Method for detecting circumferential thickness variation of a workpiece

Abstract

This invention provides a method for detecting circumferential thickness variations in a workpiece that can quickly detect thickness variations in the circumferential direction of the workpiece caused by double-sided polishing. [Solution] The upper or lower platen of the double-sided polishing apparatus has one or more workpiece thickness measuring holes that penetrate from the upper surface to the lower surface of the upper or lower platen, and the double-sided polishing method of a workpiece is characterized by comprising: a first step of performing double-sided polishing on the workpiece and obtaining workpiece thickness data by measuring the thickness of the workpiece in real time through the one or more workpiece thickness measuring holes using a workpiece thickness measuring instrument during double-sided polishing; and a second step of detecting the circumferential thickness variation of the workpiece based on the variation in the workpiece thickness data at least in the middle stage of the double-sided polishing process.

Description

Technical Field

[0001] The present invention relates to a method for detecting the thickness variation in the circumferential direction of a workpiece.

Background Art

[0002] In the manufacture of semiconductor wafers such as silicon wafers, which are typical examples of workpieces to be polished, in order to obtain higher-precision wafer flatness quality and surface roughness quality, a double-sided polishing process for polishing the front and back surfaces of the wafer simultaneously is generally adopted.

[0003] Double-sided polishing of a semiconductor wafer is performed by holding the semiconductor wafer on a carrier plate provided with one or more wafer holding holes for holding the semiconductor wafer, sandwiching the semiconductor wafer between an upper platen and a lower platen that form a rotating platen, and controlling the rotation of the carrier plate by the rotation of a sun gear provided at the center of the rotating platen and the rotation of an internal gear provided at the outer peripheral portion of the rotating platen, and relatively rotating the rotating platen and the carrier plate.

[0004] In recent years, due to the miniaturization of semiconductor elements and the increase in the diameter of semiconductor wafers, there has been a strong demand for a method for terminating polishing at an appropriate timing against the background that the flatness requirements for semiconductor wafers during exposure have become stricter.

[0005] In a typical double-sided polishing process, in the initial stages of polishing, the wafer's overall shape is convex upwards, and a significant sag is visible around the outer edge. At this point, the wafer thickness is significantly greater than the carrier plate thickness. As polishing progresses, the wafer's overall shape approaches flatness, but a sag remains around the outer edge. At this point, the wafer thickness is slightly greater than the carrier plate thickness. Further polishing results in a nearly flat wafer surface, and the amount of sag around the outer edge decreases. At this point, the wafer thickness and the carrier plate thickness are approximately equal. Subsequently, as polishing progresses, the wafer's shape gradually becomes concave in the center, and the outer edge becomes tapered. At this stage, the wafer thickness becomes thinner than the carrier plate thickness.

[0006] For the reasons described above, in order to obtain wafers with high flatness across the entire surface and outer edge, it is common practice to polish the wafer until its thickness is approximately equal to the thickness of the carrier plate, and this process has traditionally been controlled by the operator adjusting the polishing time.

[0007] However, adjusting the polishing time by the worker was heavily influenced by the polishing environment, such as the timing of changing polishing auxiliary materials and the timing of stopping the equipment. As a result, it was not always possible to accurately control the amount of polishing, and ultimately, it relied heavily on the worker's experience.

[0008] In contrast, for example, Patent Document 1 proposes a double-sided wafer polishing apparatus that can measure the thickness of the wafer being polished in real time from a workpiece thickness measuring hole above the upper platen (or below the lower platen), and determine the timing of the end of polishing based on the measurement results.

[0009] In the method described in Patent Document 1, the timing for ending double-sided polishing is determined based on the wafer thickness measurement results, so polishing can be terminated at a predetermined thickness. However, there was a problem in that the shape of the polished wafer did not match the target shape.

[0010] Therefore, in Patent Document 2, the applicant proposed a double-sided polishing apparatus that can measure the thickness of a wafer in real time during double-sided polishing, determine the overall shape index of the wafer from the measured wafer thickness, and terminate double-sided polishing at the timing when the overall shape of the wafer becomes the target shape during double-sided polishing.

[0011] Furthermore, in Patent Document 3, the applicant has further improved the invention described in Patent Document 2 and proposed a double-sided polishing apparatus and method that, taking into account the life fluctuations of auxiliary materials such as polishing pads, carrier plates, and slurry in a double-sided polishing apparatus for workpieces, can terminate double-sided polishing at the timing when the overall shape of the wafer reaches the target shape, even when batch processing of double-sided polishing of wafers is repeatedly performed. [Prior art documents] [Patent Documents]

[0012] [Patent Document 1] Japanese Patent Publication No. 2010-030019 [Patent Document 2] Japanese Patent Publication No. 2019-118975 [Patent Document 3] Japanese Patent Publication No. 2020-15122 [Overview of the project] [Problems that the invention aims to solve]

