JP4857659B2 - Film thickness evaluation method, polishing end point detection method, and device manufacturing apparatus - Google Patents

Description

  The present invention calculates the difference signal of the reflection spectrum of the polishing time t and the polishing time t−Δt separated by the time difference Δt by irradiating the surface of the film during the polishing of the film, and analyzes the difference signal. The present invention relates to a film thickness evaluation method for evaluating a film thickness at a polishing time t, a polishing end point detection method for detecting a polishing end point based on a calculated film thickness, and a device manufacturing apparatus for manufacturing a device by polishing a film.

  Currently, in the device manufacturing process, CMP (Chemical Mechanical Polishing) is used to polish an oxide film and a metal film formed on a wafer. In this polishing by CMP, it is important to accurately grasp the time point at which polishing is finished (referred to as “polishing end point”) by accurately evaluating the film thickness of the oxide film and metal film to be polished. .

  Polish an oxide film and a metal film formed on a wafer having a region where a circuit pattern is formed (referred to as “circuit region”) and a region where a circuit pattern is not formed (referred to as “non-circuit region”). When evaluating the film thickness, during the polishing of the oxide film and the metal film, the reflection spectrum by irradiating light on the surface of the oxide film and the metal film is measured, and the film is based on the period of the reflection spectrum Evaluate the thickness. However, it is difficult to evaluate the film thickness because the reflection spectrum having the scattered component from the circuit region becomes noise and it is difficult to extract only the interference component.

  Here, during the polishing of the oxide film and the metal film, the reflection spectrum of the polishing time t and the polishing time t−Δt separated by the time difference Δt by irradiating the surface of the oxide film and the metal film is measured and scattered. In order to remove the components, a technique has been proposed in which a difference signal between the reflection spectra of the polishing time t and the polishing time t−Δt is calculated, and the film thickness is evaluated based on the period of the difference signal (Japanese Patent Application No. 2004-053947). ).

  However, in the conventional technique, the waveform of the differential signal is disturbed due to short-period noise included in the reflection spectrum, fluctuations in the absolute value of the reflection spectrum, and the like, making it difficult to calculate the period of the differential signal. When the difference signal is greatly disturbed, the variation in the evaluated film thickness becomes large, and the film thickness cannot be accurately evaluated based on the period of the difference signal.

This invention is made | formed in view of such a point, and it aims at providing the film thickness evaluation method which can evaluate the film thickness in grinding | polishing correctly when grind | polishing a film | membrane.
Another object of the present invention is to provide a polishing end point detection method capable of accurately evaluating the film thickness during polishing and accurately grasping the polishing end point when polishing the film.

  Another object of the present invention is to provide a device manufacturing apparatus capable of accurately evaluating the film thickness during polishing when the film is polished.

In the present invention, in order to solve the above-mentioned problem, during polishing of the film, a difference signal between the reflection spectrum of the polishing time t and the polishing time t−Δt separated by the time difference Δt by irradiating the surface of the film with light , When the wave number of light in the medium is x, the film thickness of the film at the polishing time t is d, the film thickness of the film at the polishing time t−Δt is d + Δd, and A is a proportional constant, 2Asin (2πΔdx) × calculated using the optical model represented by sin {2π (2d + Δd) x}, the thickness evaluation method for evaluating the film thickness of the film in the polishing time t by analyzing the differential signal, said polishing A correction step for correcting the average value of the amplitude intensity of the reflection spectrum at time t-Δt so as to be aligned with the average value of the amplitude intensity of the reflection spectrum at the polishing time t, and among the data points constituting the difference signal Predetermined day The value of point, the predetermined data points, the predetermined number of data points in the immediately preceding predetermined data points, and, the suppression step of molding by substituting the average value of a predetermined number of data points immediately after the predetermined data points, the differential signal Linearly approximate the value of a predetermined data point in each data point constituting the predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined number of data points immediately after the predetermined data point; an amplitude shaping step of shaping by substituting the inclination of the straight line is a straight line approximation, the amplitude intensity and the period from the difference signal and having and a removal step of removing the defective component became other than a predetermined range A film thickness evaluation method is provided.

