Artificial intelligence anti-disturbance photoetching objective lens wave aberration measurement method

Abstract

The invention discloses an artificial intelligence anti-disturbance photoetching objective wave aberration measurement method, which is characterized in that on the aspect of a photoetching objective wave aberration measurement algorithm, a U-shaped neural network structure based on physical information is used, an integral encoder-decoder structure of a universal U-NET neural network model is reserved, and meanwhile, a pooling layer is removed. Compared with the traditional four-step phase shift (or five-step phase shift), the innovation points of the invention are as follows: (1) on hardware, the piezoelectric phase shifter only needs to move once, and only two-step phase shift needs to be obtained; compared with a traditional phase-shifting wave aberration measurement system, the phase-shifting wave aberration measurement system is higher in measurement speed. And (2) on hardware, a traditional wave aberration measurement system needs an ultrahigh-precision piezoelectric phase shifter, and the method only needs a piezoelectric phase shifter with a low-precision standard (the precision is superior to 50nm). And (3) in terms of algorithm, the two interferograms can be accurately measured to obtain the wave aberration of the photoetching objective lens. And (4) in the aspect of wave aberration reconstruction precision, because the environmental interference is reduced, the precision of the method is higher.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithography exposure, and particularly relates to a method for measuring the wave aberration of a lithography objective lens with artificial intelligence anti-disturbance. Background Art

[0002] Lithography exposure is the core subsystem in a lithography machine. An exposure system with a high numerical aperture (NA) can significantly improve the exposure resolution of the lithography machine and enhance the manufacturing process of semiconductor technology. Nowadays, most lithography exposures use high-NA objective lenses. Wave aberration detection, such as based on the five-step phase-shifting method (or other multi-step phase-shifting methods), is a complex process and is vulnerable to interference from factors such as external vibrations, air flow disturbances, and piezoelectric phase shifter errors (J. H. Bruning, D. R. Herriott, J. E. Gallagher, D.P. Rosenfeld, A. D. White, and D. J. Brangaccio, "Digital Wavefront MeasuringInterferometer for Testing Optical Surfaces and Lenses," Appl. Opt. 13, 2693-2703 (1974)).

[0003] To solve this problem, the literature (Feng Peng, Li Zhongliang, Wang Xiangchao, et al. Research on polarization phase-shifting point diffraction interference wave aberration detection technology [J]. Chinese Journal of Lasers, 2022, 49(21): 90-97.) proposed using a polarization array technology to simultaneously measure four-step phase shifting to achieve anti-interference ability. This technology uses a short coherence length light source and a single-mode polarization-maintaining fiber to generate two point sources, which can output orthogonally linearly polarized light; a quarter-wave plate is added to the optical path, and a micro-polarization array camera is used to achieve spatial synchronous phase shifting based on a single interferogram. By adjusting the light intensity ratio of the two beams of light through an attenuator, the contrast of the interference fringes can be adjusted, but polarization devices with high polarization accuracy are required, and it is not conducive to the measurement of wave aberration at the working wavelength.

[0004] From the perspective of the mathematical model, the quantitative and accurate measurement of wave aberration is an inverse problem of solving an underdetermined (or overdetermined) non-linear equation system. From a more general perspective, researchers have proposed solutions based on artificial intelligence deep learning for solving the inverse problems of these underdetermined (or overdetermined) non-linear equation systems in optical imaging, machine vision, and holography. Through the enhancement of deep learning, these underdetermined (or overdetermined) non-linear equation systems have been solved in a faster speed, higher accuracy, and with less data. In the measurement of the wave aberration of a lithography objective lens, the disturbance error sources make such non-linear equation system models more complex, increasing the difficulty of solving, resulting in a bottleneck in the measurement accuracy of the wave aberration of the lithography machine. Summary of the Invention

[0005] In view of the above deficiencies in the prior art, the present invention provides a method for measuring the wave aberration of a lithographic objective lens with artificial intelligence anti-disturbance.

