable optimal polishing conditions.

[0065] The pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h corresponds to the polishing pressure in each of the monitor regions Da to Dh. Further, the pressing force of the retainer ring 3 against the polishing pad 33 corresponds to the pressure of the compressed fluid supplied to the retainer chamber 34.

[0066] In the present embodiment, the monitor regions of the thickness of the wafer W are divided into a number of regions equal to the total number of the pressure chambers of the elastic film 5 corresponding to each of the pressure chambers 7a to 7h of the elastic film 5 (see FIGS. 6(a) and 6(b)). However, the monitor regions of the thickness of the wafer W may be further finely divided. As described above, the film thickness measuring device 8 can measure the film thickness at a plurality of measurement points in each of the monitor regions Da to Dh. Therefore, for example, the monitor regions of the thickness of the wafer W can be divided into monitor regions D1 to Dm corresponding to each measurement point MP of the film thickness by the film thickness measuring device 8 (see FIGS. 6(a) and 6(c)). Here, the subscript “m” corresponds to the number of measurement points.

[0067] In yet another method of finely dividing the monitor regions, the entire surface of the wafer or each of the monitor regions Da to Dh may be divided at equal intervals (for example, at intervals of 1 mm), and each divided region may be set as a new monitor region D1 to Dn (see FIG. 6(d)). Here, “n” is the total number of the finely divided monitor regions. When the monitor regions are divided in this way, there may be a case where there is no measurement point of the film thickness by the film thickness measuring device 8 within each monitor region. Therefore, by calculating the film thickness value at the representative point (for example, the center position of each monitor region) of each monitor region from the measured film thickness values at each measurement point by interpolation processing, the representative film thickness value (estimated film thickness value) of each finely divided monitor region can be calculated. In this way, by dividing the monitor regions of the thickness of the wafer W into more regions than the total number of the pressure chambers of the elastic film 5, the film thickness profile of the wafer W can be controlled more precisely.

[0068] When determining the monitor region of the film thickness of the wafer W, it is not necessary to select only one of the above-described monitor region setting methods, and a plurality of setting methods may be combined. For example, in the inner region of the wafer W (for example, the region corresponding to the pressure chambers 7a to 7f), the monitor region of the film thickness of the wafer W is set to the monitor regions Da to Df corresponding to the pressure chambers 7a to 7f, and in the outer region of the wafer W (for example, the region corresponding to the pressure chambers 7g to 7h), the monitor region of the film thickness of the wafer W may be set to the monitor region corresponding to each measurement point MP of the film thickness by the film thickness measuring device 8 (see FIGS. 6(a) and 6(e)).

[0069] Next, a method for creating a response model will be described. The response model is created, for example, by an experiment. In the experiment, first, a wafer for which a film thickness profile has already been acquired by the film thickness measuring device 8 is polished with a reference polishing recipe (that is, a predetermined polishing pressure and a predetermined polishing time), and the film thickness profile of the polished wafer is acquired by the film thickness measuring device 8. Next, a number of wafers different from the wafer W polished with the reference polishing recipe are polished by changing the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h of the elastic film 5 and the retainer chamber 34 from the pressure of the reference polishing recipe. At this time, the film thickness profiles before and after polishing of the number of wafers are acquired by the film thickness measuring device 8.

[0070] As described above, the film thickness measuring device 8 can measure the film thickness at a plurality of measurement points on the wafer W. Therefore, the control device 30 can calculate the polishing rate at each measurement point from the film thickness profiles before and after polishing of the wafer polished with the reference polishing recipe and the number of wafers. FIG. 7 is a conceptual diagram showing the polishing rates at each measurement point of a number of wafers in a graph. In FIG. 7, the vertical axis represents the polishing rate, and the horizontal axis represents the position in the radial direction of the wafer.

[0071] Next, the control device 30 calculates the increase amount of the polishing rate per unit polishing pressure (for example, 1 hPa) from the polishing rates at each measurement point of a large number of wafers. In this specification, the increase amount of the polishing rate per unit polishing pressure is referred to as the "response coefficient". Further, an offset amount D is calculated such that the predicted polishing amount R obtained by substituting the calculated response coefficient into the following formula (3) is equal to the actual polishing amount obtained by polishing with the above-mentioned reference polishing recipe.

[0072] Furthermore, from the calculated response coefficients and offset amounts at each measurement point, the response coefficients and offset amounts in each of the above-mentioned monitor regions are calculated by interpolation processing, thereby determining the response coefficients and offset amounts in each monitor region. FIG. 8 is a conceptual diagram showing the response coefficients at each measurement point of the wafer in a graph. In FIG. 8, the vertical axis represents the response coefficient, and the horizontal axis represents the position in the radial direction of the wafer.

[0073] The response coefficients and offset amounts calculated in this way are included in the response model. The control device 30 creates an optimal polishing recipe using the response model. Specifically, the control device 30 calculates the predicted polishing amount R from the pressures of the compressed fluids supplied to each of the pressure chambers 7a to 7h and the retainer chamber 34 and the polishing time by the following formula (3).

[0074] R = Tp·(C·X+D) ···(3)

[0075] Furthermore, by substituting the calculated predicted polishing amount R into formula (1) or formula (2), the objective function value is calculated. The control device 30 creates an optimal polishing recipe by obtaining, through optimization calculation, the pressures of the compressed fluids supplied to each of the pressure chambers 7a to 7h and the polishing time at which the objective function value is minimized. Here, since the above formula (1) or (2) can be expressed in the form of a quadratic equation of X, the optimal polishing recipe can be uniquely determined by using the quadratic programming method for the optimization calculation. In addition, other optimization calculation methods such as the gradient method like the steepest descent method or the Monte Carlo method can be used.

[0076] In Equation (3), R is a matrix consisting of the predicted polishing amounts R1 to Rm of the respective monitor regions D1 to Dm of the wafer W, Tp is the polishing time, C is a matrix consisting of the response coefficients in the respective monitor regions of the wafer W, X is a matrix consisting of the pressures of the compressed fluid supplied to the respective pressure chambers 7a to 7h, and D is a matrix consisting of the offsets of the respective monitor regions D1 to Dm of the wafer W. Here, the subscript "m" corresponds to the number of monitor regions of the wafer W.

[0077] R, C, D, and X are represented as matrices as shown in FIG. 9. In the matrices R, C, and D, each row corresponds to the respective monitor regions D1 to Dm of the wafer W. The matrix X multiplied by the matrix C corresponds to the pressure of the compressed fluid to be supplied to the respective pressure chambers 7a to 7h of the elastic film 5. Therefore, in the matrix X, "Pa" is the pressure of the compressed fluid to be supplied to the pressure chamber 7a of the elastic film 5, and "Pc" is the pressure of the compressed fluid to be supplied to the pressure chamber 7c of the elastic film 5.

[0078] Of the two subscripts of the response coefficient in the matrix C, the former subscript corresponds to the respective monitor regions D1 to Dm of the wafer W, and the latter subscript corresponds to the pressure of the compressed fluid to be supplied to the respective pressure chambers 7a to 7h. For example, the response coefficient C2b is the response coefficient with respect to the pressure of the compressed fluid supplied to the pressure chamber 7b in the monitor region D2 of the wafer W, and the response coefficient C3c is the response coefficient with respect to the pressure of the compressed fluid supplied to the pressure chamber 7c in the monitor region D3 of the wafer W.

[0079] The control device 30 calculates the pressures Pa to Ph of the compressed fluid to be supplied to the respective pressure chambers 7a to 7h of the elastic film 5 and the polishing time Tp by the above-described optimization calculation, and uses the pressures Pa to Ph of these compressed fluids and the polishing time Tp as the optimum polishing recipe. The optimum polishing recipe calculated in this way uses a plurality of film thickness measurement values in the respective monitor regions D1 to Dm of the wafer W to determine the polishing pressure of each pressure chamber 7a to 7h, so that the film thickness profile of the wafer W can be precisely controlled.

[0080] In one embodiment, an optimal polishing recipe may be calculated by adding the pressure Pi of the compressed fluid to be supplied to the retainer chamber 34 to the pressures Pa to Ph of the compressed fluid to be supplied to the pressure chambers 7a to 7h of the elastic film 5 and the polishing time Tp. In this case, instead of the response matrix C described above, a matrix C' composed of response coefficients C1a to Cmh corresponding to the pressures of the compressed fluid supplied to the pressure chambers 7a to 7h as shown in FIG. 10 and response coefficients C1i to Cmi corresponding to the pressure of the compressed fluid supplied to the retainer chamber 34 is used. Further, instead of the matrix X, a matrix X' composed of the pressures Pa to Ph of the compressed fluid to be supplied to the pressure chambers 7a to 7h of the elastic film 5 and the pressure Pi of the compressed fluid to be supplied to the retainer chamber 34 is used.

[0081] The response coefficients C1a to Cmh and the response coefficients C1i to Cmi can be determined as follows. First, in the experiment for determining the response coefficients C1a to Cmh, with the pressure of the compressed fluid supplied to the retainer chamber 34 fixed at a predetermined value, the pressures of the compressed fluid supplied to the pressure chambers 7a to 7h are changed to calculate the response coefficients C1a to Cmh. Next, with the pressures of the compressed fluid supplied to the pressure chambers 7a to 7h fixed at predetermined values, the pressure of the compressed fluid supplied to the retainer chamber 34 is changed to obtain the response coefficients C1i to Cmi.

[0082] The optimal polishing recipe obtained using the matrix C' of the response model described above takes into account not only the change in the polishing amount between the monitor regions D1 to Dm accompanying the change in the pressure of the compressed fluid supplied to the pressure chambers 7a to 7h, but also the change in the polishing amount between the monitor regions D1 to Dm accompanying the change in the pressing force of the retainer ring 3 on the polishing pad 33. Therefore, the film thickness profile of the wafer W can be controlled more precisely.

[0083] In one embodiment, the response model, that is, the matrix C (or the matrix C') composed of the response coefficients C1a to Cmh and the matrix D (or the matrix D') may be determined by simulation. Also in this case, the matrix C (or the matrix C') and the matrix D (or the matrix D') obtained by simulation are stored in the control device 30 in advance.

[0084] Returning to FIG. 5, the control device 30 conveys the wafer W to any one of the polishing units 14a to 14d and polishes the wafer W with the above-described optimal polishing recipe (see Step 3 in FIG. 5). The polishing of the wafer W is performed as follows. As shown in FIG. 2, the polishing head 37 and the polishing table 35 are rotated in the directions indicated by the arrows, and a polishing liquid (slurry) is supplied from the polishing liquid supply nozzle 38 onto the polishing pad 33. In this state, the polishing head 37 presses the wafer W against the polishing surface 33a of the polishing pad 33 for a polishing time corresponding to the optimal polishing recipe. When pressing the wafer W against the polishing pad 33, the pressures of the compressed fluid supplied to each of the pressure chambers 7a to 7h of the elastic film 5 and the retainer chamber 34 are adjusted to the pressures corresponding to the optimal polishing recipe. The surface of the wafer W is polished by the mechanical action of the abrasive grains contained in the polishing liquid and the chemical action of the polishing liquid. After the polishing is completed, dressing (conditioning) of the polishing surface 33a by the dressing device 40 is performed.

