Double-sided polishing device and double-sided polishing method

Abstract

In order to provide a double-sided polishing device capable of suppressing deviation of a workpiece shape from a target shape when ending final polishing, and obtaining a desired workpiece shape, according to the present invention, a control unit (30) for controlling a polishing machine (10) on the basis of a measurement result of a thickness measuring device (20) includes a first learner (31) that is trained regarding a correlative relation among main polishing conditions, outer peripheral flatness, and in-plane flatness, a second learner (32) that is trained regarding a correlative relation among end point polishing conditions, a degree of change in in-plane flatness when main polishing is stopped, a degree of change in outer peripheral flatness when the main polishing is stopped, and an amount of change of outer peripheral flatness during end point polishing, a main polishing condition correction unit (33) that corrects the main polishing conditions at the time of correction under a correction target value of in-plane flatness obtained from learning results of the first learner (31) and main polishing conditions obtained by inputting a target value of outer peripheral flatness to the first learner (31), and a main polishing stopping condition setting unit (34) that sets main polishing stopping conditions using amount of change in outer peripheral flatness obtained from learning results of the second learner (32).

Description

Double-sided polishing apparatus and double-sided polishing method

[0001] The present invention relates to a double-sided polishing apparatus and a double-sided polishing method for polishing the front and back surfaces of a disk-shaped workpiece.

[0002] Conventionally, when polishing the front and back surfaces of a disk-shaped workpiece such as a silicon wafer using upper and lower platens, a double-sided polishing machine is known that controls the distance between the platens based on an optimal distance obtained by inputting, for example, a desired workpiece flatness into an artificial intelligence model that has learned the relationship between the distance between the platens and the workpiece flatness (see, for example, Patent Document 1). Also known is a double-sided polishing machine that stops double-sided polishing of the workpiece when the shape index of the entire workpiece reaches a set value for the shape index of the entire workpiece (see, for example, Patent Document 2).

[0003] JP 2022-189524 A

[0004] To polish a workpiece using a double-side polishing machine, the load applied to the workpiece is first gradually increased. Once a certain condition is met, such as when the load reaches a predetermined load, the main polishing is performed, in which the workpiece is polished for a predetermined time while maintaining the load applied to the workpiece at the target load. Then, once a predetermined time has elapsed since the start of the main polishing, the main polishing is stopped. The load applied to the workpiece is then gradually reduced, and polishing of the workpiece is completed. Here, even when the load applied to the workpiece is gradually reduced after the main polishing, the shape of the workpiece still changes. Furthermore, the amount of shape change is not constant because it varies depending on the polishing conditions and the state of the secondary materials. Therefore, the operator of the polishing machine estimates the amount of shape change of the workpiece after the main polishing and manually adjusts the timing to stop the main polishing. However, because the timing to stop the main polishing is adjusted based on the operator's intuition, there is a problem of inconsistency in the finished quality of the workpiece.

[0005] On the other hand, the conventional double-sided polishing machines described in Patent Documents 1 and 2 only set polishing conditions during main polishing. That is, in conventional double-sided polishing machines, the distance between the plates is controlled during main polishing, and main polishing of the workpiece is stopped when the shape index of the entire workpiece reaches a set value. However, this does not take into consideration that the shape of the workpiece changes even when the load applied to the workpiece is gradually reduced after main polishing. As a result, the final workpiece shape at the end of polishing may not match the target shape, deviating from the target shape and making it impossible to obtain the desired workpiece shape.

[0006] The present invention has been made with an eye on the above-mentioned problems, and aims to provide a double-sided polishing apparatus and a double-sided polishing method that can prevent the workpiece shape at the end of final polishing from deviating from the target shape, thereby making it possible to obtain the desired workpiece shape.

[0007] In order to achieve the above object, the double-sided polishing apparatus of the present invention is a polishing machine that holds a disk-shaped workpiece between a lower surface plate and an upper surface plate disposed opposite the lower surface plate, and polishes the front and back surfaces of the workpiece by relatively moving the lower and upper surface plates and the workpiece while a load is applied to the workpiece; a thickness gauge that measures the thickness of the workpiece while the workpiece is being polished by the polishing machine; and a control unit that controls the polishing machine based on the measurement results of the thickness gauge, wherein the workpiece has an outer circumferential region of a predetermined range extending radially inward from its outer circumferential end and an in-plane region of a range from the outer circumferential region to the center of the workpiece, and the control unit is configured to control the polishing machine based on the measurement results of the thickness gauge. a main-polishing condition correction unit that calculates a correction target value for the in-plane flatness based on the polishing conditions for the main polishing set at the time of correction and an estimated value of the in-plane flatness obtained by inputting the outer circumferential flatness at the time of correction into the first learning unit, and corrects the polishing conditions for the main polishing set at the time of correction using the polishing conditions for the main polishing obtained by inputting the correction target value for the in-plane flatness and the target value for the outer circumferential flatness into the first learning unit; and a main-polishing stop condition setting unit that inputs the polishing conditions for the endpoint polishing, the degree of change in the in-plane flatness at the time of calculation, and the degree of change in the outer circumferential flatness at the time of calculation into the second learning unit to obtain the amount of change in the outer circumferential flatness during the endpoint polishing, and sets stop conditions for the main polishing based on the amount of change in the outer circumferential flatness and the target value for the outer circumferential flatness.

[0008] In order to achieve the above object, the double-side polishing method of the present invention includes a step of sandwiching a disk-shaped workpiece between a lower surface plate and an upper surface plate disposed opposite the lower surface plate, and polishing the front and back surfaces of the workpiece by relatively moving the lower surface plate, the upper surface plate, and the workpiece while applying a load to the workpiece, the step comprising polishing conditions for a main polishing in which the workpiece is polished while maintaining the load at a target load; a peripheral flatness, which is the flatness of a peripheral region set in a predetermined range from the outer peripheral edge of the workpiece toward the inside in the radial direction; a first learning device that has learned the correlation between the polishing conditions of the end point polishing in which the workpiece is polished while gradually reducing the load and the in-plane flatness, and a second learning device that has learned the correlation between the polishing conditions of the end point polishing in which the workpiece is polished while gradually reducing the load, the degree of change in the in-plane flatness at the time when the main polishing is stopped, the degree of change in the outer periphery flatness at the time when the main polishing is stopped, and the amount of change in the outer periphery flatness during the end point polishing, and a step of setting a target value of the outer periphery flatness and a target value of the in-plane flatness; and then performing initial polishing in which the workpiece is polished while gradually increasing the load. After the initial polishing is completed, starting the main polishing is performed. During the main polishing, a correction target value for the in-plane flatness is calculated based on the polishing conditions for the main polishing set at the time of correction and the estimated value of the in-plane flatness obtained by inputting the outer periphery flatness at the time of correction into the first learning device, and the correction target value for the in-plane flatness and the target value for the outer periphery flatness are input into the first learning device. and a step of correcting the polishing conditions for the main polishing that are set at the time of correction; a step of inputting the polishing conditions for the endpoint polishing, the degree of change in the in-plane flatness at the time of calculation, and the degree of change in the peripheral flatness at the time of calculation into the second learning device during execution of the main polishing to obtain an amount of change in the peripheral flatness during the endpoint polishing, and setting a stop condition for the main polishing based on the amount of change in the peripheral flatness and a target value for the peripheral flatness; and a step of executing the endpoint polishing after the stop condition for the main polishing is established.

[0009] The double-side polishing apparatus and double-side polishing method of the present invention can prevent the workpiece shape at the end of final polishing from deviating from the target shape, making it possible to obtain the desired workpiece shape.

[0010] FIG. 1 is an explanatory diagram showing an outline of the overall configuration of the double-sided polishing apparatus of Example 1. FIG. 2 is an explanatory diagram showing the positional relationship between the sun gear, the internal gear, and the carrier plate of Example 1. FIG. 3 is an explanatory diagram showing a passing trajectory when a measurement hole passes over a workpiece in the double-sided polishing apparatus of Example 1. FIG. 4 is an explanatory diagram showing a group of thickness data obtained by measuring the thickness of a workpiece polished by the double-sided polishing apparatus of Example 1, and an outer peripheral region and an in-plane region set on the workpiece. FIG. 5 is an explanatory diagram showing an approximated straight line set for a group of thickness data of a workpiece whose outer peripheral region has a roll-up shape. FIG. 6 is an explanatory diagram showing an approximated straight line set for a group of thickness data of a workpiece whose outer peripheral region has a flat shape. FIG. 7 is an explanatory diagram showing the innermost region and the outermost region when the outer peripheral region is divided into a plurality of regions. FIG. 8 is an explanatory diagram showing the calculated values ​​of in-plane flatness and outer peripheral flatness over time and the degree of change. FIG. 9 is an explanatory diagram showing a data set to be learned by a first learning unit of Example 1. FIG. 10 is an explanatory diagram showing a data set to be learned by a second learning unit of Example 1. FIG. 11 is a flowchart showing the flow of polishing process control executed by a control unit of Example 1. FIG. 12 is a table showing main polishing conditions for an exemplary workpiece, and the outer peripheral flatness and in-plane flatness at that time. 1 is a table showing a data set to be learned by a learning machine of a comparative example. 2 is a polishing result showing the final workpiece shape at the end of polishing when the workpiece is polished with the target value of outer peripheral flatness set to zero (nm) in the double-sided polishing apparatus of Example 1. 3 is a polishing result showing the final workpiece shape at the end of polishing in the double-sided polishing apparatus of Example 1.