[0006] A method for measuring the wave aberration of a lithographic objective lens with artificial intelligence anti-disturbance, one side of the lithographic objective lens to be measured has a mask layout surface, and the opposite side is used as a chip surface. A lithographic objective lens wave aberration measuring device is used. The lithographic objective lens wave aberration measuring device includes: A laser for emitting light for measurement; A beam splitter is connected to the laser through an incident optical fiber, and is also connected to a test optical fiber and a reference optical fiber. The beam splitter is used to divide the light emitted by the laser into two beams. One beam enters the test optical fiber as test light, and the other beam enters the reference optical fiber as reference light; A piezoelectric phase shifter is arranged in the reference optical fiber for phase-shifting the reference light; A pupil aperture stop has a small hole with a small aperture and a large hole with a large aperture. The small hole is circular, and the large hole is a regular polygon or circular; An image sensor for detecting and recording an interference pattern; The method for measuring the wave aberration of a lithographic objective lens with artificial intelligence anti-disturbance includes the following steps: (1) Align and fix the end faces of the test optical fiber and the reference optical fiber on the mask layout surface of the lithographic objective lens to be measured, and the fixed positions on the mask layout surface are symmetric about the optical axis of the lithographic objective lens to be measured; the pupil aperture stop is fixed on the chip surface of the lithographic objective lens to be measured, and the small hole and the end face of the reference optical fiber are in a conjugate imaging relationship, and the large hole and the test optical fiber are in a conjugate imaging relationship. The light emitted from the small hole and the light emitted from the large hole form an interference pattern on the image sensor, and are recorded as the first interference pattern by the image sensor; control the piezoelectric phase shifter to shift the phase by π / 2 and record it as the second interference pattern; (2) Train to obtain a PIUN network, and the PIUN network is a U-shaped neural network structure based on physical information; (3) Input the first interference pattern into the PIUN network to generate two sequences of interference patterns {I 1 , I 2 , I 3 , I 4 , I 5} and {I 1 , I’ 2 , I’ 3 , I’ 4 , I’ 5}. Among the 5 interference patterns in each sequence, I 1 represents the interference pattern when the phase of the piezoelectric phase shifter is 0, and the phase difference of the piezoelectric phase shifter between the latter interference pattern and the previous interference pattern is 2π / 5; Extract I from the sequence 2 Perform a structural similarity operation with the second interference pattern. If the similarity exceeds 0.7, then {I 1 , I 2 , I 3 , I 4 , I 5} is used as the interference pattern data and input into Equation 1; if the similarity is lower than 0.7, then {I 1 , I’ 2 , I’ 3 , I’ 4 , I’ 5} is used as the interference pattern data and input into Equation 2; Equation 1 is: , where Phi is the wave aberration; Equation 2 is: , where PhyLoss is the physical loss function, , , , represent the derivatives of the interference image along the horizontal direction of the picture, , , , k1 , I k2 , I k3 , I k4 , I k5}, and the wave aberration corresponding to the interference pattern is Phi k , where k ranges from 1 to the total number of a series of interference patterns; Construct a deep learning dataset {I k1 , I k2 ; Phi k}, divide the deep learning dataset into training data and validation data, input the training data into the initial PIUN network for training, and use the validation data for testing until the PIUN network converges. When the loss function of the PIUN network converges to the 10 -7 th power, the trained PIUN network is obtained.

[0007] Preferably, the measurement is performed in a high-vacuum and highly stable point diffraction wave aberration measurement system. High vacuum means the vacuum degree ≤ 1×10 -7Pa, high stability refers to the shock-absorbing environment of the VC-F level.

[0008] Preferably, the PIUN network consists of an image input layer, an encoder, a decoder, and an image output layer. There is 1 image input layer and 5 image output layers. Each image output layer is in a parallel relationship, and each image output layer corresponds to an independent encoder and decoder.

