Intelligent wide-spectrum film thickness real-time monitoring system and method

Abstract

The application discloses a kind of intelligent wide-spectrum film thickness real-time monitoring system and method, it is related to online monitoring technical field, the system includes: approximate model acquisition module, synchronous design parameter and match and obtain approximate film model set;Model update module, according to design parameter and material intrinsic parameter update model parameter, generate alternative model set;Spectrum acquisition module, through wide-spectrum light source collection reflected light and convert into reflected interference spectrum;Model adjustment and output module, based on nonlinear least square method iteration optimization alternative model, output the target film model of minimum residual error;Monitoring result extraction module, extract film thickness from target model as film thickness monitoring result.Further achieve the technical effect of improving thickness measurement adaptability, improve thickness measurement precision.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology, and in particular to an intelligent broadband film thickness real-time monitoring system and method. Background Technology

[0002] Current thin film thickness measurement technologies mainly rely on empirical formulas and fixed model parameters, which are difficult to adapt to complex and ever-changing thin film parameters and material properties. Furthermore, fixed model parameters need to be set in advance and cannot be dynamically adjusted to cope with changes in actual production. These technologies are insufficient in terms of accuracy and real-time performance, especially when dealing with thin films with different design parameters and intrinsic material parameters, making it difficult to achieve high-precision thickness monitoring. Summary of the Invention

[0003] This invention provides an intelligent broadband film thickness real-time monitoring system and method to solve the technical problem of poor parameter and model adaptability in the prior art, which affects the thickness measurement efficiency and accuracy, and achieve the technical effect of improving thickness measurement adaptability and thickness measurement accuracy.

[0004] In a first aspect, the present invention provides an intelligent broadband film thickness real-time monitoring system, wherein the intelligent broadband film thickness real-time monitoring system comprises:

[0005] The approximate model acquisition module is used to synchronize the target thin film design parameters and obtain an approximate thin film model set based on the target thin film design parameters.

[0006] The model update module is used to update the model parameters of multiple approximate thin film models by assigning difference values ​​based on the design parameters of the target thin film and the intrinsic parameters of the target thin film material, so as to obtain a set of candidate approximate thin film models.

[0007] The spectral acquisition module is used to activate a broadband light source to illuminate the target thin film, collect the reflected light signal, and convert it into a reflected interference spectrum.

[0008] The model adjustment and output module is used to iteratively adjust multiple candidate approximate thin film models in the candidate approximate thin film model set based on the nonlinear least squares method and the reflection interference spectrum as a reference, and output the model that minimizes the residual between the model reflection spectrum and the reflection interference spectrum as the target thin film model.

[0009] The monitoring result extraction module is used to extract the film thickness in the target thin film model as the film thickness monitoring result.

[0010] In one feasible implementation, the approximate thin film model set includes multiple physical thin film structure models, and the optical parameter combinations of the multiple physical thin film structure models are configured differently.

[0011] In one feasible implementation, the target thin film design parameters are synchronized, and an approximate thin film model set is obtained by matching the target thin film design parameters. The execution steps of the approximate model acquisition module include:

[0012] Based on the target thin film design parameters, thin film structure parameters are extracted, wherein the thin film structure parameters include at least the number of thin film layers and the layer thickness distribution.

[0013] By traversing historical monitoring records of film thickness, an approximate match is made based on the film structure parameters and the preset tolerance.

[0014] The physical thin film structure model that meets the preset tolerance in terms of both the number of layers and the thickness is extracted as the approximate thin film model, and the approximate thin film model set is obtained.

[0015] In one feasible implementation, based on the target thin film design parameters and the intrinsic parameters of the target thin film material, the model parameters of multiple approximate thin film models are updated by assigning difference values ​​to obtain a set of candidate approximate thin film models. The execution steps of the model update module include:

[0016] Based on the intrinsic parameters of the target thin film material, a difference recognition limit is defined.

[0017] The thin film model is extracted by iterating through multiple approximate thin film models and targeting the optical parameter combination that does not meet the difference recognition limit.

[0018] The optical parameter combination is extracted and updated based on the intrinsic parameters of the target thin film material, and the updated thin film model is put back into the approximate thin film model set, and the output is the candidate approximate thin film model set.

[0019] In one feasible implementation, a broadband light source is activated to illuminate the target thin film, the reflected light signal is collected and converted into a reflected interference spectrum, and the execution steps of the spectral acquisition module include:

[0020] The target thin film surface is irradiated by a broadband light source.

[0021] The reflected light signal and the reference light signal generated in the reference optical path are acquired synchronously and then normalized in intensity.