[0011] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A double-sided polishing apparatus and a double-sided polishing method according to the present invention will be described below with reference to a first embodiment shown in the drawings.

[0012] The double-sided polishing apparatus 1 of the first embodiment is a double-sided polishing apparatus for polishing both the front and back surfaces of a thin, circular workpiece W, such as a semiconductor wafer, a quartz wafer, a sapphire wafer, a glass wafer, or a ceramic wafer. As shown in FIG. 1 , the double-sided polishing apparatus 1 includes a polishing machine 10, a thickness measuring device 20, and a control unit 30.

[0013] The polishing machine 10 sandwiches a workpiece W between a lower surface plate 11 and an upper surface plate 12 disposed opposite the lower surface plate 11, and moves the lower surface plate 11, the upper surface plate 12, and the workpiece W relative to each other while a load is applied to the workpiece W, thereby simultaneously polishing the front and back surfaces of the workpiece W. The polishing machine 10 includes the donut-shaped lower surface plate 11 and the upper surface plate 12, which are concentrically arranged about an axis L1, a sun gear 13 rotatably disposed in the center of the lower surface plate 11, an internal gear 14 disposed on the outer periphery of the lower surface plate 11, and a carrier plate 15 disposed between the lower surface plate 11 and the upper surface plate 12 and having a workpiece holding hole 15a (see FIG. 2 ). A polishing pad 11a is affixed to the upper surface of the lower surface plate 11, and a polishing pad 12a is affixed to the lower surface of the upper surface plate 12. Furthermore, a plurality of supply holes (not shown) for supplying polishing slurry (hereinafter referred to as "slurry") are provided in the upper surface plate 12. Note that it may be possible to select whether or not to supply slurry for each supply hole or for each group of supply holes.

[0014] 2, the carrier plate 15 meshes with the sun gear 13 and the internal gear 14. The rotation of the sun gear 13 and the internal gear 14 causes the carrier plate 15 to rotate (revolve) around the axis L1 while rotating on its own axis.

[0015] The workpiece W is placed in a workpiece holding hole 15a of the carrier plate 15. Then, as the carrier plate 15 rotates and revolves with the workpiece W sandwiched between a polishing pad 11a affixed to the rotating lower surface plate 11 and a polishing pad 12a affixed to the rotating upper surface plate 12, the lower surface plate 11, the upper surface plate 12 and the workpiece W move relative to each other, and the workpiece W is polished by the polishing pads 11a and 12a. In other words, the surfaces of the polishing pads 11a and 12a become the polishing surfaces that polish the workpiece W.

[0016] The upper surface plate 12 is fixed to a rod 16 via support studs 16a and mounting members 16b attached to the upper surface. The rod 16 is extended and retracted in the vertical direction by a fifth driving device M5, and the upper surface plate 12 moves up and down as the rod 16 extends and retracts. A predetermined load is applied to the workpiece W from the upper surface plate 12 according to the extension length of the rod 16. In other words, the load applied to the workpiece W is adjusted by controlling the fifth driving device M5.

[0017] A first drive shaft 17a is disposed in the center of the polishing machine 10, standing along the axis L1. The first drive shaft 17a is rotated by the first drive unit M1. A driver 18 is fixed to the upper end of the first drive shaft 17a. As a result, the driver 18 rotates integrally with the first drive shaft 17a. Meanwhile, a groove (not shown) is formed on the outer circumferential surface of the driver 18, into which a hook 12b provided on the upper surface plate 12 engages. Then, the rod 16 extends, causing the upper surface plate 12 to move downward, and the hook 12b engages with the groove of the driver 18, causing the driver 18 and the upper surface plate 12 to rotate integrally. In other words, since the upper surface plate 12 rotates integrally with the first drive shaft 17a, which is rotated by the first drive unit M1, the rotation speed of the upper surface plate 12 can be adjusted by controlling the first drive unit M1.

[0018] A second drive shaft 17b is fixed and passes through a hole 13a in the center of the sun gear 13. The second drive shaft 17b is a hollow tube with both ends open, and the first drive shaft 17a passes through it so as to be able to rotate freely. The second drive shaft 17b is rotated by the second drive device M2. That is, the sun gear 13 rotates integrally with the second drive shaft 17b, which is rotated by the second drive device M2, and the rotation speed of the sun gear 13 is adjusted by controlling the second drive device M2.

[0019] A third drive shaft 17c is formed below the center of the lower surface plate 11. The third drive shaft 17c is a hollow tube with both ends open, through which the first drive shaft 17a and the second drive shaft 17b rotatably pass. The third drive shaft 17c is rotated by a third drive unit M3. That is, the lower surface plate 11 rotates integrally with the third drive shaft 17c, which is rotated by the third drive unit M3, and the rotation speed of the lower surface plate 11 is adjusted by controlling the third drive unit M3.

[0020] The internal gear 14 is also formed with a fourth drive shaft 17d. The fourth drive shaft 17d is a hollow tube open at both ends, through which the first drive shaft 17a, the second drive shaft 17b, and the third drive shaft 17c rotatably pass. The fourth drive shaft 17d is rotated by the fourth drive device M4. That is, the internal gear 14 rotates integrally with the fourth drive shaft 17d, which is rotated by the fourth drive device M4, and the rotation speed of the internal gear 14 is adjusted by controlling the fourth drive device M4.

[0021] The double-side polishing machine 1 of Example 1 polishes the workpiece W by controlling the rotation speeds of the lower platen 11, upper platen 12, sun gear 13, and internal gear 14 while extending and retracting the rod 16 to control the load applied to the workpiece W based on predetermined polishing conditions. Here, since it takes time to extend and retract the rod 16 and adjust (increase or decrease) the rotation speed of the lower platen 11, etc., the polishing process of the workpiece W is divided into initial polishing, main polishing, and end-point polishing.

[0022] The "initial polishing" is a process in which the workpiece W is polished while gradually increasing the load applied to the workpiece W after polishing has started. During the initial polishing, the rotation speeds of the lower surface plate 11, upper surface plate 12, sun gear 13, and internal gear 14 are also gradually increased. The polishing conditions for the initial polishing (initial polishing conditions), including the polishing time and termination conditions for the initial polishing, are determined by the prerequisite polishing conditions. The "prerequisite polishing conditions" are polishing conditions that are determined in advance depending on the final processing target (polishing target) of the workpiece W, the type of workpiece W, etc.

[0023] Furthermore, "main polishing" refers to a process in which, after the initial polishing is completed, the workpiece W is polished under predetermined polishing conditions while maintaining the load applied to the workpiece W at the target load. During the main polishing, the rotational speeds of the lower platen 11, upper platen 12, sun gear 13, and internal gear 14 are also maintained at their respective target rotational speeds. The polishing conditions for the main polishing (main polishing conditions), including the target load and the target rotational speeds for the main polishing, are determined by the prerequisite polishing conditions at the start of the main polishing and are repeatedly corrected during the main polishing. Furthermore, a stop condition for the main polishing (main polishing stop condition), which determines the timing for stopping the main polishing, is repeatedly determined during the main polishing, and the main polishing is stopped when the stop condition is met. The main polishing may be composed of multiple sections (multiple steps), and the main polishing conditions may include appropriate adjustments of the slurry temperature, flow rate, supply destination, etc., based on the results of monitoring the trend in the shape of the workpiece W.

[0024] "Endpoint polishing" is a process in which the workpiece W is polished while gradually reducing the load applied to the workpiece W after the main polishing has been stopped. In endpoint polishing, the rotation speeds of the lower surface plate 11, upper surface plate 12, sun gear 13, and internal gear 14 are also gradually reduced. The polishing conditions for endpoint polishing (endpoint polishing conditions), including the polishing time in endpoint polishing, are determined by the prerequisite polishing conditions. The endpoint polishing conditions may also include appropriate adjustment of the temperature, flow rate, supply destination, etc. of the slurry according to the results of monitoring the tendency of changes in the shape of the workpiece W.

[0025] A measurement hole 19 is formed in the upper surface plate 12 at a position a predetermined distance away from the center in the radial direction. The measurement hole 19 penetrates the upper surface plate 12 and the polishing pad 12a, and is fitted with a window member 19a that transmits laser light, which is measurement light.

[0026] The thickness measuring device 20 irradiates measurement light toward the workpiece W and receives the measurement light reflected by the workpiece W to measure the thickness of the workpiece W being polished (the distance between the front and back surfaces of the workpiece W). The thickness measuring device 20 of the first embodiment also digitizes the shape of the workpiece W from a group of measured thickness data of the workpiece W. The thickness measuring device 20 has a measurement unit 21, a thickness measuring section 22, and a shape calculation section 23.

[0027] The measurement unit 21 is attached to the upper surface plate 12 and rotates integrally with the upper surface plate 12. The measurement unit 21 also has a laser light source (not shown) that irradiates laser light, which is measurement light, toward the workpiece W through a window member 19a attached to the measurement hole 19 of the upper surface plate 12, and a light receiving unit (not shown) that receives light reflected from the front and back surfaces of the workpiece W as a light receiving signal. The light receiving signal received by the light receiving unit is transmitted to the thickness measurement unit 22 by a transmission unit 21a.

