Film property measuring method and film property measuring device

Abstract

[Problem] To increase the reliability and throughput of measurement of the properties of a film formed on a substrate. [Solution] A substrate is illuminated with polarization components consisting of p-polarization and s-polarization of terahertz waves with mutually orthogonal planes of vibration. Time waveforms of the electric field intensities of the p-polarization and the s-polarization are acquired from a time waveform of the electric field intensity of a polarization component reflected from the substrate. A spectrum of an actually measured amplitude ratio of the p-polarization and the s-polarization, and a spectrum of an actually measured phase difference between the p-polarization and the s-polarization are acquired. A spectrum of a calculated amplitude ratio of the p-polarization and the s-polarization, and a spectrum of a calculated phase difference between the p-polarization and the s-polarization are acquired through simulation using an optical model of the substrate. The properties of a film formed on the substrate are measured by fitting involving at least alignment of the spectrum of the calculated amplitude ratio with the spectrum of the actually measured amplitude ratio. Among a plurality of measurement values used in the measurement of the properties of the film, those with high reliability are employed.

Description

Method and apparatus for measuring film properties

[0001] This disclosure relates to a film property measurement method and a film property measurement apparatus.

[0002] In recent years, ellipsometry using terahertz waves has become known as a technique that can non-destructively measure the electrical properties of films formed on substrates, such as epitaxial films. Ellipsometry is an analytical method that measures the polarization state of reflected light produced when incident terahertz waves incident on a substrate are reflected at the substrate interface (for example, the boundary between films or the boundary between a film and a substrate) (see, for example, Patent Document 1).

[0003] Of the incident light, the polarization component that vibrates parallel to the plane of incidence perpendicular to the substrate surface is p-polarized light, and the polarization component that vibrates perpendicular to the plane of incidence is s-polarized light. The p-polarized and s-polarized light incident on the substrate are linearly polarized such that their phase and amplitude are the same. When these p-polarized and s-polarized lights are reflected at the substrate interface, their respective phases and amplitudes change depending on the optical constants of the film (refractive index and extinction coefficient for each wavelength) and the film thickness.

[0004] Then, p-polarized and s-polarized light with changed phase and amplitude are detected, and these p-polarized and s-polarized light with changed phase and amplitude constitute elliptically polarized light. Furthermore, the change in the polarization state of the reflected light is expressed as the spectrum of the amplitude ratio Ψ of the detected p-polarized and s-polarized light (hereinafter referred to as "measured amplitude ratio Ψ") and the spectrum of the phase difference Δ of the measured p-polarized and s-polarized light (hereinafter referred to as "measured phase difference Δ").

[0005] On the other hand, an optical model is constructed that assumes the film thickness and optical constants of each film to match the substrate on which the film properties are measured, and the amplitude ratio Ψ spectrum and the phase difference Δ spectrum are calculated from the optical model through simulation. In the simulation, changing the film thickness and optical constants of each film changes the calculated amplitude ratio Ψ spectrum (hereinafter referred to as "calculated amplitude ratio Ψ") and the calculated phase difference Δ spectrum (hereinafter referred to as "calculated phase difference Δ").

[0006] In ellipsometry, the simulation is fitted by changing the film thickness and optical constants of each film until the spectrum of the measured amplitude ratio Ψ and the spectrum of the calculated amplitude ratio Ψ, and the spectrum of the measured phase difference Δ and the spectrum of the calculated phase difference Δ, nearly match. Then, the film thickness of each film in the simulation when the spectra nearly match is considered to be the actual film thickness, and the optical constants of each film in the simulation are considered to be the actual optical constants of each film. Furthermore, since the electrical properties of each film can be determined by the optical constants, the electrical properties of each film can be determined based on the optical constants that have been considered to be the actual optical constants of each film through fitting. This makes it possible to measure the electrical properties of each film formed on the substrate, such as carrier concentration.

[0007] Japanese Patent Publication No. 2012-208098

[0008] The technology disclosed herein improves the reliability and throughput of characterization of films formed on substrates.

[0009] One aspect of the technology described herein is a substrate film characteristic measurement method for measuring the characteristics of a film formed on a substrate, comprising: injecting a terahertz wave incident light into the substrate, comprising a polarization component consisting of p-polarized light vibrating parallel to an incident plane perpendicular to the surface of the substrate and s-polarized light vibrating perpendicular to the incident plane; detecting the time waveform of the electric field intensity of the polarization component reflected from the substrate; obtaining the time waveforms of the electric field intensity of p-polarized and s-polarized light from the time waveform of the electric field intensity of the detected polarization component; and performing a Fourier transform on the obtained time waveforms of the electric field intensity of p-polarized and s-polarized light to obtain a spectrum of the measured amplitude ratio, which is the spectrum of the amplitude ratio of the electric field intensity of p-polarized and s-polarized light, and the phase difference spectrum of p-polarized and s-polarized light. The spectral values ​​of the measured phase difference, which is a parameter, are obtained, an optical model of the substrate is constructed, and a calculated amplitude ratio spectrum, which is the spectral values ​​of the amplitude ratio of the electric field intensities of the p-polarized and s-polarized light, and a calculated phase difference spectrum, which is the spectral values ​​of the phase difference between the p-polarized and s-polarized light, are obtained by simulation using the optical model, the properties of the film formed on the substrate are measured by fitting, which involves matching the spectral values ​​of the calculated amplitude ratio to the spectral values ​​of the measured amplitude ratio, and matching the spectral values ​​of the calculated phase difference to the spectral values ​​of the measured phase difference, and a highly reliable measured value is used from among a plurality of measured values ​​used when measuring the properties of the film.

[0010] The technology described herein can improve the reliability and throughput of characteristic measurement of films formed on a substrate.

[0011] This is a schematic diagram showing the configuration of a film characteristic measuring device according to one embodiment of the technology described herein. This is a diagram for explaining the fitting performed by the control unit of the ellipsometer. This is a diagram for explaining the fitting performed by the control unit of the ellipsometer. This is a graph showing the variation in carrier concentration measured by fitting in the ellipsometer. This is a graph plotting the measured amplitude ratio tangent tanΨ at a specific frequency selected when the tanΨ stability index g becomes smaller than the threshold α. This is a graph plotting the measured amplitude ratio tangent tanΨ at a specific frequency selected when the tanΨ stability index g becomes smaller than the threshold α. This is a flowchart showing the carrier concentration measurement process as a film characteristic measuring method according to the first embodiment. This is a diagram showing the time waveform of the electric field intensity of the polarization component detected for each rotation angle of the rotating polarizer. This is a graph showing the fluctuation of electric field intensity in the time waveform of the raw (polarized component) electric field intensity in the preceding stage, the main observation period, and the subsequent stage. This is a graph showing the fluctuation of electric field intensity in the time waveform of the raw (polarized component) electric field intensity in the preceding stage, the main observation period, and the subsequent stage. This graph shows the fluctuations in the pre-measurement, main observation period, and post-measurement phases of 128 carrier concentration measurements. This graph shows the fluctuations in the pre-measurement, main observation period, and post-measurement phases of 128 carrier concentration measurements. This graph shows the fluctuations in the pre-measurement, main observation period, and post-measurement phases of 128 carrier concentration measurements. This graph shows the relationship between the average value of the electric field intensity fluctuations for each period and the carrier concentration in 128 carrier concentration measurements. This graph shows the relationship between the average value of the electric field intensity fluctuations for each period and the carrier concentration in 128 carrier concentration measurements. This graph shows the relationship between the average value of the electric field intensity fluctuations for each period and the carrier concentration in 128 carrier concentration measurements. The change in the tanΨ stability index g and the polarization component E C This graph compares the change in the fluctuation pattern of the preceding stage in the time waveform of the electric field intensity. It shows the change in the tanΨ stability index g and the polarization component E. CThis graph compares the change in the fluctuation pattern of the preceding stage in the time waveform of the electric field intensity. This figure explains how to set the fluctuation generation threshold. This flowchart shows the time waveform detection process as a film characteristic measurement method according to the second embodiment. This graph shows the carrier concentration of the epitaxial film measured using the mercury probe C-V method and ellipsometry, respectively. This graph shows the relationship between the brightness and frequency of measured p-polarized and measured s-polarized light. This graph shows the relationship between the brightness and frequency of measured p-polarized and measured s-polarized light. This graph shows the relationship between the sensitivity of the calculated amplitude ratio tangent tanΨ to the change in carrier concentration in the optical model and the frequency. Polarization component E C This graph shows the change in brightness of measured p-polarization and measured s-polarization when detection is repeated. Polarization component E C This graph shows the change in brightness of measured p-polarization and measured s-polarization when detection is repeated. Polarization component E C This graph shows the change in brightness of measured p-polarization and measured s-polarization when detection is repeated. Polarization component E CThis graph shows the change in brightness of measured p-polarized and measured s-polarized light when detection is repeated. This figure shows an optical model consisting of an epitaxial film, a buffer layer, and a substrate. This graph shows the amount of change in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter is changed by ±10% from the median for all parameters. This graph shows the amount of change in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter is changed by ±10% from the median for all parameters. This graph shows the relationship between carrier concentration and mobility. This graph shows the amount of change in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter other than the film thickness of the epitaxial film and the carrier concentration and mobility of the substrate is changed by ±10% from the median. This graph shows the amount of change in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter other than the film thickness of the epitaxial film and the carrier concentration and mobility of the substrate is changed by ±10% from the median. This graph shows the variation in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter other than the epitaxial film thickness, the carrier concentration and mobility of the substrate, and the buffer layer thickness is changed by approximately ±10% from the median value. This graph shows the variation in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter other than the epitaxial film thickness, the carrier concentration and mobility of the substrate, and the buffer layer thickness is changed by approximately ±10% from the median value. This flowchart shows the parameter measurement process as a film characteristic measurement method according to the third embodiment. This figure shows the time waveform of the signal intensity of the measurement signal at a certain rotation angle of the rotating polarizer. This graph shows the standard deviation of the signal intensity of the measurement signal in the preceding stage portion at each rotation angle of the rotating polarizer. This graph shows the relationship between the noise characteristic value of the measurement signal in the preceding stage portion of the rotating polarizer at a certain rotation angle and the dose amount of the epitaxial film obtained by fitting, when the dose amount of the epitaxial film on the substrate was measured 128 times.This graph shows the relationship between the noise characteristic values ​​of the measurement signal in the preceding stage of a certain rotation angle of the rotating polarizer and the dose amount of the epitaxial film obtained by fitting, when the dose amount of the epitaxial film on the substrate was measured 128 times. This graph shows the relationship between the noise characteristic values ​​of the measurement signal in the preceding stage of a certain rotation angle of the rotating polarizer and the dose amount of the epitaxial film obtained by fitting, when the dose amount of the epitaxial film on the substrate was measured 128 times. This graph shows the relationship between the noise characteristic values ​​of the measurement signal in the main observation period of a certain rotation angle of the rotating polarizer and the dose amount of the epitaxial film obtained by fitting, when the dose amount of the epitaxial film on the substrate was measured 128 times. This graph shows the relationship between the noise characteristic values ​​of the measurement signal in the main observation period of a certain rotation angle of the rotating polarizer and the dose amount of the epitaxial film obtained by fitting, when the dose amount of the epitaxial film on the substrate was measured 128 times. This graph shows the relationship between the noise characteristic value of the measurement signal during the main observation period at a certain rotation angle of the rotating polarizer and the dose amount of the epitaxial film obtained by fitting, when the dose amount of the epitaxial film on the substrate was measured 128 times. This graph shows the relationship between the maximum signal intensity of the measurement signal in the preceding stage at a certain rotation angle of the rotating polarizer and the maximum signal intensity of the measurement signal in the main observation period for the same measurement run. This figure compares the distribution range of the maximum signal intensity of the measurement signal in the preceding stage at a certain rotation angle of the rotating polarizer and the distribution range of the maximum signal intensity of the measurement signal in the main observation period. This figure compares the distribution range of the maximum signal intensity of the measurement signal in the preceding stage at a certain rotation angle of the rotating polarizer and the distribution range of the maximum signal intensity of the measurement signal in the main observation period. This is a flowchart of the stable time waveform acquisition process as a film characteristic measurement method according to the fourth embodiment.

