Polishing method, polishing monitoring method and polishing monitoring device for workpiece

Abstract

Provided is a polishing method, a workpiece polishing monitoring method, and a polishing monitoring device that can determine the thickness of a polishing target layer correctly without being affected by noise. The method is a polishing method for polishing a polishing target layer of a workpiece (W), rotates a polishing table (3) that supports a polishing pad (2), polishes the polishing target layer by pressing the workpiece (W) against the polishing pad (2), irradiates light toward the workpiece (W), receives reflected light from the workpiece (W), measures the intensity of the reflected light for each wavelength, generates a spectroscopic waveform that indicates the relationship between the intensity and the wavelength of the reflected light, performs Fourier transform processing on the spectroscopic waveform, thereby generating a frequency spectrum, moves the peak search range of the frequency spectrum according to the polishing time, determines the peak of the frequency spectrum within the peak search range, and thereby determines the thickness of the polishing target layer corresponding to the determined peak.

Description

Technical Field

[0001] This invention relates to a method for grinding workpieces such as crystals, substrates, and panels, and more particularly to a technique for determining the thickness of the grinding object layer based on the spectrum of reflected light from the workpiece.

[0002] In addition, the present invention relates to a method and apparatus for grinding workpieces such as wafers, substrates, and panels, and particularly to a technique for monitoring the grinding of workpieces based on optical information contained in reflected light from the workpiece. Background Technology

[0003] The manufacturing process of semiconductor devices includes various processes such as polishing insulating films like SiO2 and polishing metal films like copper and tungsten. In the manufacturing process of back-illuminated CMOS sensors, in addition to polishing the insulating film and metal film, it also includes polishing the silicon layer (silicon wafer).

[0004] The grinding of the grinding target layer (insulating film, metal film, silicon layer, etc.) constituting the exposed surface of the workpiece ends when the thickness of the grinding target layer reaches a predetermined target value. As an example of a method for measuring the thickness of the grinding target layer during the grinding process, there is an optical monitoring method that irradiates the surface of the workpiece, performs a Fourier transform on the spectrogram of the light reflected from the workpiece, and determines the thickness based on the peak of the resulting frequency spectrum (see, for example, Patent Document 1). The peak of the frequency spectrum varies depending on the thickness of the grinding target layer. Therefore, by tracking the peak of the frequency spectrum during the grinding process of the workpiece, the thickness of the grinding target layer can be monitored.

[0005] In the semiconductor device manufacturing process, various materials are repeatedly formed into films on a silicon wafer to create a layered structure. To form this layered structure, techniques for planarizing the surface of the top layer become important. Chemical mechanical polishing (CMP) is used as one method of such planarization.

[0006] Chemical mechanical polishing (CMP) is performed by a polishing apparatus. Generally, such an apparatus includes: a polishing table supporting a polishing pad, a polishing head holding a workpiece (e.g., a wafer with a film), and a polishing slurry supply nozzle for supplying polishing slurry (e.g., a slurry) to the polishing pad. When polishing the workpiece, polishing slurry is supplied to the polishing pad from the polishing slurry supply nozzle while the surface of the workpiece is pressed against the polishing pad by the polishing head. By rotating the polishing head and polishing table respectively, the workpiece and the polishing pad are moved relative to each other, thereby polishing the layers forming the surface of the workpiece.

[0007] To determine the thickness of the polished layer, such as an insulating film or silicon layer, polishing equipment typically includes an optical thickness measuring device. This optical thickness measuring device is configured to guide light emitted from a light source to the surface of the workpiece and determine the thickness of the polished layer by analyzing the spectrum of the reflected light from the workpiece.

[0008] Patent document 2 discloses a technique for determining film thickness based on the change in the spectrum of reflected light. Figure 21 This is a graph showing the relationship between the change in spectral density and grinding time. The change in spectral density is the change in the shape of the spectrum per unit time. The spectrum of reflected light varies depending on the thickness of the layer being ground. Therefore, the change in the spectrum of reflected light corresponds to the amount of layer removed per unit time.

[0009] Figure 22 This is a graph representing the cumulative change in the spectrum as a function of grinding time. Figure 22 It can be seen that the cumulative spectral change increases approximately monotonically with grinding time. Therefore, the amount of material removed from the layer being ground can be determined based on the cumulative spectral change.

[0010] Existing technical documents

[0011] Patent documents

[0012] Patent Document 1: Japanese Patent Application Publication No. 2013-110390

[0013] Patent Document 2: Japanese Patent Application Publication No. 2015-156503

[0014] The technical problem that the invention aims to solve

[0015] However, due to the grinding environment such as the slurry, and the substrate layer existing beneath the grinding object layer, spurious peaks may appear in the frequency spectrum. In conventional optical monitoring methods, these spurious peaks have been incorrectly tracked, resulting in the inability to determine the correct thickness.

[0016] Furthermore, conventional optical monitoring methods have failed to determine the thickness of the grinding object layer when the pseudo-peak caused by noise overlaps with the target peak corresponding to the thickness of the grinding object layer.

[0017] In addition, such as Figure 23 As shown, due to the influence of patterns formed on the surface of the workpiece, the grinding environment (e.g., slurry), etc., there are cases where local noise is applied to the amount of spectral variation. Therefore, as Figure 24 As shown, the cumulative change in the spectrum does not increase monotonically, which may lead to an incorrect estimate of the amount of abrasion on the polished layer. Summary of the Invention

[0018] Therefore, the object of the present invention is to provide a grinding method that can determine the correct thickness of the grinding target layer without being affected by noise.

[0019] Furthermore, the present invention provides a method and apparatus for accurately monitoring the grinding of a workpiece, eliminating the influence of patterns formed on the surface of the workpiece, the grinding environment (e.g., slurry), etc.

[0020] Technical means for solving technical problems

[0021] In one embodiment, a grinding method is provided for grinding a grinding target layer of a workpiece. The grinding method involves rotating a grinding table supporting a grinding pad, pressing the workpiece against the grinding pad to grind the grinding target layer, irradiating the workpiece with light, receiving reflected light from the workpiece, measuring the intensity of the reflected light at each wavelength, generating a spectroscopic waveform representing the relationship between the intensity and the wavelength of the reflected light, performing a Fourier transform on the spectroscopic waveform to generate a frequency spectrum, shifting the peak search range of the frequency spectrum according to the grinding time, determining the peak of the frequency spectrum within the peak search range, and determining the thickness of the grinding target layer corresponding to the determined peak.

[0022] In one approach, the peak search range is a range of values ​​calculated based on the thickness of the grinding target layer and the grinding rate of the workpiece, as determined previously.

[0023] In one approach, the grinding rate is a preset grinding rate.

[0024] In one approach, the grinding rate is a grinding rate calculated during the grinding process of the workpiece.

[0025] In one embodiment, a grinding method is provided for grinding a grinding target layer of a workpiece. The grinding method involves rotating a grinding table supporting a grinding pad, pressing the workpiece against the grinding pad to grind the grinding target layer, irradiating the workpiece with light, receiving reflected light from the workpiece, measuring the intensity of the reflected light at each wavelength, generating a spectroscopic waveform representing the relationship between the intensity and the wavelength of the reflected light, removing noise from the spectroscopic waveform using a filter, performing a Fourier transform on the noise-removed spectroscopic waveform to generate a frequency spectrum, determining the thickness of the grinding target layer based on the peaks of the frequency spectrum, and supplementing the thickness of the grinding target layer corresponding to the peaks of the frequency spectrum that disappear due to the noise removal by interpolating multiple values ​​of the thickness of the grinding target layer obtained during the grinding process of the workpiece.

[0026] In one embodiment, the filter is configured to remove noise from the spectral waveform that has a frequency peak that does not shift with the grinding time of the workpiece.

[0027] In one approach, the noise is noise caused by light reflected from the base layer of the polishing object layer.

[0028] In one embodiment, the filter is a band-stop filter.

[0029] In one embodiment, the grinding method grinds other workpieces having the same pattern as the workpiece before grinding the workpiece, and the grinding method further includes a step of creating the filter for removing noise contained in the spectral waveform of reflected light from the other workpieces.

[0030] In one embodiment, the grinding method further includes, during the grinding process of the workpiece, a step of creating the filter for removing noise contained in the spectral waveform of reflected light from the workpiece.

