Methods and apparatus for improving the measurement accuracy and efficiency of complex transistor structures

Abstract

This invention discloses a method and apparatus for improving the measurement accuracy and efficiency of complex transistor structures. The method includes: calculating and simulating the relationship between the wide-band Mueller matrix spectrum and the near-field electric field of the transistor; determining a narrow band with high sensitivity based on the transistor's properties and the relationship between the wide-band Mueller matrix spectrum and the near-field electric field; acquiring the Mueller matrix spectrum data of the transistor in the narrow band with high sensitivity, and training the data to obtain a measurement model; and predicting the structure of the transistor under test based on the measurement model. This invention, by calculating a narrow band with high sensitivity and replacing the wide spectrum, is easier to implement. Furthermore, the spectrum of the narrow band with high sensitivity can save measurement time compared to using the wide spectrum in deep learning measurement processes, thereby improving production efficiency. In addition, this invention can reduce dispersion errors caused by optical components in wide-spectrum measurement systems, thereby improving the reliability and measurement accuracy of high-density integration.

Description

Technical Field

[0001] This invention relates to the field of key dimension measurement technology for nanostructured semiconductors, and in particular to a method and apparatus for improving the measurement accuracy and efficiency of complex transistor structures. Background Technology

[0002] In recent years, with the expansion of Moore's Law, numerous innovations in transistor structures have emerged, such as nanosheet and fork-plate field-effect transistors (FETs), gate-all-around (GAA) transistors, and complementary field-effect transistors (CFETs). The emergence of these novel and complex 3D transistor structures has presented significant challenges to online metrology and inspection. Only with substantial advances in metrology for the patterning process control of these structures can the transition from research to manufacturing be achieved. Mueller matrix spectral ellipsometry is a fast, accurate, and non-destructive online metrology technique and a potential metrology method for the mass production of ultra-high-density complex transistors. Furthermore, machine learning methods are widely applied to the online metrology of complex structures.

[0003] To increase the sample set for machine learning and improve measurement accuracy, the Mueller matrix spectral ellipsometric method selects a wider spectrum. However, a wider spectrum increases computational complexity, which reduces measurement efficiency. Furthermore, lenses in a wide-spectrum measurement optical path system can introduce significant aberrations due to dispersion, thus introducing measurement errors.

[0004] Using a higher sensitivity spectral range can improve measurement efficiency and reduce measurement errors caused by dispersion, making it necessary to establish a measurement method accordingly. Summary of the Invention

[0005] The purpose of this invention is to improve the metrological efficiency of complex transistor structures and reduce dispersion errors caused by wide spectrum in the measurement of key parameters of ultra-high density complex transistors, thereby improving the reliability of high-density integration.

[0006] To achieve the above objectives, the present invention provides a method for improving the measurement accuracy and efficiency of complex transistor structures, comprising,

[0007] The relationship between the wideband Mueller matrix spectrum and the near-field electric field of the transistor was calculated and simulated.

[0008] Based on the properties of the transistor and the relationship between the wide-band Mueller matrix spectrum and the near-field electric field, a narrow band with high sensitivity is determined.

[0009] Acquire the Mueller matrix spectral data of the transistor in a narrow band with high sensitivity, and train the data to obtain a measurement model;

[0010] The structure of the transistor under test is predicted based on the measurement model.

[0011] Furthermore, determining the narrow band with high sensitivity based on the relationship between the wide-band Mueller matrix spectrum and the near-field electric field, as well as the properties of the transistor, includes:

[0012] If the transistor is an isotropic structure, the narrow band with high sensitivity can be separated from the relationship between the wide-band Mueller matrix spectrum and the near-field electric field to generate surface plasmon polaritons and the near-field electric field above the threshold.

[0013] If the transistor has an anisotropic structure, the narrow band with high sensitivity can be separated from the relationship between the wide-band Mueller matrix spectrum and the near-field electric field to generate local surface plasmon polaritons and the near-field electric field above the threshold.

[0014] Furthermore, the statement that the near-electric field is above the threshold means that the intensity of the near-electric field after normalization is greater than 0.6.

[0015] Furthermore, the step of acquiring the transistor's narrow-band Mueller matrix spectral data with high sensitivity, and training the measurement model using the data, includes:

[0016] If the transistor has an isotropic structure, M is selected in the narrow-band Mueller matrix spectrum with high sensitivity. 12 M 22 M 33 M 34 Elements are trained;

[0017] If the transistor has an anisotropic structure, select the transistor other than M in the narrow band Mueller matrix spectrum with high sensitivity. 11 Training is performed using elements other than those mentioned above.