[0012] Figure 1 is a schematic diagram showing the configuration of a film property measurement device according to one embodiment of the technology described herein. In Figure 1, the ellipsometer 10, which is a film property measurement device that performs ellipsometry, comprises a femtosecond laser oscillator 11, a beam splitter 12, a terahertz wave generator 13, an incident light elliptical mirror 14, and an incident light polarizer 15 as its components. The ellipsometer 10 also comprises a rotating polarizer 16, a reflected light polarizer 17, a reflected light elliptical mirror 18, a terahertz wave detector 19, an optical delay mechanism 20, and a control unit 21 as other components.

[0013] The femtosecond laser oscillator 11 emits femtosecond light pulses to the beam splitter 12. The beam splitter 12 splits the laser pulses emitted from the femtosecond laser oscillator 11 into two beams: a pump beam and a probe beam. The pump beam is directed to the terahertz wave generator 13, which generates terahertz light from the pump beam and radiates the terahertz waves toward the incident light elliptical mirror 14. The frequency of the terahertz waves radiated by the terahertz wave generator 13 is, for example, 0.1 to 5 THz. The incident light elliptical mirror 14 focuses the terahertz waves radiated from the terahertz wave generator 13 toward the substrate S.

[0014] An incident light polarizer 15 is interposed between the incident light elliptical mirror 14 and the substrate S. The incident light polarizer 15 transmits the linearly polarized component of the terahertz waves emitted from the terahertz wave generator 13, which includes p-polarized light, a polarization component that vibrates parallel to the incident plane perpendicular to the surface of the substrate S, and s-polarized light, a polarization component that vibrates perpendicular to the incident plane. The incident light polarizer 15 is configured such that the phase and amplitude of the transmitted p-polarized and s-polarized light are the same.

[0015] When p-polarized light and s-polarized light are incident on the substrate S, they are reflected at the interface of the substrate S. The substrate S has, for example, a buffer layer formed on its surface and an epitaxial film made of SiC (silicon carbide) formed on the buffer layer. Therefore, the interface of the substrate S corresponds to the surface of the epitaxial film, the boundary between the epitaxial film and the buffer layer, and the boundary between the buffer layer and the surface of the substrate. At this time, the reflected light from each boundary and surface interferes, and the amplitudes of the p-polarized light and the s-polarized light change. Also, since the speed of the light transmitted through the buffer layer and the epitaxial film changes according to the refractive indices of the buffer layer and the epitaxial film, the phases of the p-polarized light and the s-polarized light also change. That is, since the reflected p-polarized light and s-polarized light have different phases and amplitudes from each other, the reflected p-polarized light and s-polarized light constitute elliptically polarized light.

[0016] The rotating polarizer 16 rotates around the optical axis of the reflected light from the substrate S, for example, in 15° increments, and transmits a polarization component according to the rotation angle from the reflected light. The reflected light polarizer 17 stabilizes the detection accuracy of the polarization component transmitted through the rotating polarizer 16.

[0017] Here, let the reflected p-polarized light be E p and the reflected s-polarized light be E s and let the rotation angle of the rotating polarizer 16 be θ B . Then, the polarization component E B transmitted through the rotating polarizer 16 is represented by the following formula.

[0018]

[0019] Also, if the transmission angle of the reflected light polarizer 17 is θ C , then among the polarization components E B , the polarization component E C transmitted through the reflected light polarizer 17 is represented by the following formula. And the reflected light elliptical mirror 18 condenses the polarization component E C toward the terahertz wave detector 19, and the terahertz wave detector 19 detects the electric field strength of the polarization component E C . C in the following formula is a stray light component that does not depend on the rotating polarizer 16.

[0020]

[0021] In the present embodiment, the transmission angle θ of the reflected light polarizer 17C This is set to -π / 4. Therefore, the above polarization component E C In the equation, cosθ c This becomes 1 / √2, and sinθ c This becomes -1 / √2, and the polarization component E C It is shown by the following formula. In the following formula, A, B, and C' are coefficients.

[0022]

[0023] Polarization component E of the above formula C In θ, B When θ is 0, it indicates p-polarization transmitted through the reflected light polarizer 17, and θ B When is π / 2, it will show s-polarized light transmitted through the reflected light polarizer 17. Therefore, by calculating coefficients A and B, the p-polarized and s-polarized light transmitted through the reflected light polarizer 17 can be determined from the above equation. That is, the polarization component E C By removing the stray light component that does not depend on the angle of the rotating polarizer 16 included in the observational data, appropriate p-polarization and s-polarization can be obtained.

[0024] Furthermore, the probe light split by the beam splitter 12 is incident on the optical delay mechanism 20. The optical delay mechanism 20 has a mirror (retro reflector) 22 that reflects and folds back the probe light, and changes the optical path length of the probe light by moving the mirror 22 parallel to the probe light. Meanwhile, the polarization component E transmitted through the reflected light polarizer 17, which is terahertz light generated from the pump light C Since the optical path length does not change, the optical delay mechanism 20 directs the polarization component E to the probe light. C A time delay is applied to it.

[0025] Then, the probe light with time delay added has polarization component E C It arrives at the terahertz wave detector 19 later than the probe light. The terahertz wave detector 19 uses the probe light with an arbitrary time delay as a reference wave and the polarization component E C The time waveform of the electric field intensity can be detected. In addition, the terahertz wave detector 19 can detect the detected polarization component E CA current proportional to the electric field strength flows, and the terahertz wave detector 19 transmits a signal converted from these currents to the control unit 21. The control unit 21 obtains the time waveforms of the p-polarized and s-polarized electric field strengths from the received signal. Subsequently, the polarization component E detected by the terahertz wave detector 19 is used. C The p-polarized and s-polarized signals obtained from these sources are referred to as "measured p-polarized signals" and "measured s-polarized signals," respectively.

[0026] In the ellipsometer 10, probe light and pump light (polarization component E C Multiple mirrors are placed along the optical path of the probe light and pump light, and by changing the direction of propagation of the probe light and pump light with each mirror, the ellipsometer 10 is prevented from becoming unnecessarily large.

[0027] The control unit 21 obtains the spectrum of the tangent tanΨ of the amplitude ratio of the electric field strengths of the measured p-polarized and measured s-polarized fields (hereinafter referred to as "measured amplitude ratio tangent tanΨ") by performing a Fourier transform on the time waveforms of the electric field strengths of the measured p-polarized and measured s-polarized fields. The control unit 21 also obtains the spectrum of the phase difference Δ between the measured p-polarized and measured s-polarized fields (hereinafter referred to as "measured phase difference Δ") by performing a Fourier transform on the time waveforms of the electric field strengths of the measured p-polarized and measured s-polarized fields.

[0028] In addition, in the ellipsometer 10, the polarization component E C Since the detection of the time waveforms of the electric field intensity (measured p-polarization and measured s-polarization) is repeated, multiple time-continuous spectra of the measured amplitude ratio tangent tanΨ are obtained, and multiple time-continuous spectra of the measured phase difference Δ are obtained.

[0029] The measured amplitude ratio tangent tanΨ and measured phase difference Δ are indicators of the change in polarization state. The measured amplitude ratio tangent tanΨ and measured phase difference Δ are defined by the following equations. Note that in the following equations, r p is the Fresnel amplitude reflection coefficient for p-polarized light, and r s is the Fresnel amplitude reflection coefficient for s-polarized light. Also, Δ rp Δ is the phase of the measured p-polarization, rs This represents the phase of the measured s-polarization.

[0030]

[0031] Furthermore, the control unit 21 constructs an optical model of the substrate S. At this time, the control unit 21 assumes the film thickness of the buffer layer and epitaxial film, as well as optical constants (refractive index and extinction coefficient for each wavelength), as fitting variables. Then, using the assumed fitting variables, the control unit 21 calculates the spectrum of the amplitude ratio tangent tanΨ and the spectrum of the phase difference Δ by simulation using the optical model. Hereafter, the amplitude ratio tangent tanΨ calculated from the optical model will be referred to as "calculated amplitude ratio tangent tanΨ," and the phase difference Δ calculated from the optical model will be referred to as "calculated phase difference Δ." Subsequently, the control unit 21 performs fitting by changing the fitting variables until the spectrum of the measured amplitude ratio tangent tanΨ and the spectrum of the calculated amplitude ratio tangent tanΨ, and the spectrum of the measured phase difference Δ and the spectrum of the calculated phase difference Δ, are in close agreement.

[0032] Figures 2A and 2B illustrate the fitting performed by the control unit 21. In Figures 2A and 2B, solid lines represent the measured amplitude ratio tangent tanΨ and measured phase difference Δ, while dashed lines represent the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ. The upper spectral curve represents the phase difference Δ, and the lower spectral curve represents the amplitude ratio tangent tanΨ.

[0033] First, the control unit 21 compares the spectrum of the calculated amplitude ratio tangent tanΨ, calculated using the assumed fitting variables, with the spectrum of the measured amplitude ratio tangent tanΨ, and also compares the spectrum of the calculated phase difference Δ with the spectrum of the measured phase difference Δ.

[0034] If the spectrum of the calculated amplitude ratio tangent tanΨ does not match the spectrum of the measured amplitude ratio tangent tanΨ, and furthermore, the spectrum of the calculated phase difference Δ does not match the spectrum of the measured phase difference Δ (Figure 2A), the fitting variables are changed. Then, the spectrum of the calculated amplitude ratio tangent tanΨ calculated using the changed fitting variables is compared with the spectrum of the measured amplitude ratio tangent tanΨ, and the spectrum of the calculated phase difference Δ is compared with the spectrum of the measured phase difference Δ. The control unit 21 repeats changing the fitting variables until the spectrum of the calculated amplitude ratio tangent tanΨ almost matches the spectrum of the measured amplitude ratio tangent tanΨ, and furthermore, the spectrum of the calculated phase difference Δ almost matches the spectrum of the measured phase difference Δ. The degree of agreement between the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the measured amplitude ratio tangent tanΨ, and the degree of agreement between the spectrum of the calculated phase difference Δ and the spectrum of the measured phase difference Δ are determined, for example, by the least squares method.

[0035] Then, when the spectrum of the calculated amplitude ratio Ψ closely matches the spectrum of the measured amplitude ratio tangent tanΨ, and the spectrum of the calculated phase difference Δ closely matches the spectrum of the measured phase difference Δ, the control unit 21 outputs the fitting variables at that time as the fitting result. The control unit 21 also identifies the film thickness and optical constants of the buffer layer and epitaxial film as the actual film thickness and optical constants of the buffer layer and epitaxial film, based on the fitting result. Since the optical constants change according to the electrical characteristics of each film, the electrical characteristics (dielectric constant and carrier concentration) of the buffer layer and epitaxial film can be calculated inversely from the identified optical constants. As a result, the control unit 21 can measure parameters such as the film thickness and carrier concentration of the buffer layer and epitaxial film.

[0036] Furthermore, the fitting described above is performed by the terahertz wave detector 19, which detects the polarization component E C The time waveform of the electric field intensity is detected, and this process is executed each time the spectrum of the measured amplitude ratio tangent tanΨ or the measured phase difference Δ is obtained.

[0037] Next, a first embodiment of the technology relating to this disclosure will be described. However, p-polarized and s-polarized light in the ellipsometer 10 is susceptible to disturbances, and for example, even minute rattle in the rotation mechanism of the rotating polarizer 16 can cause variations in the amount of light (electric field intensity) transmitted through the rotating polarizer 16. As a result, the spectra of the measured amplitude ratio tangent tanΨ and measured phase difference Δ obtained from the time waveforms of the measured p-polarized and measured s-polarized electric field intensity will vary, and consequently, the carrier concentration measured by fitting will also vary, as has been confirmed by the applicant.

