Spectral dimension reduction method, model training method, and structural parameter measurement method

Abstract

This specification provides a spectral dimensionality reduction method, a model training method, and a method for measuring structural parameters. The method includes: processing the spectrum to be processed to obtain a d-dimensional restored baseline spectrum. If the restoration accuracy of the baseline restored spectrum meets a first preset restoration accuracy condition, then the d-dimensional dimension is taken as the target spectral dimension. If not, the following steps are iteratively executed: determining k in each iteration and determining a k-dimensional restoration matrix; for the first iteration, determining a ()-dimensional restored candidate spectrum based on the baseline restored spectrum and the k-dimensional restoration matrix; for iterations beyond the first iteration, determining the candidate restored spectrum for the current iteration based on the candidate restored spectrum determined in the previous iteration and the k-dimensional restoration matrix determined in the current iteration; if the candidate restored spectrum in any iteration meets a second preset restoration accuracy condition, the iteration stops, and the candidate restored dimension of the candidate restored spectrum in any iteration is taken as the target spectral dimension.

Description

Technical Field

[0001] This specification relates to the field of semiconductor manufacturing process technology, and in particular to spectral dimensionality reduction methods, model training methods, and structural parameter measurement methods. Background Technology

[0002] With the development of the semiconductor integrated circuit manufacturing industry, the critical dimensions in the process are getting smaller and smaller, and the number of device structural parameters that need to be controlled is increasing. Traditional optical imaging analysis methods cannot meet the measurement of critical dimensions in the process.

[0003] Optical critical dimension (OCD) technology is a non-contact measurement technique used to measure the size and morphology of semiconductor microstructures. The basic measurement process involves using an OCD measuring device (such as an ellipsometer or reflectometer) to actually illuminate the sample, thereby collecting the diffracted light signal from the sample to obtain the measurement spectrum. The measured spectrum is then matched against a vast database of structural parameters and their corresponding theoretical spectral values ​​stored in a theoretical spectral database. The structural parameter corresponding to the theoretical spectrum that best matches the measured spectrum is selected as the measured value of the actual structural parameter of the sample.

[0004] As semiconductor geometries become increasingly complex and structural parameters multiply, the amount of data in theoretical spectral databases grows exponentially. Therefore, theoretical spectral databases typically require very large storage spaces.

[0005] To address the aforementioned technical issues, patent (CN119377682A) proposes a spectral calculation model. Based on structural parameters and theoretical spectra from a theoretical spectral database as training samples, the spectral calculation model is trained to output corresponding spectral values ​​for input structural parameters. In applications measuring the structural parameters of a sample, a large number of structural parameters can be temporarily sampled, and the corresponding predicted spectral values ​​can be calculated instantly based on the spectral calculation model. Then, the measured spectrum of the sample to be tested is matched with each predicted spectrum to filter out target predicted spectra that meet the matching conditions. The target structural parameter corresponding to the target predicted spectrum is then used as the measured value of the actual structural parameter of the sample to be tested. This patent "replaces" the theoretical spectral database with a spectral calculation model, saving storage space and improving the flexibility of matching.

[0006] The spectral calculation model in this patent outputs predicted spectral values ​​for the target spectral dimension after dimensionality reduction. Therefore, the target spectral dimension needs to be determined in advance before model training. A suitable target spectral dimension is found to balance model training complexity and spectral information retention. The approach is as follows: first, multiple candidate spectral dimensions are set, and calculations are performed round by round for each candidate spectral dimension to select the target spectral dimension. Specifically, for each selected candidate spectral dimension, the original spectral dimension of the spectrum to be processed needs to be reduced to the candidate spectral dimension, and then the original spectral dimension is restored. The error between the restored spectrum and the original spectrum to be processed is calculated to evaluate whether the information retention of the original spectral dimension reduced to the current candidate spectral dimension meets the restoration accuracy requirements. If it does, the current candidate spectral dimension is selected as the target spectral dimension.

[0007] However, the computational cost of each round of dimensionality reduction and restoration calculations is positively correlated with the size of the candidate spectral dimension. In conventional dimensionality reduction and restoration calculations, each round requires calculations based on all data components of the current candidate spectral dimension. Therefore, the higher the candidate spectral dimension, the greater the computational overhead and the higher the computational performance requirements of the computer hardware. Summary of the Invention

[0008] To overcome the problems existing in related technologies, this specification provides a spectral dimensionality reduction method, a model training method, and a method for measuring structural parameters.

[0009] According to a first aspect of the embodiments of this specification, a spectral dimensionality reduction method is provided, the method comprising:

[0010] For the first spectral dimension The spectrum to be processed is subjected to dimensionality reduction and restoration processing to obtain... The baseline restoration spectrum for dimensional restoration;

[0011] If the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition, then the... The dimension is the target spectral dimension after the dimensionality reduction of the spectrum to be processed.

[0012] If not satisfied, the following steps are executed iteratively:

[0013] The iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing 3D restored matrix;

[0014] For the first iteration, based on the reference restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration;

[0015] For iterations other than the first round, the candidate restoration spectra determined in the previous round and those determined in this round are used as the basis for the current round. The 3D restoration matrix determines the candidate restored spectra for this round;

[0016] If the candidate restored spectrum in any round meets the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in any round is taken as the target spectral dimension.

[0017] in, , and All are positive integers, and .

[0018] According to a second aspect of the embodiments of this specification, a method for calculating a restored spectrum is provided, the method comprising:

[0019] For the first spectral dimension The spectrum to be processed is subjected to dimensionality reduction and restoration processing to obtain... The restored spectrum of the microstructure;

[0020] Sure( The restored spectrum of the 3D restored dimension is located at the tail of the restored dimension. Restored 3D restored matrix;

[0021] Based on the above The restored spectrum of the dimensional restoration and the said Determining the dimensional restoration matrix ( The restored spectrum of 3D model;

[0022] in, , and All are positive integers, and .

[0023] According to a third aspect of the embodiments of this specification, a method for training a spectral calculation model is provided, the method comprising:

[0024] Obtain the training dataset, wherein each training sample in the training dataset includes a set of structural parameters and a first spectral dimension for describing the grating structure. The first theoretical spectrum;

[0025] The first theoretical spectrum is subjected to dimensionality reduction and restoration to obtain The baseline restoration spectrum for dimensional restoration;

[0026] If the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition, then the... The dimension is the target spectral dimension after the first theoretical spectrum is reduced in dimensionality.

[0027] If not satisfied, the following steps are executed iteratively:

[0028] The iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing 3D restored matrix;

[0029] For the first iteration, based on the reference restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration;

[0030] For iterations other than the first round, the candidate restoration spectra determined in the previous round and those determined in this round are used as the basis for the current round. The 3D restoration matrix determines the candidate restored spectra for this round;

[0031] If the candidate restored spectrum in any round meets the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in that round is taken as the target spectral dimension; wherein... , and All are positive integers, and ;

[0032] The structural parameters are input into the spectral calculation model to output the predicted spectrum of the target spectral dimension;

[0033] The spectral calculation model is trained based on the loss value between the predicted spectrum and the second theoretical spectrum of the target spectral dimension; wherein the second theoretical spectrum is obtained by dimensionality reduction of the first theoretical spectrum. According to a fourth aspect of an embodiment of this specification, a method for measuring structural parameters is provided, the method comprising:

[0034] Obtain all structural parameters generated from the structural model of the sample to be tested;

[0035] The structural parameters are respectively input into the spectral calculation model trained as in the third aspect to output the predicted spectrum corresponding to each structural parameter;

[0036] Obtain the measurement spectrum obtained from the sample to be tested;

[0037] The measured spectra are matched with each predicted spectrum to select target predicted spectra that meet the matching conditions, and the target structural parameters corresponding to the target predicted spectra are used as the measured values ​​of the actual structural parameters of the sample to be tested.

[0038] According to a fifth aspect of the embodiments of this specification, another method for measuring structural parameters is provided, the method comprising:

[0039] Obtain all structural parameters generated from the structural model of the sample to be tested;

[0040] The structural parameters are input into the spectral calculation model to output the predicted spectrum corresponding to each structural parameter.

[0041] Obtain the measurement spectrum obtained from the sample to be tested;

[0042] Determine the difference between the dimension of the measured spectrum and the dimension of the predicted spectrum. And determine the location of the measured spectrum at the tail of the dimension. Restored The dimension of each predicted spectrum is smaller than the dimension of the measured spectrum; It is a positive integer;

[0043] Based on the above The dimension-restored matrix upscals each predicted spectrum to the same dimension as the measured spectrum.

[0044] The measured spectra are matched with the predicted spectra after dimensionality upgrades to select target predicted spectra that meet the matching conditions, and the target structural parameters corresponding to the target predicted spectra are used as the measured values ​​of the actual structural parameters of the sample to be tested.

