Information processing method, computer program, and information processing device

Abstract

Provided are an information processing method, a computer program, and an information processing device capable of deriving optical characteristics of a thin film. A computer acquires a provisional film thickness of a thin film, a measured value of reflected or transmitted light from a sample including the thin film as measured by an optical film-thickness measuring device, and the wavelength and the incident angle of incident light at the time of measurement. The computer estimates optical characteristics of the thin film for which the provisional film thickness has been acquired, using a model for estimating the optical characteristics of the thin film on the basis of the film thickness of the thin film, the measured value of the reflected or transmitted light from the sample including the thin film as measured by the optical film-thickness measuring device, and the wavelength and the incident angle of the incident light at the time of measurement. The computer derives the optical characteristics of the thin film by performing, on the measured value, fitting based on the estimated optical characteristics of the thin film.

Description

Information processing method, computer program, and information processing device

[0001] The present disclosure relates to an information processing method, a computer program, and an information processing device for deriving optical properties of a thin film.

[0002] A typical analytical technique using an ellipsometer involves irradiating a thin film-formed sample with incident light, detecting the reflected or transmitted light from the sample, measuring the change in the polarization state between the incident light and the reflected or transmitted light, creating an optical model of the sample based on the measurement data, and fitting simulation data from the created optical model to the measurement data to derive the film thickness or optical properties (e.g., refractive index and extinction coefficient). Analysis using an ellipsometer requires a process called modeling, which creates an optical model from the measurement data. Accurate modeling requires appropriate assumptions about the film thickness and dielectric properties. The dielectric properties of a thin film can be expressed using a dielectric function, and the parameters included in the dielectric function must be appropriately set. Traditionally, these values ​​have been estimated by experts through trial and error, and even experts have difficulty accurately estimating these values.

[0003] Patent No. 7072924

[0004] Patent Literature 1 discloses a technology for obtaining material optical constants corresponding to an ellipsometry test parameter vector input by inputting the ellipsometry test parameter vector of a material to be processed into a machine learning model constructed based on a mapping relationship between the ellipsometry test parameter vector (Δ, Ψ) of the material and the optical constants of the material. However, in order to obtain the optical constants (optical properties) of the material, the film thickness of the material must also be input. Therefore, if the film thickness of the material cannot be obtained, the optical properties of the material cannot be properly derived.

[0005] An object of the present disclosure is to provide an information processing method, a computer program, and an information processing device that are capable of deriving the optical properties of a thin film.

[0006] An information processing method according to one embodiment of the present disclosure includes a computer executing a process to acquire a tentative film thickness of a thin film, acquire a measurement value of reflected light or transmitted light from a sample including the thin film measured by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, estimate the optical properties of the thin film from which the tentative film thickness has been acquired using a model that estimates the optical properties of the thin film based on the film thickness, the measurement value of reflected light or transmitted light from a sample including the thin film measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and derive the optical properties of the thin film by fitting the acquired measurement value based on the estimated optical properties of the thin film.

[0007] In one embodiment of the present disclosure, the optical properties of a thin film to be measured are estimated using a model that estimates the optical properties of the thin film based on the film thickness of the thin film, measurement values ​​of reflected light or transmitted light from a sample including the thin film measured by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement.The optical properties of the thin film are then derived by fitting the measurement values ​​based on the estimated optical properties of the thin film.Since the optical properties of the thin film are estimated based on the above-mentioned information, it is possible to derive the optical properties with high accuracy, and it is possible to create an optical model with appropriate optical properties.

[0008] In an information processing method according to an embodiment of the present disclosure, the computer executes a process of deriving a film thickness of the thin film based on optical properties of the thin film.

[0009] According to one aspect of the present disclosure, the film thickness of a thin film to be measured can be derived with high accuracy based on optical properties derived with high accuracy.

[0010] In an information processing method according to one embodiment of the present disclosure, the optical film thickness measurement device is a spectroscopic ellipsometer, and the computer executes a process of acquiring, for each of a plurality of wavelengths, an amplitude ratio and a phase difference between p-polarized light and s-polarized light in reflected light or transmitted light from the sample measured by the spectroscopic ellipsometer, and an angle of incidence at the time of measurement, and using the model, estimating optical properties of the thin film for each of the plurality of wavelengths based on a tentative film thickness of the thin film, the amplitude ratio and phase difference between p-polarized light and s-polarized light in reflected light or transmitted light from the sample measured by the spectroscopic ellipsometer, and the wavelength and angle of incidence of the incident light at the time of measurement, and fitting the amplitude ratio and phase difference acquired for the plurality of wavelengths based on the optical properties of the thin film estimated for the plurality of wavelengths, thereby deriving the optical properties of the thin film.

[0011] In one embodiment of the present disclosure, a spectroscopic ellipsometer is used as the optical film thickness measurement device, and the spectroscopic ellipsometer acquires the amplitude ratio and phase difference between p-polarized light and s-polarized light in reflected or transmitted light from a sample for each wavelength. Then, for each wavelength, the optical properties of the thin film are estimated using a model, and the acquired amplitude ratio and phase difference for each wavelength are fitted based on the estimated optical properties to derive the optical properties of the thin film. Therefore, high-precision measurements are possible when measuring the optical properties of a thin film using a spectroscopic ellipsometer.

[0012] In an information processing method according to one embodiment of the present disclosure, the model is a learning model that has been trained using training data including the film thickness of the thin film, the measurement values ​​of the reflected light or transmitted light from the sample measured by the optical film thickness measurement device, the wavelength and incident angle of the incident light at the time of measurement, and the optical properties of the thin film, and the computer executes a process of inputting the acquired tentative film thickness of the thin film, the measurement values ​​of the reflected light or transmitted light from the sample measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement into the learning model, and acquiring the optical properties of the thin film estimated by the learning model.

[0013] In one embodiment of the present disclosure, by using a learning model, it is possible to estimate the optical properties of a thin film with high accuracy based on the film thickness of the thin film, the measurement value obtained by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement.

[0014] In an information processing method according to an embodiment of the present disclosure, the computer executes a process of estimating optical properties of the thin film without using a dielectric function of the thin film.

[0015] In one embodiment of the present disclosure, the optical properties of a thin film are estimated using a model without using the dielectric function of the thin film. Since it is difficult to appropriately set the parameters of the dielectric function, not using such a dielectric function simplifies the work involved in estimating the optical properties of the thin film.

[0016] In an information processing method according to one embodiment of the present disclosure, the computer executes a process of deriving the optical characteristics of the thin film by performing fitting based on the estimated optical characteristics of the thin film while changing the tentative film thickness.

[0017] In one embodiment of the present disclosure, by fitting while changing the temporary film thickness of the thin film, it is possible to efficiently derive optical characteristics such that the simulation data calculated by simulation from the estimated optical characteristics is close to the measured value.

[0018] In an information processing method according to an embodiment of the present disclosure, the measured values ​​include an amplitude ratio and a phase difference between p-polarized light and s-polarized light in reflected light or transmitted light from the sample, the optical properties of the thin film estimated using the model are modified according to a predetermined rule to generate a plurality of optical property candidates, and based on each of the plurality of optical property candidates, a tentative film thickness of the thin film, and a wavelength and an incident angle of the incident light at the time of measurement, calculate an amplitude ratio and a phase difference between p-polarized light and s-polarized light in reflected light or transmitted light from the sample including the thin film having each of the optical property candidates, and The computer then identifies optical property candidates that have small differences between the amplitude ratio and phase difference between p-polarized light and s-polarized light in reflected light or transmitted light from a sample including the thin film having each of the optical property candidates and the amplitude ratio and phase difference between p-polarized light and s-polarized light in reflected light or transmitted light from the sample measured by the optical film thickness measurement device, and performs processing to derive the optical properties of the thin film by fitting based on the identified optical property candidates, the tentative film thickness of the thin film, the measured amplitude ratio and phase difference, and the wavelength and incident angle of the incident light at the time of measurement.

[0019] In one embodiment of the present disclosure, optical properties of a thin film estimated using a model are modified according to a predetermined rule to generate multiple optical property candidates, and optical property candidates that can calculate simulation data with small differences from measured values ​​are identified from among the multiple optical property candidates, thereby obtaining optical properties with reduced noise generated during measurement. By performing fitting using the noise-reduced optical properties for each wavelength, it becomes possible to derive the optical properties of the thin film with high accuracy.

[0020] In an information processing method according to one embodiment of the present disclosure, the model includes at least one of a first model that estimates the refractive index of the thin film based on the film thickness of the thin film, a measurement value of reflected light or transmitted light from the sample measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and a second model that estimates the extinction coefficient of the thin film based on the film thickness of the thin film, a measurement value of reflected light or transmitted light from the sample measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement.