  According to such a film thickness evaluation method, during polishing of the film, a difference signal is calculated from the reflection spectrum of the polishing time t and the polishing time t−Δt whose amplitude intensity is corrected, and the film thickness is evaluated from the difference signal. Therefore, the variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. In addition, since the film thickness is evaluated from the difference signal in which minute fluctuations are suppressed, the variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. In addition, since the film thickness is evaluated from the difference signal in which the centers of the amplitudes are aligned, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately evaluated when the film is polished. Further, since the film thickness is evaluated from the difference signal from which the defective component has been removed, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately evaluated when the film is polished.

Further, in the present invention, in order to solve the above-mentioned problem, during the polishing of the film, the difference signal between the reflection spectrums of the polishing time t and the polishing time t−Δt separated by the time difference Δt by irradiating the surface of the film with light. Where x is the wave number of light in the medium, d is the film thickness of the film at the polishing time t, d + Δd is the film thickness of the film at the polishing time t−Δt, and A is a proportional constant. 2πΔdx) × sin {2π (2d + Δd) x} is calculated using an optical model, and the difference signal is analyzed to evaluate the film thickness of the film at the polishing time t. In the polishing end point detection method for detecting the polishing end point based on the thickness, the average value of the amplitude intensity of the reflection spectrum at the polishing time t−Δt is corrected to be equal to the average value of the amplitude intensity of the reflection spectrum at the polishing time t. Correction steps to , The value of the predetermined data points in each data point constituting the differential signal, a predetermined data points, the predetermined number of data points in the immediately preceding predetermined data points, and, a predetermined number of data points immediately after the predetermined data points A suppression step of replacing with an average value and shaping, a value of a predetermined data point in each data point constituting the difference signal, a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined An amplitude shaping step in which a predetermined number of data points immediately after the data point are linearly approximated and replaced with the slope of the linearly approximated straight line, and a defect whose amplitude intensity and period are outside the predetermined range from the difference signal polishing end point detecting method characterized by comprising a removing step of removing the component, it is provided.

  According to such a polishing end point detection method, during polishing of the film, a difference signal is calculated from the reflection spectrum of the polishing time t and the polishing time t−Δt whose amplitude intensity is corrected, and the film thickness is evaluated from the difference signal. Therefore, the variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. In addition, since the film thickness is evaluated from the difference signal in which minute fluctuations are suppressed, the variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. In addition, since the film thickness is evaluated from the difference signal in which the centers of the amplitudes are aligned, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately evaluated when the film is polished. Further, since the film thickness is evaluated from the difference signal from which the defective component has been removed, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately evaluated when the film is polished. Therefore, the difference signal is sequentially calculated to accurately grasp the transition of the film thickness, and the polishing end point can be accurately grasped.

According to the present invention, in order to solve the above-described problem, in a device manufacturing apparatus for manufacturing a device by polishing a film, polishing means for polishing the film, and light on the surface of the film during polishing of the film Measuring means for measuring the reflection spectrum of the polishing time t and the polishing time t−Δt separated by a time difference Δt, and a difference signal between the reflection spectra of the polishing time t and the polishing time t−Δt in the medium. When the wave number of light is x, the film thickness of the film at the polishing time t is d, the film thickness of the film at the polishing time t−Δt is d + Δd, and A is a proportionality constant, 2Asin (2πΔdx) × sin { Difference signal calculation means for calculating using an optical model represented by 2π (2d + Δd) x}, film thickness calculation means for evaluating the film thickness at the polishing time t by analyzing the difference signal, has the The average value of the amplitude intensity of the reflection spectrum of the grinding time t-Delta] t, and correcting means for correcting to align the average value of the amplitude intensity of the reflection spectrum of the polishing time t, in each data point constituting the differential signal A predetermined data point, a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a suppression means that replaces and shapes the average value of the predetermined number of data points immediately after the predetermined data point ; The value of a predetermined data point in each data point constituting the difference signal is a straight line of a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined number of data points immediately after the predetermined data point. approximated, an amplitude shaping means for shaping by substituting the inclination of the linear approximation straight lines, the amplitude intensity and the period from the difference signal Device manufacturing apparatus is provided, characterized in that it comprises a removal means for removing the defective component became non constant range, the.