[0031] In one embodiment, a grinding apparatus is provided for grinding a grinding target layer of a workpiece. The grinding apparatus includes: a rotatable grinding table supporting a grinding pad; a grinding head pressing the workpiece onto the grinding pad on the grinding table; an optical sensor head irradiating light onto the workpiece held on the grinding head and receiving reflected light from the workpiece; a beam splitter measuring the intensity of the reflected light according to each wavelength; and a grinding control unit determining the thickness of the grinding target layer based on the intensity of the reflected light. The grinding control unit is configured to: generate a beam splitting waveform representing the relationship between the intensity and the wavelength of the reflected light; perform a Fourier transform on the beam splitting waveform to generate a frequency spectrum; shift the peak search range of the frequency spectrum according to grinding time; determine the peak of the frequency spectrum within the peak search range; and determine the thickness of the grinding target layer corresponding to the determined peak.

[0032] In one embodiment, a grinding apparatus is provided for grinding a grinding target layer of a workpiece. The grinding apparatus includes: a rotatable grinding table supporting a grinding pad; a grinding head pressing the workpiece against the grinding pad on the grinding table; an optical sensor head irradiating light onto the workpiece held on the grinding head and receiving reflected light from the workpiece; a beam splitter measuring the intensity of the reflected light according to each wavelength; and a grinding control unit determining the thickness of the grinding target layer based on the intensity of the reflected light. The grinding control unit is configured to: generate a beam splitting waveform representing the relationship between the intensity and the wavelength of the reflected light; remove noise from the beam splitting waveform using a filter; perform a Fourier transform on the noise-removed beam splitting waveform to generate a frequency spectrum; determine the thickness of the grinding target layer based on the peaks of the frequency spectrum; and supplement the thickness of the grinding target layer corresponding to the peaks of the frequency spectrum that disappear due to the noise removal by interpolating multiple values ​​of the thickness of the grinding target layer obtained during the grinding process of the workpiece.

[0033] In one approach, a computer-readable recording medium is provided, the program of which causes a computer to perform the following steps: sending a command to a grinding table motor to rotate a grinding table supporting a grinding pad; sending a command to a grinding head to press the workpiece against the grinding pad and grind the workpiece's grinding target layer; sending a command to a light source to irradiate light onto the workpiece; generating a spectroscopic waveform representing the relationship between the intensity of reflected light from the workpiece and the wavelength of the reflected light; performing a Fourier transform on the spectroscopic waveform to generate a frequency spectrum; shifting the peak search range of the frequency spectrum according to the grinding time; determining the peak of the frequency spectrum located within the peak search range; and determining the thickness of the grinding target layer corresponding to the determined peak.

[0034] In one approach, a computer-readable recording medium is provided, the program of which causes a computer to perform the following steps: sending a command to a grinding table motor to rotate a grinding table supporting a grinding pad; sending a command to a grinding head to press the workpiece against the grinding pad and grind the workpiece's grinding target layer; sending a command to a light source to irradiate light onto the workpiece; generating a spectroscopic waveform representing the relationship between the intensity of reflected light from the workpiece and the wavelength of the reflected light; removing noise from the spectroscopic waveform using a filter; performing a Fourier transform on the noise-removed spectroscopic waveform to generate a frequency spectrum; determining the thickness of the grinding target layer based on the peaks of the frequency spectrum; and supplementing the thickness of the grinding target layer corresponding to the peaks of the frequency spectrum that have disappeared due to the noise removal by using extrapolation of multiple values ​​of the film thickness of the grinding target layer obtained during the grinding process of the workpiece.

[0035] In one aspect, a method for monitoring the grinding of a workpiece is provided. This method involves irradiating the workpiece with light during the grinding process to generate a spectrum of reflected light from the workpiece, calculating the change in the spectrum at predetermined intervals, correcting the change in the spectrum when the change in the spectrum meets predetermined exclusion conditions, and calculating the cumulative change in the spectrum by accumulating the changes in the spectrum that do not meet the exclusion conditions and the corrected changes in the spectrum along the grinding time.

[0036] In one approach, the exclusion condition is any of the following: the change in the spectrum is greater than a threshold; the difference between the current change in the spectrum and the average of multiple changes in the spectrum obtained during the grinding process of the workpiece is greater than a threshold; the change in the spectrum is outside the range of the average of the normal distribution of multiple changes in the spectrum obtained in the past ± Zσ, where Z is a pre-defined coefficient; and the change in the spectrum is determined to be an outlier by the Smirnov-Grubbs test.

[0037] In one approach, the process of correcting the change in the spectrum is a process of correcting the change in the spectrum by extrapolation or interpolation.

[0038] In one approach, the correction by extrapolation uses multiple variations of the spectrum arranged along the grinding time, while the correction by interpolation uses multiple variations of the spectrum obtained at multiple measurement points arranged on the workpiece.

[0039] In one embodiment, a grinding monitoring device is provided, which is a workpiece grinding monitoring device, comprising: an optical sensor head that illuminates light onto the workpiece during the grinding process and receives reflected light from the workpiece; a spectrometer that measures the intensity of the reflected light at each wavelength; and a data processing unit having a storage device storing a program and an arithmetic unit that performs calculations according to commands contained in the program, the data processing unit being configured to: generate a spectrum of the reflected light based on the measured data of the intensity of the reflected light; calculate the change in the spectrum at each predetermined time interval; correct the change in the spectrum when the change in the spectrum meets a predetermined exclusion condition; and calculate a cumulative change in the spectrum by accumulating the changes in the spectrum that do not meet the exclusion condition and the corrected changes in the spectrum along the grinding time.

[0040] In one approach, the exclusion condition is any of the following: the change in the spectrum is greater than a threshold; the difference between the current change in the spectrum and the average of multiple changes in the spectrum obtained during the grinding process of the workpiece is greater than a threshold; the change in the spectrum is outside the range of the average of the normal distribution of multiple changes in the spectrum obtained in the past ± Zσ, where Z is a pre-defined coefficient; and the change in the spectrum is determined to be an outlier by the Smirnov-Grubbs test.

[0041] In one embodiment, the data processing unit is configured to correct the change in the spectrum by extrapolation or interpolation.

[0042] In one embodiment, the data processing unit is configured to perform corrections via interpolation using multiple variations of the spectrum arranged along the grinding time, and to perform corrections via interpolation using multiple variations of the spectrum obtained at multiple measurement points arranged on the workpiece.

[0043] The effects of the invention

[0044] According to the present invention, the grinding method can determine the correct thickness of the grinding target layer without tracking spurious peaks caused by noise by shifting the search range of the peaks of the frequency spectrum according to the grinding time.

[0045] Moreover, the grinding method can determine the correct thickness of the grinding layer by using a filter to remove noise from the beam waveform generated based on the reflected light from the workpiece.

[0046] Moreover, according to the present invention, the amount of spectral variation that satisfies the exclusion condition (i.e., includes noise) is corrected, enabling accurate monitoring of the grinding of the workpiece. Attached Figure Description

[0047] Figure 1This is a schematic diagram illustrating one embodiment of the grinding apparatus.

[0048] Figure 2 This is a schematic diagram illustrating the principle of an optical film thickness measuring device.

[0049] Figure 3 It is a top view showing the positional relationship between the workpiece and the grinding table.

[0050] Figure 4 This is a diagram showing an example of a beam splitting waveform generated by the grinding control unit.

[0051] Figure 5 This is a diagram showing an example of the frequency spectrum generated by the grinding control unit.

[0052] Figure 6 This is a graph illustrating the peak search range of the Nth measurement.

[0053] Figure 7 This is a graph illustrating the peak search range of the N+1th measurement.

[0054] Figure 8 This is a diagram illustrating how the peak search range shifts depending on the grinding time.

[0055] Figure 9 This is a flowchart illustrating an example of a process for determining the thickness of a layer to be ground by shifting the peak search range.

[0056] Figure 10 It is a graph showing the frequency spectrum before filter processing.

[0057] Figure 11 It is a graph showing the frequency spectrum after filter processing.

[0058] Figure 12 This is a diagram illustrating the situation where the extrapolation peaks disappear.

[0059] Figure 13 This is a flowchart illustrating an example of the process of using a filter to remove noise.

[0060] Figure 14 This is a schematic diagram illustrating one embodiment of the grinding apparatus.

[0061] Figure 15 This is a diagram showing an example of a spectrum generated by the data processing unit.

