Plasma processing apparatus, plasma processing method, and etching system

Abstract

Provided are a plasma processing apparatus, a plasma processing method, and an etching processing system that achieve highly accurate film-thickness monitoring and end point determination, regardless of a database acquisition method or signal processing to be executed.　A plasma processing apparatus (70), which measures wafer reflection light reflected from a processing target wafer during etching processing and estimates the film thickness and the depth of the wafer from the wafer reflection light by using a waveform pattern database (122), comprises a film-thickness / depth amount decision unit (102) that executes data processing for matching the correspondence relationship between the film thickness and the depth for use in the waveform pattern database (122) and the intensity of light before and after second signal processing is performed on the intensity of light.

Description

Plasma processing apparatus, plasma processing method, and etching system

[0001] The present invention relates to a plasma processing apparatus, a plasma processing method, and an etching system.

[0002] In the manufacture of semiconductor devices, various components and interconnections are formed on the surface of a wafer. These devices are formed by repeatedly depositing films of various materials, including conductors, semiconductors, and insulators, followed by the removal of unwanted portions. Dry etching using plasma (also known as plasma etching) is widely used as a process for removing unwanted portions. In plasma etching, gas introduced into the processing chamber of an etching device is converted into plasma using a high-frequency power source, and the wafer is then exposed to the plasma gas. During this process, anisotropic and isotropic etching occurs due to sputtering by ions in the plasma and chemical reactions by radicals. By using these methods appropriately, components and interconnections with various structures can be formed on the wafer surface.

[0003] If the processed shape by plasma etching differs from the design, the various components formed will not be able to function as intended. For this reason, many process monitor technologies have been proposed to monitor and stabilize the etching process. Process monitors that measure the thickness of the film formed on the wafer and the depth of the trenches and holes formed on the wafer by measuring the light reflected from the wafer during processing are called film thickness / depth monitors, and have been used to determine the endpoint of the etching process.

[0004] Patent Document 1 describes a method for improving processing accuracy using a film thickness / depth monitor. This document describes a method for detecting when the target film is about to be completely removed using a film thickness / depth monitor that uses plasma light as a light source, and terminating the etching process. It also describes that the etching process is then performed by switching to conditions that allow highly selective etching of the target and non-target portions.

[0005] Patent Document 2 describes a technique for improving the accuracy of film thickness and depth measurement by a film thickness and depth monitor. This document describes the use of an external light source instead of plasma light as the light source for irradiating the wafer.

[0006] Patent Document 3 describes a method for improving the accuracy of film thickness and depth estimation by using multiple waveform pattern databases that show the correspondence between reflected light from a wafer (wafer reflected light) and film thickness / depth. This document uses multiple waveform pattern databases and calculates estimated values ​​of film thickness and depth by calculation using weighting based on the difference between measured wafer reflected light and the multiple waveform pattern databases.

[0007] Japanese Patent Application Laid-Open No. 2006-119145 Japanese Patent Application Laid-Open No. 2004-507070 Japanese Patent Application Laid-Open No. 2022-058184

[0008] In a method for calculating film thickness and depth using multiple waveform pattern databases (hereinafter simply referred to as databases), multiple databases must be prepared in advance. These databases contain data in which the wafer reflected light (light intensity) acquired at each time during wafer processing in the target processing step is associated with the film thickness and depth during processing. For example, the database can be acquired by etching the wafer in the processing step, acquiring the wafer reflected light, measuring the film thickness and depth before and after etching, and linearly interpolating the film thickness and depth before and after etching to assign them to the film thickness and depth at each time during etching. Alternatively, if the relationship between the wafer film thickness and depth and the wafer reflected light in the processing step can be calculated, the database can be acquired by calculation.

[0009] The estimation of film thickness and depth using these databases is performed based on the wafer reflected light measured during processing of the wafer to be estimated. At this time, the wafer reflected light is subjected to various signal processing to remove noise components, which are various variations in light intensity that are added during measurement. For example, an LPF (Low Pass Filter) is used to remove random variations in light intensity over time, and differential value calculations and difference calculations between measurement times are performed to remove light intensity and offset components. Since the film thickness and depth are calculated based on the wafer reflected light that has undergone these signal processings, the same signal processing is also performed on the normal database.

[0010] Here, in determining the etching endpoint, it is necessary to estimate the film thickness and depth in real time from the light reflected from the wafer during etching, determine the etching endpoint based on the estimated values, and terminate the etching operation. Therefore, the signal processing performed on the wafer reflected light and database is limited to signal processing that can be performed in real time. In the time-domain calculation processing of the signal processing, such as the LPF and differential value calculation, the correspondence relationship between the time and the measured wafer reflected light before and after the signal processing may change. Due to this change in the correspondence relationship, for example, when estimating film thickness and depth using multiple databases, inconsistencies may occur between the correspondence relationships between film thickness and depth and the wafer reflected light among the multiple databases, resulting in a decrease in the accuracy of the film thickness and depth estimation.

[0011] Furthermore, changes in the correspondence between time and the reflected light from the measurement wafer before and after the signal processing on the wafer being etched can reduce the accuracy of etching endpoint determination. For example, even if the correspondence between the film thickness and depth and the reflected light from the wafer before signal processing is the same, the correspondence between the film thickness and depth and the reflected light from the wafer after signal processing will be different depending on whether the etching rate of the wafer for which the film thickness and depth are to be estimated is fast or slow. In this case, if the film thickness and depth are estimated based on the reflected light from the wafer after signal processing using the same database or multiple databases with consistent (consistent) correspondence between film thickness and depth and the reflected light from the wafer, the actual film thickness and depth will differ even if the estimated film thickness and estimated depth are the same. As a result, wafers processed using endpoint determination will have variations in film thickness and depth after processing from wafer to wafer.

[0012] The present invention solves the above-mentioned problems and provides, for example, a plasma processing apparatus, a plasma processing method, and an etching processing system that realize highly accurate film thickness monitoring and endpoint determination, regardless of the database acquisition method and signal processing executed.

[0013] According to one embodiment of the present disclosure, a plasma processing apparatus measures wafer reflected light from a wafer being processed during an etching process and estimates a film thickness and depth of the wafer being processed from the wafer reflected light using a database, the plasma processing apparatus includes a data processor that performs data processing to match the correspondence between the film thickness and depth and the light intensity used in the database before and after performing second signal processing on the light intensity, and a film thickness and depth corrector that corrects the film thickness and depth estimated in real time from the wafer reflected light during etching using a deviation amount between the correspondence between the film thickness and depth and the light intensity before and after performing the second signal processing, the deviation amount being acquired in the data processing to match the correspondence between the film thickness and depth and the light intensity before and after performing the second signal processing.

[0014] According to a representative embodiment of the present disclosure, the correspondence between film thickness or depth and light intensity in the database remains constant (unchanging) regardless of the measurement conditions of the database or the conditions of the second signal processing performed on the database, and there is no longer any inconsistency in the correspondence between film thickness or depth and light intensity between multiple databases.

[0015] Furthermore, regardless of differences in the amount of deviation in the correspondence between film thickness / depth and light intensity generated by the first signal processing between wafers being etched, the correspondence between the estimated film thickness / depth and the actually processed film thickness / depth is always constant.

[0016] Therefore, it is possible to estimate the film thickness and depth with high accuracy, and to determine the end point with high accuracy using the estimated film thickness and depth. Problems, configurations, and effects other than those described above will become clear from the following description of the embodiment.