[0085] The dressing of the polishing pad 33 is performed as follows. While the dresser 41 rotates around the dresser shaft 45, pure water is supplied from the polishing liquid supply nozzle 38 onto the polishing pad 33. In this state, the dresser 41 is pressed against the polishing pad 33 by the air cylinder 47, and its dressing surface 41a is brought into sliding contact with the polishing surface 33a of the polishing pad 33. Further, the dresser arm 48 is swung around the dresser swing axis 49 to swing the dresser 41 in the radial direction of the polishing pad 33. In this way, the polishing pad 33 is slightly shaved off by the dresser 41, and its surface 33a is dressed (regenerated).

[0086] Next, the control device 30 conveys the polished wafer W to the first cleaning unit 16 and / or the second cleaning unit 18 for cleaning, and further conveys the cleaned wafer W to the drying unit 20 for drying. Further, the control device 30 conveys the polished wafer W to the film thickness measuring device 8 to obtain the film thickness profile of the polished wafer W (see step 4 in FIG. 5). The control device 30 stores the film thickness profiles of the wafer W before and after polishing and the optimal polishing recipe for polishing this wafer W.

[0087] Next, the control device 30 corrects the stored response model in order to create an optimal polishing recipe for polishing the next wafer W (see step 5 in FIG. 5). The correction of the response model is performed as follows.

[0088] The control device 30 calculates the actual polishing amount Rac in each monitor region D1 to Dm of the wafer W from the film thickness profiles of the wafer W before and after polishing that has already been polished. When calculating the actual polishing amount Rac, the control device 30 subtracts, from the initial film thickness values T1 to Tm in each monitor region D1 to Dm, the post-polishing film thickness values T1' to Tm' that are representative values of a plurality of film thickness measurement values corresponding to each monitor region D1 to Dm of the polished wafer W. The post-polishing film thickness values T1' to Tm' are, for example, the average values of a plurality of film thickness measurement values in each monitor region D1 to Dm, or the film thickness values at representative points of each monitor region calculated by interpolation from the post-polishing film thickness measurement values. The control device 30 calculates the actual polishing amounts Rac1 to Racm in each monitor region D1 to Dm by subtracting the post-polishing film thickness values T1' to Tm' from the initial film thicknesses T1 to Tm.

[0089] Next, the control device 30 calculates a correction coefficient K such that the predicted polishing amount R and the actual polishing amount Rac in the above formula (3) satisfy the following formula (4).

[0090] Rac = K·R ···(4)

[0091] Here, Rac is a matrix consisting of the actual polishing amounts Rac1 to Racm of each monitor region D1 to Dm, and K is a matrix consisting of correction factors corresponding to each monitor region D1 to Dm.

[0092] Next, as shown in the following equations (5) and (6), the control device 30 multiplies the above-described matrix C and matrix D by K obtained from equation (4), and calculates and stores a corrected response coefficient matrix Cadj and a corrected offset amount matrix Dadj.

[0093] Cadj = K·C ···(5)

[0094] Dadj = K·D ···(6)

[0095] Next, the control device 30 conveys the next wafer W to the film thickness measuring device 8, and acquires the film thickness profile of the next wafer W before polishing (see step 6 in FIG. 5).

[0096] Next, the control device 30 calculates the pressure of the compressed fluid supplied to each pressure chamber 7a to 7h and the retainer chamber 34 and the polishing time by the above-described optimization calculation using the corrected predicted polishing amount Radj obtained from the following equation (7). Radj = Tp·(Cadj·X + Dadj) ···(7) In equation (7), Radj is a matrix consisting of the predicted polishing amounts of each monitor region D1 to Dm of the next wafer W, and Tp is the polishing time of the next wafer W.

[0097] In the present embodiment, an optimum polishing recipe used for polishing the next wafer W is created using a response model corrected based on the film thickness profiles before and after polishing the previous wafer W. The film thickness profiles before and after polishing the previous wafer W are data reflecting the state of the polishing unit that actually polished the wafer W (for example, the surface properties of the polishing pad 33). Therefore, by creating an optimum polishing recipe used for polishing the next wafer W using a response model corrected based on the film thickness profiles before and after polishing the previous wafer W, the film thickness profile of the next wafer W can be controlled more precisely.

[0098] Next, the control device 30 polishes the next wafer W with the created optimal polishing recipe (see step 8 in FIG. 5). Next, the control device 30 transports the next wafer W after polishing to the first cleaning unit 16 and / or the second cleaning unit 18 for cleaning, and further transports the cleaned next wafer W to the drying unit 20 for drying. Further, the control device 30 transports the next wafer W after polishing to the film thickness measuring device 8 to obtain the film thickness profile of the wafer W after polishing (see step 9 in FIG. 5).

[0099] Furthermore, the control device 30 repeats steps 5 to 9 in FIG. 5. That is, before polishing the next wafer W, the control device 30 corrects the response model using the film thickness profiles before and after polishing the next wafer W and the optimal polishing recipe. Next, the film thickness measuring device 8 obtains the film thickness profile of the further next wafer W before polishing. Next, based on the corrected response model, an optimal polishing recipe for polishing the further next wafer W is created. Next, the control device 30 polishes the further next wafer W with the created optimal polishing recipe and obtains the film thickness profile of the further next wafer W after polishing. In this way, by correcting the response model every time the wafer W is polished, the film thickness profile of the next wafer W can be controlled more precisely.

[0100] In one embodiment, the control device 30 may be connected to be able to transmit and receive data to and from an arithmetic device 70 (refer to the dotted line in FIG. 1) provided outside the polishing device. In this case, the response model is pre-stored in the arithmetic device 70, and the control device 30 transmits the film thickness profile before polishing and the target film thickness to the arithmetic device 70. The arithmetic device 70 creates an optimal polishing recipe based on the polishing amount, which is the difference between the transmitted film thickness profile and the target film thickness, and the response model. Next, the arithmetic device 70 transmits the optimal polishing recipe to the control device 30 of the polishing device, and the control device 30 polishes the wafer W according to the received optimal polishing recipe. Further, the control device 30 sends the film thickness profile of the wafer W after polishing to the arithmetic device 70, and the arithmetic device 70 stores the film thickness profiles of the wafer W before and after polishing and the optimal polishing recipe used for polishing this wafer W. Further, the control device 30 corrects the response model using the film thickness profiles of the wafer W before and after polishing and the optimal polishing recipe used for polishing this wafer W.

[0101] When polishing the next wafer W, the control device 30 sends the film thickness profile of the next wafer W before polishing to the arithmetic device 70. The arithmetic device 70 creates an optimal polishing recipe for polishing the next wafer W based on the target polishing amount, which is the difference between the film thickness profile of the next wafer W before polishing and the target film thickness, and the corrected response model, and transmits it to the control device 30. The control device 30 polishes the next wafer W with the transmitted optimal polishing recipe. Further, the control device 30 transmits the film thickness profile of the next wafer W after polishing to the arithmetic device 70, and the arithmetic device 70 stores the film thickness profiles of the next wafer W before and after polishing and the optimal polishing recipe used for polishing the next wafer W. The arithmetic device 70 uses the film thickness profiles of the next wafer W before and after polishing and the optimal polishing recipe used for polishing the next wafer W for correcting the response model to be used when polishing the further next wafer W.

[0102] FIG. 11 is a perspective view schematically showing a polishing head according to another embodiment. Since the configuration of this embodiment that is not particularly described is the same as that of the above-described embodiment, the overlapping description thereof is omitted. Hereinafter, an example in which the polishing head 37 described with reference to FIG. 11 is mounted on the polishing unit 14a shown in FIG. 1 will be described, but this polishing head 37 may be mounted on the polishing units 14b to 14d.

[0103] The polishing head 37 shown in FIG. 11 includes a head body 2 that presses the wafer W against the polishing pad 33, and a retainer ring 3 disposed so as to surround the wafer W. The retainer ring 3 is configured to be vertically movable independently of the head body 2. The retainer ring 3 projects radially outward from the head body 2. During polishing of the wafer W, the retainer ring 3 contacts the polishing surface 33a of the polishing pad 33 and presses the polishing pad 33 outside the wafer W while rotating.

[0104] The polishing head 37 further includes a rotating ring 71 in which a plurality of rollers are disposed inside, and a stationary ring 81. The rotating ring 71 is fixed to the upper surface of the retainer ring 3 and is configured to be rotatable together with the retainer ring 3. The stationary ring 81 is disposed on the rotating ring 71. The rotating ring 71 rotates together with the retainer ring 3, but the stationary ring 81 does not rotate and remains stationary.

[0105] The polishing unit 14a includes a plurality of local load applying devices that apply local loads to a part of the retainer ring 3. In the illustrated example, the polishing unit 14a includes two local load applying devices, namely, a first local load applying device 83A and a second local load applying device 83B. The local load applying devices 83A and 83B are disposed above the retainer ring 3. The local load applying devices 83A and 83B are fixed to the head arm 42 (see FIG. 2). The retainer ring 3 during polishing rotates around its axis, but the local load applying devices 83A and 83B do not rotate integrally with the retainer ring 3 and are stationary. The stationary ring 81 is connected to the local load applying devices 83A and 83B. The first local load applying device 83A is disposed on the upstream side of the retainer ring 3 (one side of the retainer ring 3 into which the polishing surface 33a flows) in the advancing direction of the polishing surface 33a of the polishing pad 33, and the second local load applying device 83B is disposed on the downstream side of the retainer ring 3 (the opposite side of the retainer ring 3 from which the polishing surface 33a flows out) in the advancing direction of the polishing surface 33a of the polishing pad 33.

[0106] The plurality of local load applying devices 83A and 83B include a plurality of pressing members 84A and 84B that apply downward local loads to the stationary ring 81, a plurality of bridges 85A and 85B, a plurality of air cylinders 86A and 86B that generate downward forces, a plurality of pressure regulators R10 and R11 that adjust the pressure of the compressed fluid in the air cylinders 86A and 86B, a plurality of linear guides 87A and 87B, a plurality of guide rods 88A and 88B, and a plurality of unit bases 89A and 89B.

[0107] Specifically, the first local load applying device 83A includes a first pressing member 84A, a first bridge 85A, a first air cylinder 86A, a first pressure regulator R10, a first linear guide 87A, a first guide rod 88A, and a first unit base 89A. The second local load applying device 83B includes a second pressing member 84B, a second bridge 85B, a second air cylinder 86B, a second pressure regulator R11, a second linear guide 87B, a second guide rod 88B, and a second unit base 89B.