[0062] Figure 16 It is a graph representing the change in the spectrum per unit time.

[0063] Figure 17 It is a graph showing the change in spectral density along the grinding time during the grinding of the workpiece.

[0064] Figure 18 It is a graph representing the cumulative change in the spectrum obtained by accumulating the change in the spectrum along the grinding time.

[0065] Figure 19A It is a graph showing the amount of spectral variation, including noise, after it has been removed.

[0066] Figure 19B It is a graph that replaces the change in the spectrum that was lost due to removal with the change in the spectrum that does not contain noise.

[0067] Figure 20 It is a diagram representing an approximation generated based on multiple changes in the spectrum obtained during the grinding process of the workpiece.

[0068] Figure 21 It is a graph showing the relationship between the change in the spectrum and the grinding time.

[0069] Figure 22 It is a graph representing the cumulative change in the spectrum obtained by accumulating the change in the spectrum along the grinding time.

[0070] Figure 23 It is a graph showing the relationship between the change in the spectrum containing local noise and the grinding time.

[0071] Figure 24 It means to Figure 23 The graph shows the relationship between the cumulative change in the spectrum and the grinding time, obtained by accumulating the changes in the spectrum.

[0072] Symbol Explanation

[0073] 1 Grinding head

[0074] 2. Grinding pad

[0075] 2a Grinding surface

[0076] 3 Grinding table

[0077] 5 Grinding fluid supply nozzle

[0078] 6 Grinding table motor

[0079] 7. Optical sensor head

[0080] 10-head shaft

[0081] 31. Optical fiber cable for light transmission

[0082] 32 Optical fiber cable for receiving light

[0083] 40 Optical film thickness measuring device

[0084] 44 Light Source

[0085] 47 beam splitter

[0086] 49 Grinding Control Department

[0087] 449a Storage device

[0088] 49b Processing device

[0089] Peaks P1, P2, P3, and P4

[0090] Pf1, Pf2, Pf3, Pf4 pseudo-peaks

[0091] Pu lower peak

[0092] R1, R2, R3, R4 peak search range

[0093] 101 Grinding Head

[0094] 102 Grinding Pad

[0095] 102a Grinding surface

[0096] 103 Grinding Table

[0097] 105 Grinding fluid supply nozzle

[0098] 106 Grinding Table Motor

[0099] 107 Optical Sensor Head

[0100] 110 head shaft

[0101] 140 Optical Grinding Monitoring Device

[0102] 144 Light Source

[0103] 147 beam splitter

[0104] 149 Data Processing Department

[0105] 149a Storage device

[0106] 149b Computing Device Detailed Implementation

[0107] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0108] Figure 1 This is a schematic diagram illustrating one embodiment of the grinding apparatus. (As shown) Figure 1As shown, the polishing apparatus includes a polishing table 3 that supports the polishing pad 2, a polishing head 1 that presses the workpiece W with the polishing target layer onto the polishing pad 2, a polishing table motor 6 that rotates the polishing table 3, a polishing fluid supply nozzle 5 for supplying polishing fluid such as slurry to the polishing pad 2, and a polishing control unit 49 for controlling the operation of the polishing apparatus. The upper surface of the polishing pad 2 forms a polishing surface 2a for polishing the workpiece W. Examples of workpiece W include wafers, substrates, and panels.

[0109] The grinding head 1 is connected to the head shaft 10, which in turn is connected to a grinding head motor (not shown). The grinding head motor causes the grinding head 1 and the head shaft 10 to rotate together in the direction indicated by the arrow. The grinding table 3 is connected to a grinding table motor 6, which is configured to rotate the grinding table 3 and the grinding pad 2 in the direction indicated by the arrow. The grinding head 1, the grinding head motor, and the grinding table motor 6 are connected to the grinding control unit 49.

[0110] The workpiece W is ground as follows. While the grinding table 3 and the grinding head 1 are moving towards... Figure 1 The grinding head rotates in the direction indicated by the arrow, while the grinding fluid is supplied from the grinding fluid supply nozzle 5 to the grinding surface 2a of the grinding pad 2 on the grinding table 3. As the workpiece W rotates through the grinding head 1, and with the grinding fluid present on the grinding pad 2, the workpiece W is pressed against the grinding surface 2a of the grinding pad 2 by the grinding head 1. The surface of the workpiece W is ground by the chemical action of the grinding fluid, the abrasive particles contained in the grinding fluid, and the mechanical action of the grinding pad 2.

[0111] The grinding apparatus includes an optical film thickness measuring device 40 for measuring the thickness of the grinding target layer on the workpiece W. The optical film thickness measuring device 40 includes a light source 44, a beam splitter 47, and an optical sensor head 7 connected to the light source 44 and beam splitter 47. The light source 44 and beam splitter 47 are connected to a grinding control unit 49. The optical sensor head 7, the light source 44, and the beam splitter 47 are mounted on the grinding table 3 and rotate integrally with the grinding table 3 and the grinding pad 2. The position of the optical sensor head 7 is the position across the surface of the workpiece W on the grinding pad 2 each time the grinding table 3 and the grinding pad 2 rotate once.

[0112] Light emitted from light source 44 is transmitted to optical sensor head 7 and then shines onto the surface of workpiece W. The light is reflected from workpiece W, and the reflected light from workpiece W is received by optical sensor head 7 and sent to beam splitter 47. Beam splitter 47 decomposes the reflected light according to wavelength and measures the intensity of the reflected light at each wavelength. The intensity measurement data of the reflected light is sent to grinding control unit 49.

[0113] The grinding control unit 49 is configured to generate a spectrum of reflected light based on the intensity measurement data of the reflected light. The spectrum of reflected light is represented as a curve (i.e., a spectroscopic waveform) showing the relationship between the wavelength and intensity of the reflected light. The intensity of the reflected light can also be expressed as a relative value such as reflectance or relative reflectance.

[0114] Figure 2 This is a schematic diagram illustrating the principle of the optical film thickness measuring device 40. Figure 3 This is a top view showing the positional relationship between the workpiece W and the grinding table 3. Figure 2 In the example shown, the workpiece W has a base layer and a grinding target layer formed on the base layer. The grinding target layer is, for example, a silicon layer or an insulating film.

[0115] The optical sensor head 7 is composed of the tips of a light-emitting optical fiber cable 31 and a light-receiving optical fiber cable 32, and is disposed opposite to the surface of the workpiece W. Each time the grinding table 3 rotates once, the optical sensor head 7 illuminates the workpiece W and receives the reflected light from the workpiece W.

[0116] The light illuminating the workpiece W in the medium ( Figure 2 In the example, the light reflected from the interface between water and the workpiece layer, and from the interface between the workpiece layer and the substrate layer, interferes with each other. The manner of this light interference varies depending on the thickness of the workpiece layer (i.e., the optical path length). Therefore, the spectrum generated from the reflected light from the workpiece W varies depending on the thickness of the workpiece layer. The polishing control unit 49 performs Fourier transform processing (or high-speed Fourier transform processing) on ​​the split waveform to generate a frequency spectrum, and determines the thickness of the workpiece layer based on the peaks of the frequency spectrum. In cases where the workpiece layer is a silicon layer and... Figure 2 In the case where the medium is water, as shown, it is preferable to use light with a wavelength of less than 1100 nm to prevent light from being absorbed by the water.

[0117] During the grinding process of workpiece W, the optical sensor head 7 moves lateral to workpiece W for each revolution of the grinding table 3. When the optical sensor head 7 is below workpiece W, the light source 44 emits light. The light is transmitted through the projection fiber optic cable 31 and shines from the optical sensor head 7 onto the surface of workpiece W (the surface to be ground). The reflected light from workpiece W is received by the optical sensor head 7 and sent to the beam splitter 47 through the light receiving fiber optic cable 32. The beam splitter 47 measures the intensity of the reflected light at each wavelength across a specified wavelength range and sends the intensity measurement data to the grinding control unit 49. The grinding control unit 49 generates a spectrum (i.e., a beam splitting waveform) of the reflected light representing the intensity of light at each wavelength based on the intensity measurement data.

[0118] Figure 4 This is a diagram showing an example of the beam-splitting waveform generated by the polishing control unit 49. Figure 4In the diagram, the horizontal axis represents the wavelength of the reflected light from the workpiece W, and the vertical axis represents the relative reflectivity derived from the intensity of the reflected light. Relative reflectivity is an indicator of the intensity of reflected light, and is the ratio of the light intensity to a specified reference intensity. By dividing the intensity (measured intensity) of light at each wavelength by the specified reference intensity, unwanted noise such as fluctuations in the optical system of the device and the inherent intensity of the light source can be removed from the measured intensity.