[0017] 1 is a diagram showing an example of a plasma processing apparatus; FIG. 2 is a block diagram of an example of a film thickness / depth calculation unit; FIG. 3 is a block diagram of an example of a film thickness / depth determination unit; FIG. 4 is a diagram showing an example of table data indicating the content of a spectrum measured at one time, and FIG. 5 is an example of table data indicating the content of a database; FIG. 5 is a flowchart showing an example of a procedure for creating a database, and FIG. 5 is an example of table data acquired in creating the database; FIG. 6 is a flowchart showing an example of a procedure for correcting the correspondence between film thickness / depth and spectrum in the database, and FIG. 6 is an example of table data acquired in correcting the database; FIG. 7 is a flowchart showing an example of a procedure for calculating a time delay Δt used for correcting the database using a test signal; FIG. 8 is a flowchart showing an example of a procedure for calculating a time delay Δt used for correcting the database using an actual measurement signal; FIG. 9 is a diagram showing an example of the contents of a plurality of databases; and FIG. 10 is an example of a diagram showing the results of measuring a time delay Δt using a test signal. 1A is a diagram showing an example of the results of film thickness estimation using each corrected database, and FIG. 1B is a diagram showing an example of the results of film thickness estimation using each database before correction. FIG. 1A is a diagram showing an example of the results of film thickness estimation for multiple wafers using the corrected database, and FIG. 1B is a diagram showing an example of the results of film thickness estimation for multiple wafers using the database before correction. FIG. 1B is a diagram showing an example of the results of measuring a time delay Δt using an actual measurement signal. FIG. 1C is a diagram showing an example of the results of performing signal processing on a database using signal processing without time delay. FIG. 1D is a diagram showing an example of the results of film thickness estimation using each database created using signal processing without time delay. FIG. 1E is a block diagram of an example of a film thickness and depth correction unit. FIG. 1F is a diagram showing an example of a comparison of the film thickness estimation error amount for an evaluation target wafer with and without correction of the determined film thickness and depth values. FIG. 1G is a diagram showing an example of a similar time path calculated based on the similarity between the database and the spectrum of the evaluation target wafer. FIG. 1H is a diagram showing an example of a comparison of the film thickness estimation error amount for an evaluation target wafer with and without correction of the determined film thickness and depth values ​​when the correspondence between the film thickness in the database and the spectrum is not corrected.

[0018] In the following embodiments, when necessary for convenience, the description will be divided into multiple sections or embodiments, but unless otherwise specified, they are not unrelated to each other, and one is a partial or complete modification, detail, supplementary explanation, etc. of the other. Furthermore, in the following embodiments, when the number of elements (including the number, numerical value, amount, range, etc.) is mentioned, it is not limited to that specific number, and may be more or less than the specific number, unless otherwise specified or when it is clearly limited to a specific number in principle.

[0019] Furthermore, in the following embodiments, it goes without saying that the components (including element steps, etc.) are not necessarily essential unless otherwise specified or considered to be clearly essential in principle. Similarly, in the following embodiments, when referring to the shape, positional relationship, etc. of components, etc., it is intended to include those that are substantially similar or similar to the shape, etc., unless otherwise specified or considered to be clearly not essential in principle. The same applies to the above numerical values ​​and ranges.

[0020] Hereinafter, a plasma processing apparatus, a plasma processing method, and an etching system according to embodiments of the present invention will be described with reference to the drawings. First, the configuration of a plasma processing apparatus that performs etching processing and is equipped with a film thickness / depth estimation means will be described, and then a method for estimating film thickness / depth during etching processing in a plasma processing apparatus (plasma processing method) will be described as an embodiment of the present invention.

[0021] A schematic diagram of the plasma processing apparatus used in this embodiment is shown in Figure 1. The plasma processing apparatus 70 includes a plasma processing unit 55. The plasma processing unit 55 includes a vacuum processing chamber 10, in which an etching gas introduced into the vacuum processing chamber 10 from a gas introduction means is excited and decomposed by electric power or microwaves generated by a high-frequency power source or the like to form plasma 12. This plasma 12 performs an etching process (plasma processing) on ​​a processing object 16, such as a semiconductor wafer, placed on a sample stage 14. The plasma processing apparatus 70 may also be referred to as an etching processing apparatus.

[0022] 1, the plasma processing apparatus 70 includes, in addition to the plasma processing unit 55, a film thickness / depth amount calculation unit 30, a control unit 40, and an end point determination unit 44. The film thickness / depth amount calculation unit 30 and the end point determination unit 44 will be described later.

[0023] The control unit 40 controls the introduction of gas into the vacuum processing chamber 10, the generation and control of the plasma 12, and the application of voltage to the processing target by a high-frequency power supply or the like. These controls adjust the timing between the various devices so that the desired etching process is achieved. When the plasma 12 is pulsed, the pulsation control is also performed by the control unit 40. At this time, the plasma 12 is pulsed by switching on and off the voltage application by a high-frequency power supply or the like that converts the etching gas into plasma, or by modulating microwave irradiation, etc. The plasma can also be pulsed by time-modulating the introduction of the etching gas.

[0024] The plasma processing unit 55 is equipped with a mechanism for measuring the etching amount, film thickness, and depth of the processing object 16. Light emitted from the light source unit 18 is introduced into the vacuum processing chamber 10 via the optical system 50, and irradiates the processing object 16 with irradiation light 22. The light source unit 18 uses continuous light ranging from ultraviolet to infrared, but any wavelength necessary for measuring the film thickness and depth of the processing object 16 may be used, and continuous light in a specific wavelength band or a specific wavelength may also be used. Reflected light 24 from the processing object 16 is introduced into the detection unit 28 via the optical system 50.

[0025] Here, the light-irradiating optical port in optical system 50 that irradiates light into vacuum processing chamber 10 is composed of, for example, an optical fiber that guides light from light source unit 18 and a lens that irradiates light toward processing target 16 in vacuum processing chamber 10. Furthermore, the light-receiving optical port in optical system 50 that receives reflected light 24 is composed of, for example, a lens that collects reflected light 24 and an optical fiber that guides the collected light to detection unit 28. The light-irradiating optical port and the light-receiving optical port are not limited to the above configurations, and when these ports are configured as a single port, they may be configured with a bundle fiber having two fibers and one fiber at both ends, and one lens.

[0026] Furthermore, multiple light-irradiating optical ports and / or multiple light-receiving optical ports may be installed when measuring reflected light 24 at multiple positions on the processing object 16. The configuration of these optical ports is not limited to a configuration using fibers and lenses, and the light-irradiating optical port may be configured with a planar light-emitting element, and the light-receiving optical port may be configured with a camera using a CCD or CMOS sensor.

[0027] When film thickness and depth estimation is performed using a specific wavelength band or multiple wavelengths, the detection unit 28 is configured as a spectrometer, which disperses the introduced light and detects the light intensity for each wavelength. When film thickness and depth estimation is performed using specific wavelengths, the detection unit 28 is not limited to a spectrometer and may be a photodetector or the like. In this case, if only the light introduced into the detection unit 28 is of the desired specific wavelength, a photodetector can be used directly. If continuous light is introduced, a mechanism for selecting only specific wavelengths, such as a monochromator, can be provided upstream of the photodetector. This light detection is performed continuously at regular intervals during etching, for example, at 2 Hz or 10 Hz. The output data from the detection unit 28 is input to the film thickness / depth calculation unit 30, where the film thickness and depth are determined (estimated). In other words, the detection unit 28 functions as a light receiver that receives light reflected from the wafer during plasma processing.

[0028] In the above description, the case where light is incident from an external light source unit 18 as a light source has been described, but if light from the plasma 12 is used as a light source, it is not necessary to use the light source unit 18. Even when the plasma 12 is used as a light source, the light emitted from the plasma 12 is reflected by the processing object 16, and the reflected light 24 is detected in the same way as when the light source unit 18 is used.

[0029] An example of the functional block configuration of the film thickness / depth calculation unit 30 is shown in Figure 2. The film thickness / depth calculation unit 30 includes a digital signal processing unit 100, a film thickness / depth determination unit 102, a film thickness / depth correction unit 104, and a waveform pattern database 122. Note that "film thickness / depth" refers to the film thickness of a film formed on a wafer (processing object 16) and the depth of a groove or hole formed on the wafer. Similarly to "film thickness / depth," the "film thickness" described in the claims is not limited to the film thickness of a film formed on a wafer, but also includes the depth of a groove or hole formed on the wafer.

[0030] The time-series data D1 of the light intensity of each wavelength introduced from the detection unit 28 to the film thickness / depth calculation unit 30 is subjected to removal or correction of various noises and fluctuations by the digital signal processing unit 100, and is then supplied to the film thickness / depth determination unit 102 as time-series data D2.

[0031] The digital signal processing unit 100 performs predetermined signal processing (first signal processing) on ​​the output data of the detection unit 28 (the intensity of the reflected light received by the photoreceiver) and outputs the result. For example, when removing random noise from the time-series data of the light amount, the digital signal processing unit 100 uses a low-pass filter (LPF). For example, the relational expression between the time-series data D1 and D2 before and after signal processing in a Butterworth LPF is as shown in Equation 1 below.