[0021] In one embodiment of the present disclosure, a first model is used to estimate the refractive index of a thin film based on the film thickness of the thin film, the measurement value obtained by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and / or a second model is used to estimate the extinction coefficient of the thin film based on the film thickness of the thin film, the measurement value obtained by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement. By using a model for estimating the refractive index and a model for estimating the extinction coefficient separately, it is possible to estimate the refractive index and the extinction coefficient with higher accuracy.

[0022] In an information processing method according to one embodiment of the present disclosure, the model includes at least one of a first model that estimates the refractive index of the thin film based on the film thickness of the thin film, a measurement value of reflected light or transmitted light from the sample measured by the optical film thickness measurement device, the wavelength and incident angle of the incident light at the time of measurement, and the extinction coefficient of the thin film, and a second model that estimates the extinction coefficient of the thin film based on the film thickness of the thin film, a measurement value of reflected light or transmitted light from the sample measured by the optical film thickness measurement device, the wavelength and incident angle of the incident light at the time of measurement, and the refractive index of the thin film.

[0023] In one embodiment of the present disclosure, a first model is used that estimates the refractive index of a thin film based on the film thickness of the thin film, measurements made by an optical film thickness measurement device, the wavelength and incident angle of incident light at the time of measurement, and the extinction coefficient of the thin film, and / or a second model is used that estimates the extinction coefficient of a thin film based on the film thickness of the thin film, measurements made by an optical film thickness measurement device, the wavelength and incident angle of incident light at the time of measurement, and the refractive index of the thin film. Including the extinction coefficient of the thin film in the input data of the first model for estimating the refractive index enables the refractive index to be estimated with higher accuracy, and including the refractive index of the thin film in the input data of the second model for estimating the extinction coefficient enables the extinction coefficient to be estimated with higher accuracy.

[0024] In an information processing method according to one embodiment of the present disclosure, the computer performs a process of deriving the optical properties of the thin film by weighting the residual between the measurement values ​​obtained for each wavelength and simulation data calculated using the optical properties of the thin film estimated for each wavelength and fitting the residual.

[0025] In one embodiment of the present disclosure, residuals between measured values ​​for each wavelength and simulation data calculated using the optical properties of the thin film estimated for each wavelength are weighted for each wavelength and then fitted. The simulation data calculated using the estimated optical properties has different errors (residuals) from the measured values ​​depending on the wavelength. However, the influence of these errors can be reduced, making it possible to derive the optical properties of the thin film with high accuracy. For example, the error in the optical properties increases in the low energy band where the phase film thickness tends to be small, and also in the energy band where the S / N ratio of the lamp for incident light is low. By weighting for each wavelength and fitting, it is possible to reduce the influence of such errors.

[0026] In an information processing method according to one embodiment of the present disclosure, the computer performs a process of deriving the optical properties of the thin film by fitting based on the residual and a weighted sum of the absolute value or square value of the difference in the optical properties of the thin film estimated for adjacent wavelengths.

[0027] In one embodiment of the present disclosure, fitting is performed based on the residual and a weighted sum of the absolute values ​​or square values ​​of the differences in the optical properties of the thin film estimated for adjacent wavelengths, thereby improving the accuracy of fitting and enabling the optical properties of the thin film to be derived with high accuracy.

[0028] In an information processing method according to one embodiment of the present disclosure, the computer executes a process of estimating the optical properties of the thin film based on a value obtained by multiplying the tentative film thickness of the thin film by the reciprocal of the wavelength of the incident light at the time of measurement, the measurement value of the sample measured by the optical film thickness measurement device, and the angle of incidence at the time of measurement.

[0029] In one embodiment of the present disclosure, the optical properties of a thin film are estimated using a value obtained by multiplying the tentative film thickness of the thin film by the inverse of the wavelength of the incident light at the time of measurement as one parameter, thereby making it possible to reduce the number of parameters to be processed.

[0030] In an information processing method according to one embodiment of the present disclosure, the model estimates the optical properties of the thin film based on the film thickness of the thin film, the measurement value of the reflected light or transmitted light from the sample measured by the optical film thickness measurement device, the wavelength and incident angle of the incident light at the time of the measurement, and the optical properties of a substrate on which the thin film is formed in the sample.

[0031] According to one embodiment of the present disclosure, the optical properties of the thin film are estimated taking into consideration the optical properties of the substrate on which the thin film is formed, thereby enabling the optical properties of the thin film to be estimated with higher accuracy.

[0032] In an information processing method according to an embodiment of the present disclosure, the computer executes a process of acquiring a tentative film thickness of the thin film from an apparatus involved in the process of forming the thin film.

[0033] In one aspect of the present disclosure, the tentative thickness of the thin film is acquired from an apparatus involved in the process of forming the thin film, thereby eliminating the need for a measurer to input the tentative thickness, for example, and improving operability.

[0034] In the information processing method according to one embodiment of the present disclosure, the computer executes a process of displaying the derived optical properties of the thin film on a display unit.

[0035] In one aspect of the present disclosure, the derived optical properties of the thin film can be displayed on a display unit, thereby making it possible to notify the measurer.

[0036] A computer program according to one embodiment of the present disclosure causes a computer to execute a process of acquiring a tentative film thickness of a thin film, acquiring a measurement value of reflected light or transmitted light from a sample including the thin film measured by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, estimating the optical properties of the thin film from which the tentative film thickness has been acquired using a model that estimates the optical properties of the thin film based on the film thickness of the thin film, the measurement value of reflected light or transmitted light from a sample including the thin film measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and deriving the optical properties of the thin film by fitting the acquired measurement value based on the estimated optical properties of the thin film.

[0037] In one embodiment of the present disclosure, a computer program estimates the optical properties of a thin film using a model that estimates the optical properties of a thin film based on the film thickness of the thin film, the measurement value obtained by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and the optical properties of the thin film are derived based on the estimated optical properties of the thin film.

[0038] An information processing device according to one embodiment of the present disclosure is an information processing device having a control unit, in which the control unit acquires a tentative film thickness of a thin film, acquires a measurement value of reflected light or transmitted light from a sample including the thin film measured by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, estimates the optical properties of the thin film from which the tentative film thickness has been acquired using a model that estimates the optical properties of the thin film based on the film thickness of the thin film, the measurement value of reflected light or transmitted light from a sample including the thin film measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and performs processing to derive the optical properties of the thin film by fitting the acquired measurement value based on the estimated optical properties of the thin film.

[0039] In one embodiment of the present disclosure, an information processing device estimates the optical properties of a thin film using a model that estimates the optical properties of a thin film based on the film thickness of the thin film, the measurement value obtained by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and the optical properties of the thin film are derived based on the estimated optical properties of the thin film.

[0040] According to the present disclosure, the optical properties of a thin film can be derived with high accuracy.

[0041] 1 is a block diagram showing an example of the hardware configuration of an optical film thickness measurement device; FIG. 2 is a schematic cross-sectional view showing paths of incident light and reflected light; FIG. 3 is an explanatory diagram showing the polarization state of reflected light; FIG. 4 is an explanatory diagram of a calculation process for an amplitude reflection coefficient; FIG. 5 is a block diagram showing an example of the internal configuration of a data processing device; FIG. 6 is an explanatory diagram showing an example of the configuration of a learning model; FIG. 7 is a functional block diagram showing an example of the functional configuration of a calculation unit; FIG. 8 is a flowchart showing an example of a measurement process procedure; FIG. 9 is an explanatory diagram showing an example of a refractive index adjustment process; FIG. 10 is an explanatory diagram showing the influence of measurement noise on an estimated refractive index; FIG. 11 is an explanatory diagram of a method for setting refractive index candidates; FIG. 12 is a flowchart showing an example of a measurement process procedure of embodiment 2; and FIG. 13 is a graph showing a comparison result between a true value and a calculated value.

[0042] Hereinafter, an information processing method, a computer program, and an information processing device according to the present disclosure will be specifically described with reference to the drawings illustrating embodiments thereof.

[0043] (Embodiment 1) FIG. 1 is a block diagram showing an example of the hardware configuration of an optical film thickness measurement device. The optical film thickness measurement device uses a sample 20, for example, a glass substrate 21 on which a thin film 22 is formed, irradiates the sample 20 with light, and measures the film thickness and optical characteristics of the thin film 22 based on the reflected light or transmitted light from the sample 20. Note that the measurement target of the optical film thickness measurement device is not limited to a thin film formed on a glass substrate, but can also be a thin film formed on various substrates, such as a thin film formed on a semiconductor substrate (e.g., a silicon substrate), or a thin film alone (a thin film not formed on a substrate). Furthermore, the substrate 21 is not limited to a single-layer substrate, but can also be a substrate on which multiple layers are stacked. In this case, the substrate on which the multiple layers are formed can be treated as a single virtual substrate.