[0045] According to a sixth aspect of the embodiments of this specification, a spectral dimensionality reduction device is provided, the device comprising:

[0046] The first reference restored spectrum determination module is used for determining the first spectral dimension. The spectrum to be processed is subjected to dimensionality reduction and restoration processing to obtain... The baseline restoration spectrum for dimensional restoration;

[0047] The first target spectral dimension determination module is used to determine the target spectral dimension if the restoration accuracy of the reference restored spectrum meets a first preset restoration accuracy condition. The dimension is taken as the target spectral dimension after dimensionality reduction of the spectrum to be processed; if this condition is not met, the following steps are performed iteratively: the iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing The dimensional restoration matrix; for the first iteration, based on the baseline restored spectrum and the... Determining the dimensional restoration matrix ( The process involves several iterations: 1) determining the candidate restored spectra for k-dimensional restoration; 2) determining the candidate restored spectra for the current iteration based on the candidate restored spectra determined in the previous iteration and the k-dimensional restoration matrix determined in the current iteration; 3) stopping the iteration if the candidate restored spectra in any iteration satisfy the second preset restoration accuracy condition, and using the candidate restored dimension of the candidate restored spectra in that iteration as the target spectral dimension; where... , and All are positive integers, and .

[0048] According to a seventh aspect of the embodiments of this specification, a training apparatus for a spectral calculation model is provided, the apparatus comprising:

[0049] The training dataset acquisition module is used to acquire the training dataset, wherein each training sample in the training dataset includes a set of structural parameters describing the grating structure and a first spectral dimension. The first theoretical spectrum;

[0050] The second reference restored spectrum determination module is used to perform dimensionality reduction and restoration processing on the first theoretical spectrum to obtain... The baseline restoration spectrum for dimensional restoration;

[0051] The second target spectral dimension determination module is used to determine the target spectral dimension if the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition. The dimension is taken as the target spectral dimension after the dimensionality reduction of the first theoretical spectrum; if this condition is not met, the following steps are performed iteratively: the iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing The dimensional restoration matrix; for the first iteration, based on the baseline restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration; for non-first iterations, based on the candidate restored spectra determined in the previous round and the current round... The dimensional restoration matrix determines the candidate restored spectra for this round; if the candidate restored spectra in any round satisfy the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectra in that round is taken as the target spectral dimension; wherein, , and All are positive integers, and ;

[0052] A predicted spectrum output module is used to input the structural parameters into the spectral calculation model to output a predicted spectrum of the target spectral dimension;

[0053] The model parameter training module is used to train the spectral calculation model based on the loss value between the predicted spectrum and the second theoretical spectrum of the target spectral dimension; wherein the second theoretical spectrum is obtained by dimensionality reduction of the first theoretical spectrum.

[0054] According to an eighth aspect of the embodiments of this specification, a device for measuring structural parameters is provided, the device comprising:

[0055] The structural parameter acquisition module is used to acquire all structural parameters generated from the structural model of the sample under test.

[0056] The predictive spectrum module is used to input the structural parameters into the spectral calculation model trained as in the third aspect to output the predicted spectrum corresponding to each structural parameter.

[0057] The measurement spectrum acquisition module is used to acquire the measurement spectrum obtained by measuring the sample to be tested;

[0058] The sample structure measurement module is used to match the measured spectrum with each predicted spectrum to screen out the target predicted spectrum that meets the matching conditions, and use the target structural parameters corresponding to the target predicted spectrum as the measured values ​​of the actual structural parameters of the sample to be tested.

[0059] According to a ninth aspect of the embodiments of this specification, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method as described in any one of the first to ninth aspects.

[0060] According to an eighth aspect of the embodiments of this specification, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method as described in any one of the first to ninth aspects.

[0061] The technical solutions provided in the embodiments of this specification may include the following beneficial effects:

[0062] In the embodiments of this specification, in the application of selecting the target spectral dimension for the spectral calculation model, this scheme can, in each iteration, base the candidate restored spectrum obtained from the previous round of calculation, and combine it with the iteration step size. Restored The dimensional restoration matrix is ​​used to jointly determine ( Candidate restoration spectra for 3D restoration. That is, only one restoration spectrum needs to be restored per round. Dimensionality reduction and restoration calculations are performed on the dimensionality components without requiring a complete ( Dimensionality reduction and restoration calculations are performed on the data components of 3D.

[0063] As can be seen, this solution can reduce the performance requirements of computer hardware by reducing the computational load of dimensionality reduction and restoration calculations.

[0064] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description

[0065] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this specification and, together with the description, serve to explain the principles of this specification.

[0066] Figure 1 This is a schematic diagram illustrating a conventional method for finding the spectral dimension of a target according to an exemplary embodiment of this specification.

[0067] Figure 2 This is a flowchart illustrating a spectral dimensionality reduction method according to an exemplary embodiment of this specification.

[0068] Figure 3 This is a schematic diagram illustrating, according to an exemplary embodiment, a method for calculating the restored spectrum of a P-dimensional restoration.

[0069] Figure 4 This is a schematic diagram illustrating, according to an exemplary embodiment, a method for comparing conventional methods of calculating reconstructed spectra and an improved method.

[0070] Figure 5 This is a flowchart illustrating a training method for a spectral calculation model according to an exemplary embodiment of this specification.

[0071] Figure 6 This is a schematic diagram illustrating a conventional method for calculating calibration evaluation metrics according to an exemplary embodiment of this specification.

[0072] Figure 7 This is a schematic diagram illustrating an improved method for calculating calibration evaluation metrics according to an exemplary embodiment of this specification.

[0073] Figure 8 This is a flowchart illustrating a method for measuring structural parameters according to an exemplary embodiment of this specification.

[0074] Figure 9 This is a flowchart illustrating a method for calculating a restored spectrum according to an exemplary embodiment of this specification.

[0075] Figure 10This is a flowchart illustrating another method for measuring structural parameters according to an exemplary embodiment of this specification.

[0076] Figure 11 This is a schematic diagram of the structure of an electronic device according to an exemplary embodiment of this specification.

[0077] Figure 12 This is a block diagram illustrating a spectral dimensionality reduction device according to an exemplary embodiment of this specification.

[0078] Figure 13 This is a block diagram of a training apparatus for a spectral calculation model, as illustrated in this specification according to an exemplary embodiment.

[0079] Figure 14 This is a block diagram illustrating a structural parameter measuring device according to an exemplary embodiment of this specification. Detailed Implementation

[0080] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this specification as detailed in the appended claims.

[0081] The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of this specification. The singular forms “a,” “the,” and “the” as used in this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0082] It should be understood that although the terms first, second, third, etc., may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this specification, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0083] In optical critical dimension (OCD) measurement of a grating structure under test, before constructing a theoretical spectral database, it is necessary to determine the grating structure based on the design drawings and manufacturing process information of the product under test, thereby establishing a structural model of the structure under test. From this structural model, key structural parameters to be measured can be selected. For example, for a grating structure under test in an integrated circuit, its key structural parameters may include critical dimensions (CD), such as the top and bottom widths of the grating lines; sidewall angles (SWA) relative to the substrate surface; and the total height (HT) of the grating structure. Of course, other structural parameters may also be included; the measured variables differ for different structures under test, and will not be elaborated upon here. A structural parameter vector x = (CD, SWA, HT, ...) can represent a set of all structural parameters of the grating structure under test.

[0084] Engineers can set the value range and variation step size for each structural parameter, and iterate through all possible combinations of these structural parameters. For each set of structural parameters, the corresponding theoretical spectrum can be calculated using the Rigorous Coupled Wave Analysis (RCWA) algorithm.

[0085] Finally, a theoretical spectral database is established that stores these structural parameters and their corresponding theoretical spectra.

[0086] As semiconductor geometries become increasingly complex and the number of structural parameters grows, the data volume of theoretical spectral databases is increasing exponentially. This is primarily because each factor—the number of structural parameters, the number of step points for structural parameters, the number of wavelength points, and the spectral dimensions—causes a significant increase in the data volume of the theoretical spectral database. For example, adding even a single structural parameter, even with a step size of only two points (upper and lower limits), doubles the data volume of the theoretical spectral database, and so on. The amount of data in the theoretical spectral database at each step point becomes the original Even if the step size of a single structural parameter is doubled, the amount of data in the theoretical spectral database will double. The structural parameters are The impact of changing the number of wavelength points on the data volume of a theoretical spectral database is similar to that of changing the number of steps for a single structural parameter; the spectral dimension doubles for every two-fold increase. Therefore, theoretical spectral databases typically require a very large amount of storage space, some requiring storage in server clusters, resulting in poor portability, inconvenience for users, and significant time-consuming data extraction, which can easily lead to memory overflow.

[0087] The relevant patents propose a spectral calculation model that can "replace" theoretical spectral databases, saving storage space, improving matching flexibility, and solving the problems of difficult storage and application migration of theoretical spectral databases.

[0088] In addition, other technical solutions have proposed a structural parameter calculation model, which directly outputs the structural parameters corresponding to the measured spectrum based on the input measured spectrum. This allows engineers to directly predict structural parameters based on the structural parameter calculation model, instead of matching pre-set structural parameters from a theoretical spectral database, thus saving the matching step.

[0089] Compared to structural parameter calculation models, spectral calculation models still use the matching approach of "theoretical spectral databases" but add the advantage of reverse verification. That is, the structural parameters of the sample can be accurately measured and input into the spectral calculation model to output the predicted spectrum. Finally, the output predicted spectrum can be compared with the measured spectrum of the actual sample. The smaller the error between the two, the higher the accuracy of the spectral calculation model, thus allowing reverse verification of the model accuracy.