[0119] The reference intensity is the intensity of light measured in advance for each wavelength, and the relative reflectance is calculated for each wavelength. Specifically, the relative reflectance is obtained by dividing the intensity (measured intensity) of light at each wavelength by the corresponding reference intensity. For example, the reference intensity can be obtained by directly measuring the intensity of light emitted from the optical sensor head 7, or by illuminating light from the optical sensor head 7 onto a mirror and measuring the intensity of the reflected light from the mirror. Alternatively, the reference intensity can also be the intensity of reflected light from the silicon substrate measured by the beam splitter 47 when the uncoated silicon substrate (bare substrate) is water-polished on the polishing pad 2, or when the aforementioned silicon substrate (bare substrate) is placed on the polishing pad 2.

[0120] In actual grinding, the corrected measured intensity is obtained by subtracting the dark intensity (the background intensity obtained under shading conditions) from the measured intensity. Then, the corrected reference intensity is obtained by subtracting the dark intensity from the reference intensity. Finally, the relative reflectance is obtained by dividing the corrected measured intensity by the corrected reference intensity. Specifically, the relative reflectance R(λ) can be obtained by the following formula (1).

[0121] (Equation 1)

[0122]

[0123] Here, λ is the wavelength of the light reflected from the workpiece W, E(λ) is the intensity at wavelength λ, B(λ) is the reference intensity at wavelength λ, and D(λ) is the background intensity (dark intensity) at wavelength λ measured under shading conditions.

[0124] The grinding control unit 49 sends a command to the light source 44 to emit light every time the grinding table 3 rotates one revolution. The optical sensor head 7, optically connected to the light source 44, illuminates the surface of the workpiece W (the surface to be ground), and receives the reflected light from the workpiece W. The reflected light is sent to a beam splitter 47, optically connected to the optical sensor head 7. The beam splitter 47 decomposes the reflected light according to wavelength and measures the intensity of the reflected light at each wavelength. The intensity measurement data of the reflected light is sent to the grinding control unit 49, which generates a generator based on the intensity measurement data. Figure 4 The spectrum shown. Figure 4In the example shown, the spectrum of the reflected light is a spectroscopic waveform that represents the relationship between the relative reflectivity and the wavelength of the reflected light. However, the spectrum of the reflected light can also be a spectroscopic waveform that represents the relationship between the intensity of the reflected light itself and the wavelength of the reflected light.

[0125] The polishing control unit 49 performs Fourier transform processing (typically high-speed Fourier transform processing) on ​​the obtained spectroscopic waveform to analyze it. More specifically, the polishing control unit 49 performs Fourier transform processing (or high-speed Fourier transform processing) on ​​the obtained spectroscopic waveform to extract the frequency components and their intensities contained in the spectroscopic waveform, and uses a prescribed formula to convert the obtained frequency components into the thickness of the polishing target layer. Then, it generates a frequency spectrum representing the relationship between the thickness of the polishing target layer and the intensity of the frequency components. The prescribed formula is a function representing the thickness of the polishing target layer with the frequency components as variables, and can be obtained from measured results, optical film thickness measurement simulations, theoretical formulas, etc.

[0126] Figure 5 This is a graph showing the frequency spectrum generated by the grinding control unit 49. In Figure 5 In the diagram, the vertical axis represents the intensity of the frequency components contained in the split waveform, and the horizontal axis represents the thickness of the grinding object layer. Figure 5 It can be seen that the frequency spectrum has a peak at thickness t1. That is, this frequency spectrum represents the thickness of the material being polished as t1. Therefore, the thickness of the material being polished can be determined based on the peak of the frequency spectrum.

[0127] The polishing control unit 49 includes: a storage device 49a storing a program for determining the thickness of the polishing target layer; and a processing device 49b performing calculations according to the instructions contained in the program. The polishing control unit 49 is configured as at least one computer. The storage device 49a includes a main storage device such as RAM and an auxiliary storage device such as a hard disk drive (HDD) or a solid-state drive (SSD). Examples of the processing device 49b include a CPU (central processing unit) and a GPU (graphics processing unit). However, the specific structure of the polishing control unit 49 is not limited to these examples.

[0128] The grinding control unit 49 determines the grinding endpoint based on the determined thickness of the grinding target layer and controls the operation of the grinding apparatus. For example, the grinding control unit 49 determines the grinding endpoint as the time point at which the determined thickness of the grinding target layer reaches a target value. In one embodiment, the grinding endpoint may also be determined by measuring the combined thickness of the grinding target layer and the base layer. The grinding control unit for determining the thickness of the grinding target layer and the control unit for controlling the grinding operation of the workpiece W may also be configured separately.

[0129] Figure 6 This is a graph illustrating the peak search range R1 for the Nth measurement. Figure 6 In the diagram, the vertical axis represents the intensity of the frequency components contained in the beam splitting waveform, and the horizontal axis represents the thickness of the grinding object layer. Figure 6 Each of the multiple frequency spectra shown is generated by the grinding control unit 49 during the grinding process of workpiece W, each rotation of the grinding table 3. Peak P1 is the peak that appears in the Nth measurement (when the grinding table 3 rotates for the Nth time), peak P2 is the peak that appears in the (N+1)th measurement (when the grinding table 3 rotates for the N+1th time), peak P3 is the peak that appears in the (N+2)th measurement (when the grinding table 3 rotates for the N+2th time), and peak P4 is the peak that appears in the (N+3)th measurement (when the grinding table 3 rotates for the N+3rd time). Figure 6 In this representation, peaks with lower frequency component intensities are omitted. The symbol N represents a natural number, such as 1.

[0130] like Figure 6 As shown, the peaks of the frequency spectrum shift with the progress of grinding workpiece W. During one revolution of the grinding table 3, the grinding target layer of workpiece W is also ground. Therefore, the thickness of the grinding target layer decreases with each revolution of the grinding table 3. The relationship between the thickness t1 corresponding to peak P1, the thickness t2 corresponding to peak P2, the thickness t3 corresponding to peak P3, and the thickness t4 corresponding to peak P4 is t1>t2>t3>t4.

[0131] Each rotation of the grinding table 3 determines the thickness of the layer to be ground by the optical film thickness measuring device 40 based on the peaks of the frequency spectrum. However, due to spurious peaks generated by noise during the grinding process of the workpiece W (noise from the grinding environment caused by the slurry, noise from the substrate layer, etc.), the grinding control unit 49 may incorrectly determine the thickness of the layer to be ground. Figure 6 In this case, the spurious peak Pf1 is a noise peak generated during the grinding process of workpiece W, and it is the spurious peak that appears in the Nth measurement. In this case, in the Nth measurement, the thickness tf1 corresponding to the spurious peak Pf1, which has a greater intensity than peak P1, is incorrectly determined as the thickness of the grinding object layer.

[0132] Therefore, the grinding control unit 49 is configured to search for the peak of the frequency spectrum within the peak search range R1. Figure 6 In the example shown, the polishing control unit 49 determines that peak P1 is within the peak search range R1, and determines the thickness t1 (i.e., the correct thickness) corresponding to the determined peak P1 as the thickness of the polishing target layer. At this time, since the spurious peak Pf1 is not within the peak search range R1, the spurious peak Pf1 will not be incorrectly determined.

[0133] exist Figure 6In this context, the peak search range R1 represents the peak search range measured in the Nth measurement. In one embodiment, the grinding control unit 49 determines the peak search range R1 based on the initial thickness value of the grinding target layer. For example, the peak search range R1 may also be a range including the initial thickness value of the grinding target layer of the workpiece W. The initial thickness of the grinding target layer is the thickness of the grinding target layer before the grinding of the workpiece W begins, which is pre-measured by an independent film thickness measuring device (not shown) or given as basic information of the workpiece W.

[0134] Figure 7 This is a graph illustrating the peak search range R2 of the (N+1)th measurement. Figure 7 In the middle, peaks P1 to P4 are similar to the reference peaks. Figure 6 The peaks described are the same. The spurious peak Pf2 represents the peak caused by noise generated during the grinding process of workpiece W, and it is a spurious peak that appears during the N+1th measurement. Since this spurious peak Pf2 is within the peak search range R1 of the previous Nth measurement, the grinding control unit 49 incorrectly determines the thickness tf2 corresponding to the spurious peak Pf2 as the thickness of the grinding target layer.