[0009] More preferably, each encoder consists of 10 encoding modules, and each decoder consists of 10 decoding modules. Each encoding module is composed of a 3×3 convolutional module, a normalization module, and a Leaky Relu activation module. Each decoding module is composed of a 3×3 deconvolutional module, a normalization module, and a Leaky Relu activation module.

[0010] Preferably, an energy regulator for adjusting the light energy is provided in each of the test optical fiber and the reference optical fiber to cope with the measured lithography objective lens with different transmittances and make the interference fringe contrast the highest.

[0011] More preferably, after the beam splitter splits the beam, the energy ratio of the reference light to the test light is 1∶7~11, and most preferably 1∶9. The energy ratio is achieved by adjusting the two energy regulators.

[0012] Preferably, the phase shift accuracy of the piezoelectric phase shifter ≤50nm. In this application, only a piezoelectric phase shifter with a low-precision standard is required, which reduces the requirements for the piezoelectric phase shifter in terms of hardware.

[0013] Preferably, the diameter of the small hole is 50%~95% of the wavelength value of the light emitted by the laser. For example, when the wavelength of the laser is 633nm, the diameter of the small hole can be 500nm. When the large hole is a regular polygon, the area of the inscribed circle of the regular polygon is greater than or equal to the pupil area of the measured lithography objective lens. When the large hole is a circle, the circular area is greater than or equal to the pupil area of the measured lithography objective lens.

[0014] In the wavefront aberration measurement algorithm of a lithographic objective lens, the present invention innovatively proposes a physics-informed U-shaped neural network structure, with the English word being Physics-informed U-Net, abbreviated as the PIUN network. In the general U-NET neural network model, its overall encoder-decoder structure is retained, while the pooling layer is removed. We only use convolution and deconvolution operations to adjust the feature dimensions of the image. To better adapt to computers with different performances, we use a "batch" normalization layer in the encoder instead of a "layer" normalization layer. Innovatively, in the present invention, the input and output of the U-NET neural network are not "one-to-one"; instead, they are "one-to-five". Innovatively, the function of the network model not only includes predicting the phase-shifting interferogram, but also includes denoising and equalizing the light intensity.

[0015] Compared with the traditional four-step phase shift (or five-step phase shift), the innovation points of the present application include: (1) In terms of hardware, the piezoelectric phase shifter only needs to be moved once, and only two-step phase shifts need to be obtained. Compared with the traditional phase shift wavefront aberration measurement system, the measurement speed of the present invention is faster. (2) In terms of hardware, the traditional wavefront aberration measurement system requires a piezoelectric phase shifter with ultra-high precision, while the present invention only requires a piezoelectric phase shifter with low precision standards (precision better than 50 nm). (3) In terms of the algorithm, the wavefront aberration of the lithographic objective lens can be accurately measured from two interferograms in the present invention. (4) In terms of the wavefront aberration reconstruction accuracy, due to reducing environmental interference, the accuracy of the present invention is higher. Brief Description of the Drawings

[0016] Figure 1 It is a schematic structural diagram of the wavefront aberration measurement device for the lithographic objective lens of the present invention.

[0017] Figure 2 It is a schematic structural diagram of the aperture stop.

[0018] Figure 3 It is a schematic diagram of the artificial intelligence anti-disturbance wavefront aberration measurement method targeted by the present invention.

[0019] Figure 4 It is the deep learning network architecture of the present invention.

[0020] Figure 5 It is the structure of the PIUN network.

[0021] Figure 6 It is the structure of the encoder.

[0022] Figure 7 It is the structure of the decoder.

[0023] Figure 8 It is the structure of each encoding module.

[0024] Figure 9 It is the structure of each decoding module.

[0025] Figure 10 is the prior loss function of the physical function of the present invention.

[0026] Figure 11 are the test results of the experimental data for performance improvement of the present invention.