  According to such a device manufacturing apparatus, during polishing of the film, the difference signal is calculated from the reflection spectrum of the polishing time t and the polishing time t−Δt whose amplitude intensity is corrected, and the film thickness is evaluated from the difference signal. The variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. In addition, since the film thickness is evaluated from the difference signal in which minute fluctuations are suppressed, the variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. In addition, since the film thickness is evaluated from the difference signal in which the centers of the amplitudes are aligned, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately evaluated when the film is polished. Further, since the film thickness is evaluated from the difference signal from which the defective component has been removed, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately evaluated when the film is polished.

  In the present invention, during polishing of the film, a difference signal is calculated from the reflection spectrum of the polishing time t and the polishing time t−Δt whose amplitude intensity is corrected, and the film thickness is evaluated from the difference signal. The variation in thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. In addition, since the film thickness is evaluated from the difference signal in which minute fluctuations are suppressed, the variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. In addition, since the film thickness is evaluated from the difference signal in which the centers of the amplitudes are aligned, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately evaluated when the film is polished. Further, since the film thickness is evaluated from the difference signal from which the defective component has been removed, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately evaluated when the film is polished. Therefore, the difference signal is sequentially calculated to accurately grasp the transition of the film thickness, and the polishing end point can be accurately grasped.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings, taking as an example the case of polishing an oxide film, a metal film, and the like formed on a wafer in a device manufacturing process.
First, an optical model used for evaluating the film thickness during polishing when polishing an oxide film, a metal film, and the like formed on the wafer will be described.

It is assumed that the film thickness at the polishing time t is d, and the reflectance with respect to light of wave number x in the medium at the polishing time t is R ₁ (x). Further, the film thickness at the polishing time t−Δt that is separated from the polishing time t by a relatively short time difference Δt is d + Δd, and the reflectance with respect to light of wave number x in the medium at the polishing time t−Δt is R ₀ (x). . These R ₁ (x) and R ₀ (x) are represented by optical models of the following formulas (1) and (2), respectively. Here, A and B are proportional constants.

R ₁ (x) = Acos (4πdx) + B (1)
R ₀ (x) = Acos {4π (d + Δd) x} + B (2)
When polishing an oxide film, a metal film, or the like formed on a wafer in a device manufacturing process, the proportionality constant B has a dependency on the wave number x, but changes slowly with respect to the film thickness difference Δd during polishing. Therefore, the difference signal R ₁ (x) −R ₀ (x) represented by the optical model of the following equation (3) can be removed.

R ₁ (x) -R ₀ (x)
= Acos (4πdx) −Acos {4π (d + Δd) x}
= 2Asin (2πΔdx) × sin {2π (2d + Δd) x} (3)
Here, in the optical model of Expression (3), a short-period sine curve with 1 / (2d + Δd) as one cycle overlaps a long-period sine curve with 1 / Δd as one cycle. Therefore, for the difference signal R ₁ (x) −R ₀ (x), 2d + Δd is calculated from this short period, and the film thickness is calculated by subtracting the predetermined Δd from this and multiplying by 1/2.

Next, a device manufacturing apparatus that is a CMP apparatus having a reflectance measurement function used in the embodiment of the present invention will be described. FIG. 1 is a diagram illustrating a device manufacturing apparatus.
The device manufacturing apparatus 1 includes a rotary table 3 having a polishing pad 2 disposed on the front surface side, and a rotatable polishing head 4 facing the polishing pad 2. A wafer 10 is disposed on the polishing pad 2 side of the polishing head 4.

  During polishing, the slurry 10 is dropped on the polishing pad 2 and the wafer 10 is pressed against the polishing pad 2 to rotate both the rotary table 3 and the polishing head 4 to polish the surface of the wafer 10.