[0135] Therefore, in this embodiment, the peak search range of the frequency spectrum is configured to shift according to the grinding time. As described above, as the workpiece W is ground, the thickness of the grinding target layer decreases, i.e., the peak of the frequency spectrum shifts. By shifting the peak search range to follow the change in the thickness of the grinding target layer, the grinding control unit 49 can accurately determine the thickness of the grinding target layer even if spurious peaks exist.

[0136] exist Figure 7 In the diagram, the peak search range R2 represents the peak search range for the (N+1)th measurement. The grinding control unit 49 moves the peak search range to follow the thickness of the grinding target layer, which varies according to the grinding time. In the (N+1)th measurement, the peak search range is moved from... Figure 6 The peak search range R1 is moved to the peak search range R2. For example, as described below, the peak search range R2 can also be calculated based on the previously determined thickness t1 of the grinding object layer and the grinding rate of the workpiece W.

[0137] The grinding control unit 49 determines the peak P2 located within the peak search range R2, and sets t2 (i.e., the correct thickness) corresponding to the determined peak P2 as the thickness of the grinding target layer. At this time, since the spurious peak Pf2 is not within the peak search range R2, the spurious peak Pf2 will not be incorrectly determined. Similarly, in measurements after the N+2th time, by shifting the peak search range with the grinding time, the correct thickness of the grinding target layer can also be determined.

[0138] Figure 8This is a graph illustrating how the peak search range shifts based on the grinding time. Figure 8 In the middle, peaks P1~P4, pseudo-peaks Pf1, Pf2, and peak search ranges R1, R2 and Figure 6 , Figure 7 The same symbols are shown. Figure 8 The graph shows the relationship between grinding time and the thickness corresponding to each peak. Pseudo-peak Pf3 represents the pseudo-peak that appeared in the (N+2)th measurement. Pseudo-peak Pf4 represents the pseudo-peak that appeared in the (N+3)th measurement. Peak search range R3 represents the peak search range that shifted during the (N+2)th measurement. Peak search range R4 represents the peak search range that shifted during the (N+3)th measurement.

[0139] exist Figure 8 In this context, the initial value t0 represents the thickness of the grinding target layer of the workpiece W before grinding. The initial value t0 is the initial thickness of the grinding target layer that is pre-measured before grinding the workpiece W, or a value given by the user. In one embodiment, the peak search range R1 at the Nth measurement is set based on the initial value t0. Specifically, the peak search range R1 can be set using the following formula (2).

[0140] The peak search range R1 = initial value t0 ± first set value X (2)

[0141] Here, the first setting value X can be freely set by the user.

[0142] In one embodiment, the peak search range R2 is set to include a range of values ​​calculated based on the thickness t1 of the previously determined grinding target layer and the grinding rate PR of the workpiece W. Specifically, the peak search range R2 can be set using the following formula (3).

[0143] The peak search range R2 = (the previously determined thickness t1 - (the time interval dT of the grinding rate PRx)) ± the second set value Y (3)

[0144] Here, the grinding rate is also called the removal rate. The time interval dT is equivalent to the time required for the grinding table 3 to rotate one revolution. Typically, the time interval dT is constant during the grinding process of the workpiece W. The grinding rate PR can also be a preset grinding rate. The second setting value Y can be freely set by the user. The second setting value Y can be smaller than the first setting value X, or it can be the same.

[0145] The peak search ranges R3 and R4 are the same as the peak search range R2, and are calculated using equation (3). In this case, the grinding rate PR in equation (3) can be either a pre-set grinding rate or a grinding rate calculated during the grinding process of the workpiece W. The grinding rate can be calculated based on multiple values ​​of the thickness of the grinding object layer obtained during the grinding process of the workpiece W and the time required to obtain these multiple values.

[0146] Figure 9 This is a flowchart illustrating an example of a process for determining the thickness of a layer to be ground by shifting the peak search range.

[0147] In step S101, the grinding table motor 6 causes the grinding table 3 supporting the grinding pad 2 to rotate.

[0148] In step S102, the grinding head 1 presses the workpiece W onto the grinding surface 2a of the grinding pad 2 and begins grinding the workpiece W. At this time, as shown in reference... Figure 1 As explained, while the workpiece W rotates through the grinding head 1, and with the grinding fluid present on the grinding pad 2, the workpiece W is pressed by the grinding head 1 onto the grinding surface 2a of the grinding pad 2.

[0149] Next, during the grinding process of workpiece W, the thickness of the grinding object layer is measured by an optical film thickness measuring device 40.

[0150] In step S103, the light source 44 emits light, illuminating the surface of the workpiece W from the optical sensor head 7.

[0151] In step S104, the optical sensor head 7 receives reflected light from the workpiece W.

[0152] In step S105, the beam splitter 47 measures the intensity of the reflected light from the workpiece W according to each wavelength.

[0153] In step S106, the grinding control unit 49 generates a beam splitting waveform based on the intensity measurement data of the reflected light.

[0154] In step S107, the grinding control unit 49 performs Fourier transform processing on the spectroscopic waveform to generate a frequency spectrum.

[0155] In step S108, the grinding control unit 49 moves the peak search range according to the grinding time. (This is achieved by referring to...) Figure 8 The method described calculates the peak search range and moves the peak search range accordingly.

[0156] In step S109, the grinding control unit 49 determines the peak of the frequency spectrum that is within the peak search range.

[0157] In step S110, the polishing control unit 49 determines the thickness of the polishing target layer corresponding to the determined wave peak.

[0158] The grinding control unit 49 operates according to commands contained in a program stored in the storage device 49a. Specifically, the grinding control unit 49 performs the following steps: sending a command to the grinding table motor 6 to rotate the grinding table 3 supporting the grinding pad 2 (see step S101); sending a command to the grinding head 1 to begin grinding the workpiece W (see step S102); sending a command to the light source 44 to irradiate light onto the workpiece W (see step S103); generating a spectroscopic waveform based on intensity measurement data of reflected light from the workpiece W (see step S106); performing a Fourier transform on the spectroscopic waveform to generate a frequency spectrum (see step S107); shifting the peak search range according to the grinding time (see step S108); determining the peak of the frequency spectrum located within the peak search range (see step S109); and determining the thickness of the grinding target layer corresponding to the determined peak (see step S110).

[0159] The program used to cause the polishing control unit 49 to perform these steps is recorded on a computer-readable recording medium that is a non-transitory tangible object, and is provided to the polishing control unit 49 via the recording medium. Alternatively, the program may also be input to the polishing control unit 49 via a communication network such as the Internet or a local area network.

[0160] Next, an implementation method for accurately determining the thickness of the grinding object layer by using a filter to remove noise will be described.

[0161] Figure 10 This is a graph showing the frequency spectrum before filtering. Figure 10 In the diagram, the vertical axis represents the intensity of the frequency components contained in the beam splitting waveform, and the horizontal axis represents the thickness of the grinding object layer. Figure 10 Each of the multiple frequency spectra shown is generated by the grinding control unit 49 during the grinding process of workpiece W, each rotation of the grinding table 3. Peak P1 is the peak that appears in the Nth measurement (when the grinding table 3 rotates for the Nth time), peak P2 is the peak that appears in the (N+1)th measurement (when the grinding table 3 rotates for the N+1th time), peak P3 is the peak that appears in the (N+2)th measurement (when the grinding table 3 rotates for the N+2th time), and peak P4 is the peak that appears in the (N+3)th measurement (when the grinding table 3 rotates for the N+3rd time). Figure 10 In this representation, peaks with lower frequency component intensities are omitted. The symbol N represents a natural number, such as 1.

[0162] The thickness of the workpiece layer is determined by an optical film thickness measuring device 40 based on the peak of the frequency component per revolution of the polishing table 3. However, due to the influence of noise from the substrate layer beneath the workpiece layer, the thickness of the workpiece layer may not be accurately determined. Figure 10In the grinding process, the lower peak Pu is a peak formed by reflected light from the substrate layer of the workpiece being ground, and it represents noise unnecessary for determining the thickness of the workpiece layer. During the grinding of workpiece W, since the substrate layer is not ground, the position of the lower peak Pu remains unchanged even as the workpiece layer is being ground. Figure 10 In the example shown, since the lower peak Pu appears at the same position as peak P3, the polishing control unit 49 cannot accurately determine the thickness of the polishing target layer corresponding to peak P3.