[0028] The thickness measuring unit 22 measures the thickness of the workpiece W by, for example, optical reflection interference. The thickness measuring unit 22 has a receiving unit 22a that receives a light receiving signal transmitted from the measuring unit 21, and acquires thickness data of the workpiece W based on the light receiving signal received by the receiving unit 22a.

[0029] As shown in FIG. 3A , the upper surface plate 12 rotates, and while the measurement hole 19 passes over the surface of the workpiece W, the laser light from the measurement unit 21 is continuously irradiated onto the surface of the workpiece W. Therefore, the thickness measurement unit 22 continuously measures the thickness of each in-plane position of the workpiece W along the measurement hole 19's path Na-Nc. The thickness measurement unit 22 outputs a group of thickness data consisting of a number of consecutive thickness data points each time the measurement hole 19 passes over each path Na-Nc (during the period during which the measurement hole 19 passes from one end W1a-W3a of the workpiece W to the other end W1b-W3b). As a result, the thickness measurement unit 22 outputs a group of thickness data consisting of multiple consecutive data points obtained by measuring the thickness of each in-plane position of the workpiece W each time the measurement hole 19 passes over the surface of the workpiece W (see FIG. 3B ). The group of thickness data output from the thickness measurement unit 22 is input to the shape calculation unit 23.

[0030] Furthermore, when inputting the thickness data group, the correlation between the measurement data (e.g., GBIR, ESFQD, etc.) measured by an external measuring device (external measuring device) capable of measuring the thickness (shape) of the workpiece W and the thickness data group may be separately calculated, and the correction value obtained from the correlation may be added and input. Note that "GBIR (Global Backside Ideal Range)" is a value indicating the difference between the maximum and minimum values ​​of the thickness distribution. Also, "ESFQD (Edge Site Front Least Square)" is a value that outputs the larger absolute value of the maximum and minimum values ​​of the distance from the reference surface at the outer periphery of the wafer.

[0031] Furthermore, the thickness data group and the numerical values ​​of the in-plane flatness and peripheral flatness described below are not necessarily highly correlated with the measurement data of an external measuring instrument. For this reason, the thickness data group may be corrected as needed or at a desired timing for the thickness data group, the in-plane flatness, and the peripheral flatness based on a value obtained from a difference database that stores the differences between the thickness data group and the measurement data of an external measuring instrument, or a value obtained by inputting the obtained value into a conversion formula, or a value obtained by inputting measurement data measured by an external measuring instrument into a conversion formula.

[0032] The difference database may be updated as needed, which allows the correlation between the thickness data measured by the thickness measurement unit 22 and the measurement data of the external measuring device to be understood, and makes it possible to input the target value of the work shape to the control unit 30 in accordance with the measurement standard of the external measuring device, for example.

[0033] The shape calculation unit 23 converts the shape of the workpiece W (hereinafter referred to as "workpiece shape") into numerical information based on the group of thickness data of the workpiece W measured by the thickness measurement unit 22. In Example 1, the workpiece shape is represented by in-plane flatness and peripheral flatness. The information on in-plane flatness and peripheral flatness output from the shape calculation unit 23 is input to the control unit 30.

[0034] Here, "in-plane flatness" refers to the flatness of an in-plane region G set on the workpiece W, and is the slope of an approximate straight line (the approximate straight line indicated by A in FIGS. 4A and 4B) set by, for example, the least squares method for the group of thickness data in the in-plane region G. Note that the approximate straight line indicated by A' in FIGS. 4A and 4B is also an approximate straight line set for the group of thickness data in the in-plane region G. However, when the shape calculation unit 23 converts the workpiece shape into numerical information, the region to the right of the workpiece center Wo (in-plane region G', outer peripheral region E') is ignored when viewed in a cross section of the workpiece W broken along a line passing through the workpiece center Wo.

[0035] The approximate line is expressed by the following formula (1), and the slope of the approximate line is "a" in the formula (1). Approximate line: Y = aX + b (1), where a is the slope and b is the intercept.

[0036] When the in-plane flatness is a positive value, the workpiece shape is a convex shape with the center of the in-plane region G protruding. When the in-plane flatness is a negative value, the workpiece shape is a concave shape with the center of the in-plane region G recessed.

[0037] Furthermore, "peripheral flatness" refers to the flatness of the outer peripheral region E set on the workpiece W. The outer peripheral flatness is determined by performing a continuity process on the gradient of an approximate straight line set on a group of thickness data for the outer peripheral region E based on the in-plane flatness. Here, "a continuity process on the gradient of an approximate straight line set on a group of thickness data for the outer peripheral region E based on the in-plane flatness" refers to a process of adding or subtracting the gradient of an approximate straight line set on a group of thickness data in the outer peripheral region E to the in-plane flatness. In other words, the "continuity process" includes a process of adding the in-plane flatness and the gradient of the approximate straight line set on the group of thickness data for the outer peripheral region E (a process of calculating a sum value), and a process of calculating the difference between the in-plane flatness and the gradient of the approximate straight line set on the group of thickness data for the outer peripheral region E (a process of calculating a difference value).

[0038] The peripheral flatness in Example 1 is calculated by adding the in-plane flatness to the slope of the approximate straight line (the line indicated by B in Figures 4A and 4B) set for the thickness data group in the peripheral region E, and is the sum of the in-plane flatness and the slope of the approximate straight line in the peripheral region E.

[0039] When the outer circumferential flatness is a positive value, the workpiece shape is a rolled-up shape in which the outer circumferential region E bounces up relative to the in-plane region G. When the outer circumferential flatness is a negative value, the workpiece shape is a rolled-off shape in which the outer circumferential region E sags relative to the in-plane region G.

[0040] As shown in Figures 3A and 3B, the "outer peripheral region E" is a predetermined region extending radially inward from the outer peripheral edge We of the workpiece W. As shown in Figures 3A and 3B, the "in-plane region G" is a region extending from the radially inner edge of the outer peripheral region E to the workpiece center Wo. That is, as shown in Figure 3A, the outer peripheral region E is an annular region between the outer peripheral edge We of the workpiece W and a circle indicated by a dashed line set within the workpiece W, and is set at the periphery of the workpiece W. Furthermore, the in-plane region G is a circular region surrounded by a circle indicated by a dashed line set within the workpiece W, and is set at the center of the workpiece W. As shown in Figure 3B, when viewed in a cross section of the workpiece W cut along a straight line passing through the workpiece center Wo, the in-plane region G and the outer peripheral region E are set symmetrically with respect to the workpiece center Wo. Furthermore, the set range of the outer peripheral region E can be determined arbitrarily.

[0041] 4A shows an approximate straight line created based on the thickness data group of the first example workpiece W. In the case shown in FIG. 4A, the slope value of the approximate straight line A in the in-plane region G is +50 (nm: nanometers), and the slope value of the approximate straight line B in the outer periphery region E is zero (nm). Therefore, the in-plane flatness of the first example workpiece W is +50 (nm), and the outer periphery flatness is +50 (= +50 + zero) (nm).

[0042] 4B also shows an approximate line created based on the thickness data group of the second example workpiece W. In the case shown in FIG. 4B, the slope value of the approximate line A in the in-plane region G is +30 (nm), and the slope value of the approximate line B in the outer periphery region E is −30 (nm). Therefore, the in-plane flatness of the second example workpiece W is +30 (nm), and the outer periphery flatness is zero (= +30 + (−30)) (nm).

[0043] 5, the outer peripheral region E may be divided into multiple sections (three in the example shown in FIG. 5) along the radial direction of the workpiece W. In this case, the outer peripheral flatness is calculated for each of the sections Ei, Eo1, and Eo2 of the outer peripheral region E.

[0044] 5, a predetermined section of the outer peripheral region E adjacent to the in-plane region G is defined as the innermost section Ei. A section adjacent to the innermost section Ei and radially outward of the innermost section Ei is defined as the first outer section Eo1. A section adjacent to the first outer section Eo1 and radially outward of the first outer section Eo1 is defined as the second outer section Eo2.

[0045] The flatness of the innermost section Ei is calculated by performing a continuation process on the slope of the approximation line set for the thickness data set of the innermost section Ei based on the in-plane flatness. The flatness of the innermost section Ei is calculated by adding the slope of the approximation line set for the thickness data set of the innermost section Ei to the in-plane flatness. In other words, the flatness of the innermost section Ei in Example 1 is calculated as the sum of the in-plane flatness and the slope of the approximation line for the innermost section Ei.

[0046] The flatness of the first outer section Eo1 is calculated by performing a continuation process on the slope of an approximation line set for the thickness data set of the first outer section Eo1, using the flatness of the section (innermost section Ei) adjacent to the first outer section Eo1 radially inward as a reference. The flatness of the first outer section Eo1 is calculated by adding the slope of the approximation line set for the thickness data set of the first outer section Eo1 to the flatness of the innermost section Ei. In other words, the flatness of the first outer section Eo1 in Example 1 is calculated as the sum of the flatness of the innermost section Ei and the slope of the approximation line of the first outer section Eo1.