  In addition, the polishing pad 2 of the device manufacturing apparatus 1 is provided with a viewing window 2a made of quartz at a portion where the wafer 10 is disposed during polishing. A through hole 3a is provided in a portion of the turntable 3 corresponding to the viewing window 2a. On the back side of the turntable 3, an optical head 6 facing the viewing window 2a is disposed. The optical head 6 is connected to a spectral reflectance measuring device 7, and the spectral reflectance measuring device 7 is connected to a computer 8.

  During polishing, the surface of the oxide film and metal film formed on the wafer 10 is irradiated with light, and the reflection spectrum is measured. Usually, the measurement is performed once every time the rotary table 3 and the polishing head 4 are rotated, and data processing and data storage accompanying this measurement are executed by the computer 8.

  Hereinafter, a sequence for evaluating the film thickness during polishing when polishing an oxide film and a metal film of an IC (Integrated Circuit) having a chip size of 3.1 mm × 3.8 mm will be described. This IC is equipped with an analog circuit composed of a trench MOS (Metal Oxide Semiconductor) for power and a digital circuit for control. The film thickness of the oxide film and metal film of the IC before polishing is about 2.5 μm. The polishing conditions are that the rotation speed of the turntable 3 is 90 rpm and the polishing rate is 750 nm / min.

First, the spectral data regarding the wavelength is converted into spectral data regarding the wave number. This conversion is intended to simplify optical calculations.
Here, when the refractive index is n, the wavelength is λ, and the calculated wave number is k,
k = 2πn / λ (4)
It becomes. Note that data processing is performed with the unit of wavenumber k as cm ⁻¹ .

  Next, the reflection spectrum by irradiating light on the surface of the oxide film, the metal film, etc. at the polishing time t and the polishing time t−Δt separated by the time difference Δt is measured, and the reflection spectrum at the polishing time t and the polishing time t− A difference signal from the reflection spectrum of Δt is calculated. FIG. 2 is a diagram illustrating the difference signal. In FIG. 2, the horizontal axis represents the wave number, and the vertical axis represents the difference signal.

Here, the m-th reflection spectrum is R _m (k), the measurement m−x-th reflection spectrum is R _mx (k), and the differential signal calculated at the measurement m-th is D _m (k). In terms of
D _m (k) = R _m (k) −R _mx (k) (5)
It becomes. For example, the differential signal D _m (k) illustrated in FIG. 2 is a differential signal between a predetermined reflection spectrum after the twelfth measurement and a reflection spectrum that has been traced 11 times from the predetermined reflection spectrum. x is “11”.

Next, the calculated difference signal D _m (k) is corrected.
Here, the reflection spectra R _m (k) and R _mx (k) have a fluctuation of an absolute value of about 10% or less due to the problem of the measurement method. In order to suppress the influence of fluctuation of the absolute value of the reflection spectrum, correction processing is performed so that the absolute values of the reflection spectrum R _m (k) and the reflection spectrum R _mx (k) are aligned. Specifically, in equation (5), so that the average value of the amplitude intensities of R _mx (k) is aligned to the average value of the amplitude intensity of R _m (k), the average value of the amplitude intensities of R _mx (k) The correction processing is performed by multiplying by a correction coefficient. When the correction coefficient is C and the corrected differential signal is expressed as D _m ′ (k),
D _m ′ (k) = R _m (k) −R _mx (k) × C (6)
It becomes.

Then, the noise of a short cycle having the reflection spectrum was removed, 'to suppress the minute fluctuation of (k), the correction processed difference signal D _m' difference signal D _m to smooth (k). FIG. 3 is a diagram illustrating the smoothed difference signal. In FIG. 3, the horizontal axis represents the wave number, and the vertical axis represents the difference signal.

Here, the value of a predetermined data point in each data point of the difference signal D _m ′ (k) is set to the predetermined data point, three data points immediately before the predetermined data point, and immediately after the predetermined data point. And smoothing by replacing with the average value of the three data points.