[0090] The applicant discovered in practice that the original theoretical spectrum has too high a spectral dimensionality, and the number of parameters that the spectral calculation model needs to handle increases explosively with the increase in dimensionality. To address this issue, the conventional approach is to determine the target spectral dimensionality after dimensionality reduction before model training, and then the spectral calculation model outputs the predicted spectrum with the target spectral dimensionality instead of the original spectral dimensionality. This optimization of the spectral calculation model's performance is achieved by compressing the dimensionality of the data.

[0091] However, compressing the spectral dimension to a smaller value is not always better. Excessive dimensionality reduction leads to significant loss of spectral information, while excessively high spectral dimensions increase model complexity and computational cost. Therefore, it is necessary to find a suitable target spectral dimension to balance model complexity and spectral information retention, while minimizing computation time while ensuring computational accuracy.

[0092] like Figure 1 As shown, a large number of theoretical spectra can be obtained from the theoretical spectral database as the original spectra to be processed. Assume the size of the original spectral dimension of the original spectra to be processed is... Dimensions can be found to The process iterates through the dimensions. For any dimension traversed, the original spectrum to be processed is reduced in dimension and restored. The process then determines whether the spectrum restored based on the current dimension meets the restoration accuracy.

[0093] For example, in the first round, it can be initialized. Dimensionality is used as the target for dimensionality reduction; the original spectrum to be processed is first reduced to dimensionality. The original spectral dimension is then restored through restoration calculations. The accuracy of the restoration is determined by comparing the error between the restored spectrum and the original spectrum.

[0094] If the requirements are not met, additional steps can be taken based on the first round. In the second round, we can ( Using dimensionality reduction as the target, the original spectrum to be processed is first reduced to ( ). The original spectral dimension is then restored through restoration calculations. The accuracy of the restoration is determined by comparing the error between the restored spectrum and the original spectrum.

[0095] And so on, if the ( If the condition is not met in round (), then it can continue in round (). Add to the base of the wheel Dimension. Assume that in the 1st... In the wheel, based on ( If the error between the restored spectrum and the original spectrum meets the restoration accuracy, then the dimension of the target spectrum can be determined as ( ). )dimension.

[0096] While related technologies have proposed improvements to the methods for finding the target spectral dimension, their focus is on reducing the number of traversals required. For example, using a binary search method to quickly narrow down the traversal range can reduce the number of traversals required to find the target spectral dimension from 100 to 60.

[0097] However, for each round of dimensionality reduction and restoration calculations, the relevant techniques still rely on the selected full dimensionality for the current round. For example, suppose the dimensionality reduction calculation is to ( Then restore the calculation to the original value. Dimension, the related technologies are still based on ( Dimensionality reduction and restoration calculations are performed on a 3D data matrix.

[0098] Therefore, the applicant found that there is still room for improvement in the way the dimensionality reduction and restoration calculations are performed in each round, and hopes to reduce the overall computational cost of finding the target spectral dimension by reducing the computational cost of each round of dimensionality reduction and restoration calculations.

[0099] To address the aforementioned technical problems, this specification provides a spectral dimensionality reduction method to solve the issues of high computational overhead and high performance requirements for computer hardware when selecting a suitable target spectral dimension for a spectral calculation model.

[0100] The embodiments described in this specification will now be described in detail.

[0101] like Figure 2 As shown, Figure 2 This is a flowchart illustrating a spectral dimensionality reduction method according to an exemplary embodiment, comprising the following steps:

[0102] For the first spectral dimension The spectrum to be processed is subjected to dimensionality reduction and restoration processing to obtain... The baseline restoration spectrum for dimensional restoration;

[0103] If the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition, then the... The dimension is the target spectral dimension after the dimensionality reduction of the spectrum to be processed.

[0104] If not satisfied, the following steps are executed iteratively:

[0105] The iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing 3D restored matrix;

[0106] For the first iteration, based on the reference restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration;

[0107] For iterations other than the first round, the candidate restoration spectra determined in the previous round and those determined in this round are used as the basis for the current round. The 3D restoration matrix determines the candidate restored spectra for this round;

[0108] If the candidate restored spectrum in any round meets the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in any round is taken as the target spectral dimension.

[0109] in, , and All are positive integers, and .

[0110] In applications where target spectral dimensions are selected for spectral computation models, this approach, in each iteration, can use candidate reconstructed spectra obtained from the previous round of calculation, combined with the iteration step size. Restored The dimensional restoration matrix is ​​used to jointly determine ( Candidate restoration spectra for 3D restoration. That is, only one restoration spectrum needs to be restored per round. Dimensionality reduction and restoration calculations are performed on the dimensionality components without requiring a complete ( Dimensionality reduction and restoration calculations are performed on the data components of 3D.

[0111] As can be seen, this solution can reduce the performance requirements of computer hardware by reducing the computational load of dimensionality reduction and restoration calculations.

[0112] The spectrum to be processed can be selected from theoretical spectra stored in a theoretical spectral database. By selecting real theoretical spectra as the object to be processed, it can be ensured that the final selected target spectral dimension can accurately reflect the information retention of the real theoretical spectrum.

[0113] like Figure 3 As shown, The meaning of the restored spectrum in the first spectral dimension is that the spectrum to be processed in the first spectral dimension has been reduced to the second spectral dimension by dimensionality reduction calculation. P The spectrum of the first spectral dimension is obtained, and then it is restored to the spectrum of the first spectral dimension through restoration calculation.

[0114] P The amount of information that the restored spectrum retains depends on the algorithm used for dimensionality reduction and restoration, as well as the size of the spectral dimension. Specifically, the algorithm used for dimensionality reduction and restoration can be based on the principles of Principal Component Analysis (PCA).

[0115] The generality of the above naming conventions is not difficult to understand. The baseline restored spectrum for dimensionality reduction is the spectrum to be processed in the first spectral dimension, which is then processed by dimensionality reduction calculation. The spectrum of the first spectral dimension is obtained, and then it is restored to the spectrum of the first spectral dimension through restoration calculation. Similarly, ( The candidate restored spectrum for dimensionality reduction is the spectrum to be processed in the first spectral dimension, which is then processed by dimensionality reduction calculation. The spectrum of the first spectral dimension is obtained by performing a restoration calculation, and then the spectrum of the first spectral dimension is restored to its original value.

[0116] ( Candidate restoration spectra of 3D restoration Candidate restoration spectra or The candidate restoration spectrum for dimensionality reduction is determined by the traversal direction. If the traversal is performed from low to high dimensionality reduction, then in the previous round... On the basis of That is, what is calculated in the current round is Candidate restored spectra for dimensionality reduction; if the traversal is performed from high to low dimensionality, then in the previous round... Reduce on the basis of maintenance That is, what is calculated in the current round is Candidate restoration spectra for visceral restoration.

[0117] The applicant proved through theoretical deduction that, Candidate restoration spectra and There is a certain mathematical relationship between the baseline restored spectrum and the standard restored spectrum. Based on the known baseline restored spectrum, it is not necessary to perform a complete calculation (…). To obtain the 3D restoration matrix ( The candidate restored spectra for 3D restoration only require calculation. The 3D restoration matrix, combined with the baseline restoration spectrum calculated in the previous round, can be used to calculate ( Candidate restoration spectra for 3D restoration.

[0118] Next, let's take an example of traversing the dimensions from smallest to largest. The mathematical relationship described above is derived using the formula, and the derivation process is as follows:

[0119] Let the data matrix of the spectrum to be processed be . The centered data matrix after centering the spectrum to be processed is: ,in For sample size, This is the first spectral dimension. The data matrix after dimensionality reduction and compression of the spectral center to be processed is... ,in The dimension after dimensionality reduction is an uncertain value and needs to be determined based on the accuracy of the restoration. Let's assume the dimension has now been reduced to... Wei, then Specifically .

[0120] The restoration calculation formula is:

[0121]

[0122] for The data matrix of the baseline restored spectrum. It is the mean data matrix of the spectrum to be processed. For the spectrum to be processed Principal component loading matrix of dimension, for The transpose of .

[0123] Now we need to calculate the increase. Candidate restored spectra of dimensionality, i.e., based on Data matrix of candidate restored spectra obtained by dimensional component restoration The calculation formula is:

[0124]

[0125]

[0126] in, Candidate restoration dimensions based on candidate restored spectra Tail The restoration obtained The dimensional restoration matrix will be the... The dimensional restoration matrix as 3D restored matrix for( The restoration dimension of the candidate restored spectrum. Tail Principal component loading matrix composed of dimensional elements, for The transpose of .

[0127] It is easy to conclude that, ( Data matrix of candidate restored spectra in 3D Theoretically, Data matrix of the baseline restored spectrum of the dimensional reconstruction and 3D restored matrix sum.

[0128] For example, such as Figure 4 As shown, assuming we need to calculate ( The data matrix of candidate restored spectra in dimensionality is given. Existing solutions (represented by dashed lines) involve dimensionality reduction of the spectra to be processed. The spectrum to be processed is ) dimension, and then the original calculation is performed to restore it to ( The restored spectrum is obtained from the 1980-dimensional reconstruction. Therefore, existing methods aim to calculate (the restored spectrum). The data matrix of candidate restored spectra for 3D restoration needs to be calculated. This solution only requires calculation. And combined with the calculations already made in the previous round The data matrix of the baseline restored spectrum can then be obtained ( The data matrix of candidate restored spectra in 3D reduces the number of spectral depths. The computational cost of the data components in a dimension.