[0027] Reference numerals: laser 1, incident optical fiber 2, beam splitter 3, test optical fiber 4, energy regulator 5, reference optical fiber 6, energy regulator 7, piezoelectric phase shifter 8, lithography objective lens 9, mask layout 10, aperture stop 11, small hole 12, large hole 13, image sensor 14. Detailed implementation manners

[0028] As Figure 1 shown is a lithography objective wave aberration measuring device used in the lithography objective wave aberration measuring method for artificial intelligence anti-disturbance of the present invention, including a laser 1, an incident optical fiber 2, a beam splitter 3, a test optical fiber 4, an energy regulator 5, a reference optical fiber 6, an energy regulator 7, a piezoelectric phase shifter 8, an aperture stop 11, and an image sensor 14.

[0029] The laser 1 is used to emit light for measurement. The beam splitter 3 is connected to the laser 1 through the incident optical fiber 2. The beam splitter 3 is also connected to the test optical fiber 4 and the reference optical fiber 6 at the same time. The beam splitter 3 is used to divide the light emitted by the laser 1 into two beams, one beam enters the test optical fiber 4 as the test light, and the other beam enters the reference optical fiber 6 as the reference light.

[0030] An energy regulator 5 is provided in the test optical fiber 4, and an energy regulator 7 is provided in the reference optical fiber 6. The energy regulator is used to cope with the lithography objective to be measured with different transmittances to make the interference fringe contrast the highest. After the beam splitter 3 divides the beam, the energy ratio of the reference light to the test light is 1:7 to 11, and the most preferably is 1:9. The energy ratio is achieved by adjusting the two energy regulators.

[0031] The piezoelectric phase shifter 8 is provided in the reference optical fiber 6 and is used to perform phase shift on the reference light. In a preferred implementation manner, the phase shift accuracy of the piezoelectric phase shifter is ≤50 nm. In the present application, only a piezoelectric phase shifter with a low-precision standard is required, which reduces the requirements for the piezoelectric phase shifter in terms of hardware. Of course, if a piezoelectric phase shifter with a higher phase shift accuracy is used, it is naturally applicable, but the investment in hardware will increase in this case.

[0032] As Figure 2As shown, the aperture stop 11 has a small hole 12 with a smaller aperture and a large hole 13 with a larger aperture. The small hole 12 is circular, and the large hole 13 is a regular polygon or circular. The diameter of the small hole 12 is 50% - 95% of the wavelength value of the light emitted by the laser 1. When the large hole 13 is a regular polygon, the area of the inscribed circle of the regular polygon is greater than or equal to the pupil area of the lithography objective 9 to be measured. When the large hole 13 is circular, the circular area is greater than or equal to the pupil area of the lithography objective 9 to be measured. In one embodiment, the small hole 12 is circular with a diameter of 500 nm; the large hole 13 is square with a side length of 10 mm.

[0033] The image sensor 14 is used to detect and record the interference pattern.

[0034] One side of the lithography objective 9 to be measured has a mask layout 10, and the opposite side serves as the chip side.

[0035] During detection, the end faces of the test optical fiber 4 and the reference optical fiber 6 are aligned and fixed on the mask layout 10 of the lithography objective 9 to be measured, and the positions fixed on the mask layout 10 are axisymmetric with respect to the optical axis of the lithography objective 9 to be measured; the aperture stop 11 is fixed on the chip side of the lithography objective 9 to be measured, and the small hole 12 and the end face of the reference optical fiber 6 are in a conjugate imaging relationship, and the large hole 13 and the test optical fiber 4 are in a conjugate imaging relationship. The light emitted from the small hole 12 and the light emitted from the large hole 13 form an interference pattern on the image sensor 14.