[0108] The piston rod 101a of the first air cylinder 86A is connected to the first pressing member 84A via the first bridge 85A, and the end of the first pressing member 84A is connected to the stationary ring 81. Therefore, the force generated by the first air cylinder 86A is transmitted to the first pressing member 84A, and the first pressing member 84A applies a local load to a part of the stationary ring 81. Similarly, the piston rod 101b of the second air cylinder 86B is connected to the second pressing member 84B via the second bridge 85B, and the end of the second pressing member 84B is connected to the stationary ring 81. Therefore, the force generated by the second air cylinder 86B is transmitted to the second pressing member 84B, and the second pressing member 84B applies a local load to a part of the stationary ring 81.

[0109] In the present embodiment, the combination of the first air cylinder 86A and the first pressure regulator R10 constitutes a first actuator 90A that adjusts the local load applied from the first pressing member 84A to the stationary ring 81, and the combination of the second air cylinder 86B and the second pressure regulator R11 constitutes a second actuator 90B that adjusts the local load applied from the second pressing member 84B to the stationary ring 81. In one embodiment, each of the first actuator 90A and the second actuator 90B may be composed of a combination of a servo motor, a ball screw mechanism, and a motor driver.

[0110] The first pressing member 84A includes two pressing rods 103a, and the second pressing member 84B includes two pressing rods 103b. The pressing rods 103a and 103b are connected to the stationary ring 81. The first pressing member 84A is configured to apply a local load to the upstream portion of the stationary ring 81 in the advancing direction of the polishing surface 33a of the polishing pad 33, and the second pressing member 84B is configured to apply a local load to the downstream portion of the stationary ring 81 in the advancing direction of the polishing surface 33a of the polishing pad 33.

[0111] The local load applying devices 83A and 83B are fixed to the head arm 42 (see FIG. 2) via the unit bases 89A and 89B. Therefore, during the polishing of the wafer W, while the polishing head 37 and the wafer W are rotating, the local load applying devices 83A and 83B are stationary. Similarly, during the polishing of the wafer W, while the rotating ring 71 is rotating together with the polishing head 37, the stationary ring 81 is stationary.

[0112] The local load applying devices 83A and 83B have the same configuration. The following description relates to the first local load applying device 83A, but is similarly applicable to the second local load applying device 83B. The first unit base 89A is attached with a first air cylinder 86A and a first linear guide 87A. The piston rod 101a of the first air cylinder 86A and the first guide rod 88A are connected to the first bridge 85A. The first guide rod 88A is supported by the first linear guide 87A so as to be vertically movable with low friction. By the first linear guide 87A, the first bridge 85A can move smoothly up and down without tilting.

[0113] The air cylinders 86A and 86B are connected to a compressed fluid supply source 32 (see FIG. 4) through the gas transfer lines F1 and F2. The pressure regulators R10 and R11 are respectively provided in the gas transfer lines F1 and F2 and are arranged in the pressure adjusting device 65 shown in FIG. 4. The compressed fluid from the compressed fluid supply source is independently supplied to the air cylinders 86A and 86B through the pressure regulators R10 and R11.

[0114] The pressure regulators R10 and R11 can independently adjust the pressure of the compressed fluid in the air cylinders 86A and 86B, whereby the air cylinders 86A and 86B can generate forces independently of each other.

[0115] The pressure regulators R10 and R11 are electrically connected to the control device 30 shown in FIG. 1. During the polishing of the wafer W, the control device 30 issues a command to any one of the pressure regulators R10 and R11 to adjust the pressure of the compressed fluid in the air cylinder 86A or the air cylinder 86B.

[0116] The forces generated by the air cylinders 86A and 86B are transmitted to the bridges 85A and 85B, respectively. The bridges 85A and 85B are connected to the stationary ring 81 via the pressing members 84A and 84B, and the pressing members 84A and 84B transmit the forces of the air cylinders 86A and 86B applied to the bridges 85A and 85B to the stationary ring 81. That is, the first pressing member 84A presses a part of the stationary ring 81 with a local load corresponding to the force generated by the first air cylinder 86A, and the second pressing member 84B presses a part of the stationary ring 81 with a local load corresponding to the force generated by the second air cylinder 86B.

[0117] Each of the local load applying devices 83A and 83B applies a downward local load to a part of the retainer ring 3 via the stationary ring 81 and the rotating ring 71. That is, the downward local load is transmitted to the retainer ring 3 through the stationary ring 81 and the rotating ring 71.

[0118] The polishing device polishes the wafer W while rotating the rotating ring 71 fixed to the retainer ring 3 together with the retainer ring 3 and applying a local load from the first pressing member 84A or the second pressing member 84B to the stationary ring 81. During the polishing of the wafer W, the retainer ring 3 contacts the polishing surface 33a of the polishing pad 33, presses the polishing pad 33 outside the wafer W while rotating, and applies a downward local load to a part of the polishing surface 33a.

[0119] FIG. 12 is a longitudinal sectional view schematically showing a state when the retainer ring presses against the polishing surface. As shown in FIG. 12, when the retainer ring 3 applies a downward local load to a part of the polishing surface 33a, a part of the polishing surface 33a bulges upward. The polishing surface 33a that bulges upward applies a local upward force to the wafer W. In this specification, this local upward force is referred to as a local repulsive force. In FIG. 12, for the sake of explanation, only the bulged part of the polishing surface 33a is in contact with the wafer W, but during actual polishing, the entire lower surface (surface to be polished) of the wafer W is in contact with the polishing surface 33a. The polishing rate of the part of the wafer W that receives the local repulsive force increases. The magnitude of the local repulsive force depends on the magnitude of the force with which the retainer ring 3 presses against the polishing pad 33, and the polishing rate changes depending on the magnitude of the local repulsive force. That is, the larger the local repulsive force, the larger the polishing rate. The position where the local repulsive force is generated depends on the position of the local load applied by the retainer ring 3 to the polishing surface 33a.

[0120] Therefore, by polishing the wafer W while applying a local load from the first pressing member 84A or the second pressing member 84B to the stationary ring 81, a local repulsive force corresponding to each local load can be generated, and the polishing rate of the part of the wafer W that receives the local repulsive force can be changed. For example, when the control device 30 wants to increase the local load applied by the first pressing member 84A, it issues a command to the pressure regulator R10 to increase the pressure of the compressed fluid in the air cylinder 86A. When the control device 30 wants to increase the local load applied by the second pressing member 84B, it issues a command to the pressure regulator R11 to increase the pressure of the compressed fluid in the air cylinder 86B.

[0121] Thus, the local loads applied by the local load applying devices 83A and 83B to the retainer ring 3 (in this embodiment, corresponding to the pressures of the compressed fluids supplied to the air cylinders 86A and 86B) also become factors that affect the film thickness profile of the wafer W after polishing.

[0122] Therefore, in the present embodiment, the control device 30 (see FIG. 1) calculates the above-described response model (i.e., the matrix C composed of the response coefficients C1a to Cmh and the matrix D composed of the offset values) in consideration of the local load in addition to the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h and the retainer chamber 34. Further, the control device 30 creates an optimal polishing recipe using the response model in which the local load is also taken into account (see step 3 in FIG. 5).

[0123] The thus obtained optimal polishing recipe takes into account the change in the polishing amount between the monitor regions D1 to Dm accompanying the change in the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h, the change in the polishing amount between the monitor regions D1 to Dm accompanying the change in the pressing force of the retainer ring 3 against the polishing pad 33, and the change in the polishing amount between the monitor regions D1 to Dm accompanying the change in the local load. Therefore, the film thickness profile of the wafer W can be controlled more precisely.

[0124] Furthermore, the correction response model used for calculating the optimal polishing recipe when polishing the next wafer W also takes into account the change in the polishing amount between the monitor regions D1 to Dm accompanying the change in the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h, the change in the polishing amount between the monitor regions D1 to Dm accompanying the change in the pressing force of the retainer ring 3 against the polishing pad 33, and the change in the polishing amount between the monitor regions D1 to Dm accompanying the change in the local load. Therefore, the film thickness profile of the next wafer W polished with the optimal polishing recipe created based on the correction response model can also be controlled more precisely.

[0125] FIG. 13 is a graph showing the transition of the in-plane uniformity in each embodiment when a plurality of wafers are continuously polished using the optimal polishing recipes according to a plurality of embodiments. In FIG. 13, the vertical axis represents the in-plane uniformity, and the horizontal axis represents the order of the wafers polished continuously. The in-plane uniformity shown in FIG. 13 is represented by the difference between the maximum value and the minimum value of the film thickness measurement values of the polished wafer W.

[0126] In FIG. 13, the dotted line shows the transition of in-plane uniformity when creating an optimal polishing recipe and correcting the response model using a response model created considering only the change in the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h (hereinafter referred to as Example 1). The thick solid line shows the transition of in-plane uniformity when creating an optimal polishing recipe and correcting the response model using a response model created considering the change in the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h and the change in the pressure of the compressed fluid supplied to the retainer chamber 34 (hereinafter referred to as Example 2). The thin solid line shows the transition of in-plane uniformity when creating an optimal polishing recipe and correcting the response model using a response model created considering the change in the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h, the change in the pressure of the compressed fluid supplied to the retainer chamber 34, and the change in the local load (hereinafter referred to as Example 3).

[0127] Note that the graph of the two-dot chain line in FIG. 13 is a graph showing a reference example, and is a graph representing the transition of in-plane uniformity when a plurality of wafers W are continuously polished using a conventional polishing pressure adjustment method. In the conventional polishing pressure adjustment method, the difference between the film thickness and the target film thickness in the monitor regions Da to Dh corresponding to each of the pressure chambers 7a to 7h of the wafer W after polishing is calculated, and the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h is changed so that the difference in each of the monitor regions Da to Dh becomes 0, and the next wafer W is polished. Further, in the conventional polishing pressure adjustment method, the above-described response model considering the change in the polishing amount between the monitor regions Da to Dh is not created, and it is assumed that the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h and the polishing amount of each of the monitor regions Da to Dh correspond one-to-one. Therefore, the next wafer W is polished with a polishing recipe that does not consider the change in the polishing amount between the monitor regions Da to Dh accompanying the change in the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h.

[0128] As can be seen from FIG. 13, the in-plane uniformity of Examples 1 to 3 is significantly improved compared to the in-plane uniformity of the reference example. Therefore, it can be understood that by using the response model described above and performing continuous polishing of the wafer W while creating an optimal polishing recipe and correcting the response model, the film thickness profile can be precisely controlled.

[0129] Furthermore, it can be seen that the in-plane uniformity of Example 2 and Example 3 is superior to the in-plane uniformity of Example 1. Therefore, it can be understood that at least by using a response model created in consideration of the change in the pressure of the compressed fluid supplied to each of the pressure chambers 7a to 7h and the change in the pressure of the compressed fluid supplied to the retainer chamber 34, the film thickness profile can be precisely controlled.