[0032]

[0033] Equation 1 shows the LPF method using data from three times, where λ is wavelength, t is time, and a and b are coefficients determined by the sampling rate of the time series data and the cutoff frequency of the LPF, etc. For example, when the sampling rate is 0.5 sec and the cutoff frequency is 0.1 Hz, {a0, a1, a2, b1, b2} = {0.0201, 0.0402, 0.0201, -1.5610, 0.6414}.

[0034] Furthermore, for example, when removing a light intensity offset from time-series data of light intensity, signal processing is used to calculate the amount of change in light intensity between times and a differential value. For example, Savitzky-Golay is used to calculate the differential value, and the relational expression between the time-series data D1 and D2 before and after signal processing is as shown in Equation 2 below.

[0035]

[0036] Equation 2 shows a method for calculating the first derivative using data from five time periods, where a is a coefficient determined by the number of time periods used, the derivative order, etc. For example, when calculating the first derivative using data from five time periods, {a0, a1, a2, a3, a4} = {2, 1, 0, -1, -2}.

[0037] Furthermore, for example, when removing and correcting the magnification fluctuation of the light amount synchronous with all wavelengths from the spectroscopic measurement data (spectral spectrum) at each time, signal processing is applied to normalize the light amount of each wavelength by the average value or the sum of absolute values ​​of the light amount of all wavelengths. Normalization by the sum of absolute values ​​is performed, for example, using the following equation 3.

[0038]

[0039] The signal processing and data manipulation in the digital signal processing unit 100 are not limited to those described above, and signal processing and data manipulation required for the film thickness and depth determination unit 102 to determine the film thickness and depth, such as data reduction, thinning, interpolation, smoothing, statistical processing, and comparison, may also be performed.

[0040] The film thickness and depth determining unit 102 determines (estimates) the film thickness and depth using the supplied time-series data D2 and data supplied from the waveform pattern database 122. An example of the functional block configuration of the film thickness and depth determining unit 102 is shown in Fig. 3. That is, the film thickness and depth determining unit 102 functions as a determiner that determines the thickness of the film on the wafer from the intensity of reflected light that has been signal-processed by a signal processor, based on the database.

[0041] The film thickness / depth amount determining unit 102 includes a waveform comparator 302, a film thickness / depth amount determiner 304, an optimum database determiner 306, a compatible database calculator 308, and a film thickness / depth calculation model generator 310.

[0042] For example, the film thickness and depth can be determined by comparing a spectrum 320 (i.e., time-series data D2) acquired during etching and supplied from the digital signal processor 100 with a database 322 supplied from the waveform pattern database 122. An example of the content of the spectrum 320 is shown in Figure 4(a). The spectrum sp(t) at each time t is a data table that associates wavelengths w1 to wM with light intensities I1,t to IM,t.

[0043] 4B shows an example of the contents of the database 322. One database is a data table that associates film thicknesses and depths d1 to dN with spectra db-sp1 to db-spN. That is, the database 322 associates the intensity of light at a predetermined wavelength with the thickness of a film formed on a reference wafer, which serves as a basis for determining the film thickness.

[0044] When determining the film thickness and depth using one database 322, the spectrum 320 and the database 322 are directly supplied to the waveform comparator 302. The waveform comparator 302 compares the spectrum sp(t) at each time with each spectrum db-sp in the database 322, and supplies the calculated error amount (or similarity amount) to the film thickness and depth determiner 304. The film thickness and depth determiner 304 determines an optimal spectrum from the spectrum db-sp in the database based on the error amount (or similarity amount), and determines the film thickness and depth associated with the determined spectrum as the film thickness and depth for the spectrum sp(t). The determined film thickness and depth are supplied to the film thickness and depth correction unit 104. By determining the film thickness and depth using one database 322, it is possible to reduce the capacity of, for example, a storage device or the like that constitutes the waveform pattern database 122.

[0045] Furthermore, for example, the film thickness and depth may be determined by comparing the spectrum 320 supplied from the digital signal processing unit 100 with a plurality of databases 322 supplied from the waveform pattern database 122. In other words, a plurality of databases may be stored in the waveform pattern database 122. For example, the spectrum 320 and the plurality of databases 322 are directly supplied to the waveform comparator 302, and the spectrum sp(t) at each time is compared with all the spectra db-sp of the plurality of databases 322, and the amount of error therebetween is calculated. The calculated amount of error is then supplied to the film thickness and depth determiner 304.

[0046] The calculated error amount is also supplied to an optimum database determiner 306, which determines one optimum database from among the multiple databases based on the error amount of the spectrum in each database, and supplies identification information of the determined database to a film thickness / depth amount determiner 304.

[0047] The film thickness / depth determiner 304 determines the database to be used using the identification information of the supplied database. Then, it determines the spectrum of the optimal database using the error amount between the supplied database and the spectrum sp(t). Then, it determines the film thickness / depth associated with the determined spectrum as the film thickness / depth amount in the spectrum sp(t). The determined film thickness / depth amount is supplied to the film thickness / depth amount correction unit 104. By using multiple databases, even if the data in a database is noisy, for example, it is possible to increase the number of databases and ensure a certain level of accuracy.

[0048] Here, the matching database calculator 308 can generate a composite database having an error amount lower than the supplied error amount by weighting and adding the plurality of databases 322 using the error amounts of the plurality of databases 322 supplied and the waveform comparator 302. The matching database calculator 308 has a function of adding the generated composite database to the plurality of databases. To improve the accuracy of estimating the film thickness and depth, the matching database calculator 308 may be used to generate the composite database before estimating the film thickness and depth.

[0049] Furthermore, for example, the film thickness and depth may be determined using a spectrum 320 supplied from the digital signal processing unit 100 and a learning model (linear / nonlinear regression function, machine learning, deep learning, etc.) of the spectrum and film thickness and depth created using a plurality of databases 322 supplied from the waveform pattern database 122. For example, the plurality of databases 322 are supplied to a film thickness and depth calculation model generator 310, and a learning model is created in the film thickness and depth calculation model generator 310.

[0050] The learning model created from the correspondence between spectra and film thickness / depth in multiple databases is, for example, a regression function created using kernel ridge regression or support vector machine, an analysis / classification model created using random forest or k-nearest neighbor method, or a learning model created using a neural network. The learning model created by the film thickness / depth calculation model generator 310 and the spectrum 320 supplied from the digital signal processing unit 100 are supplied to the film thickness / depth amount determiner 304. The film thickness / depth amount determiner 304 determines the film thickness / depth amount corresponding to the spectrum sp(t) using the learning model. The determined film thickness / depth amount is supplied to the film thickness / depth amount correction unit 104. Determining the film thickness / depth amount using the learning model enables accurate and rapid processing of the learned data.

[0051] The film thickness and depth supplied to the film thickness and depth correction unit 104 as described above are time-series data D3 in FIG. 2. The film thickness and depth correction unit 104 can correct the film thickness and depth at each time based on the calculated film thickness and depth time-series data D3. For example, if the calculated film thickness and depth time-series data D3 contains fluctuations due to noise or the like, the film thickness and depth at each time are corrected by linearly approximating the time progression of the film thickness and depth. The film thickness and depth corrected by the film thickness and depth correction unit 104 are output to the outside as film thickness and depth data D4.

[0052] The plasma processing apparatus 70 of FIG. 1 performs endpoint determination in the endpoint determination unit 44 using the film thickness and depth data supplied from the film thickness and depth calculation unit 30. The endpoint determination unit 44 compares the acquired film thickness and depth data with predetermined target film thicknesses and depths to determine whether the current film thickness and depth have reached the targets. If it determines that the targets have been reached, it supplies an etching process stop signal to the control unit 40. The control unit 40 stops the etching process based on the etching stop signal. This operation enables the plasma processing apparatus 70 to perform endpoint determination using a film thickness and depth monitor. In other words, the endpoint determination unit 44 functions as a determiner that determines the process endpoint based on the film thickness determined by the determiner.

[0053] As described above, the digital signal processing unit 100 executes a signal processing step, the film thickness and depth determination unit 102 executes a determination step, and the end point determination unit 44 executes a determination step. In addition, as will be described later, the film thickness and depth determination unit 102 executes a data processing step.

[0054] 2 used in determining the film thickness and depth is acquired before estimating the film thickness and depth of the wafer to be processed. For example, FIG. 5A shows an example of a method for acquiring the database by etching a test sample. In this method, the film thickness and depth of the test sample are first measured in advance before etching (S11). If non-destructive measurement is possible, the test sample is evaluated, and if destructive measurement is possible, a test sample equivalent to the test sample is measured.