[0163] Therefore, in this embodiment, the grinding control unit 49 removes noise caused by the substrate layer from the spectral waveform of the reflected light from the workpiece W by using a filter, and further supplements the thickness of the grinding target layer corresponding to the peak of the frequency spectrum that disappears due to the removal of noise by interpolation, thereby determining the thickness of the grinding target layer.

[0164] A filter for removing noise from the spectroscopic waveform of reflected light is pre-made as follows: Before grinding workpiece W, another workpiece W' with the same pattern as workpiece W is ground. During the grinding of the other workpiece W', the thickness of the grinding target layer is measured by an optical film thickness measuring device 40. The grinding control unit 49 determines the lower peak of the noise based on the frequency spectrum generated from the spectroscopic waveform of the reflected light from the other workpiece W', thereby creating a filter for removing the specific noise. Specifically, the grinding control unit 49 determines the peak of the frequency spectrum that does not shift with the grinding time of the other workpiece W', thereby creating a filter that removes the component (noise) with the frequency of the specific peak from the spectroscopic waveform of the reflected light. This filter is a digital filter, for example, a band-stop filter.

[0165] In one embodiment, instead of using other workpieces W' with the same pattern as workpiece W, the grinding control unit 49 can create a filter by specifically targeting the lower peak of noise during the initial stage of grinding workpiece W. Specifically, during the grinding process of workpiece W, the grinding control unit 49 targets the peak of the frequency spectrum that does not shift with the grinding time, thereby creating a filter that removes the component (noise) with the frequency of the specific peak from the spectral waveform of the reflected light.

[0166] Figure 11 This is a graph showing the frequency spectrum after filtering. The grinding control unit 49 applies a pre-made filter to the beam waveform of the reflected light from the workpiece W, thereby removing noise. Figure 11The multiple frequency spectra shown are frequency spectra generated by the polishing control unit 49 through Fourier transform processing (or high-speed Fourier transform processing) of the noise-removed beam splitting waveform. By removing noise from the beam splitting waveform of the reflected light, peak P3, which is located at a position that overlaps with the lower peak Pu, also disappears along with the lower peak Pu. Therefore, the polishing control unit 49 cannot accurately determine the thickness of the polishing target layer corresponding to peak P3.

[0167] Therefore, the polishing control unit 49 is configured to supplement the thickness of the polishing target layer corresponding to the wave peak P3 by interpolation. Figure 12 This is a graph illustrating the situation where the extrapolation peak P3 disappears. In Figure 12 In the middle, peaks P1 to P4 and Figure 10 , Figure 11 The same symbols shown correspond to, Figure 12 The chart shown illustrates the relationship between grinding time and the thickness corresponding to each peak. (See reference...) Figure 11 As explained, when peak P3 disappears due to the removal of the lower peak Pu, which serves as noise, the polishing control unit 49 supplements the thickness of the polishing target layer corresponding to the portion where peak P3 disappears by interpolating the thickness of the polishing target layer corresponding to peaks P1 and P2. Figures 10 to 12 Although the disappearance of peak P3 is explained in the text, it is not limited to this. As long as multiple values ​​of the thickness of the grinding object layer are obtained during the grinding process of workpiece W, the thickness of the grinding object layer can be supplemented by interpolation using the obtained multiple values.

[0168] In the above embodiments, although the wave peak (noise) that does not change with the polishing time originates from the base layer of the polishing target layer, the present invention is not limited thereto. For example, the wave peak (noise) that does not change with the polishing time may also be the inherent noise of devices such as the light source 44 or the beam splitter 47.

[0169] Figure 13 This is a flowchart illustrating an example of the process of using a filter to remove noise.

[0170] In step S201, the grinding device grinds other workpieces W' that have the same pattern as the workpiece W to be ground.

[0171] In step S202, during the grinding process of other workpieces W', the grinding control unit 49 measures the thickness of the grinding target layer using an optical film thickness measuring device 40. The grinding control unit 49 generates a frequency spectrum by performing a Fourier transform on the spectroscopic waveform obtained from the reflected light from the other workpieces W'. The processes of steps S201 to S202 are similar to... Figure 9 The steps S101 to S107 are the same.

[0172] In step S203, the grinding control unit 49 specifies the peak of the frequency spectrum that does not change with the grinding time.

[0173] In step S204, the grinding control unit 49 creates a filter that removes noise with a specific peak frequency from the beam waveform of the reflected light from other workpieces W'.

[0174] In step S205, the grinding device begins grinding the workpiece W to be ground.

[0175] In step S206, during the grinding process of workpiece W, the thickness of the grinding target layer is measured by an optical film thickness measuring device 40. The grinding control unit 49 uses the filter created in step S204 above to remove noise from the beam splitting waveform of the reflected light from workpiece W. The processes from step S205 to S206 up to the generation of the beam splitting waveform are similar to... Figure 9 The steps S101 to S106 are the same.

[0176] In step S207, the grinding control unit 49 performs Fourier transform processing on the noise-removed spectroscopic waveform to generate a frequency spectrum.

[0177] In step S208, the polishing control unit 49 determines the thickness of the polishing target layer based on the peak of the frequency spectrum.

[0178] In step S209, the polishing control unit 49 supplements the thickness of the polishing target layer corresponding to the peak of the frequency spectrum that has disappeared due to noise removal by interpolation.

[0179] The grinding control unit 49 operates according to commands contained in a program stored in the storage device 49a. Specifically, the grinding control unit 49 performs the following steps: sending a command to the grinding table motor 6 to rotate the grinding table 3 supporting the grinding pad 2; sending a command to the grinding head 1 to begin grinding the workpiece W (see step S205); sending a command to the light source 44 to irradiate light onto the workpiece W; generating a beam splitting waveform based on intensity measurement data of the reflected light from the workpiece W; removing noise from the beam splitting waveform using a filter (see step S206); performing a Fourier transform on the noise-removed beam splitting waveform to generate a frequency spectrum (see step S207); determining the thickness of the grinding target layer based on the peaks of the frequency spectrum (see step S208); and supplementing the thickness of the grinding target layer corresponding to the peaks of the frequency spectrum that disappeared due to noise removal by interpolation (see step S209).

[0180] The program used to cause the polishing control unit 49 to perform these steps is recorded on a computer-readable recording medium that is a non-transitory tangible object, and is provided to the polishing control unit 49 via the recording medium. Alternatively, the program may also be input to the polishing control unit 49 via a communication network such as the Internet or a local area network.

[0181] Figures 6 to 9 The described implementation methods and Figures 10 to 13 The described implementation methods can also be combined. That is, a filter can be used to remove noise from the spectroscopic waveform to generate a frequency spectrum, and the peak search range of the frequency spectrum can be shifted to determine the thickness of the grinding object layer.

[0182] Figure 14 This is a schematic diagram illustrating one embodiment of the grinding apparatus. (As shown) Figure 14 As shown, the polishing apparatus includes a polishing table 103 supporting a polishing pad 102, a polishing head 101 pressing a workpiece W, such as a wafer, substrate, or panel having a polishing layer, onto the polishing pad 102, a polishing table motor 106 rotating the polishing table 103, and a polishing slurry supply nozzle 105 for supplying polishing slurry or other polishing fluid to the polishing pad 102. The upper surface of the polishing pad 102 forms a polishing surface 102a for polishing the workpiece W.

[0183] The grinding head 101 is connected to the head shaft 110, which is connected to a grinding head motor (not shown). The grinding head motor causes the grinding head 101 and the head shaft 110 to rotate together in the direction shown by the arrow. The grinding table 103 is connected to a grinding table motor 106, which is configured to rotate the grinding table 103 and the grinding pad 102 in the direction shown by the arrow.

[0184] The workpiece W is ground as follows. While the grinding table 103 and the grinding head 101 are moving towards... Figure 14 The grinding head rotates in the direction indicated by the arrow, supplying grinding fluid from the grinding fluid supply nozzle 105 to the grinding surface 102a of the grinding pad 102 on the grinding table 103. As the workpiece W rotates through the grinding head 101, it is pressed against the grinding surface 102a of the grinding pad 102 by the grinding head 101 while the grinding fluid is present on the grinding pad 102. The exposed surface of the workpiece W is ground by the chemical action of the grinding fluid and the mechanical action of the abrasive particles contained in the grinding fluid and the grinding pad 102. Examples of the ground surface of the workpiece W include insulating films, silicon layers, etc., but are not limited to these.