[0022] Based on the intensity-normalized reflected light signal and the reference light signal, optical signal interference operation is performed to generate a reflection interference spectrum.

[0023] In one feasible implementation, based on the nonlinear least squares method, multiple candidate approximate thin film models in the candidate approximate thin film model set are iteratively adjusted using the reflection interference spectrum as a reference. The target thin film model is the one whose residual between the model reflection spectrum and the reflection interference spectrum is minimized. The execution steps of the model adjustment and output module include:

[0024] Using each candidate approximate thin film model as initialization input, a theoretical reflection spectrum is generated through simulation.

[0025] The theoretical reflection spectrum and the reflection interference spectrum are aligned in the frequency domain to construct a residual function. The model parameters are then iteratively optimized using a nonlinear least squares algorithm to minimize the residual function. The model parameters include a combination of thin film thickness and optical parameters.

[0026] The candidate approximate thin film model set is traversed, and the candidate approximate thin film model corresponding to the smallest residual function value is extracted as the target thin film model.

[0027] In one feasible implementation, the execution steps also include:

[0028] Obtain the optical parameter combination for each target thin film model to acquire intrinsic material verification data.

[0029] The intrinsic verification data of the materials are classified and divided using the material ID as an index.

[0030] Based on the classification results, the intrinsic verification indicators of each material ID are statistically analyzed using the accumulator. The intrinsic verification indicators include at least one of the following: intrinsic parameter offset rate, intrinsic parameter cumulative offset, and intrinsic parameter continuous offset.

[0031] Material intrinsic verification is triggered based on the material intrinsic verification indicators and the preset material intrinsic verification constraints.

[0032] Secondly, the present invention also provides a method for real-time monitoring of intelligent broadband film thickness, wherein the method includes:

[0033] Synchronize the target thin film design parameters and obtain an approximate thin film model set based on the target thin film design parameters.

[0034] Based on the target thin film design parameters and the intrinsic parameters of the target thin film material, the model parameters of multiple approximate thin film models are updated by assigning differences to obtain a set of candidate approximate thin film models.

[0035] A broadband light source is activated to illuminate the target thin film, and the reflected light signal is collected and converted into a reflection interference spectrum.

[0036] Based on the nonlinear least squares method, multiple candidate approximate thin film models in the candidate approximate thin film model set are iteratively adjusted with the reflection interference spectrum as the benchmark, and the model that minimizes the residual between the reflection spectrum and the reflection interference spectrum is the target thin film model.

[0037] The film thickness in the target thin film model is extracted as the film thickness monitoring result.

[0038] This invention discloses an intelligent broadband film thickness real-time monitoring system and method, comprising: an approximate model acquisition module, used to synchronously receive the design parameters of the target thin film and match and search in a preset model library based on the design parameters to obtain a corresponding set of approximate thin film models; a model update module, used to combine the design parameters of the target thin film with its corresponding intrinsic material parameters to differentiate and adjust the model parameters in the set of approximate thin film models, and construct a set of candidate approximate thin film models; a spectrum acquisition module, used to control a broadband light source to irradiate the target thin film, acquire the reflected light signal and perform spectral conversion to obtain the corresponding reflection interference spectrum; a model adjustment and output module, used to use the reflection interference spectrum as the fitting target, perform parameter iterative optimization on multiple models in the set of candidate approximate thin film models based on a nonlinear least squares algorithm, and select the model with the smallest residual between the reflection spectrum and the reflection interference spectrum as the target thin film model output; and a monitoring result extraction module, used to extract the corresponding film thickness information from the target thin film model as the film thickness monitoring result output. This invention solves the technical problem of poor parameter and model adaptability affecting thickness measurement efficiency and accuracy, and achieves the technical effects of improving thickness measurement adaptability and accuracy. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the structure of an intelligent broadband film thickness real-time monitoring system according to the present invention.

[0040] Figure 2 This is a flowchart illustrating a smart broadband film thickness real-time monitoring method according to the present invention.

[0041] Figure labeling: Approximate model acquisition module 11, model update module 12, spectrum acquisition module 13, model adjustment and output module 14, monitoring result extraction module 15. Detailed Implementation

[0042] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.

[0043] Example 1, as Figure 1This is a flowchart illustrating a real-time intelligent broadband film thickness monitoring system according to the present invention, wherein the intelligent broadband film thickness monitoring system includes:

[0044] The approximate model acquisition module 11 is used to synchronize the target thin film design parameters and obtain an approximate thin film model set based on the target thin film design parameters.

[0045] Specifically, synchronizing the target thin film design parameters includes obtaining key design data such as the number of thin film layers, thickness, and material. Matching and obtaining a set of approximate thin film models involves finding similar thin film structure models in a pre-set model library based on the above design parameters, thereby providing an initial model foundation for subsequent analysis and calculations. This ensures that subsequent model adjustments have a suitable starting point, reduces computational load, and improves monitoring efficiency.