[0013] The methods described in Patent Documents 2 and 3 allow for the completion of double-sided polishing at the timing when the overall shape of the wafer reaches the target shape. However, when double-sided polishing of a semiconductor wafer is repeatedly performed, the rotation of the wafer during double-sided polishing may be inhibited depending on the condition of the polishing auxiliary materials (e.g., polishing pad). If the rotation of the wafer stops during double-sided polishing, the distance traveled per unit time differs between the outer and inner circumferences of the carrier plate. As a result, polishing progresses more rapidly in the portion of the semiconductor wafer located on the outer circumference of the carrier plate than in the portion located on the inner circumference, causing variations in the wafer thickness in the circumferential direction (variations when measured along the circumference of the wafer).

[0014] Conventionally, the above-mentioned variation in wafer thickness in the circumferential direction was detected by measuring the flatness of the semiconductor wafer using a surface inspection device after the completion of the double-sided polishing process. Therefore, even if variation in wafer thickness in the circumferential direction occurs due to polishing auxiliary materials, the occurrence of the thickness variation cannot be detected until the wafer flatness is measured. As a result, there was a problem in that the double-sided polishing process of semiconductor wafers was continued in an environment where thickness variation could occur.

[0015] Furthermore, the flatness measurement described above is not performed on all semiconductor wafers that have undergone the double-sided polishing process; rather, it is performed on only a select few semiconductor wafers, for example, one in 25 or one in 50. Therefore, even when flatness measurements are performed, semiconductor wafers with thickness variations in the circumferential direction may be missed.

[0016] The present invention has been made in view of the above problems, and its objective is to propose a method for detecting circumferential variation in a workpiece that can quickly detect the occurrence of thickness variation in the circumferential direction of the workpiece during a double-sided polishing process. [Means for solving the problem]

[0017] The present invention, which solves the above problems, is as follows.

[0018] [1] A method for double-sided polishing a workpiece, comprising: holding a workpiece in a carrier plate provided with one or more workpiece holding holes for holding a workpiece; sandwiching the workpiece between a rotating platen consisting of an upper platen and a lower platen; controlling the rotation and revolution of the carrier plate by the rotation of a sun gear provided in the center of the rotating platen and the rotation of an internal gear provided on the outer circumference of the rotating platen; thereby causing the rotating platen and the carrier plate to rotate relative to each other and simultaneously polishing both sides of the workpiece, The upper or lower platen has one or more workpiece thickness measuring holes that penetrate from the upper surface to the lower surface of the upper or lower platen, The method for polishing both sides of the workpiece is as follows: The first step involves performing double-sided polishing on the workpiece, and during double-sided polishing, measuring the thickness of the workpiece in real time using one or more workpiece thickness measuring instruments that can measure the thickness of the workpiece in real time through one or more workpiece thickness measuring holes, thereby acquiring workpiece thickness data. A second step of detecting the occurrence of circumferential thickness variation of the workpiece based on the variation in the thickness data of the workpiece at least in the middle stage of the double-sided polishing process, A method for detecting circumferential thickness variation of a workpiece, characterized by comprising the following:

[0019] [2] The method for detecting circumferential thickness variation of a workpiece according to [1], wherein the detection of the occurrence of thickness variation in the second step is performed based on an increase in the thickness variation of the workpiece in the middle stage of the double-sided polishing process and a decrease in the thickness variation of the workpiece from the middle stage onward of the double-sided polishing process.

[0020] [3] The method for detecting circumferential thickness variation of a workpiece according to [1] or [2], wherein the detection of the occurrence of thickness variation in the second step is performed based on the Mean Square Error (MSE) value, which is the value of the mean square error (MSE) between the obtained approximation curve and the thickness data of the workpiece.

[0021] [4] The detection of the occurrence of thickness variation in the second step is performed based on the value of the mean square error obtained by dividing the double-sided polishing step into a plurality of time zones and obtaining the value of the mean square error for each time zone, and is the method for detecting the thickness variation in the circumferential direction of the workpiece according to [3] above.

[0022] [5] When the number of divisions of the double-sided polishing step is 6, and the difference between the maximum value of the mean square error in the six time zones and the maximum value of the values of the mean square error in the first time zone and the last time zone of the double-sided polishing step is greater than 0.10 μm, it is determined that variation has occurred in the circumferential direction of the workpiece, according to the method described in [4] above.

[0023] [6] Further comprising a third step of classifying the obtained thickness data of the workpiece for each workpiece between the first step and the second step, and the second step is performed for each workpiece, which is the method for detecting the thickness variation in the circumferential direction of the workpiece according to any one of [1] to [5] above.

[0024] [7] The workpiece is a silicon wafer, which is the method for detecting the thickness variation in the circumferential direction of the workpiece according to any one of [1] to [6] above.