[0055] Next, the test sample is etched, and the optical spectrum during etching is measured (acquired) (S12). Then, the film thickness and depth of the test sample after etching are measured non-destructively or destructively (S13). Using the measured spectrum and the film thickness and depth before and after etching, film thickness and depth are assigned to the spectrum at each time during the etching (S14), and a database is created and acquired by performing desired digital signal processing on the spectral data (S15). That is, the optical intensity and film thickness included in the database before correction are obtained based on the results of processing using a test sample (a sample wafer prepared in advance as a reference wafer). Furthermore, by generating the database using the test sample, each piece of data can be reliably acquired.

[0056] The method of assigning film thickness and depth to the spectrum at each time in Fig. 5(a) (S14) and the method of signal processing the spectrum data (S15) will be described with reference to Fig. 5(b). Column c1 in the table of Fig. 5(b) shows the time when the spectrum was acquired during etching, column c2 shows the acquired spectrum, column c3 shows the film thickness and depth corresponding to each time, and column c4 shows the results of performing signal processing (second signal processing) on ​​the spectrum. Here, the film thickness and depth in column c3 are determined by, for example, linearly interpolating the initial value d1 measured before etching and the endpoint value dN measured after etching and assigning them to each time.

[0057] The signal processing method (second signal processing) for determining the spectral data in column c4 may be, for example, the same as the signal processing executed by the digital signal processor 100 on the time-series data D1 in FIG. 2 . That is, the first signal processing and the second signal processing may have the same processing content. In this case, the same signal processing is executed on the spectra and database used during etching to determine the film thickness and depth, making it possible to accurately compare them. Furthermore, for example, signal processing different from that executed by the digital signal processor 100 may be used as long as it is possible to determine the film thickness and depth. The database created in this manner becomes table data (database before correction) in which column c3 of the film thickness and depth in FIG. 5( b) corresponds to column c4 of the spectrum.

[0058] Here, the database can be created not only by etching the test sample shown in Fig. 5A, but also by simulation using optical calculations, for example. In this case, the relationship between the film thickness / depth and the spectrum is known, so the database can be created without performing steps S11 to S14 shown in Fig. 5A.

[0059] In the database acquired by the above-described method, the correspondence relationship between the film thickness / depth and the spectrum is corrected as necessary. For example, in the signal processing of S15 in FIG. 5A, if a difference occurs in the correspondence relationship between the time and / or film thickness / depth and the spectrum before and after the signal processing, the correspondence relationship may be corrected after the signal processing. A method for correcting the correspondence relationship between the time and / or film thickness / depth and the spectrum is shown in FIG. 6A.

[0060] For example, the time delay Δt due to the signal processing is calculated (S21), and the correspondence between time and spectrum in the database is shifted by Δt (S22). Figure 6(b) shows the result of correcting the database using the calculated time delay Δt. Column c4 in Figure 6(b) is the same as column c4 in Figure 5(b). Here, the time t corresponding to each spectrum column c4 in the database is uniformly shifted by t - Δt, and a new corresponding time column c5 is calculated.

[0061] Next, using the correspondence between columns c1 and c3 in FIG. 5(b), d' corresponding to each t-Δt in column c5 in FIG. 6(b) is calculated to determine column c6. Then, the correspondence between column c6 of film thickness / depth and column c4 of spectrum is determined as the corrected database. That is, the time delay due to signal processing is calculated, and the correspondence between light intensity and film thickness is corrected based on that time delay. This correction makes it possible to match (constantly) the correspondence between film thickness / depth and spectrum in the database, which is caused by changes in the correspondence between time and spectrum before and after signal processing. This eliminates inconsistencies between databases, particularly in film thickness / depth estimations using multiple databases, and suppresses a decrease in the estimation accuracy.

[0062] In the above correction, the calculation of the time delay Δt in S21 of Fig. 6(a) is performed, for example, by a method using a test signal as shown in Fig. 7. In Fig. 7, for example, a sine wave of a specific frequency, in which a time delay due to signal processing can be observed, is prepared as the test signal I(t) (S31). Next, signal processing is performed on the test signal I(t) to obtain a post-processing test signal I'(t) (S32). Then, Δt is calculated by comparing the test signals I(t) before and after signal processing with I'(t) (S33).

[0063] Furthermore, for example, the time delay Δt can also be calculated by a method using the measured signal shown in Fig. 8. Here, a database is first created according to the procedure in Fig. 5(a), and a database DB is prepared in which column c3 (film thickness / depth) and column c2 (spectrum) in Fig. 5(b) correspond to each other (S41).

[0064] A database DB' is also prepared in which column c3 (film thickness / depth) and column c4 (spectrum after signal processing) are associated with each other (S42). Then, film thickness / depth estimation is performed using DB as the database for film thickness / depth estimation and DB' as the spectrum to be estimated for film thickness / depth estimation (S43).

[0065] Using the film thickness and depth d'i obtained by film thickness and depth estimation and the film thickness and depth di in DB', the film thickness and depth delay amount is calculated by Δdi = (d'i - di), and the time delay amount Δt is calculated from the film thickness and depth delay amount Δdi using the relationship between column c1 (time) and column c3 (film thickness and depth amount) in Figure 5 (b) (S44). The time delay amount Δt is calculated, for example, by using the average value Δdave of the film thickness and depth delay amount Δdi, by Δdave × (tN - t1) / (dN - d1).

[0066] By using the time delay Δt calculated by these methods, the correspondence between the film thickness / depth and the spectrum in the database can be corrected.

[0067] The above description has been made with respect to the case where the signal processing in S15 in Fig. 5A causes a difference in the correspondence relationship between the film thickness / depth and the spectrum, but if signal processing in S15 is used that does not cause a difference in the correspondence relationship between the film thickness / depth and the spectrum, the correction of the correspondence relationship between the film thickness / depth and the spectrum in the database as described above in Fig. 5A is not necessary. In other words, the signal processing itself may be data processing that leaves the correspondence relationship between the light intensity and the film thickness unchanged.

[0068] For example, in the low-pass filter of Equation 1, after performing a low-pass filter calculation process on the data time series from time t1 to time tN in the database, the low-pass filter calculation process is performed again on the data series from time tN to time t1. As a result, the time delay occurring in the first calculation process is offset by the second calculation process, and the time delay amount of the time series data after the second calculation process becomes zero.

[0069] Furthermore, for example, in the calculation of the differential value of Equation 2, the time delay amount becomes zero by changing the calculation process to Equation 4 below.

[0070]

[0071] In this calculation process, data from times tn+1 and tn+2, which are earlier than time tn, are used to calculate the differential value at time tn. Therefore, if this calculation process is used for data acquired in real time during etching, the real-time nature of the output will be reduced. On the other hand, since the database contains data that has already been acquired, using this calculation process makes it possible to zero the time delay of the data after differential value calculation.

[0072] The above describes a method for making the correspondence between the film thickness / depth and the spectrum in the database consistent (unchanged) before and after signal processing. However, it is also possible to reduce the difference in the film thickness estimate depending on the database used by using a database in which the correspondence between the film thickness / depth and the spectrum has changed due to signal processing, and correcting the output estimated film thickness value.

[0073] For example, for each database, a time delay Δt is calculated in step S21 of FIG. 6A, and the etching rate calculated from columns c1 (time) and c3 (film thickness / depth) in the data table of FIG. 5B for that database is used to determine the film thickness / depth change Δd corresponding to that time delay Δt. In film thickness / depth estimation using each database, the film thickness / depth estimate value acquired at each time is corrected by the film thickness / depth change Δd in that database. As a result, the corrected film thickness / depth estimate value at each time is unique regardless of the database used. Even when using this method, it is possible to obtain an effect equivalent to that of a method in which the correspondence between the film thickness / depth in the database and the spectrum is matched before and after signal processing.

[0074] Signal processing that does not cause a difference in the correspondence between film thickness / depth and spectrum is not limited to the above, and any signal processing that does not change the correspondence between film thickness / depth and spectrum before and after signal processing such as a zero-phase filter can be used.

