Waveform aberration detection method for EUV lithography projection lens based on principal component analysis of aerial images

Abstract

Extreme ultraviolet (EUV) lithography projection objective wave aberration detection method based on principal component analysis of aerial images. The present invention adopts isolated spaces in the two directions of 0° and 90° as detection marks, and obtains the principal component coefficients and wave aberrations of the measured aerial image by principal component fitting. The method first performs principal component analysis and linear regression on the simulated aerial image set to obtain the principal components and regression matrix, and then fits the measured aerial image with the principal components to obtain the corresponding principal component coefficients. According to the regression matrix and the principal component coefficients, the least squares method is used to fit and solve the wave aberration. The present invention is the first to extend the lithography projection objective wave aberration detection method based on principal component analysis of aerial images (AMAI‑PCA) to EUV lithography, optimize the mask marks, reduce the influence of the three-dimensional mask effect on aberration detection, and has the characteristics of fast process and high aberration detection accuracy.

Description

Technical Field

[0001] The invention relates to a wave aberration detection technology for a lithography projection objective lens, and in particular to a wave aberration detection method for an extreme ultraviolet lithography projection objective lens based on principal component analysis of an aerial image. Background Art

[0002] Photolithography is a key technology for the manufacture of very large-scale integrated circuits. The resolution of photolithography determines the characteristic size of integrated circuits. The wave aberration of the projection lens will cause the contrast of photolithography imaging to decrease, seriously affecting the quality of photolithography imaging. In order to ensure the imaging quality of the photolithography machine, high-precision detection of image quality parameters is required during the manufacturing of projection lenses, the whole machine integration calibration of the photolithography machine, exposure, and periodic repair and maintenance.

[0003] In deep ultraviolet (DUV) lithography, the existing wave aberration detection technologies mainly include photoresist pattern-based wave aberration detection technology, pupil-based wave aberration detection technology, and aerial image-based wave aberration detection technology. Among them, the aerial image-based wave aberration detection technology only needs to use mask marks and aerial image sensors to realize in-situ detection of wave aberration, which has the advantages of low cost and easy operation.

[0004] Among the many wave aberration detection technologies based on aerial images, TAMIS technology is a representative one (see prior art 1, H. van der Laan, M. Dierichs, H. van Greevenbroek, E. McCoo, F. Stoffels, R. Pongers and R. Willekers, "Aerial image measurement methods for fastaberration set-up and illumination pupil verification", Proc. SPIE 4346, 394-407 (2001)). TAMIS technology uses the imaging position offset and the best focal plane offset of the detection mark aerial image to extract the wave aberration, but it requires multiple illumination modes and can only measure some low-order aberrations, which has certain limitations.

[0005] The AMAI-PCA technology is another wave aberration detection method based on spatial images. It has the characteristics of simple process, fast speed and high aberration solution accuracy. In DUV lithography, by establishing a detection model based on vector lithography imaging theory and optimizing the detection mark, image space sampling method and illumination method, the application scope of the existing AMAI-PCA technology has been expanded from dry DUV lithography to immersion DUV lithography, which can achieve Z5~Z 6460 Zernike coefficients are detected, and the detection accuracy of a single Zernike coefficient is better than 1.03mλ (see prior art 2, Wang Xiangchao, Dai Fengzhao, et al., "Photolithography Image Quality Detection Technology. Volume 1", Beijing: Science Press, 2021.3, 496-528).