[0038] Figure 3 is a graph showing the variation in carrier concentration measured by fitting in the ellipsometer 10. The graph in Figure 3 plots the carrier concentration when the carrier concentration was measured 128 times by fitting. As shown in the graph in Figure 3, even when using the same substrate S, the minimum measured carrier concentration was approximately 3.7 × E 15 (cm -3 ) whereas the maximum carrier concentration measured was approximately 8.2 × E 15 (cm -3 As a result, the measured carrier concentrations vary considerably.

[0039] Conventionally, the ellipsometer 10 was used to measure numerous carrier concentrations by repeatedly performing fittings, and the average value of all measured carrier concentrations was obtained as the carrier concentration measurement result. This method is expected to reduce the influence of disturbances on carrier concentrations through averaging, thereby improving the reliability of the carrier concentration measurement results.

[0040] However, this method has the drawback of reducing throughput because it involves numerous fittings. On the other hand, simply reducing the number of carrier concentration measurements by fitting in order to suppress the decrease in throughput also presents a problem: the reliability of the carrier concentration measurement results obtained by averaging decreases.

[0041] In response to this, the film characteristic measurement method according to the first embodiment selects the measured amplitude ratio tangent tanΨ (measured value) at a specific frequency that is not significantly affected by external disturbances.

[0042] In the ellipsometer 10, when fitting is repeated, the polarization component E C The time waveform of the electric field intensity is repeatedly detected, and the spectra of the measured amplitude ratio tangent tanΨ and the measured phase difference Δ are also repeatedly obtained. At this time, for example, a certain polarization component E C If the system is significantly affected by external disturbances, it is thought that the spectra of multiple measured amplitude ratio tangents tanΨ and measured phase difference Δ, which are acquired continuously over time, will vary. Therefore, the applicant considered using the spectrum of the measured amplitude ratio tangent tanΨ with less variation among the multiple spectra of measured amplitude ratio tangents tanΨ acquired continuously over time, that is, the spectrum of the measured amplitude ratio tangent tanΨ that is not significantly affected by external disturbances.

[0043] Specifically, the applicant focused on the measured amplitude ratio tangent tanΨ at a specific frequency in the spectrum of the measured amplitude ratio tangent tanΨ, and considered fitting using the measured amplitude ratio tangent tanΨ of the median specific frequency when the amount of change in the measured amplitude ratio tangent tanΨ of three temporally consecutive specific frequencies is below a threshold. In this case, the amount of change in the measured amplitude ratio tangent tanΨ of three temporally consecutive specific frequencies used in the study (hereinafter referred to as the "tanΨ stability index") g(m) is the sum of the absolute values ​​of the differences between the measured amplitude ratio tangent tanΨ of the median specific frequency and the measured amplitude ratio tangent tanΨ of the preceding and succeeding specific frequencies, and corresponds to the variability of the measured amplitude ratio tangent tanΨ of the median specific frequency, and is expressed by the following formula. Here, m is the measurement time, ABS indicates the absolute value, and tanΨ(m) is the measured amplitude ratio tangent tanΨ at a specific frequency in the spectrum of the measured amplitude ratio tangent tanΨ for measurement time m.

[0044]

[0045] Then, the measured amplitude ratio tangent tanΨ(m) at specific frequencies was selected when the tanΨ stability index g(m) became smaller than the threshold α. Because the tanΨ stability index g(m) for these selected specific frequencies is small, it is considered that they are not significantly affected by disturbances.

[0046] Figures 4A and 4B are graphs plotting selected measured amplitude ratio tangent tanΨ at specific frequencies when the tanΨ stability index g is smaller than the threshold α. Figure 4A shows the case where the threshold α is set to 0.002, and Figure 4B shows the case where the threshold α is set to 0.001. As shown in Figures 4A and 4B, the premise for selecting measured amplitude ratio tangent tanΨ at specific frequencies is that the polarization component E C The time waveform of the electric field intensity was detected 128 times, and the spectrum of the measured amplitude ratio tangent tanΨ was also acquired 128 times.

[0047] As shown in Figure 4A, when the threshold α was set to 0.002, the variation (difference between maximum and minimum values) of the measured amplitude ratio tangent tanΨ for the selected specific frequencies was 0.0047. At this time, the variation (difference between maximum and minimum values) of the carrier concentration measured by fitting using the spectrum corresponding to the measured amplitude ratio tangent tanΨ for the selected specific frequencies was ±0.15 × E 15 (cm -3 )

[0048] Furthermore, as shown in Figure 4B, when the threshold α was set to 0.001, the variation (difference between maximum and minimum values) of the measured amplitude ratio tangent tanΨ for the selected specific frequencies was 0.0023. At this time, the variation (difference between maximum and minimum values) of the carrier concentration measured by fitting using the spectrum corresponding to the measured amplitude ratio tangent tanΨ for the selected specific frequencies was ±0.07 × E 15 (cm -3 )

[0049] On the other hand, polarization component E CThe variation (difference between maximum and minimum values) of the measured amplitude ratio tangent tanΨ at a specific frequency in the spectrum of all measured amplitude ratio tangent tanΨ obtained by repeating the detection of the time waveform of the electric field intensity 128 times was 0.009. At this time, the variation (difference between maximum and minimum values) of the carrier concentration measured by fitting using the spectrum of all measured amplitude ratio tangent tanΨ was ±0.28 × E 15 (cm -3 )

[0050] From the above, it was confirmed that the variation in the measured amplitude ratio tangent tanΨ of a specific frequency selected based on the tanΨ stability index g was smaller than the variation in the measured amplitude ratio tangent tanΨ of all specific frequencies obtained. Furthermore, it was confirmed that the variation in carrier concentration measured using the spectrum corresponding to the measured amplitude ratio tangent tanΨ of a specific frequency selected based on the tanΨ stability index g was smaller than the variation in carrier concentration measured using the spectrum of all measured amplitude ratio tangent tanΨ. In other words, it was found that by selecting the measured amplitude ratio tangent tanΨ of a specific frequency that is not significantly affected by disturbances based on the tanΨ stability index g, the variation in the measured carrier concentration can be suppressed and the reliability of the carrier concentration measurement results can be improved.

[0051] Furthermore, the reliability of the measured amplitude ratio tangent tanΨ at a specific frequency that falls within the variation of the measured amplitude ratio tangent tanΨ at multiple specific frequencies selected based on the tanΨ stability index g described above is considered high. Therefore, once the variation of the measured amplitude ratio tangent tanΨ at a selected specific frequency is obtained, it is considered that the reliability of the measured carrier concentration can be improved by performing fitting using the spectrum corresponding to the measured amplitude ratio tangent tanΨ at a specific frequency that falls within that variation in subsequent measurements.

[0052] Figure 5 is a flowchart showing the carrier concentration measurement process as a film characteristic measurement method according to the first embodiment. The carrier concentration measurement process in Figure 5 is performed in the ellipsometer 10 by the control unit 21 executing a program. In addition, before executing the carrier concentration measurement process in Figure 5, the control unit 21 determines the polarization component E C The time waveform of the electric field intensity is detected many times to obtain the spectra of many measured amplitude ratio tangents tanΨ, and the variation of measured amplitude ratio tangents tanΨ at a specific frequency that is not significantly affected by disturbances, selected based on the tanΨ stability index g, is obtained (hereinafter referred to as the "reference variation of tanΨ"). The threshold α of the tanΨ stability index g used to select measured amplitude ratio tangents tanΨ at a specific frequency for obtaining the reference variation of tanΨ is not particularly limited, but for example, 0.001 or 0.002 as described above can be used.

[0053] In the process shown in Figure 5, first, the femtosecond laser oscillator 11 emits a femtosecond light pulse, and the polarization component E generated at this time C Step S51 detects the time waveform of the electric field intensity and then obtains the time waveforms of the electric field intensity in p-polarization and s-polarization. Next, a new spectrum of the measured amplitude ratio tangent tanΨ is obtained from the time waveforms of the electric field intensity in p-polarization and s-polarization, and the measured amplitude ratio tangent tanΨ at a specific frequency is obtained from the obtained spectrum of the measured amplitude ratio tangent tanΨ (Step S52). Finally, it is determined whether the obtained measured amplitude ratio tangent tanΨ at a specific frequency falls within the reference variation of tanΨ (Step S53).

[0054] In step S53, if it is determined that the measured amplitude ratio tangent tanΨ at a specific frequency obtained does not fall within the reference variation of tanΨ, the measurement cycle is increased by one and the process returns to step S51 (step S54). Steps S51 and S52 are repeated to obtain the measured amplitude ratio tangent tanΨ at a specific frequency from the spectrum of the new measured amplitude ratio tangent tanΨ.

[0055] In step S53, if it is determined that the measured amplitude ratio tangent tanΨ at a specific frequency obtained falls within the reference variation of tanΨ, fitting is performed using the spectrum corresponding to the measured amplitude ratio tangent tanΨ at the specific frequency obtained, and the carrier concentration is measured (step S55). After that, this process is terminated.

[0056] According to the carrier concentration measurement process shown in Figure 5, fitting is performed using the spectrum corresponding to the measured amplitude ratio tangent tanΨ at a specific frequency that falls within the variation of the measured amplitude ratio tangent tanΨ at that specific frequency, which is not significantly affected by disturbances. In other words, fitting is performed using the spectrum of a highly reliable measured amplitude ratio tangent tanΨ. This improves the reliability of carrier concentration measurements.

[0057] Furthermore, according to the carrier concentration measurement process shown in Figure 5, if an actual amplitude ratio tangent tanΨ that falls within the reference variation of tanΨ is obtained, subsequent acquisition of the actual amplitude ratio tangent tanΨ is not performed, thus improving throughput.

[0058] Incidentally, as mentioned above, since the spectrum of the measured amplitude ratio tangent tanΨ is obtained from the time waveforms of the electric field intensity of measured p-polarized and measured s-polarized signals, the applicant also confirmed the relationship between the stability of the signal values ​​(luminance) of measured p-polarized and measured s-polarized signals and the reliability of the measured amplitude ratio tangent tanΨ at a specific frequency. Specifically, the stability indices of measured p-polarized and measured s-polarized signals corresponding to the measured amplitude ratio tangent tanΨ at a specific frequency where the tanΨ stability index g is smallest were confirmed. The stability index of measured p-polarized signals is the amount of change in luminance of three measured p-polarized signals obtained continuously in time, and, similar to the tanΨ stability index g, is expressed as the sum of the absolute values ​​of the differences between the luminance of the median measured p-polarized signal and the luminances of the preceding and succeeding measured p-polarized signals. Furthermore, the stability index for measured s-polarization is the change in brightness of three measured s-polarizations acquired consecutively over time, and, similar to the tanΨ stability index g, is expressed as the sum of the absolute values ​​of the differences between the brightness of the median measured s-polarization and the preceding and succeeding measured s-polarizations.

[0059] At this time, the applicant confirmed that the stability indices of the measured p-polarized and measured s-polarized light corresponding to the measured amplitude ratio tangent tanΨ with the smallest tanΨ stability index g were the smallest. In other words, the applicant confirmed that there is a correlation between the luminance stability of the measured p-polarized and measured s-polarized light and the reliability of the measured amplitude ratio tangent tanΨ at a specific frequency. Therefore, in the film characteristic measurement method according to this embodiment, it was considered that a highly reliable carrier concentration can be measured by selecting measured p-polarized and measured s-polarized light with high luminance stability and fitting the spectra of the measured amplitude ratio tangent tanΨ corresponding to these measured p-polarized and measured s-polarized light.

[0060] Furthermore, in this embodiment, a highly reliable spectrum of the measured amplitude ratio tangent tanΨ was obtained based on the variation of the measured amplitude ratio tangent tanΨ at specific frequencies that are not significantly affected by disturbances, selected based on the tanΨ stability index g, and fitting was performed using this spectrum of the measured amplitude ratio tangent tanΨ.

[0061] However, fitting for measuring carrier concentration can also be performed using the spectrum of the measured phase difference Δ. Therefore, first, measured phase difference Δ at specific frequencies that are not significantly affected by disturbances is selected based on the stability index of the measured phase difference Δ. Then, based on the variability of the selected measured phase difference Δ, a highly reliable spectrum of the measured phase difference Δ is obtained, and by performing fitting using the obtained highly reliable spectrum of the measured phase difference Δ, it was thought that a reliable carrier concentration could be measured.