[0046] Specifically, an approximate thin film model set refers to a collection of physical thin film structure models that have similar design parameters to the target thin film. These models have different combinations of optical parameters to accommodate the characteristics of different thin films.

[0047] Through the above process, a targeted initial model foundation is provided for the entire monitoring system, ensuring that the system can quickly focus on a model range that closely resembles the actual thin film properties. This helps improve the initial accuracy of model matching and reduces the computational workload of subsequent adjustments.

[0048] In some embodiments, the approximate thin film model set includes multiple physical thin film structure models, and the optical parameter combinations of the multiple physical thin film structure models are configured differently.

[0049] Specifically, the approximate thin film model set contains multiple physical thin film structure models, and each model represents a possible thin film structure. The differentiated configuration of optical parameters means that the multiple approximate thin film models differ in optical characteristic parameters such as the refractive index and optical thickness of the thin film material. This can cover the range of optical characteristics under different thin film monitoring needs, and thus provide a diverse model basis for accurately matching the actual thin film characteristics.

[0050] In some embodiments, the target thin film design parameters are synchronized, and an approximate thin film model set is obtained by matching the target thin film design parameters. The execution steps of the approximate model acquisition module 11 include:

[0051] Based on the target thin film design parameters, thin film structure parameters are extracted, wherein the thin film structure parameters include at least the number of thin film layers and the layer thickness distribution; historical monitoring records of film thickness are traversed, and the thin film structure parameters are approximated by matching them with a preset tolerance; physical thin film structure models that satisfy the preset tolerance in terms of both the number of layers and the thickness are extracted as the approximate thin film models, and the approximate thin film model set is obtained.

[0052] Specifically, target thin film design parameters refer to the design specifications of the optical thin film expected to be achieved in the process, including material type, number of layers, thickness of each layer, refractive index distribution, etc., to achieve specific spectral performance. Thin film structure parameters are the key physical features extracted from the design parameters for matching, mainly including the number of thin film layers (i.e., the number of layers stacked) and the thickness of each layer (i.e., layer thickness distribution).

[0053] Specifically, film thickness history monitoring records refer to historical process data recorded during actual production, including the actual number and thickness of deposited layers in different batches of films, and the physical film structure model determined during monitoring. Preset tolerance refers to the allowable range of parameter deviations, used to determine whether a historical record can be considered approximately approximating the target design. The approximate film model set refers to a set of physical models selected from historical data that are sufficiently close to the target design in terms of structural parameters.

[0054] Specifically, the design parameters of the target thin film are first synchronized, and the thin film structure parameters for matching are extracted from them. For example, the design target is a two-layer structure with layer thicknesses of [80, 120] nm. Then, the historical film thickness monitoring database is traversed, and each historical record is compared with the target structure to determine if the number of layers is consistent, and to calculate whether the deviation of each layer thickness from the target value is within a preset tolerance range, such as ±3 nm. If the thickness error of all layers is within the tolerance range, the historical record can be considered a valid approximate thin film model, and the corresponding physical thin film structure model determined in the historical record is extracted. Further, the above process is repeated until the physical thin film structure models of all historical records that meet the conditions are finally extracted, forming an approximate thin film model set.

[0055] The above process enables the rapid extraction of realistic physical thin film models similar to the target design from historical data without re-simulation or fabrication. This ensures that a similar initial model can be found when dealing with target thin films with different optical properties. This helps improve the initial accuracy and adaptability of model matching, reduces the computational scope and difficulty of subsequent model adjustments, and thus effectively improves the thickness measurement accuracy and efficiency of the entire monitoring system.

[0056] The model update module 12 is used to update the model parameters of multiple approximate thin film models by assigning difference values ​​based on the design parameters of the target thin film and the intrinsic parameters of the target thin film material, so as to obtain a set of candidate approximate thin film models.

[0057] Specifically, the intrinsic parameters of the target thin film material involve the inherent optical properties of the material, such as refractive index and absorption coefficient, while the model parameters are the combination of parameters used to describe the optical properties of the thin film. Based on the specific parameters of the target thin film, the optical parameters of the initial model can be adjusted to generate an alternative model that is closer to the properties of the target thin film.