[0047] The flatness of the second outer section Eo2 is calculated by performing a continuation process on the slope of an approximation line set for the thickness data set of the second outer section Eo2, using the flatness of the section (first outer section Eo1) adjacent to the second outer section Eo2 radially inward as a reference. The flatness of the second outer section Eo2 is calculated by adding the slope of the approximation line set for the thickness data set of the second outer section Eo2 to the flatness of the first outer section Eo1. In other words, the flatness of the second outer section Eo2 in Example 1 is calculated as the sum of the flatness of the first outer section Eo1 and the slope of the approximation line of the second outer section Eo2.

[0048] Furthermore, the shape calculation unit 23 continuously calculates numerical information indicating the workpiece shape (in-plane flatness and peripheral flatness) at any interval (for example, every 10 to 15 seconds) during polishing of the workpiece W. This allows the control unit 30 to obtain the degree of change in the in-plane flatness and the degree of change in the peripheral flatness during polishing.

[0049] Here, the degree of change in in-plane flatness is represented by the gradient of an approximate straight line set, for example, by the least squares method, for the group of in-plane flatness data indicated by ○ in Fig. 6, i.e., the rate of change in in-plane flatness over a predetermined time interval. Furthermore, the degree of change in peripheral flatness is represented by the gradient of an approximate straight line set, for example, by the least squares method, for the group of peripheral flatness data indicated by ● in Fig. 6, i.e., the rate of change in peripheral flatness over a predetermined time interval. The control unit 30 then determines the trend of change in the workpiece shape from the degree of change in in-plane flatness and the degree of change in peripheral flatness.

[0050] That is, when the degree of change in the in-plane flatness is a positive value, the control unit 30 determines that the change trend of the in-plane area G is a tendency to change in a convex direction. On the other hand, when the degree of change in the in-plane flatness is a negative value, the control unit 30 determines that the change trend of the in-plane area G is a tendency to change in a concave direction. For example, in the example shown in FIG. 6 , the degree of change in the in-plane flatness at time t0 (the slope of the approximation line set to the group of in-plane flatness data for a very short time period including time t0) is a negative value, as shown by the dashed dotted line. Therefore, the change trend of the in-plane area G after time t0 is determined to be a tendency to change in a concave direction.

[0051] Furthermore, if the degree of change in the peripheral flatness is a positive value, the control unit 30 determines that the change trend in the peripheral region E is a tendency toward a roll-up (jump-up) direction. If the degree of change in the peripheral flatness is a negative value, the control unit 30 determines that the change trend in the peripheral region E is a tendency toward a roll-off (sagging) direction. For example, in the example shown in FIG. 6 , the degree of change in the peripheral flatness at time t0 (the slope of the approximate line set to the group of peripheral flatness data over a very short time period including time t0) is a positive value, as indicated by the two-dot chain line. Therefore, the change trend in the peripheral region E after time t0 is determined to be a tendency toward a roll-up direction.

[0052] The control unit 30 is composed of a CPU (Central Processing Unit) and the like, and is provided with a first learning unit 31, a second learning unit 32, a main polishing condition correction unit 33, a main polishing stop condition setting unit 34, a control calculation unit 35, and a memory 36. In addition, the control unit 30 is connected to an input device 41 that can be operated by an operator of the double-sided polishing machine 1 and a display 42 that can be seen by the operator.

[0053] The control unit 30 outputs control commands from the control calculation unit 35 to the first drive unit M1 to the fifth drive unit M5, etc., based on the measurement results of the workpiece W obtained by the thickness measuring device 20 (which may include a correction value obtained by calculating a correlation with measurement data obtained by measuring the shape of the workpiece W using a separate external measuring device), the processing target of the workpiece W, the prerequisite polishing conditions, conditions such as auxiliary materials, information about the device status of the polishing machine 10, the program stored in the memory 36, the main polishing conditions reset by the main polishing condition correction unit 33, the main polishing stop condition set by the main polishing stop condition setting unit 34, etc., to control the operation of the polishing machine 10. Note that information about the processing target of the workpiece W, the prerequisite polishing conditions, conditions such as auxiliary materials, and the device status of the polishing machine 10 may be input by an operator via the input device 41 or may be pre-stored in the memory 36. Furthermore, the control unit 30 appropriately displays necessary information on the display 42 while controlling the operation of the polishing machine 10.

[0054] The first learning device 31 is a learning device that has learned the correlation between the main polishing conditions at any timing during the main polishing, the outer periphery flatness acquired at that timing, and the in-plane flatness acquired at that timing. That is, the control unit 30 causes the first learning device 31 of Example 1 to learn the main polishing conditions at any timing during the main polishing, the outer periphery flatness included in the main polishing conditions, disturbance factors, and the in-plane flatness acquired at the same timing as a data set, as shown in Fig. 7, for example. Here, the main polishing conditions include, for example, the rotation speed of the lower platen 11, the rotation speed of the upper platen 12, the rotation speed of the sun gear 13, the rotation speed of the internal gear 14, the revolution speed of the carrier plate 15, the rotation speed of the carrier plate 15, the load applied to the workpiece W, the flow rate of the slurry, the type of slurry, etc. The disturbance factors include, for example, the condition of the secondary materials, such as the period of use (carrier life) of the carrier plate 15, the period of use (pad life) of the polishing pads 11a and 12a, and dressing conditions, and the equipment condition includes, for example, the load factor of each drive device, the temperature of the area (drive chamber) where each drive device is located, the temperature of the polishing pads 11a and 12a, the temperature of the slurry, and fluctuations in the load applied to the workpiece W. The main polishing conditions and the disturbance factors may be input by the operator via the input device 41, or may be detected by a sensor or the like.

[0055] The second learning device 32 is a learning device that learns the correlation between the endpoint polishing conditions, the degree of change in in-plane flatness when the main polishing is stopped, the degree of change in peripheral flatness when the main polishing is stopped, and the amount of change in peripheral flatness during endpoint polishing. That is, the control unit 30 causes the second learning device 32 of Example 1 to learn, as a data set, the predetermined endpoint polishing conditions, the degree of change in in-plane flatness when the main polishing is stopped, the degree of change in peripheral flatness when the main polishing is stopped, and the amount of change in peripheral flatness during endpoint polishing performed under the endpoint polishing conditions based on the degree of change, as shown in FIG. 8 . Here, the endpoint polishing conditions include, for example, the rotation speed of the lower platen 11, the rotation speed of the upper platen 12, the rotation speed of the sun gear 13, the rotation speed of the internal gear 14, the revolution speed of the carrier plate 15, the rotation speed of the carrier plate 15, the load applied to the workpiece W, the flow rate of the slurry, the type of slurry, and the endpoint polishing time (deceleration time). The end point polishing conditions may be input by an operator via the input device 41 or may be detected by a sensor or the like.

[0056] Furthermore, the "degree of change in in-plane flatness when main polishing is stopped" is the gradient of an approximate straight line set to a group of data on in-plane flatness over a very short time interval that includes the time point at which main polishing is stopped (the timing at which main polishing is stopped). The "degree of change in peripheral flatness when main polishing is stopped" is the gradient of an approximate straight line set to a group of data on peripheral flatness over a very short time interval that includes the time point at which main polishing is stopped. The "amount of change in peripheral flatness during endpoint polishing" is the difference between the peripheral flatness at the time point at which main polishing is stopped and the peripheral flatness at the end of endpoint polishing.

[0057] During the main polishing, the main polishing condition correction unit 33 determines a correction target value for in-wafer flatness to bring the calculated value of in-wafer flatness closer to the target value, based on an estimated value of in-wafer flatness obtained by inputting the main polishing conditions set at the time of correction and the calculated value of the peripheral flatness at the time of correction into the first learning unit 31. The main polishing condition correction unit 33 then corrects the main polishing conditions set at the time of correction using the main polishing conditions obtained by inputting the correction target value of in-wafer flatness and the final target value of the peripheral flatness at the end of polishing into the first learning unit 31. In other words, the main polishing condition correction unit 33 uses the first learning unit 31 to acquire main polishing conditions that satisfy the target value and correction target value of the peripheral flatness, and replaces the main polishing conditions set at the time of correction with the acquired main polishing conditions and resets them as new main polishing conditions. In addition, the newly reset main polishing conditions may change at least one of, for example, the rotation speed of the lower platen 11, the rotation speed of the upper platen 12, the rotation speed of the sun gear 13, the rotation speed of the internal gear 14, the revolution speed of the carrier plate 15, the rotation speed of the carrier plate 15, the load applied to the workpiece W, the flow rate of the slurry, the type, the supply destination, etc.

[0058] Here, the "correction target value of the in-plane flatness for bringing the calculated value of the in-plane flatness closer to the target value" is obtained by the following procedure. That is, the main polishing condition correcting unit 33 first inputs the main polishing conditions at the time of correction and the calculated value of the outer periphery flatness at the time of correction to the first learning device 31 to obtain an estimated value of the in-plane flatness. Next, the main polishing condition correcting unit 33 calculates the difference value between the calculated value of the in-plane flatness at the time of correction and the estimated value of the in-plane flatness. Then, the main polishing condition correcting unit 33 sets the calculated value by subtracting the difference value of the in-plane flatness from the final target value of the in-plane flatness at the end of polishing as the "correction target value."