[0044] In this embodiment, optical constants, such as refractive index and extinction coefficient, are used as the optical characteristics. The optical film thickness measurement device of this embodiment measures the film thickness and optical constants of the thin film 22, for example, when the thin film 22 is formed on the glass substrate 21. Therefore, when multiple layers are stacked on the glass substrate 21, the optical film thickness measurement device can obtain the film thickness and optical constants of each layer by measuring the film thickness and optical constants of the formed layer (i.e., the top layer) when each layer is formed. The thin film 22 to be measured can be formed by various film formation methods, such as CVD (Chemical Vapor Deposition) and PVD (Physical Vapor Deposition) methods such as sputtering. The optical film thickness measurement device is, for example, a polarimeter or ellipsometer. Below, the optical film thickness measurement device is described as a spectroscopic ellipsometer (hereinafter referred to as ellipsometer 10) capable of measuring light of multiple wavelengths. The ellipsometer 10 may be a phase modulation type, a rotating analyzer type, or a rotating compensator type. In this embodiment, the reflective ellipsometer 10 that captures reflected light from the sample 20 as shown in FIG. 1 will be described as an example, but a transmissive ellipsometer that captures transmitted light that has passed through the sample 20 may also be used.

[0045] The ellipsometer 10 shown in FIG. 1 includes a light irradiator 2 that irradiates a sample 20 with light, a sample stage 11 on which the sample 20 is placed, and a light acquirer 3 that acquires light reflected by the sample 20. FIG. 2 is a schematic cross-sectional view showing the paths of incident and reflected light. Arrows indicate the light in the figure. Incident light K irradiated from the light irradiator 2 is reflected by the top surface P1 of the thin film 22 and enters the thin film 22, passes through the thin film 22, is reflected by the bottom surface P2 of the thin film 22 (the top surface of the glass substrate 21), enters the glass substrate 21, passes through the glass substrate 21, and is reflected by the bottom surface P3 of the glass substrate 21. Reflected light K1 reflected by the top surface P1 of the thin film 22 directly enters the light acquirer 3, reflected light K2 reflected by the bottom surface P2 of the thin film 22 passes through the thin film 22 and enters the light acquirer 3, and reflected light K3 reflected by the bottom surface P3 of the glass substrate 21 passes through the glass substrate 21 and the thin film 22 and enters the light acquirer 3.

[0046] The ellipsometer 10 has a xenon lamp 1, which is connected to a light irradiator 2 via a first optical fiber cable 1a. The xenon lamp 1 is a light source that generates white light containing multiple wavelength components and transmits the generated white light to the light irradiator 2 via the first optical fiber cable 1a. The light irradiator 2 has an internal polarizer 2a that polarizes the white light and irradiates the polarized light onto a sample 20 placed on a sample stage 11. Note that the ellipsometer 10 may be configured to hold the sample 20 by a method other than placing it on the sample stage 11.

[0047] Light emitted from the light irradiator 2 is reflected by the sample 20, and the reflected light is captured by the light acquirer 3. The light acquirer 3 acquires the reflected light from the sample 20 and measures the polarization state of the acquired light. The light acquirer 3 incorporates a PEM (Photo Elastic Modulator) 3a and an analyzer 3b, and guides the light reflected by the sample 20 to the analyzer 3b via the PEM 3a. The PEM 3a incorporated in the light acquirer 3 phase-modulates the captured light at a required frequency (e.g., 50 kHz) to obtain elliptically polarized light from linearly polarized light. The analyzer 3b selectively acquires and measures polarization from the various polarized light phase-modulated by the PEM 3a. In this embodiment, elliptically polarized light can also be obtained using a rotating analyzer (RAE, RPE). However, using a PEM allows for the detection of sin Δ in addition to cos Δ, thereby improving measurement accuracy.

[0048] The ellipsometer 10 has a spectrometer 4, which is connected to the light acquirer 3 via a second optical fiber cable 4a. The light acquirer 3 transmits reflected light from the sample 20 acquired using the PEM 3a and the analyzer 3b to the spectrometer 4 via the second optical fiber cable 4a. The spectrometer 4 incorporates a reflection mirror, a diffraction grating, a PMT (photomultiplier tube), a control unit, and other components (not shown). The light transmitted from the light acquirer 3 via the second optical fiber cable 4a is reflected by the reflection mirror and guided to the diffraction grating. The spectrometer 4 has a configuration in which a total of 32 PMTs are arranged in a fan shape around the diffraction grating. The diffraction grating reflects the light guided thereto via a switch and a reflection mirror (not shown) toward each PMT, and distributes the reflection direction according to the wavelength of the light. Each PMT measures a specific wavelength reflected by the diffraction grating, and the spectrometer 4 has a total of 32 PMTs, enabling simultaneous measurement of 32 channels. Signals related to the content measured by each PMT are sent to the data acquisition device 5. When a polarimeter is used in the optical film thickness measurement device, it is also possible to configure it in combination with a photodiode array (PDA).

[0049] The ellipsometer 10 has a data acquisition device 5 and a data processing device 30, and the data acquisition device 5 is connected to the spectrometer 4, and the data processing device 30 is connected to the data acquisition device 5. The spectrometer 4 performs measurements for each wavelength and transmits the measurement results to the data acquisition device 5 as analog signals. The data acquisition device 5 calculates the amplitude ratio Ψ and phase difference Δ of the polarization states (p-polarized light and s-polarized light) of the reflected light from the sample 20 based on the signal from the spectrometer 4, and sends the calculated results to the data processing device 30. The light intensity signal I(t) measured by the spectrometer 4 (PMT) is expressed by the following equation (1) using the phase difference δ(t) generated in the PEM 3a. I in equation (1) S , I C is the coefficient of sin δ(t) and cos δ(t). S , I C has a relationship expressed by the following formula (2) with the amplitude ratio Ψ and phase difference Δ of p-polarized light and s-polarized light in the measured reflected light.

[0050]

[0051] FIG. 3 is an explanatory diagram showing the polarization state of reflected light. rp and E rs are the electric field intensities of the p-polarized and s-polarized light reflected by the reflected light, and the amplitude ratio Ψ and phase difference Δ of the p-polarized and s-polarized light reflected by the reflected light are expressed by the following equation (3). Note that rp and rs are the amplitude reflection coefficients of the p-polarized and s-polarized light reflected by the reflected light, respectively. tan Ψ = |rp| / |rs|, Δ = Δp - Δs (3)

[0052] 4 is an explanatory diagram of the calculation process of the amplitude reflection coefficient. When a thin film 22 is formed on a glass substrate 21, the amplitude reflection coefficient r 01 is calculated by the following equation (4): 12 is calculated by the following formula (5): 012 is calculated by the following formula (6). 01 denotes the amplitude transmission coefficient when light is incident on the thin film 22, and t 12 denotes the amplitude transmission coefficient when light is incident on the glass substrate 21 from the thin film 22.

[0053]

[0054] φ in Equations (4) and (5) 0 denotes the angle of incidence onto the thin film 22, and φ 1 denotes the refraction angle at the top surface of the thin film 22, and φ 2 represents the refraction angle at the bottom surface of the thin film 22 (top surface of the glass substrate 21). 0 , N 1 , N 2 represents the complex refractive index of the medium (ambient atmosphere), the thin film 22, and the glass substrate 21, and the complex refractive index N satisfies the relationship of the following equation (7) when the refractive index of the layer to be analyzed (medium, thin film 22, glass substrate 21) is n and the extinction coefficient is k. The refractive index n and the extinction coefficient k are the real part and the imaginary part of the complex refractive index, respectively. N=n-ik (7)

[0055] The data processing device 30 is configured using a computer such as a personal computer or a server computer. The data processing device 30 analyzes the sample 20 based on the amplitude ratio Ψ and phase difference Δ of the polarization state of the reflected light obtained by the data acquisition device 5 and an optical model corresponding to the sample 20. Specifically, the data processing device 30 estimates the optical constants (optical characteristics) of the thin film 22 to be measured based on the data from the data acquisition device 5. Furthermore, based on the estimated optical characteristics, the data processing device 30 performs fitting to identify an optical model (best model) from among pre-prepared optical models that minimizes the difference between the amplitude ratio and phase difference obtained by simulation calculation according to each optical model and the amplitude ratio and phase difference that are the measurement results obtained from the data acquisition device 5. The data processing device 30 then derives the film thickness and optical constants of the thin film 22 based on the best model.