Next, the difference signal D _m ′ (k) is differentiated in order to remove the upward or downward slope of the difference signal D _m ′ (k) and align the center of the amplitude of the difference signal D _m ′ (k). To process. FIG. 4 is a diagram showing a differential signal subjected to differentiation processing. In FIG. 4, the horizontal axis represents the wave number, and the vertical axis represents the difference signal.

Here, the value of a predetermined data point in each data point of the difference signal D _m ′ (k) is set to the predetermined data point, three data points immediately before the predetermined data point, and immediately after the predetermined data point. These three data points are approximated by a straight line, and the slope of this straight line is calculated as a differential value and replaced to perform differential processing. By performing the differentiation process, the period of the wave number k of the difference signal D _m ′ (k) can be easily calculated. When differential processing is performed, the amplitude intensity of the difference signal D _m ′ (k) changes and the phase is shifted by 90 °, but the period of the wave number k does not change, so there is no problem even if differential processing is performed.

For convenience of explanation, the waveform of FIG. 4 is a normal waveform, but actually has an amplitude strength defect and a period defect.
Next, the difference signal D _m ′ (k) is screened in order to remove the amplitude strength defect. FIG. 5 is a diagram illustrating a differential signal having an amplitude strength defect. In FIG. 5, the horizontal axis represents the wave number and the vertical axis represents the difference signal.

Here, for example, when the average value of the amplitude intensity of the difference signal D _m ′ (k) is within a predetermined range in the measurement wave number region, it is assumed that the amplitude intensity is sufficient and not defective data. If it is outside this predetermined range, the amplitude intensity is insufficient and is removed as defective data. Assuming that the average value is A and the predetermined value is “1” for example,
A ≧ 1 (7)
It becomes. By screening, it becomes possible to calculate the normal waveform of FIG.

Further, the differential signal D _m ′ (k) is screened in order to remove periodic defects. FIG. 6 is a diagram illustrating a differential signal having a periodic failure. In FIG. 6, the horizontal axis represents the wave number and the vertical axis represents the difference signal.

Here, for example, when the error between the cycle of the highest wave number and all the average cycles is within a predetermined range in the measurement wave number region, the cycle is assumed to be stable and not defective data. If it is outside this predetermined range, the period is not stabilized and is removed as defective data. If the period of the highest wave number is K1, all the average periods are K2, and a predetermined range of the period K1 is expressed by an expression, for example, from “0.8 × K2” to “1.2 × K2”,
0.8 <K1 / K2 <1.2 (8)
It becomes. By performing the screening, it becomes possible to calculate the normal waveform in FIG.

  The transition of the film thickness according to the increase in the number of polishings will be described before and after the screening. FIG. 7 is a diagram showing the transition of the film thickness before screening. FIG. 8 is a diagram showing the transition of the film thickness after screening. 7 and 8, the horizontal axis represents the number of polishing times, and the vertical axis represents the film thickness.

  Before the screening, as illustrated in FIG. 7, the film thickness indicated by a circle in the figure decreases with an increase in the number of polishings, and the film thickness varies due to an amplitude strength defect and a cycle defect. After the screening, as illustrated in FIG. 8, the film thickness does not vary and it is easy to grasp the transition of the film thickness. In addition, the film thickness shown by the square mark in a figure was measured with the optical film thickness meter (Nanometrics 9100 by Nanometrics), and respond | corresponds with transition of the film thickness shown by the circle mark in the figure. .

Next, the film thicknesses of the oxide film, the metal film, etc. during the polishing time t are evaluated from the period of the wave number k of the difference signal D _m ′ (k).
Here, when the approximate polishing rate evaluated in advance is represented by r and the calculated film thickness is represented by d (t),
d (t) = π / k−rΔt / 2 (9)
It becomes.