[0130] FIG. 14 is a flowchart showing a polishing method according to another embodiment. Since the steps of this embodiment not particularly described are the same as the steps of the flowchart shown in FIG. 5, the overlapping description thereof is omitted. The polishing method shown in FIG. 14 is preferably performed using a polishing apparatus having a control device 30 connected to an arithmetic unit 70 (see the dotted line in FIG. 1) because a large amount of data processing may be required. Therefore, in the following description, a method of polishing the wafer W using the arithmetic unit 70 and the control device 30 will be described. However, when the control device 30 has sufficient data processing capabilities, the polishing method shown in FIG. 14 may be executed only by the control device 30 without going through the arithmetic unit 70. In this case, “arithmetic unit 70” in the following description may be appropriately read as “control device 30”.

[0131] As shown in FIG. 14, in this embodiment, first, the film thickness profiles before polishing of a plurality of wafers W and a response model are collected and stored in a storage (not shown) provided in the arithmetic unit 70 (see step 1 in FIG. 14). Further, the arithmetic unit 70 classifies the plurality of film thickness profiles before polishing into a plurality of groups to which film thickness profiles similar to each other belong (see step 2 in FIG. 14). The film thickness profile before polishing of the wafer W and the response model are associated with the group to which the wafer W belongs.

[0132] Each time the control device 30 performs polishing of the wafer W, it transmits the film thickness profile before and after polishing of the wafer W and the optimal polishing recipe used for polishing the wafer W to the arithmetic unit 70. Each time a combination of the film thickness profile before and after polishing of the wafer W and the optimal polishing recipe is sent from the control device 30 to the arithmetic unit 70, the arithmetic unit 70 corrects the response model using the film thickness profile before and after polishing and the optimal polishing recipe. Further, the arithmetic unit 70 classifies and accumulates the combination of the film thickness profile before polishing of the wafer W and the corrected response model into any one of a plurality of groups.

[0133] For classification based on the film thickness profile before polishing, for example, a shape matching index obtained by calculation can be used. Examples of the shape matching index include absolute mean, root mean square, average film thickness difference, correlation coefficient, and GOF (Good of Fitting) value. The shape matching index is an index for obtaining the degree of shape matching (similarity) between two film thickness profiles. When the first film thickness profile is T1 to Tm, the average value of T1 to Tm is Tave, the second film thickness profile is T'1 to T'm, and the average value of T'1 to T'm is T'ave, they can be calculated by any one of the following formulas (8) to (11).

[0134]

Number

[0135]

Number

[0136] Average film thickness difference = |Tave - T'ave| ···(10)

[0137]

Number

[0138] In addition, GOF is a commonly used indicator that shows the degree of match between two profiles. The absolute mean, root mean square, and average film thickness difference indicate a higher shape match (more similar shapes) as the values are smaller, while the correlation coefficient and GOF indicate a higher shape match as the values are larger.

[0139] The arithmetic unit 70 classifies the film thickness profile using at least one of these shape match indices. Specifically, the arithmetic unit 70 obtains in advance the film thickness profiles representing each group, calculates the shape match indices between the film thickness profile before polishing and the film thickness profiles representing each group, and classifies the film thickness profile before polishing into the group that exceeds a preset threshold value and has the highest shape match degree. As the film thickness profile representing each group, for example, an average film thickness profile obtained by averaging the film thickness values at each measurement point of the film thickness profiles classified into each group can be used.

[0140] When there is no group whose shape match degree exceeds the threshold value, that is, when there is no group to be classified based on the shape match index, the arithmetic unit 70 creates a new group. By this operation, a plurality of groups in which similar film thickness profiles are gathered are created.

[0141] In one embodiment, a machine learning device (not shown) may be provided in the arithmetic unit 70, and the machine learning device may be used to classify the film thickness profile of the wafer W before polishing. In this case, the film thickness profile of the wafer W before polishing is input to the machine learning device. The machine learning device outputs the group to which the input film thickness profile should belong. When the machine learning device determines that there is no group to which the input film thickness profile should belong, the machine learning device outputs a command for the arithmetic unit 70 to create a new group.

[0142] Next, the control device 30 takes out the wafer W from the substrate cassette placed on the load port 12 (see FIG. 1), transports it to the film thickness measuring device 8, and acquires the film thickness profile of the wafer W before polishing (see step 3 in FIG. 14). This step 3 corresponds to step 1 in FIG. 5. Next, the control device 30 transmits the film thickness profile of the wafer W before polishing to the arithmetic device 70, and the arithmetic device 70 selects the group to which the received film thickness profile (i.e., the wafer to be polished) should belong (see step 4 in FIG. 14).

[0143] When selecting a group, the arithmetic device 70 uses the shape matching index described above. Specifically, the arithmetic device 70 calculates the shape matching index for each group using the received film thickness profile and the representative film thickness profile of each group, and selects the group with the shape matching degree exceeding the threshold value and being the highest as the group to which the film thickness profile should belong.

[0144] Next, the arithmetic device 70 creates an optimal polishing recipe using the response model of the group to which the film thickness profile belongs (see step 5 in FIG. 14). Since this step 5 corresponds to step 2 in FIG. 5, the description of the method for creating the optimal polishing recipe is omitted.

[0145] In step 4, if there is no group with the shape matching degree exceeding the threshold value, the arithmetic device 70 selects the group having the highest threshold value of the shape matching index, creates an optimal polishing recipe using the response model of this group. Further, the arithmetic device 70 creates a new group to which this film thickness profile belongs.

[0146] Next, the arithmetic device 70 transmits the created optimal polishing recipe to the control device 30, and the control device 30 executes the polishing of the wafer W based on the received polishing recipe (see step 6 in FIG. 14).

[0147] Next, the control device 30 transports the polished wafer W to the first cleaning unit 16 and / or the second cleaning unit 18 for cleaning, and further transports the cleaned wafer W to the drying unit 20 for drying. Further, the control device 30 transports the polished wafer W to the film thickness measuring device 8 to obtain the film thickness profile of the polished wafer W (see step 7 in FIG. 14). The control device 30 corrects the response model using the film thickness profiles of the wafer W before and after polishing and the optimal polishing recipe, and further records it in association with the group to which the film thickness profile belongs (see step 8 in FIG. 12).

[0148] The steps from step 9 to step 12 in FIG. 12 are the same as the steps from step 6 to step 9 in FIG. 5, so duplicate explanations are omitted.

[0149] According to the present embodiment, the optimal polishing model of the first wafer is created using the optimal polishing recipe and the response model belonging to the group having a film thickness profile similar to the film thickness profile of the first wafer W. Therefore, the film thickness profile of the first wafer W can be precisely controlled.

[0150] FIG. 15 is a schematic diagram showing a polishing unit according to another embodiment. The configuration of this embodiment not particularly described is the same as that of the embodiment described with reference to FIG. 2, so duplicate explanations are omitted. In FIG. 15, the dressing device 40 (see FIG. 2) is not shown. Hereinafter, an example in which the polishing unit 14b shown in FIG. 1 is the polishing unit described with reference to FIG. 13 will be described. However, the polishing unit shown in FIG. 15 may be arranged in at least any one of the polishing units 14a, 14c, 14d of the polishing apparatus.

[0151] The configuration of the polishing unit 14b shown in FIG. 15 is different from that of the polishing unit 14a shown in FIG. 2 in that it includes a film thickness sensor 52 that acquires a film thickness signal that changes according to the film thickness of the wafer W. The film thickness sensor 52 is installed inside the polishing table 35, and every time the polishing table 35 makes one rotation, it acquires film thickness signals at a plurality of measurement points in each of a plurality of monitor regions D1 to Dm of the wafer W. Examples of the film thickness sensor 52 include an optical sensor and an eddy current sensor.

[0152] During the polishing of the wafer W, the film thickness sensor 52 rotates together with the polishing table 35 and acquires a film thickness signal while crossing the surface of the wafer W as shown by the symbol A. This film thickness signal is an index value that directly or indirectly indicates the film thickness and changes according to the decrease in the film thickness of the wafer W. The film thickness sensor 52 is connected to the control device 30, and the film thickness signal is sent to the control device 30. The control device 30 can acquire the film thickness profile of the wafer W from the film thickness signal sent from the film thickness sensor 52.

[0153] With the high integration and high density of semiconductor devices, wirings having a multilayer structure are formed on the wafer W. Therefore, if the wafer W is polished with a single polishing unit until a desired film is exposed, the polishing time becomes long, and as a result, defects in the wafer are caused due to an increase in the polishing temperature or the deposition of by-products on the polishing pad, or the flatness of the wafer surface decreases. Depending on the film types of the multilayer structure formed on the wafer W, a plurality of polishing steps may be performed over a plurality of polishing units. For example, after polishing the top metal film of the wafer W with the first polishing unit, the dielectric film layer formed under the metal film may be polished with the second polishing unit.

[0154] When performing a plurality of polishing steps continuously in this way, if the film thickness profile after polishing is acquired with the film thickness measuring instrument 8 as shown in FIG. 2 every time one polishing step is completed, the throughput will decrease. Therefore, in this embodiment, the film thickness sensor 52 is used to acquire the film thickness profile of the wafer W before polishing and / or the film thickness profile of the wafer W after polishing.

[0155] FIG. 16 is a flowchart showing a polishing method according to still another embodiment. Since the steps of this embodiment not particularly described are the same as the steps of the flowchart shown in FIG. 5, duplicate explanations thereof are omitted.

[0156] As shown in FIG. 16, the control device 30 first takes out the wafer W from the substrate cassette placed on the load port 12 (see FIG. 1) and transports it to the first polishing unit (for example, the polishing unit 14b) having the film thickness sensor 52, and first polishes the wafer W according to a predetermined polishing recipe in the first polishing unit (see step 1 in FIG. 16). During the first polishing of the wafer W, the control device 30 monitors the film thickness of the wafer W obtained from the measured value of the film thickness sensor 52, and stops the first polishing when the film thickness reaches a predetermined threshold value (that is, when the thickness of the uppermost layer film of the wafer W reaches a predetermined target value).

[0157] When the first polishing of the wafer W is completed, the control device 30 performs water polishing in which the wafer W is polished while supplying pure water to the polishing pad 33 on the polishing table 35, and while performing this water polishing, the film thickness profile of the wafer W after the first polishing is acquired by the film thickness sensor 52 (see step 2 in FIG. 16). During the water polishing, the polishing of the wafer W does not substantially progress. By performing water polishing, it is possible to remove the polishing liquid, polishing debris, by-products, etc. on the polishing pad 33, so that an accurate film thickness profile can be acquired even after the first polishing of the wafer W. The film thickness sensor 52 functions as a film thickness measuring device that acquires the film thickness profile of the wafer W before the second polishing, which is required to create an optimal polishing recipe for the second polishing of the wafer W.