[0075] Although the above describes a method in which the correspondence between the film thickness / depth and the spectrum in the database matches before and after signal processing, it is also desirable that the correspondence between the measurement time and the spectrum does not change before and after signal processing for the spectrum 320 that is the target for film thickness / depth estimation. This is because, even for the spectrum 320 that is the target for film thickness / depth estimation, a difference (time delay) occurs in the correspondence between the time and the spectrum when signal processing is performed, and the correspondence between the time and the spectrum, that is, the correspondence between the film thickness / depth and the spectrum, may change depending on the amount of time delay.

[0076] As a result, the film thickness and depth values ​​estimated from the spectrum change depending on the amount of time delay. Here, the correspondence between the measurement time and the spectrum refers to, for example, the relationship between the peaks and troughs in the time-series change in light intensity shown in FIG. 10 and the time. Signal processing that maintains the correspondence between the measurement time and the spectrum 320 can be performed using, for example, the above-mentioned low-pass filter twice in ascending and descending order of the time axis, a method using the differential value calculation of Equation 4, or a zero-phase filter. While this may reduce the real-time nature of film thickness and depth estimation, it is possible to estimate the film thickness and depth at the current time by, for example, accurately outputting film thickness and depth estimates at times prior to the current time during etching and extrapolating the film thickness and depth at the current time from the output film thickness and depth estimates.

[0077] The above-mentioned arithmetic processing such as database acquisition and database correction (FIGS. 5 to 8) is mainly executed by the film thickness and depth calculation unit 30. That is, the film thickness and depth calculation unit 30 performs signal processing (second signal processing) on ​​the light intensity included in the waveform pattern database 122, and functions as a data processor that executes data processing to match the correspondence between the light intensity and the film thickness before and after the signal processing.

[0078] Furthermore, the calculation processes such as database acquisition and database correction may not necessarily be performed by the plasma processing apparatus 70 (film thickness / depth amount calculation unit 30) but may be performed by, for example, the host computer 60 in Fig. 1. Reference numeral 80 in Fig. 1 also indicates an etching system made up of the host computer 60 and the plasma processing apparatus 70.

[0079] The host computer 60 is connected to the plasma processing apparatus 70 via a network such as a communication cable or wirelessly, and creates a database using data such as spectra supplied from the plasma processing apparatus 70. The host computer 60 supplies the created database to the plasma processing apparatus 70. The host computer 60 also has the function of generating a learning model to be used for estimating film thickness and depth, and can generate a learning model for estimating film thickness and depth using the database supplied from the plasma processing apparatus 70 and supply the created learning model to the plasma processing apparatus 70. In other words, the host computer 60 functions as a data processing apparatus that generates the database, and performs data processing during database generation to match the correspondence between light intensity and film thickness before and after the second signal processing.

[0080] 2 has been described as a function of the film thickness / depth correction unit 104, which is a method for reducing fluctuations in the determined film thickness / depth due to noise, etc. However, the correction of film thickness / depth is not limited to this. For example, if the correspondence between the determined film thickness / depth and the actually processed film thickness / depth changes depending on the etching speed of the wafer during etching, the determined film thickness / depth may be changed by a predetermined calculation.

[0081] For example, the determined film thickness and depth are corrected using a time delay Δt due to signal processing, which is used in data processing to match the correspondence between the film thickness and depth in the database and the spectrum before and after the signal processing. An example of the functional blocks of the film thickness and depth correction unit 104 is shown in Figure 16. The determined film thickness and depth values ​​1620 up to the current time, supplied from the film thickness and depth determination unit 102, are input to an etching rate detector 1602 of the film thickness and depth correction unit 104.

[0082] The etching rate detector 1602 calculates the etching rate using the film thickness and depth at each time. For example, the correspondence relationship between time and film thickness / depth is linearly approximated, and the etching rate is calculated from the slope of the linear approximation. Alternatively, for example, if the time-series data D2, which is the measured spectrum data during etching in FIG. 2 , and the spectrum data from the waveform pattern database 122 are supplied to the etching rate detector 1602 via a path not shown, the etching rate may be calculated using the measured spectrum data and the spectrum data in the database. For example, the time range of the spectrum data in the database that is most similar to the measured spectrum data may be detected using dynamic time warping or the like, and the etching rate of the measured spectrum data may be calculated using the width of the time range of the measured spectrum data and the width of the detected time range of the database.

[0083] The etching rate calculated by the etching rate detector 1602 is supplied to a film thickness / depth corrector 1604. The film thickness / depth corrector 1604 corrects the determined film thickness / depth values ​​using the supplied etching rate and the time delay Δt due to the signal processing, which is supplied via a path not shown. For example, the correction is performed by adding the product of the etching rate ER(t) and the time delay Δt to the determined film thickness / depth value d(t) at the current time. This is because a spectrum at a past time corresponds to each time through signal processing, and the film thickness / depth determined from the spectrum at the past time is corrected to the film thickness / depth at the current time.

[0084] The film thickness and depth corrected by the film thickness and depth corrector 1604 are supplied as film thickness and depth data to the end point determination unit 44 in Fig. 1, and the end point determination operation is carried out in the same manner as described above. Furthermore, with regard to these calculation processes, data processing may be carried out by the host computer 60 in the same manner as data processing in generating the database.

[0085] Next, the results of determining the end point by estimating the film thickness and depth of the present invention using the plasma processing apparatus 70 will be described below.

[0086] In this example, multiple databases are obtained by actual measurements, and the correspondence between the film thickness and depth in the database and the spectrum is corrected using a test simulation signal. The results of film thickness and depth estimation using these multiple databases are described.

[0087] An example of a database acquired by actual measurement is shown in Figure 9. There are five databases, and the etching process performed when the databases were acquired ranged from a film thickness of approximately 230 nm to approximately 200 nm, but the etching rate during processing differed for each database. The digital signal processing performed by the digital signal processing unit 100 in Figure 2 was a combination of sequentially performing a low-pass filter according to Equation 1, a first-order differential calculation according to Equation 2, a low-pass filter according to Equation 1, and spectrum normalization according to Equation 3. The method for determining the film thickness and depth in the film thickness and depth amount determiner 102 in Figure 2 was a method in which the film thickness and depth were determined by the film thickness and depth amount determiner 304 via the waveform comparator 302 using multiple databases 322 in Figure 3.

[0088] The correlation between the film thickness and the spectrum in each database was corrected by calculating the time delay Δt due to signal processing using a test signal created through simulation. The signal processing used is the same as the signal processing performed by the digital signal processor 100 on the target film thickness estimation target. A sine wave was used as the test signal, and the results of comparing the time transition of the test signal before and after signal processing are shown in Figure 10.

[0089] It can be seen that the time when the light intensity reaches a minimum, around 30 to 35 seconds, is shifted later after signal processing (After_Proc) compared to before signal processing (Before_Proc). This shift is the time delay, from which the time delay amount Δt was calculated. The calculated time delay amount Δt was used to correct the correspondence between film thickness and spectrum in each database.

[0090] Figure 11(a) shows the results of film thickness estimation performed using each corrected database on a spectrum acquired during etching of one sample. It can be seen that the film thickness estimates at each time during etching are consistent regardless of the database used. For comparison, Figure 11(b) shows the results of film thickness estimation performed using each uncorrected database. It can be seen that the film thickness estimates at each time during etching differ depending on the database used. Thus, when using the uncorrected database, the film thickness estimates vary depending on the database, making it impossible to accurately estimate film thickness using multiple databases. On the other hand, when using the corrected database, the film thickness estimates remain consistent regardless of the database, confirming that film thickness estimation can be performed correctly using multiple databases.

[0091] In the endpoint determination using the film thickness estimation based on the above-mentioned corrected multiple databases, there is no variation in the film thickness estimation value at each time, so it is possible to accurately determine whether the film thickness estimation value has reached the target film thickness, thereby achieving highly accurate endpoint determination.

[0092] In this example, a description will be given of the results of film thickness and depth estimation and endpoint determination when a learning model of the spectrum and film thickness / depth created by kernel ridge regression is used in the film thickness / depth estimation method in Example 1. Other aspects are the same as in Example 1, and their description will be omitted in this example. It goes without saying that in this example to Example 6, the learning model, database creation, correction, and other processes can be executed by the film thickness / depth amount determination unit 102 and the host computer 60 described above.