Advantages of the Invention

[0025] According to the present invention, it is possible to quickly detect the occurrence of thickness variation in the circumferential direction of the workpiece in the double-sided polishing step.

Brief Description of the Drawings

[0026] [Figure 1] It is a diagram showing the relationship between the polishing time and the thickness of the wafer, and is for (a) the case where there is no thickness variation in the circumferential direction of the wafer, and (b) the case where there is thickness variation in the circumferential direction of the wafer. [Figure 2] It is a top view of an example of a double-sided polishing apparatus for a workpiece that can be used in the present invention. [Figure 3] It is a cross-sectional view taken along line A-A in FIG. 1. [Figure 4]This figure shows the relationship between polishing time and MSE value for silicon wafers with no thickness variation in the circumferential direction and silicon wafers with thickness variation in the circumferential direction, when the number of divisions in the double-sided polishing process is set to 6. [Figure 5] This figure shows the relationship between the difference in mean squared error values ​​and GBIR-GFLR. [Figure 6] This figure shows the occurrence rate of thickness variation in the wafer circumferential direction for double-sided polishing machines A to F. [Modes for carrying out the invention]

[0027] Embodiments of the present invention will be described below with reference to the drawings. The method for detecting wafer thickness variation in the circumferential direction according to the present invention is a method for polishing both sides of a workpiece, in which a workpiece is held in a carrier plate provided with one or more workpiece holding holes for holding the workpiece, the workpiece is sandwiched between a rotating platen consisting of an upper platen and a lower platen, the rotation and revolution of the carrier plate are controlled by the rotation of a sun gear provided in the center of the rotating platen and the rotation of an internal gear provided on the outer circumference of the rotating platen, thereby causing the rotating platen and the carrier plate to rotate relative to each other and polishing both sides of the workpiece simultaneously. Here, the upper platen or the lower platen has one or more work thickness measuring holes that penetrate from the upper surface to the lower surface of the upper platen or the lower platen, and the double-sided polishing method for the workpiece is characterized by comprising: a first step of performing double-sided polishing on the workpiece and obtaining workpiece thickness data by measuring the thickness of the workpiece in real time using one or more work thickness measuring instruments that can measure the thickness of the workpiece in real time from the one or more work thickness measuring holes during double-sided polishing; and a second step of detecting the occurrence of circumferential thickness variation of the workpiece based on the variation of the workpiece thickness data at least in the middle stage of the double-sided polishing process.

[0028] As described above, when both sides of a workpiece, such as a semiconductor wafer, are repeatedly polished, the rotation of the workpiece may be inhibited depending on the condition of the polishing auxiliary materials (e.g., polishing pad), which can cause variations in the thickness of the workpiece in the circumferential direction. Conventionally, such variations in the thickness of the workpiece in the circumferential direction were detected by measuring the flatness of the workpiece after the double-sided polishing process. However, this method was slow to detect, and because only some workpieces were measured, workpieces with variations in thickness in the circumferential direction were missed.

[0029] The inventors diligently investigated methods for detecting thickness variations in the circumferential direction of a workpiece more quickly than in conventional methods. To this end, they focused on thickness data acquired during the double-sided polishing process and conducted a detailed investigation into the differences between workpieces that exhibit thickness variations in the circumferential direction and those that do not. As a result, they found that there is a significant difference in the behavior of workpiece thickness during the double-sided polishing process between workpieces that exhibit thickness variations in the circumferential direction and those that do not.

[0030] Figure 1 shows the relationship between polishing time and the thickness of the silicon wafer used as a workpiece. (a) represents a wafer with no thickness variation in the circumferential direction, and (b) represents a wafer with thickness variation in the circumferential direction. As shown in Figure 1(a), when there is no (small) thickness variation in the circumferential direction of the wafer, the thickness variation of the silicon wafer remains generally constant from the start to the end of the double-sided polishing process. In contrast, as shown in Figure 1(b), when there is (large) thickness variation in the circumferential direction of the wafer, the thickness variation of the silicon wafer increases around the middle stage of the double-sided polishing process. Then, from the middle stage onward of the double-sided polishing process (i.e., the middle and final stages), the increased thickness variation of the silicon wafer decreases.

[0031] The reason why the thickness of a silicon wafer behaves as described above when circumferential thickness variation occurs is thought to be as follows: Silicon wafers subjected to the double-sided polishing process have a degree of flatness sufficient to satisfy a predetermined flatness quality for the final product. Therefore, even if the rotation of the silicon wafer is inhibited in the early stages of the double-sided polishing process, causing circumferential thickness variation, it is not noticeable. However, as the double-sided polishing process progresses to around the middle stage, it is thought that the circumferential thickness variation accumulates and becomes apparent.