[0056] The ellipsometer 10 further includes a control unit 6, an input unit 7, and a display unit 8. Although some connections are not shown in FIG. 1 , the control unit 6 is connected to the above-mentioned components of the ellipsometer 10 and controls the operation of each component. The control unit 6 is configured, for example, using a computer having a calculation unit and memory. The input unit 7 and the display unit 8 are connected to the control unit 6 and the data processing device 30. The input unit 7 accepts operation inputs from a user and sends control signals corresponding to the operation content to the control unit 6 or the data processing device 30. The input unit 7 is, for example, a touch panel, a keyboard, or a pointing device. The display unit 8 displays various information in accordance with instructions from the control unit 6 or the data processing device 30. The display unit 8 is, for example, a liquid crystal display or an organic electroluminescent (EL) display. The input unit 7 and the display unit 8 may be integrated into a touch panel.

[0057] FIG. 5 is a block diagram showing an example of the internal configuration of the data processing device 30. The data processing device 30 is an information processing device that executes an information processing method. The data processing device 30 includes a calculation unit 31 (control unit), a memory 32, an interface unit 33, a storage unit 34, a reading unit 35, etc., and these units, as well as the input unit 7 and the display unit 8, are connected via an internal bus. The calculation unit 31 is configured using one or more processors (arithmetic processing devices), such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a multi-threaded CPU, or a multi-core CPU. The calculation unit 31 may also be configured using a quantum computer. The calculation unit 31 executes processing to be performed by the data processing device 30 by appropriately reading a program P stored in the storage unit 34 into the memory 32 and executing it. When the calculation unit 31 includes multiple processors, each process may be executed by the same processor, or each process may be executed by a different processor.

[0058] The memory 32 stores temporary data generated in association with the calculations performed by the calculation unit 31. The memory 32 is, for example, a RAM (Random Access Memory). The interface unit 33 is connected to the data acquisition device 5 and the control unit 6. The data processing device 30 receives information from the data acquisition device 5 and the control unit 6 through the interface unit 33.

[0059] The storage unit 34 is nonvolatile and may be, for example, a hard disk, a solid-state drive (SSD), or a nonvolatile semiconductor memory. The storage unit 34 stores the program P (program product, computer program) executed by the calculation unit 31 and various data required for executing the program P. The program P includes a computer program for sample analysis. The storage unit 34 also stores menu image data for displaying various information on the display unit 8, known data related to the sample 20, optical models of different structures, created optical models, and reference data for various samples. The storage unit 34 also stores a learning model M that has learned training data through, for example, machine learning. The learning model M is expected to be used as a program module constituting artificial intelligence software. The learning model M performs a predetermined calculation on input values ​​and outputs the calculation results. Data such as coefficients and thresholds of functions that define this calculation are stored in the storage unit 34 as the learning model M. Instead of storing the learning model M in the storage unit 34, the data processing device 30 may access a server that stores the learning model M and read it. The memory unit 34 may be composed of multiple memory devices, and part of the memory unit 34 may be another memory device connected to the data processing device 30, or another memory device with which the data processing device 30 can communicate.

[0060] The reading unit 35 reads information stored on a recording medium 30a such as an optical disk or a portable memory. The program P and various data stored in the storage unit 34 may be read by the calculation unit 31 from the recording medium 30a via the reading unit 35 and stored in the storage unit 34. The program P and various data may be written to the storage unit 34 during the manufacturing stage of the data processing device 30, or may be downloaded from outside the data processing device 30 and stored in the storage unit 34.

[0061] The data processing device 30 is not limited to a single computer, but may be a multi-computer including multiple computers. Furthermore, the data processing device 30 may be a virtual machine virtually constructed within a single device by software, or may be a cloud server. In the following description, the data processing device 30 is described as a single computer. Furthermore, the program P may be deployed and executed on a single computer, or on a single site, or may be distributed across multiple sites and deployed to be executed on multiple computers interconnected via a communications network. The data processing device 30 and the control unit 6 may be configured as the same computer. Furthermore, some or all of the processing performed by the ellipsometer 10 may be executed on a computer external to the ellipsometer 10. In this case, the data processing device 30 may transmit information required for the processing to the external computer and obtain the processing results from the external computer.

[0062] FIG. 6 is an explanatory diagram showing an example of the configuration of the learning model M. The learning model M shown in FIG. 6 is a model that receives as input the tentative film thickness d (predicted film thickness) of the thin film 22, the wavelength λ and incident angle φ of the incident light during measurement by the ellipsometer 10, the optical constants (refractive index n0 and extinction coefficient k0) of the glass substrate 21, and measurement data from the ellipsometer 10 (the amplitude ratio Ψ and phase difference Δ of p-polarized and s-polarized light reflected from the sample 20), performs a calculation to estimate the optical constants (refractive index n1 and extinction coefficient k1) of the thin film 22 based on the input data, and outputs the calculation results. The learning model M can be configured using algorithms such as an SVM (support vector machine), a decision tree, a random forest, a CNN (convolutional neural network), or a Transformer, or it may be configured by combining multiple algorithms. The output data, the refractive index n1 and extinction coefficient k1 of the thin film 22, may be any numerical value or may be within a range of values ​​that each value can take.

[0063] The learning model M is generated by machine learning using training data that associates each piece of training data (the provisional film thickness d of the thin film 22, the wavelength λ and incident angle φ of the light incident on the sample 20, the refractive index n0 and extinction coefficient k0 of the glass substrate 21, and the amplitude ratio Ψ and phase difference Δ of p-polarized light and s-polarized light reflected from the sample 20 by the incident light) with correct answer data (the refractive index n1 and extinction coefficient k1 of the thin film 22). Each piece of training data can be generated by optical calculation (simulation) defined in a computer program stored in the storage unit 34. For example, each piece of training data can be generated by performing a simulation to calculate the amplitude ratio Ψ and phase difference Δ of p-polarized light and s-polarized light reflected from the sample 20 by the incident light, based on the provisional film thickness d of the thin film 22, the wavelength λ and incident angle φ of the light incident on the sample 20, the refractive index n0 and extinction coefficient k0 of the glass substrate 21, and the refractive index n1 and extinction coefficient k1 of the thin film 22.

[0064] The learning model M learns to output correct data (the refractive index n1 and extinction coefficient k1 of the thin film 22) when input data included in the training data (the tentative film thickness d of the thin film 22, the wavelength λ and incident angle φ of the incident light during measurement, the refractive index n0 and extinction coefficient k0 of the glass substrate 21, and the amplitude ratio Ψ and phase difference Δ of the p-polarized light and the s-polarized light reflected from the sample 20 by the incident light) are input. Specifically, the learning model M performs calculations based on the input information to estimate the refractive index and extinction coefficient of the thin film 22 and obtain the estimated results. The learning model M then compares the estimated refractive index and extinction coefficient with the correct refractive index n1 and extinction coefficient k1, and optimizes parameters such as weights (coupling coefficients) between neurons in the learning model M so that they approximate each other. The parameter optimization method can be a steepest descent method, an error backpropagation method, or the like. This results in a learning model M that outputs the refractive index n1 and extinction coefficient k1 of the thin film 22 when the temporary film thickness d of the thin film 22, the wavelength λ and incident angle φ of the incident light during measurement by the ellipsometer 10, the refractive index n0 and extinction coefficient k0 of the glass substrate 21, and the amplitude ratio Ψ and phase difference Δ of the p-polarized light and s-polarized light reflected from the sample 20 by the incident light (measurement data by the ellipsometer 10) are input.

[0065] The refractive index n1 and extinction coefficient k1 of the thin film 22 output by the learning model M shown in Figure 6 are numerical values ​​expressed as continuous values, and the learning model M outputs one of the continuous values ​​as a so-called regression problem. However, it may also be configured to consider it as a classification problem and determine one of selectable numerical values ​​or numerical ranges (numeric values ​​or numerical ranges that can be taken as the refractive index n1 and the extinction coefficient k1) for the refractive index n1 and the extinction coefficient k1. In this case, the learning model M is configured to determine the optimal numerical value or numerical range for each of the refractive index n1 and the extinction coefficient k1 from a plurality of preset numerical values ​​or numerical ranges and output the determination result.

[0066] The generation of training data and the learning of the learning model M may be performed by the data processing device 30 or another learning device. When learning is performed by another learning device, the learning model M generated by the other learning device is downloaded from the learning device to the data processing device 30 via a network or a recording medium 30a and stored in the memory unit 34. The learning model M is not limited to the configuration shown in FIG. 6. For example, the learning model M may be configured to input the temporary film thickness d of the thin film 22 and the wavelength λ of the incident light during measurement separately, or may be configured to input the result (product) of multiplying the temporary film thickness d by the reciprocal of the wavelength λ. Furthermore, the learning model M may be configured not to include the optical properties (refractive index n0 and extinction coefficient k0) of the glass substrate 21 in the input data. That is, the learning model M may be configured to estimate the optical properties (refractive index n1 and extinction coefficient k1) of the thin film 22 based on the temporary film thickness d, the wavelength λ and incident angle φ of the incident light, and the measurement data (amplitude ratio Ψ and phase difference Δ). The learning model M may be configured to have two models, namely, a first model for estimating the refractive index n1 of the thin film 22 and a second model for estimating the extinction coefficient k1 of the thin film 22, or may be configured to have either one of the models. Furthermore, the first model for estimating the refractive index n1 of the thin film 22 may be configured to include the extinction coefficient of the thin film 22 in the input data, and the second model for estimating the extinction coefficient k1 of the thin film 22 may be configured to include the refractive index n1 of the thin film 22 in the input data.