[0036] A method for measuring the wave aberration of a lithography objective with artificial intelligence anti - disturbance includes the following steps: (1) Align and fix the end faces of the test optical fiber and the reference optical fiber on the mask layout of the lithography objective to be measured, and the positions fixed on the mask layout are axisymmetric with respect to the optical axis of the lithography objective to be measured; fix the aperture stop on the chip side of the lithography objective to be measured, and the small hole and the end face of the reference optical fiber are in a conjugate imaging relationship, and the large hole and the test optical fiber are in a conjugate imaging relationship. The light emitted from the small hole and the light emitted from the large hole form an interference pattern on the image sensor, which is recorded as the first interference pattern by the image sensor; control the piezoelectric phase shifter to shift the phase by π / 2 and record it as the second interference pattern.

[0037] (2) Train to obtain a PIUN network, and the PIUN network is a U - shaped neural network structure based on physical information.

[0038] In step (2), a series of interference patterns and corresponding wave aberrations are measured. The interference patterns are denoted as {I k1 , I k2 , I k3 , I k4 ,I k5}, and the wave aberrations corresponding to the interference patterns are Phi k , where k takes values from 1 to the total number of a series of interference patterns; Construct a deep - learning data set {Ik1 , I k2 ; Phi k}, divide the deep learning dataset into training data and validation data, input the training data into the initial PIUN network for training, and use the validation data for testing until the PIUN network converges. When the loss function of the PIUN network converges to 10 -7 to the power of, the trained PIUN network is obtained.

[0039] The measurement is carried out in a high-vacuum and high-stability point diffraction wave aberration measurement system. High vacuum means the vacuum degree ≤ 1×10 -7 Pa, and high stability means a VC-F grade shock-absorbing environment.

[0040] The PIUN network consists of an image input layer, an encoder, a decoder, and an image output layer. There is 1 image input layer and 5 image output layers. Each image output layer is in a parallel relationship, and each image output layer corresponds to an independent encoder and decoder.

[0041] Each encoder consists of 10 encoding modules, each decoder consists of 10 decoding modules. Each encoding module is composed of a 3×3 convolutional module, a normalization module, and a Leaky Relu activation module; each decoding module is composed of a 3×3 deconvolutional module, a normalization module, and a Leaky Relu activation module.

[0042] (3) Input the first interferogram into the PIUN network to generate two sequences of interferograms {I 1 , I 2 , I 3 , I 4 , I 5} and {I 1 , I’ 2 , I’ 3 , I’ 4 , I’ 5}. Among the 5 interferograms in each sequence, I 1 represents the interferogram when the phase of the piezoelectric phase shifter is 0, and the phase difference of the piezoelectric phase shifter between the latter interferogram and the previous interferogram is 2π / 5; Take out I 2 in the sequence and perform a structural similarity operation with the second interferogram. If the similarity exceeds 0.7, then {I 1 ,I 2 , I 3 , I 4 , I 5} is used as interferogram data and input into Equation 1; if the similarity is lower than 0.7, then {I 1 , I’2 , I’ 3 , I’ 4 ,I’ 5} As interferogram data, input Equation 2; Equation 1 is: , wherein, Phi is the wave aberration; Equation 2 is: , wherein, PhyLoss is the physical loss function, i.e., the interference Figure 2 dimensional difference modulus value (Loss Function).

[0043] , , , represents the derivative of the interferogram along the horizontal direction of the picture, , , , represents the derivative of the interferogram along the vertical direction of the picture.

[0044] Embodiment 1

[0045] The following is a case of measuring a DUV lithography objective lens using the present invention.