[0059] During the main polishing, the main polishing stop condition setting unit 34 inputs the end point polishing conditions, the degree of change in in-wafer flatness at the time of calculation, and the degree of change in in-wafer flatness at the time of calculation to the second learning device 32, and acquires the amount of change in peripheral flatness during the end point polishing. Then, based on the acquired amount of change in peripheral flatness and the target value of the final peripheral flatness at the end of polishing, it sets the stop condition of the main polishing (main polishing stop condition). Here, the main polishing stop condition is defined by the peripheral flatness at the time of stopping the main polishing.

[0060] That is, the main polishing stop condition setting unit 34 inputs the endpoint polishing condition, the degree of change in in-plane flatness at the time of calculation (the slope of the approximate straight line set to the data group of in-plane flatness for a very short time interval including the calculation time point), and the degree of change in periphery flatness at the time of calculation (the slope of the approximate straight line set to the data group of periphery flatness for a very short time interval including the calculation time point) to the second learning unit 32, and acquires the amount of change in periphery flatness during endpoint polishing based on the degree of change. Next, the main polishing stop condition setting unit 34 calculates the target value of periphery flatness at the time of stopping main polishing by back-calculating from the acquired amount of change in periphery flatness, the final target value of periphery flatness at the time of polishing completion, and the endpoint polishing time. Then, the calculated target value of periphery flatness is specified as the main polishing stop condition.

[0061] During the execution of the initial polishing, the control unit 30 outputs control commands from the control calculation unit 35 to the first to fifth driving units M1 to M5 according to the initial polishing conditions set in advance as the prerequisite polishing conditions.

[0062] Furthermore, at the start of the main polishing, the control unit 30 outputs control commands from the control calculation unit 35 to the first to fifth driving units M1 to M5 according to the main polishing conditions preset in the prerequisite polishing conditions.

[0063] Then, during execution of the main polishing, the control unit 30 corrects the main polishing conditions using the first learner 31 and the main polishing condition correction unit 33. When the main polishing conditions are corrected, the control unit 30 outputs control commands according to the corrected main polishing conditions (reset new main polishing conditions) from the control calculation unit 35 to the first to fifth drive units M1 to M5.

[0064] Furthermore, during the execution of the main polishing, the control unit 30 sets a main polishing stop condition using the second learner 32 and the main polishing stop condition setting unit 34. When the main polishing stop condition is established, the control unit 30 outputs a control command from the control calculation unit 35 to the first to fifth drive units M1 to M5 to stop the main polishing and execute endpoint polishing.

[0065] During the execution of the endpoint polishing, the control unit 30 outputs control commands from the control calculation unit 35 to the first to fifth driving units M1 to M5 according to the endpoint polishing conditions set in advance as the prerequisite polishing conditions.

[0066] 9 is a flowchart showing the flow of the polishing process executed by the control unit 30 of the double-sided polishing machine 1 of Example 1. The flow of the polishing process of Example 1 will be described below with reference to FIG. The polishing process is executed with the carrier plate 15 and workpiece W set in the polishing machine 10. During the polishing process (from the start of initial polishing to the end of endpoint polishing), the thickness measuring device 20 continuously inputs shape information of the workpiece W (information on in-plane flatness and peripheral flatness) to the control unit 30.

[0067] In step S1, the control unit 30 sets a final target value for in-plane flatness at the end of end-point polishing and a final target value for peripheral flatness at the end of end-point polishing, which will be the final processing targets for the workpiece W, and then proceeds to step S2. Here, the target values ​​for in-plane flatness and peripheral flatness are input by the operator via the input device 41. Note that the processing targets for the workpiece W may be set in advance for each type of workpiece W, for example, and stored in the memory 36. In this case, the control unit 30 reads and sets the processing targets for the workpiece W from the memory 36 based on the type of workpiece W input by the operator.

[0068] In step S2, following the setting of the workpiece processing target in step S1, the control unit 30 sets prerequisite polishing conditions, and then proceeds to step S3. Here, the prerequisite polishing conditions are various conditions that are prerequisites for polishing the workpiece W, which are set in advance for each processing target of the workpiece W and each type of workpiece W. The prerequisite polishing conditions include, for example, initial polishing conditions including the polishing time and polishing end conditions for the initial polishing, main polishing conditions at the start of main polishing, end-point polishing conditions including the polishing time and polishing end conditions for end-point polishing, information about the slurry such as the slurry flow rate, slurry type, and slurry supply destination, and information about the status of consumable secondary materials such as carrier life. The prerequisite polishing conditions are input by the operator via the input device 41 or read out from the memory 36.

[0069] In step S3, following the setting of the prerequisite polishing conditions in step S2 or the determination in step S4 that the initial polishing is not yet completed, the control unit 30 outputs control commands according to the initial polishing conditions defined by the prerequisite polishing conditions from the control calculation unit 35 to the first drive unit M1 to the fifth drive unit M5, executes the initial polishing, and proceeds to step S4. Note that during the execution of the initial polishing, the polisher 10 polishes the workpiece W while gradually increasing the load applied to the workpiece W and gradually increasing the rotation speeds of the lower platen 11 and the upper platen 12, the rotation speed of the sun gear 13, and the rotation speed of the internal gear 14.

[0070] In step S4, following the execution of the initial polishing in step S3, the control unit 30 determines whether the initial polishing has been completed. If the control unit 30 determines YES (initial polishing has been completed), the process proceeds to step S5, and if the control unit 30 determines NO (initial polishing has not been completed), the process returns to step S3. Here, the control unit 30 determines that the initial polishing has been completed when the completion condition for the initial polishing specified in the prerequisite polishing conditions is met, such as when the load applied to the workpiece W reaches a predetermined load or when the rotation speed of the lower surface plate 11 or the like reaches a predetermined rotation speed.

[0071] In step S5, following the determination in step S4 that the initial polishing has ended, the control unit 30 outputs control commands from the control calculation unit 35 to the first drive unit M1 to the fifth drive unit M5 according to the main polishing conditions defined by the prerequisite polishing conditions, starts the execution of the main polishing, and proceeds to steps S6 and S7. During the execution of the main polishing, the polisher 10 polishes the workpiece W while adjusting the load applied to the workpiece W to the target load defined by the main polishing conditions, and adjusting the rotation speeds of the lower platen 11 and the upper platen 12, the rotation speed of the sun gear 13, and the rotation speed of the internal gear 14 to their target rotation speeds.

[0072] In step S6, following the start of main polishing in step S5, the control unit 30 corrects the main polishing conditions using the main polishing condition correcting unit 33. The correction of the main polishing conditions is performed in parallel with the setting of main polishing stop conditions in step S7, which will be described later, and is repeatedly performed at regular intervals (e.g., intervals of about 300 seconds) until it is determined in step S8, which will be described later, that main polishing has been stopped.

[0073] Here, the main polishing condition correcting unit 33 corrects the main polishing conditions in the following procedure: (1) The main polishing conditions at the time of correction and the calculated value of the peripheral flatness at the time of correction are input to the first learning device 31, and an estimated value of the in-plane flatness at the time of correction under the main polishing conditions is obtained. (2) The estimated value of the in-plane flatness obtained in (1) is subtracted from the calculated value of the in-plane flatness at the time of correction to calculate a difference value of the in-plane flatness (the difference between the estimated value of the in-plane flatness and the current value). (3) A value is calculated by subtracting the difference value of the in-plane flatness calculated in (2) from the target value of the in-plane flatness set in step S1, and the result is set as a "corrected target value." (4) The "corrected target value" set in (3) and the target value of the peripheral flatness set in step S1 are input to the first learning device 31, and the main polishing conditions are obtained. (5) The main polishing conditions acquired in (4) are reset as new main polishing conditions in place of the main polishing conditions set at the time of correction, and the main polishing conditions are corrected.

[0074] In step S7, following the start of main polishing in step S5, the control unit 30 sets a main polishing stop condition using the main polishing stop condition setting unit 34. Note that the setting of the main polishing stop condition is executed in parallel with the correction of the main polishing conditions in step S6 as described above, and is repeatedly executed at regular intervals (e.g., intervals of about 1 second) until it is determined in step S8, which will be described later, that main polishing has been stopped.

[0075] The main polishing stop condition setting unit 34 sets the main polishing stop condition in the following procedure. (1) The endpoint polishing conditions determined based on the premise polishing conditions set in step S2, the degree of change in in-wafer flatness at the time of calculation, and the degree of change in peripheral flatness at the time of calculation are input to the second learning unit 32 to obtain the amount of change in peripheral flatness during endpoint polishing (note that the "time of calculation" here refers to the time when the main polishing stop condition is set). (2) The amount of change in peripheral flatness during endpoint polishing obtained in step S1, the target value of peripheral flatness set in step S1, and the polishing time for endpoint polishing are used to calculate backward to obtain the peripheral flatness at the time of stopping main polishing (the target value of peripheral flatness at the time of stopping main polishing) necessary for the peripheral flatness to satisfy the final target value at the end of polishing. (3) The peripheral flatness obtained in step (2) is set as the main polishing stop condition. Note that the calculation of peripheral flatness in step (2) is performed continuously at regular intervals. Therefore, the main polishing stop condition is updated each time the peripheral flatness is calculated.