[0058] In some embodiments, based on the target thin film design parameters and the intrinsic parameters of the target thin film material, the model parameters of multiple approximate thin film models are updated by difference assignment to obtain a set of candidate approximate thin film models. The execution steps of the model update module 12 include:

[0059] Based on the intrinsic parameters of the target thin film material, a difference recognition limit is defined; multiple approximate thin film models are traversed, and thin film models are extracted with the goal of optical parameter combinations that do not meet the difference recognition limit; the optical parameter combinations of the extracted thin film models are updated according to the intrinsic parameters of the target thin film material, and the updated thin film models are put back into the approximate thin film model set, and the output is the candidate approximate thin film model set.

[0060] Specifically, the difference recognition limit is a threshold standard used to determine whether there are significant differences between the approximate thin film model and the target design at the optical parameter level. It is usually based on the deviation of refractive index, extinction coefficient, or a combination thereof at a specific wavelength. For example, the refractive index deviation at a certain wavelength can be set to no more than ±0.05 as the recognition limit. The optical parameter combination is a set of parameters used to describe the optical behavior of a thin film material, including refractive index, extinction coefficient, dispersion model parameters, etc. at different wavelengths.

[0061] Specifically, firstly, a difference recognition limit is set based on the intrinsic parameters of the target thin film material. For example, at a wavelength of 550 nm, the refractive index of the target material H is 2.35, and the system sets the recognition limit to ±0.03, meaning that any model in the historical model where the refractive index of this layer is outside the range [2.32, 2.38] is considered to have a significant difference. Then, multiple approximate thin film models are iterated, and the optical parameter combinations of each layer are compared to extract models where the refractive index or extinction coefficient of at least one layer exceeds the recognition limit. These models are considered to be those requiring updating. Next, the optical parameters in these models that do not meet the recognition limit are updated according to the intrinsic parameters of the corresponding material in the target design. For example, the refractive index of a certain layer is updated from 2.28 to the target value of 2.35, or its dispersion curve is replaced entirely with the fitted model of the target material. The updated models are then reintroduced into the approximate thin film model set, ultimately forming a candidate approximate thin film model set for subsequent simulation and analysis.

[0062] Through the above process, a dual matching from structural approximation to accurate optical parameters can be achieved. This ensures that the candidate model not only closely approximates the target design in structural parameters such as layer thickness and number of layers, but also maintains consistency with the target material in terms of optical performance. In other words, the above process effectively narrows the range of differences in model parameters, improves the accuracy and adaptability of the model, provides a better initial model foundation for subsequent precise model adjustments based on reflection interferometry, and provides more reliable support for accurately extracting film thickness.

[0063] The spectral acquisition module 13 is used to activate a broadband light source to irradiate the target thin film, acquire the reflected light signal, and convert it into a reflected interference spectrum.

[0064] Specifically, acquiring reflected light signals involves using a photodetector to capture the light reflected from the thin film surface and converting it into an electrical signal. Converting it into a reflection interference spectrum involves processing the acquired reflected light signal using an interferometer, that is, using the principle of light interference to convert the reflected light signal into spectral data in the form of interference fringes for subsequent analysis.

[0065] In some embodiments, a broadband light source is activated to illuminate the target thin film, the reflected light signal is collected and converted into a reflected interference spectrum, and the execution steps of the spectral acquisition module 13 include:

[0066] A broadband light source is controlled to irradiate the surface of the target thin film; the reflected light signal and the reference light signal generated in the reference optical path are simultaneously acquired and their intensity is normalized; based on the intensity-normalized reflected light signal and the reference light signal, optical signal interference operation is performed to generate a reflection interference spectrum.

[0067] Specifically, a broadband light source refers to a light source that emits a continuous spectrum covering a wide wavelength range, exemplified by halogen lamps, supercontinuum lasers, or white LEDs. The reflected light signal refers to the light signal reflected back from the surface and internal interfaces of the target thin film after broadband light illuminates it; it contains information about the intensity changes caused by multilayer interference. The reference light signal originates from the reference optical path and is acquired synchronously with the measured optical path, used for phase reference or background correction.

[0068] Specifically, intensity normalization refers to the scaling or difference processing of the reflected light signal and the reference light signal to eliminate the effects of light source power fluctuations and system response inconsistencies. The reflection interferometry spectrum is a spectral signal obtained through interferometric calculations, reflecting the interference modulation characteristics of the thin film structure on light, and is used for subsequent inversion of film thickness and optical parameters.

[0069] Specifically, the spectral acquisition module 13 first controls a broadband light source to illuminate the target thin film. This light source covers the designed wavelength range, such as 400–900 nm, to ensure the acquisition of complete optical interference information. After the broadband light illuminates the thin film, multiple reflections and interferences occur at the interfaces of each layer, thus forming a specific reflection interference pattern. The system can simultaneously acquire optical signals from two channels: the signal reflected from the target thin film and the reference optical signal obtained through the reference optical path.