[0006] In extreme ultraviolet (EUV) lithography, existing wave aberration detection technologies include detection technology based on interferometry, detection technology based on Hartmann wavefront sensor, and detection technology based on spatial image measurement. Among them, the detection technology based on interferometry mainly includes point diffraction interferometry detection and shearing interferometry detection. Point diffraction interferometry detection technology has very high measurement accuracy, but its measurement range is very small, its measurement adaptability to high NA optical systems is poor, and it has high requirements for the spatial coherence of the light source. Lateral shearing interferometry detection technology has a large measurement range and does not require additional reference light, and the requirements for the spatial coherence of the light source are significantly reduced, but because of its low tolerance for grating position and grating tilt, it is also difficult to apply to large numerical aperture optical systems. Compared with the interferometric measurement technology, the detection technology based on the Hartmann wavefront sensor has the advantages of simple structure, low requirements on the spatial and temporal coherence of the light source, and a large measurement range. However, its disadvantages are also obvious, such as low spatial resolution, unadjustable measurement sensitivity, and difficulty in applying to high NA extreme ultraviolet lithography projection objective wave aberration detection (see prior art 3, Wang Xiangchao, Dai Fengzhao, et al., "Photolithography Image Quality Detection Technology. Volume 2", Beijing: Science Press, 2021.3, 174-277). In order to detect the wave aberration of large numerical aperture extreme ultraviolet lithography projection objectives, it is of great significance to develop a wave aberration detection technology based on spatial image measurement. Summary of the invention

[0007] The purpose of the present invention is to expand the AMAI-PCA technology to extreme ultraviolet lithography and maintain its high-precision characteristics of aberration detection. The shadow effect of the EUV mask causes the width of the spatial image of the 90° direction detection mark to be narrowed to a certain extent, while the detection mark in the 0° direction is not affected. This narrowing will cause the spatial image widths of the two detection marks to be different, thereby affecting the aberration detection accuracy of the AMAI-PCA technology. The present invention optimizes the 90° direction detection mark so that the widths of the two detection mark spatial images match, thereby expanding the AMAI-PCA technology to EUV lithography and retaining its high-precision characteristics.

[0008] The technical solution of the present invention is as follows:

[0009] The method for detecting wavefront aberration of an extreme ultraviolet lithography projection objective lens based on principal component analysis of an aerial image comprises the following steps:

[0010] Step 1. Correction of mask marks:

[0011] ①Set simulation conditions:

[0012] Set the lighting conditions, the partial coherence factor of traditional lighting is σ, and the partial coherence factor of annular lighting is [σ in ,σ out ],σ in represents the internal coherence factor, σ out represents the external coherence factor;

[0013] Set the numerical aperture NA of the projection objective lens, the mask period p, the mask marks are isolated spaces in the 0 degree direction and the 90 degree direction, and the width of the mask mark in the 0 degree direction is w0, and the width of the mask mark in the 90 degree direction is w1, w0=w1;

[0014] Set the vertical axis acquisition length x of the spatial image, the vertical axis sampling interval dx, the acquisition range is symmetrical with the center of the workpiece stage, the axial direction acquisition length f of the spatial image, the axial sampling interval df, and the acquisition range is symmetrical with the axial center;

[0015] ② Correction mask mark:

[0016] Without considering the influence of photoresist, the mask imaging graphic size (CD) is usually obtained by a constant aerial image threshold. Due to the existence of the shadow effect, when the aerial image threshold is a constant value, the mask imaging graphic size CD will change with the direction of the incident light. The detection mark in the 0° direction is in the same direction as the incident light and will not be affected by the shadow effect. The detection mark in the 90° direction is perpendicular to the incident light direction. Due to the influence of the shadow effect, the mask imaging graphic size CD of the detection mark in this direction will be reduced. In order to obtain the same mask imaging graphic size CD, the mask mark can be corrected by the following steps:

[0017] Simulate a 0-degree spatial image without aberration, denoted as AI0;

[0018] Simulate a 90-degree spatial image without aberration, denoted as AI1;

[0019] Set the threshold T of the mask imaging pattern size CD;

[0020] The threshold T is used to determine the mask imaging pattern size CD0 in the 0 degree direction and the mask imaging pattern size CD1 in the 90 degree direction for the aerial image AI0 in the 0 degree direction and the aerial image AI1 in the 90 degree direction respectively;