[0129] Without loss of generality, traverse in descending order of dimension. Data matrix of candidate restored spectra in 3D Theoretically, Data matrix of the baseline restored spectrum of the dimensional reconstruction Reconstruction dimension compared to the baseline reconstructed spectrum Tail Restored 3D restored matrix difference.

[0130] Therefore, the dimensionality reduction and restoration calculation formulas in this scheme can be applied to traversal methods with dimensions from high to low and from low to high.

[0131] Next, to verify the correctness of the above theoretical derivation, the applicant used data from multiple theoretical spectral databases as training datasets to verify the application of... Figure 4 The differences in computational speed between existing schemes and the improved method presented in this paper are shown in Table 1. The final results are shown in Table 1.

[0132] Table 1

[0133]

[0134] As shown in Table 1, NM7755, NM0316, NM003, and NM931 represent different data sources, such as data from theoretical spectral databases corresponding to different semiconductor structures. 10, 20, 50, and 100 represent the number of iterations. `old` represents the time used for restoration calculations by existing methods, and `new` represents the time used for restoration calculations by this method. `Speed` is (old-new) / old, which indicates the improvement in calculation speed compared to existing methods.

[0135] It is not difficult to conclude, through practice, that this method can reduce the massive amount of computation in existing solutions, improve the speed of restoration calculation, and the time saved becomes more significant as the number of traversals increases. It is especially suitable for scenarios with excessively high spectral dimensions that require multiple traversals to find the target spectral dimension.

[0136] Next, this manual will illustrate the practical application of the above-derived conclusions by traversing from low to high dimension and from high to low dimension respectively:

[0137] (1) Traverse in the direction from low to high dimension.

[0138] In one embodiment, following the dimensionality from low to high, if the restoration accuracy of the reference restored spectrum is not less than a preset restoration accuracy and the restoration accuracy of the restored spectrum in the next lower dimension of the reference restored spectrum is less than the preset restoration accuracy, then the... Dimension is used as the target spectral dimension;

[0139] If the restoration accuracy of the reference restored spectrum is less than the preset restoration accuracy, then the following steps are executed iteratively:

[0140] The iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing 3D restored matrix;

[0141] For the first iteration, the reference restored spectrum and the... The sum of the dimensional restored matrices is used as ( Candidate restored spectra for 3D restoration;

[0142] For iterations other than the first round, the candidate restored spectra determined in the previous round and the spectra determined in this round will be used together. The sum of the dimensional restoration matrices serves as the candidate restored spectra for this round;

[0143] If the restoration accuracy of the candidate restored spectrum in any round is not less than the preset restoration accuracy and the restoration accuracy of the restored spectrum in the next lower dimension of the candidate restored spectrum is less than the preset restoration accuracy, the iteration stops and the candidate restoration dimension of the candidate restored spectrum in any round is taken as the target spectral dimension.

[0144] in, .

[0145] Specifically, in In this case, the first preset restoration accuracy condition is that the restoration accuracy of the reference restored spectrum is not less than the preset restoration accuracy, and the second preset restoration accuracy condition is that the restoration accuracy of any round of candidate restored spectra is not less than the preset restoration accuracy, and the restoration accuracy of the restored spectrum in the next lower dimension of the candidate restored spectrum is less than the preset restoration accuracy. In the case of the first preset restoration accuracy condition and the second preset restoration accuracy condition, the restoration accuracy of the current round's restored spectrum is not less than the preset restoration accuracy, and the restoration accuracy of the lower dimension of the current round's restored spectrum is less than the preset restoration accuracy.

[0146] The accuracy of restoration can be measured by the error between the restored spectrum of the current cycle and the corresponding spectrum to be processed. For example, mean squared error (MSE) can be used. Of course, this specification does not limit the method of measuring restoration accuracy; for example, mean absolute error (MAE) and the square of the correlation coefficient can also be used.

[0147] For example, suppose the first spectral dimension of the spectrum to be processed is Then it can be from 1 to Select the initial reference spectral dimension For example, one can judge based on experience. Which value is closest to the target spectral dimension to reduce the number of iterations? Of course, it's also possible to... That is, traversing from the beginning. This specification does not restrict the determination of... The way of value.

[0148] In obtaining After restoring the baseline restored spectrum, if If the value is greater than 1, it can be determined whether the restoration accuracy of the reference restored spectrum is not less than the preset restoration accuracy and whether the restoration accuracy of the lower dimension of the reference restored spectrum is less than the preset restoration accuracy. If Then it can be determined whether the restoration accuracy of the reference restored spectrum is not less than the preset restoration accuracy.

[0149] The judgment result falls into three categories:

[0150] ① The restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition.

[0151] At this point, we can stop the iteration and... The dimension is the target spectral dimension after the dimensionality reduction of the spectrum to be processed.

[0152] ② The restoration accuracy of the reference restored spectrum is not less than the preset restoration accuracy, and the restoration accuracy of the restored spectrum in the next lower dimension is not less than the preset restoration accuracy.

[0153] If the restoration accuracy of the reference restored spectrum is not less than the preset restoration accuracy, it means that the restoration accuracy of the reference restored spectrum in the current dimension meets the restoration accuracy requirements. If the restoration accuracy of the restored spectrum in the next lower dimension of the reference restored spectrum is also not less than the preset restoration accuracy, it means that, under the premise of meeting the restoration accuracy requirements, the spectral dimension can be further reduced to find the smallest spectral dimension that just meets the preset restoration accuracy as the target spectral dimension.

[0154] At this point, it can be confirmed that the target spectral dimension must be between 1 and... Between dimensions, it can be seen from Start with one dimension and traverse in descending order of dimension. Alternatively, you can choose a new, smaller dimension. The value is then recalculated in the new... If the restoration accuracy of the reference restored spectrum is less than the preset restoration accuracy, then from... Start with dimension 1 and traverse in ascending order of dimension 2.

[0155] ③ The restoration accuracy of the reference restored spectrum is less than the preset restoration accuracy.

[0156] current If the accuracy of the restored reference spectrum does not meet the accuracy requirements, then... The restored spectrum from dimensions lower than a certain dimension will inevitably fail to meet the required restoration accuracy. At this point, we can continue searching in higher dimensions, traversing the spectrum from lower to higher dimensions.

[0157] You can first determine the iteration step size. And determine the candidate restored spectrum of this round in the candidate restored dimension ( ) tail Restored 3D restored matrix.

[0158] Therefore, the reference restored spectrum and The sum of the dimensional restored matrices is determined as ( Candidate restoration spectra for 3D restoration.

[0159] like( If the restoration accuracy of the candidate restored spectra in 3D meets the restoration accuracy requirement, the iteration can be stopped, and ( ) dimension as the target spectral dimension.

[0160] If the condition is not met, a new iteration step size can be determined. Then determine the current round's ( The candidate restored spectrum of the dimensional restoration is located in the candidate restoration dimension ( (tail) Restored 3D restored matrix.

[0161] Then, the previous round determined ( Candidate restoration spectra and those determined in this round of analysis. The sum of the dimensional restored matrices is determined as ( Candidate restoration spectra for 3D restoration.

[0162] like( If the restoration accuracy of the candidate restored spectra in 3D meets the restoration accuracy requirement, the iteration can be stopped, and ( ) dimension as the target spectral dimension.

[0163] If this condition is not met, the process continues to determine a new iteration step size. Then, a new calculation begins, with each round of calculation requiring only one calculation. The system can obtain the restoration matrix in one dimension, without having to calculate the restoration matrix in all candidate restoration dimensions of the candidate restoration spectrum in the current round, thus reducing the computational cost of the restoration matrix.

[0164] It should be noted that the iteration step size is determined in each iteration. It can be a fixed value or a variable value. For example, the iteration step size between different rounds. They can be different.

[0165] Assumption , You can first initialize a reference restoration spectrum for 3D restoration, the restoration accuracy of which is less than the preset restoration accuracy.

[0166] ① It is a fixed value.

[0167] For example, It can take the value 1.

[0168] In each subsequent iteration, candidate restored spectra for 4D and 5D restoration can be calculated sequentially. Since the iteration step size is 1, when determining whether the candidate restored spectra in any round meet the second preset restoration accuracy condition, it is sufficient to determine whether the restoration accuracy of the candidate restored spectra in the current round is not less than the preset restoration accuracy.

[0169] ② The value is a change, and Initialized to 4.

[0170] In each subsequent iteration, candidate restored spectra for 7D and 11D can be calculated sequentially. Assuming the restoration accuracy of the candidate restored spectra for 7D is less than a preset accuracy, and the candidate restored spectra for 11D and 10D are not less than the preset accuracy, then the target spectral dimension will likely fall between 8D and 10D. In this case, the iteration step size can be reduced to [missing value]. =1, starting from 8 dimensions and traversing towards 10 dimensions in a direction from low to high, or starting from 10 dimensions and traversing towards 8 dimensions in a direction from high to low.

[0171] (2) Traverse in the direction from high to low dimension.