[0046] The light emitted by a 532 nm measurement laser is coupled into a polarization-maintaining single-mode fiber. After passing through a beam splitter, the fiber is "split into two" and enters two fibers. The two fibers are respectively denoted as "test fiber" and "reference fiber". The energy ratio of the "test fiber" to the "reference fiber" is 9:1. The test fiber contains an energy regulator. The reference fiber contains an energy regulator and a piezoelectric phase shifter with low precision requirements. The phase shift accuracy of the piezoelectric phase shifter is better than 50 nm (≤50 nm). The end face of the test fiber is flush with the end face of the reference fiber and is fixed on the mask surface of the lithography objective lens. The two are symmetric about the optical axis of the lithography objective lens ("one on the left and one on the right", "equidistant"). The center distance between the end faces of the two fibers is 1 mm. The aperture stop consists of a small hole with a smaller aperture and a large hole with a larger aperture. The diameter of the small hole is 500 nm. The large hole is a square hole with a side length of 10 mm. The aperture stop is fixed on the chip surface of the lithography objective lens. Among them, the small hole and the end face of the reference fiber are in a conjugate imaging relationship; the large hole and the end face of the test fiber are in a conjugate imaging relationship. The light emitted from the small hole interferes with the light emitted from the large hole to form an interferogram on the image sensor. The computer drives the piezoelectric phase shifter to shift the phase by π / 2, and the interferogram is recorded again.

[0047] As Figure 3As shown in the figure, the artificial intelligence anti-disturbance wave aberration measurement method targeted by the present invention includes: (1) Training stage, obtaining a PIUN network (Physics-informed U-Net, abbreviated as PIUN network); (2) Application stage, quickly realizing the solution of the wave aberration of the lithography objective lens with anti-disturbance.

[0048] The method in the training stage: In a point diffraction wave aberration measurement system with high vacuum and high stability, a series of interference patterns and their corresponding wave aberrations are measured. High vacuum means the vacuum degree is 1×10 -7 Pa; high stability means a VC-F level shock absorption environment. Taking a certain measurement as an example, the interference patterns are {I 1 , I 2 , I 3 , I 4 , I 5}. Among them, I 1 represents the interference pattern when the phase of the piezoelectric phase shifter is 0, and the phase difference of the piezoelectric phase shifter corresponding to the interference patterns of I 2 and I 1 is 2π / 5; and so on, the phase difference of the piezoelectric phase shifter corresponding to adjacent serial numbers of the interference patterns is 2π / 5. The formula for calculating the wave aberration Phi (denoted as formula 1) is: , The steps in the training stage, as Figure 2 shown, include: Step 1: Under the conditions of vacuum (1×10 -7 Pa) and high stability (VC-F), 2000 groups of interference patterns are measured. Each group of interference patterns is denoted as {I k1 , I k2 , I k3 , I k4 , I k5}, where k takes integer values from 1 to 2000. Based on the 2000 groups of interference patterns, the corresponding 2000 groups of objective lens wave aberrations are solved, and each wave aberration is Phi k .

[0049] Step 2: Based on the deep learning dataset {I k1 , I k2 ; Phi k} constructed in Step 1, the number is 2000 groups, the same as in Step 1.

[0050] Step 3: Based on the data in Step 2, 1800 groups of data are used as training data and 200 groups of data are used as test data. Based on as Figure 3As shown, 1800 groups of data are input into the PIUN network for training, and 200 groups of data are used for testing until the PIUN network converges.

[0051] Step 4: When the loss function of the PIUN network converges to 10 -7 to the power of, the obtained PIUN network model parameters are the results of the deep learning training stage. In the application stage, only two interferograms need to be input to obtain the wave aberration of the lithographic objective.

[0052] The application stage described above is as Figure 4 shown. Taking a certain measurement as an example, Figure 1 two frames of interferograms are collected in the system, marked as: the first interferogram and the second interferogram. Taking the first interferogram as the input and PIUN as the deep network, two sequences of interferograms {I 1 , I 2 , I 3 , I 4 , I 5} and {I 1 , I’ 2 , I’ 3 , I’ 4 , I’ 5} will be generated. Take out I 2 in this sequence and perform "structural similarity" (SSIM) operation with the second interferogram. If the similarity exceeds 0.7, then {I 1 , I 2 , I 3 , I 4 , I 5} is used as interferogram data and input into formula (1). If the similarity is lower than 0.7, then {I 1 , I’ 2 , I’ 3 , I’ 4 , I’ 5} is used as interferogram data and input into formula (2) (see formula (2) below).