[0021] The difference between the mask imaging pattern size CD0 in the 0 degree direction and the mask imaging pattern size CD1 in the 90 degree direction is recorded as the compensation amount ΔCD, then the corresponding mask mark width compensation amount Δw=4*ΔCD, and the width of the mask mark in the 90 degree direction w1=w0+Δw;

[0022] Step 2. Acquisition of spatial image set:

[0023] ① Set the projection objective to solve the six wave aberrations, namely Z7~Z9 and Z 14 ~Z 16 , and the amplitude range of each wave aberration is [-a,a];

[0024] ② Constructing matrix D: For the six wave aberrations, Box_Behnken Design statistical sampling is used to generate 54 combinations as 54 rows of matrix D, and the six wave aberrations are used as 6 columns of matrix D;

[0025] Calculate the Zernike coefficient combination A, the formula is as follows:

[0026] A=a·D;

[0027] ③ The mask marks in the 0-degree direction and the 90-degree direction are simulated respectively. The 54 spatial images in the 0-degree direction constitute the spatial image set IM0 in the 0-degree direction, and the 54 spatial images in the 90-degree direction constitute the spatial image set IM1 in the 90-degree direction;

[0028] Step 3. Generation of principal component matrix and regression matrix:

[0029] ① Perform principal component analysis on the spatial image set IM0 in the 0-degree direction to obtain the principal component matrix PC0 and principal component coefficient matrix C0 corresponding to the spatial image in the 0-degree direction;

[0030] Perform principal component analysis on the aerial image set IM1 at 90 degrees to obtain the principal component matrix PC1 and principal component coefficient matrix C1 corresponding to the aerial image at 90 degrees;

[0031] ② Perform linear regression analysis on the principal component coefficient matrix C0 and the Zernike coefficient combination A to obtain the regression matrix RM0 in the 0 degree direction;

[0032] Perform linear regression analysis on the principal component coefficient matrix C1 and the Zernike coefficient combination A to obtain the regression matrix RM1 in the 90-degree direction;

[0033] Step 4. Collection of aerial images:

[0034] Set the parameters of the projection objective lens of the lithography machine to be inspected. The parameters of illumination, numerical aperture, mask period and spatial image scanning range are set the same as in step 1. The mask mark adopts the corrected width.

[0035] The photolithography machine is started, and the illumination light emitted by the light source is modulated by the illumination system to obtain the corresponding illumination mode, and then illuminates the test mask on the mask stage. The workpiece stage drives the spatial image sensor to complete the acquisition of the actual spatial image, that is, the spatial image to be tested in the 0-degree direction and the spatial image to be tested in the 90-degree direction;

[0036] Step 5. Fitting and solving of the spatial image:

[0037] ① Fit the principal component matrix PC0 to obtain the principal component coefficient matrix V0 of the spatial image to be measured in the 0 degree direction. The formula is as follows:

[0038] V0=PC0 / AI hor

[0039] In the formula, AI hor is the column vector of each pixel point in the measured spatial image in the 0 degree direction arranged in sequence;

[0040] The principal component matrix PC1 is fitted to obtain the principal component coefficient matrix V1 of the aerial image to be measured at 90 degrees, and the formula is as follows:

[0041] V1=PC1 / AI ver

[0042] In the formula, AI ver is a column vector of pixels in the spatial image to be measured at 90 degrees arranged in sequence;

[0043] ②According to the regression matrix RM0 and RM1, the total regression matrix RM is obtained, and the formula is as follows:

[0044]

[0045] According to the principal component coefficient matrices V0 and V1, the total principal component coefficient matrix V is obtained by combining them. The formula is as follows:

[0046]

[0047] The least squares fitting of the regression matrix RM and the principal component coefficient matrix V is used to obtain the wave aberration Z of the lithography machine projection lens, and the formula is as follows:

[0048] Z = RM\V.

[0049] Among them, the numerical aperture NA ranges from 0.33 to 0.55.

[0050] Among them, the lighting mode is various lighting modes supported by the lithography machine, including traditional lighting and ring lighting.