[0018] Furthermore, when acquiring data for anisotropic transistors, the incident light azimuth angle should be measured to avoid being perpendicular or parallel to the transistor structure.

[0019] Furthermore, a rigorous coupled-wave analysis algorithm was used to calculate and simulate the relationship between the wideband Mueller matrix spectrum and the near-field electric field of the transistor.

[0020] This invention also provides an apparatus for improving the measurement accuracy and efficiency of complex transistor structures, comprising,

[0021] The calculation module is used to calculate and simulate the relationship between the wideband Mueller matrix spectrum and the near-field electric field of the transistor.

[0022] The determination module is used to determine a narrow band with high sensitivity based on the properties of the transistor and the relationship between the wide-band Mueller matrix spectrum and the near-field electric field.

[0023] The training module is used to acquire the Mueller matrix spectral data of the transistor in a narrow band with high sensitivity, and to train the measurement model using the data.

[0024] The prediction module is used to predict the structure of the transistor under test based on the measurement model.

[0025] The present invention also provides a measurement system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the method described above.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] This invention calculates a highly sensitive narrow-band spectrum and replaces the broad spectrum, making it easier to implement. Furthermore, the highly sensitive narrow-band spectrum saves measurement time compared to using a broad spectrum during deep learning measurements, thus improving production efficiency. In addition, this invention reduces dispersion errors caused by optical components in broad-spectrum measurement systems, thereby improving the reliability and measurement accuracy of high-density integration. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A flowchart of the method for improving the measurement accuracy and efficiency of complex transistor structures according to the present invention is shown;

[0030] Figure 2 The diagram shows the structural schematics of isotropic and anisotropic transistors analyzed in this invention.

[0031] Figure 3 A schematic diagram is shown showing the calculation of the wideband Mueller matrix spectrum and near-field electric field of a transistor using the RCWA algorithm;

[0032] Figure 4 The graph showing the relationship between the wideband Mueller matrix spectrum and the near-field electric field of the isotropic transistor of the present invention is illustrated.

[0033] Figure 5 The results of cross-entropy loss and verification accuracy in different wavelength ranges are shown in the thickness training network of the isotropic structure transistor SiGe of the present invention.

[0034] Figure 6 A scatter plot of the thickness distribution of SiGe according to the present invention is shown;

[0035] Figure 7The graph showing the relationship between the wideband Mueller matrix spectrum and the near-field electric field of the anisotropic structure transistor of the present invention is illustrated.

[0036] Figure 8 The results of cross-entropy loss and verification accuracy in different wavelength ranges in the training network for etching thickness of the anisotropic structure transistor of the present invention are shown.

[0037] Figure 9 A scatter plot of the etching thickness distribution of the present invention is shown;

[0038] Figure 10 A schematic diagram of the device for improving the measurement accuracy and efficiency of complex transistor structures according to the present invention is shown. Detailed Implementation

[0039] The method for improving the measurement accuracy and efficiency of ultra-high density complex transistor structures involved in this invention is unique in that it calculates the near-field electric field of the complex transistor structure using near-field optics theory, and separates the wavelength band with higher near-field electric field for Mueller matrix spectroscopy measurement. Specifically, Mueller matrix spectroscopy measurement can obtain the spectra of 16 Mueller matrix elements of the sample under test.

[0040] Muller matrix spectrum

[0041] The near-field electric field is mainly the electric field at the critical dimensions of nanoscale transistors.

[0042] The technical solutions of the present invention will be clearly and completely described below with reference to specific embodiments and accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example 1

[0044] like Figure 1 As shown, a method for improving the measurement accuracy and efficiency of complex transistor structures includes the following steps:

[0045] S1. Classify transistors: Distinguish between isotropic and anisotropic (i.e., non-isotropic) transistors, and use the rigorous coupled-wave analysis (RCWA) algorithm to simulate the relationship between the wide-band Mueller matrix spectrum and the near-field electric field of the transistor.

[0046] S2. Determine the narrow band with high sensitivity: If the transistor has an isotropic structure, the band that generates surface plasmons and whose near-field electric field is higher than the threshold is the narrow band with high sensitivity, separated from the relationship between the wide-band Mueller matrix spectrum and the near-field electric field; if the transistor has an anisotropic structure, the band that generates local surface plasmons and whose near-field electric field is higher than the threshold is the narrow band with high sensitivity, separated from the relationship between the wide-band Mueller matrix spectrum and the near-field electric field; the near-field electric field higher than the threshold means that the intensity of the near-field electric field after normalization is greater than 0.6.