[0076] In step S8, following the setting of the main polishing stop condition in step S7, the control unit 30 determines whether or not to stop the main polishing. If the control unit 30 determines YES (main polishing is stopped), the process proceeds to step S9, and if the control unit 30 determines NO (main polishing is continued), the process returns to step S7. Here, the main polishing stop condition is that the calculated value of the peripheral flatness has reached the peripheral flatness set as the main polishing stop condition in step S7. Therefore, in step S8, it is determined whether or not the current peripheral flatness matches the peripheral flatness calculated in step S7.

[0077] In step S9, following the determination in step S8 that the main polishing has been stopped, the control unit 30 stops the main polishing, outputs control commands from the control calculation unit 35 to the first drive unit M1 to the fifth drive unit M5 according to the endpoint polishing conditions defined by the prerequisite polishing conditions, starts execution of endpoint polishing, and proceeds to step S10. Note that during execution of endpoint polishing, the polishing machine 10 polishes the workpiece W while gradually reducing the load applied to the workpiece W and reducing the rotation speeds of the lower platen 11 and upper platen 12, the rotation speed of the sun gear 13, and the rotation speed of the internal gear 14.

[0078] In step S10, following the start of endpoint polishing in step S9, the control unit 30 determines whether endpoint polishing has been completed. If the control unit 30 determines YES (endpoint polishing has been completed), the process proceeds to step S11, and if the control unit 30 determines NO (endpoint polishing has not been completed), the process returns to step S9. Here, the control unit 30 determines that endpoint polishing has been completed when the end condition for endpoint polishing specified in the prerequisite polishing conditions is met, such as when the endpoint polishing time has elapsed or when the load applied to the workpiece W has become equal to or less than a predetermined value.

[0079] In step S11, following the determination in step S10 that the end point polishing has been completed, the control unit 30 ends polishing of the workpiece W by the polishing machine 10, records various polishing data in the memory 36, and proceeds to END.

[0080] The operation of the double-side polishing apparatus 1 of the first embodiment will be explained below by dividing it into "main polishing condition correction control operation," "main polishing stop condition control operation," and "other control operations."

[0081] [Correction Control of Main Polishing Conditions] In the double-sided polishing apparatus 1 of Example 1, the in-plane flatness and the outer circumferential flatness each change during the main polishing. Therefore, unless the main polishing conditions are set taking into consideration the correlation between the in-plane flatness and the outer circumferential flatness, it is difficult to control both the in-plane flatness and the outer circumferential flatness with high precision.

[0082] That is, in a double-sided polishing machine (hereinafter referred to as a "comparative double-sided polishing machine") that learns only the correlation between the in-plane flatness when the peripheral flatness is zero (nm) and the main polishing conditions in that case, and obtains the main polishing conditions by inputting a target value for the in-plane flatness into the comparative learning machine, the state of the peripheral flatness is not taken into consideration, and therefore an appropriate workpiece shape cannot be obtained.

[0083] To be more specific, assume that there is a workpiece W (hereinafter referred to as "example workpiece W'") whose shape changes for each main polishing condition as shown in Figure 10. That is, when main polishing is performed on example workpiece W' under main polishing condition A, the in-plane flatness is +80 (nm) when the peripheral flatness is zero (nm), +90 (nm) when the peripheral flatness is -20 (nm), and +100 (nm) when the peripheral flatness is -50 (nm). Also, when main polishing is performed under main polishing condition B, the in-plane flatness is zero (nm) when the peripheral flatness is zero (nm), +10 (nm) when the peripheral flatness is -20 (nm), and +20 (nm) when the peripheral flatness is -50 (nm). Furthermore, when the main polishing is performed under the main polishing condition C, when the peripheral flatness is zero (nm), the in-plane flatness is −20 (nm), when the peripheral flatness is −20 (nm), the in-plane flatness is zero (nm), and when the peripheral flatness is −50 (nm), the in-plane flatness is +10 (nm).

[0084] In contrast, in the double-sided polishing machine of the comparative example, as described above, the learning machine of the comparative example is made to learn only the correlation between the in-plane flatness when the peripheral flatness is zero (nm) and the main polishing conditions. Therefore, as shown in Figure 11, the data set that is made to be learned by the learning machine of the comparative example is an in-plane flatness of +80 (nm) for main polishing condition A, an in-plane flatness of zero (nm) for main polishing condition B, and an in-plane flatness of -20 (nm) for main polishing condition C.

[0085] Consider a case where, during main polishing, an exemplary workpiece W' is polished so that its in-plane flatness becomes zero (nm) when its peripheral flatness is -20 (nm). Because the comparative double-sided polishing machine cannot take peripheral flatness into account, the peripheral flatness is ignored and the target value for in-plane flatness (here, zero (nm)) is input into the comparative learning machine. This results in a result equivalent to "main polishing condition B." However, when main polishing of the exemplary workpiece W' is performed under main polishing condition B, the comparative double-sided polishing machine is expected to achieve an in-plane flatness of +10 (nm) when the peripheral flatness is -20 (nm) during main polishing. Furthermore, if the in-plane flatness is set to zero (nm), the peripheral flatness is also expected to become zero (nm). In other words, the comparative double-sided polishing machine is unable to accurately match both the in-plane flatness and the peripheral flatness to their target values.

[0086] In contrast, the double-sided polishing apparatus 1 of Example 1 includes a first learning unit 31 and a main polishing condition correction unit 33. The first learning unit 31 is a learning unit that has learned the correlation between the main polishing conditions, the peripheral flatness, and the in-plane flatness. That is, the data set that the first learning unit 31 of Example 1 learns includes various main polishing conditions and the peripheral flatness and in-plane flatness under those conditions, as shown in FIG. 10 . The main polishing condition correction unit 33 corrects the main polishing conditions during the main polishing using a correction target value calculated based on an estimated value of in-plane flatness obtained by inputting the main polishing conditions set at the time of correction and the calculated value of the peripheral flatness at the time of correction into the first learning unit 31, and the main polishing conditions obtained by inputting the final target value of the peripheral flatness into the first learning unit 31.

[0087] That is, in the double-sided polishing machine 1 of Example 1, during main polishing, the main polishing conditions at the time of correction and the calculated value of the outer periphery flatness at the time of correction are input to the first learning device 31. Next, the double-sided polishing machine 1 of Example 1 obtains an estimated value of the in-plane flatness at the time of correction from the input information to the first learning device 31 and the learning result of the first learning device 31, and calculates the difference value from the actual in-plane flatness at the time of correction. Then, the double-sided polishing machine 1 of Example 1 subtracts the difference value from the target value of the in-plane flatness to obtain a "correction target value." Then, the "correction target value" of the in-plane flatness and the target value of the outer periphery flatness are input to the first learning device 31 to obtain new main polishing conditions, and the main polishing conditions set at the time of correction are corrected (step S6).

[0088] Below, assuming that the data set to be learned by the first learning device 31 is the content shown in Figure 10, we will specifically explain the example of main polishing of a specified workpiece W in the double-sided polishing apparatus 1 of Example 1, with the goal of achieving an in-plane flatness of zero (nm) and an outer peripheral flatness of zero (nm) at the final end of polishing.

[0089] In this case, since the target values ​​at the final end of polishing are zero (nm) for the in-plane flatness and zero (nm) for the peripheral flatness, the main polishing is started under main polishing conditions B. If the main polishing conditions are corrected during the main polishing under main polishing conditions B when the calculated value of the peripheral flatness reaches −20 (nm), the main polishing conditions at the time of correction, “main polishing conditions B,” and the calculated value of the peripheral flatness at the time of correction, “−20 (nm),” are input to the first learning device 31 (see FIG. 10 ), and the estimated value of the in-plane flatness obtained as a result becomes “+10 (nm).”

[0090] On the other hand, it is assumed that the calculated value of the in-plane flatness (actual in-plane flatness) is +30 (nm) due to factors such as disturbance.

[0091] In this case, since the estimated value of the in-plane flatness is "+10 (nm)", while the calculated value of the in-plane flatness is "+30 (nm)", it can be inferred that the actual in-plane flatness during the main polishing will be +20 (nm) larger than the learning result of the first learning device 31.

[0092] Therefore, the double-sided polishing machine 1 of Example 1 calculates the difference (+20 (nm)) between the estimated value of in-plane flatness "+10 (nm)" and the calculated value of in-plane flatness "+30 (nm)." Then, the difference (+20 (nm)) is subtracted from the target value of in-plane flatness (zero (nm)) to obtain a "correction target value (-20 (nm))." The double-sided polishing machine 1 of Example 1 then inputs the correction target value of in-plane flatness (-20 (nm)) and the target value of peripheral flatness (zero (nm))) to the first learning device 31 (see FIG. 10 ). As a result, new main polishing conditions, "main polishing conditions C," are acquired. Then, "main polishing conditions C" are set as new main polishing conditions in place of the main polishing conditions (main polishing conditions B) set at the time of correction.