[0093] As in Example 1, one sample was etched, and the spectra acquired during etching were used to estimate film thickness using the corrected databases. The results are shown in Figure 11(a) above. Since there is no contradiction between the film thickness and the spectra in each database, the estimated film thickness at each time point is constant regardless of the database. Using these databases, a learning model was created by kernel ridge regression in the film thickness / depth calculation model generator 310 in Figure 2.

[0094] For example, the correspondence data between spectra and film thicknesses in the entire database is integrated, and a nonlinear or linear regression model is optimized using the spectra as explanatory variables and the film thickness as the target variable. In the optimization, for example, the correspondence data between all spectra and film thicknesses is used to calculate the optimal values ​​of the parameters used in the regression model through cross-validation. The regression model created in this way is a learning model that takes spectra as input and outputs film thicknesses.

[0095] In this embodiment, the etching endpoint is determined by estimating the film thickness from the spectrum during etching using the learning model created above. Since the correspondence between film thickness and spectrum is consistent in the multiple databases used to create the learning model, as described above, the film thickness output value from the spectrum using the created learning model is always unique. Therefore, in endpoint determination using film thickness estimation by this learning model, there is no variation in the film thickness estimate at each time, so it is possible to accurately determine whether the film thickness estimate has reached the target film thickness, thereby achieving highly accurate endpoint determination.

[0096] In this embodiment, the learning model was created using kernel ridge regression, but the method for creating the learning model is not limited to this. For example, it is possible to use a regression function created using kernel ridge regression or a support vector machine, an analysis / classification model created using a random forest or k-nearest neighbor method, or a neural network. Even when these methods are used, the correspondence between film thickness and spectrum is consistent in the multiple corrected databases used to create the learning model, so the created learning model can output a unique film thickness for the input spectrum. Therefore, it is clear that highly accurate film thickness estimation and endpoint determination can be achieved.

[0097] In this example, the results of film thickness and depth estimation and endpoint determination will be described when a method of determining film thickness and depth in film thickness and depth determiner 304 via waveform comparator 302 using one database 322 in Fig. 3 is used in place of the method of estimating film thickness and depth in Example 1. The remaining parts are the same as in Example 1, and their description will be omitted in this example.

[0098] 9 was used as the database, and the time delay Δt was calculated by signal processing in the same manner as in Example 1, and the correspondence between the film thickness and the spectrum in the database was corrected. Ten wafers were evaluated for film thickness estimation using the database, and all wafers were etched so that the film thickness after etching would be approximately 201 nm. Each evaluation sample had different recipe conditions during the etching process.

[0099] FIG. 12( a) shows the results of film thickness estimation for each evaluation wafer. The horizontal axis of FIG. 12( a) represents the evaluation wafer number, and the vertical axis represents the estimated film thickness at the etching endpoint. The estimated film thickness values ​​were the same for all evaluation wafers, with a very small variation of 0.3 nm. This demonstrates that the film thickness estimation of this embodiment can be used to calculate highly accurate film thickness estimates regardless of the evaluation wafer. As a comparative example, FIG. 12( b) shows the results of film thickness estimation using the uncorrected database. The estimated film thickness values ​​at the etching endpoint differed for each evaluation wafer, with a variation of 3.3 nm, demonstrating that accurate film thickness estimation was not possible for each evaluation wafer. Thus, even when using a single database for film thickness estimation, it was confirmed that highly accurate film thickness estimation could be achieved regardless of the evaluation wafer by correcting the correspondence between the film thickness in the database and the spectrum.

[0100] In endpoint determination using the film thickness estimate based on the single corrected database, there is no variation in the film thickness estimates for the wafers being evaluated at the etching endpoint and at each time during etching, making it possible to accurately determine whether the film thickness estimate has reached the target film thickness, thereby achieving highly accurate endpoint determination.

[0101] In this example, a description will be given of the results of film thickness and depth estimation and end point determination when a plurality of databases in Example 1 are acquired by simulation. Other aspects are the same as in Example 1, and therefore will not be described in this example.

[0102] The data showing the correspondence between film thickness / depth and spectrum for each database was created by calculating the spectrum for each structure through optical simulation using information on the structure of the sample being evaluated and the structural changes caused by etching. Multiple databases were obtained by performing optical simulations while changing the etching rate and structural changes caused by etching. In other words, the light intensity and film thickness included in the database before correction were obtained based on the simulation results.

[0103] By acquiring the database in this manner, steps S11 to S14 in the database acquisition procedure of FIG. 5A are self-evident and can be omitted, and the acquired database undergoes signal processing in S15. Therefore, the number of steps required for database generation is reduced, and the setting of conditions is also simplified. Thereafter, as in Example 1, the correspondence between film thickness / depth and spectrum in the database is corrected based on the time delay Δt, and the correction is used for film thickness / depth estimation. Therefore, there is no inconsistency between the correspondence between film thickness / depth and spectrum in multiple databases, and when film thickness / depth estimation is performed using each database, the estimated film thickness / depth values ​​for the same evaluation sample at a specific time are consistent, as in Example 1.

[0104] Therefore, in the endpoint determination using the film thickness estimates based on the multiple databases obtained by the above simulation, there is no variation in the film thickness estimates for the wafers being evaluated at the etching endpoint and at each time during etching, so it is possible to accurately determine whether the film thickness estimate has reached the target film thickness, and highly accurate endpoint determination is achieved.

[0105] In this example, a description will be given of the results of film thickness and depth estimation and end point determination when actual measurement data is used to calculate the time delay Δt in Example 1. Other aspects are the same as in Example 1, and therefore will not be described in this example.

[0106] The correspondence between the film thickness and the spectrum in the multiple databases shown in Fig. 9 was corrected using the procedure in Fig. 8. DB1 was used as the test waveform pattern database in S41. The results of film thickness estimation using DB1 before and after signal processing in accordance with S42 and S43 are shown in Fig. 13. The film thickness at each time in the database after signal processing is shown as Actual, and the results of film thickness estimation in the database after signal processing using the database before signal processing are shown as Estimated.

[0107] By estimating the film thickness using the above combination, it can be seen that an error occurs between Actual and Estimated. The time delay Δt in S44 is calculated by measuring the time difference between Actual and Estimated for each film thickness. In this example, since an estimation error occurs in the film thickness estimate (Estimated) immediately after the start of etching due to the influence of signal processing, the average time difference from 20 seconds onwards was calculated as the time delay Δt. The correspondence between film thickness and spectrum in multiple databases was corrected using the calculated time delay Δt.

[0108] Since the time delay Δt calculated in this embodiment is substantially the same as the time delay Δt calculated in embodiment 1, it can be said that the corrected databases obtained in this embodiment are equivalent to those in embodiment 1. Therefore, in endpoint determination using film thickness estimation based on multiple databases that calculate the time delay Δt using the above-mentioned actual measurement data, there is no variation in the film thickness estimates for the wafer being evaluated at the etching endpoint and at each time during etching. Therefore, it is possible to accurately determine whether the estimated film thickness has reached the target film thickness, and highly accurate endpoint determination is achieved.

[0109] In this example, we will describe the results of film thickness / depth estimation and end point determination when the database in Example 1 is created using signal processing that does not cause any difference in the correspondence between film thickness / depth and spectrum. The rest of the process is the same as in Example 1, and so the description will be omitted in this example.

[0110] In Example 1, the signal processing performed on the database is the same as the signal processing performed by the digital signal processing unit 100 in Fig. 2 on the film thickness estimation target, and is a combination of sequentially performing a low-pass filter of Equation 1, a first-order differential calculation of Equation 2, a low-pass filter of Equation 1, and spectrum normalization of Equation 3. In this example, signal processing is performed that performs the same function as this signal processing combination and does not cause a difference in the correspondence between film thickness and spectrum before and after the signal processing.

[0111] Specifically, the low-pass filter in Equation 1 performs signal processing once in the forward direction and once in the reverse direction of the time series in which the spectra in each database were acquired. The calculation of the first derivative in Equation 2 is performed using Equation 4. Equation 3 is signal processing performed on the spectrum at each time point, and since it does not cause any difference in the correspondence between film thickness and spectrum, it was decided to use it as is.

[0112] The results of performing the above-described signal processing on the light intensity data at each time point at a specific wavelength in a specific database are shown in Fig. 14. In Fig. 14, symbol a indicates the result before signal processing (no signal processing), symbol b indicates the result after performing the signal processing of this embodiment (signal processing without time-delay), and symbol c indicates the result after performing the signal processing of Example 1 (signal processing with time-delay).