[0070] Furthermore, the reflected light signal and the reference light signal are subjected to intensity normalization to remove the influence of light source instability or optical path loss. Then, based on the normalized reflected signal and the reference signal, an interference operation is performed to generate a reflection interference spectrum. This spectrum exhibits interference fringes that vary with wavelength, and its period and amplitude are related to the thickness and refractive index of the thin film.

[0071] The above process can obtain high-quality and accurate reflection interference spectra, thus providing basic data support for subsequent inversion calculations of thin film parameters.

[0072] The model adjustment and output module 14 is used to iteratively adjust multiple candidate approximate thin film models in the candidate approximate thin film model set based on the nonlinear least squares method and the reflection interference spectrum as a reference, and output the model that minimizes the residual between the model reflection spectrum and the reflection interference spectrum as the target thin film model.

[0073] Specifically, nonlinear least squares is a mathematical method that minimizes the sum of squared residuals between model predictions and actual observations through iterative optimization. This includes the Levenberg-Marquardt (LM) algorithm and trust region methods. In this embodiment, by iteratively adjusting multiple models in the candidate approximate thin film model set based on the reflection interferogram, the best-fit model between the model reflection spectrum and the reflection interferogram can be found. The film thickness is then obtained based on the obtained best-fit model (i.e., the target thin film model). Here, the model reflection spectrum is the theoretical spectrum calculated based on the current model parameters, while the reflection interferogram is the actual acquired optical property data of the target thin film. The residual refers to the difference between the two, and the model with the smallest residual is the target thin film model.

[0074] In some embodiments, based on the nonlinear least squares method, multiple candidate approximate thin film models in the candidate approximate thin film model set are iteratively adjusted using the reflection interference spectrum as a reference, and the model that minimizes the residual between the model reflection spectrum and the reflection interference spectrum is output as the target thin film model. The execution steps of the model adjustment and output module 14 include:

[0075] Using each candidate approximate thin film model as initialization input, a theoretical reflection spectrum is generated through simulation. The theoretical reflection spectrum and the reflection interference spectrum are aligned in the frequency domain to construct a residual function. The model parameters are then iteratively optimized using a nonlinear least squares algorithm to minimize the residual function. The model parameters include a combination of thin film thickness and optical parameters. The candidate approximate thin film model with the smallest residual function value is extracted as the target thin film model and the set of candidate approximate thin film models is traversed.

[0076] Specifically, the reflection interference spectrum is spectral data collected from actual thin film samples, containing information on the interference modulation of light by the thin film structure. The theoretical reflection spectrum is the reflection spectrum calculated through optical simulation based on candidate thin film models (including film structure, thickness, and material optical parameters).

[0077] Specifically, firstly, each candidate approximate thin-film model is used as the initial input, and the corresponding theoretical reflection spectrum is generated using optical simulation software. Next, the theoretical reflection spectrum is aligned with the actually acquired reflection interference spectrum in the frequency domain (or wavelength domain) to ensure that both have the same sampling accuracy and starting wavelength.

[0078] Furthermore, a residual function is constructed, which can be defined as the sum of the squares of the light intensity differences between the two at each wavelength. Then, a nonlinear least squares algorithm is used to iteratively adjust the model parameters, such as the film thickness and refractive index of each layer. After each iteration, the theoretical spectrum is recalculated and the residual is updated until the residual function converges or reaches a set minimum value. For example, after multiple iterations, the model parameters are updated to a thickness of 102 nm and a refractive index of 1.52, at which point the residual function value decreases significantly. All candidate models are traversed, and the model with the smallest residual is finally selected as the target thin film model, i.e., the physical model that best matches the actual measurement data.

[0079] Through the above process, the optimal model is automatically selected from multiple approximate models. Furthermore, a nonlinear least squares optimization algorithm is used to fine-tune the model parameters, ensuring that the final output target thin film model matches the experimental data in terms of reflectance spectrum. This helps improve the accuracy and reliability of thin film modeling and avoids the inefficiency and subjectivity of manual parameter tuning.

[0080] The monitoring result extraction module 15 is used to extract the film thickness in the target film model as the film thickness monitoring result.

[0081] Specifically, the film thickness parameters are obtained from the best-fit model obtained after iterative optimization using the nonlinear least squares method, and these parameters are determined as the actual thickness of the current film.

[0082] Specifically, the target thin film model obtained after model adjustment contains parameter information that most closely approximates the actual thin film properties, and the thin film thickness is the result that the monitoring system needs to measure. For example, when monitoring a three-layer thin film structure, if the target thin film model after iterative adjustment shows that the thicknesses of each layer are 102 nm, 201 nm, and 149 nm, then these values ​​can be directly used as the real-time thickness monitoring data output for the thin film, providing real-time and accurate thickness monitoring results for the thin film manufacturing process.