[0093] Then, if the main polishing is performed by correcting the main polishing conditions to main polishing conditions C, the learning result of the first learning device 31 estimates that when the peripheral flatness is zero (nm), the in-plane flatness will be −20 (nm). However, as described above, it is estimated that the actual in-plane flatness during the main polishing will be +20 (nm) greater than the learning result of the first learning device 31. Therefore, the double-sided polishing apparatus 1 can set the in-plane flatness to zero (nm) when the peripheral flatness is zero (nm). In this way, the double-sided polishing apparatus 1 of Example 1 can correct the main polishing conditions during the main polishing in consideration of the correlation between the in-plane flatness and the peripheral flatness, and can accurately control both the in-plane flatness and the peripheral flatness.

[0094] "Control of Main Polishing Stop Conditions" It is known that in the double-sided polishing apparatus 1 of Example 1, the workpiece shape, particularly the peripheral flatness, changes due to the end point polishing. Therefore, even if the workpiece shape reaches the target shape when the main polishing is stopped, the peripheral flatness may change due to the end point polishing, and the final workpiece shape may deviate from the target shape. Furthermore, if the degree of change in the workpiece shape at the end point polishing is different, the amount of change in peripheral flatness during the end point polishing will differ even if the end point polishing is performed under the same end point polishing conditions. As a result, it is expected that the final workpiece shape will differ from the workpiece shape at the end point polishing, and will deviate significantly from the target shape. Therefore, it is necessary to stop the main polishing taking into account the degree of change in the workpiece shape and the amount of change in peripheral flatness.

[0095] In contrast, the double-sided polishing apparatus 1 of Example 1 is equipped with a second learning device 32 and a main-polishing stop condition setting unit 34. Here, the second learning device 32 is a learning device that has learned the correlations among the end-point polishing conditions, the degree of change in in-plane flatness at the time of stopping the main polishing, the degree of change in peripheral flatness at the time of stopping the main polishing, and the amount of change in peripheral flatness during the end-point polishing. Furthermore, the main-polishing stop condition setting unit 34 sets, as the stop condition for the main polishing (main-polishing stop condition), that the peripheral flatness of the workpiece W becomes a target peripheral flatness calculated based on the end-point polishing conditions, the degree of change in in-plane flatness at the time of calculation, and the amount of change in peripheral flatness during the end-point polishing obtained by inputting the end-point polishing conditions, the degree of change in in-plane flatness at the time of calculation, and the degree of change in peripheral flatness at the time of calculation into the second learning device 32, and the target value of peripheral flatness at the end of the end-point polishing.

[0096] That is, in the double-sided polishing machine 1 of Example 1, when setting the main polishing stop condition, the endpoint polishing condition, the degree of change in in-plane flatness at the time of calculation, and the degree of change in peripheral flatness at the time of calculation are input to the second learning unit 32. Then, the double-sided polishing machine 1 of Example 1 obtains the amount of change in peripheral flatness during endpoint polishing from the input information to the second learning unit 32 and the learning results of the second learning unit 32. Then, by back-calculating from the amount of change in peripheral flatness during endpoint polishing, the final target value of peripheral flatness, and the endpoint polishing time, the target value of peripheral flatness (target peripheral flatness) at the time of stopping main polishing that is necessary to satisfy the target value of peripheral flatness is determined. Then, the main polishing stop condition is specified based on the target peripheral flatness (step S7).

[0097] In this way, the double-sided polishing apparatus 1 of Example 1 can stop the main polishing so that the peripheral flatness at the time the main polishing is stopped will have the required shape, taking into account the amount of change in peripheral flatness due to end-point polishing, taking into account the degree of change in in-plane flatness and peripheral flatness. Therefore, even if the shape of the workpiece W changes due to end-point polishing, the double-sided polishing apparatus 1 of Example 1 can prevent the workpiece shape from significantly deviating from the target shape at the end of end-point polishing (final end of polishing).

[0098] 12 shows the results of the shapes of the workpieces W after endpoint polishing was performed in the double-sided polishing machine 1 of Example 1, with the target value of peripheral flatness at the end of polishing set to zero (nm). As shown in FIG. 12, the first sample workpiece had an in-plane flatness of +34 (nm) and a peripheral flatness of -2 (nm). The second sample workpiece had an in-plane flatness of -4 (nm) and a peripheral flatness of -5 (nm). The third sample workpiece had an in-plane flatness of +40 (nm) and a peripheral flatness of -3 (nm). The fourth sample workpiece had an in-plane flatness of +62 (nm) and a peripheral flatness of -3 (nm).

[0099] These results show that the double-sided polishing machine 1 of Example 1 can reduce the difference between the actual peripheral flatness and the target value (zero (nm)) regardless of the magnitude of the in-plane flatness. Therefore, it can be seen that the double-sided polishing machine 1 of Example 1 can prevent the workpiece shape from significantly deviating from the target shape at the end point of polishing (final end point of polishing).

[0100] FIG. 13 shows the polishing results of the workpiece W at the final polishing completion point when the workpiece W was polished using the double-sided polishing apparatus 1 of Example 1. As shown in FIG. 13 , for the fifth sample workpiece, the target values ​​for in-plane flatness and peripheral flatness were set to zero (nm) and zero (nm), respectively. The actual polishing results were an in-plane flatness of −1 (nm) and a peripheral flatness of +3 (nm). For the sixth sample workpiece, the target values ​​for in-plane flatness and peripheral flatness were set to zero (nm), respectively. The actual polishing results were an in-plane flatness of −4 (nm) and a peripheral flatness of −20 (nm). For the seventh sample workpiece, the target values ​​for in-plane flatness and peripheral flatness were set to −30 (nm) and zero (nm), respectively. The actual polishing results were an in-plane flatness of −30 (nm) and a peripheral flatness of +2 (nm). For the eighth sample workpiece, the target value for in-plane flatness was set to +30 (nm) and the target value for peripheral flatness was set to zero (nm), and the actual polishing result was an in-plane flatness of +30 (nm) and a peripheral flatness of +3 (nm). For the ninth sample workpiece, the target value for in-plane flatness was set to zero (nm) and the target value for peripheral flatness was set to zero (nm), and the actual polishing result was an in-plane flatness of -5 (nm) and a peripheral flatness of -4 (nm).

[0101] The results shown in Figure 13 reveal that the double-sided polishing apparatus 1 of Example 1 does not significantly deviate from the target values ​​for both the in-plane flatness and the peripheral flatness, and can prevent the workpiece shape at the end of final polishing from deviating from the target shape.

[0102] [Other Control Functions] In the double-sided polishing apparatus 1 of Example 1, an in-plane region G and an outer peripheral region E are defined on the workpiece W, and the workpiece shape is quantified and displayed based on the in-plane flatness and outer peripheral flatness. That is, in Example 1, the workpiece shape is divided into multiple regions and controlled. Meanwhile, the workpiece W polished by the double-sided polishing apparatus 1 of Example 1 may be used as a substrate for a semiconductor device having a fine electronic circuit formed on its surface. Here, the electronic circuit formed on the surface of the workpiece W may straddle the boundary between the in-plane region G and the outer peripheral region E. Therefore, it is desirable that the boundary between the in-plane region G and the outer peripheral region E be a smooth surface so that no inflection point is formed at the boundary between the in-plane region G and the outer peripheral region E. To control the workpiece shape so that no inflection point is formed at the boundary between the in-plane region G and the outer peripheral region E, it is necessary to numerically express the flatness (smoothness of the workpiece shape) at the boundary between different regions, such as the in-plane region G and the outer peripheral region E, i.e., the change in the inclination of the workpiece shape at the boundary between the in-plane region G and the outer peripheral region E.

[0103] In contrast, in the double-sided polishing apparatus 1 of Example 1, the in-plane flatness is determined as the gradient of an approximate straight line set for the group of thickness data for the in-plane region G. Moreover, the outer periphery flatness is determined by performing a continuation process on the gradient of an approximate straight line set for the group of thickness data for the outer periphery region E, with the in-plane flatness as the reference. In other words, in the double-sided polishing apparatus 1 of Example 1, the outer periphery flatness is determined as the sum of the gradient of the approximate straight line set for the group of thickness data for the outer periphery region E and the in-plane flatness.

[0104] In this way, the double-sided polishing machine 1 of Example 1 calculates the in-plane flatness and the peripheral flatness separately, but by calculating the peripheral flatness by performing a continuation process on the slope of an approximate line set for the thickness data group of the peripheral region E based on the in-plane flatness, the slope of the approximate line set for the data group of the in-plane region G can be used as a reference line, and the slope of the approximate line set for the data group of the peripheral region E relative to this reference line can be quantified as the peripheral flatness. Therefore, the peripheral flatness can be expressed as an index expressing the degree of roll-off with respect to the in-plane flatness, and the flatness (smoothness of the workpiece shape) at the boundary between different regions such as the in-plane region G and the peripheral region E can be expressed numerically. Furthermore, the double-sided polishing machine 1 of Example 1 can numerically indicate the shape of the workpiece W, making it easier for operators and the like to understand the workpiece shape.

[0105] The outer peripheral flatness may be calculated by subtracting from the in-plane flatness the gradient of the approximation line set for the thickness data group of the outer peripheral region E. In other words, the outer peripheral flatness may be the difference between the in-plane flatness and the gradient of the approximation line set for the thickness data group of the outer peripheral region E.