[0172] In one embodiment, when traversing from high to low dimensions, if the restoration accuracy of the reference restored spectrum is not less than a preset restoration accuracy and the restoration accuracy of the restored spectrum in the next lower dimension of the reference restored spectrum is less than the preset restoration accuracy, then the... Dimension is used as the target spectral dimension;

[0173] If not satisfied, iteratively execute the following steps:

[0174] The iteration step size is determined in each iteration. And by using the candidate restoration spectrum from the previous round at the tail of the candidate restoration dimension Wei Fuxing 3D restored matrix.

[0175] For the first iteration, the reference restored spectrum and the... The difference between the dimensional restored matrices is used as ( Candidate restored spectra for 3D reconstruction. At this time... The dimensional restoration matrix is ​​located at the reference restored spectrum. Tail Restored 3D restored matrix.

[0176] For iterations other than the first round, the candidate restored spectra determined in the previous round and the spectra determined in this round will be used together. The difference between the dimensional restoration matrices is used as the candidate restored spectrum for this round. At this point, the spectrum determined in this round... The dimensional restoration matrix is ​​the candidate restored spectrum from the previous round located at the tail of the candidate restoration dimension. The restoration obtained 3D restored matrix.

[0177] If the restoration accuracy of the candidate restored spectrum in any round is not less than the preset restoration accuracy and the restoration accuracy of the restored spectrum in the next lower dimension of the candidate restored spectrum is less than the preset restoration accuracy, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in any round is taken as the target spectral dimension.

[0178] in, .

[0179] exist In this case, the iteration direction must be from high to low.

[0180] exist In this case, the iteration direction proceeds from high to low. For example, the restoration accuracy of the reference restored spectrum is not less than the preset restoration accuracy and ( If the restoration accuracy of the restored spectrum in dimension 1 is not less than the preset restoration accuracy, then iterative processing can be performed from high to low to find a suitable target spectral dimension.

[0181] There is a special case where the initial selection... The value is too small, causing the target spectral dimension to be missed, i.e. The accuracy of the restored baseline spectrum is less than the preset accuracy, and the target spectral dimension is greater than [missing value]. At this point, a larger one can be selected. The value is set such that the restoration accuracy of the newly selected reference restored spectrum is not less than the preset restoration accuracy, and then the traversal continues in the direction of descending dimension. Alternatively, it can start from the current reference restored spectrum and... to Traverse the spectral range from low to high dimensions until the target spectral dimension is found.

[0182] Similar to the aforementioned implementation that traverses dimensions from low to high, each iteration... The value can be 1, or it can be greater than 1 in the early stages, quickly traversing to the vicinity of the target spectral dimension, and then narrowing down. The value of is determined to accurately locate the position of the target spectral dimension.

[0183] It should be noted that the aforementioned traversal from high to low dimension or from low to high dimension is not mutually exclusive in practice, and can be used in combination. For example, when traversing from low to high dimension, due to the iteration step size... If the value is too large and the target spectral dimension is missed, it means the target spectral dimension is before the current spectral dimension. Therefore, we can traverse backwards from the currently determined spectral dimension in a descending order. For example, when traversing in a descending order of dimensions... If the value is too small, the target spectral dimension will also be missed. This means that the target spectral dimension is after the current spectral dimension. Therefore, we can traverse backward from the currently determined spectral dimension in the direction from low to high.

[0184] Furthermore, after missing the target spectral dimension, the dimensional range of the target spectral dimension can be determined, and then the iteration step size can be reduced. The value of is used to more accurately find the target spectral dimension within that range.

[0185] Next, this manual will continue to describe the application of the above-mentioned spectral dimensionality reduction method in the training process of spectral calculation models.

[0186] like Figure 5 As shown, Figure 5 This is a flowchart illustrating a training method for a spectral calculation model according to an exemplary embodiment, comprising the following steps:

[0187] Obtain the training dataset, wherein each training sample in the training dataset includes a set of structural parameters and a first spectral dimension for describing the grating structure. The first theoretical spectrum.

[0188] The first theoretical spectrum is subjected to dimensionality reduction and restoration to obtain The baseline restored spectrum of the dimensional restoration.

[0189] If the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition, then the... The dimension is the target spectral dimension after the first theoretical spectrum is reduced in dimension.

[0190] If not satisfied, the following steps are executed iteratively:

[0191] The iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing 3D restored matrix;

[0192] For the first iteration, based on the reference restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration;

[0193] For iterations other than the first round, the candidate restoration spectra determined in the previous round and those determined in this round are used as the basis for the current round. The 3D restoration matrix determines the candidate restored spectra for this round;

[0194] If the candidate restored spectrum in any round meets the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in that round is taken as the target spectral dimension; wherein... , and All are positive integers, and ;

[0195] The structural parameters are input into the spectral calculation model to output the predicted spectrum of the target spectral dimension.

[0196] The spectral calculation model is trained based on the loss value between the predicted spectrum and the second theoretical spectrum of the target spectral dimension; wherein the second theoretical spectrum is obtained by dimensionality reduction of the first theoretical spectrum.

[0197] In this embodiment, a target spectral dimension is selected before training the spectral calculation model. In finding a suitable target spectral dimension, this scheme, in each iteration, can use the candidate restored spectra obtained from the previous iteration, combined with the iteration step size. Restored The dimensional restoration matrix is ​​used to jointly determine ( Candidate restored spectra for k-dimensional restoration. That is, each round only requires dimensionality reduction and restoration calculations for the k-dimensional data components, without needing to perform dimensionality reduction and restoration calculations on the complete (k-dimensional) data. The dimensionality reduction and restoration calculations are performed on the 3D data components. Therefore, this scheme can reduce the computational load of dimensionality reduction and restoration calculations during training, thereby lowering the performance requirements of computer hardware.

[0198] Since the target spectral dimension of the predicted spectrum output by the spectral calculation model is smaller than the first spectral dimension of the first theoretical spectrum, the output of the spectral calculation model is the dimensionality-reduced spectral dimension.

[0199] When verifying the training effect of the spectral calculation model, in order to accurately evaluate the training effect of the spectral calculation model, it is necessary to restore the dimension of the predicted spectral values ​​after dimensionality reduction to the same dimension as the theoretical spectrum through restoration calculation before evaluating the error between the two.

[0200] For example, such as Figure 6As shown, assume that a set of training samples obtained from a theoretical spectral database includes a set of structural parameters and a corresponding first theoretical spectrum of a first spectral dimension. After inputting the structural parameters into the spectral computation model, the output is a predicted spectrum of a target spectral dimension, which is smaller than the first spectral dimension. To calculate the loss between the predicted spectrum and the theoretical spectrum, the first theoretical spectrum needs to be dimensionalized into a second theoretical spectrum with the same spectral dimension as the predicted spectrum. Then, the loss between the predicted spectrum and the second theoretical spectrum is calculated, and this loss is used to train the parameters of the spectral computation model. To accurately evaluate the training effect of the spectral computation model, existing solutions require restoring the current spectral dimension of the predicted spectrum to the first spectral dimension through reconstruction calculation, and using the error between the first theoretical spectrum and the predicted spectrum of the first spectral dimension as a correction evaluation index to measure the training effect of the model.

[0201] Therefore, before each evaluation of training effectiveness, it is necessary to perform restoration calculations on the predicted spectral values ​​to obtain the restored predicted spectrum, and then calculate the correction evaluation index between the restored predicted spectrum and the first theoretical spectrum. This repeated restoration calculation results in significant computational overhead and places high demands on computer hardware performance.

[0202] To address the aforementioned technical problems, in one embodiment, the loss value is multiplied by a correction factor to obtain a correction evaluation index; wherein the correction factor is the ratio of the target spectral dimension to the first spectral dimension. Training the spectral calculation model based on the loss value between the predicted spectrum and the second theoretical spectrum of the target spectral dimension includes: if the correction evaluation index does not meet preset performance conditions, then training the spectral calculation model based on the loss value; if it does meet the conditions, then stopping the training of the spectral calculation model.

[0203] For example, the first theoretical spectrum can be dimensionality reduced based on the Principal Component Analysis (PCA) algorithm to obtain the second theoretical spectrum.

[0204] The loss between the predicted spectrum and the second theoretical spectrum of the target spectral dimension can be measured by mean squared error (MSE).

[0205] The applicant discovered through mathematical derivation that, for example Figure 7 As shown, Figure 6 There is a certain numerical relationship between the correction evaluation index and the loss value. That is, the correction evaluation index can be obtained by multiplying the loss value by the correction factor, which is the ratio of the target spectral dimension to the first spectral dimension.

[0206] This scheme utilizes a correction factor derived from mathematical theory. Multiplying this correction factor by the loss value yields a correction evaluation index used to assess the training effectiveness of the spectral calculation model. The value of this correction evaluation index is equivalent to the error between the predicted spectrum and the theoretical spectrum after restoration calculation.

[0207] As can be seen, this scheme can obtain the correction evaluation index without recalculating the predicted spectral values ​​before each evaluation of the training effect, thus saving computational costs while accurately evaluating the training effect.

[0208] Next, we will introduce the mathematical derivation process:

[0209] Let the data matrix of the predicted spectrum in the first spectral dimension be . The data matrix of the first theoretical spectrum is ,in, For sample size, This is the first spectral dimension.