[0158] Next, in order to perform the second polishing of the wafer W, the control device 30 transports the wafer W after the first polishing to a second polishing unit (for example, polishing unit 14a) other than the first polishing unit (see step 3 in FIG. 16). At this time, the control device 30 creates an optimal recipe for the second polishing of the wafer W based on the film thickness profile of the wafer W before the second polishing (that is, the film thickness profile obtained by the film thickness sensor 52 after the first polishing) and the above-described response model (see step 4 in FIG. 16). The method for creating this optimal recipe for the second polishing is performed in the same manner as step 2 in FIG. 5 described above.

[0159] Next, the control device 30 performs the second polishing of the wafer W in accordance with the optimal recipe for the second polishing in the second polishing unit to which the wafer W was transported in step 3 (see step 5 in FIG. 16). Next, the control device 30 transports the wafer W after the second polishing to the first cleaning unit 16 and / or the second cleaning unit 18 for cleaning, and further transports the cleaned wafer W to the drying unit 20 for drying. Further, the control device 30 transports the wafer W after the second polishing to the film thickness measuring device 8 (see FIG. 1) to obtain the film thickness profile of the wafer W after the second polishing (see step 6 in FIG. 16).

[0160] Next, the control device 30 transports the next wafer W to the first polishing unit and executes the first polishing of the next wafer W (see step 7 in FIG. 16). After the completion of the first polishing of the next wafer W, the control device 30 obtains the film thickness profile of the next wafer W after the first polishing by the film thickness sensor 52 while performing the water polishing of the next wafer W (see step 8 in FIG. 16).

[0161] Next, in order to perform the second polishing of the next wafer W, the control device 30 transports the next wafer W after the first polishing to the second polishing unit (see step 9 in FIG. 16). At this time, the control device 30 corrects the second polishing response model in order to create an optimal recipe for the second polishing of the next wafer W (see step 9 in FIG. 16). The correction of the second polishing response model for the next wafer W is performed based on the optimal recipe for the second polishing of the previous wafer W and the film thickness profiles before and after the second polishing, as described in step 5 of FIG. 5.

[0162] Specifically, the control device 30 calculates the actual polishing amount Rac of the second polishing in each monitor region D1 to Dm of the wafer W from the film thickness profiles before and after the second polishing of the wafer W that has already been polished for the first and second times, and calculates the correction coefficient K so that the predicted polishing amount R and the actual polishing amount Rac in the above formula (3) satisfy the above formula (4). Next, the control device 30 multiplies the above-described matrix C and matrix D by K obtained from formula (4), and calculates and stores the corrected response coefficient matrix Cadj and the corrected offset amount matrix Dadj corrected by the above formulas (5) and (6).

[0163] Next, the control device 30 calculates, by the above-described optimization calculation, the optimum recipe for the second polishing of the next wafer W, including at least the pressures of the compressed fluid supplied to each of the pressure chambers 7a to 7h and the retainer chamber 34, and the polishing time, from the obtained Cadj, Dadj, and the target polishing amount R' of the second polishing of the next wafer W.

[0164] Next, the control device 30 performs the second polishing on the next wafer W according to the calculated optimum recipe for the second polishing (see step 12 in FIG. 16). Next, the control device 30 conveys the next wafer W after the second polishing to the first cleaning unit 16 and / or the second cleaning unit 18 for cleaning, and further conveys the cleaned next wafer W to the drying unit 20 for drying. Further, the control device 30 conveys the next wafer W after the second polishing to the film thickness measuring device 8 to obtain the film thickness profile of the wafer W after the second polishing (see step 13 in FIG. 16).

[0165] Furthermore, the control device 30 repeats steps 7 to 13 in FIG. 16. That is, before further polishing the next wafer W with the second polishing unit, the control device 30 further uses the film thickness sensor 52 of the first polishing unit to obtain the film thickness profile of the next wafer W before the second polishing. Further, the control device 30 corrects the second polishing response model used for polishing the next wafer W based on the film thickness profiles of the next wafer W before and after the second polishing. Next, the control device 30 creates an optimal recipe for the second polishing of the next wafer W based on the corrected response model. Next, the control device 30 polishes the next wafer W with the optimal recipe for the second polishing calculated based on the corrected response model, and obtains the film thickness profile of the next wafer W after the second polishing. In this way, by correcting the second polishing response model every time the wafer W is polished and polishing with the optimal recipe for the second polishing calculated based thereon, the film thickness profile of the next wafer W can be controlled more precisely.

[0166] According to the present embodiment, even when it is necessary to perform a plurality of polishing steps, it is possible to precisely control the film thickness profile of the wafer W while suppressing a decrease in throughput.

[0167] FIG. 17 is a flowchart showing a polishing method according to still another embodiment. Since the steps of this embodiment not specifically described are the same as the steps of the flowchart shown in FIG. 16, the overlapping description thereof is omitted.

[0168] In the polishing method shown in the flowchart of FIG. 17, the control device 30 first takes out the wafer W from the substrate cassette placed on the load port 12 (see FIG. 1), conveys it to the film thickness measuring device 8, and obtains the film thickness profile of the wafer W before the first polishing (see step 1 in FIG. 17).

[0169] Next, the control device 30 creates an optimal recipe for the first polishing of the wafer W based on the film thickness profile of the wafer W before the first polishing and the response model described above (see step 2 in FIG. 17). The method for creating this optimal recipe for the first polishing is performed in the same manner as step 2 in FIG. 5 described above. Next, the control device 30 performs the first polishing on the wafer W according to the optimal recipe for the first polishing (see step 3 in FIG. 17), and transports the wafer W to a second polishing unit (for example, the polishing unit 14b) having a film thickness sensor 52 (see step 4 in FIG. 17).

[0170] Next, before performing the second polishing, the control device 30 obtains the film thickness profile of the wafer W after the first polishing by the film thickness sensor 52 while performing the water polishing described above (see step 5 in FIG. 17). Next, the control device 30 performs the second polishing of the wafer W in the second polishing unit according to a predetermined polishing recipe (see step 6 in FIG. 17). Next, the control device 30 transports the wafer W after the second polishing to the first cleaning unit 16 and / or the second cleaning unit 18 for cleaning, and further transports the next cleaned wafer W to the drying unit 20 for drying.

[0171] Next, the control device 30 transports the next wafer W to the film thickness measuring device 8 and obtains the film thickness profile of the next wafer W before the first polishing (see step 7 in FIG. 15). Next, the control device 30 corrects the first polishing response model based on the optimal recipe for the first polishing of the wafer W and the film thickness profiles before and after the first polishing (see step 8 in FIG. 17). Next, the control device 30 creates an optimal polishing recipe for polishing the next wafer W using the corrected first polishing response model (see step 9 in FIG. 17).

[0172] Next, the control device 30 performs the first polishing on the next wafer W according to the created optimal polishing recipe (see step 10 in FIG. 17), and transports the next wafer W after the first polishing to the second polishing unit (see step 11 in FIG. 17). In the second polishing unit, the film thickness profile of the next wafer W after the first polishing is obtained by the film thickness sensor 52 during water polishing (see step 12 in FIG. 17), and then the second polishing of the next wafer W is performed (see step 13 in FIG. 17).

[0173] Furthermore, the control device 30 repeats steps 7 to 13 in FIG. 17. That is, the control device 30 corrects the first polishing response model using the first polishing optimal recipe for the next wafer W and the film thickness profiles before and after the first polishing. Further, before the next wafer W is first polished by the first polishing unit, the film thickness measuring device 8 is used to obtain the film thickness profile of the next wafer W before the first polishing. Further, the control device 30 creates a first polishing optimal recipe for polishing the next wafer W based on the corrected first polishing response model and the film thickness profile of the next wafer W before the first polishing. Next, the control device 30 first polishes the next wafer W with the created first polishing optimal recipe and obtains the film thickness profile of the next wafer W after the first polishing. In this way, each time the wafer W is polished, the first polishing response model is corrected, and by polishing with the first polishing optimal recipe calculated based on the corrected model, the film thickness profile of the next wafer W can be controlled more precisely.

[0174] Also in this embodiment, when it is necessary to perform a plurality of polishing steps, it is possible to precisely control the film thickness profile of the wafer W while suppressing a decrease in throughput.

[0175] FIG. 18 is the first half of a flowchart showing a polishing method according to still another embodiment, and FIG. 19 is the second half of a flowchart showing a polishing method according to still another embodiment. Since the steps of this embodiment not specifically described are the same as the steps of the flowcharts shown in FIGS. 16 and 17, the overlapping descriptions are omitted. In the polishing method shown in the flowcharts of FIGS. 18 and 19, the first polishing and the second polishing are performed in one polishing unit having the film thickness sensor 52. Therefore, the polishing apparatus can omit the film thickness measuring device 8 (see FIG. 1).

[0176] As shown in FIG. 18, first, the control device 30 takes out the wafer W from the substrate cassette placed on the load port 12 (see FIG. 1), transports it to the first polishing unit having the film thickness sensor 52, and while performing water polishing in the first polishing unit, the film thickness profile of the wafer W before the first polishing is acquired by the film thickness sensor 52 (see step 1 in FIG. 18).

[0177] Next, the control device 30 creates the first polishing optimal recipe for the wafer W based on the film thickness profile of the wafer W before the first polishing and the first polishing response model described above (see step 2 in FIG. 18). The method for creating this first polishing optimal recipe is performed in the same manner as step 2 in FIG. 5 described above. After finishing the water polishing, the control device 30 performs the first polishing of the wafer W according to the first polishing optimal recipe (see step 3 in FIG. 18).

[0178] Next, after the first polishing is completed, the control device 30 restarts the water polishing and acquires the film thickness profile of the wafer W after the first polishing during the water polishing (see step 4 in FIG. 18). The film thickness profile acquired after this first polishing corresponds to the film thickness profile before the second polishing. Therefore, the control device 30 creates the second polishing optimal recipe for the wafer W based on the film thickness profile acquired in step 4 and the second polishing response model described above (see step 5 in FIG. 18), and performs the second polishing of the wafer W with the second polishing optimal recipe (see step 6 in FIG. 18).

[0179] When the second polishing is completed, the control device 30 starts the water polishing and acquires the film thickness profile of the wafer W after the second polishing during the water polishing (see step 7 in FIG. 18). Next, the control device 30 transports the next wafer W after the second polishing to the first cleaning unit 16 and / or the second cleaning unit 18 for cleaning, and further transports the cleaned next wafer W to the drying unit 20 for drying.

[0180] Next, the control device 30 conveys the next wafer W to the polishing unit and acquires the film thickness profile of the next wafer W before the first polishing using the film thickness sensor 52 (see step 8 in FIG. 19). Further, the control device 30 corrects the first polishing response model using the first polishing optimal recipe for the wafer W and the film thickness profiles before and after the first polishing (see step 9 in FIG. 19), and creates a first polishing optimal recipe for polishing the next wafer W (see step 10 in FIG. 19). The creation of the first polishing optimal recipe for the next wafer W is performed based on the previously corrected response model for the first polishing and the target polishing amount for the first polishing of the wafer W.