With the above sequence, the thickness d (t) during polishing is calculated when polishing the oxide film, the metal film, and the like.
In addition, when the period of the wave number k of the difference signal D _m ′ (k) becomes large and is about the same as the measurement wave number region, and it becomes difficult to calculate the period of the wave number k, the calculated difference signal D _m ′ (k) Is compared with the sine curve stored in the database, and the period and amplitude intensity corresponding to the matched database sine curve are obtained as the period and amplitude intensity of the calculated differential signal D _m ′ (k) May be. This database stores a plurality of sine curves using the period and amplitude intensity as parameters.

Next, a method for determining the time difference Δt will be described. FIG. 9 is a film thickness pass rate at each phase difference. In FIG. 9, the horizontal axis represents the phase difference and the vertical axis represents the screening pass rate.
Here, the time difference Δt is converted into a phase difference that is more universal than the time difference Δt. When the central wavelength of reflectance measurement is λ _C , the refractive index of the film to be polished is n, and the converted phase difference is φ,
φ = 4nπrΔt / λ _C (10)
It becomes.

As illustrated in FIG. 9, the pass rate is maximized in the vicinity of the phase difference where the phase is reversed. Specifically, the pass rate becomes maximum when the phase difference φ is around 180 ° and 540 °. Moreover, when the phase difference φ is in the vicinity of 90 ° to 270 °, the pass rate is about 80% of the maximum pass rate. Further, there is periodicity every 360 °. Therefore, it is preferable to determine the time difference Δt in the range where the phase difference φ is 90 ° to 270 °. Expressed as a formula:
λ _C / 8n ≦ rΔt ≦ 3λ _C / 8n (11)
It becomes.

In this way, during the polishing of the film, the difference signal D _m ′ (k) is calculated from the reflection spectrum of the polishing time t and the polishing time t−Δt whose amplitude intensity is corrected, and this difference signal D _m ′ (k ), The variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. Further, since the film thickness is evaluated from the differential signal D _m ′ (k) in which minute fluctuations are suppressed, the variation in the evaluated film thickness is reduced, and the film thickness during polishing is accurately evaluated when the film is polished. it can. Further, since the film thickness is evaluated from the difference signal D _m ′ (k) in which the centers of the amplitudes are aligned, the variation in the evaluated film thickness is reduced, and the film thickness being polished can be accurately determined when polishing the film. Can be evaluated. Further, since the film thickness is evaluated from the differential signal D _m ′ (k) from which the defective component has been removed, the variation in the evaluated film thickness is reduced, and the film thickness being polished is accurately evaluated when the film is polished. it can. Therefore, the difference signal D _m ′ (k) can be sequentially calculated to accurately grasp the transition of the film thickness, and the polishing end point can be accurately grasped. In addition, variation in device characteristics is reduced, and device yield is improved.

  Here, when polishing an oxide film, a metal film, or the like having a step, there is a tendency that the number of data to be removed by the above-described screening tends to increase, resulting in large variations in film thickness. FIG. 10 is a diagram showing a film having a step. FIG. 11 is a diagram illustrating the transition of the film thickness when the number of missing data is large. In FIG. 11, the horizontal axis represents the number of polishing times and the vertical axis represents the film thickness.

  As illustrated in FIG. 10, the buried oxide film 23 is formed on the silicon substrate 24, and the polysilicon 22 is formed on the buried oxide film 23. An oxide film 21 is formed on a silicon substrate 24 including the polysilicon 22.

When the oxide film and the metal film having such a step are polished, for example, as illustrated in FIG. 11, the number of missing data increases in the vicinity of 20 to 40 polishing times.
In order to reduce the number of missing data, a plurality of time differences Δt are prepared instead of one, and a plurality of differential signals D _m ′ (k) of the reflection spectra of the polishing time t and the polishing times t−Δt are calculated. For example, five time differences Δt are prepared, and five difference signals D _m ′ (k) are calculated. FIG. 12 is a diagram illustrating a differential signal having a phase difference of 128 °. FIG. 13 is a diagram illustrating a differential signal having a phase difference of 157 °. FIG. 14 is a diagram illustrating a differential signal having a phase difference of 171 °. FIG. 15 is a diagram illustrating a differential signal having a phase difference of 185 °. FIG. 16 is a diagram illustrating a differential signal having a phase difference of 257 °. FIG. 17 is a diagram showing the difference signals of FIGS. 12 to 17, the horizontal axis represents the number of polishing times and the vertical axis represents the film thickness.