[0051] Among them, when the lighting method is traditional lighting, the partial coherence factor σ ranges from 0.1 to 0.9.

[0052] Among them, when the lighting mode is ring lighting, the internal coherence factor σ in The range is 0.1~0.8, and the external coherence factor σ out The range is 0.2~0.9, where σ out -σin Not less than 0.1.

[0053] The acquisition length x in the vertical axis direction is any value between 100 nm and 3000 nm; and the acquisition step length dx in the vertical axis direction is any value between 0.5 nm and 100 nm.

[0054] The collection length f in the axial direction is any value between 1000 nm and 8000 nm; and the collection step length df in the axial direction is any value between 1 nm and 125 nm.

[0055] The width w0 of the mask mark in the 0 degree direction is any value between 50 nm and 400 nm; and the mask period p is any value between 500 nm and 5000 nm.

[0056] Compared with the prior art, the present invention has the following advantages:

[0057] 1. Expand the application scope of the traditional principal component analysis-based lithography projection objective wave aberration detection technology to EUV lithography.

[0058] 2. By optimizing the mask markings, the interference of the EUV mask shadow effect on aberration detection is reduced, and the accuracy of wave aberration solution in EUV lithography is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Technical flow chart of the present invention.

[0060] Figure 2 Schematic diagram of the lighting method adopted by the present invention.

[0061] Figure 3 Schematic diagram of the mask mark used in the present invention, wherein (a) is the mask mark in the 0 degree direction, and (b) is the mask mark in the 90 degree direction.

[0062] Figure 4 When the mask mark is not optimized, the solution results of the six Zernike aberrations (30 groups of tested spatial images with random aberrations are used, and the solution accuracy is expressed as the average value + 3*standard deviation, the same below).

[0063] Figure 5 The technical solution of the present invention is adopted to optimize the 90° direction detection mark, and the accuracy of Z9 changes.

[0064] Figure 6 The technical solution of the present invention is used to optimize the 90° direction detection mark, Z 16 The accuracy changes. DETAILED DESCRIPTION

[0065] The present invention will be further described below in conjunction with embodiments and drawings, but the protection scope of the present invention should not be limited by these embodiments.

[0066] The lighting method is as follows Figure 2 As shown, the conventional lighting 1 on the left and the annular lighting 2 on the right are included. This example uses conventional lighting, where the partial coherence factor σ = 0.65.

[0067] The mask markings are as follows Figure 3 As shown, it includes a detection mark 3 in the 0 degree direction for detecting coma and spherical aberration in the 0 degree direction, and a detection mark 4 in the 90 degree direction for detecting coma and spherical aberration in the 90 degree direction.

[0068] The specific measurement includes the following five steps:

[0069] Step 1. Correction of mask marks:

[0070] First, the imaging formula of spatial image is introduced.

[0071] The spatial image of the lithography machine projection lens can be given by the following lithography imaging formula:

[0072]

[0073] According to this formula, the spatial image intensity distribution under certain lighting and mask conditions can be calculated. Among them, I represents the light intensity in the image space. x, y are the coordinates in the image space, and z is the defocus. J(f, g) represents the distribution of the light source, and (f, g) is the coordinate of the point on the light source. O(f′, g′) represents the Fourier spectrum of the mask, O * (f″,g″) represents the conjugate of the mask Fourier spectrum. k′ zp and k" zp represents the propagation vector in the z direction. According to the angular spectrum theory, the propagation vector in the z direction can be obtained as:

[0074]

[0075] H is the pupil function, expressed as:

[0076]

[0077] in, represents the effect of defocus, φ(f,g) represents the wavefront aberration, and the wavefront aberration can be expressed as a combination of various aberrations using the Zernike coefficients:

[0078]

[0079] Among them, (ρ, θ) is the normalized polar coordinate of the objective lens exit pupil surface, Z7 and Z 14Respectively represent the third and fifth order coma in the 0 degree direction, Z8 and Z 15 Represents the third and fifth order coma in the 90 degree direction, Z9 and Z 16 represent the third-order and fifth-order spherical aberration respectively.