[0181] Next, the control device 30 polishes the next wafer W with the corrected first polishing optimal recipe (see step 11 in FIG. 19). Further, the control device 30 starts water polishing, and during the water polishing, acquires the film thickness profile of the next wafer W after the first polishing (see step 12 in FIG. 19).

[0182] Next, the control device 30 corrects the second polishing response model using the second polishing optimal recipe for the wafer W and the film thickness profiles before and after the second polishing (see step 13 in FIG. 19), and creates a second polishing optimal recipe for polishing the next wafer W (see step 14 in FIG. 19). The creation of the optimal polishing recipe for the next wafer W is performed based on the previously corrected response model for the second polishing and the target polishing amount for the second polishing of the wafer W. Next, the control device 30 polishes the next wafer W with the corrected second polishing optimal recipe (see step 15 in FIG. 19). After the completion of the second polishing, the control device 30 starts water polishing, and during the water polishing, acquires the film thickness profile of the next wafer W after the second polishing using the film thickness sensor 52 (see step 16 in FIG. 19). Further, the control device 30 repeats steps 8 to 16 in FIG. 19.

[0183] According to this embodiment, since a plurality of polishing processes can be performed by one polishing unit, it is possible to suppress a decrease in throughput as much as possible while precisely controlling the film thickness profile. Further, since a generally expensive film thickness measuring instrument 8 (see FIG. 1) can be omitted, a polishing apparatus capable of precisely controlling the film thickness profile can be provided at low cost.

[0184] The above-described embodiment is described for the purpose of enabling a person having ordinary knowledge in the technical field to which the present invention pertains to practice the present invention. Various modifications of the above embodiment can be naturally made by those skilled in the art, and the technical idea of the present invention can be applied to other embodiments. Therefore, the present invention is not limited to the described embodiments, but is construed in the broadest scope in accordance with the technical idea defined by the claims.

Explanation of Reference Numerals

[0185] 2 Head body 3 Retainer ring 5 Elastic film 7a, 7b, 7c, 7d, 7e, 7f, 7g, 7h Pressure chamber 8 Film thickness measuring instrument 10 Housing 12 Load port 14a, 14b, 14c, 14d Polishing unit 16, 18 Cleaning unit 20 Drying unit 30 Control device 33 Polishing pad 34 Retainer chamber 37 Polishing head (substrate holding device) 38 Polishing liquid supply nozzle 52 Film thickness sensor (film thickness measuring instrument) 65 Pressure adjusting device 70 Arithmetic unit 83A, 83B Local load applying device 86A, 86B Air cylinder 90A, 90B Actuator

Claims

1. At least one polishing unit including a polishing table for supporting a polishing pad and a substrate holding device for pressing a substrate against the polishing pad; A film thickness measuring device for measuring the film thickness profile of the substrate; A control device for at least controlling the operations of the polishing unit and the film thickness measuring device; and The substrate holding device includes: An elastic film forming a plurality of pressure chambers for pressing the substrate; A head body to which the elastic film is attached; and A retainer ring disposed so as to surround the substrate; The control device stores in advance a response model for calculating a predicted polishing amount in each monitor region, which is created in consideration of a change in the polishing amount between a plurality of monitor regions of the substrate accompanying a change in the pressure in each pressure chamber; When the response model represents a matrix R consisting of the predicted polishing amounts of each monitor region of the substrate, represents the polishing time of the substrate as Tp, represents a matrix C consisting of the response coefficients of each monitor region of the substrate, represents a matrix X consisting of the pressures of the compressed fluid supplied to each pressure chamber, and represents a matrix D consisting of the offset amounts of each monitor region of the substrate, it is expressed by the formula R = Tp · (C · X + D); The response coefficient is an increase amount of the polishing rate per unit polishing pressure; The offset amount is calculated so that the predicted polishing amount becomes equal to the actual polishing amount obtained by polishing with a reference polishing recipe that polishes at a predetermined polishing pressure and a predetermined polishing time; The control device: Acquires the film thickness profile of the substrate before polishing using the film thickness measuring device; Polishes the substrate with an optimum polishing recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount, which is the difference between the film thickness profile of the substrate before polishing and the target film thickness of the substrate, and the response model; Polishes the next substrate with a new optimum polishing recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount of the next substrate, the optimum polishing recipe, and the response model corrected so that the predicted polishing amount matches the actual polishing amount using the film thickness profiles of the substrate before and after polishing; The correction of the response model: Calculates the actual polishing amount in each monitor region of the substrate from the film thickness profiles of the substrate before and after polishing; When the actual polishing amount is represented as Rac, a correction coefficient K is calculated so that the predicted polishing amount and the actual polishing amount satisfy Rac = K・R. A polishing apparatus, wherein the correction coefficient is multiplied by a matrix C composed of response coefficients of each monitor region of the substrate and a matrix D composed of offset amounts of each monitor region of the substrate. **Claim 2** The polishing apparatus according to claim 1, wherein the optimum polishing recipe is created using an optimization calculation that minimizes an objective function including at least the target polishing amount and a term of a difference between the predicted polishing amount calculated using the response model. **Claim 3** The polishing apparatus according to claim 2, wherein the objective function further includes a term of a difference between the compression fluid pressure of the optimum polishing recipe and a preset reference compression fluid pressure, and / or a term of a difference between the compression fluid pressure of the optimum polishing recipe and the compression fluid pressure of the optimum polishing recipe of a wafer polished previously. **Claim 4** The polishing apparatus according to claim 2 or 3, wherein the optimization calculation is a quadratic programming method. **Claim 5** The polishing apparatus according to any one of claims 1 to 4, wherein the number of the plurality of monitor regions is larger than the number of the plurality of pressure chambers. **Claim 6** The response model is a response model created in consideration of changes in the polishing amount between a plurality of monitor regions of the substrate accompanying changes in the pressing force of the retainering against the polishing pad. The polishing apparatus according to claim 1, wherein the optimum polishing recipe further includes the pressing force of the retainering. **Claim 7** The polishing apparatus according to any one of claims 1 to 6, wherein the film thickness measuring device is configured to be able to measure the film thickness at a plurality of measurement points set in each of the plurality of monitor regions. **Claim 8** The polishing apparatus according to any one of claims 1 to 7, wherein the response model includes a response coefficient representing an increase amount of the polishing rate per unit polishing pressure in each of the plurality of monitor regions. **Claim 9** The polishing apparatus further includes a plurality of local load applying devices that apply a local load to a part of the retainering. The polishing apparatus according to any one of claims 1 to 8, wherein the response model is created in further consideration of changes in the polishing amount between the plurality of monitor regions accompanying changes in the local load. **Claim 10** At least one polishing unit including a polishing table for supporting a polishing pad and a substrate holding device for pressing a substrate against the polishing pad. A film thickness measuring device for measuring a film thickness profile of the substrate. A control device that at least controls the operations of the polishing unit and the film thickness measuring device, The substrate holding device includes An elastic film that forms a plurality of pressure chambers for pressing the substrate, A head body to which the elastic film is attached, A retainer ring arranged to surround the substrate, The control device Pre-accumulates a film thickness profile before polishing of a plurality of substrates and a response model when each of the plurality of substrates is polished, the response model being created in consideration of changes in the polishing amounts between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber, and calculating a predicted polishing amount in each monitor region, Pre-classifies the film thickness profiles before polishing of the plurality of substrates into a plurality of groups to which film thickness profiles similar to each other belong, The response model is represented by the formula R = Tp·(C·X + D), where R represents a matrix composed of the predicted polishing amounts in each monitor region of the substrate, Tp represents the polishing time of the substrate, C represents a matrix composed of the response coefficients in each monitor region of the substrate, X represents a matrix composed of the pressures of the compressed fluid supplied to each pressure chamber, and D represents a matrix composed of the offset amounts in each monitor region of the substrate, The response coefficient is the increase amount of the polishing rate per unit polishing pressure, The offset amount is calculated so that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe at a predetermined polishing pressure and a predetermined polishing time, The control device Uses the film thickness measuring device to obtain the film thickness profile of the substrate before polishing, Determines the group to which the film thickness profile of the substrate before polishing belongs from the plurality of groups, Polishes the substrate with an optimal polishing recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, the optimal polishing recipe being created based on the target polishing amount, which is the difference between the film thickness profile of the substrate before polishing and the target film thickness of the substrate, and the response model associated with the determined group, Polishes the next substrate with a new optimal polishing recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, the new optimal polishing recipe being created based on the target polishing amount of the next substrate, the optimal polishing recipe, and the response model corrected so that the predicted polishing amount matches the actual polishing amount using the film thickness profiles of the substrate before and after polishing, The correction of the response model Calculate the actual polishing amount in each monitor region of the substrate from the film thickness profiles before and after polishing the substrate. When the actual polishing amount is represented as Rac, calculate a correction coefficient K such that the predicted polishing amount and the actual polishing amount satisfy Rac = K·R. A polishing apparatus, wherein the correction coefficient is multiplied by a matrix C composed of response coefficients of each monitor region of the substrate and a matrix D composed of offset amounts of each monitor region of the substrate.

11. A plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing a substrate against the polishing pad. A control device for at least controlling the operation of the polishing unit. The substrate holding device includes: An elastic film forming a plurality of pressure chambers for pressing the substrate. A head body to which the elastic film is attached. A retainer ring disposed so as to surround the substrate. The substrate is a substrate polished by a plurality of polishing steps including a first polishing and a second polishing performed by a polishing unit different from the polishing unit where the first polishing is performed. The polishing unit that performs the first polishing has a film thickness sensor capable of measuring the film thickness profile of the substrate. The control device stores in advance a second polishing response model for calculating a predicted polishing amount of the second polishing in each monitor region, which is created in consideration of changes in the polishing amounts between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber. The second polishing response model is expressed by the formula R = Tp·(C·X + D), where a matrix composed of the predicted polishing amounts of each monitor region of the substrate is represented as R, the polishing time of the substrate is represented as Tp, a matrix composed of the response coefficients of each monitor region of the substrate is represented as C, a matrix composed of the pressures of the compressed fluid supplied to each pressure chamber is represented as X, and a matrix composed of the offset amounts of each monitor region of the substrate is represented as D. The response coefficient is the increase amount of the polishing rate per unit polishing pressure. The offset amount is calculated such that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe at a predetermined polishing pressure and a predetermined polishing time. The control device: After performing the first polishing, use the film thickness sensor to obtain the film thickness profile of the substrate before the second polishing. Based on the target polishing amount of the second polishing, which is the difference between the film thickness profile of the substrate before the second polishing and the target film thickness of the second polishing of the substrate, and the second polishing response model, the substrate is second polished with a second polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time. Using the target polishing amount of the second polishing of the next substrate, the second polishing response model corrected so that the predicted polishing amount of the second polishing matches the actual polishing amount using the second polishing optimal recipe and the film thickness profiles of the substrate before and after the second polishing of the substrate, the next substrate is second polished with a new second polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time. The correction of the second polishing response model is as follows: Calculate the actual polishing amount in each monitor area of the substrate from the film thickness profiles of the substrate before and after the second polishing. When the actual polishing amount is represented as Rac, calculate a correction coefficient K such that the predicted polishing amount and the actual polishing amount satisfy Rac = K·R. A polishing apparatus, wherein the correction coefficient is multiplied by a matrix C composed of response coefficients of each monitor area of the substrate and a matrix D composed of offset amounts of each monitor area of the substrate.