[0083] In some embodiments, the execution steps of the intelligent broadband film thickness real-time monitoring system further include:

[0084] Obtain the optical parameter combination for each target thin film model to acquire intrinsic material verification data; classify the intrinsic material verification data using material ID as an index; based on the classification results, combine the accumulator to statistically analyze the intrinsic material verification indicators for each material ID, wherein the intrinsic material verification indicators include at least one of intrinsic material parameter offset rate, cumulative intrinsic material parameter offset, and continuous intrinsic material parameter offset; trigger intrinsic material verification based on the intrinsic material verification indicators and preset intrinsic material verification constraints.

[0085] Specifically, optical parameter combinations typically refer to the optical constants of the materials in each layer of a thin film model; intrinsic material verification data refers to historical or real-time collected reference data related to the optical properties of materials, used to determine whether a material has experienced performance drift under different process batches or conditions. A material ID is a unique identifier for different materials in the database.

[0086] Specifically, intrinsic material verification indicators refer to statistical quantities extracted from multiple target thin film models to reflect the stability and variation trend of material optical parameters, preferably including:

[0087] Parameter offset rate is the relative rate of change of the current parameter relative to the reference value; cumulative offset is the sum of multiple historical offsets; consecutive offset is the total offset value when the model parameter offset directions are consistent for several consecutive times.

[0088] Specifically, the intrinsic material verification constraints are preset judgment criteria used to trigger the verification mechanism, such as an offset rate exceeding 2% or more than 3 consecutive offsets. These intrinsic material verification constraints correspond one-to-one with the aforementioned intrinsic material verification indicators.

[0089] Specifically, after acquiring each target thin film model, its corresponding optical parameter combinations are extracted, such as the refractive index of materials in different models. Then, the data is categorized by material ID, and the changes in optical parameters for each type of material at different time points and in different process batches are statistically analyzed. Next, using an accumulator mechanism, the parameter change trend of each material ID is continuously tracked, and the aforementioned intrinsic verification indicators are calculated. For example, if the refractive index of a material is more than 2% higher than the reference value in five consecutive measurements, its continuous offset can be recorded as positive, and the cumulative offset continuously increases.

[0090] Furthermore, when any intrinsic verification indicator of a material exceeds the preset verification constraint threshold, such as a continuous offset exceeding 0.15, intrinsic verification of the material is automatically triggered to indicate that the material may have undergone changes in physical composition, process conditions, or equipment status, and further analysis or calibration is required.

[0091] Through the above process, long-term monitoring and intelligent evaluation of the optical properties of materials in actual production environments can be achieved. This not only allows for the early identification of minute drifts in the intrinsic parameters of materials, but also avoids model misjudgments or film thickness measurement errors caused by changes in material properties.

[0092] In summary, the intelligent broadband film thickness real-time monitoring system provided by this invention has the following technical effects:

[0093] The approximate model acquisition module 11 is used to synchronously receive the design parameters of the target thin film and match and search in the preset model library based on the design parameters to obtain the corresponding approximate thin film model set; the model update module 12 is used to combine the design parameters of the target thin film with its corresponding material intrinsic parameters to differentiate and adjust the model parameters in the approximate thin film model set to construct a candidate approximate thin film model set; the spectrum acquisition module 13 is used to control a broadband light source to irradiate the target thin film, collect the reflected light signal and perform spectral conversion to obtain the corresponding reflection interference spectrum; the model adjustment and output module 14 is used to use the reflection interference spectrum as the fitting target, perform parameter iterative optimization on multiple models in the candidate approximate thin film model set based on the nonlinear least squares algorithm, and select the model with the smallest residual between the reflection spectrum and the reflection interference spectrum as the target thin film model output; the monitoring result extraction module 15 is used to extract the corresponding film thickness information from the target thin film model as the thin film thickness monitoring result output, thereby achieving the technical effects of improving thickness measurement adaptability and thickness measurement accuracy.

[0094] Example 2, as Figure 2 This is a flowchart illustrating a real-time intelligent broadband film thickness monitoring method according to the present invention. For example, Figure 1 The schematic diagram of the intelligent broadband film thickness real-time monitoring system of the present invention can be used to achieve, for example... Figure 2 The process is shown below.

[0095] Based on the same concept as the intelligent broadband film thickness real-time monitoring system in the above embodiment, the present invention also provides an intelligent broadband film thickness real-time monitoring method comprising:

[0096] S100: Synchronize the target thin film design parameters and obtain an approximate thin film model set based on the target thin film design parameters.