[0032] As the double-sided polishing process progresses to its final stages, the thickness of the silicon wafer approaches the thickness of the carrier plate. As a result, the polishing rate of the silicon wafer located on the outer edge, where polishing had progressed more, decreases compared to the polishing rate of the portion located on the inner edge of the carrier plate. This is thought to be the reason why the thickness variation decreased as the polishing progressed more in the portion located on the inner edge of the carrier plate.

[0033] Thus, the inventors have discovered that it is possible to detect circumferential thickness variations of a workpiece based on variations in workpiece thickness data at least in the middle stages of the double-sided polishing process, and have completed the present invention.

[0034] Figure 2 shows a top view of an example of a workpiece double-sided polishing apparatus that can be used in the method of the present invention. Figure 3 is a cross-sectional view of AA in Figure 2. As shown in Figures 2 and 3, the double-sided polishing apparatus 1 comprises a rotating platen 4 having an upper platen 2 and a lower platen 3 opposite to it, a sun gear 5 provided at the rotation center of the rotating platen 4, and an internal gear 6 provided in an annular shape on the outer circumference of the rotating platen 4. As shown in Figure 3, polishing pads 7 are attached to the opposing surfaces of the upper and lower rotating platens 4, that is, the lower surface side which is the polishing surface of the upper platen 2 and the upper surface side which is the polishing surface of the lower platen 3.

[0035] Furthermore, as shown in Figures 2 and 3, the apparatus 1 is provided between the upper platen 2 and the lower platen 3 and includes a plurality of carrier plates 9, each having one or more (one in the illustrated example) workpiece holding holes 8 for holding workpieces W. Note that only one of the plurality of carrier plates 9 is shown in Figure 2. The number of workpiece holding holes 8 can be one or more; for example, there can be three. In the illustrated example, workpieces W (for example, semiconductor wafers such as silicon wafers) are held in the workpiece holding holes 8.

[0036] The above-described apparatus 1 is a planetary gear type double-sided polishing apparatus that can cause the carrier plate 9 to undergo planetary motion, including orbital and rotational motion, by rotating the sun gear 5 and the internal gear 6. That is, while supplying polishing slurry, the carrier plate 9 is made to undergo planetary motion, and at the same time, the upper platen 2 and the lower platen 3 are rotated relative to the carrier plate 9. As a result, while the workpiece W is rotated within the workpiece holding hole 8, the polishing pads 7 attached to the upper and lower rotating platens 4 slide against both sides of the workpiece W held in the workpiece holding hole 8 of the carrier plate 9, allowing both sides of the workpiece W to be polished simultaneously.

[0037] Furthermore, as shown in Figures 2 and 3, in the apparatus 1, the upper platen 2 is provided with one or more wafer thickness measuring holes (hereinafter also simply referred to as "holes") 10 that penetrate from the upper surface of the upper platen 2 to the lower surface which is the polishing surface. In the illustrated example, one hole 10 is positioned to pass near the center of the workpiece W. In this example, the hole 10 is provided in the upper platen 2, but it may also be provided in the lower platen 3, and it is sufficient to provide one or more holes 10 in either the upper platen 2 or the lower platen 3. Also, in the example shown in Figures 2 and 3, one hole 10 is provided, but multiple holes may be arranged around the periphery of the upper platen 2 (on the dashed line in Figure 1). Here, as shown in Figure 3, the hole 10 penetrates from the upper surface of the upper platen 2 to the lower surface of the polishing pad 7.

[0038] Furthermore, as shown in Figure 3, the apparatus 1 is equipped with one or more (one in the illustrated example) work thickness measuring instruments 11 above the upper platen 2 in the illustrated example, which can measure the thickness of the workpiece W in real time from one or more (one in the illustrated example) work thickness measuring holes 10 while the workpiece W is being polished on both sides. In this example, the work thickness measuring instrument 11 is a tunable infrared laser device. For example, this work thickness measuring instrument 11 may include an optical unit that irradiates the workpiece W with laser light, a detection unit that detects the laser light reflected from the workpiece W, and a calculation unit that calculates the thickness of the workpiece W from the detected laser light. With such a work thickness measuring instrument 11, the thickness of the workpiece W can be calculated from the difference in optical path length between the reflected light reflected from the front surface of the workpiece W and the reflected light reflected from the back surface of the workpiece W. Note that the work thickness measuring instrument 11 only needs to be capable of measuring the thickness of the workpiece W in real time and is not particularly limited to those using infrared lasers as described above. Furthermore, the workpiece thickness measuring instrument 11 is not fixed to the upper platen 2 (or the lower platen 3 if the hole 10 is located on the lower platen 3) which has the hole 10, and does not rotate together with the upper platen 2 (or the lower platen 3 if the hole 10 is located on the lower platen 3).

[0039] Furthermore, as shown in Figure 3, the device 1 includes a control unit 12. As shown in Figure 3, in this example, the control unit 12 is connected to the upper and lower platens 2 and 3, the sun gear 5, the internal gear 6, and the workpiece thickness measuring instrument 11.