[0080] ①Set simulation conditions:

[0081] The lighting condition is set to conventional lighting, where the partial coherence factor σ=0.65.

[0082] The numerical aperture of the projection objective is set to NA = 0.33.

[0083] The mask period p is set to 1600nm, the mask marks are isolated spaces in the 0-degree and 90-degree directions, the width of the mask mark in the 0-degree direction is w0=88nm, and the width of the mask mark in the 90-degree direction is w1=88nm.

[0084] The acquisition length of the spatial image in the vertical axis direction is set to x=300nm, the sampling interval is set to dx=3nm, and the acquisition range is symmetrical with the center of the workpiece stage; the acquisition length of the spatial image in the axial direction is set to f=2800nm, the sampling interval is set to df=40nm, and the acquisition range is symmetrical with the axial center.

[0085] ② Correction mask mark:

[0086] According to the above conditions, simulate an aberration-free 0-degree aerial image, recorded as AI0; simulate an aberration-free 90-degree aerial image, recorded as AI1; set the threshold T of the mask imaging pattern size CD to 0.25; use the threshold T to determine the mask imaging pattern size CD0 in the 0-degree direction and the mask imaging pattern size CD1 in the 90-degree direction for the aerial image AI0 in the 0-degree direction and the aerial image AI1 in the 90-degree direction, respectively. The specific steps are as follows:

[0087] Take out a row of light intensity data corresponding to the best focal position of the two aerial images, and use the threshold T to judge each pixel in a row of data. If the intensity value of the point is greater than the threshold T, the point data is recorded as 1, otherwise it is recorded as 0. Using the threshold segmentation method, the CD of the aerial image can be determined by calculating the number of data points that are 1. The size of the mask imaging pattern in the 0-degree direction is CD0 = 18nm, and the size of the mask imaging pattern in the 90-degree direction is CD1 = 12nm.

[0088] The difference between the mask imaging pattern size CD0 in the 0 degree direction and the mask imaging pattern size CD1 in the 90 degree direction is recorded as the compensation amount ΔCD, ΔCD = 6nm, then the corresponding mask mark width compensation amount is Δw = 4*ΔCD = 24nm. The width of the mask mark in the 90 degree direction w1 = w0 + Δw = 112nm.

[0089] Step 2. Acquisition of spatial image set:

[0090] Set the projection objective to solve the aberrations Z7~Z9 and Z 14 ~Z 16 There are 6 types of wave aberrations, and the amplitude range of each wave aberration is [-a, a], a = 0.1λ. The combination of Zernike coefficients is designed using the Box_Behnken Design statistical sampling method to obtain a uniform and sufficient set of spatial images.

[0091] Use the bbdesign function in MATLAB. The input is the number of Zernike aberrations, and the output is the desired design combination. When the wave aberration to be solved is 6, the input parameter of the bbdesign function is 6, and the output matrix is ​​as follows:

[0092]

[0093] The matrix has six columns, corresponding to six Zernike aberrations, and 54 rows, representing 54 combinations. By multiplying the wave aberration amplitude a on the basis of the matrix, the Zernike coefficient combination A = a·D is obtained.

[0094] The two mask marks were simulated using commercial simulation software. First, the 0-degree direction aerial image was simulated. According to the above Zernike coefficient combination, each group of Zernike coefficient combination generates an aerial image, and a total of N = 54 aerial images are generated. Each aerial image has 101 points in the vertical axis direction and 71 points in the axial direction, that is, the pixel specification is M = 71 × 101. These aerial images are recorded as Among them, each pixel point of the i-th spatial image is expressed as follows:

[0095]

[0096] Take the pixels in the spatial image from left to right in columns, connect them end to end, and arrange them into a new column. The vector I i ,have:

[0097] I i =[a 1,1 ,a 2,1 ,…,a 71,1 ,a 1,2 ,a 2,2 ,…,a 71,2 ,…,a 1,101 ,a 2,101 ,…,a 71,101 ] T

[0098] The 54 spatial images correspond to 54 vectors I, forming the 0-degree spatial image set IM0:

[0099] IM0=[I1,I2,…,I 54 ] M×N

[0100] Next, simulate the 90-degree spatial image in the same way and construct a set of 90-degree spatial images:

[0101] IM1=[I′1,I′2,…,I′ 54 ] M×N

[0102] Step 3. Generation of principal component matrix and regression matrix:

[0103] Taking the processing of the 0-degree spatial image set as an example, the specific implementation process is explained. Perform principal component analysis on the 0-degree spatial image set IM0 to obtain the principal component matrix PC0 and principal component coefficient matrix C0 corresponding to the 0-degree spatial image. In MATLAB, this step can be completed by calling the function pca, and the calling format is as follows:

[0104] [C0,PC0,~]=pca(IM0)

[0105] Among them, C0 is the principal component coefficient matrix, with a size of N × N. PC0 is the principal component matrix, with a size of M × N.

[0106] The principal component coefficient matrix C0 can be expressed as follows:

[0107]

[0108] Each column in the matrix C0 represents a principal component coefficient, and is arranged from left to right in descending order of eigenvalue. In order to balance the operation speed and accuracy, we take the first 30 principal components to participate in the operation, that is, the first 30 columns of the matrix C0:

[0109] C′0=C0(:,1:30)

[0110] Linear regression analysis of the principal component coefficients and the wave aberration coefficients used for modeling can be performed by calling the function regress in MATLAB. The calling format is as follows:

[0111] b i =regress(C′ 0i ,[1A])

[0112] Among them, C′ 0iis the i-th principal component coefficient (1≤i≤30), A is the wavefront aberration combination matrix obtained using the Box_Behnken Design method, b i is the corresponding regression coefficient. Combining the first 30 regression coefficients, we get the regression matrix RM0 in the 0 degree direction:

[0113] RM0=[b1,b2,…,b 30 ] T

[0114] In the same way as the above steps, the regression matrix RM1 corresponding to the 90-degree directional spatial image can be obtained.

[0115] Step 4. Collection of aerial images:

[0116] The parameters of the projection objective lens of the lithography machine to be tested are set. The parameters such as illumination, numerical aperture, mask period and spatial image scanning range are set in the same way as in step 1. The mask mark adopts the width after the graphic size correction. The lithography machine is started. The illumination light emitted by the light source is modulated by the illumination system to obtain the corresponding illumination mode, which is then irradiated to the test mask on the mask stage. The workpiece stage is used to drive the spatial image sensor to complete the actual spatial image acquisition, i.e., the spatial image to be tested in the 0-degree direction and the spatial image to be tested in the 90-degree direction.

[0117] In order to simulate the detection accuracy of the method in the projection lens of the lithography machine, 30 groups of Zernike coefficient combinations are randomly generated within the range of the single Zernike coefficient amplitude a to simulate the wave aberration distribution of the projection lens of the lithography machine, which is represented by the matrix B:

[0118]

[0119] Matrix B has 6 columns, corresponding to the six Zernike aberrations, and a total of 30 rows, corresponding to 30 sets of Zernike coefficient design combinations.

[0120] In the above combination, 30 groups of spatial images to be tested are generated to evaluate the detection accuracy of the present technology, wherein each group of spatial images to be tested includes one spatial image to be tested in the 0 degree direction and one spatial image to be tested in the 90 degree direction.

[0121] Step 5. Fitting and solving of the spatial image:

[0122] For the spatial image to be measured in the 0 degree direction, the pixel size is 71×101, with a total of 7171 pixels. Take the pixels by column and connect them end to end to get a column vector AI hor , AI hor The specifications are 7171×1.