[0113] The time when the light intensity was at its minimum before signal processing and in this embodiment are the same, and it can be seen that no time delay occurs due to the signal processing in this embodiment. Even when compared with Example 1 (symbol c) shown for reference, it is clear that there is no time delay due to the signal processing in this embodiment.

[0114] Figure 15 shows the results of creating multiple databases using signal processing that does not result in a difference in the correspondence between film thickness and spectrum before and after signal processing, and performing film thickness estimation using each database on spectra acquired during the etching of one sample. It can be seen that the film thickness estimates at each time during etching are consistent regardless of the database used. This demonstrates that by creating databases using signal processing that does not result in a difference in the correspondence between film thickness and spectrum before and after signal processing, film thickness estimates remain consistent regardless of the database, and film thickness estimation using multiple databases can be performed correctly.

[0115] In the endpoint determination using film thickness estimation based on multiple databases corrected by the method of this embodiment, there is no variation in the film thickness estimation value at each time, so it is possible to accurately determine whether the film thickness estimation value has reached the target film thickness, thereby achieving highly accurate endpoint determination.

[0116] In this example, the method for estimating film thickness and depth in Example 1 uses one database in the waveform pattern database 122 in Fig. 2, calculates determined values ​​of film thickness and depth in the film thickness and depth determination unit 102, and corrects the determined values ​​of film thickness and depth shown in Fig. 16 in the film thickness and depth correction unit 104. The rest of the process is the same as in Example 1, and the description thereof will be omitted in this example.

[0117] 9 was used as the database, and the time delay Δt was calculated by signal processing in the same manner as in Example 1, and the correspondence between the film thickness and the spectrum in the database was corrected. Ten wafers were evaluated for film thickness estimation using the database, and each wafer was etched so that the film thickness after etching would be approximately 200 nm. Each evaluation sample had different recipe conditions during the etching process, resulting in different etching rates.

[0118] In estimating the film thickness of each evaluation wafer, the determined values ​​of film thickness and depth shown in FIG. 16 were corrected. The etching rate detection unit 1602 calculated the etching rate ER(t) at each time during etching using the determined film thickness values ​​determined at that time and the time before that time. In this example, the correlation between time and the determined film thickness value was linearly approximated, and the etching rate ER(t) was calculated from the slope of the obtained approximate line. The film thickness and depth corrector 1604 corrected the determined film thickness value d(t) using the calculated etching rate ER(t) and the above-mentioned time delay Δt. The correction was performed at each time using the determined film thickness value d(t), the etching rate ER(t), and the time delay Δt, as calculated in Equation 5 below.

[0119]

[0120] Here, d'(t) is the determined film thickness value after correction, and in the end point determination, this corrected determined film thickness value d'(t) is used as the film thickness estimate value and compared with the target film thickness for the end point determination.

[0121] The results of film thickness estimation for each evaluation wafer using these methods are shown in Figure 17. The horizontal axis in Figure 17 represents the evaluation wafer number, and the numbers in parentheses represent the etching rate index for that wafer. The vertical axis represents the film thickness estimation error at the etching endpoint, which is the error between the estimated film thickness and the actual film thickness. The solid bars in Figure 17 represent the results of applying the correction of the film thickness estimate of the present invention, while the hatched bars represent the results of estimation using the conventional method without applying the correction of the film thickness estimate of the present invention. When the film thickness estimate correction of the present invention was applied, the film thickness estimation error was small for all wafers, demonstrating that highly accurate film thickness estimation can be achieved regardless of recipe conditions such as the wafer etching rate. On the other hand, when using the conventional method without applying the correction of the estimated film thickness, the film thickness estimation error varied depending on the wafer etching rate, indicating that accurate film thickness estimation was not achieved.

[0122] As a result, in the film thickness estimation using the correction of the determined values ​​of film thickness and depth described above, the estimated film thickness value and the actual film thickness value correspond with high precision, and the endpoint determination using the film thickness estimated value of this method eliminates film thickness variations between wafers after processing, thereby achieving high-precision processing.

[0123] In this example, a description will be given of the results of estimating film thickness and depth when the calculation method of the time delay Δt and the etching rate ER(t) is changed in the film thickness and depth estimation method using correction of the determined film thickness and depth values ​​of Example 7. The rest of the process is the same as in Example 1, and the description will be omitted in this example.

[0124] 9 was used as the database, and the time delay Δt was calculated by signal processing using DB2, which was the actual measurement data, and the correspondence between the film thickness in the database and the spectrum was corrected, as in Example 5. The wafer to be evaluated, whose film thickness is estimated using this database, is the same as in Example 7.

[0125] In estimating the film thickness of each evaluation wafer, the determined values ​​of film thickness and depth shown in FIG. 16 were corrected. The etching rate detection unit 1602 calculated the etching rate ER(t) using spectral time series data from a digital signal processing unit (not shown) and spectral time series data from a waveform pattern database. In this example, dynamic time warping (DTW), a dynamic time warping method, was used to calculate similar time paths in which the spectra of the DB and the evaluation wafer were similar, and the etching rate ER(t) was calculated. Here, in this example, the spectral similarity was quantified using cosine similarity. However, the L1 norm or L2 norm of Euclidean distance, or the error of feature values ​​calculated using a dimension reduction technique or machine learning, may also be used. FIG. 18 shows an example of calculation of similar time paths at a time when the number of time data points for the evaluation wafer is 100. The horizontal and vertical axes represent the time numbers (sampling data numbers) in the spectral time series data of the DB and the evaluation wafer, respectively. The straight line portion of this similar time path is the section where similar spectra are found, and the width of this section at DB is calculated as ΔNdb(t), and the width at the wafer under evaluation is calculated as ΔNtest(t). The etching rate ER(t) at that time is calculated using the etching rate ERdb at DB using the following equation 6.

[0126]

[0127] Using the calculated etching rate ER(t) and the above-mentioned time delay Δt, the determined film thickness value d(t) was corrected in the film thickness / depth corrector 1604. The correction method was the same as in the seventh embodiment.

[0128] The error between the determined film thickness value corrected by the above method and the actual film thickness value is the same as in Fig. 17. As described above, in film thickness estimation using the above-mentioned correction of the determined film thickness and depth values, the estimated film thickness value and the actual film thickness value correspond with high precision, and end point determination using the estimated film thickness value of this method eliminates film thickness variations between wafers after processing, thereby achieving high-precision processing.

[0129] In this example, a description will be given of the results of estimating film thickness and depth when the correspondence between the film thickness and spectrum in the database is not corrected in the method of estimating film thickness and depth using correction of the determined values ​​of film thickness and depth in Example 7. The rest of the process is the same as in Example 1, and the description will be omitted in this example.

[0130] 9 is used as the database, and the time delay Δt is calculated by signal processing in the same manner as in Example 7. However, the correspondence between the film thickness and the spectrum in the database is not corrected and is used to estimate the film thickness and depth. The wafer to be evaluated for film thickness estimation using the database is the same as in Example 7.

[0131] In the film thickness estimation for each evaluation wafer, the correction of the determined film thickness and depth values ​​shown in FIG. 16 was performed in the same manner as in Example 7. The results of calculating the film thickness estimation error at the end point for each evaluation wafer are shown in FIG. 19 . When the correction of the determined film thickness values ​​of the present invention was performed, the film thickness estimation error averaged approximately 2 nm, but the variation was within 1 nm. On the other hand, when the correction of the determined film thickness values ​​was not performed, the average error was approximately 2 nm, and the variation was also large. Therefore, even when the correspondence between the film thickness in the database and the spectrum was not corrected as in this example, the correction of the determined film thickness values ​​of the present invention ensured that the estimated film thickness value and the actual film thickness corresponded with high accuracy, and endpoint determination using the corrected determined film thickness value eliminated film thickness variation between wafers after processing, thereby achieving high-precision processing.

[0132] Here, when the correspondence relationship between the film thickness and spectrum in the database is not corrected, an offset error occurs in the determined film thickness value. This is because not correcting the database results in a shift in the correspondence relationship between the film thickness and spectrum in the database, whereas correcting the determined film thickness value results in no shift in the correspondence relationship between the film thickness and spectrum on the evaluation target wafer. As a result, the shift in the correspondence relationship between the film thickness and spectrum in the database appears as an offset in the determined film thickness value.