[0040] The apparatus 1 includes a calculation unit 13 that determines the timing to end the double-sided polishing of the workpiece W during the double-sided polishing process. The calculation unit 13 is connected to the control unit 12. The calculation unit 13 acquires workpiece thickness data measured by the workpiece thickness measuring instrument 11 and determines the timing to end the double-sided polishing of the workpiece W based on a predetermined algorithm.

[0041] As described in Patent Document 2, the above-mentioned predetermined algorithm can be one that determines the overall shape index of the wafer from the measured wafer thickness data and terminates the double-sided polishing process at the timing when the overall shape of the wafer becomes the target shape.

[0042] Furthermore, as described in Patent Document 3, the above-mentioned predetermined algorithm can be one that takes into account the life fluctuations of auxiliary materials such as polishing pads, carrier plates, and slurry in a double-sided polishing apparatus for workpieces, and terminates double-sided polishing at the timing when the overall shape of the wafer reaches the target shape, even when batch processing of double-sided polishing of wafers is repeatedly performed.

[0043] <1st process> First, the workpiece W is polished on both sides, and during the double-sided polishing process, the thickness of the workpiece W is measured in real time using one or more workpiece thickness measuring instruments 11 that can measure the thickness of the workpiece W in real time from one or more workpiece thickness measuring holes 10, and the thickness data of the workpiece W is acquired (first step).

[0044] Examples of workpieces W suitable for double-sided polishing include semiconductor wafers such as silicon wafers, germanium wafers, and gallium arsenide wafers. In particular, silicon wafers can be polished well on both sides.

[0045] Furthermore, double-sided polishing of the workpiece W and real-time measurement of the thickness of the workpiece W during the double-sided polishing process can be performed, for example, using the apparatus 1 shown in Figures 2 and 3.

[0046] <Second process> Next, the circumferential thickness variation of the workpiece W is detected based on the variation in the workpiece thickness data at least in the middle stages of the double-sided polishing process (second step).

[0047] As described above, if there is variation in the thickness data in the circumferential direction of the workpiece, the variation in the thickness data of the workpiece W will increase in the middle stage of the double-sided polishing process. Then, in the middle and final stages of the double-sided polishing process, the increased variation in the thickness of the workpiece W will decrease. Therefore, the second step is performed based on the variation in the thickness data of the workpiece W at least in the middle stage of the double-sided polishing process.

[0048] In this specification, the entire double-sided polishing process is divided into three equal parts, with the first stage referred to as the "initial stage," the second stage as the "intermediate stage," and the third stage as the "final stage."

[0049] Specifically, the detection of the workpiece circumferential thickness variation described above can be performed based on an increase in the thickness variation of the workpiece W during the middle stage of the double-sided polishing process, and a decrease in the thickness variation of the workpiece W after the middle stage of the double-sided polishing process.

[0050] Furthermore, the detection of circumferential thickness variation of the workpiece W can be performed more specifically by approximating the thickness data of the workpiece W with a polynomial and basing the detection on the Mean Square Error (MSE) value, which is the value of the error between the obtained approximation curve and the thickness data of the workpiece W.

[0051] The thickness data of the workpiece W can be well fitted with a polynomial. Therefore, by approximating the thickness data of the workpiece W with a polynomial, the circumferential thickness variation of the workpiece W can be detected based on the MSE value between the obtained approximation curve and the thickness data of the workpiece W. As described in Patent Document 2, the thickness data of the workpiece W can be best fitted with a quadratic function. For this reason, it is preferable to use a quadratic polynomial.

[0052] Furthermore, the detection of variations in the circumferential thickness of the workpiece W can be performed more specifically by dividing the double-sided polishing process into multiple time periods, determining the MSE value for each time period, and then using the determined MSE value as the basis for detection.

[0053] If the number of divisions in the double-sided polishing process described above is too small, thickness variations caused by factors other than thickness variations in the circumferential direction of the wafer will become apparent. On the other hand, if the number of divisions is too large, it becomes difficult to detect the thickness variations in the circumferential direction of the wafer itself. Therefore, it is preferable that the number of divisions in the double-sided polishing process be between 3 and 9, and most preferably 6.

[0054] Figure 4 shows the relationship between polishing time and MSE value for silicon wafers with no thickness variation in the circumferential direction and silicon wafers with thickness variation in the circumferential direction, when the number of divisions in the double-sided polishing process is set to 6. In Figure 4, the MSE value is for each time period (i.e., the value obtained for each time period).

[0055] As shown in Figure 4, for silicon wafers with no thickness variation in the circumferential direction, the MSE value is small at all time points. In contrast, for silicon wafers with no thickness variation in the circumferential direction, the MSE value increases towards the fourth time point and then decreases towards the sixth time point.