[0185] The grinding apparatus includes an optical grinding monitoring device 140 for monitoring the grinding of workpiece W. The optical grinding monitoring device 140 includes a light source 144, a beam splitter 147, an optical sensor head 107 optically connected to the light source 144 and the beam splitter 147, and a data processing unit 149 connected to the beam splitter 147. The optical sensor head 107, the light source 144, and the beam splitter 147 are mounted on the grinding table 103 and rotate integrally with the grinding table 103 and the grinding pad 102. The position of the optical sensor head 107 is the position of the workpiece W across the grinding pad 102 during each rotation of the grinding table 103 and the grinding pad 102.

[0186] The data processing unit 149 includes: a storage device 149a storing a program for performing the generation of spectra and monitoring of the grinding of workpiece W (described later); and a processing unit 149b performing calculations according to the instructions contained in the program. The data processing unit 149 is configured as at least one computer. The storage device 149a includes a main storage device such as RAM and an auxiliary storage device such as a hard disk drive (HDD) or a solid-state drive (SSD). Examples of the processing unit 149b include a CPU (central processing unit) and a GPU (graphics processing unit). However, the specific structure of the data processing unit 149 is not limited to these examples.

[0187] Light emitted from light source 144 is transmitted to optical sensor head 107 and guided from optical sensor head 107 to workpiece W. The light is reflected at workpiece W, and the reflected light from workpiece W is received by optical sensor head 107 and sent to beam splitter 147. Beam splitter 147 decomposes the reflected light according to wavelength and measures the intensity of the reflected light at each wavelength. The intensity measurement data of the reflected light is sent to data processing unit 149.

[0188] The data processing unit 149 is configured to generate a spectrum of reflected light based on the intensity measurement data of the reflected light. The spectrum of the reflected light is represented as a graph (i.e., a spectroscopic waveform) showing the relationship between the wavelength and intensity of the reflected light. The intensity of the reflected light can also be expressed as a relative value such as reflectance or relative reflectance.

[0189] Figure 15 This is a graph showing an example of a spectrum generated by the data processing unit 149. The spectrum is represented as a curve (i.e., a spectroscopic waveform) showing the relationship between the wavelength and intensity of light. Figure 15 In the diagram, the horizontal axis represents the wavelength of the reflected light from the workpiece W, and the vertical axis represents the relative reflectivity derived from the intensity of the reflected light. Relative reflectivity is an indicator of the intensity of reflected light, and is the ratio of the light intensity to a specified reference intensity. By dividing the intensity (measured intensity) of light at each wavelength by the specified reference intensity, unwanted noise such as fluctuations in the optical system of the device and the inherent intensity of the light source can be removed from the measured intensity.

[0190] The reference intensity is the intensity of light measured in advance for each wavelength, and the relative reflectance is calculated for each wavelength. Specifically, the relative reflectance is obtained by dividing the intensity (measured intensity) of light at each wavelength by the corresponding reference intensity. For example, the reference intensity can be obtained by directly measuring the intensity of light emitted from the optical sensor head 107, or by illuminating light from the optical sensor head 107 onto a mirror and measuring the intensity of the reflected light from the mirror. Alternatively, the reference intensity can also be the intensity of reflected light from the silicon substrate measured by the beam splitter 147 when the uncoated silicon substrate (bare substrate) is water-polished on the polishing pad 102 in the presence of water, or when the aforementioned silicon substrate (bare substrate) is placed on the polishing pad 102.

[0191] In actual grinding, the corrected measured intensity is obtained by subtracting the dark intensity (the background intensity obtained under shading conditions) from the measured intensity. Then, the corrected reference intensity is obtained by subtracting the dark intensity from the reference intensity. Finally, the relative reflectance is obtained by dividing the corrected measured intensity by the corrected reference intensity. Specifically, the relative reflectance R(λ) can be obtained by the following formula (4).

[0192]

[0193] Here, λ is the wavelength of the reflected light from the workpiece W, E(λ) is the intensity at wavelength λ, B(λ) is the reference intensity at wavelength λ, and D(λ) is the background intensity (dark intensity) at wavelength λ measured under shading conditions.

[0194] The optical sensor head 107 traverses the workpiece W with each revolution of the grinding table 103, simultaneously guiding light to multiple measurement points on the workpiece W and receiving reflected light from these measurement points. The reflected light is sent to a beam splitter 147. The beam splitter 147 decomposes the reflected light from each measurement point according to wavelength and measures the intensity of the reflected light at each wavelength. The intensity measurement data of the reflected light is sent to a data processing unit 149, which generates a data processing unit such as... Figure 15 The spectrum shown. Figure 15 In the example shown, the spectrum of the reflected light is a spectroscopic waveform that represents the relationship between the relative reflectivity and the wavelength of the reflected light. However, the spectrum of the reflected light can also be a spectroscopic waveform that represents the relationship between the intensity of the reflected light itself and the wavelength of the reflected light.

[0195] The spectrum of the reflected light varies depending on the thickness of the polished layer of the workpiece W. Therefore, the data processing unit 149 monitors the polishing of the workpiece W based on the spectrum of the reflected light as follows. The data processing unit 149 uses the following formula (5) to calculate the spectral change ΔS(T) per unit time.

[0196]

[0197] Here, λ is the wavelength of the reflected light, λ1 and λ2 are the minimum and maximum wavelengths of the wavelength range of the spectrum specified as the object being monitored, T is the grinding time, ΔT is the specified unit time, and R(λ, T) is the relative reflectance (intensity of the reflected light) of wavelength λ at time T. ΔT can be, for example, the time required for the grinding table to rotate p times (p is a small natural number (e.g., 1)).

[0198] Figure 16 This represents the change in spectral density within a unit time ΔT. The spectral change ΔS(T) calculated using equation (5) above corresponds to the region (shown by the cross-section) enclosed by two spectra obtained at two different time points. The area of ​​this region corresponds to the change in the thickness of the polished layer per unit time ΔT. Therefore, by accumulating the spectral change ΔS(T) during the polishing process, it is expected that the change in the thickness of the polished layer can be captured.

[0199] Figure 17 This is a graph showing the spectral change ΔS(T) of the workpiece W during grinding, along the grinding time. For example... Figure 17 As shown, although the spectral variation ΔS(T) varies periodically, the average level of the spectral variation ΔS(T) is almost constant. The data processing unit 149 uses the following formula (6) to calculate the cumulative spectral variation ΔS(T) as the cumulative value along the grinding time.

[0200]

[0201] Here, T0 is the time at which monitoring of film thickness changes begins.

[0202] Figure 18 This is a graph representing the cumulative spectral change A(T) calculated using equation (6) above. As mentioned above, since the spectral change varies periodically, the error from the average level due to this variation hardly accumulates. Therefore, as... Figure 18 As shown, the cumulative spectral change A(T) increases almost linearly with the grinding time. The cumulative spectral change A(T) corresponds to the amount of workpiece W being ground (i.e., the amount of layer removed by the grinding process). The data processing unit 149 calculates the cumulative spectral change during the grinding process of workpiece W and monitors the grinding progress of workpiece W based on the cumulative spectral change. Furthermore, the data processing unit 149 can also determine the grinding endpoint of workpiece W based on the cumulative spectral change. The grinding endpoint can be the time point at which the cumulative spectral change reaches a predetermined target value.

[0203] For reference Figure 23As explained, due to the influence of patterns formed on the surface of workpiece W, the grinding environment (e.g., slurry), etc., there are localized noises that affect the amount of spectral variation. Such noises hinder accurate monitoring of the grinding of workpiece W.

[0204] Therefore, the data processing unit 149 is configured to correct for changes in the applied noisy spectrum. That is, the data processing unit 149 is configured as follows: Figure 19A The amount of spectral variation removed, including noise, is shown. Figure 19B The variation in the spectrum that is missing due to removal is replaced with a variation in the spectrum that does not contain noise. More specifically, the data processing unit 149 is configured to correct the variation in the spectrum when the variation in the spectrum meets the prescribed exclusion conditions, and to calculate the cumulative variation in the spectrum by accumulating the variation in the spectrum that does not meet the exclusion conditions and the above-mentioned corrected variation in the spectrum over the grinding time.