[0123] The principal component matrix PC0 in the 0-degree direction obtained during the modeling process is used for fitting to obtain the principal component coefficient matrix of the spatial image to be measured:

[0124] V0=PC0 / AI hor

[0125] For the aerial image to be measured in the 90-degree direction, the principal component coefficient V1 can be obtained in the same way.

[0126] According to the regression matrix RM0 and RM1, the total regression matrix RM is obtained:

[0127]

[0128] According to the principal component coefficient matrices V0 and V1, the total principal component coefficient matrix V is obtained:

[0129]

[0130] The least squares method is used to fit the regression matrix RM and the principal component coefficient matrix V to obtain the wave aberration of the measured spatial image:

[0131] Z=RM\V

[0132] Figure 4 The figure shows the detection results of the six Zernike aberrations when the mask mark is not optimized, that is, w0=w1=88nm. The mask mark size optimization method proposed in this technology can effectively improve the detection accuracy of Z9 and Z under the condition that the detection accuracy of other Zernike aberrations remains basically unchanged. 16 The accuracy of solving two aberrations, such as Figure 5 and Figure 6 shown.

Claims

1. A method for detecting wavefront aberration of an extreme ultraviolet lithography projection objective based on principal component analysis of an aerial image, characterized in that: The steps include: Step 1. Correction of mask marks: ①Set simulation conditions: Set the lighting conditions, the partial coherence factor of traditional lighting is σ, and the partial coherence factor of annular lighting is [σ in , σ out ],σ in represents the internal coherence factor, σ out represents the external coherence factor; Set the numerical aperture NA of the projection objective lens, the mask period p, the mask marks are isolated spaces in the 0 degree direction and the 90 degree direction, and the width of the mask mark in the 0 degree direction is w0, and the width of the mask mark in the 90 degree direction is w1, w0=w1; Set the vertical axis acquisition length x of the spatial image, the vertical axis sampling interval dx, the acquisition range is symmetrical with the center of the workpiece stage, the axial direction acquisition length f of the spatial image, the axial sampling interval df, and the acquisition range is symmetrical with the axial center; ② Correction mask mark: Simulate a 0-degree spatial image without aberration, denoted as AI0; Simulate a 90-degree spatial image without aberration, denoted as AI1; Set the threshold T of the mask imaging pattern size CD; The threshold T is used to determine the mask imaging pattern size CD0 in the 0 degree direction and the mask imaging pattern size CD1 in the 90 degree direction for the aerial image AI0 in the 0 degree direction and the aerial image AI1 in the 90 degree direction respectively; The difference between the mask imaging pattern size CD0 in the 0 degree direction and the mask imaging pattern size CD1 in the 90 degree direction is recorded as the compensation amount ΔCD, then the corresponding mask mark width compensation amount Δw=4*ΔCD, and the width of the mask mark in the 90 degree direction w1=w0+Δw; Step 2. Acquisition of spatial image set: ① Set the projection objective to solve the six wave aberrations, namely Z7~Z9 and Z 14 ~Z 16 , and the amplitude range of each wave aberration is [-a,a]; ② Constructing matrix D: For the six wave aberrations, Box_Behnken Design statistical sampling is used to generate 54 combinations as 54 rows of matrix D, and the six wave aberrations are used as 6 columns of matrix D; Calculate the Zernike coefficient combination A, the formula is as follows: A=a·D; ③ The mask marks in the 0-degree direction and the 90-degree direction are simulated respectively. The 54 spatial images in the 0-degree direction constitute the spatial image set IM0 in the 0-degree direction, and the 54 spatial images in the 90-degree direction constitute the spatial image set IM1 in the 90-degree direction; Step 3. Generation of principal component matrix and regression matrix: ① Perform principal component analysis on the spatial image set IM0 in the 0-degree direction to obtain the principal component matrix PC0 and principal component coefficient matrix C0 corresponding to the spatial image in the 0-degree direction; Perform principal component analysis on the aerial image set IM1 at 90 degrees to obtain the principal component matrix PC1 and principal component coefficient matrix C1 corresponding to the aerial image at 90 degrees; ② Perform linear regression analysis on the principal component coefficient matrix C0 and the Zernike coefficient combination A to obtain the regression matrix RM0 in the 0 degree direction; Perform linear regression analysis on the principal component coefficient matrix C1 and the Zernike coefficient combination A to obtain the regression matrix RM1 in the 90-degree direction; Step 4. Collection of aerial images: Set the parameters of the projection objective lens of the lithography machine to be inspected. The parameters of illumination, numerical aperture, mask period and spatial image scanning range are set the same as in step 1. The mask mark adopts the corrected width. The photolithography machine is started, and the illumination light emitted by the light source is modulated by the illumination system to obtain the corresponding illumination mode, and then illuminates the test mask on the mask stage. The workpiece stage drives the spatial image sensor to complete the acquisition of the actual spatial image, that is, the spatial image to be tested in the 0-degree direction and the spatial image to be tested in the 90-degree direction; Step 5. Fitting and solving of the spatial image: ① Fit the principal component matrix PC0 to obtain the principal component coefficient matrix V0 of the spatial image to be measured in the 0 degree direction. The formula is as follows: V0=PC0 / AI hor In the formula, AI hor is the column vector of each pixel point in the measured spatial image in the 0 degree direction arranged in sequence; The principal component matrix PC1 is fitted to obtain the principal component coefficient matrix V1 of the aerial image to be measured at 90 degrees, and the formula is as follows: V1=PC1 / AI ver In the formula, AI ver is a column vector of pixels in the spatial image to be measured at 90 degrees arranged in sequence; ②According to the regression matrix RM0 and RM1, the total regression matrix RM is obtained, and the formula is as follows: According to the principal component coefficient matrices V0 and V1, the total principal component coefficient matrix V is obtained by combining them. The formula is as follows: The least squares fitting of the regression matrix RM and the principal component coefficient matrix V is used to obtain the wave aberration Z of the lithography machine projection lens, and the formula is as follows: Z = RM\V.