As illustrated in FIGS. 12 to 16, when the film thickness is evaluated from one difference signal, the number of missing data increases, but as illustrated in FIG. 17, a plurality of difference signals D _m ′ (k ), The number of missing data is reduced.

In this way, since a plurality of difference signals D _m ′ (k) are calculated, the number of data of the difference signals D _m ′ (k) increases. Therefore, the number of data does not decrease even when screening is performed for the increased amount, the transition of the film thickness can be accurately grasped, and the polishing end point can be accurately grasped. In addition, variation in device characteristics is reduced, and device yield is improved.

In addition, since a plurality of difference signals D _m ′ (k) are calculated, the film thickness can be evaluated from a plurality of difference signals instead of one difference signal, and the film thickness reliability is improved.
When 10 samples were actually evaluated for film thickness during polishing using the film thickness evaluation method described above, when one time difference Δt was prepared, the film thickness was 93 nm at the maximum from the set value. The minimum variation was −68 nm. When five time differences Δt were prepared, the film thickness varied from the set value to 42 nm at the maximum and −24 nm from the minimum. The variation from the set value was 100 nm to −100 nm, which was confirmed to be effective in the device manufacturing process.

  In the description of the sequence for evaluating the film thickness during polishing, the correction process, the smoothing process, the differential process, the amplitude intensity defect removal and the periodic defect removal steps are described as being executed continuously. Even if the step is executed alone, the variation in the evaluated film thickness is reduced, and the film thickness during polishing can be accurately evaluated when the film is polished. Moreover, even if any step is omitted from the steps of correction processing, smoothing, differentiation processing, amplitude strength defect removal and period defect removal, the evaluated film thickness variation is reduced, and the film is polished. In addition, the film thickness during polishing can be accurately evaluated.

It is a figure which shows a device manufacturing apparatus. It is a figure which shows a difference signal. It is a figure which shows the smoothed difference signal. It is a figure which shows the differential signal by which the differential process was carried out. It is a figure which shows the difference signal which has an amplitude strength defect. It is a figure which shows the difference signal which has a periodic defect. It is a figure which shows transition of the film thickness before screening. It is a figure which shows transition of the film thickness after screening. It is the pass rate of the film thickness at each phase difference. It is a figure which shows the film | membrane which has a level | step difference. It is a figure which shows transition of the film thickness when there are many missing data numbers. It is a figure which shows the difference signal of phase difference 128 degrees. It is a figure which shows the difference signal of phase difference 157 degrees. It is a figure which shows the difference signal of 171 degrees of phase differences. It is a figure which shows the difference signal of phase difference 185 degrees. It is a figure which shows the difference signal of phase difference 257 degrees. It is a figure which superimposes and shows the difference signal of FIGS.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Device manufacturing apparatus 2 Polishing pad 2a Viewing window 3 Rotary table 3a Through-hole 4 Polishing head 5 Slurry 6 Optical head 7 Spectral reflectance measuring device 8 Computer 10 Wafer

Claims (6)

