A large numerical aperture lithography projection objective wave aberration splicing device and method

Abstract

The application discloses a large numerical aperture lithography projection objective wave aberration splicing and stitching device and method, which comprises the following in sequence along the illumination direction of an illumination system: a mask, a mask table, a large numerical aperture lithography projection objective, a Shack-Hartmann sensor for aligning an actual image point on an imaging surface after magnification of a mask pattern by the large numerical aperture lithography projection objective, a sensor motion table capable of six-dimensional motion for bearing the Shack-Hartmann sensor, and an interferometer measurement system for measuring alignment position coordinate information and returning three-dimensional coordinates of the alignment position. The sensor motion table is used to align the Shack-Hartmann sensor with the actual image point and scan the large numerical aperture lithography projection objective wavefront information in the form of a Hadamard measurement matrix code along an I-shaped beam, differential measurement is used to separate measurement errors caused by translation and tilt adjustment errors, wavelet transform is used to improve signal sparsity, and a compressed sensing algorithm is used to reconstruct an actual wavefront, so that the large numerical aperture wave aberration splicing and stitching technology is realized.

Description

Technical Field

[0001] This invention belongs to the fields of optical measurement and lithography technology, specifically a device and method for stitching together wavefront aberrations of large numerical aperture lithography projection lenses. Background Technology

[0002] In today's semiconductor field, as feature sizes continue to shrink, the number of transistors that can be integrated on a chip increases in accordance with Moore's Law. The continuous shrinking of feature sizes requires shorter wavelengths and larger numerical apertures. Lithography equipment has evolved from visible light g-line (436nm), ultraviolet i-line (UV 365nm), deep ultraviolet KrF (DUV 248nm), ArF (DUV 193nm) to the most advanced extreme ultraviolet (EUV 13.5nm) lithography equipment. After adopting immersion technology, the NA value of lithography projection lenses has increased from 0.50-0.93 to 0.85-1.35, thereby further improving the resolution. However, the requirements for wavefront aberration of lithography projection lenses have also increased. Therefore, wavefront aberration detection for large numerical aperture lithography projection lenses has become very important.

[0003] Hartmann and Shack's Shack-Hartmann sensor uses a condenser lens to collimate the test wavefront, then uses a microlens array to segment the collimated wavefront into sub-wavefronts. The centroid coordinates of the resulting spot on a CCD array are acquired to calculate the offset, which is then used to reconstruct the wavefront tilt and obtain the wavefront information. However, testing large-aperture wavefronts with the Shack-Hartmann sensor requires more microlens arrays, increasing manufacturing errors. It also necessitates higher CCD array resolution, increased storage space, and a higher sampling rate, leading to increased cost and computational storage requirements. Compressed sensing algorithms were systematically proposed by Donoho, Candes, and Chinese-American mathematician Terence Tao. The traditional Nyquist theorem tells us that in order to completely preserve the information in the original signal after sampling, the sampling frequency must be greater than twice the highest frequency in the signal. However, for compressible sparse signals, compressed sensing algorithms can undersample the data in a way that is far below the standard of the traditional Nyquist theorem. When the signal under test has sparse or compressible characteristics, the signal can be accurately reconstructed with a small number of signal measurements while reducing the sampling rate. Compressed sensing is increasingly widely used in image processing, optics, infrared, terahertz, and radar imaging.

[0004] To reduce sampling rate, save costs, and conserve computational storage space by using a small-aperture Shake-Hartmann sensor to test large-aperture wavefront aberrations, this study proposes a large numerical aperture lithography projection lens wavefront aberration stitching device and method. This device uses a sensor stage to move and support the Shake-Hartmann sensor, aligning it with the actual image point along an I-shape and scanning the large-aperture wavefront information in the form of a Hadamard measurement matrix. Differential measurement is used to separate the measurement errors caused by translation and tilt adjustment errors. Wavelet transform is used to improve signal sparsity, and compressed sensing algorithms are used to reconstruct the actual wavefront, thus realizing the large numerical aperture wavefront aberration stitching technique. Summary of the Invention

[0005] To test large-aperture wavefront aberrations using a small-aperture Shaker-Hartmann sensor, thereby reducing the sampling rate, saving costs, and conserving computational storage space, this study proposes a large numerical aperture lithography projection lens wavefront aberration stitching device and method.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A wavelet aberration stitching device for a large numerical aperture lithography projection lens is disclosed. The device includes an illumination system, a mask, a mask stage, a large numerical aperture lithography projection lens, a Shaker-Hartmann sensor, a sensor motion stage, and an interferometer.