[0205] The specified exclusion criteria are used to determine whether the calculated spectral variation contains noise. The following specific examples can be cited as examples of specified exclusion criteria.

[0206] (i) The change in the spectrum is greater than the threshold.

[0207] (ii) The difference between the current spectral change and the average of multiple spectral changes obtained during the grinding process of workpiece W is greater than the threshold.

[0208] (iii) The change in the spectrum is outside the range of the average of the normal distributions of multiple changes in the spectrum obtained in the past ± Zσ (Z is a pre-defined coefficient).

[0209] (iv) The spectral variation was identified as an outlier by the Smirnov-Grubbs test.

[0210] Regarding (i) above, the threshold is a pre-set fixed value, determined based on data of past spectral variations. For example, the threshold is a value that can intentionally divide the data of past spectral variations into groups of spectral variations without noise and groups of spectral variations with noise.

[0211] Regarding (ii) above, the multiple spectral variations obtained during the grinding process of workpiece W are, for example, the variations of the most recent L spectra that do not include the current spectral variation. The number L is a preset value. When the difference between the current spectral variation and the average of the multiple spectral variations obtained is greater than a threshold, the current spectral variation is more likely to contain noise. This threshold can also be a fixed value. Alternatively, the threshold can be a value that varies based on the aforementioned average. For example, the threshold can also be a value that deviates from the aforementioned average by a preset percentage (%).

[0212] Regarding (iii) above, the multiple changes in the spectrum obtained in the past may also be multiple changes in the spectrum obtained when grinding at least one workpiece having the same structure as workpiece W, or multiple changes in the spectrum already obtained during the grinding process of workpiece W.

[0213] Regarding (iv) above, the sample data used in the Smirnov-Grubbs test may also be multiple changes in the spectrum obtained when grinding at least one workpiece having the same structure as workpiece W, or multiple changes in the spectrum already obtained during the grinding process of workpiece W.

[0214] The spectral change that satisfies any one of the exclusion conditions (i) to (iv) above is presumed to contain noise. Therefore, when the spectral change calculated during the grinding process of workpiece W satisfies any one of the above exclusion conditions, the data processing unit 149 corrects the spectral change. This correction of the spectral change is a noise removal process. The exclusion condition may also be one of the above exclusion conditions (i) to (iv) selected in advance.

[0215] In one embodiment, the data processing unit 149 corrects for the spectral variations that satisfy the exclusion criteria by extrapolation or interpolation. More specifically, the data processing unit 149 excludes spectral variations that satisfy the exclusion criteria (i.e., contain noise), and obtains spectral variations that do not satisfy the exclusion criteria (i.e., do not contain noise) corresponding to the excluded spectral variations by extrapolation or interpolation. The correction by extrapolation uses multiple spectral variations arranged along the grinding time, while the correction by interpolation uses multiple spectral variations obtained at multiple measurement points arranged on the workpiece W.

[0216] The data processing unit 149 is configured to correct the change in the spectrum that satisfies the exclusion conditions by performing any of the following actions.

[0217] (I) The action of replacing the change in spectrum that satisfies the exclusion criteria (i.e., includes noise) with the change in spectrum that does not satisfy the exclusion criteria (i.e. does not include noise).

[0218] (II) The action of replacing the spectral change that meets the exclusion criteria with the average of multiple spectral changes already obtained during the grinding process of workpiece W.

[0219] (III) The action of replacing the spectral change that meets the exclusion criteria with the average of the spectral changes obtained from multiple adjacent measurement points on the workpiece W.

[0220] (IV) The action of generating an approximate formula based on multiple spectral changes obtained during the grinding process of workpiece W, and replacing the spectral changes that satisfy the exclusion conditions with the values ​​obtained from the approximate formula.

[0221] Regarding (I) above, the amount of spectral change that does not meet the exclusion criteria (i.e., includes noise) is the amount of spectral change that has already been obtained during the grinding process of the workpiece W. In one embodiment, the amount of spectral change that does not meet the exclusion criteria (i.e., does not include noise) is the most recent spectral change obtained before the amount of spectral change that meets the exclusion criteria is obtained.

[0222] Regarding (II) above, the multiple spectral changes obtained during the grinding process of workpiece W are multiple spectral changes that do not meet the exclusion criteria (i.e., do not contain noise). In one embodiment, the multiple spectral changes obtained during the grinding process of workpiece W are the changes of the most recent M spectra that do not include the current spectral change. The number M is a preset value.

[0223] Regarding (III) above, adjacent measurement points on the workpiece W are multiple measurement points adjacent to the measurement point on the workpiece W where the spectral change that satisfies the exclusion condition is acquired. As described above, the optical sensor head 107 traverses the workpiece W with each revolution of the grinding table 103, simultaneously guiding light to multiple measurement points on the workpiece W and receiving reflected light from these multiple measurement points. Adjacent measurement points on the workpiece W may, for example, be multiple measurement points located on either side of the measurement point where the spectral change that satisfies the exclusion condition (i.e., does not contain noise) is acquired. The spectral change acquired at each adjacent measurement point is the spectral change that does not satisfy the exclusion condition (i.e., does not contain noise).

[0224] Regarding the above (IV), as Figure 20 As shown, the approximation is a formula on a coordinate system with axes representing the spectral variation and the grinding time. The approximation uses the grinding time as a variable. The approximation is determined based on multiple data points on a coordinate system specific to multiple spectral variations that do not meet the exclusion criteria (i.e., do not contain noise) and the corresponding multiple grinding times. The approximation can also be expressed as a polynomial. The method for obtaining the approximation is not particularly limited. The data processing unit 149 generates the approximation during the grinding process of the workpiece W, and can correct the spectral variation by inputting the grinding time obtained from the spectral variation that meets the exclusion criteria (i.e., contains noise) into the approximation.

[0225] The data processing unit 149 calculates the cumulative spectral change based on the cumulative spectral change over the grinding time that does not meet the exclusion criteria (i.e., does not contain noise) and the aforementioned corrected spectral change, and monitors the grinding amount of the workpiece W based on the cumulative spectral change. Since the cumulative spectral change is the cumulative value of the spectral change that does not contain noise, the data processing unit 149 is able to accurately monitor the grinding amount of the workpiece W.

[0226] The above embodiments are described with the aim of enabling those skilled in the art to carry out the present invention. Various modifications of the above embodiments can obviously be implemented by those skilled in the art, and the technical concept of the present invention can also be applied to other embodiments. Therefore, the present invention is not limited to the described embodiments, and the technical concept is interpreted in the broadest sense according to the scope of the claims.

Claims

1. A grinding method for grinding a grinding target layer of a workpiece, characterized in that, Rotate the grinding table that supports the grinding pad. The workpiece is pressed onto the abrasive pad to abrade the target layer. Irradiate light onto the workpiece. Receive reflected light from the workpiece. The intensity of the reflected light is measured for each wavelength. Generate a spectroscopic waveform representing the relationship between the intensity and the wavelength of the reflected light. Noise is removed from the split waveform using a filter. The noise-removed spectroscopic waveform is subjected to Fourier transform processing to generate a frequency spectrum. The thickness of the polishing layer is determined based on the peaks of the frequency spectrum. The thickness of the grinding object layer corresponding to the peak of the frequency spectrum that disappears due to the removal of the noise is supplemented by interpolation of multiple values ​​of the thickness of the grinding object layer obtained during the grinding process of the workpiece.

2. The grinding method according to claim 1, characterized in that, The filter is configured to remove noise from the spectral waveform that has a frequency peak that does not shift with the grinding time of the workpiece.

3. The grinding method according to claim 1 or 2, characterized in that, The noise is caused by light reflected from the base layer of the polished object layer.

4. The grinding method according to claim 1 or 2, characterized in that, The filter is a band-stop filter.

5. The grinding method according to claim 1 or 2, characterized in that, Before grinding the workpiece, other workpieces with the same pattern as the workpiece are ground. The grinding method further includes the step of creating the filter for removing noise contained in the spectral waveform of reflected light from the other workpiece.

6. The grinding method according to claim 1 or 2, characterized in that, The grinding process of the workpiece further includes a step of creating the filter for removing noise contained in the spectral waveform of the reflected light from the workpiece.

Images