[0062] Next, a second embodiment of the technology relating to this disclosure will be described. In the ellipsometer 10, in one measurement of the carrier concentration in the terahertz wave detector 19, the rotating polarizer 16 rotates once in increments of 15°, for example. At this time, the mirror 22 of the optical delay mechanism 20 is moved for each rotation angle of the rotating polarizer 16, thereby polarizing the polarization component E C The time waveform of the electric field intensity (raw electric field intensity) is detected. Therefore, when the rotating polarizer 16 rotates once during one carrier concentration measurement, 25 polarization components E are detected, as shown in Figure 6. CThe time waveform of the electric field intensity is detected. When the carrier concentration is measured 128 times in the ellipsometer 10, a total of 3200 polarization components E are detected. C The time waveform of the electric field intensity is detected.

[0063] Polarization component E C The time waveform of the electric field intensity is composed of a preliminary stage detected before the incident light enters the substrate S, a main observation period stage detected when the incident light enters the substrate S, and a post-stage stage detected after the incident light enters the substrate S, in accordance with the movement of the mirror 22 of the optical delay mechanism 20 (Figure 6). Here, detection of the preliminary stage requires 8 ps (picoseconds), detection of the main observation period stage requires 10 ps, ​​and detection of the post-stage stage requires 10 ps.

[0064] Incidentally, since the object to which the incident light is irradiated does not change in the preceding stage, the main observation period, and the subsequent stage, the electric field strength (measured value) is theoretically constant in each of these stages. However, due to minute rattles in the rotation mechanism of the rotating polarizer 16, the polarization component E C Large fluctuations in electric field strength may occur in the preceding, main observation period, and subsequent parts of the time waveform of the electric field strength. C Large fluctuations in electric field intensity that occur in the time waveform of the electric field intensity contribute to variations in the brightness of p-polarized and s-polarized light, the spectra of the measured amplitude ratio tangent tanΨ, the spectra of the measured phase difference Δ, and the carrier concentration obtained afterward.

[0065] Therefore, in a single carrier concentration measurement, after rotating the rotating polarizer 16 once (360° rotation), the polarization component E detected at any rotation angle is... C If it is found that there are large fluctuations in the electric field intensity in the time waveform of the electric field intensity, the rotation angle of the rotating polarizer 16 is initialized, and the rotating polarizer 16 is rotated again to determine the polarization component E at each rotation angle. C It was necessary to re-detect the time waveform of the electric field intensity. As a result, there was a problem of reduced throughput.

[0066] In response to this, in the film characteristic measurement method according to the second embodiment, the polarization component E C Based on the preceding portion of the time waveform of the electric field strength, it is determined whether or not large fluctuations in the electric field strength occur during the main observation period.

[0067] By the way, in the previous stage, the incident light does not enter the substrate S, so the polarization component E C The time waveform of the electric field intensity reflects the fundamental state of the ellipsometer 10 excluding the substrate S. Therefore, the polarization component E C If a large fluctuation in electric field strength occurs in the preceding stage of the time waveform of the electric field strength, it is considered that there is some kind of abnormality in the ellipsometer 10, for example, an abnormality in the rotation mechanism of the rotating polarizer 16. Furthermore, in such a case, it is considered that the abnormality in the rotation mechanism of the rotating polarizer 16 will continue in the subsequent stages, and therefore the detected polarization component E C It is also thought that large fluctuations in electric field intensity are likely to occur in the time waveform of the electric field intensity. Therefore, the applicant has proposed that the polarization component E C We investigated whether it is possible to predict the occurrence of fluctuations in electric field strength during the main observation period using the preceding portion of the time waveform of the electric field strength.

[0068] The applicant first analyzes the polarization component E of the preceding stage, the main observation period, and the subsequent stage. C The change in the fluctuation of electric field intensity in the time waveform of the electric field intensity was confirmed. Figures 7A and 7B show the raw (polarization component E) of the preceding stage, the main observation period, and the subsequent stage. C This graph shows the fluctuations in electric field intensity in the time waveform of the electric field intensity. In Figures 7A and 7B, the data number on the horizontal axis corresponds to the polarization component E for each rotation angle of the rotating polarizer 16. CThis corresponds to the detection of the time waveform of the electric field intensity. The fluctuation of the electric field intensity on the vertical axis is the standard deviation of the electric field intensity during the preceding stage, the main observation period, and the subsequent stage. Since the 25 data numbers correspond to the number of detections in one rotation of the rotating polarizer 16, the graphs in Figures 7A and 7B show data for four consecutive rotations of the rotating polarizer 16. Figure 7A shows the fluctuation of the electric field intensity during the main observation period (hereinafter abbreviated as "fluctuation of the main observation period"). Figure 7B shows the fluctuation of the electric field intensity during the preceding and subsequent stages (hereinafter abbreviated as "fluctuation of the preceding stage" and "fluctuation of the subsequent stage," respectively).

[0069] As shown in Figures 7A and 7B, it was confirmed that the patterns of change in fluctuations during the main observation period, the preceding fluctuations, and the subsequent fluctuations were similar.

[0070] Furthermore, the applicant plotted and compared all the fluctuations detected in the preceding stage, the main observation period, and the subsequent stage during 128 carrier concentration measurements. Figures 8A to 8C are graphs showing the fluctuations in the preceding stage, the main observation period, and the subsequent stage during 128 carrier concentration measurements. As described above, 25 polarization components E were detected in one carrier concentration measurement. C Since the time waveform of the electric field intensity is detected, 3200 fluctuations of electric field intensity are plotted in the graphs of Figures 8A to 8C. Furthermore, Figure 8A shows the fluctuations in the preceding stage, Figure 8B shows the fluctuations during the main observation period, and Figure 8C shows the fluctuations in the subsequent stage.

[0071] As shown in Figures 8A to 8C, it was confirmed that fluctuations in the electric field intensity for data numbers 200 to 275 and 700 to 750 were consistently reduced in each graph, for example, in the fluctuations during the main observation period, the preceding stage, and the subsequent stage (see the white arrows in each figure). Therefore, it was confirmed from the graphs in Figures 8A to 8C that the patterns of change in the fluctuations during the main observation period, the preceding stage, and the subsequent stage are similar.

[0072] Furthermore, the applicant confirmed the relationship between the average value of the electric field intensity fluctuations for each period in 128 carrier concentration measurements and the measured carrier concentration. Figures 9A to 9C are graphs showing the relationship between the average value of the electric field intensity fluctuations for each period in 128 carrier concentration measurements and the carrier concentration. Figure 9A shows the relationship between the average value of the fluctuations in the preceding stage and the carrier concentration, Figure 9B shows the relationship between the average value of the fluctuations in the main observation period and the carrier concentration, and Figure 9C shows the relationship between the average value of the fluctuations in the subsequent stage and the carrier concentration.

[0073] As shown in Figures 9A to 9C, it was confirmed that a common trend was observed in each graph: a smaller average value of electric field intensity fluctuations correlated with smaller carrier concentration variability. Considering the similar distribution patterns of carrier concentrations in each graph, it was concluded that the fluctuation patterns in the main observation period, the preceding fluctuations, and the subsequent fluctuations were similar.

[0074] Based on the results of the verification of each graph in Figures 7 to 9 above, the applicant has determined that the polarization component E C We have found that if a large fluctuation in electric field strength occurs in the preceding part of the time waveform of the electric field strength, there is a high probability that a large fluctuation in electric field strength will also occur in the main observation period. In other words, the applicant has found that the polarization component E C We obtained the insight that the occurrence of fluctuations in electric field strength during the main observation period can be predicted based on the preceding portion of the time waveform of the electric field strength.

[0075] Furthermore, the applicant uses the stability index g of the measured amplitude ratio tangent tanΨ at a specific frequency obtained in 128 carrier concentration measurements, and the polarization component E detected in 128 carrier concentration measurements. C The relationship between the time waveform of the electric field intensity and the preceding fluctuations was also confirmed. Figures 10A and 10B show the change in the stability index g of tanΨ and the polarization component E C This graph compares the change in the fluctuation pattern of the preceding stage in the time waveform of the electric field intensity. Figure 10A shows the change in the tanΨ stability index g, and Figure 10B shows the change in the fluctuation pattern of the preceding stage. The horizontal axis of Figure 10A represents the number of times the carrier concentration was measured, and the horizontal axis of Figure 10B represents the polarization component E for each rotation angle of the rotating polarizer 16.C The data numbers corresponding to the detection of the time waveform of the electric field strength are shown, with 25 times the measurement count in Figure 10A corresponding to the data numbers. As shown in Figures 10A and 10B, it was confirmed that when the tanΨ stability index g decreases, the fluctuations in the preceding stage also decrease. Therefore, it was thought that when the fluctuation of the electric field strength is small, the tanΨ stability index g decreases, and the reliability of the measured amplitude ratio tangent tanΨ obtained at that time increases, and thus the reliability of the measured carrier concentration also improves.

[0076] Furthermore, based on the above-mentioned findings, the applicant has determined that when the rotating polarizer 16 is rotated, the polarization component E detected at a certain rotation angle... C If a large fluctuation in electric field strength occurs in the preceding part of the time waveform of the electric field strength, the polarization component E C Before detecting the portion of the time waveform of the electric field intensity during the main observation period, the polarization component E at the rotation angle is detected. C We devised a method to re-detect the time waveform of the electric field intensity.

[0077] In this case, it is necessary to set a threshold (hereinafter referred to as the "fluctuation generation threshold") to determine whether or not a large fluctuation in electric field strength occurred in the preceding stage, and the applicant set this threshold based on the carrier concentration.

[0078] Figure 11 is a diagram illustrating the method for setting the fluctuation generation threshold. Figure 11 shows the relationship between the average value of the preliminary fluctuations in 128 carrier concentration measurements and the amount of shift of each carrier concentration from the average value of the 128 measured carrier concentrations.

[0079] As shown in Figure 11, the amount of shift from the average value of the carrier concentration that is acceptable as a measurement result is set as the acceptable shift amount. In addition, a regression line is obtained between the average value of the fluctuations in the previous stage and the shift amount of each carrier concentration, and the average value of the fluctuations in the previous stage corresponding to the acceptable shift amount is obtained from this regression line. Then, the average value of the fluctuations in the previous stage corresponding to the acceptable shift amount is set as the fluctuation generation threshold. Note that the method of setting the fluctuation generation threshold shown in Figure 11 is just one example, and the fluctuation generation threshold may be set by other methods.

[0080] Figure 12 is a flowchart showing the time waveform detection process as a film characteristic measurement method according to the second embodiment. The time waveform detection process in Figure 12 is performed when the carrier concentration is measured once in the ellipsometer 10, by the control unit 21 executing a program.

[0081] First, the rotation angle of the rotating polarizer 16 is initialized and set to 0° (step S121). Then, the polarization component E is detected by the terahertz wave detector 19. C The preceding portion of the time waveform of the electric field intensity is detected (step S122).

[0082] Next, it is determined whether the detected fluctuation in the preceding stage is smaller than the fluctuation generation threshold (step S123). If the fluctuation in the preceding stage in the detected time waveform is greater than or equal to the fluctuation generation threshold, the process returns to step S122, and the polarization component E C The detection of the time waveform of the electric field intensity is restarted from the previous step.

[0083] In step S123, if the preceding fluctuation in the detected time waveform is smaller than the fluctuation generation threshold, then the polarization component E C The portion of the time waveform of the electric field intensity during the main observation period is detected (step S124), and further, the polarization component E C The later stages of the time waveform of the electric field intensity are detected. If the time required to detect the main observation period portion is short, the detection of the main observation period portion may be performed without waiting for the determination result of step S123 (step S123 may be skipped). In this case, if the fluctuation of the main observation period portion in the detected time waveform is greater than or equal to the fluctuation generation threshold, the detected time waveform may be discarded and the process may return to step S122.