[0097] S200: Based on the target thin film design parameters and the intrinsic parameters of the target thin film material, the model parameters of multiple approximate thin film models are updated by assigning differences to obtain a set of candidate approximate thin film models.

[0098] S300: Activates a broadband light source to illuminate the target thin film, collects the reflected light signal, and converts it into a reflected interference spectrum.

[0099] S400: Based on the nonlinear least squares method, multiple candidate approximate thin film models in the candidate approximate thin film model set are iteratively adjusted with the reflection interference spectrum as the benchmark, and the model with the smallest residual between the reflection spectrum and the reflection interference spectrum is the target thin film model.

[0100] S500: Extract the film thickness from the target thin film model as the film thickness monitoring result.

[0101] In some embodiments, the approximate thin film model set includes multiple physical thin film structure models, and the optical parameter combinations of the multiple physical thin film structure models are configured differently.

[0102] In some embodiments, synchronizing target thin film design parameters and obtaining an approximate thin film model set based on the target thin film design parameters includes:

[0103] Based on the target thin film design parameters, thin film structure parameters are extracted, wherein the thin film structure parameters include at least the number of thin film layers and the layer thickness distribution; historical monitoring records of film thickness are traversed, and the thin film structure parameters are approximated by matching them with a preset tolerance; physical thin film structure models that satisfy the preset tolerance in terms of both the number of layers and the thickness are extracted as the approximate thin film models, and the approximate thin film model set is obtained.

[0104] In some embodiments, the model parameters of multiple approximate thin film models are updated by assigning differences based on the target thin film design parameters and the intrinsic parameters of the target thin film material to obtain a set of candidate approximate thin film models, including:

[0105] Based on the intrinsic parameters of the target thin film material, a difference recognition limit is defined; multiple approximate thin film models are traversed, and thin film models are extracted with the goal of optical parameter combinations that do not meet the difference recognition limit; the optical parameter combinations of the extracted thin film models are updated according to the intrinsic parameters of the target thin film material, and the updated thin film models are put back into the approximate thin film model set, and the output is the candidate approximate thin film model set.

[0106] In some embodiments, activating a broadband light source to illuminate the target thin film, collecting the reflected light signal, and converting it into a reflected interference spectrum includes:

[0107] A broadband light source is controlled to irradiate the surface of the target thin film; the reflected light signal and the reference light signal generated in the reference optical path are simultaneously acquired and their intensity is normalized; based on the intensity-normalized reflected light signal and the reference light signal, optical signal interference operation is performed to generate a reflection interference spectrum.

[0108] In some embodiments, based on the nonlinear least squares method, multiple candidate approximate thin film models in the candidate approximate thin film model set are iteratively adjusted using the reflection interferogram as a reference, and the model that minimizes the residual between the reflection spectrum and the reflection interferogram is output as the target thin film model, including:

[0109] Using each candidate approximate thin film model as initialization input, a theoretical reflection spectrum is generated through simulation. The theoretical reflection spectrum and the reflection interference spectrum are aligned in the frequency domain to construct a residual function. The model parameters are then iteratively optimized using a nonlinear least squares algorithm to minimize the residual function. The model parameters include a combination of thin film thickness and optical parameters. The candidate approximate thin film model with the smallest residual function value is extracted as the target thin film model and the set of candidate approximate thin film models is traversed.

[0110] In some embodiments, the intelligent broadband film thickness real-time monitoring method further includes:

[0111] Obtain the optical parameter combination for each target thin film model to acquire intrinsic material verification data; classify the intrinsic material verification data using material ID as an index; based on the classification results, combine the accumulator to statistically analyze the intrinsic material verification indicators for each material ID, wherein the intrinsic material verification indicators include at least one of intrinsic material parameter offset rate, cumulative intrinsic material parameter offset, and continuous intrinsic material parameter offset; trigger intrinsic material verification based on the intrinsic material verification indicators and preset intrinsic material verification constraints.

[0112] It should be understood that the focus of the embodiments mentioned in this specification is their difference from other embodiments. The specific embodiments in the aforementioned Embodiment 1 are also applicable to the intelligent broadband film thickness real-time monitoring method described in Embodiment 2. For the sake of brevity, they will not be further elaborated here.

[0113] It should be understood that the embodiments disclosed in this invention and the above description enable those skilled in the art to implement this invention. However, this invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention, and should all be included within the protection scope of this invention.