[0008] The illumination system provides a laser beam for illumination. The mask and Shak-Hartmann sensor are mounted on a mask stage and a sensor motion stage with six degrees of freedom, designed to facilitate adjustment of measurement errors caused by translation and tilt during testing. The sensor motion stage aligns the Shak-Hartmann sensor with the actual image point and scans along an I-shape, scanning the wavefront information of the large numerical aperture lithography projection lens in the form of a Hadamard measurement matrix. An interferometer measures the three-dimensional coordinates of the alignment position. Differential measurements are used to separate the measurement errors caused by translation and tilt adjustment errors. Wavelet transform is used to improve signal sparsity, and the actual wavefront is reconstructed using a compressed sensing algorithm to achieve large numerical aperture wavefront aberration stitching technology.

[0009] The Shaker-Hartmann sensor divides the large numerical aperture test wavefront into a set of sub-wavefronts using a microlens array, calculates the wavefront tilt by acquiring the offset of the spot centroid using a CCD, and then reconstructs the large numerical aperture wavefront aberration.

[0010] After the sensor motion stage carries the Shaker-Hartmann sensor and aligns it with the actual image point, it scans along an I-shape and scans the wavefront information of the large numerical aperture lithography projection lens in the form of a Hadamard measurement matrix. The wavefront information of the large numerical aperture lithography projection is then stored in the form of a Hadamard measurement matrix encoding.

[0011] Wavelet transform is used to improve signal sparsity and reduce aliasing error, thereby improving the accuracy of compressed sensing algorithms in reconstructing wavefront information of large numerical aperture lithography projection lenses.

[0012] A method for stitching together wavefront aberrations of large numerical aperture lithography projection lenses includes the following steps:

[0013] Step 1: The light from the illumination system passes through the mask pattern along the projection direction and is scaled and imaged at the optimal focal plane by a large numerical aperture lithography projection lens at a certain magnification. The sensor motion stage carries the Shaker-Hartmann sensor to find the actual imaging position of the mask pattern on the imaging surface.

[0014] Step 2: Since the sensor aperture is small, it is impossible to detect the complete wavefront information. Therefore, the sensor motion stage scans the entire large-aperture wavefront information along the I-shape using a preset Hadamard measurement matrix.

[0015] Step 3: By using multiple measurements of the small aperture wavefront information, adjustment errors caused by translation and tilt during motion scanning can be eliminated;

[0016] Step 4: Project the sampled data onto the orthogonal basis neighborhood of the wavelet to make the signal sparse in that neighborhood, while the wavelet transform improves the sparsity of the data.

[0017] Step 5: Input the measurement signal, measurement matrix, orthogonal basis, and other parameters. Set the residual r. Perform an inner product of the residual and the observation matrix to find the best matching column and store it. Clear the matching column of the observation matrix and update the residual. When the residual is less than the threshold or the number of iterations has reached the upper limit, perform an inverse wavelet transform on the stored matrix and use least squares recovery to reconstruct the wavefront aberration matrix of the large numerical aperture lithography projection lens. This enables the compressed sensing algorithm to stitch together the collected small aperture wavefront information into the actual wavefront aberration information of the large numerical aperture lithography projection lens.

[0018] The principle of this invention is as follows: the sensor motion stage aligns the Shaker-Hartmann sensor with the actual image point and scans along an I-shape, scanning the wavefront information of the large numerical aperture lithography projection lens in the form of a Hadamard measurement matrix. The interferometer measures the three-dimensional coordinates of the alignment position. Differential measurement is used to separate the measurement errors caused by translation and tilt adjustment errors. Wavelet transform is used to improve the signal sparsity and the actual wavefront is reconstructed using a compressed sensing algorithm to realize the large numerical aperture wave aberration stitching technology.

[0019] Compared with the prior art, the advantages of this invention are: the device and method have a simple structure, reduce the cost of large aperture wave aberration detection, and reduce the sampling rate through compressed sensing algorithm, saving computing and storage space. At the same time, it separates the adjustment error caused by translation and tilt, improves the measurement accuracy, and realizes a low-cost, computing and storage-saving large numerical aperture wave aberration stitching technology. Attached Figure Description

[0020] Figure 1 This invention relates to a device for stitching together wavefront aberrations of large numerical aperture photolithography projection lenses.

[0021] Figure 2 This is a schematic diagram of a Shaker-Hartmann sensor.

[0022] Figure 3 This is a schematic diagram of wavelet aberration splicing and stitching technology.

[0023] Figure 4 Test procedure for wavefront aberration stitching method of large numerical aperture lithography projection lens. Detailed Implementation

[0024] To better illustrate the purpose, technical solution, and advantages of the present invention, a further detailed description is provided below in conjunction with the accompanying drawings, but this should not limit the scope of protection of the present invention.

[0025] like Figure 1 The large numerical aperture lithography projection lens wavefront aberration stitching device shown includes an illumination system 1, a mask 2, a mask stage 3, a large numerical aperture lithography projection lens 4, a Shaker-Hartmann sensor 5, a sensor motion stage 6, and an interferometer 7.