2. The method for detecting wavefront aberration of an EUV lithography projection objective lens based on principal component analysis of an aerial image according to claim 1, characterized in that: The numerical aperture NA ranges from 0.33 to 0.

55.

3. The method for detecting wavefront aberration of an EUV lithography projection objective lens based on principal component analysis of an aerial image according to claim 1, characterized in that: The lighting method described is a lighting method supported by a photolithography machine, including traditional lighting and annular lighting.

4. The method for detecting wavefront aberration of an EUV lithography projection objective lens based on principal component analysis of an aerial image according to claim 1, characterized in that: When the lighting method is traditional lighting, the partial coherence factor σ ranges from 0.1 to 0.

9.

5. The method for detecting wavefront aberration of an EUV lithography projection objective lens based on principal component analysis of an aerial image according to claim 1, characterized in that: When the lighting mode is annular lighting, the internal coherence factor σ in The range is 0.1~0.8, and the external coherence factor σ out The range is 0.2~0.9, where σ out -σ in Not less than 0.

1.

6. The method for detecting wavefront aberration of an EUV lithography projection objective lens based on principal component analysis of an aerial image according to claim 1, characterized in that: The acquisition length x in the vertical axis direction is any value between 100 nm and 3000 nm; the acquisition step length dx in the vertical axis direction is any value between 0.5 nm and 100 nm.

7. The method for detecting wavefront aberration of an EUV lithography projection objective lens based on principal component analysis of an aerial image according to claim 1, characterized in that: The collection length f in the axial direction is any value between 1000nm and 8000nm; the collection step length df in the axial direction is any value between 1nm and 125nm.

8. The method for detecting wavefront aberration of an EUV lithography projection objective lens based on principal component analysis of an aerial image according to claim 1, characterized in that: The width w0 of the mask mark in the 0 degree direction is any value between 50nm and 400nm; the mask period p is any value between 500nm and 5000nm.

Images