[0053] As Figure 5 shown, the PIUN network consists of an image input layer, an encoder, a decoder, and an image output layer. There is only 1 image input layer and 5 image output layers. Each image output layer is in a parallel relationship. Each image output layer corresponds to an independent encoder and decoder. As Figure 6 shown, each encoder consists of 10 encoding modules. As Figure 7 shown, each decoder consists of 10 decoding modules. As Figure 8As shown, each encoding module consists of a 3×3 convolutional module, a normalization module, and a Leaky Relu activation module. Similarly, as Figure 9 shown, each decoding module consists of a 3×3 deconvolutional module, a normalization module, and a Leaky Relu activation module.

[0054] The innovative loss function of the PIUN network described above is as Figure 10 shown. The loss function (LOSS) includes: the MSE loss function (MSELoss) and the PhyLoss physical loss function (PhyLoss). The MSE loss function is consistent with the definition of the MSE loss function commonly used in the field of artificial intelligence. The PhyLoss physical loss function of the present invention is constructed by the derivative operation of the interference pattern, as shown in the following formula (denoted as Formula 2): , where I 1 , I 2 , I 3 , I 4 are interference patterns, , , , represent the derivatives of the interference image along the horizontal direction of the picture, , , , represent the derivatives of the interference image along the vertical direction of the picture.

[0055] The experimental test results of the accuracy of the artificial intelligence anti-disturbance based on the present invention are as Figure 11 shown. Figure 11 This is a comparison of the experimental data between the present invention and the traditional time series piezoelectric phase shift method. The experimental results show that under the same laboratory conditions, compared with the time phase shift method (five-step phase shift wavefront aberration measurement method, PSI5), the RMS repeatability of the wavefront aberration measurement results obtained by the PIUN method is improved by 88.7%.

[0056] The specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, but these descriptions should not be construed as limiting the scope of the present invention. The protection scope of the present invention is defined by the appended claims, and any modification based on the claims of the present invention is within the protection scope of the present invention.

Claims

1. An artificial intelligence anti-disturbance lithography objective wave aberration measurement method, wherein one side of the lithography objective to be measured has a mask surface, and the other side opposite to it is used as a chip surface, characterized in that: A photolithography objective wave aberration measuring device is used, and the photolithography objective wave aberration measuring device comprises: A laser, which emits light for measurement; A beam splitter is connected to the laser through an incident optical fiber, and is also connected to a test optical fiber and a reference optical fiber, and is used to split the light emitted by the laser into two beams, one beam enters the test optical fiber as a test light, and the other beam enters the reference optical fiber as a reference light; A piezoelectric phase shifter, provided in the reference optical fiber, for performing phase shifting on the reference light; An aperture stop having a small hole with a smaller aperture and a large hole with a larger aperture, wherein the small hole is circular and the large hole is a regular polygon or a circle; an image sensor for detecting and recording the interference pattern; The artificial intelligence anti-disturbance lithography objective lens wave aberration measurement method comprises the following steps: (1) The end faces of the test optical fiber and the reference optical fiber are flush and fixed on the mask plate of the photolithography objective to be tested, and the positions of the fixed on the mask plate are symmetrical with respect to the optical axis of the photolithography objective to be tested; the aperture stop is fixed on the chip surface of the photolithography objective to be tested, and the small hole and the end face of the reference optical fiber are in a conjugate imaging relationship, and the large hole and the test optical fiber are in a conjugate imaging relationship, and the light emitted from the small hole and the light emitted from the large hole form an interference pattern on the image sensor, which is recorded as a first interference pattern by the image sensor; the piezoelectric phase shifter is controlled to shift the phase by π / 2, which is recorded as a second interference pattern; (2) Train to obtain the PIUN network, which is a U-shaped neural network structure based on physical information; (3) Input the first interferogram into the PIUN network to generate two sequences of interferograms {I1, I2, I3, I4, I5} and {I1, I'2, I'3, I'4, I'5}. In each sequence of five interferograms, I1 represents the interferogram when the phase of the piezoelectric phase shifter is 0. The phase difference between the piezoelectric phase shifter corresponding to the latter interferogram and the previous interferogram is 2π / 5. Take out I2 from the sequence and perform structural similarity calculation with the second interference pattern. If the similarity exceeds 0.7, {I1, I2, I3, I4, I5} is used as interference pattern data and input into formula 1. If the similarity is less than 0.7, {I1, I'2, I'3, I'4, I'5} is used as interference pattern data and input into formula 2. Formula 1 is: ， in, Phi is the wave aberration; Formula 2 is: ， in, PhyLoss is the physical loss function, , , , represents the derivative of the interference image along the horizontal direction of the image, , , , Represents the derivative of the interference pattern along the vertical direction of the image.