[0067] Furthermore, multiple learning models M may be prepared for each possible angle of incidence φ. In this case, the learning model M is configured to input the tentative film thickness d, the wavelength λ of the incident light, the optical properties of the glass substrate 21 (refractive index n0 and extinction coefficient k0), and the measurement data (amplitude ratio Ψ and phase difference Δ). By using the learning model M corresponding to the incident angle φ, it is possible to estimate the optical properties of the thin film 22 (refractive index n1 and extinction coefficient k1). Note that, if the incident angle φ is limited to one type, it is sufficient to generate one learning model M using the tentative film thickness d, the wavelength λ of the incident light, the optical properties of the glass substrate 21, and the measurement data as input data. Similarly, a learning model M may be configured for each combination of possible values ​​of the wavelength λ and incident angle φ of the incident light, the optical properties of the glass substrate 21 (refractive index n0 and extinction coefficient k0), and the measurement data (amplitude ratio Ψ and phase difference Δ). In this case, the learning model M is configured to receive only the tentative film thickness d of the thin film 22 as input, and a learning model M corresponding to the wavelength λ and incident angle φ of the incident light, the optical characteristics of the glass substrate 21, and the measurement data is used, and the optical characteristics (refractive index n1 and extinction coefficient k1) of the thin film 22 can be estimated by inputting the tentative film thickness d of the thin film 22 into this learning model M. Therefore, by generating a learning model M for each combination of possible values ​​for any two or more of the tentative film thickness d of the thin film 22, the wavelength λ and incident angle φ of the incident light, the optical characteristics of the glass substrate 21, and the measurement data, a learning model M capable of estimating the optical characteristics of the thin film 22 using other data (at least one of the above data) as input data can be realized.

[0068] The data processing device 30 prepares the learning model M as described above in advance, and when the ellipsometer 10 is used to measure the amplitude ratio and phase difference between p-polarized light and s-polarized light reflected from the sample 20, the data processing device 30 uses the learning model M to estimate the refractive index n1 and extinction coefficient k1 of the thin film 22. The data processing device 30 also performs a simulation using the estimation results to obtain the amplitude ratio and phase difference (simulation data) at each wavelength, and derives the film thickness of the thin film 22 by fitting the obtained simulation data to the measurement data obtained by the ellipsometer 10 while changing the temporary film thickness of the thin film 22. Note that when the data processing device 30 can derive the film thickness of the thin film 22, it can also derive the optical properties (refractive index and extinction coefficient) of the thin film 22.

[0069] FIG. 7 is a functional block diagram showing an example of the functional configuration of the calculation unit 31 of the data processing device 30. The calculation unit 31 appropriately loads and executes a program P stored in the storage unit 34 into the memory 32, thereby implementing the functions of the data acquisition unit 31a, optical constant estimation unit 31b, amplitude ratio / phase difference calculation unit 31c, Is / Ic calculation unit 31d, and fitting unit 31e. The data acquisition unit 31a acquires various data used in the process of determining the film thickness and optical constants of the thin film to be measured. In this embodiment, the data acquisition unit 31a acquires measurement data obtained by the ellipsometer 10 (the amplitude ratio and phase difference between p-polarized light and s-polarized light reflected from the sample 20 for each wavelength of incident light), the angle of incidence during measurement, the provisional film thickness of the thin film 22 to be measured, and the optical constants of the glass substrate 21. The measurement data can be acquired via the data acquisition device 5. The angle of incidence can be acquired, for example, from the control unit 6. The provisional film thickness can be acquired, for example, via the input unit 7. The optical constants of the glass substrate 21 are preset, for example, in a computer program for sample analysis.

[0070] The optical constant estimation unit 31b estimates the optical constants (refractive index n1 and extinction coefficient k1) of the thin film 22 using the data acquired by the data acquisition unit 31a. The optical constant estimation unit 31b estimates the optical constants of the thin film 22 using, for example, a learning model M. The amplitude ratio and phase difference calculation unit 31c calculates simulation data (theoretical values) of the amplitude ratio and phase difference obtained when the sample 20 having the thin film 22 with the optical constants estimated by the optical constant estimation unit 31b is measured using a simulation in accordance with a computer program for sample analysis. The Is / Ic calculation unit 31d calculates I according to the above formula (2) based on the simulation data of the amplitude ratio and phase difference calculated by the amplitude ratio and phase difference calculation unit 31c. S and I C Calculate.

[0071] The fitting unit 31e applies the I calculated by the Is / Ic calculation unit 31d to the measurement data acquired by the data acquisition unit 31a. S and I C The fitting unit 31e performs fitting based on the I of the measurement data while changing the temporary film thickness and / or optical constants of the thin film 22 to be measured. S , I C and the calculated data I S , I C The difference between the measured data and S , I C and the calculated data I S , I C The temporary film thickness and / or optical constants that minimize the difference from the measured film thickness are derived. S , I C and the calculated data I S , I C The optical constants estimated when the tentative film thickness and / or optical constants that minimize the difference from the measured value can be derived become the optical constants of the thin film 22 to be measured.

[0072] The following describes the data analysis process that the data processing device 30 of this embodiment executes to derive the film thickness and optical constants (refractive index and extinction coefficient) of the thin film 22. Fig. 8 is a flowchart showing an example of the measurement process procedure. The ellipsometer 10 of this embodiment executes the following process, for example, when the thin film 22 is formed on the glass substrate 21, to measure the film thickness and optical constants of the thin film to be measured.

[0073] When the ellipsometer 10 performs measurement processing under the control of the control unit 6, the calculation unit 31 of the data processing device 30 acquires measurement data from the ellipsometer 10 via the data acquisition device 5 (S11). The measurement data is the amplitude ratio Ψ and phase difference Δ at each wavelength between p-polarized light and s-polarized light reflected from the sample 20, and the amplitude ratio Ψ is E (λ) and phase difference Δ E (λ).

[0074] The calculation unit 31 acquires the incident angle φ of the light irradiated by the light irradiator 2 onto the sample 20 when the measurement data is measured (S12). The incident angle can be controlled by the control unit 6, so the calculation unit 31 can acquire the incident angle from the control unit 6. The calculation unit 31 acquires the temporary film thickness of the thin film 22 (S13). For example, the calculation unit 31 acquires the temporary film thickness by inputting data via the input unit 7. If the ellipsometer 10 is communicably connected to an apparatus (e.g., a film forming apparatus) that performs a process related to the formation of the thin film 22, the calculation unit 31 may acquire the film thickness set as a target in the film forming process from the apparatus. Alternatively, the memory unit 34 may store multiple selectable values ​​for the temporary film thickness of the thin film 22, and the calculation unit 31 may display the multiple temporary film thicknesses on the display unit 8, accept the selection of one of the displayed temporary film thicknesses, and acquire the accepted temporary film thickness. The calculation unit 31 stores the acquired measurement data, incident angle, and temporary film thickness in the memory 32 or the memory unit 34.

[0075] The calculation unit 31 identifies a wavelength to be processed (S14) and estimates the refractive index n1(λ) and extinction coefficient k1(λ) of the thin film 22 for the wavelength λ to be processed (S15). Here, the calculation unit 31 identifies one of the wavelengths included in the measured reflected light, and inputs the tentative film thickness of the thin film 22 acquired in step S13, the wavelength to be processed, the angle of incidence acquired in step S12, the optical constants (refractive index and extinction coefficient) of the glass substrate 21, and the amplitude ratio and phase difference of the wavelength to be processed extracted from the measurement data acquired in step S11 into the learning model M, and obtains the refractive index and extinction coefficient of the thin film 22 as output values ​​from the learning model M. The tentative film thickness of the thin film 22 and the wavelength to be processed may be input separately to the learning model M, or the product of the tentative film thickness and the reciprocal of the wavelength (tentative film thickness × reciprocal of the wavelength) may be input to the learning model M. The optical constants (refractive index n0(λ) and extinction coefficient k0(λ)) of the glass substrate 21 are preset in, for example, a computer program for sample analysis. The calculation unit 31 stores the estimated refractive index n1(λ) and extinction coefficient k1(λ) of the thin film 22 in the storage unit 34 in association with the wavelength to be processed.