[0047] S3. Set up the optical path system for measuring the specified band and acquire data: Adjust the measuring instrument according to the narrow band with high sensitivity determined in step S2. Accordingly, acquire the Mueller matrix spectral data of the transistor in the narrow band with high sensitivity. When acquiring the data of anisotropic transistors, avoid measuring the incident light azimuth angle perpendicular or parallel to the transistor structure to increase the numerical difference of the Mueller matrix sub-diagonal and increase the spectral difference characteristics between structures.

[0048] S4. Measurement Model Training: If the transistor has an isotropic structure, select M in the narrow-band Mueller matrix spectrum with high sensitivity. 12 M 22 M 33 M 34 Elements are trained; if the transistor has an anisotropic structure, select elements other than M from the narrow-band Mueller matrix spectrum with high sensitivity. 11 The model is trained using 15 elements other than those used in the deep learning training to obtain a measurement model.

[0049] S5. Structure Prediction: Based on the measurement model trained in S4, predict the structure of the transistor under test and output the prediction results.

[0050] To better understand the technical solution of the present invention, the solution of the present invention will be further explained in conjunction with a specific transistor.

[0051] like Figure 2 As shown in (a), the isotropic transistor is a multilayer superlattice Si / SiGe grown pair transistor; the thickness of the SiGe layer needs to be controlled, as the final thickness directly determines the height of the nanosheet channel and its electrostatic properties. Figure 2 As shown in (b), the anisotropic transistor is a partially recessed transistor with a sacrificial SiGe layer, including SiO2 to separate adjacent transistors, with a transistor spacing of 30 nm. The gate (Poly-Si) period is 40 nm, a 10 nm SiO2 layer is attached to the gate surface as an isolation layer (Low k), and a 30 nm thick Si3N4 layer is attached above the gate as a hard mask. Figure 3A schematic diagram is shown showing the calculation of the wideband Mueller matrix spectrum and near-field electric field of a transistor using the RCWA algorithm.

[0052] Figure 4 The graph shows the relationship between the broadband Mueller matrix spectrum and the near-field electric field of the corresponding isotropic transistor, where the near-field electric field has been normalized using conventional methods, and the SiGe thickness step size is 2 nm. Figure 4 As can be seen, when the near-field electric field strength is above the threshold of 0.6, the near-field electric field (E) in the wavelength ranges of λ1 (0.68μm~0.83μm) and λ2 (0.96μm~1.175μm) is... Near-field The presence of a distinct peak indicates surface plasmon resonance at this wavelength. The thickness of SiGe exhibits significant differences within the λ1 and λ2 wavelength ranges. Therefore, λ1 and λ2 are considered the bands with higher sensitivity for measuring Si / SiGe. The Mueller matrix spectra in the λ1 and λ2 wavelength ranges were used for machine learning training to measure the thickness of Si / SiGe, and the results were compared with those from the full wavelength range λ3 (0.2–1.5 μm). The thickness of SiGe ranged from 10–20 nm (with a step size of 2 nm). Spectral data with 1000 noise samples generated for each structure were used as the dataset for the artificial neural network (ANN).

[0053] Figure 5 (a) and Figure 5 (b) The results of cross-entropy loss and validation accuracy in different wavelength ranges of the training network for isotropic SiGe transistors are shown. It can be seen that as the number of epochs increases, the training convergence speed of λ3 is 2–7 epochs earlier than the other two bands. This is because the amount of data for λ3 is greater than that for λ1 and λ2. This means that for the same number of epochs, λ3 takes more time than λ1 and λ2. The thickness of SiGe is analyzed using t-distributed random neighborhood embedding (t-SNE), as shown in... Figure 6 As shown, different scatter points represent different SiGe thicknesses, with each point representing simulated data with 20% noise, randomly generated from the Mueller matrix spectrum of the λ1 band. The three scatter points are clearly dispersed, indicating that the SiGe thickness can be accurately distinguished. The thickness classification accuracy for λ1, λ2, and λ3 is 100%.

[0054] Figure 7 The graph shows the relationship between the broadband Mueller matrix spectrum and the near-field electric field of the corresponding anisotropic transistor, where the near-field electric field has been normalized using conventional methods, and the SiGe etching step size is 2 nm. Figure 7As can be seen, there are obvious peaks at wavelengths of 0.55 μm and 1.03 μm, indicating the generation of localized surface plasmon resonances (LPRs) in the anisotropic GAA nanowires. However, the broadening at the peak (λ = 0.55 μm) is relatively small, resulting in insufficient data for ANN. Therefore, the wavelength range of λ4 is expanded to 0.46–0.58 μm, where over 50% of the wavelength range has normalized E... Near-field Greater than 0.6. To increase the difference in the sub-diagonal of the Mueller matrix, the azimuth angle was measured at 45° relative to the gate.