2. The artificial intelligence anti-disturbance lithography objective wave aberration measurement method according to claim 1, characterized in that: In step (2), a series of interference patterns and corresponding wave aberrations are measured, and the interference patterns are denoted as {I k1 , I k2 , I k3 , I k4 , I k5 }, the wave aberration corresponding to the interference pattern is Phi k , where k ranges from 1 to the total number of interference patterns; Building a Deep Learning Dataset k1 , I k2 ;Phi k }, the deep learning dataset is divided into training data and validation data, the training data is input into the initial PIUN network for training, and the validation data is used for testing until the PIUN network converges. When the loss function of the PIUN network converges to 10 -7 When it is raised to the power, the trained PIUN network is obtained.

3. The artificial intelligence anti-disturbance lithography objective wave aberration measurement method according to claim 2, characterized in that: The measurement is carried out in a high vacuum and high stability point diffraction wave aberration measurement system. High vacuum refers to a vacuum degree of ≤1×10 -7 Pa, high stability refers to VC-F level shock absorption environment.

4. The artificial intelligence anti-disturbance lithography objective wave aberration measurement method according to claim 1, characterized in that: The PIUN network consists of an image input layer, an encoder, a decoder, and an image output layer. There is one image input layer and five image output layers. Each image output layer is in parallel, and each image output layer corresponds to an independent encoder and decoder.

5. The artificial intelligence anti-disturbance lithography objective wave aberration measurement method according to claim 4, characterized in that: Each encoder consists of 10 encoding modules, and each decoder consists of 10 decoding modules. Each encoding module consists of a 3×3 convolution module, a normalization module, and a Leaky Relu activation module; Each decoding module consists of a 3×3 deconvolution module, a normalization module, and a Leaky Relu activation module.

6. The artificial intelligence anti-disturbance lithography objective wave aberration measurement method according to claim 1, characterized in that: The test optical fiber and the reference optical fiber are each provided with an energy regulator for regulating light energy.

7. The artificial intelligence anti-disturbance lithography objective wave aberration measurement method according to claim 6, characterized in that: After the beam splitter splits the light, the energy ratio of the reference light to the test light is 1:7-11.

8. The artificial intelligence anti-disturbance lithography objective wave aberration measurement method according to claim 1, characterized in that: The phase shift accuracy of the piezoelectric phase shifter is ≤50nm.

9. The artificial intelligence anti-disturbance lithography objective wave aberration measurement method according to claim 1, characterized in that: The diameter of the small hole is 50% to 95% of the wavelength of the light emitted by the laser; when the large hole is a regular polygon, the area of ​​the inscribed circle of the regular polygon is greater than or equal to the pupil area of ​​the photolithography objective to be measured; when the large hole is a circle, the area of ​​the circle is greater than or equal to the pupil area of ​​the photolithography objective to be measured.

Images