12. A plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing the substrate against the polishing pad. A film thickness measuring device for measuring the film thickness profile of the substrate. Comprising a control device for at least controlling the operations of the polishing unit and the film thickness measuring device. The substrate holding device includes: An elastic membrane forming a plurality of pressure chambers for pressing the substrate. A head body to which the elastic membrane is attached. A retainer ring arranged to surround the substrate. The substrate is a substrate polished by a plurality of polishing steps including a first polishing and a second polishing performed by a polishing unit different from the polishing unit where the first polishing is performed. The polishing unit for performing the second polishing has a film thickness sensor capable of measuring the film thickness profile of the substrate. The control device stores in advance a first polishing response model for calculating the predicted polishing amount of the first polishing in each monitor area, which is created in consideration of the change in the polishing amount between the plurality of monitor areas of the substrate accompanying the change in the pressure in each pressure chamber. The first polishing response model represents a matrix R composed of predicted polishing amounts in each monitor region of the substrate, represents the polishing time of the substrate as Tp, represents a matrix C composed of response coefficients in each monitor region of the substrate, represents a matrix X composed of pressures of compressed fluid supplied to each pressure chamber, and represents a matrix D composed of offset amounts in each monitor region of the substrate. When represented by the formula R = Tp · (C · X + D), The response coefficient is the increase amount of the polishing rate per unit polishing pressure, The offset amount is calculated such that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe at a predetermined polishing pressure and a predetermined polishing time, The control device, uses the film thickness measuring device to obtain the film thickness profile of the substrate before the first polishing, Based on the target polishing amount of the first polishing, which is the difference between the film thickness profile of the substrate before the first polishing and the target film thickness of the first polishing of the substrate, and the first polishing response model, the substrate is first polished with a first polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, After performing the first polishing, the substrate is transported to the polishing unit having the film thickness sensor, and the film thickness profile of the substrate after the first polishing is obtained using the film thickness sensor, Based on the target polishing amount of the first polishing of the next substrate, the first polishing response model corrected so that the predicted polishing amount of the first polishing matches the actual polishing amount using the first polishing optimal recipe and the film thickness profiles of the substrate before and after the first polishing, the next substrate is first polished with a new first polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, The correction of the first polishing response model, calculates the actual polishing amount in each monitor region of the substrate from the film thickness profiles of the substrate before and after the first polishing, When the actual polishing amount is represented as Rac, a correction coefficient K is calculated so that the predicted polishing amount and the actual polishing amount satisfy Rac = K · R, The correction coefficient is multiplied by a matrix C composed of response coefficients in each monitor region of the substrate and a matrix D composed of offset amounts in each monitor region of the substrate, a polishing apparatus.

13. A polishing table for supporting a polishing pad, a substrate holding device for pressing a substrate against the polishing pad, and a film thickness sensor capable of measuring the film thickness profile of the substrate, a plurality of polishing units including the same, A control device for at least controlling the operation of the polishing unit, The substrate holding device, An elastic film forming a plurality of pressure chambers for pressing the substrate, A head body to which the elastic film is attached, A retainer ring arranged so as to surround the substrate, The substrate is a substrate polished by a plurality of polishing steps including a first polishing and a second polishing performed in a polishing unit different from the polishing unit in which the first polishing is performed, The control device stores in advance a first polishing response model and a second polishing response model that respectively calculate the predicted polishing amount of the first polishing and the predicted polishing amount of the second polishing in each monitor region, which are created in consideration of the change in the mutual polishing amount in a plurality of monitor regions of the substrate accompanying the change in the pressure in each pressure chamber. When the first polishing response model and the second polishing response model are respectively represented by a matrix R composed of the predicted polishing amounts of each monitor region of the substrate, the polishing time of the substrate is represented by Tp, a matrix C composed of the response coefficients of each monitor region of the substrate is represented, a matrix X composed of the pressures of the compressed fluid supplied to each pressure chamber is represented, and a matrix D composed of the offset amounts of each monitor region of the substrate is represented, they are represented by the formula R = Tp·(C·X + D). The response coefficient is the increase amount of the polishing rate per unit polishing pressure. The offset amount is calculated so that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe that polishes at a predetermined polishing pressure and a predetermined polishing time. The control device, Conveys the substrate to any one of the plurality of polishing units, and uses the film thickness sensor to obtain the film thickness profile of the substrate before the first polishing. Based on the target polishing amount of the first polishing, which is the difference between the film thickness profile of the substrate before the first polishing and the target film thickness of the first polishing of the substrate, and the first polishing response model, the substrate is first polished with a first polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time. Using the film thickness sensor, the film thickness profile of the substrate before the second polishing is obtained. The substrate is subjected to second polishing with a second polishing optimal recipe that includes at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, the second polishing target polishing amount being the difference between the film thickness profile of the substrate before the second polishing and the target film thickness of the second polishing of the substrate, and being created based on the second polishing response model. The film thickness profile of the substrate after the second polishing is obtained using the film thickness sensor. A new first polishing optimal recipe that includes at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time is created based on the target polishing amount of the first polishing of the next substrate, the first polishing optimal recipe, and the film thickness profiles of the substrate before and after the first polishing, and the first polishing response model corrected so that the predicted polishing amount of the first polishing matches the actual polishing amount, and the next substrate is subjected to first polishing with the new first polishing optimal recipe. A new second polishing optimal recipe that includes at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time is created based on the target polishing amount of the second polishing of the next substrate, the second polishing optimal recipe, and the film thickness profiles of the substrate before and after the second polishing, and the second polishing response model corrected so that the predicted polishing amount of the second polishing matches the actual polishing amount, and the next substrate is subjected to second polishing with the new second polishing optimal recipe. The correction of the first polishing response model and the correction of the second polishing response model are each The actual polishing amount in each monitor region of the substrate is calculated from the film thickness profiles of the substrate before and after the first polishing of the substrate. When the actual polishing amount is represented as Rac, a correction coefficient K is calculated so that the predicted polishing amount and the actual polishing amount satisfy Rac = K·R. A polishing apparatus, wherein the correction is performed by multiplying the correction coefficient by a matrix C composed of response coefficients of each monitor region of the substrate and a matrix D composed of offset amounts of each monitor region of the substrate.

14. A polishing method in which a substrate held by a substrate holding device having an elastic film that forms a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate is pressed against a polishing pad supported by a polishing table for polishing, The film thickness profile of the substrate before polishing is obtained using a film thickness measuring device. Polishing the substrate with an optimal polishing recipe that includes at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount, which is the difference between the film thickness profile of the substrate before polishing and the target film thickness of the substrate, and the response model; Polishing the next substrate with a new optimal polishing recipe that includes at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount of the next substrate and the response model corrected so that the predicted polishing amount matches the actual polishing amount using the optimal polishing recipe and the film thickness profiles of the substrate before and after polishing; The response model is a model for calculating the predicted polishing amount of each monitoring region of the substrate, which is created in consideration of the change in the polishing amount between the plurality of monitoring regions of the substrate accompanying the change in the pressure of each pressure chamber. When a matrix consisting of the predicted polishing amounts of each monitoring region of the substrate is represented by R, the polishing time of the substrate is represented by Tp, a matrix consisting of the response coefficients of each monitoring region of the substrate is represented by C, a matrix consisting of the pressures of the compressed fluid supplied to each pressure chamber is represented by X, and a matrix consisting of the offset amounts of each monitoring region of the substrate is represented by D, it is expressed by the formula R = Tp · (C · X + D); The response coefficient is the increase in the polishing rate per unit polishing pressure; The offset amount is calculated so that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe at a predetermined polishing pressure and a predetermined polishing time; The correction of the response model is Calculating the actual polishing amount in each monitoring region of the substrate from the film thickness profiles of the substrate before and after polishing; When the actual polishing amount is represented by Rac, calculating a correction coefficient K so that the predicted polishing amount and the actual polishing amount satisfy Rac = K · R; A polishing method, which is performed by multiplying the correction coefficient by a matrix C consisting of the response coefficients of each monitoring region of the substrate and a matrix D consisting of the offset amounts of each monitoring region of the substrate.

15. The polishing method according to claim 14, wherein the optimal polishing recipe is created using an optimization calculation that minimizes an objective function including at least the target polishing amount and a term of the difference between the predicted polishing amount calculated using the response model.

16. The polishing method according to claim 15, wherein the objective function further includes a term of the difference between the compression fluid pressure of the optimal polishing recipe and a preset reference compression fluid pressure, and / or a term of the difference between the compression fluid pressure of the optimal polishing recipe and the compression fluid pressure of the optimal polishing recipe of the wafer polished previously.

17. The polishing method according to claim 15 or 16, wherein the optimization calculation is a quadratic programming method.

18. The polishing method according to any one of claims 14 to 17, wherein the number of the plurality of monitor regions is larger than the number of the plurality of pressure chambers.

19. The response model is a response model created in consideration of the change in the polishing amount between the plurality of monitor regions of the substrate accompanying the change in the pressing force of the retainering to the polishing pad, The polishing method according to claim 14, wherein the optimal polishing recipe further includes the pressing force of the retainering.

20. The polishing method according to any one of claims 14 to 19, wherein the film thickness measuring device measures the film thickness at a plurality of measurement points set for each of the plurality of monitor regions.

21. The polishing method according to any one of claims 14 to 20, wherein the response model includes a response coefficient representing an increase amount of the polishing rate per unit polishing pressure in each of the plurality of monitor regions.

22. The polishing method according to any one of claims 14 to 21, wherein the response model is created in further consideration of the change in the polishing amount between the plurality of monitor regions accompanying the change in the local load applied to a part of the retainering by a plurality of local load applying devices.