[0056] As shown in the embodiment described later, when the number of divisions is set to 6, it is preferable to determine that variation in the circumferential direction of the workpiece W occurs when the difference between the maximum value of MSE in the 6 time zones (i.e., the MSE value in the 4th time zone) and the maximum value of the MSE in the first time zone and the MSE in the last time zone of the double-sided polishing process (i.e., the MSE value in the 6th time zone) is greater than 0.10 μm. This allows for a good correlation with the variation in wafer thickness in the circumferential direction obtained by measuring the flatness after the double-sided polishing process.

[0057] In this way, variations in the circumferential thickness of the workpiece caused by double-sided polishing can be quickly detected. If variations in the circumferential thickness of the workpiece are detected, the auxiliary materials can be replaced or adjusted to address the issue. More specifically, a dressing treatment can be applied to the polishing pad 7.

[0058] Furthermore, when polishing multiple workpieces W on both sides in the first step, the acquired thickness data includes the thickness data of multiple workpieces W being polished simultaneously. As mentioned above, the thickness variation in the circumferential direction of the workpiece is caused by the obstruction of the wafer's rotation due to the condition of the polishing auxiliary materials (e.g., polishing pad). Therefore, it is not always necessary to identify the workpiece W in which the thickness variation in the circumferential direction occurred, but by identifying the workpiece W, the cause of the thickness variation in the circumferential direction can be investigated in more detail. For this reason, it is preferable to further include a third step between the first and second steps in which the thickness data of the workpieces W acquired in the first step is classified for each workpiece W.

[0059] The following describes in detail a method for measuring the thickness of a workpiece W and a method for classifying the acquired workpiece W thickness data for each workpiece W, using as an example a case in which there is one workpiece thickness measuring instrument 11 composed of an infrared laser, five workpiece thickness measuring holes 10 are provided on the upper platen 2 at equal intervals in the circumferential direction of the upper platen 2, there are five carrier plates 9, and one workpiece W is held on each carrier plate 9.

[0060] First, the thickness of the workpiece W is measured correctly by the workpiece thickness measuring instrument 11 when the laser beam emitted from the workpiece thickness measuring instrument 11 passes through the hole 10 in the upper platen 2 and shines on the surface of the workpiece W.

[0061] In contrast, if the laser beam does not pass through the hole 10 but is irradiated onto the upper surface of the upper platen 2, or if the laser beam passes through the hole 10 but is irradiated onto the surface of the carrier plate 9 instead of the surface of the workpiece W, the thickness of the workpiece W will not be obtained. Hereinafter, the time-continuous interval in which the thickness of the workpiece W is measured by the workpiece thickness measuring instrument 11 will be referred to as the "measurable interval," and the interval in which the thickness of the workpiece W was not measured correctly will be referred to as the "unmeasurable interval."

[0062] Even with data measured within the above-mentioned measurable interval, there may be significant variability in the measured data, making it impossible to accurately evaluate the shape of the workpiece W. In such cases, the shape of the workpiece W can be evaluated by averaging the data measured within the measurable interval for each of the 10 holes.

[0063] Specifically, as described above, the upper platen 2 has five holes 10 for thickness measurement. Therefore, when the upper platen 2 is rotated, for example, at 20 rpm (3-second cycle), the laser beam from the workpiece thickness measuring instrument 11 passes through the holes 10 at a cycle of 0.6 seconds. Also, if the time required to pass through the diameter of the hole 10 (for example, 15 mm) is 0.01 seconds, the time interval between the measurable section of one hole 10 and the next measurable section, i.e., the unmeasurable section, will be between 0.01 seconds and 0.59 seconds. Therefore, when the unmeasurable section becomes between 0.01 seconds and 0.59 seconds, the continuous data measured up to that point is treated as data continuously measured in one of the holes 10, averaged, and it is determined that the measurement has moved to the adjacent hole 10. Furthermore, even if a hole 10 passes directly beneath the workpiece thickness measuring instrument 11, it may become an unmeasurable section because the workpiece W is not present. Therefore, when moving from the currently measured hole 10 to the next hole 10, the time interval between the measurable interval and the next measurable interval, i.e., the unmeasurable interval, will be between 0.59 seconds and 1.19 seconds.

[0064] Furthermore, as mentioned above, even averaged data may contain outliers, for example, when the thickness of the outermost part of the workpiece is measured. When such outliers are present, it may not be possible to correctly evaluate the shape of the workpiece W. Therefore, it is preferable to first remove outliers from the measured thickness data.

[0065] The removal of the above-mentioned outliers can be performed based on the initial thickness of the carrier plate 9, the initial thickness of the workpiece W, etc. Furthermore, once a certain amount of workpiece thickness measurement data has been obtained, data exceeding a predetermined standard deviation (e.g., 0.2 μm) can be statistically removed as outliers. Hereinafter, the values ​​from which outliers have been removed will be referred to as "normal values." Figure 1(a) shows an example of workpiece W thickness data from which outliers have been removed.