[0055] Figure 8 (a) and Figure 8 (b) shows the cross-entropy loss and verification accuracy results for different wavelength ranges in the training network for anisotropic transistors, with two E values. Near-field The wavebands near the peak (λ4 and λ5) and the entire waveband (λ3) both converged rapidly, with 100% accuracy in both tests. The etching thickness results are as follows: Figure 9 As shown, this demonstrates that the etching thickness can be accurately distinguished.

[0056] The training times of the isotropic and anisotropic transistors analyzed above are summarized in Table 1.

[0057] Table 1

[0058]

[0059]

[0060] As can be seen from Table 1, the training time for both isotropic and anisotropic transistor ANNs is reduced by approximately 60% compared to the training time for broadband.

[0061] Example 2

[0062] like Figure 10 As shown, a device for improving the measurement accuracy and efficiency of complex transistor structures includes,

[0063] The calculation module is used to calculate and simulate the relationship between the wideband Mueller matrix spectrum and the near-field electric field of the transistor.

[0064] The determination module is used to determine a narrow band with high sensitivity based on the properties of the transistor and the relationship between the wide-band Mueller matrix spectrum and the near-field electric field.

[0065] The training module is used to acquire the Mueller matrix spectral data of the transistor in a narrow band with high sensitivity, and to train the measurement model using the data.

[0066] The prediction module is used to predict the structure of the transistor under test based on the measurement model.

[0067] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for improving the measurement accuracy and efficiency of complex transistor structures, characterized in that, include, The relationship between the wideband Mueller matrix spectrum and the near-field electric field of the transistor was calculated and simulated. Based on the properties of the transistor and the relationship between the wide-band Mueller matrix spectrum and the near-field electric field, a narrow band with high sensitivity is determined. Acquire the Mueller matrix spectral data of the transistor in a narrow band with high sensitivity, and train the data to obtain a measurement model; The structure of the transistor under test is predicted based on the measurement model.

2. The method for improving the measurement accuracy and efficiency of complex transistor structures according to claim 1, characterized in that, The method for determining high-sensitivity narrow-band frequencies based on the relationship between the wide-band Mueller matrix spectrum and the near-field electric field, as well as the properties of the transistor, includes: If the transistor is an isotropic structure, the narrow band with high sensitivity can be separated from the relationship between the wide-band Mueller matrix spectrum and the near-field electric field to generate surface plasmon polaritons and the near-field electric field above the threshold. If the transistor has an anisotropic structure, the narrow band with high sensitivity can be separated from the relationship between the wide-band Mueller matrix spectrum and the near-field electric field to generate local surface plasmon polaritons and the near-field electric field above the threshold.

3. The method for improving the measurement accuracy and efficiency of complex transistor structures according to claim 2, characterized in that, The near-field electric field being higher than the threshold refers to the near-field electric field having a strength greater than 0.6 after normalization.

4. The method for improving the measurement accuracy and efficiency of complex transistor structures according to claim 3, characterized in that, The process of acquiring the transistor's narrow-band Mueller matrix spectral data for high sensitivity, and training the measurement model using that data, includes... If the transistor has an isotropic structure, M is selected in the narrow-band Mueller matrix spectrum with high sensitivity. 12 M 22 M 33 M 34 Elements are trained; If the transistor has an anisotropic structure, select the transistor other than M in the narrow band Mueller matrix spectrum with high sensitivity. 11 Training is performed using elements other than those mentioned above.

5. The method for improving the measurement accuracy and efficiency of complex transistor structures according to claim 4, characterized in that, When acquiring data for anisotropic transistors, the incident light azimuth angle should be measured to avoid being perpendicular or parallel to the transistor structure.

6. The method for improving the measurement accuracy and efficiency of complex transistor structures according to claim 1, characterized in that, The relationship between the wideband Mueller matrix spectrum and the near-field electric field of the transistor was calculated and simulated using a rigorous coupled-wave analysis algorithm.

7. A device for improving the measurement accuracy and efficiency of complex transistor structures, characterized in that, include, The calculation module is used to calculate and simulate the relationship between the wideband Mueller matrix spectrum and the near-field electric field of the transistor. The determination module is used to determine a narrow band with high sensitivity based on the properties of the transistor and the relationship between the wide-band Mueller matrix spectrum and the near-field electric field. The training module is used to acquire the Mueller matrix spectral data of the transistor in a narrow band with high sensitivity, and to train the measurement model using the data. The prediction module is used to predict the structure of the transistor under test based on the measurement model.

8. A measurement system, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the method according to any one of claims 1 to 6.

Images