[0076] The process of estimating the refractive index and extinction coefficient of the thin film 22 from the provisional film thickness of the thin film 22, the wavelength to be processed, the angle of incidence, the optical constants of the glass substrate 21, and the measurement data (amplitude ratio and phase difference) of the wavelength to be processed is not limited to processing using the learning model M, but may also be rule-based processing. For example, each set of the provisional film thickness of the thin film 22, the wavelength to be processed, the angle of incidence, the optical constants of the glass substrate 21, and the measurement data (amplitude ratio and phase difference) of the wavelength to be processed may be associated in advance with the refractive index and extinction coefficient of the thin film 22 and stored in a database (DB) prepared in the storage unit 34. By referencing the DB, the refractive index and extinction coefficient of the thin film 22 corresponding to the provisional film thickness of the thin film 22, the wavelength to be processed, the angle of incidence, the optical constants of the glass substrate 21, and the measurement data (amplitude ratio and phase difference) of the wavelength to be processed can be estimated. Furthermore, a plurality of learning models M and DBs may be prepared for each possible angle of incidence. In this case, the calculation unit 31 identifies a learning model M or DB corresponding to the incident angle of the object to be processed, and using the identified learning model M or DB, it can estimate the refractive index and extinction coefficient of the thin film 22 from the tentative film thickness of the thin film 22, the wavelength of the object to be processed, the optical constants of the glass substrate 21, and the measurement data of the object to be processed.

[0077] The calculation unit 31 calculates the amplitude ratio Ψ and phase difference Δ between p-polarized light and s-polarized light in the light reflected from the sample 20 based on the estimated refractive index and extinction coefficient (optical constant) of the thin film 22 for the wavelength to be processed (S16). Here, the calculation unit 31 executes a simulation in accordance with a computer program for sample analysis stored in the storage unit 34, and calculates the amplitude ratio Ψ by theoretical calculation processing related to the simulation. M (λ) and phase difference Δ M The calculation unit 31 calculates the theoretical value (simulation data) of the amplitude ratio Ψ (λ). M (λ) and phase difference Δ M Based on (λ), I is calculated according to the above equation (2). S and I C is calculated (S17), and I S , I C Get.

[0078] The calculation unit 31 determines whether the processing of steps S14 to S17 has been completed for all wavelengths in the measurement data acquired in step S11 (S18). If it determines that the processing has not been completed (S18: NO), the calculation unit 31 returns to step S14 and repeats the processing of steps S14 to S17 for unprocessed wavelengths. If it determines that the processing has been completed for all wavelengths (S18: YES), the calculation unit 31 calculates the difference between the measurement data and the calculated data (the sum of squares of the residuals for the measurement data) (S19). For example, the calculation unit 31 calculates the amplitude ratio Ψ of the measurement data. E (λ) and phase difference Δ E (λ), I of each wavelength is calculated according to the above formula (2). S and I C (Measurement data I S , I C ) is calculated, and for each wavelength, the I of the measurement data is calculated. S , I C and the calculated data I S , I C The calculation unit 31 calculates the difference (sum of squares of residuals) by calculating the square of the difference between the wavelengths and the refractive index n1(λ) and the extinction coefficient k1(λ) (optical constants) estimated in step S15, and stores the difference calculated in step S19 in the memory 32 or the storage unit 34 in association with the tentative film thickness of the thin film 22 (step S20).

[0079] The calculation unit 31 performs the above-described process for each temporary film thickness while changing the temporary film thickness of the thin film 22, and calculates the I S , I C and the measurement data I S , I C The difference between the measured values ​​and the measured values ​​is calculated, and the tentative film thickness that minimizes the difference is determined as the film thickness of the thin film 22. S , I C The calculated data I that has the smallest difference from S , I C Identify the specified calculated data I S , I C The process of specifying the calculated temporary film thickness and / or optical constants is called fitting. Here, the calculation unit 31 sets a processing range from a temporary film thickness obtained in step S13 that is, for example, reduced by 10% to a temporary film thickness that is increased by 10%, sets a plurality of temporary film thicknesses within the processing range (for example, set in increments of 5%), performs the processes of steps S14 to S20 for each temporary film thickness, and calculates the calculated data I for each temporary film thickness. S , I C and the measurement data I S , I C The temporary film thickness that results in the smallest difference (sum of squares of residuals) between the calculated film thickness and the target film thickness is identified. Specifically, the calculation unit 31 determines whether the processing of steps S14 to S20 has been completed for all the target film thicknesses (S21), and if it determines that the processing has not been completed (S21: NO), changes the target film thickness (S22). Here, the calculation unit 31 sets other temporary film thicknesses within the processing range by adding or subtracting a predetermined value to the temporary film thickness used in the most recent processing. The calculation unit 31 may also set the target film thickness using an optimization method such as Bayesian optimization or reinforcement learning. Furthermore, the calculation unit 31 calculates the I of the calculated data. S , I C and the measurement data I S , I C By fitting with the above equation, a value closer to the true value (the actual film thickness) than the processed tentative film thickness may be derived and used as the tentative film thickness of the processing target.

[0080] Then, the calculation unit 31 returns to step S14 and repeats the processes of steps S14 to S20 for the changed temporary film thickness. If it is determined that the processes have been completed for all temporary film thicknesses (S21: YES), the calculation unit 31 identifies the temporary film thickness that has the smallest difference among the differences corresponding to the temporary film thicknesses stored in step S20 and the optical constants corresponding to this temporary film thickness (S23). The calculation unit 31 specifies the identified temporary film thickness and optical constants as the film thickness and optical constants of the thin film 22, and stores the identified film thickness and optical constants in the memory 32 or the storage unit 34 (S24). Through the above-mentioned process, the calculation unit 31 repeats the processes of steps S14 to S20 while changing the temporary film thickness, and calculates the I of the calculated data. S , I C and the measurement data IS , I C The calculation unit 31 may display the film thickness and / or optical constants of the thin film 22 determined by the above-described process on the display unit 8 to present them to the measurer.

[0081] According to the above-described process, in this embodiment, the optical properties (refractive index and extinction coefficient) of the thin film 22 are estimated using the learning model M without using a dielectric function. This eliminates the need to identify parameters of the dielectric function, enabling simple modeling. Therefore, even for thin films formed of unknown materials, the optical properties can be estimated easily and accurately, and fitting can be performed efficiently based on the highly accurately estimated optical properties. Furthermore, in this embodiment, the optical properties of the thin film are estimated for each wavelength, allowing estimation using the same learning model M even when the measurement wavelength range or measurement wavelength resolution of the ellipsometer 10 is different. Furthermore, the dielectric properties (e.g., dielectric constant) of the target layer can be calculated using the dielectric function, and the optical properties (e.g., optical constants) of the target layer can be calculated from the dielectric properties. Since this embodiment does not use a dielectric function, the dispersion formula used to calculate optical properties that can be calculated from the dielectric function is also not used.

[0082] The ellipsometer 10 of this embodiment is configured to specify the film thickness and optical constants (refractive index n1 and extinction coefficient k1) of the thin film 22, but may be configured to specify only one of them.

[0083] (Embodiment 2) This section describes an ellipsometer 10 that performs processing to reduce the influence of measurement noise and the like on the refractive index n1(λ) and extinction coefficient k1(λ) of the thin film 22 estimated for each wavelength in the measurement processing by the ellipsometer 10 of Embodiment 1. The ellipsometer 10 of this embodiment can be realized by a configuration similar to that of the ellipsometer 10 of Embodiment 1 shown in Figures 1 and 5, and therefore a description of the configuration of each part will be omitted.

[0084] FIG. 9 is an explanatory diagram showing an example of a refractive index adjustment process, FIG. 10 is an explanatory diagram showing the effect of measurement noise on the estimated refractive index, and FIG. 11 is an explanatory diagram showing a method for setting a refractive index candidate. The graph shown in FIG. 9 shows light energy (a value inversely proportional to wavelength) on the horizontal axis and the refractive index n1(En) of the thin film 22 on the vertical axis. The upper graph in FIG. 9 shows the estimated value of the refractive index n1(En) of the thin film 22 estimated using the learning model M. The data processing device 30 of this embodiment performs a correction process on the estimated value of the refractive index n1(En) shown in the upper graph in FIG. 9 to reduce the machine learning estimation error for each light energy, and calculates a corrected value of the refractive index n1(En) with the reduced estimation error, as shown in the center graph in FIG. 9. The data processing device 30 also performs a similar process on the estimated value of the extinction coefficient k1(En) to calculate a corrected value of the extinction coefficient k1(En) with the reduced machine learning estimation error. Then, the data processing device 30 performs fitting based on the calculated correction value of the refractive index n1(En) and the correction value of the extinction coefficient k1(En), thereby identifying the refractive index n1(En) and the extinction coefficient k1(En) that are close to the true values ​​(the original values ​​of the refractive index and extinction coefficient) of the thin film 22 being measured.