[0066] When both sides of a workpiece W are polished under normal polishing conditions, a measurable thickness section appears, followed by an unmeasurable thickness section, and then another measurable thickness section appears again. This alternating pattern of measurable and unmeasurable thickness sections repeats. Here, the appearance of an unmeasurable thickness section indicates that the workpiece W being irradiated with the laser light has changed. Therefore, the appearance of such unmeasurable thickness sections can be used as an indicator to classify the thickness data measured within the measurable thickness section for each workpiece W.

[0067] Furthermore, as a result of the inventors' investigation, it was found that when the thickness of a workpiece W held on a carrier plate 9 is measured in a measurable section, and then an unmeasurable section appears, the workpiece W whose thickness is measured in the next measurable section is not necessarily held on an adjacent carrier plate 9, but may be held on two or more carrier plates 9 that are far apart.

[0068] Specifically, consider a case where carrier plates 9, each labeled A, B, C, D, and E, are arranged in a ring and revolve toward the workpiece thickness measuring instrument 11 in the order A, B, C, D, E, A, B…. When measuring the thickness of a workpiece W held on carrier plate 9 labeled A, a measurement impossible interval may occur. In the subsequent measurement possible interval, the workpiece W measured may be the workpiece W held on carrier plate 9 labeled C, which is two positions away. In this case, the measurement impossible interval is longer than when measuring a workpiece W on an adjacent carrier plate 9.

[0069] Therefore, based on the time interval of the unmeasurable interval, or in other words, the time interval between measurable intervals, it is possible to determine, for example, whether the thickness of the workpiece W of the carrier plate 9 with label B was measured after the workpiece W of the carrier plate 9 with label A, or whether the workpiece W of the carrier plate 9 with labels C or D was measured. In this way, the thickness data of the workpiece W can be correctly classified for each workpiece W. [Examples]

[0070] The following describes examples of the present invention, but the present invention is not limited to these examples.

[0071] Using the double-sided polishing apparatus 1 shown in Figures 2 and 3, a total of 35,000 silicon wafers were polished on both sides, with 5 wafers per batch. During the double-sided polishing of each silicon wafer, the thickness of the silicon wafer was measured in real time and thickness data was acquired. After the completion of the double-sided polishing process, the acquired thickness data was divided into six time periods, as shown in Figure 4, and the MSE value was calculated for each time period. Then, the difference between the maximum value of the MSE value in the six time periods and the maximum value of the MSE value in the first time period and the MSE value in the last time period of the double-sided polishing process was calculated.

[0072] Furthermore, for each silicon wafer after double-sided polishing, the GBIR (Global Backside Ideal Range) and GFLR (Global Front Least-squares Range) were determined using a surface inspection device (WaferSight2, manufactured by KLA-Tencor). GBIR indicates the overall flatness of the wafer, while GFLR indicates the overall flatness of the wafer after correcting for uneven wear, i.e., variations in the circumferential thickness of the wafer. Therefore, the GBIR-GFLR value can be used as an indicator of the variation in the circumferential thickness of the wafer.

[0073] Figure 5 shows the relationship between the difference in MSE values ​​and GBIR-GFLR. In the flatness measurement using the surface inspection device described above, silicon wafers with a GBIR-GFLR value exceeding 0.1 μm showed variations in thickness in the wafer circumferential direction. As is clear from Figure 5, when the difference in MSE values ​​exceeds 0.1 μm, the GBIR-GFLR value also exceeds 0.1 μm. Therefore, it is possible to determine whether or not variations in wafer circumferential thickness occur based on whether or not the difference in MSE values ​​exceeds 0.1 μm.

[0074] (Example of an invention) Using the double-sided polishing apparatus 1 (apparatus A) shown in Figures 2 and 3, a total of 4500 silicon wafers, 5 wafers per batch, were polished on both sides. During the double-sided polishing of each silicon wafer, the thickness of the silicon wafer was measured in real time and thickness data was acquired. After the completion of the double-sided polishing process, the acquired thickness data was divided into six time periods, as shown in Figure 4, and the MSE value was calculated for each time period. The difference between the maximum value of the MSE value in the six time periods and the maximum value of the MSE value in the first and last time periods of the double-sided polishing process was calculated, and whether or not a thickness variation in the circumferential direction of the wafer occurred was detected based on whether or not this difference exceeded 0.10 μm. If a thickness variation in the circumferential direction of the wafer was detected, double-sided polishing of the next batch was not performed, and the polishing pads 7 attached to the upper platen 2 and lower platen 3 of apparatus 1 were dressed. Then, double-sided polishing of subsequent batches was performed using the polishing pads 7 after the dressing process. The same double-sided polishing of the silicon wafers described above was performed using five different apparatuses 1 (apparatuses B to F), distinct from apparatus A. Approximately 5,000 silicon wafers were polished using each apparatus B to F.