[0084] Next, it is determined whether the rotation angle of the rotating polarizer 16 has reached 360° (step S125). If the rotation angle of the rotating polarizer 16 has not reached 360°, the rotation angle of the rotating polarizer 16 is changed so that it increases by 15° (step S126). Then, the process returns to step S122, and at the changed rotation angle, the polarization component E CThe preceding portion of the time waveform of the electric field intensity is detected. In step S125, if the rotation angle of the rotating polarizer 16 reaches 360°, this process is terminated.

[0085] According to the time waveform detection process in Figure 12, if the detected fluctuation in the preceding stage is above the fluctuation generation threshold, the subsequent polarization component E C Without detecting the main observation period portion or subsequent stages of the time waveform, the polarization component E C The detection of the time waveform of the electric field intensity is restarted from the previous step. If a large fluctuation occurs in the detected time waveform, the polarization component E is checked once. C After waiting for the detection of the later part of the time waveform to be completed, the polarization component E C This eliminates the need to recalculate the time waveform. Therefore, throughput can be improved.

[0086] Furthermore, as a result of performing the time waveform detection process shown in Figure 12, time waveforms containing large fluctuations in electric field strength are no longer detected. This improves the reliability of the measured amplitude ratio tangent tanΨ and measured phase difference Δ obtained by the Fourier transform of the electric field strength time waveform. Moreover, it improves the reliability of the carrier concentration measurement results.

[0087] In the time waveform detection process shown in Figure 12, a fluctuation threshold was set to determine whether a large fluctuation in electric field strength occurred in the preceding stage. However, it is also possible to use a trained model to have artificial intelligence determine whether a large fluctuation in electric field strength occurred in the preceding stage. In this case, for example, a trained model configured as a convolutional neural network is generated by machine learning using training data. The training data here includes fluctuations in the preceding stage of numerous detected time waveforms of electric field strength, and each of these preceding stage fluctuations is associated with the result of whether or not each preceding stage fluctuation is acceptable from a reliability standpoint. When a preceding stage fluctuation is input to the generated trained model, it outputs a result indicating whether or not that preceding stage fluctuation is acceptable from a reliability standpoint.

[0088] Next, a third embodiment of the technology relating to this disclosure will be described. Conventionally, the carrier concentration of an epitaxial film made of SiC on a substrate S is measured by the mercury probe C-V method or the non-contact C-V method. For example, in the mercury probe C-V method, a mercury probe electrode is brought into direct contact with the epitaxial film on the substrate S to form a Schottky junction, and the capacitance (C)-reverse voltage (V) characteristics are measured across this Schottky junction. The carrier concentration is then calculated from the measurement results of the C-V characteristics. In the non-contact C-V method, a pseudo-electrode created by corona discharge is used to measure the change in depletion capacitance of a P-N junction in a non-contact manner.

[0089] On the other hand, as mentioned above, in recent years, the carrier concentration of epitaxial films is measured by ellipsometry using terahertz waves. In ellipsometry, the carrier concentration is measured by fitting the measured amplitude ratio tangent tanΨ and measured phase difference Δ spectra with the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ spectra calculated from the optical model of the substrate S.

[0090] The applicant measured the carrier concentration of an epitaxial film on the same substrate S using both the mercury probe C-V method and ellipsometry. Figure 13 is a graph showing the carrier concentrations of the epitaxial film measured using the mercury probe C-V method and ellipsometry, respectively. In this case, as shown in Figure 13, the carrier concentration measured using ellipsometry varied with respect to the X direction of the substrate S, and there was a large difference (up to approximately 1.2 × E) between it and the carrier concentration measured using the mercury probe C-V method. 16 (cm -3 )) occurred partially.

[0091] In ellipsometry, it was considered that the reason for the difference shown in FIG. 13 was that the incident p-polarized light and s-polarized light were reflected at a plurality of interfaces of the substrate S. Therefore, the applicant first performed ellipsometry using only the substrate S on which no buffer layer or epitaxial film was formed. Further, ellipsometry was performed using the substrate S on which only the buffer layer was formed. Then, when ellipsometry was performed using the substrate S on which the epitaxial film was formed in consideration of the parameters (carrier concentration, etc.) (characteristics) of the substrate S and the buffer layer obtained at this time, it was confirmed that the large difference as described above did not occur.

[0092] However, in this method, not only ellipsometry using the substrate S on which the epitaxial film is formed but also ellipsometry using the substrate S alone or the substrate S on which only the buffer layer is formed needs to be performed. Therefore, it takes a very long time to measure the carrier concentration, and there is a problem that the throughput decreases.

[0093] In response to this, in the film property measurement method according to the third embodiment, only by performing ellipsometry using the substrate S on which the buffer layer or the epitaxial film is formed, the carrier concentration of the epitaxial film with suppressed variation is obtained.

[0094] By the way, the applicant first confirmed the relationship between the signal values (luminance) and frequencies of the measured p-polarized light and the measured s-polarized light obtained from the polarization component E C FIGS. 14A and 14B are graphs showing the relationship between the luminance and frequency of the measured p-polarized light and the measured s-polarized light. FIG. 14A shows the measured p-polarized light, and FIG. 14B shows the measured s-polarized light. As shown in FIGS. 14A and 14B, when the frequency changes, the luminance of the measured p-polarized light and the measured s-polarized light changes, but the values of the luminance of the measured p-polarized light and the measured s-polarized light are much larger in the vicinity of 1 THz than in the vicinity of 2 THz. Therefore, at first glance, it is considered preferable to measure the buffer concentration using the measured p-polarized light or the measured s-polarized light in the vicinity of 1 THz.

[0095] On the other hand, the applicant used an optical model of the substrate S and performed simulations by changing the fitting variables to confirm the relationship between the sensitivity of the calculated amplitude ratio tangent tanΨ to changes in carrier concentration (calculated amplitude ratio tangent tanΨ divided by carrier concentration) and frequency. Figure 15 is a graph showing the relationship between the sensitivity of the calculated amplitude ratio tangent tanΨ to changes in carrier concentration and frequency in the optical model. As shown in Figure 15, the sensitivity of the calculated amplitude ratio tangent tanΨ to changes in carrier concentration is much greater near 2 THz than near 1 THz.

[0096] In fitting using an optical model, the fitting variable is changed to match the spectrum of the calculated amplitude ratio tangent tanΨ to the spectrum of the measured amplitude ratio tangent tanΨ. Here, in fitting that matches spectra across all frequencies (0.5 THz to 3 THz), the fitting is performed not only in the frequency range where the calculated amplitude ratio tangent tanΨ is highly sensitive to changes in carrier concentration, but also in the frequency range where it is less sensitive to changes in carrier concentration. Therefore, when the fitting variable is changed in minute increments, the spectrum of the calculated amplitude ratio tangent tanΨ does not change dynamically, and a fine-tuning of the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the measured amplitude ratio tangent tanΨ is not possible. As a result, the reliability of the fitting variable when the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the measured amplitude ratio tangent tanΨ almost match cannot be said to be very high.

[0097] On the other hand, in fitting the spectra of the calculated amplitude ratio tangent tanΨ in the frequency range where it is highly sensitive to changes in carrier concentration, the spectrum of the calculated amplitude ratio tangent tanΨ changes dynamically even with only a small change in the fitting variable. This allows for fine-tuning of the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the measured amplitude ratio tangent tanΨ. As a result, the reliability of the fitting variable is considered to be high when the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the measured amplitude ratio tangent tanΨ are in close agreement.

[0098] Therefore, based on the example in the graph of Figure 15, it is considered preferable to perform fitting in the frequency range near 2 THz from the viewpoint of obtaining highly reliable fitting variables, and consequently, from the viewpoint of reliable carrier concentration measurement.

[0099] Furthermore, the applicant states that the polarization component E C The frequency-dependent changes in the measured p-polarized and s-polarized signal values ​​(luminance) were confirmed when detection was repeated. Figures 16A to 16D show the polarization component E C These graphs show the changes in brightness of measured p-polarized and measured s-polarized signals when detection is repeated. Figure 16A shows the changes in brightness of measured p-polarized signals at 1 THz, Figure 16B shows the changes in brightness of measured s-polarized signals at 1 THz, Figure 16C shows the changes in brightness of measured p-polarized signals at 2 THz, and Figure 16D shows the changes in brightness of measured s-polarized signals at 2 THz. For example, comparing the graphs in Figure 16A and Figure 16C, the variation in brightness of measured p-polarized signals is smaller at 2 THz than at 1 THz. Also, comparing the graphs in Figure 16B and Figure 16D, the variation in brightness of measured s-polarized signals is also smaller at 2 THz than at 1 THz. From these findings, it was confirmed that measured s-polarized and measured p-polarized signals in the frequency range near 2 THz are more reliable than measured s-polarized and measured p-polarized signals in the frequency range near 1 THz.

[0100] Incidentally, the spectra corresponding to the measured amplitude ratio tangent tanΨ obtained from the time waveforms of the electric field intensity of highly reliable measured s-polarized and measured p-polarized fields are also considered to be highly reliable, and it is expected that a highly reliable carrier concentration can be measured by fitting using the obtained spectra corresponding to the measured amplitude ratio tangent tanΨ. Therefore, in order to measure a highly reliable carrier concentration, it is considered preferable to perform fitting in the frequency range near 2 THz, from the viewpoint of the variability of the brightness of the measured s-polarized and measured p-polarized fields.

[0101] From the above, it was found that the spectrum of the calculated amplitude ratio tangent tanΨ has a frequency region that is highly sensitive to changes in parameters such as carrier concentration. Similarly, it is thought that the spectrum of the calculated phase difference Δ also has a frequency region that is highly sensitive to changes in parameters such as carrier concentration.

[0102] Here, if the spectra (measured values) of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ can be divided into frequency regions that are highly sensitive to changes in each parameter, then it is thought that each parameter can be measured accurately by fitting using each divided frequency region.

[0103] Furthermore, in ellipsometry using a substrate S on which a buffer layer or epitaxial film is formed, if the spectra of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ can be separated for each parameter into frequency regions that are highly sensitive to changes in each parameter, it is conceivable that it would be possible to eliminate the need to perform ellipsometry using only a substrate S on which no buffer layer or epitaxial film is formed, or using a substrate S on which only a buffer layer is formed.

[0104] Therefore, the applicant used an optical model to simulate and confirm whether there are frequency regions that are highly sensitive to changes in each parameter in the spectra of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ. Specifically, the applicant first constructed an optical model consisting of an epitaxial film, a buffer layer, and a substrate, as shown in Figure 17. Using this optical model, the applicant calculated the amount of variation in the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ when each parameter was changed.

[0105] The parameters used to be varied were the carrier concentration, mobility, and scattering time of the substrate; the carrier concentration, mobility, scattering time, and film thickness of the buffer layer; and the carrier concentration, mobility, scattering time, and film thickness of the epitaxial film. In the simulation, as shown in Table 1 below, a median value (reference value) was set for each parameter, and the amount of variation in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ was calculated when each parameter was changed by approximately ±10% from the median value.

[0106]

[0107] The variation in the calculated amplitude ratio tangent tanΨ is shown as the increase or decrease in the calculated amplitude ratio tangent tanΨ at the upper and lower limits of the range of change for each parameter, with the calculated amplitude ratio tangent tanΨ at the median value of each parameter set to 1. Similarly, the variation in the calculated phase difference Δ is shown as the increase or decrease in the calculated phase difference Δ at the upper and lower limits of the range of change for each parameter, with the calculated phase difference Δ at the median value of each parameter set to 1. Therefore, the variation in the calculated amplitude ratio tangent tanΨ and the variation in the calculated phase difference Δ correspond to the sensitivity of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ when each parameter is changed.

[0108] Figures 18A and 18B are graphs showing the variation in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter is changed by approximately ±10% from the median value. Figure 18A shows the variation in the calculated amplitude ratio tangent tanΨ, and Figure 18B shows the variation in the calculated phase difference Δ. Note that scattering time is proportional to mobility and shows the same trend, so it is omitted from the following Figures 18A, 18B, 20A, 20B, 21A, and 21B.