During polishing of the film, the difference between the reflection spectrum of the polishing time t and the polishing time t−Δt separated by a time difference Δt by irradiating light on the surface of the film, the wave number of light in the medium x, and the polishing time When the film thickness of the film at t is d, the film thickness of the film at the polishing time t−Δt is d + Δd, and A is a proportionality constant, it is expressed as 2Asin (2πΔdx) × sin {2π (2d + Δd) x}. In the film thickness evaluation method for evaluating the film thickness at the polishing time t by calculating using an optical model and analyzing the difference signal,
A correction step of correcting the average value of the amplitude intensity of the reflection spectrum at the polishing time t-Δt so as to be aligned with the average value of the amplitude intensity of the reflection spectrum at the polishing time t ;
The value of a predetermined data point in each data point constituting the differential signal is an average of a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined number of data points immediately after the predetermined data point. A suppression step of forming by replacing with a value ;
The value of a predetermined data point in each data point constituting the difference signal is a straight line of a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined number of data points immediately after the predetermined data point. Amplitude shaping step that approximates and replaces with the straight line approximation of the straight line, and shaping ,
A removal step of removing a defective component whose amplitude intensity and period are outside a predetermined range from the difference signal ;
The film thickness evaluation method characterized by having. The film thickness evaluation method according to claim 1, wherein a plurality of difference signals of reflection spectra of the polishing time t and the polishing times t−Δt are calculated. The rate at which the film is polished is r, and the center wavelength in the wavelength range where the reflection spectrum can be measured is λ. _C When the refractive index of the film is n, the time difference Δt is expressed as λ _C / 8n ≦ rΔt ≦ 3λ _C The film thickness evaluation method according to claim 1, wherein the film thickness is set in a range of / 8n. A plurality of sine curves with parameters of period and amplitude intensity are stored in advance in the database, and the period and amplitude intensity of the difference signal are obtained by comparing the sine curve of the difference signal with the sine curve stored in the database. The film thickness evaluation method according to claim 1, wherein the film thickness is evaluated. During polishing of the film, the difference between the reflection spectrum of the polishing time t and the polishing time t−Δt separated by a time difference Δt by irradiating light on the surface of the film, the wave number of light in the medium x, and the polishing time When the film thickness of the film at t is d, the film thickness of the film at the polishing time t−Δt is d + Δd, and A is a proportional constant, it is expressed by 2Asin (2πΔdx) × sin {2π (2d + Δd) x}. In the polishing end point detection method, wherein the film thickness of the film at the polishing time t is evaluated by calculating using the optical model and analyzing the difference signal, and the polishing end point is detected based on the calculated film thickness. ,
  A correction step of correcting the average value of the amplitude intensity of the reflection spectrum at the polishing time t-Δt so as to be aligned with the average value of the amplitude intensity of the reflection spectrum at the polishing time t;
  The value of a predetermined data point in each data point constituting the differential signal is an average of a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined number of data points immediately after the predetermined data point. A suppression step of forming by replacing with a value;
  The value of a predetermined data point in each data point constituting the difference signal is a straight line of a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined number of data points immediately after the predetermined data point. Amplitude shaping step that approximates and replaces with the straight line approximation of the straight line, and shaping,
  A removal step of removing a defective component whose amplitude intensity and period are outside a predetermined range from the difference signal;
  A method for detecting a polishing end point, comprising: In a device manufacturing apparatus for manufacturing a device by polishing a film,
Polishing means for polishing the film;
A measuring means for irradiating the surface of the film with light during polishing of the film and measuring a reflection spectrum of the polishing time t and the polishing time t−Δt separated by a time difference Δt;
The difference signal of the reflection spectrum between the polishing time t and the polishing time t−Δt, the wave number of light in the medium is x, the film thickness of the film at the polishing time t is d, and the film at the polishing time t−Δt. A differential signal calculation means for calculating using an optical model represented by 2A sin (2πΔdx) × sin {2π (2d + Δd) x}, where d + Δd and A is a proportional constant;
A film thickness calculating means for evaluating the film thickness of the film at the polishing time t by analyzing the difference signal;
Correction means for correcting the average value of the amplitude intensity of the reflection spectrum at the polishing time t−Δt so as to be aligned with the average value of the amplitude intensity of the reflection spectrum at the polishing time t;
The value of a predetermined data point in each data point constituting the differential signal is an average of a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined number of data points immediately after the predetermined data point. A suppression means for substituting with a value and molding;
The value of a predetermined data point in each data point constituting the difference signal is a straight line of a predetermined data point, a predetermined number of data points immediately before the predetermined data point, and a predetermined number of data points immediately after the predetermined data point. Amplitude shaping means for approximating and substituting with the slope of the straight line approximated,
Removing means for removing a defective component whose amplitude intensity and period are outside the predetermined range from the difference signal;
A device manufacturing apparatus comprising:

Images