[0106] Furthermore, the double-sided polishing apparatus 1 of Example 1 may divide the outer peripheral region E into multiple sections along the radial direction of the workpiece W, and calculate the outer peripheral flatness for each of the divided outer peripheral regions (innermost section Ei, outer section Eo). By dividing the outer peripheral region E into multiple sections, the double-sided polishing apparatus 1 of Example 1 can correct the main polishing conditions and set main polishing stop conditions in accordance with changes in the shape of each section. This allows the workpiece shape at the end of polishing to match the target shape with even greater accuracy.

[0107] Furthermore, in the double-sided polishing apparatus 1 of Example 1, when the outer peripheral region E is divided into multiple sections, the section adjacent to the in-plane region G is the innermost section Ei, and the section radially outward of the innermost section Ei is the outer section Eo.

[0108] The flatness of the innermost section Ei is determined by performing a continuation process on the slope of an approximation line set for the thickness data group of the innermost section Ei, with the in-plane flatness used as a reference. The flatness of the innermost section Ei in Example 1 is the sum of the slope of the approximation line set for the thickness data group of the innermost section Ei and the in-plane flatness. The flatness of the outer section Eo is determined by performing a continuation process on the slope of the approximation line set for the thickness data group of the outer section Eo, with the flatness of the section adjacent to the outer section Eo radially inward (e.g., the innermost section Ei) used as a reference. The flatness of the outer section Eo in Example 1 is the sum of the slope of the approximation line set for the thickness data group of the outer section Eo and the flatness of the section adjacent to the outer section Eo radially inward (e.g., the innermost section Ei).

[0109] As a result, in the double-sided polishing apparatus 1 of Example 1, even if the outer peripheral region E is divided into multiple sections, the flatness of the outer peripheral region E (smoothness of the workpiece shape) can be appropriately expressed, and the workpiece shape can be precisely controlled.

[0110] The flatness of the innermost section Ei may be calculated by subtracting the slope of the approximation line set for the thickness data group of the innermost section Ei from the in-plane flatness. In other words, the flatness of the innermost section Ei may be the difference between the in-plane flatness and the slope of the approximation line set for the thickness data group of the innermost section Ei.

[0111] The flatness of the outer section Eo may be calculated by subtracting the slope of the approximation line set for the thickness data set of the outer section Eo from the flatness of the section adjacent to the outer section Eo on the radial inside (e.g., the innermost section Ei). In other words, the flatness of the outer section Eo may be the difference between the flatness of the section adjacent to the outer section Eo on the radial inside (e.g., the innermost section Ei) and the slope of the approximation line set for the thickness data set of the outer section Eo.

[0112] Furthermore, the outer section Eo in Example 1 is divided into two sections, a first outer section Eo1 and a second outer section Eo2. However, the outer section Eo may not be divided, or may be divided into three or more sections.

[0113] The double-side polishing machine of the present invention has been described above based on Example 1, but the specific configuration is not limited to this Example, and design changes and additions are permitted as long as they do not deviate from the gist of the invention according to each claim.

[0114] In the double-sided polishing apparatus 1 of Example 1, when setting the target values ​​for in-plane flatness and peripheral flatness in step S1 of the polishing process control shown in Figure 9, an example is shown in which the final target value for in-plane flatness at the end of polishing endpoint and the final target value for peripheral flatness at the end of polishing endpoint, which are the final processing targets for the workpiece W, are set. However, the target values ​​for in-plane flatness and peripheral flatness are not limited to this. For example, multiple target values ​​may be set according to the polishing status of the workpiece W, and the target values ​​may be appropriately changed when correcting the main polishing conditions and when setting the main polishing stop conditions. CROSS-REFERENCE TO RELATED APPLICATIONS

[0115] This application claims priority based on Japanese Patent Application No. 2023-204053, filed with the Japan Patent Office on December 1, 2023, the entire disclosure of which is incorporated herein by reference in its entirety.

Claims

1. A double-sided polishing apparatus comprising: a polishing machine that holds a disk-shaped workpiece between a lower platen and an upper platen arranged opposite the lower platen, and polishes the front and back surfaces of the workpiece by moving the lower platen, the upper platen and the workpiece relatively while a load is applied to the workpiece; a thickness gauge that measures the thickness of the workpiece while the workpiece is being polished by the polishing machine; and a control unit that controls the polishing machine based on the measurement results of the thickness gauge, wherein the workpiece has an outer circumferential region of a predetermined range extending radially inward from its outer circumferential end, and an in-plane region of a range extending from the outer circumferential region to the center of the workpiece, and the control unit comprises: a first learning device that has been made to learn the correlation between polishing conditions for main polishing in which the workpiece is polished while maintaining the load at a target load, and the outer circumferential flatness, which is the flatness of the outer circumferential region, and the in-plane flatness, which is the flatness of the in-plane region; a second learning device that has learned correlations between polishing conditions for end point polishing in which the workpiece is polished while gradually reducing the load, a degree of change in the in-plane flatness at the time when the main polishing is stopped, a degree of change in the outer periphery flatness at the time when the main polishing is stopped, and an amount of change in the outer periphery flatness during the end point polishing; a main polishing condition correction unit that obtains a correction target value for the in-plane flatness based on the polishing conditions for the main polishing set at the time of correction and an estimated value of the in-plane flatness obtained by inputting the outer periphery flatness at the time of correction to the first learning device, and corrects the polishing conditions for the main polishing set at the time of correction using the polishing conditions for the main polishing obtained by inputting the correction target value for the in-plane flatness and the target value for the outer periphery flatness to the first learning device; a main polishing stop condition setting unit that inputs the polishing conditions of the endpoint polishing, the degree of change in the in-plane flatness at the time of calculation, and the degree of change in the peripheral flatness at the time of calculation to the second learning device to obtain an amount of change in the peripheral flatness during the endpoint polishing, and sets a stop condition for the main polishing based on the amount of change in the peripheral flatness and a target value of the peripheral flatness.

2. A double-sided polishing apparatus as described in claim 1, characterized in that the in-plane flatness is the gradient of an approximation line set for a group of thickness data of the in-plane region, and the outer periphery flatness is determined by performing a continuum process on the gradient of an approximation line set for a group of thickness data of the outer periphery region based on the in-plane flatness.

3. A double-sided polishing machine according to claim 1 or 2, characterized in that the outer peripheral region is divided into a plurality of sections along the radial direction of the workpiece, and the outer peripheral flatness is calculated for each section of the divided outer peripheral region.

4. A double-sided polishing apparatus as described in claim 3, characterized in that, when a partition adjacent to the in-plane region is defined as an innermost partition and a partition radially outward of the innermost partition is defined as an outer partition, the flatness of the innermost partition is determined by performing a continuum process on the slope value of an approximation straight line set for a group of thickness data of the innermost partition based on the in-plane flatness, and the flatness of the outer partition is determined by performing a continuum process on the slope value of an approximation straight line set for a group of thickness data of the outer partition based on the flatness of a partition adjacent to the radially inward side of the outer partition.

5. A double-sided polishing method in which a disk-shaped workpiece is sandwiched between a lower platen and an upper platen arranged opposite the lower platen, and the lower platen, the upper platen, and the workpiece are moved relatively while a load is applied to the workpiece to polish the front and back surfaces of the workpiece, using a first learning device that has learned the correlation between polishing conditions for main polishing in which the workpiece is polished while maintaining the load at a target load, peripheral flatness, which is the flatness of an outer peripheral region set in a predetermined range from the outer peripheral end of the workpiece toward the inside in the radial direction, and in-plane flatness, which is the flatness of an in-plane region of the workpiece in the range from the outer peripheral region to the center of the workpiece, and a second learning device that has learned the correlation between polishing conditions for end-point polishing in which the workpiece is polished while gradually reducing the load, the degree of change in the in-plane flatness at the time when the main polishing is stopped, the degree of change in the outer peripheral flatness at the time when the main polishing is stopped, and the amount of change in the outer peripheral flatness during the end-point polishing, a step of setting a target value for the outer circumferential flatness and a target value for the in-plane flatness; a step of performing initial polishing in which the workpiece is polished while gradually increasing the load after setting the target values ​​for the outer circumferential flatness and the in-plane flatness; a step of starting the main polishing after the initial polishing is completed; and a step of, during the execution of the main polishing, determining a correction target value for the in-plane flatness based on the polishing conditions for the main polishing set at the time of correction and an estimated value of the in-plane flatness obtained by inputting the outer circumferential flatness at the time of correction into the first learning device, and correcting the polishing conditions for the main polishing set at the time of correction using the polishing conditions for the main polishing obtained by inputting the correction target value for the in-plane flatness and the target value for the outer circumferential flatness into the first learning device. a step of: inputting, during execution of the main polishing, polishing conditions for the endpoint polishing, a degree of change in the in-plane flatness at the time of calculation, and a degree of change in the peripheral flatness at the time of calculation into the second learning device to obtain an amount of change in the peripheral flatness during the endpoint polishing, and setting a stop condition for the main polishing based on the amount of change in the peripheral flatness and a target value for the peripheral flatness; and a step of executing the endpoint polishing after the stop condition for the main polishing is established.

Images