[0109] As shown in Figure 18A, it was confirmed that in the frequency range below 1 THz, the amount of variation in the calculated amplitude ratio tangent tanΨ was large when the carrier concentration and mobility of the substrate were changed, while the amount of variation in the calculated amplitude ratio tangent tanΨ was almost zero when other parameters were changed. Furthermore, as shown in Figure 18B, it was confirmed that the amount of variation in the calculated phase difference Δ was large in the entire frequency range when the thickness of the epitaxial film was changed.

[0110] Therefore, it was thought that the thickness of the epitaxial film could be accurately measured by fitting the calculated phase difference Δ spectrum across the entire frequency range. However, when the thickness of the epitaxial film is changed, the amount of variation in the calculated phase difference Δ shows a peak between 1.4 THz and 2.4 THz. Near this peak, the calculated phase difference Δ is thought to fluctuate extremely in response to changes in the thickness of the epitaxial film, i.e., it fluctuates nonlinearly, making fitting somewhat difficult. Also, the formation of the peak may be strongly influenced by changes in other parameters. Therefore, it was thought that the thickness of the epitaxial film could be accurately measured by fitting the calculated phase difference Δ spectrum using frequencies between 1.0 THz and 1.3 THz and above 2.5 THz, excluding the frequency range between 1.4 THz and 2.4 THz.

[0111] Furthermore, it was considered that the carrier concentration and mobility of the substrate could be accurately measured by fitting the spectrum of the calculated amplitude ratio tangent tanΨ in the frequency range below 1 THz, for example, in the frequency range between 0.8 THz and 1.0 THz.

[0112] Incidentally, as shown in Figure 19, the relationship between carrier concentration and mobility is publicly known from literature, etc., so for example, after measuring the carrier concentration of the substrate by fitting, the mobility of the substrate can also be determined from the measured carrier concentration of the substrate based on the relationship between carrier concentration and mobility shown in Figure 19.

[0113] In this manner, after measuring the thickness of the epitaxial film and the carrier concentration and mobility of the substrate by fitting the spectra of the divided calculated amplitude ratio tangent tanΨ and calculated phase difference Δ, the applicant focused on the amount of variation in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when the remaining parameters, excluding the thickness of the epitaxial film and the carrier concentration and mobility of the substrate, were changed.

[0114] Figures 20A and 20B are graphs showing the variation in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter other than the epitaxial film thickness and the carrier concentration and mobility of the substrate is changed by approximately ±10% from the median value.

[0115] As shown in Figure 20B, it was confirmed that the amount of variation in the calculated phase difference Δ when the thickness of the buffer layer is changed is large across the entire frequency range. Therefore, it was thought that the thickness of the buffer layer could be accurately measured by fitting the spectrum of the calculated phase difference Δ across the entire frequency range. However, the amount of variation in the calculated phase difference Δ when the thickness of the buffer layer is changed also has a peak between 1.4 THz and 2.4 THz. Therefore, it was thought that the thickness of the buffer layer could be accurately measured by fitting the spectrum of the calculated phase difference Δ in the frequency range between 1.0 THz and 1.3 THz and above 2.5 THz, excluding the frequency range between 1.4 THz and 2.4 THz.

[0116] Incidentally, using the calculated phase difference Δ spectrum in the frequency range between 1.0 THz and 1.3 THz or above 2.5 THz is common to the measurement of epitaxial film thickness described above. Therefore, it is also possible that it is difficult to distinguish between the measurement of buffer layer thickness and the measurement of epitaxial film thickness using only fitting with the calculated phase difference Δ spectrum in the frequency range between 1.0 THz and 1.3 THz or above 2.5 THz.

[0117] On the other hand, comparing Figure 18A and Figure 20A, in the frequency range of 2.4 THz and above, the pattern of change in the calculated amplitude ratio tangent tanΨ when the thickness of the epitaxial film is changed is different from the pattern of change in the calculated amplitude ratio tangent tanΨ when the thickness of the buffer layer is changed.

[0118] Therefore, in order to distinguish it from the measurement of the epitaxial film thickness, the measurement of the buffer layer thickness involves not only fitting using the spectrum of the calculated phase difference Δ in the frequency range between 1.0 THz and 1.3 THz and above 2.5 THz, but also fitting using the spectrum of the calculated amplitude ratio tangent tanΨ in the frequency range above 2.4 THz.

[0119] In this manner, after measuring the thickness of the buffer layer by fitting using the spectra of the divided calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ, the applicant focused on the amount of variation in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when the remaining parameters, excluding the thickness of the buffer layer, were changed.

[0120] Figures 21A and 21B are graphs showing the variation in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter other than the epitaxial film thickness, substrate carrier concentration and mobility, and buffer layer thickness is changed by approximately ±10% from the median value.

[0121] As shown in Figure 21A, it was confirmed that the amount of variation in the calculated amplitude ratio tangent tanΨ when the carrier concentration of the epitaxial film is changed is large in the frequency range between 1 THz and 1.8 THz. Also, as shown in Figure 21B, it was confirmed that the amount of variation in the calculated phase difference Δ when the carrier concentration of the epitaxial film is changed is also large in this frequency range between 1 THz and 1.8 THz.

[0122] This suggests that the carrier concentration of the epitaxial film can be accurately measured by fitting the spectrum of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ in the frequency range between 1 THz and 1.8 THz.

[0123] From the above, we found that in the spectra of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ obtained using an optical model consisting of an epitaxial film, a buffer layer, and a substrate, there exists a frequency region that is highly sensitive to changes in each parameter. Therefore, it was found that the spectra of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ obtained in ellipsometry using a substrate S on which a buffer layer and an epitaxial film are formed can be divided into frequency regions that are highly sensitive to changes in each parameter, for each parameter.

[0124] Figure 22 is a flowchart showing the parameter measurement process as a film characteristic measurement method according to the third embodiment. The parameter measurement process in Figure 22 is based on the finding that, in the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the calculated phase difference Δ described above, there exists a frequency range for each parameter that is highly sensitive to changes in each parameter. Furthermore, the parameter measurement process in Figure 22 is executed in the ellipsometer 10 by the control unit 21 executing a program. Note that the execution of the parameter measurement process in Figure 22 is performed in the ellipsometer 10 by the polarization component E of the reflected light from the substrate S on which the buffer layer and epitaxial film are formed. C It is assumed that the function has been detected and that the spectra of the measured amplitude ratio tangent tanΨ and measured phase difference Δ have been obtained.

[0125] First, the thickness of the epitaxial film is measured by fitting using the spectrum of the calculated phase difference Δ in the frequency range between 1.0 THz and 1.3 THz and above 2.5 THz (step S221).

[0126] Next, fitting is performed using the spectrum of the calculated amplitude ratio tangent tanΨ in the frequency domain between 0.8 THz and 1.0 THz to measure the carrier concentration of the substrate. At this time, the mobility of the substrate is calculated from the carrier concentration of the substrate based on the relationship between carrier concentration and mobility shown in Figure 19 (step S222).

[0127] Next, fitting is performed using the spectrum of the calculated phase difference Δ in the frequency range between 1.0 THz and 1.3 THz and above 2.5 THz, and fitting is also performed using the spectrum of the calculated amplitude ratio tangent tanΨ in the frequency range above 2.4 THz, and the thickness of the buffer layer is measured (step S223). In addition, when fitting using the spectrum of the calculated phase difference Δ in the frequency range between 1.0 THz and 1.3 THz and above 2.5 THz in order to measure the thickness of the buffer layer, the thickness of the epitaxial film is also measured again (step S223). The reason why the thickness of the epitaxial film is measured again in step S223, even though the thickness of the epitaxial film has been measured once in step S221, is that the thickness of the epitaxial film is affected by the thickness of the buffer layer, and also because, taking into account the other parameters measured up to step S222, the thickness of the epitaxial film measured by fitting using the spectrum of the calculated phase difference Δ may change.

[0128] Next, the carrier concentration of the epitaxial film is measured by fitting using the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the calculated phase difference Δ in the frequency domain between 1 THz and 1.8 THz (step S224), and this process is completed.

[0129] According to this embodiment, the spectra of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ in ellipsometry using a substrate S on which a buffer layer and an epitaxial film are formed are divided into frequency regions that are highly sensitive to changes in each parameter. By performing fitting using the spectra of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ in the divided frequency regions, the parameters of the substrate, buffer layer, and epitaxial film can be measured. Therefore, when measuring the parameters of the substrate, buffer layer, and epitaxial film, it is not necessary to perform ellipsometry using the substrate S alone or a substrate S on which only a buffer layer is formed, thereby improving throughput.

[0130] Furthermore, according to this embodiment, the spectra of the calculated amplitude ratio tangent tanΨ and the calculated phase difference Δ are divided into frequency regions that are highly sensitive to changes in each parameter and fitted accordingly. This allows for fine-tuning of the spectra of the calculated amplitude ratio tangent tanΨ and the measured amplitude ratio tangent tanΨ, as well as fine-tuning of the spectra of the calculated phase difference Δ and the measured phase difference Δ, during the fitting process for measuring each parameter. As a result, the reliability of the parameters as measurement results can be improved.

[0131] Note that the parameter measurement process in Figure 22 is merely an example. When the thickness of the epitaxial film or the buffer layer changes, the shape of the fluctuations in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter is changed from its median value changes, so the parameter measurement process in Figure 22 cannot be used as is. When the thickness of the epitaxial film or the buffer layer changes, it is necessary to recalculate the fluctuations in the calculated amplitude ratio tangent tanΨ and calculated phase difference Δ when each parameter is changed from its median value using an optical model, and to find a new frequency range that is highly sensitive to changes in each parameter.

[0132] However, the essence of the technology according to this embodiment is to divide the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the calculated phase difference Δ into frequency regions that are highly sensitive to changes in each parameter and then perform fitting. Therefore, even if the frequency region that is highly sensitive to changes in each parameter changes in the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the calculated phase difference Δ, as long as the spectrum of the calculated amplitude ratio tangent tanΨ and the spectrum of the calculated phase difference Δ are divided into frequency regions that are highly sensitive to changes in each parameter, the technology does not deviate from the essence of the technology according to this embodiment.

[0133] Further, in the present embodiment, for example, when executing the parameter measurement process of FIG. 22, the measurement result by ellipsometry of the substrate S on which the epitaxial film to be measured is formed can be substituted by the measurement result by ellipsometry of the back surface of the wafer made of SiC. Further, when executing the parameter measurement process of FIG. 22, a learned model generated by machine learning using the measurement results of the parameters of the substrate S measured by the existing mercury probe C-V method or non-contact C-V method as teacher data may be used in combination. Thereby, the reliability of the measurement results of each parameter can be improved. However, it should be noted that it is difficult to analyze a multilayer film by the known mercury probe C-V method.

[0134] Next, a fourth embodiment of the technology according to the present disclosure will be described. By the way, in the above-described first embodiment, since the spectra of the measured amplitude ratio tangent tan Ψ and the measured phase difference Δ vary due to minute rattling of the rotation mechanism of the rotating polarizer 16, for example, until the acquired measured amplitude ratio tangent tan Ψ falls within the reference variation of tan Ψ, the detection of the time waveform of the electric field strength of the polarization component E C is repeated.

[0135] Further, as described above, in one measurement of the parameters of each layer of the substrate S in the terahertz wave detector 19, every time the rotating polarizer 16 rotates by 15°, the time waveform of the electric field strength of the polarization component E C is detected. However, again, fluctuations may occur in the time waveform of the electric field strength of the polarization component E C due to minute rattling of the rotation mechanism of the rotating polarizer 16. Therefore, in the second embodiment, the detection of the time waveform of the electric field strength of the polarization component E C is repeated until the fluctuation in the previous stage of the time waveform of the electric field strength of the polarization component E C becomes smaller than the fluctuation generation threshold value.

[0136] As described above, in the first embodiment and the second embodiment, the detection of the time waveform of the electric field strength of the stable polarization component E C is repeated until the time waveform of the electric field strength of the polarization component E C is acquired. In the present embodiment, correspondingly, the polarization component E CWithout repeatedly detecting the time waveform of the electric field intensity, a stable polarization component E C The time waveform of the electric field intensity is obtained.