Claims

1. An intelligent wide spectrum film thickness real time monitoring system characterized in that, The method comprises the following steps: An approximate model acquisition module is used to synchronize target thin film design parameters and acquire an approximate thin film model set according to the target thin film design parameters; A model updating module is used to perform difference assignment updating on model parameters of multiple approximate thin film models according to the target thin film design parameters and intrinsic parameters of a target thin film material, and acquire a candidate approximate thin film model set; A spectrum acquisition module is used to activate a wide-spectrum light source to irradiate a target thin film, acquire a reflected light signal and convert it into a reflected interference spectrum; A model adjustment and output module is used to iteratively adjust multiple candidate approximate thin film models in the candidate approximate thin film model set based on a nonlinear least square method and taking the reflected interference spectrum as a reference, and output a target thin film model that minimizes the residual error between a model reflection spectrum and the reflected interference spectrum; A monitoring result extraction module is used to extract a thin film thickness in the target thin film model as a film thickness monitoring result; The execution steps of the model adjustment and output module include: Taking each candidate approximate thin film model as an initialization input, a theoretical reflection spectrum is simulated and generated; The theoretical reflection spectrum and the reflected interference spectrum are aligned in a frequency domain, a residual error function is constructed, and model parameters are iteratively optimized based on a nonlinear least square algorithm to minimize the residual error function; wherein the model parameters include a combination of a thin film thickness and optical parameters; The candidate approximate thin film model set is traversed, and a candidate approximate thin film model corresponding to the minimum residual error function value is extracted as the target thin film model.

2. An intelligent wide spectrum film thickness real time monitoring system as claimed in claim 1, wherein, The approximate thin film model set includes multiple physical thin film structure models, and the optical parameter combinations of the multiple physical thin film structure models are differentially configured.

3. An intelligent wide spectrum film thickness real time monitoring system as claimed in claim 2, wherein, The execution steps of the approximate model acquisition module include: Based on the target thin film design parameters, thin film structure parameters are extracted, wherein the thin film structure parameters at least include the number of thin film layers and the layer thickness distribution; The film thickness historical monitoring records are traversed, and approximate matching is performed based on the thin film structure parameters and a preset tolerance; The physical thin film structure model that satisfies the preset tolerance in both the number of layers and the thickness is extracted as the approximate thin film model, and the approximate thin film model set is acquired.

4. An intelligent wide spectrum film thickness real time monitoring system as claimed in claim 3, wherein, The execution steps of the model updating module include: According to the intrinsic parameters of the target thin film material, a difference identification limit is defined; The multiple approximate thin film models are traversed, and the thin film model extraction is performed with the optical parameter combination not satisfying the difference identification limit as the target; The assignment updating of the optical parameter combination is performed on the thin film model extraction according to the intrinsic parameters of the target thin film material, and the thin film model after the assignment updating is put back into the approximate thin film model set and output as the candidate approximate thin film model set.

5. An intelligent wide spectrum film thickness real time monitoring system as claimed in claim 1, wherein, The wide-spectrum light source is activated to irradiate the target film, the reflected light signal is collected and converted into a reflection interference spectrum, and the execution steps of the spectrum collection module include: controlling the wide-spectrum light source to irradiate the surface of the target film; synchronously collecting the reflected light signal and a reference light signal generated in a reference light path, and performing intensity normalization processing; based on the reflected light signal and the reference light signal after the intensity normalization processing, performing light signal interference operation to generate a reflection interference spectrum.

6. An intelligent wide spectrum film thickness real time monitoring system as claimed in claim 1, wherein, The execution steps further include: obtaining an optical parameter combination of each target film model, and obtaining material intrinsic verification data; classifying the material intrinsic verification data according to the material ID as an index; according to the classification result, combining the accumulator to statistically obtain a material intrinsic verification indication of each material ID, wherein the material intrinsic verification indication includes at least one of a material intrinsic parameter offset rate, a material intrinsic parameter cumulative offset amount, and a material intrinsic parameter continuous offset amount; according to the material intrinsic verification indication and a preset material intrinsic verification constraint, triggering material intrinsic verification.

7. A method for intelligent wide spectrum film thickness real time monitoring characterized in that, The method is applied to the system of any one of claims 1-6, and includes: synchronizing target film design parameters, and obtaining an approximate film model set according to the target film design parameters; according to the target film design parameters and target film material intrinsic parameters, performing difference assignment update on model parameters of multiple approximate film models to obtain a candidate approximate film model set; activating the wide-spectrum light source to irradiate the target film, collecting the reflected light signal and converting it into a reflection interference spectrum; based on a nonlinear least square method, iteratively adjusting multiple candidate approximate film models in the candidate approximate film model set according to the reflection interference spectrum as a reference, and outputting one of the candidate approximate film models as the target film model, which minimizes the residual error between the model reflection spectrum and the reflection interference spectrum; extracting the film thickness in the target film model as a film thickness monitoring result.

Images