[0085] Specifically, as shown in the upper graph of FIG. 11 , the data processing device 30 determines a range for each light energy based on the estimated value of the refractive index n1(En), and sets multiple refractive index candidates (optical property candidates) within the range. The range may be, for example, a range from a refractive index smaller by a predetermined amount or a predetermined percentage to a refractive index larger by a predetermined amount or a predetermined percentage than the estimated value of the refractive index n1(En), or may be a range corresponding to the variation (standard deviation) of the refractive index n1(En), or may be set according to a predetermined rule. The range may also be a range corresponding to the absolute value of the difference between the refractive index n1(En) for a nearby wavelength. The refractive index candidates within the range may be set randomly within the range, or may be set at predetermined intervals, or may be set using an optimization method such as Bayesian optimization or reinforcement learning. In the center diagram of FIG. 11 , the estimated value of the refractive index n1(En) is indicated by a black circle, and refractive index candidates other than the estimated value are indicated by a white circle. The data processing device 30 also performs similar processing on the estimated value of the extinction coefficient k1(En) to set multiple candidates for the extinction coefficient.

[0086] The data processing device 30 calculates the amplitude ratio and phase difference of the p-polarized light and the s-polarized light reflected from the sample 20 by simulation based on the optical constants of each refractive index candidate (including an estimated value of the refractive index n1(λ)) and each extinction coefficient candidate (including an estimated value of the extinction coefficient k1(λ)) set for each light energy, the tentative film thickness of the thin film 22, the wavelength and angle of incidence of the measurement light, and the optical constants of the glass substrate 21 (refractive index n0 and extinction coefficient k0). The data processing device 30 then identifies, from the calculated amplitude ratio and phase difference, the calculated values ​​of the amplitude ratio and phase difference that minimize the difference from the values ​​measured by the ellipsometer 10, and uses the optical constants identified based on the identified calculated values ​​as the correction values ​​for the refractive index n1(En) and the extinction coefficient k1(En). By performing the above-described processing for each light energy, corrected values ​​of the refractive index n1(En) and the extinction coefficient k1(En) with reduced estimation errors by machine learning are obtained, as shown in the lower graph of FIG. 11 . The data processing device 30 performs fitting based on the corrected values ​​of the refractive index n1(En) and the extinction coefficient k1(En), thereby obtaining the refractive index n1(En) and the extinction coefficient k1(En) (fitted values) as shown in the lower graph of Figure 9.

[0087] In this embodiment, weighting is performed for each wavelength on the error function used to calculate the difference between the estimated and measured values ​​of the refractive index and extinction coefficient during fitting. The graph shown in FIG. 10 shows light energy (a value inversely proportional to wavelength) on the horizontal axis and the refractive index n1 (En) of the thin film 22 on the vertical axis, with the true value of the refractive index indicated by the solid line and the estimated value of the refractive index indicated by the dashed line. As shown by the closed curve in the upper graph of FIG. 10 , for example, when the film thickness of the thin film 22 is less than 100 angstroms, the phase film thickness tends to be sufficiently small compared to the wavelength of the measurement light, and therefore the influence of measurement noise increases in the low-energy band (1.5 eV to 2.5 eV). Furthermore, as shown by the closed curve in the lower graph of FIG. 10 , the influence of measurement noise due to the optical characteristics of the xenon lamp 1 increases in the high-energy band (over 4.0 eV). Therefore, in this embodiment, the weighting is reduced in the low-energy band and the high-energy band, thereby reducing the influence of measurement noise in the low-energy band and the high-energy band. The weighted error function is expressed by the following equation (8).

[0088]

[0089] In Equation (8), i represents the wavelength index, and Wi represents the residual weight for each wavelength index. The difference in Equation (8) represents the difference between the estimated values ​​for adjacent wavelengths, and may be a difference calculated using a first-order difference method, a second-order central difference method, or the like, or may be a difference calculated using a difference method that combines multiple difference methods. Furthermore, in Equation (8), instead of the difference, the absolute value of the difference may be used, or a weighted sum of the squared values ​​of the differences calculated for each wavelength may be used. α in Equation (8) is a variable for controlling the weights for the residual and difference with respect to the measurement value. When emphasis is placed on the residual with respect to the measurement value, a small value is set to α, and when emphasis is placed on the difference, a large value is set to α. The residual weight Wi for each wavelength index may be determined, for example, based on the standard deviation of the measurement value, or may be determined based on simulation results (calculated data).

[0090] Fig. 12 is a flowchart showing an example of a measurement processing procedure according to embodiment 2. The processing shown in Fig. 12 is obtained by adding steps S31 to S35 between steps S15 and S16 in the processing shown in Fig. 8. Explanations of the same steps as in Fig. 8 will be omitted.

[0091] In this embodiment, the calculation unit 31 of the data processing device 30 estimates the refractive index n1(λ) and extinction coefficient k1(λ) of the thin film 22 for a wavelength to be processed (S15), and then determines a processing range for the estimated refractive index n1(λ) and extinction coefficient k1(λ) for the wavelength to be processed (S31). The calculation unit 31 then sets multiple candidate refractive indexes and candidate extinction coefficients within the determined range (S32). The calculation unit 31 calculates the amplitude ratio and phase difference of the p-polarized light and the s-polarized light reflected from the sample 20 by simulation based on the optical constants determined by the candidate refractive indexes and candidate extinction coefficients set within the range, the tentative film thickness of the thin film 22, the wavelength and angle of incidence of the measurement light, and the optical constants of the glass substrate 21 (S33). The calculation unit 31 then calculates the difference between the calculated amplitude ratio and phase difference and the measured amplitude ratio and phase difference values ​​obtained by the ellipsometer 10 (S34).

[0092] The calculation unit 31 identifies the smallest difference among the differences calculated for each optical constant using each refractive index candidate and each extinction coefficient candidate, identifies the amplitude ratio and phase difference resulting from the identified difference, and specifies the refractive index candidate and extinction coefficient candidate from which the identified amplitude ratio and phase difference were calculated as the refractive index and extinction coefficient at the wavelength to be processed (S35). The identified refractive index and extinction coefficient are used as correction values ​​for the refractive index n1(λ) and extinction coefficient k1(λ). The calculation unit 31 then proceeds to step S16. In step S16, the calculation unit 31 calculates the amplitude ratio Ψ and phase difference Δ between p-polarized light and s-polarized light reflected from the sample 20 for the wavelength to be processed based on the identified correction values ​​for the refractive index n1(λ) and extinction coefficient k1(λ).

[0093] Thereafter, the calculation unit 31 executes the processes from step S17 onwards. Therefore, in this embodiment, the estimated values ​​of the refractive index and extinction coefficient of the thin film 22 estimated using the learning model M are corrected to values ​​closer to the measured data, and then fitting is performed. In step S19, the calculation unit 31 calculates the difference between the measured data and the calculated data (estimated value) (the sum of squares of the residuals for the measured data) using the error function expressed by the above formula (8). Specifically, the calculation unit 31 calculates the amplitude ratio Ψ of the measured data E (λ) and phase difference Δ E (λ) for each wavelength S and I C (Measurement data I S , I C ) is calculated, and for each wavelength, the I of the measurement data is calculated. S , I C and the calculated data I S , I C The difference here is calculated by calculating the sum of squares of the residuals between the measured values ​​and the calculated data for adjacent wavelengths, and substituting the sum of squares of the residuals and the difference into Equation (8). The above-mentioned process calculates the difference taking into account measurement noise that differs depending on the wavelength, making it possible to perform fitting with high accuracy and derive more appropriate film thickness and optical constants for the thin film 22 to be measured.

[0094] The following describes the estimation accuracy of the optical constants of the thin film 22 estimated from the measurement data in the measurement process of the present disclosure. FIG. 13 is a graph showing the comparison results between true values ​​and calculated values. The true values ​​indicate the actual optical property values ​​of the thin film 22, and the calculated values ​​indicate fitting values ​​obtained after processing to reduce the influence of measurement noise, etc., on the estimated values ​​of the optical constants of the thin film 22 estimated from the measurement data in the measurement process of the present disclosure. The graph shown in FIG. 13 shows light energy on the horizontal axis and the refractive index n1(λ) of the thin film 22 on the vertical axis. The true values ​​of the refractive index are shown by solid lines, and the calculated values ​​of the refractive index are shown by dashed lines. Note that the upper graph in FIG. 13 shows the comparison results for the thin film 22 made of Si3N4 (silicon nitride), and the lower graph in FIG. 13 shows the comparison results for the thin film 22 made of HfO2 (hafnium oxide).