[0137] Furthermore, the terahertz wave detector 19 detects the polarization component E C In order to detect the time waveform of the electric field intensity as a signal, in this embodiment, from now on, "polarization component E C The electric field strength of the measurement signal E C This is referred to as "signal strength."

[0138] Here, the applicant has found that the spectral variations of the measured amplitude ratio tangent tanΨ and measured phase difference Δ, as well as the measured signal E, are important. C The cause of fluctuations in the time waveform of the signal intensity is noise originating from the measurement system of the ellipsometer 10, such as minute rotational rattle of the rotation mechanism of the rotating polarizer 16, which affects the measured signal E C I thought it was about being included in it.

[0139] Therefore, the applicant claims that noise is measured signal E C We checked whether or not it was included. Figure 23 shows the measurement signal E at a certain rotation angle of the rotating polarizer 16. C This figure shows the time waveform of the signal intensity. Figure 23 shows 128 measured signals E of a certain rotation angle of the rotating polarizer 16, detected when the parameter was measured 128 times in the ellipsometer 10. C The time waveform of the signal intensity is superimposed, and in part A of the figure, the measured signal E from the previous stage is shown. C The time waveform of the signal intensity is partially magnified and shown, with the measured signal E in portion B of the figure representing the main observation period. C The time waveform of the signal intensity is shown partially magnified.

[0140] As shown in section A, the measurement signal E from the previous stage C The signal intensity is not constant and changes, but as mentioned above, the preceding stage is the part detected before the incident light enters the substrate S, so ideally, the measured signal E of the preceding stage C The signal strength is 0. Therefore, the measured signal E from the previous stage is 0. C The change in signal intensity is measured by the E CThis is thought to be noise (noise originating from the measurement system of the ellipsometer 10).

[0141] Figure 24 shows the measured signal E of the preceding stage at each rotation angle of the rotating polarizer 16. C This is a graph showing the standard deviation of the signal intensity. Figure 24 shows the 128 preceding stage measurement signals E detected when the parameter was measured 128 times. C The standard deviation of the signal intensity is displayed. As shown in Figure 24, even if the number of parameter measurements changes, the measured signal E is different at each rotation angle of the rotating polarizer 16. C The standard deviation of the signal intensity remains almost unchanged. Here, the measured signal E from the previous stage C The standard deviation of the signal intensity is the measured signal E from the previous stage. C Since these are characteristic noise values, Figure 24 shows the measurement signal E from the previous stage at each rotation angle of the rotating polarizer 16, even if the number of parameter measurements changes. C This indicates that the noise remains almost unchanged. In other words, the measured signal E from the previous stage C This means that the noise is highly reproducible, but since the 128 parameter measurements are performed on the same ellipsometer 10, the measurement signal E from the previous stage C The noise is thought to originate from the measurement system of the ellipsometer 10.

[0142] Furthermore, as shown in part B of Figure 23, the measurement signal E of the main observation period. C The signal intensity also does not change smoothly, therefore the measured signal E in the main observation period portion C It is thought that this includes some noise in addition to the signal corresponding to the reflected light from the substrate S (elliptical polarization composed of p-polarized and s-polarized light). And the measurement signal E from the previous stage C The time waveform of the signal intensity and the measured signal E during the main observation period. C The time waveform of the signal intensity is detected by the terahertz wave detector 19 of the same ellipsometer 10, so the measured signal E of the preceding stage C Noise similar to that in the measured signal E during the main observation period C It is also thought to be included in this.

[0143] Therefore, the applicant uses the measurement signal E of the preceding stage. C The noise generation pattern and the measured signal E during the main observation period. C We attempted to compare the noise generation patterns. Specifically, when the dose amount of the epitaxial film on the substrate S was measured 128 times in the ellipsometer 10, the measurement signal E of the rotating polarizer 16 in the stage preceding a certain rotation angle was examined. C The relationship between the noise characteristic values ​​and the dose amount of the epitaxial film obtained (measured) by fitting (hereinafter abbreviated as "relationship in the previous stage"), and the measured signal E for the portion of the main observation period at the same rotation angle of the rotating polarizer 16. C We obtained the relationship between the noise characteristics and the dose amount of the epitaxial film acquired (measured) by fitting (hereinafter abbreviated as "relationship during the main observation period"), and attempted to compare the relationship in the previous stage with the relationship during the main observation period.

[0144] Figures 25A to 25C are graphs showing the relationship in the previous stage, and Figures 26A to 26C are graphs showing the relationship during the main observation period. Here, the measured signal E C The noise characteristics are the measured signal E for each period (preliminary stage, main observation period). C The maximum, minimum, or standard deviation of the signal intensity was used. Figure 25A shows the measured signal E C The noise characteristic value is the measured signal E C Figure 25B shows the relationship in the preceding stage when the maximum signal intensity is used, and the measured signal E C The noise characteristic value is the measured signal E C Figure 25C shows the relationship in the preceding stage when the minimum signal intensity is used, and the measured signal E C The noise characteristic value is the measured signal E C Figure 26A shows the relationship in the preceding stage when using the standard deviation of the signal intensity. Also, the measured signal E C The noise characteristic value is the measured signal E C Figure 26B shows the relationship during the main observation period when the maximum signal intensity of is used. C The noise characteristic value is the measured signal E CFigure 26C shows the relationship during the main observation period when the minimum signal intensity of is used. C The noise characteristic value is the measured signal E C This graph shows the relationship during the main observation period when using the standard deviation of signal intensity. The vertical dashed lines in each graph represent the average dose amount obtained.

[0145] From the graph in Figure 25A, in the preceding stage, the measured signal E C It can be seen that as the maximum value of the signal intensity increases, the deviation from the average value of each dose also increases. On the other hand, from the graph in Figure 26A, even in the main observation period, the measured signal E C It can be seen that as the maximum value of the signal intensity increases, the deviation from the average value of each dose also increases. That is, the measured signal E C The relationship between the maximum signal intensity and the deviation from the average value for each dose was found to be similar in the preceding stage and the main observation period.

[0146] Furthermore, from the graph in Figure 25B, in the preceding stage, the measured signal E C It can be seen that as the minimum value of the signal intensity decreases (the absolute value of the minimum value increases), the deviation from the average value of each dose increases. On the other hand, from the graph in Figure 26B, it can be seen that even in the main observation period, the measured signal E C It can be seen that as the minimum value of the signal intensity decreases (the absolute value of the minimum value increases), the deviation from the average value of each dose increases. That is, the measured signal E C The relationship between the minimum signal intensity and the deviation from the average value for each dose was found to be similar in the preliminary stage and the main observation period.

[0147] Furthermore, from the graph in Figure 25C, in the preceding stage, the measured signal E C It can be seen that as the standard deviation of the signal intensity increases, the deviation from the mean value for each dose also increases. On the other hand, from the graph in Figure 26C, even in the main observation period, the measured signal E C It can be seen that as the standard deviation of the signal intensity increases, the deviation from the average value of each dose also increases. That is, the measured signal E CThe relationship between the standard deviation of signal intensity and the deviation from the mean value for each dose was found to be similar in the preceding stage and the main observation period.

[0148] As mentioned above, the measured signal E C The maximum, minimum, and standard deviation of the signal intensity are all measured from the E C Since this is a characteristic value of the noise, the measurement signal E from the preceding stage is used for the deviation from the average value of each dose. C The noise and the measured signal E from the main observation period. C It was confirmed that the noise behaves similarly. That is, the measured signal E in the main observation period. C This is the measurement signal E from the previous stage. C It was confirmed that it contains noise similar to that of [another source].

[0149] Furthermore, the applicant obtained the measurement signal E from the preceding stage when the dose amount of the epitaxial film on the substrate S was measured 128 times in the ellipsometer 10. C The characteristic values ​​of the noise and the measured signal E during the main observation period. C We also checked the relationship with the noise characteristic values. Specifically, the measurement signal E of the portion preceding a certain rotation angle of the rotating polarizer 16 in the same measurement run. C The maximum value of the signal intensity (hereinafter abbreviated as "maximum value of the previous stage") and the measured signal E of the portion of the main observation period at the same rotation angle of the rotating polarizer 16. C We examined the relationship with the maximum value of the signal intensity (hereinafter abbreviated as "maximum value during the main observation period").

[0150] Figure 27 is a graph showing the relationship between the maximum value of a certain rotation angle of the rotating polarizer 16 in the preceding stage and the maximum value during the main observation period in the same measurement run. From the graph in Figure 27, it was confirmed that there is a correlation between the maximum value of the preceding stage and the maximum value during the main observation period (see dashed line in the figure).

[0151] Furthermore, although not shown in the figure, the applicant also obtained the measurement signal E of a portion of the rotation angle preceding a certain rotation angle of the rotating polarizer 16 in the same measurement run. C The minimum value of the signal intensity and the measured signal E for the portion of the main observation period at the same rotation angle of the rotating polarizer 16. CWe also checked the relationship with the minimum value of the signal intensity. And the measured signal E from the previous stage C The minimum signal intensity and the measured signal E during the main observation period. C We also confirmed that the minimum signal intensity is correlated.

[0152] Furthermore, although not shown in the figure, the measurement signal E of the preceding stage of a certain rotation angle of the rotating polarizer 16 during the same measurement cycle. C The standard deviation of the signal intensity and the measured signal E for the portion of the main observation period at the same rotation angle of the rotating polarizer 16. C We also checked the relationship with the standard deviation of the signal intensity. And the measured signal E from the previous stage C The standard deviation of the signal intensity and the measured signal E for the main observation period. C We also confirmed that the standard deviation of the signal intensity is correlated. Here, as mentioned above, the measured signal E C The maximum, minimum, and standard deviation of the signal intensity are all measured from the E C Since it is a characteristic value of the noise, the measured signal E from the previous stage C Noise and the measured signal E of the main observation period C It was found that the noise was correlated. This is because the measured signal E in the main observation period. C However, the measurement signal E from the preceding stage C This was considered evidence that it contained noise similar to that of [another source].

[0153] As explained above, the measurement signal E from the previous stage C Noise similar to that in the measured signal E during the main observation period C Since it is also included in the measurement signal E of the main observation period C Measurement signal E from the preceding stage C If noise similar to the above noise is removed, the measurement signal E for the main observation period portion corresponds only to the reflected light from the substrate S. C That is, the measured signal E from the stable main observation period portion (free of noise). C It is thought that the time waveform of the signal intensity can be obtained.

[0154] However, as shown in Figures 23, 25A to 25C and 26A to 26C, the measurement signal E of the main observation periodC The signal strength is the measured signal E from the previous stage. C Because it is amplified from the signal intensity, the measured signal E for the main observation period C The noise is also from the measurement signal E in the preceding stage. C It is thought to be amplified from the noise.

[0155] Here, the measurement signal E from the previous stage. C Noise and the measured signal E of the main observation period C The noise ratio is the same as the measured signal E from the previous stage at the same rotation angle. C The distribution range (variability) of the noise feature values ​​and the measured signal E during the main observation period. C This is thought to be reflected in the ratio of the distribution range (variability) of the noise's feature values.

[0156] Therefore, in this embodiment, for example, the ratio of the distribution range of the maximum value in the previous stage and the distribution range of the maximum value in the main observation period at the same rotation angle (hereinafter referred to as the "correction coefficient") is obtained, and the measurement signal E of the previous stage is used. C From the noise, the measured signal E of the main observation period C Estimate the noise.

[0157] Figures 28A and 28B compare the distribution range of the maximum value in the previous stage and the distribution range of the maximum value during the main observation period at a certain rotation angle of the rotating polarizer 16. In Figures 28A and 28B, the distribution range of the maximum value in the previous stage is, for example, 0.006 (Figure 28A), and the distribution range of the maximum value during the main observation period is, for example, 0.25 (Figure 28B). Therefore, a correction coefficient of approximately 41.7 (= 0.25 / 0.006) is obtained at this rotation angle.