[0210] Let the data matrix of the predicted spectrum of the target spectral dimension be denoted as The data matrix of the second theoretical spectrum is ,in, The target spectral dimension.

[0211] The loss calculated based on MSE between the predicted spectrum of the target spectral dimension and the second theoretical spectrum is:

[0212]

[0213] The correction evaluation index based on MSE calculation between the predicted spectrum and the first theoretical spectrum in the first spectral dimension is:

[0214]

[0215] Comparing the two calculation formulas, the difference lies in calculating the ratio of the squared distance to the respective dimension. Without loss of generality, for a centered random vector... and ,have:

[0216]

[0217] It represents the statistical expectation of a random vector.

[0218] Assuming a centered random vector covariance matrix Has characteristic roots ,consider Linear combinations:

[0219]

[0220]

[0221]

[0222]

[0223] These are called principal components.

[0224] because If it is centralized, then Expectations Similarly,

[0225]

[0226] in .

[0227] By the principal component analysis theorem,

[0228]

[0229]

[0230] and

[0231]

[0232]

[0233] so

[0234]

[0235] Take the front that meets the restoration accuracy requirements One principal component, then

[0236]

[0237] Therefore, the loss value and the correction evaluation index have the following relationship:

[0238]

[0239] To verify the correctness of the above theoretical derivation, the applicant used multiple data sources, respectively... Figure 6 Existing solutions and Figure 7 The improved method was used to calculate the correction evaluation index, and the final results are shown in Table 2:

[0240] Table 2

[0241]

[0242] As shown in Table 2, "Dataset" is the name of the training dataset, "num" is the number of spectra, "dim" is the first spectral dimension, "after" is the target spectral dimension, "paranum" is the number of structural parameters, "eva time" is the time spent processing the predicted spectrum of the target spectral dimension into the predicted spectrum of the first spectral dimension through restoration calculation, "epoch time" is the time spent in one training epoch, "Time Ratio" is the proportion of time spent on restoration calculation in that training epoch, and "real mse" indicates the time spent on restoration calculation. Figure 7 The improved method calculates the value of the correction evaluation index, "eva mse" indicating that through Figure 6 The value of the corrected evaluation index is calculated using the method described above.

[0243] It is evident that this scheme saves the time spent on restoration calculations, and the accuracy of the calculated correction evaluation index is almost the same as that of the results calculated by existing schemes, proving the feasibility of the derivation results of this scheme.

[0244] Next, this instruction manual will continue to introduce... Figure 5 The application of the spectral calculation model trained by this method in the scenario of measuring the structural parameters of samples.

[0245] like Figure 8 As shown, Figure 8 This is a flowchart illustrating a method for measuring structural parameters according to an exemplary embodiment, including steps 801-804:

[0246] Step 801: Obtain all structural parameters generated from the structural model of the sample to be tested.

[0247] Step 802: Input the structural parameters into the following... Figure 5 The method trains a spectral calculation model to output the predicted spectrum corresponding to each structural parameter.

[0248] Step 803: Obtain the measurement spectrum obtained by measuring the sample to be tested.

[0249] Step 804: Match the measured spectrum with each predicted spectrum to select the target predicted spectrum that meets the matching conditions, and use the target structural parameter corresponding to the target predicted spectrum as the measured value of the actual structural parameter of the sample to be tested.

[0250] As mentioned earlier, different products under test correspond to different structural models. A structural model describes the specific types of structural parameters of the product under test. After determining the value range of all structural parameters in the structural model, structural parameters with different values ​​can be randomly generated.

[0251] Compared to the structural parameters and theoretical spectra stored in theoretical spectral databases, the spectral calculation model in this scheme can predict the predicted spectra corresponding to all structural parameters in a very short time. Therefore, in practical applications, it is not necessary to store massive amounts of structural parameters and theoretical spectra locally like a theoretical spectral database. Instead, when there is a measurement requirement, a large number of structural parameters can be temporarily sampled, and the corresponding predicted spectral values ​​can be calculated instantly based on the spectral calculation model.

[0252] like Figure 9 As shown, Figure 9 This is a flowchart illustrating a method for calculating a reconstructed spectrum according to an exemplary embodiment, including steps 901-903:

[0253] Step 901: For the first spectral dimension The spectrum to be processed is subjected to dimensionality reduction and restoration processing to obtain... The restored spectrum of the microstructure.

[0254] Step 902: Through ( The restored spectrum of the 3D restored dimension is located at the tail of the restored dimension. Wei Fuxing 3D restored matrix.

[0255] Step 903: Based on the above The restored spectrum of the dimensional restoration and the said Determining the dimensional restoration matrix ( The restored spectrum of 3D.

[0256] in, , and All are positive integers, and .

[0257] In this embodiment, the solution is an application of the aforementioned derivation formula. Assume that it is necessary to calculate ( The restored spectrum of the 3D model only requires calculation in this scheme. The dimensional restoration matrix, and combined with The restoration spectrum of the microstructure can then be obtained, reducing the need for further analysis. The computational cost of the data components in a dimension.

[0258] In one embodiment, the method further includes: iteratively determining candidate restoration spectra for candidate restoration dimensions; wherein the candidate restoration spectra are derived from the candidate restoration spectra determined in the previous round and the candidate restoration spectra in the current round, with the spectrum located at the tail of the restoration dimension. Restored The dimensional restoration matrix is ​​determined, the The value is the difference between the candidate restoration dimension of the previous round of candidate restored spectra and the candidate restoration dimension of the current round of candidate restored spectra; if the candidate restored spectrum of the candidate restoration dimension in any round meets the preset restoration accuracy condition, the candidate restoration dimension is taken as the target spectral dimension.

[0259] In this embodiment, the preset restoration accuracy condition may include that the restoration accuracy of the candidate restored spectrum in any round is not less than the preset restoration accuracy and that the restoration accuracy of the restored spectrum in the next lower dimension of the candidate restored spectrum is less than the preset restoration accuracy.

[0260] In one embodiment, the low-dimensional spectrum can be obtained by principal component analysis of the high-dimensional spectrum, or the low-dimensional spectrum can be obtained by a spectral calculation model.

[0261] like Figure 10 As shown, Figure 10 This is a flowchart illustrating another method for measuring structural parameters according to an exemplary embodiment of this specification, including steps 1001-1006:

[0262] Step 1001: Obtain all structural parameters generated from the structural model of the sample to be tested.

[0263] Step 1002: Input the structural parameters into the spectral calculation model to output the predicted spectrum corresponding to each structural parameter.

[0264] Step 1003: Obtain the measurement spectrum obtained by measuring the sample to be tested.

[0265] Step 1004: Determine the difference between the dimension of the measured spectrum and the dimension of the predicted spectrum. And determine the location of the measured spectrum at the tail of the dimension. Restored The dimension of the predicted spectrum is smaller than that of the measured spectrum.

[0266] Step 1005: Based on the above The dimension restoration matrix upscales each predicted spectrum to the same dimension as the measured spectrum.

[0267] Step 1006: Match the measured spectrum with each of the upscaled predicted spectra to select the target predicted spectra that meet the matching conditions, and use the target structural parameters corresponding to the target predicted spectra as the measured values ​​of the actual structural parameters of the sample to be tested.

[0268] In this embodiment, by utilizing the result of the formula derived above, only calculation is required. The dimension restoration matrix can be used to upgrade the measured spectrum to the same dimension as the measured spectrum, saving computational overhead and reducing the performance requirements of computer hardware.

[0269] Corresponding to the embodiments of the foregoing methods, this specification also provides embodiments of the apparatus and the terminal to which it is applied.

[0270] Figure 11 This is a schematic diagram illustrating the structure of an electronic device according to an exemplary embodiment. Figure 11 As shown, at the hardware level, the electronic device 1100 includes a processor 1102, an internal bus 1104, a network interface 1106, memory 1108, and non-volatile memory 1110, and may also include other hardware required for business operations. One or more embodiments of this specification can be implemented in software, for example, the processor 1102 reads the corresponding computer program from the non-volatile memory 1110 into the memory 1108 and then runs it. Of course, in addition to software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to each logic module, but can also be hardware or logic devices.

[0271] Figure 12 This is a block diagram illustrating a spectral dimensionality reduction device according to an exemplary embodiment of this specification. Figure 12 As shown, this device can be applied to, for example Figure 11 The electronic device 1100 shown implements the technical solution of this specification. The device includes:

[0272] The first reference restored spectrum determination module 1202 is used for determining the first spectral dimension. The spectrum to be processed is subjected to dimensionality reduction and restoration processing to obtain... The baseline restored spectrum of the dimensional restoration.

[0273] The first target spectral dimension determination module 1204 is used to determine the target spectral dimension if the restoration accuracy of the reference restored spectrum meets a first preset restoration accuracy condition. The dimension is taken as the target spectral dimension after dimensionality reduction of the spectrum to be processed; if this condition is not met, the following steps are performed iteratively: the iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing The dimensional restoration matrix; for the first iteration, based on the baseline restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration; for non-first iterations, based on the candidate restored spectra determined in the previous round and the current round... The dimensional restoration matrix determines the candidate restored spectra for this round; if the candidate restored spectra in any round satisfy the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectra in that round is taken as the target spectral dimension; wherein, , and All are positive integers, and .