[0095] 13, by using the measurement process of the present disclosure, it is possible to calculate calculated values ​​(estimated values) of the refractive index and extinction coefficient that are small in difference from the actual values ​​of the optical properties (refractive index and extinction coefficient) of the thin film 22. By being able to identify optical constants that are close to the true values ​​(optical constants of the actual thin film) of the thin film 22 to be measured in this way, it is possible to derive with high accuracy the film thickness and optical properties of the thin film 22 to be measured based on the optical constants identified with high accuracy.

[0096] The matters described in each embodiment can be combined with each other. Furthermore, the independent claims and dependent claims described in the claims can be combined with each other in any combination, regardless of the reference format. Furthermore, the claims use a format in which a claim references two or more other claims (multiple claim format), but this is not limited to this. A multiple claim (multi-multi claim) that references at least one other multiple claim may also be used.

[0097] The embodiments disclosed herein are illustrative in all respects and should not be considered limiting. The scope of the present invention is defined by the claims, not by the meaning described above, and is intended to include all modifications within the meaning and scope of the claims. In other words, embodiments obtained by combining technical means appropriately modified within the scope of the claims are also included in the technical scope of the present invention.

[0098] 7 Input unit 8 Display unit 10 Ellipsometer 20 Sample 30 Data processing device 31 Calculation unit 34 Storage unit

Claims

1. An information processing method in which a computer executes a process of acquiring a tentative film thickness of a thin film, acquiring measured values ​​of reflected or transmitted light from a sample containing said thin film measured by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, estimating the optical properties of the thin film from which the tentative film thickness has been acquired using a model that estimates the optical properties of the thin film based on the film thickness, the measured values ​​of reflected or transmitted light from a sample containing said thin film measured by said optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and deriving the optical properties of the thin film by fitting the acquired measured values ​​based on the estimated optical properties of the thin film.

2. The information processing method according to claim 1, wherein the computer executes a process for deriving the film thickness of the thin film based on the optical characteristics of the thin film.

3. The information processing method according to claim 1 or 2, wherein the optical film thickness measurement device is a spectroscopic ellipsometer, and the computer executes a process of: acquiring, for each of a plurality of wavelengths, the amplitude ratio and phase difference between p-polarized light and s-polarized light in the reflected light or transmitted light from the sample measured by the spectroscopic ellipsometer; and the angle of incidence at the time of measurement; using the model, estimating, for each of the plurality of wavelengths, the optical characteristics of the thin film based on the tentative film thickness of the thin film, the amplitude ratio and phase difference between p-polarized light and s-polarized light in the reflected light or transmitted light from the sample measured by the spectroscopic ellipsometer; and the wavelength and angle of incidence of the incident light at the time of measurement; and deriving the optical characteristics of the thin film by fitting the amplitude ratio and phase difference acquired for the plurality of wavelengths based on the optical characteristics of the thin film estimated for the plurality of wavelengths.

4. The model is a learning model that has been trained using training data including the film thickness of the thin film, the measured values ​​of the reflected light or transmitted light from the sample measured by the optical film thickness measuring device, the wavelength and incident angle of the incident light at the time of measurement, and the optical properties of the thin film; and the information processing method described in any one of claims 1 to 3, wherein the computer inputs the acquired tentative film thickness of the thin film, the measured values ​​of the reflected light or transmitted light from the sample measured by the optical film thickness measuring device, and the wavelength and incident angle of the incident light at the time of measurement into the learning model, and executes a process to obtain the optical properties of the thin film estimated by the learning model.

5. The information processing method according to any one of claims 1 to 4, wherein the computer executes a process for estimating the optical properties of the thin film without using the dielectric function of the thin film.

6. An information processing method according to any one of claims 1 to 5, wherein the computer executes a process of deriving the optical characteristics of the thin film by performing fitting based on the estimated optical characteristics of the thin film while changing the temporary film thickness.

7. The measured values ​​include an amplitude ratio and a phase difference between p-polarized light and s-polarized light in the reflected light or transmitted light from the sample, and the optical properties of the thin film estimated using the model are modified according to a predetermined rule to generate a plurality of optical property candidates, and the amplitude ratio and phase difference between p-polarized light and s-polarized light in the reflected light or transmitted light from the sample including the thin film having each of the optical property candidates are calculated based on each of the plurality of optical property candidates, the tentative film thickness of the thin film, and the wavelength and incident angle of the incident light at the time of measurement, and from among the plurality of optical property candidates, optical property candidates are identified that have small differences between the amplitude ratio and phase difference between p-polarized light and s-polarized light in the reflected light or transmitted light from the sample including the thin film having each of the optical property candidates and the amplitude ratio and phase difference between p-polarized light and s-polarized light in the reflected light or transmitted light from the sample measured by the optical film thickness measurement device, The information processing method according to any one of claims 1 to 6, wherein the computer executes a process of deriving the optical properties of the thin film by fitting based on the identified optical property candidate, the tentative film thickness of the thin film, the measured amplitude ratio and phase difference, and the wavelength and incident angle of the incident light at the time of measurement.

8. An information processing method according to any one of claims 1 to 7, wherein the model includes at least one of: a first model that estimates the refractive index of the thin film based on the film thickness of the thin film, the measured value of reflected light or transmitted light from the sample measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement; and a second model that estimates the extinction coefficient of the thin film based on the film thickness of the thin film, the measured value of reflected light or transmitted light from the sample measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement.

9. An information processing method according to any one of claims 1 to 8, wherein the model includes at least one of: a first model that estimates the refractive index of the thin film based on the film thickness of the thin film, the measured value of the reflected light or transmitted light from the sample measured by the optical film thickness measurement device, the wavelength and incident angle of the incident light at the time of measurement, and the extinction coefficient of the thin film; and a second model that estimates the extinction coefficient of the thin film based on the film thickness of the thin film, the measured value of the reflected light or transmitted light from the sample measured by the optical film thickness measurement device, the wavelength and incident angle of the incident light at the time of measurement, and the refractive index of the thin film.

10. An information processing method according to any one of claims 1 to 9, wherein the computer executes a process of deriving the optical characteristics of the thin film by weighting the residual between the measurement values ​​acquired for each wavelength and simulation data calculated using the optical characteristics of the thin film estimated for each wavelength and fitting the residual.

11. The information processing method according to claim 10, wherein the computer executes a process of deriving the optical properties of the thin film by fitting based on the residual and a weighted sum of the absolute values ​​or square values ​​of the differences in the optical properties of the thin film estimated for adjacent wavelengths.

12. An information processing method according to any one of claims 1 to 11, wherein the computer executes a process of estimating the optical characteristics of the thin film based on a value obtained by multiplying the tentative film thickness of the thin film by the reciprocal of the wavelength of the incident light at the time of the measurement, the measurement value of the sample measured by the optical film thickness measurement device, and the angle of incidence at the time of the measurement.

13. An information processing method according to any one of claims 1 to 12, wherein the model estimates the optical characteristics of the thin film based on the film thickness of the thin film, the measured values ​​of reflected or transmitted light from the sample measured by the optical film thickness measurement device, the wavelength and incident angle of the incident light at the time of the measurement, and the optical characteristics of a substrate on which the thin film is formed in the sample.

14. The information processing method according to any one of claims 1 to 13, wherein the computer executes a process of acquiring a tentative film thickness of the thin film from an apparatus involved in the process of forming the thin film.

15. The information processing method according to any one of claims 1 to 14, wherein the computer executes a process for displaying the derived optical characteristics of the thin film on a display unit.

16. A computer program that causes a computer to execute the following process: acquiring a tentative film thickness of a thin film; acquiring measured values ​​of reflected or transmitted light from a sample containing said thin film measured by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement; estimating the optical properties of the thin film from which the tentative film thickness has been acquired using a model that estimates the optical properties of the thin film based on the film thickness, the measured values ​​of reflected or transmitted light from a sample containing said thin film measured by said optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement; and deriving the optical properties of the thin film by fitting the acquired measured values ​​based on the estimated optical properties of the thin film.

17. An information processing device having a control unit, wherein the control unit acquires a tentative film thickness of a thin film, acquires measured values ​​of reflected or transmitted light from a sample including the thin film measured by an optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, estimates the optical properties of the thin film from which the tentative film thickness has been acquired using a model that estimates the optical properties of the thin film based on the film thickness of the thin film, the measured values ​​of reflected or transmitted light from a sample including the thin film measured by the optical film thickness measurement device, and the wavelength and incident angle of the incident light at the time of measurement, and performs processing to derive the optical properties of the thin film by fitting the acquired measured values ​​based on the estimated optical properties of the thin film.

Images