23. A polishing method in which a substrate held by a substrate holding device having an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainering disposed so as to surround the substrate is pressed against a polishing pad supported by a polishing table for polishing, A film thickness profile before polishing of a plurality of substrates and a response model when each of the plurality of substrates is polished are preliminarily stored, the response model being created in consideration of the change in the polishing amount between the plurality of monitor regions of the substrate accompanying the change in the pressure in each pressure chamber, and calculating a predicted polishing amount of each monitor region in the plurality of monitor regions of the substrate, The film thickness profiles before polishing of the plurality of substrates are preliminarily classified into a plurality of groups to which film thickness profiles similar to each other belong, Obtain the film thickness profile of the substrate before polishing, and determine the group to which the film thickness profile belongs from the plurality of groups. Based on the target polishing amount, which is the difference between the film thickness profile of the substrate before polishing and the target film thickness of the substrate, and the response model associated with the determined group, polish the substrate with an optimal polishing recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time. Based on the target polishing amount of the next substrate, the optimal polishing recipe, and the film thickness profiles of the substrate before and after polishing, polish the next substrate with a new optimal polishing recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the response model corrected so that the predicted polishing amount matches the actual polishing amount. The response model is a model for calculating the predicted polishing amount of each monitoring region of the substrate, which is created considering the change in the polishing amount between the plurality of monitoring regions of the substrate accompanying the change in the pressure of each pressure chamber. When a matrix consisting of the predicted polishing amounts of each monitoring region of the substrate is represented as R, the polishing time of the substrate is represented as Tp, a matrix consisting of the response coefficients of each monitoring region of the substrate is represented as C, a matrix consisting of the pressures of the compressed fluid supplied to each pressure chamber is represented as X, and a matrix consisting of the offset amounts of each monitoring region of the substrate is represented as D, it is expressed by the formula R = Tp · (C · X + D). The response coefficient is the increase amount of the polishing rate per unit polishing pressure. The offset amount is calculated so that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe at a predetermined polishing pressure and a predetermined polishing time. The correction of the response model is Calculate the actual polishing amount in each monitoring region of the substrate from the film thickness profiles of the substrate before and after polishing. When the actual polishing amount is represented as Rac, calculate a correction coefficient K so that the predicted polishing amount and the actual polishing amount satisfy Rac = K · R. A polishing method performed by multiplying the correction coefficient by a matrix C consisting of the response coefficients of each monitoring region of the substrate and a matrix D consisting of the offset amounts of each monitoring region of the substrate.

24. A method of polishing a substrate by a plurality of polishing steps performed in a plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing the substrate against the polishing pad, wherein the substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate, wherein the plurality of polishing steps include a first polishing and a second polishing performed in a polishing unit different from the polishing unit in which the first polishing is performed, the polishing unit performing the first polishing has a film thickness sensor capable of measuring a film thickness profile of the substrate, preparing in advance a second polishing response model for calculating a predicted polishing amount of the second polishing in each monitor region, which is created in consideration of changes in the polishing amount between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber, after performing the first polishing, using the film thickness sensor to obtain a film thickness profile of the substrate before the second polishing, polishing the substrate by a second polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount of the second polishing, which is the difference between the film thickness profile of the substrate before the second polishing and the target film thickness of the second polishing of the substrate, and the second polishing response model, polishing the next substrate by a new second polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount of the second polishing of the next substrate, the second polishing optimum recipe, and the second polishing response model corrected so that the predicted polishing amount of the second polishing matches the actual polishing amount using the film thickness profiles of the substrate before and after the second polishing, The second polishing response model is a model for calculating the predicted polishing amount of each monitoring region of the substrate, which is created by taking into account the change in the polishing amount between the plurality of monitoring regions of the substrate accompanying the change in the pressure of each pressure chamber. When a matrix composed of the predicted polishing amounts of each monitoring region of the substrate is represented by R, the polishing time of the substrate is represented by Tp, a matrix composed of the response coefficients of each monitoring region of the substrate is represented by C, a matrix composed of the pressures of the compressed fluid supplied to each pressure chamber is represented by X, and a matrix composed of the offset amounts of each monitoring region of the substrate is represented by D, it is expressed by the formula R = Tp·(C·X + D). The response coefficient is the increase amount of the polishing rate per unit polishing pressure. The offset amount is calculated so that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe at a predetermined polishing pressure and a predetermined polishing time. The correction of the second polishing response model calculates the actual polishing amount in each monitoring region of the substrate from the film thickness profiles before and after the second polishing of the substrate. When the actual polishing amount is represented by Rac, a correction coefficient K is calculated so that the predicted polishing amount and the actual polishing amount satisfy Rac = K·R. The polishing method is performed by multiplying the correction coefficient by a matrix C composed of the response coefficients of each monitoring region of the substrate and a matrix D composed of the offset amounts of each monitoring region of the substrate.

25. A method for polishing a substrate by a plurality of polishing steps performed in a plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing the substrate against the polishing pad, wherein the substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring arranged to surround the substrate. The plurality of polishing steps include a first polishing and a second polishing performed in a polishing unit different from the polishing unit in which the first polishing is performed. The polishing unit for performing the second polishing has a film thickness sensor capable of measuring the film thickness profile of the substrate. A first polishing response model for calculating the predicted polishing amount of the first polishing in each monitoring region is prepared in advance, which is created by taking into account the change in the polishing amount between the plurality of monitoring regions of the substrate accompanying the change in the pressure in each pressure chamber. Using a film thickness measuring device, obtain the film thickness profile of the substrate before the first polishing; Based on the target polishing amount of the first polishing, which is the difference between the film thickness profile of the substrate before the first polishing and the target film thickness of the first polishing of the substrate, and the first polishing response model, polish the substrate with a first polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time; After performing the first polishing, transfer the substrate to a polishing unit having the film thickness sensor; Using the film thickness sensor, obtain the film thickness profile of the substrate after the first polishing; Based on the target polishing amount of the first polishing of the next substrate, the first polishing optimal recipe, and the film thickness profiles of the substrate before and after the first polishing, create a new first polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, based on the first polishing response model corrected so that the predicted polishing amount of the first polishing matches the actual polishing amount, and polish the next substrate with the new first polishing optimal recipe; The first polishing response model is a model for calculating the predicted polishing amount of each monitoring region of the substrate, created by considering the change in the polishing amount between the plurality of monitoring regions of the substrate accompanying the change in the pressure of each pressure chamber. When a matrix composed of the predicted polishing amounts of each monitoring region of the substrate is represented as R, the polishing time of the substrate is represented as Tp, a matrix composed of the response coefficients of each monitoring region of the substrate is represented as C, a matrix composed of the pressures of the compressed fluid supplied to each pressure chamber is represented as X, and a matrix composed of the offset amounts of each monitoring region of the substrate is represented as D, it is expressed by the formula R = Tp·(C·X + D); The response coefficient is the increase amount of the polishing rate per unit polishing pressure; The offset amount is calculated so that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe at a predetermined polishing pressure and a predetermined polishing time; The correction of the first polishing response model is as follows: Calculate the actual polishing amount in each monitoring region of the substrate from the film thickness profiles of the substrate before and after the first polishing; When the actual polishing amount is represented as Rac, calculate a correction coefficient K so that the predicted polishing amount and the actual polishing amount satisfy Rac = K·R; A polishing method in which the correction coefficient is multiplied by a matrix C composed of response coefficients of each monitor region of the substrate and a matrix D composed of offset amounts of each monitor region of the substrate.

26. A method of polishing a substrate by a plurality of polishing steps performed by a plurality of polishing units including a polishing table for supporting a polishing pad and a substrate holding device for pressing the substrate against the polishing pad, wherein the substrate holding device includes an elastic film forming a plurality of pressure chambers for pressing the substrate, a head body to which the elastic film is attached, and a retainer ring disposed so as to surround the substrate. The plurality of polishing steps include a first polishing and a second polishing performed in a polishing unit different from the polishing unit in which the first polishing is performed. The polishing unit performing the first polishing and the polishing unit performing the second polishing each have a film thickness sensor capable of measuring a film thickness profile of the substrate. A first polishing response model and a second polishing response model for calculating, respectively, a predicted polishing amount of the first polishing and a predicted polishing amount of the second polishing in each monitor region, which are created in consideration of changes in the polishing amounts between a plurality of monitor regions of the substrate accompanying changes in the pressure in each pressure chamber, are prepared in advance. The substrate is transported to the polishing unit performing the first polishing, and the film thickness profile of the substrate before the first polishing is obtained using the film thickness sensor. The substrate is first polished with a first polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount of the first polishing, which is the difference between the film thickness profile of the substrate before the first polishing and the target film thickness of the first polishing of the substrate, and the first polishing response model. The film thickness profile of the substrate before the second polishing is obtained using the film thickness sensor. The substrate is second polished with a second polishing optimum recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, which is created based on the target polishing amount of the second polishing, which is the difference between the film thickness profile of the substrate before the second polishing and the target film thickness of the second polishing of the substrate, and the second polishing response model. The film thickness profile of the substrate after the second polishing is obtained using the film thickness sensor. Using the target polishing amount of the first polishing of the next substrate, the first polishing optimal recipe, and the film thickness profiles of the substrate before and after the first polishing, the first polishing response model corrected so that the predicted polishing amount of the first polishing matches the actual polishing amount, and based on this, polishing the next substrate with a new first polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, Using the target polishing amount of the second polishing of the next substrate, the second polishing optimal recipe, and the film thickness profiles of the substrate before and after the second polishing, the second polishing response model corrected so that the predicted polishing amount of the second polishing matches the actual polishing amount, and based on this, polishing the next substrate with a new second polishing optimal recipe including at least the pressure of the compressed fluid supplied to the plurality of pressure chambers and the polishing time, The first polishing response model and the second polishing response model are each models for calculating the predicted polishing amount of each monitoring region of the substrate, which are created in consideration of the change in the polishing amount between the plurality of monitoring regions of the substrate accompanying the change in the pressure of each pressure chamber. When a matrix composed of the predicted polishing amounts of each monitoring region of the substrate is represented by R, the polishing time of the substrate is represented by Tp, a matrix composed of the response coefficients of each monitoring region of the substrate is represented by C, a matrix composed of the pressures of the compressed fluid supplied to each pressure chamber is represented by X, and a matrix composed of the offset amounts of each monitoring region of the substrate is represented by D, it is represented by the formula R = Tp·(C·X + D). The response coefficient is the increase amount of the polishing rate per unit polishing pressure, The offset amount is calculated so that the predicted polishing amount is equal to the actual polishing amount obtained by polishing with a reference polishing recipe at a predetermined polishing pressure and a predetermined polishing time, The correction of the first polishing response model is Calculating the actual polishing amount in each monitoring region of the substrate from the film thickness profiles of the substrate before and after the first polishing, When the actual polishing amount is represented by Rac, calculating a correction coefficient K so that the predicted polishing amount and the actual polishing amount satisfy Rac = K·R, The correction is performed by multiplying the correction coefficient by a matrix C composed of the response coefficients of each monitoring region of the substrate and a matrix D composed of the offset amounts of each monitoring region of the substrate, The correction of the second polishing response model is Calculate the actual polishing amount in each monitor region of the substrate from the film thickness profiles before and after the second polishing of the substrate, When the actual polishing amount is represented as Rac, calculate a correction coefficient K such that the predicted polishing amount and the actual polishing amount satisfy Rac = K·R, A polishing method performed by multiplying the correction coefficient by a matrix C composed of response coefficients of each monitor region of the substrate and a matrix D composed of offset amounts of each monitor region of the substrate.

Images