[0133] In this example, a description will be given of the results of estimating film thickness and depth using a plurality of databases in the method for estimating film thickness and depth using correction of the determined values ​​of film thickness and depth in Example 7. Other aspects are the same as in Example 1, and therefore description thereof will be omitted in this example.

[0134] In this example, the five databases shown in FIG. 9 were used, and the time delay Δt was calculated by signal processing in the same manner as in Example 7, and the correspondence between the film thickness and the spectrum in each database was corrected. The wafer to be evaluated, for which film thickness was estimated using the database, was the same as in Example 7. In this example, film thickness estimation using multiple databases was performed in the following procedure. First, multiple pattern databases and spectral data of the wafer to be evaluated, supplied from the digital signal processor, were input to waveform comparator 302 in FIG. 3, and the waveform comparison results between each database and the spectral data of the wafer to be evaluated were supplied to optimal database determiner 306. The optimal database determiner 306 determined the optimal database to use for estimating the film thickness of the wafer to be evaluated, and information on the determination result was supplied to film thickness / depth determiner 304, which calculated the determined film thickness value at each time during etching.

[0135] The results of calculating the film thickness estimation error at the end point for each evaluation wafer are the same as those in Fig. 17. As described above, even in film thickness estimation using multiple databases, by using the correction of the determined values ​​of film thickness and depth according to the present invention, the estimated film thickness value and the actual film thickness value correspond with high precision, and end point determination using the film thickness estimated value of this method eliminates film thickness variations between wafers after processing, thereby achieving high-precision processing.

[0136] The invention made by the inventor has been specifically described above based on an embodiment, but it goes without saying that the present invention is not limited to the above embodiment and can be modified in various ways without departing from the gist of the invention.

[0137] 10 Vacuum processing chamber 12 Plasma 14 Sample stage 16 Processing object 18 Light source unit 22 Irradiation light 24 Reflected light 28 Detection unit 30 Film thickness / depth calculation unit 40 Control unit 44 End point determination unit 50 Optical system 55 Plasma processing unit 60 Host computer 70 Plasma processing device 80 Etching system 100 Digital signal processing unit 102 Film thickness / depth determination unit 104 Film thickness / depth correction unit 122 Waveform pattern database D1 Time series data supplied from the detection unit D2 Time series data supplied from the digital signal processing unit D3 Film thickness. Time series data supplied from the depth amount determination unit D4 Time series data supplied from the film thickness and depth amount correction unit 302 Waveform comparator 304 Film thickness and depth amount determiner 306 Optimum database determiner 308 Matching database calculator 310 Film thickness and depth calculation model generator 320 Spectrum supplied from the digital signal processing unit 322 Database supplied from the waveform pattern database

Claims

1. A plasma processing apparatus that processes a wafer to be processed, which is placed in a processing chamber inside a vacuum vessel, using plasma formed in the processing chamber, comprising: a photoreceiver that receives light reflected from the wafer to be processed during processing using the plasma; a signal processor that performs first signal processing on the intensity of the reflected light received by the photoreceiver and outputs the result; a database in which the intensity of light of a predetermined wavelength corresponds to the thickness of a film formed on a reference wafer that serves as a basis for determining film thickness; a data processor that performs second signal processing on the light intensity contained in the database and executes data processing to match the correspondence between the light intensity and the film thickness before and after the second signal processing; a determiner that determines the film thickness of the wafer to be processed from the intensity of the reflected light that has been subjected to the first signal processing by the signal processor based on the database; and a determiner that determines the end point of processing using the plasma based on the film thickness of the wafer to be processed determined by the determiner.

2. A plasma processing apparatus according to claim 1, wherein the data processor calculates the amount of time delay caused by the second signal processing, and corrects the correspondence between the light intensity and the film thickness based on the amount of time delay.

3. A plasma processing apparatus according to claim 2, wherein the light intensity and the film thickness contained in the database before correction are obtained based on the results of processing using the plasma on a sample wafer prepared in advance as the reference wafer.

4. A plasma processing apparatus according to claim 2, wherein the light intensity and the film thickness contained in the database before correction are obtained based on simulation results.

5. A plasma processing apparatus according to claim 1, wherein the determiner compares the light intensity contained in the plurality of databases with the intensity of the reflected light after the first signal processing to determine one database with a small amount of error, and determines the film thickness corresponding to the light intensity with a small amount of error from the intensity of the reflected light after the first signal processing in the determined one database as the film thickness of the wafer to be processed.

6. A plasma processing apparatus according to claim 1, wherein the determiner compares the light intensity contained in one of the databases with the intensity of the reflected light that has been subjected to the first signal processing, and determines the film thickness corresponding to the light intensity with the smallest amount of error as the film thickness of the wafer to be processed.

7. A plasma processing apparatus according to claim 1, further comprising a generator that generates a learning model based on a plurality of said databases, and said determiner determines the thickness of the film on the wafer to be processed based on said learning model.

8. A plasma processing apparatus that processes a wafer to be processed placed in a processing chamber inside a vacuum vessel using plasma formed in the processing chamber, comprising: a photoreceiver that receives reflected light from the wafer to be processed at a predetermined time during the processing; a detector that detects the remaining thickness of the film during processing of the wafer to be processed at the predetermined time by comparing the intensity of the reflected light of a predetermined wavelength received by the photoreceiver with a database in which the remaining thickness of a film formed on a reference wafer and the intensity of the predetermined wavelength of the reflected light from the reference wafer obtained during processing of the reference wafer are correlated; and a determiner that determines the end point of processing using the plasma based on the thickness of the film on the wafer to be processed detected by the detector, wherein the detector detects a second remaining thickness of the film by correcting a first remaining thickness of the film obtained based on a comparison of the intensity of the reflected light obtained during the processing with the database in accordance with the difference in the correspondence between the intensity of the reflected light and the thickness of the film before and after signal processing performed on the intensity of the reflected light.

9. A plasma processing apparatus according to claim 8, wherein the detector corrects the first remaining thickness based on the calculated amount of time delay caused by the signal processing.

10. A plasma processing apparatus according to claim 9, wherein the processing is an etching processing, and the detector corrects the first remaining thickness using a value of the product of the etching rate of the wafer being processed during the processing before the predetermined time and the time delay.

11. A plasma processing apparatus according to claim 8 or 9, wherein the reflected light intensity, the remaining film thickness and the correspondence therebetween contained in the database are obtained based on the results of performing the processing using the plasma on the reference wafer.

12. A plasma processing apparatus according to claim 8 or 9, wherein the intensity of the reflected light and the thickness of the film contained in the database before correction are obtained based on simulation results.

13. A plasma processing method performed in a plasma processing apparatus that processes a wafer to be processed placed in a processing chamber inside a vacuum vessel using plasma formed in the processing chamber, comprising: a signal processing step of performing first signal processing on the intensity of reflected light from the wafer to be processed obtained during processing using the plasma and outputting the result; a determination step of determining the thickness of a film on the wafer to be processed from the intensity of the reflected light that has been subjected to the first signal processing in the signal processing step, based on a database in which the intensity of light of a predetermined wavelength corresponds to the thickness of a film formed on a reference wafer that serves as a basis for film thickness determination; a determination step of judging the end point of processing using the plasma based on the thickness of the film on the wafer to be processed determined in the determination step; and a data processing step of performing second signal processing on the light intensity included in the database, and performing data processing to match the correspondence between the light intensity and the film thickness before and after the second signal processing.

14. An etching system comprising: an etching processing apparatus comprising: a light receiver that receives reflected light from a wafer to be processed during processing using plasma; a signal processor that performs first signal processing on the intensity of the reflected light received by the light receiver and outputs the result; a database in which the intensity of light of a predetermined wavelength corresponds to the thickness of a film formed on a reference wafer that is used as a basis for film thickness determination; a determiner that determines the thickness of the film on the wafer to be processed from the intensity of the reflected light that has been subjected to the first signal processing by the signal processor based on the database; and a determiner that judges the endpoint of the etching processing based on the thickness of the film on the wafer to be processed determined by the determiner; and a data processing apparatus that acquires the intensity of light of the predetermined wavelength and the thickness of the film formed on the reference wafer and generates the database, wherein the data processing apparatus performs second signal processing on the intensity of the light when generating the database, and executes data processing to match the correspondence between the intensity of the light and the thickness of the film before and after the second signal processing.