[0274] Optionally, the first reference restored spectrum determination module 1202 is specifically used to, when traversing in a direction from low to high dimension, if the restoration accuracy of the reference restored spectrum is not less than a preset restoration accuracy and the restoration accuracy of the restored spectrum in the next lower dimension of the reference restored spectrum is less than the preset restoration accuracy, then... The first target spectral dimension determination module 1204 is specifically used to iteratively execute the following steps if the restoration accuracy of the reference restored spectrum is less than the preset restoration accuracy: determining the iteration step size in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing A 3D restoration matrix; for the first iteration, the reference restored spectrum and the... The sum of the dimensional restored matrices is used as ( Candidate restored spectra for 3D restoration; for iterations other than the first round, the candidate restored spectra determined in the previous round and the candidate restored spectra determined in the current round are combined. The sum of the dimensional restoration matrices is used as the candidate restored spectrum for this round; if the restoration accuracy of the candidate restored spectrum in any round is not less than the preset restoration accuracy and the restoration accuracy of the restored spectrum in the next lower dimension of the candidate restored spectrum is less than the preset restoration accuracy, the iteration stops, and the candidate restored dimension of the candidate restored spectrum in that round is used as the target spectral dimension; where, .

[0275] Optionally, the formula used to determine the reference restored spectrum is:

[0276]

[0277] in, The data matrix of the reference restored spectrum, For sample size, This is the first spectral dimension; This is the centered data matrix of the spectrum to be processed; For the spectrum to be processed Principal component loading matrix of dimension, for Transpose of; This is the mean data matrix of the spectrum to be processed.

[0278] Sure The formula used for the k-dimensional restoration matrix of the candidate restored spectrum is:

[0279]

[0280] in, Candidate restoration dimensions based on candidate restored spectra k-dimensional reconstruction of the tail The dimensional restoration matrix as 3D restored matrix; Candidate restoration dimension of the candidate restored spectrum Tail Principal component loading matrix composed of dimensional elements, for The transpose of .

[0281] The reference restored spectrum and based on The component restoration obtained by the above The sum of the dimensional restored matrices is based on ( The formula used to obtain the candidate restored spectra from the 1) dimensional component restoration is:

[0282]

[0283] in, For the basis of ( The data matrix of candidate restored spectra obtained by 3D component restoration.

[0284] Optionally, the first target spectral dimension determination module 1204 is specifically used to iteratively execute the following steps when traversing in a direction from high to low dimension: determining the iteration step size in each iteration. And by using the candidate restoration spectrum from the previous round at the tail of the candidate restoration dimension Wei Fuxing A 3D restoration matrix; for the first iteration, the reference restored spectrum and the... The difference between the dimensional restored matrices is used as ( Candidate restored spectra for 3D restoration; for iterations other than the first round, the candidate restored spectra determined in the previous round and the candidate restored spectra determined in the current round are combined. The difference between the two restored matrices is used as the candidate restored spectrum for this round. If the restoration accuracy of the candidate restored spectrum in any round is not less than the preset restoration accuracy, and the restoration accuracy of the restored spectrum in the next lower dimension of the candidate restored spectrum is less than the preset restoration accuracy, the iteration stops, and the candidate restored dimension of the candidate restored spectrum in that round is used as the target spectral dimension. .

[0285] Figure 13 This is a block diagram illustrating a training apparatus for a spectral calculation model according to an exemplary embodiment of this specification. Figure 13 As shown, this device can be applied to, for example Figure 11 The electronic device 1100 shown implements the technical solution of this specification. The device includes:

[0286] Training dataset acquisition module 1302 is used to acquire a training dataset, wherein each training sample in the training dataset includes a set of structural parameters describing the grating structure and a first spectral dimension. The first theoretical spectrum.

[0287] The second reference restored spectrum determination module 1304 is used to perform dimensionality reduction and restoration processing on the first theoretical spectrum to obtain... The baseline restored spectrum of the dimensional restoration.

[0288] The second target spectral dimension determination module 1306 is used to determine the target spectral dimension if the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition. The dimension is taken as the target spectral dimension after the dimensionality reduction of the first theoretical spectrum; if this condition is not met, the following steps are performed iteratively: the iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing The dimensional restoration matrix; for the first iteration, based on the baseline restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration; for non-first iterations, based on the candidate restored spectra determined in the previous round and the current round... The dimensional restoration matrix determines the candidate restored spectra for this round; if the candidate restored spectra in any round satisfy the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectra in that round is taken as the target spectral dimension; wherein, , and All are positive integers, and .

[0289] The predicted spectrum output module 1308 is used to input the structural parameters into the spectral calculation model to output the predicted spectrum of the target spectral dimension.

[0290] The model parameter training module 1310 is used to train the spectral calculation model based on the loss value between the predicted spectrum and the second theoretical spectrum of the target spectral dimension; wherein the second theoretical spectrum is obtained by dimensionality reduction of the first theoretical spectrum.

[0291] Figure 14 This is a block diagram illustrating a structural parameter measuring device according to an exemplary embodiment of this specification. Figure 14 As shown, this device can be applied to, for example Figure 11 The electronic device 1100 shown implements the technical solution of this specification. The device includes:

[0292] The structural parameter acquisition module 1402 is used to acquire all structural parameters generated from the structural model of the sample to be tested.

[0293] The predictive spectrum module 1404 is used to input the structural parameters into the spectral calculation model to output the predicted spectrum corresponding to each structural parameter.

[0294] The measurement spectrum acquisition module 1406 is used to acquire the measurement spectrum obtained by measuring the sample to be tested.

[0295] The sample structure measurement module 1408 is used to match the measured spectrum with each predicted spectrum to screen out the target predicted spectrum that meets the matching conditions, and use the target structural parameter corresponding to the target predicted spectrum as the measured value of the actual structural parameter of the sample to be tested.

[0296] The specific implementation process of the functions and roles of each module in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0297] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the solution in this specification according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0298] This specification also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the aforementioned methods provided in this application.

[0299] Specifically, computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks.

[0300] This specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of any of the aforementioned methods.

Claims

1. A spectral dimensionality reduction method, characterized in that, The method includes: For the spectrum to be processed in the first spectral dimension, dimensionality reduction and restoration processing are performed on the spectrum to obtain... The baseline restored spectrum for the first spectral dimension; the value of the first spectral dimension is... ; If the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition, then the... The dimension is the target spectral dimension after the dimensionality reduction of the spectrum to be processed. If not satisfied, the following steps are executed iteratively: The iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing 3D restored matrix; Dimension and iteration step size The values ​​are the same and all are ; For the first iteration, based on the reference restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration; For iterations other than the first round, the candidate restoration spectra determined in the previous round and those determined in this round are used as the basis for the current round. The 3D restoration matrix determines the candidate restored spectra for this round; If the candidate restored spectrum in any round meets the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in any round is taken as the target spectral dimension. in, , and All are positive integers, and .

2. The method according to claim 1, characterized in that, If the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition, then the... The dimension is used as the target spectral dimension; if the reference restored spectrum does not meet the first preset restoration accuracy condition, and the traversal is performed in the direction from low to high dimension, the following steps are executed iteratively: The iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing 3D restored matrix; For the first iteration, the reference restored spectrum and the... The sum of the dimensional restored matrices is used as ( Candidate restored spectra for 3D restoration; For iterations other than the first round, the candidate restored spectra determined in the previous round and the spectra determined in this round will be used together. The sum of the dimensional restoration matrices serves as the candidate restored spectra for this round; If the candidate restored spectrum in any round meets the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in any round is taken as the target spectral dimension. in, ; exist In the case of the first preset restoration accuracy condition and the second preset restoration accuracy condition, the restoration accuracy of the current round's restored spectrum is not less than the preset restoration accuracy, and the restoration accuracy of the lower dimension of the current round's restored spectrum is less than the preset restoration accuracy.

3. The method according to claim 2, characterized in that, The formula used to determine the reference restored spectrum is: in, The data matrix of the reference restored spectrum, For sample size, This is the first spectral dimension; This is the centered data matrix of the spectrum to be processed; For the spectrum to be processed Principal component loading matrix of dimension, for Transpose of; This is the mean data matrix of the spectrum to be processed; Sure Candidate restored spectra of the visceral restoration The formula used for restoring the 3D matrix is: in, Candidate restoration dimensions based on candidate restored spectra Tail The restoration obtained The dimensional restoration matrix will be the... The dimensional restoration matrix as 3D restored matrix; Candidate restoration dimension of the candidate restored spectrum Tail Principal component loading matrix composed of dimensional elements, for Transpose of; The reference restored spectrum and based on The component restoration obtained by the above The sum of the dimensional restored matrices is based on ( The formula used to obtain the candidate restored spectra from the 1) dimensional component restoration is: in, For the basis of ( The data matrix of candidate restored spectra obtained by 3D component restoration; The candidate restored spectrum obtained by dimensional component restoration is the spectrum to be processed in the first spectral dimension. After dimensionality reduction calculation, it becomes ( The spectrum of the first spectral dimension is obtained by performing a restoration calculation, and then the spectrum of the first spectral dimension is restored to its original value.