[0075] (Conventional example) Similar to the inventive example, both sides of a silicon wafer were polished. However, instead of detecting variations in wafer thickness in the circumferential direction using thickness data acquired during the double-sided polishing process, flatness was measured using a surface inspection device after the double-sided polishing process to detect whether or not variations in wafer thickness in the circumferential direction had occurred. If variations in wafer thickness in the circumferential direction were detected, double-sided polishing of the next batch was not performed, and the polishing pads 7 attached to the upper platen 2 and lower platen 3 of the apparatus 1 were dressed. Then, double-sided polishing of subsequent batches was performed using the polishing pads 7 after the dressing process. All other conditions were the same as in the inventive example.

[0076] Figure 6 shows the incidence rate of wafer circumferential thickness variation for double-sided polishing apparatuses A to F. As is clear from Figure 6, the present invention can reduce the incidence rate of wafer circumferential thickness variation for all apparatuses. This is because, in the example invention, the occurrence of wafer circumferential thickness variation could be detected immediately after the double-sided polishing process and addressed promptly by dressing the polishing pad 7, whereas in the comparative example, the occurrence of wafer circumferential thickness variation could not be detected until flatness measurement was performed using a surface inspection apparatus, resulting in a delay in addressing the issue. [Industrial applicability]

[0077] According to the present invention, variations in the circumferential thickness of a workpiece caused by double-sided polishing can be quickly detected, improving productivity. This reduces the amount of resources required for product manufacturing, thereby reducing the environmental burden and improving industrial sustainability. [Explanation of Symbols]

[0078] 1. Double-sided polishing device 2 Upper surface plate 3 Lower surface plate 4 Rotating surface plate 5 Sun Gear 6 Internal Gear 7 Polishing pads 8 Workpiece holding holes 9. Carrier Plate 10 Holes for measuring workpiece thickness 11. Workpiece thickness measuring instrument 12 Control Unit 13 Arithmetic section Double job

Claims

1. In a method for polishing both sides of a workpiece, a workpiece is held in a carrier plate provided with one or more workpiece holding holes, the workpiece is sandwiched between a rotating platen consisting of an upper platen and a lower platen, and the rotation and revolution of the carrier plate are controlled by the rotation of a sun gear provided in the center of the rotating platen and the rotation of an internal gear provided on the outer circumference of the rotating platen, thereby causing the rotating platen and the carrier plate to rotate relative to each other and polish both sides of the workpiece simultaneously, The upper or lower platen has one or more workpiece thickness measuring holes that penetrate from the upper surface to the lower surface of the upper or lower platen, The method for polishing both sides of the workpiece is as follows: The first step involves performing double-sided polishing on the workpiece, and during double-sided polishing, measuring the thickness of the workpiece in real time using one or more workpiece thickness measuring instruments that can measure the thickness of the workpiece in real time through one or more workpiece thickness measuring holes, thereby acquiring workpiece thickness data. A second step of detecting the occurrence of circumferential thickness variation of the workpiece based on the variation in the thickness data of the workpiece at least in the middle stage of the double-sided polishing process, A method for detecting circumferential thickness variation of a workpiece, characterized by comprising the following:

2. The method for detecting circumferential thickness variation of a workpiece according to claim 1, wherein the detection of the occurrence of thickness variation in the second step is performed based on an increase in the thickness variation of the workpiece in the middle stage of the double-sided polishing step and a decrease in the thickness variation of the workpiece from the middle stage onward of the double-sided polishing step.

3. The method for detecting circumferential thickness variation of a workpiece according to claim 1 or 2, wherein the detection of the occurrence of thickness variation in the second step is performed based on the Mean Square Error (MSE) value, which is the value of the mean square error (MSE) between the obtained approximation curve and the thickness data of the workpiece, obtained by approximating the thickness data of the workpiece with a polynomial.

4. The method for detecting circumferential thickness variation of a workpiece according to claim 3, wherein the detection of the occurrence of thickness variation in the second step is performed by dividing the double-sided polishing step into multiple time periods, determining the mean squared error value for each time period, and based on the determined mean squared error value.

5. The method according to claim 4, wherein the number of divisions in the double-sided polishing process is set to six, and it is determined that variation occurs in the circumferential direction of the workpiece when the difference between the maximum value of the mean square error in the six time periods and the maximum value of the mean square error in the first time period and the mean square error in the last time period of the double-sided polishing process is greater than 0.10 μm.

6. The method for detecting circumferential thickness variation of a workpiece according to claim 1 or 2, further comprising a third step of classifying acquired workpiece thickness data for each workpiece between the first step and the second step, wherein the second step is performed for each workpiece.

7. The method for detecting circumferential thickness variation of a workpiece according to claim 1 or 2, wherein the workpiece is a silicon wafer.

Images