[0158] Furthermore, when obtaining the correction coefficient, not only the ratio of the distribution range of the maximum value in the previous stage to the distribution range of the maximum value in the main observation period is considered, but also the measurement signal E from the previous stage. C Distribution range of the minimum signal intensity and the measured signal E during the main observation period. C The ratio of the minimum signal intensity to the distribution range, and the measured signal E of the preceding stage. C Distribution range of the standard deviation of the signal intensity and the measured signal E for the main observation period. CThe ratio of the standard deviation of the signal intensity to the distribution range can be used.

[0159] In this embodiment, the correction coefficient obtained in this manner is used in the measurement signal E of the preceding step. C By multiplying by the noise characteristic value, the measured signal E for the main observation period is obtained. C The characteristic values ​​of the noise are estimated, and the measured signal E of the main observation period is obtained. C The measured signal E of the main observation period estimated from C This reduces the noise characteristics. This results in a stable measurement signal E for the main observation period. C The time waveform of the signal intensity is obtained.

[0160] By the way, as shown in the graph in Figure 24, when the rotation angle of the rotating polarizer 16 changes, the measurement signal E of the previous stage changes. C The standard deviation of the signal intensity of the measured signal E fluctuates. As mentioned above, C The standard deviation of the signal intensity is the measured signal E C Since this is a characteristic value of the noise, as can be seen from the graph in Figure 24, when the rotation angle of the rotating polarizer 16 changes, the measured signal E of the previous stage changes. C The noise is thought to change. Therefore, the measurement signal E of the preceding stage C The noise and correction coefficient need to be acquired for each rotation angle of the rotating polarizer 16. Therefore, in this embodiment, the measurement signal E of the stable main observation period is obtained. C Before acquiring the time waveform of the signal intensity, numerous parameter measurements are performed in the ellipsometer 10, and the measurement signal E of the preceding stage acquired at that time is obtained. C The distribution range of the noise feature values ​​and the measured signal E during the main observation period. C Correction coefficients are obtained in advance for each rotation angle of the rotating polarizer 16 from the distribution range of the noise characteristic values.

[0161] Figure 29 is a flowchart showing the process for acquiring a stable time waveform as a film characteristic measurement method according to the fourth embodiment. The time waveform detection process in Figure 29 is performed when the carrier concentration is measured once in the ellipsometer 10, by the control unit 21 executing a program. Note that the measurement signal E in this process CThe characteristic value of the noise is the measured signal E C This could be the maximum value, minimum value, or standard deviation of the signal intensity.

[0162] First, the rotation angle of the rotating polarizer 16 is initialized and set to 0° (step S291). Then, the terahertz wave detector 19 measures the signal E C The time waveform of the signal intensity is detected (step S292).

[0163] Next, the measurement signal E of the preceding stage is obtained from the detected time waveform. C The noise is acquired (step S293), and furthermore, the measurement signal E of the acquired portion from the previous stage is obtained. C The noise characteristic value is multiplied by a correction coefficient corresponding to the rotation angle of the current rotating polarizer 16 to obtain the measured signal E for the main observation period. C The characteristic values ​​of the noise are estimated (step S294).

[0164] Next, the measurement signal E for the main observation period. C The measured signal E of the main observation period estimated from C By reducing the noise features, the measured signal E of the main observation period is obtained. C Correction is performed (step S295), and the measurement signal E of the stable main observation period portion is obtained. C The time waveform of the signal intensity is obtained.

[0165] Next, it is determined whether the rotation angle of the rotating polarizer 16 has reached 360° (step S296). If the rotation angle of the rotating polarizer 16 has not reached 360°, the rotation angle of the rotating polarizer 16 is changed so that it increases by 15° (step S297). Then, the process returns to step S292, and steps S292 to S295 are executed at the changed rotation angle to obtain the measurement signal E for the stable main observation period. C The time waveform of the signal intensity is obtained. In step S296, if the rotation angle of the rotating polarizer 16 reaches 360°, this process is terminated.

[0166] According to this embodiment, the measurement signal E of the preceding stage is measured for each rotation angle of the rotating polarizer 16. C From the noise characteristic values, the measured signal E of the main observation period CThe characteristic values ​​of the noise are estimated, and the measured signal E of the main observation period is obtained. C The measured signal E of the main observation period estimated from C By reducing the noise feature values, the measurement signal E of the stable main observation period is obtained. C The time waveform of the signal intensity is obtained. That is, the measured signal E during the stable main observation period. C When acquiring the time waveform of the signal intensity, the polarization component E C Since it is not necessary to repeatedly detect the time waveform of the electric field strength, the throughput of acquiring the time waveform can be improved.

[0167] Furthermore, in this embodiment, the measurement signal E of the acquired main observation period portion C Since noise is no longer included, the measurement signal E of the main observation period is C This can further improve its reliability.

[0168] Furthermore, the measurement signal E C Since the noise originating from the measurement system of the ellipsometer 10, which is included as noise, is thought to include noise caused by the measurement environment, such as temperature, it is thought that by performing the stable time waveform acquisition process shown in Figure 29, noise caused by the environment in which the ellipsometer 10 is placed can also be removed. Therefore, for example, even if parameter measurements are performed when the temperature of the environment in which the ellipsometer 10 is placed or the temperature of the ellipsometer 10 itself is unstable, the measurement signal E of the stable main observation period can be removed. C It is thought that the time waveform of the signal intensity can be obtained.

[0169] While preferred embodiments of this disclosure have been described above, this disclosure is not limited to the embodiments described above, and various modifications and changes are possible within the scope of its essence.

[0170] This application claims priority based on Japanese Patent Application No. 2024-180327, filed on 15 October 2024, and Japanese Patent Application No. 2025-106608, filed on 24 June 2025, and all of the contents of those applications are incorporated herein by reference.

[0171] S Substrate 10 Ellipsometer 16 Rotating polarizer 19 Terahertz wave detector 20 Optical delay mechanism 21 Control unit

Claims

1. A method for measuring the properties of a film formed on a substrate, comprising: Incidenting a terahertz wave onto the substrate a polarization component consisting of p-polarized light vibrating parallel to an incident plane perpendicular to the surface of the substrate and s-polarized light vibrating perpendicular to the incident plane; detecting the time waveform of the electric field intensity of the polarization component reflected from the substrate; obtaining the time waveforms of the electric field intensity of p-polarized and s-polarized light from the time waveform of the electric field intensity of the detected polarization component; performing a Fourier transform on the obtained time waveforms of the electric field intensity of p-polarized and s-polarized light to obtain a spectrum of measured amplitude ratio, which is the spectrum of the amplitude ratio of the electric field intensity of the p-polarized and s-polarized light, and a spectrum of measured phase difference, which is the spectrum of the phase difference between the p-polarized and s-polarized light; constructing an optical model of the substrate; and obtaining a calculated amplitude ratio spectrum, which is the spectrum of the amplitude ratio of the electric field intensity of p-polarized and s-polarized light, and a calculated phase difference spectrum, which is the spectrum of the phase difference between the p-polarized and s-polarized light, by simulation using the optical model. A method for measuring the properties of a film formed on a substrate by performing a fitting that includes fitting the spectrum of the calculated amplitude ratio to the spectrum of the measured amplitude ratio, and fitting the spectrum of the calculated phase difference to the spectrum of the measured phase difference, wherein a highly reliable measurement value is used from among a plurality of measurement values ​​used when measuring the properties of the film.

2. A method for measuring film characteristics according to claim 1, wherein a plurality of temporally continuous spectra of the measured amplitude ratios are obtained, the measured value is the measured amplitude ratio of a specific frequency in each of the plurality of spectra of the measured amplitude ratios, and the measured amplitude ratio of a specific frequency that is not significantly affected by disturbances is selected from the plurality of measured amplitude ratios of the specific frequency.

3. The film characteristic measurement method according to claim 2, wherein the measured amplitude ratio of the selected specific frequency is the median measured amplitude ratio of the specific frequency when the amount of change in the measured amplitude ratio of three temporally consecutive specific frequencies is less than or equal to a threshold.

4. The film characteristic measurement method according to claim 3, wherein the amount of change in the measured amplitude ratio of three time-consecutive specific frequencies is the sum of the absolute values ​​of the difference between the measured amplitude ratio of the median specific frequency and the measured amplitude ratios of the preceding and succeeding specific frequencies.

5. The film characteristic measurement method according to claim 4, wherein the threshold is 0.002 or 0.

001.

6. The film characteristic measurement method according to claim 2, comprising: obtaining the variation of measured amplitude ratios of a plurality of selected specific frequencies; obtaining a new spectrum of the measured amplitude ratios in subsequent measurements; and, if the measured amplitude ratio of the specific frequency in the newly obtained spectrum of the measured amplitude ratios falls within the obtained variation, performing the fitting using the newly obtained spectrum of the measured amplitude ratios.

7. The film characteristic measurement method according to claim 6, wherein if the measured amplitude ratio of the specific frequency of the newly acquired measured amplitude ratio spectrum falls within the acquired variation, no further measured amplitude ratio spectra are acquired thereafter.

8. The film characteristic measurement method according to claim 1, wherein the measured value is the electric field intensity of the polarized component reflected from the substrate, the time waveform of the electric field intensity of the polarized component has a preceding stage portion and a main observation period portion following the preceding stage portion, and it is determined whether or not a large fluctuation in electric field intensity occurs in the main observation period portion based on the preceding stage portion of the time waveform of the electric field intensity of the polarized component.

9. The film characteristic measurement method according to claim 8, wherein if it is determined that a large fluctuation in electric field intensity occurs in the portion of the main observation period, the detection of the time waveform of the electric field intensity of the polarization component is repeated.

10. The film characteristic measurement method according to claim 8, wherein a trained model is used to determine whether or not a large fluctuation in electric field strength occurred in the preceding stage.

11. The film characteristic measurement method according to claim 1, wherein the measured values ​​are the spectrum of the calculated amplitude ratio and the spectrum of the calculated phase difference, the spectrum of the calculated amplitude ratio and the spectrum of the calculated phase difference are divided into frequency regions that are highly sensitive to changes in the characteristics of the substrate and the film, and the fitting is performed using the spectrum of the calculated amplitude ratio and the spectrum of the calculated phase difference in the divided frequency regions.

12. The film characteristic measurement method according to claim 1, wherein the film formed on the substrate is made of silicon carbide, and the film characteristic to be measured is the carrier concentration.

13. A substrate film characteristic measuring device for measuring the characteristics of a film formed on a substrate, comprising a plurality of components and a control unit, wherein the control unit controls the operation of the plurality of components of the substrate film characteristic measuring device, thereby incident a polarization component consisting of p-polarized light vibrating parallel to an incident plane perpendicular to the surface of the substrate and s-polarized light vibrating perpendicular to the incident plane onto the substrate from the incident terahertz wave, detecting the time waveform of the electric field intensity of the polarization component reflected from the substrate, obtaining the time waveforms of the electric field intensity of p-polarized and s-polarized light from the time waveform of the electric field intensity of the detected polarization component, performing a Fourier transform on the time waveforms of the electric field intensity of p-polarized and s-polarized light to obtain a spectrum of the measured amplitude ratio, which is the spectrum of the amplitude ratio of the electric field intensity of p-polarized and s-polarized light, and a spectrum of the measured phase difference, which is the spectrum of the phase difference between p-polarized and s-polarized light, and constructing an optical model of the substrate. A film characteristic measuring device that obtains a calculated amplitude ratio spectrum, which is the spectrum of the amplitude ratio of the electric field intensities of the p-polarized and s-polarized light, and a calculated phase difference spectrum, which is the spectrum of the phase difference between the p-polarized and s-polarized light, by simulation using the optical model, measures the characteristics of the film formed on the substrate by fitting, which involves matching the calculated amplitude ratio spectrum to the measured amplitude ratio spectrum and matching the calculated phase difference spectrum to the measured phase difference spectrum, and the control unit uses a highly reliable measurement value from among a plurality of measurement values ​​used when measuring the characteristics of the film.

Images