4. The method according to claim 1, characterized in that, When traversing from high to low dimension, the following steps are performed iteratively: The iteration step size is determined in each iteration. And by using the candidate restoration spectrum from the previous round at the tail of the candidate restoration dimension Wei Fuxing 3D restored matrix; For the first iteration, the reference restored spectrum and the... The difference between the dimensional restored matrices is used as ( Candidate restored spectra for 3D restoration; For iterations other than the first round, the candidate restored spectra determined in the previous round and the spectra determined in this round will be used together. The difference between the dimensional restoration matrices is used as the candidate restored spectra for this round; If the candidate restored spectrum in any round meets the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in any round is taken as the target spectral dimension. The second preset restoration accuracy condition is that the restoration accuracy of the candidate restored spectrum in any round is not less than the preset restoration accuracy and the restoration accuracy of the restored spectrum in the lower dimension of the candidate restored spectrum is less than the preset restoration accuracy. .

5. A method for calculating a restored spectrum, characterized in that, The method includes: For the spectrum to be processed in the first spectral dimension, dimensionality reduction and restoration processing are performed on the spectrum to obtain... The restored spectrum of the first spectral dimension; the value of the first spectral dimension is... ; pass( The restored spectrum of the 3D restored dimension is located at the tail of the restored dimension. Wei Fuxing 3D restored matrix; Based on the above The restored spectrum of the dimensional restoration and the said Determining the dimensional restoration matrix ( The restored spectrum of 3D model; in, , and All are positive integers, and .

6. The method according to claim 5, characterized in that, The method further includes: The candidate restored spectra for the candidate restored dimensions are determined iteratively; wherein, the candidate restored spectra are composed of the candidate restored spectra determined in the previous round and the candidate restored spectra obtained in the current round, which are located at the tail of the restored dimension. Wei Fuxing The dimensional restoration matrix is ​​determined, the The value is the difference between the candidate restoration dimension of the previous round of candidate restored spectra and the candidate restoration dimension of the current round of candidate restored spectra. If the candidate restored spectrum of any candidate restored dimension in any round meets the preset restoration accuracy condition, the candidate restored dimension is taken as the target spectral dimension; the preset restoration accuracy condition includes that the restoration accuracy of the candidate restored spectrum in any round is not less than the preset restoration accuracy and the restoration accuracy of the restored spectrum of the next lower dimension of the candidate restored spectrum is less than the preset restoration accuracy.

7. The method according to claim 5, characterized in that, Low-dimensional spectra can be obtained through principal component analysis of high-dimensional spectra, or low-dimensional spectra can be obtained through spectral calculation models.

8. A training method for a spectral calculation model, characterized in that, The method includes: Obtain a training dataset, wherein each training sample in the training dataset includes a set of structural parameters describing the grating structure and a first theoretical spectrum of a first spectral dimension; the value of the first spectral dimension is... ; The first theoretical spectrum is subjected to dimensionality reduction and restoration to obtain The baseline restoration spectrum for dimensional restoration; If the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition, then the... The dimension is the target spectral dimension after the first theoretical spectrum is reduced in dimensionality. If not satisfied, the following steps are executed iteratively: The iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing 3D restored matrix; Dimension and iteration step size The values ​​are the same and all are ; For the first iteration, based on the reference restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration; For iterations other than the first round, the candidate restoration spectra determined in the previous round and those determined in this round are used as the basis for the current round. The 3D restoration matrix determines the candidate restored spectra for this round; If the candidate restored spectrum in any round meets the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectrum in that round is taken as the target spectral dimension; wherein... , and All are positive integers, and ; The structural parameters are input into the spectral calculation model to output the predicted spectrum of the target spectral dimension; The spectral calculation model is trained based on the loss value between the predicted spectrum and the second theoretical spectrum of the target spectral dimension; wherein the second theoretical spectrum is obtained by dimensionality reduction of the first theoretical spectrum.

9. The method according to claim 8, characterized in that, The method further includes: The loss value is multiplied by the correction factor to obtain the correction evaluation index; wherein the correction factor is the ratio of the target spectral dimension to the first spectral dimension; Training the spectral calculation model based on the loss value between the predicted spectrum and the second theoretical spectrum of the target spectral dimension includes: If the correction evaluation index does not meet the preset performance conditions, then the spectral calculation model is trained based on the loss value; if it does meet the conditions, then the training of the spectral calculation model is stopped.

10. A method for measuring structural parameters, characterized in that, The method includes: Obtain all structural parameters generated from the structural model of the sample to be tested; The structural parameters are respectively input into the spectral calculation model trained as described in claim 8 to output the predicted spectrum corresponding to each structural parameter; Obtain the measurement spectrum obtained from the sample to be tested; The measured spectra are matched with each predicted spectrum to select target predicted spectra that meet the matching conditions, and the target structural parameters corresponding to the target predicted spectra are used as the measured values ​​of the actual structural parameters of the sample to be tested.

11. A method for measuring structural parameters, characterized in that, The method includes: Obtain all structural parameters generated from the structural model of the sample to be tested; The structural parameters are input into the spectral calculation model to output the predicted spectrum corresponding to each structural parameter. Obtain the measurement spectrum obtained from the sample to be tested; Determine the difference between the dimension of the measured spectrum and the dimension of the predicted spectrum. And through the measured spectrum located at the tail of the dimension Wei Fuxing The dimension of each predicted spectrum is smaller than the dimension of the measured spectrum; It is a positive integer; Based on the above The dimension-restored matrix upscals each predicted spectrum to the same dimension as the measured spectrum. The measured spectra are matched with the predicted spectra after dimensionality upgrades to select target predicted spectra that meet the matching conditions, and the target structural parameters corresponding to the target predicted spectra are used as the measured values ​​of the actual structural parameters of the sample to be tested.

12. A spectral dimensionality reduction device, characterized in that, The device includes: The first reference restored spectrum determination module is used to perform dimensionality reduction and restoration processing on the spectrum to be processed in the first spectral dimension to obtain... The baseline restored spectrum for the first spectral dimension; the value of the first spectral dimension is... ; The first target spectral dimension determination module is used to determine the target spectral dimension if the restoration accuracy of the reference restored spectrum meets a first preset restoration accuracy condition. The dimension is taken as the target spectral dimension after dimensionality reduction of the spectrum to be processed; if this condition is not met, the following steps are performed iteratively: the iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing The dimensional restoration matrix; for the first iteration, based on the baseline restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration; for non-first iterations, based on the candidate restored spectra determined in the previous round and the current round... The dimensional restoration matrix determines the candidate restored spectra for this round; if the candidate restored spectra in any round satisfy the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectra in that round is taken as the target spectral dimension; wherein, Dimension and iteration step size The values ​​are the same and all are ; , and All are positive integers, and .

13. A training device for a spectral calculation model, characterized in that, The device includes: The training dataset acquisition module is used to acquire a training dataset, wherein each training sample in the training dataset includes a set of structural parameters describing the grating structure and a first theoretical spectrum of the first spectral dimension; the value of the first spectral dimension is... ; The second reference restored spectrum determination module is used to perform dimensionality reduction and restoration processing on the first theoretical spectrum to obtain... The baseline restoration spectrum for dimensional restoration; The second target spectral dimension determination module is used to determine the target spectral dimension if the restoration accuracy of the reference restored spectrum meets the first preset restoration accuracy condition. The dimension is taken as the target spectral dimension after the dimensionality reduction of the first theoretical spectrum; if this condition is not met, the following steps are performed iteratively: the iteration step size is determined in each iteration. And by using the candidate restoration spectrum located at the tail of the candidate restoration dimension in each round Wei Fuxing The dimensional restoration matrix; for the first iteration, based on the baseline restored spectrum and the... Determining the dimensional restoration matrix ( Candidate restored spectra for 3D restoration; for non-first iterations, based on the candidate restored spectra determined in the previous round and the current round... The dimensional restoration matrix determines the candidate restored spectra for this round; if the candidate restored spectra in any round satisfy the second preset restoration accuracy condition, the iteration stops, and the candidate restoration dimension of the candidate restored spectra in that round is taken as the target spectral dimension; wherein, Dimension and iteration step size The values ​​are the same and all are ; , and All are positive integers, and ; A predicted spectrum output module is used to input the structural parameters into the spectral calculation model to output a predicted spectrum of the target spectral dimension; The model parameter training module is used to train the spectral calculation model based on the loss value between the predicted spectrum and the second theoretical spectrum of the target spectral dimension; wherein the second theoretical spectrum is obtained by dimensionality reduction of the first theoretical spectrum.

14. A device for measuring structural parameters, characterized in that, The device includes: The structural parameter acquisition module is used to acquire all structural parameters generated from the structural model of the sample under test. The predictive spectrum module is used to input the structural parameters into the spectral calculation model trained as described in claim 8 to output the predicted spectrum corresponding to each structural parameter. The measurement spectrum acquisition module is used to acquire the measurement spectrum obtained by measuring the sample to be tested; The sample structure measurement module is used to match the measured spectrum with each predicted spectrum to filter out the target predicted spectrum that meets the matching conditions, and use the target structural parameters corresponding to the target predicted spectrum as the measured values ​​of the actual structural parameters of the sample to be tested.

15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-11.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1-11.

Images