[0026] The illumination system 1 provides a laser beam for illumination. The mask 2 and the Shak-Hartmann sensor 5 are respectively mounted on the mask stage 3 and the sensor motion stage 6, which have six degrees of freedom. This is to facilitate the adjustment of measurement errors caused by translation and tilt during the test. The sensor motion stage 6 aligns the Shak-Hartmann sensor 5 with the actual image point and scans along an I-shape, scanning the wavefront information of the large numerical aperture lithography projection lens 4 in the form of a Hadamard measurement matrix. The interferometer 7 measures the three-dimensional coordinates of the alignment position. Differential measurement is used to separate the measurement errors caused by translation and tilt adjustment errors. Wavelet transform is used to improve signal sparsity, and the actual wavefront is reconstructed using a compressed sensing algorithm to realize the large numerical aperture wavefront aberration stitching technique.

[0027] Figure 2 This is a schematic diagram of a Shaker-Hartmann sensor. The large numerical aperture test wavefront is divided into sub-wavefronts by a microlens array. The offset of the centroid of the light spot is collected by a CCD to calculate the wavefront tilt, and then the large numerical aperture wavefront aberration is reconstructed.

[0028]

[0029] A = (Z) T Z) -1 Z T G

[0030] in, These are the slopes of the m-th sub-aperture spot in the x and y directions, respectively, according to the formula (3) above;

[0031] These are the slopes of the nth-order Zernike polynomial of the sub-aperture spot in the x and y directions, respectively.

[0032] a n These are the coefficients of the nth Zernike polynomial;

[0033] m is the number of sub-aperture light spots;

[0034] n is the order of the Zernike polynomial;

[0035] Waveform aberration stitching method as follows Figure 3 As shown:

[0036] Figure 3 The lower right corner shows the sub-aperture sampling area of ​​the Shaker-Hartmann sensor. It is much smaller than a large numerical aperture lithography objective, but by scanning with the sensor's motion stage, a small aperture can cover the entire large aperture. Then, through algorithmic stitching, each sub-aperture within the aperture is sampled in the form of a Hadamard matrix encoding. The small white squares represent '-1', and the small black squares represent '+1'. The Hadamard matrix exhibits good orthogonality and satisfies the finite isometric property (RIP).

[0037]

[0038] Where 0 < δK < 1, K represents sparsity, x represents the original signal, Φ represents the Hadamard matrix, and ||·|| represents the norm. From an energy perspective, the RIP property means that the energy of the transformed signal cannot exceed twice the energy of the original signal. There are many studies on the RIP property of measurement matrices. Measurement matrices that have been proven to have the RIP property include Gaussian matrices, Fourier matrices, random Bernoulli matrices, and random matrices.

[0039] Compressed sensing theory consists of three main elements: signal sparsity, measurement matrix, and signal recovery algorithm. Since some signals are not sparse in nature, it is generally necessary to transform the original signal using orthogonal bases. This involves projecting the original signal through an orthogonal base to make the signal exhibit sparsity in certain specific transform domains. Therefore, wavelet transform is used to improve the sparsity of the original signal.

[0040] x=Ψs (3)

[0041] Where Ψ is the wavelet transform orthogonal basis, and s is the sparse signal after the original signal is transformed by the wavelet transform.

[0042] Therefore, we can conclude that:

[0043] G = Φx = ΦΨs = As (4)

[0044] Where, G represents a slope matrix of M×1 dimension, Φ represents a Hadamard measurement matrix of M×N dimension, Ψ represents an orthogonal basis of wavelet transform of N×N dimension, s is a sparse signal of N×1 dimension after wavelet transform of the original signal, and A represents an observation matrix of M×N dimension composed of the measurement matrix and the wavelet orthogonal basis. M represents the number of Hadamard matrix encoding measurements, N represents the number of Shack-Hartmann sub-aperture sampling times, M < N. Assuming that the sampling rate of the traditional Nyquist sampling theorem is 1, then the compressive sensing sampling rate is reduced from 1 to M / N. The smaller M is, the lower the sampling rate, and the more computing storage space can be saved, but the algorithm difficulty will also increase accordingly.

[0045] During the process of the sensor moving stage carrying the Shack-Hartmann sensor, adjustment errors caused by translation and tilt will occur:

[0046]

[0047] Where, W i represents the wavefront aberration measured by the i-th sub-aperture, represents the wavefront aberration of the central sub-aperture, a, b, c represent translation and tilt coefficients, i = (1, 2,... M), so we can get:

[0048]

[0049] As a constant term, c does not affect the calculation of the slope, so it can be ignored.

[0050] The test process is as Figure 4 shown:

[0051] Step 1: The light of the illumination system passes through the mask pattern along the projection direction and is scaled and imaged at the best focus plane with a certain magnification by a large numerical aperture lithography projection objective. The sensor moving stage carries the Shack-Hartmann sensor to find the actual imaging position of the mask pattern on the imaging plane;

[0052] The sensor moving stage scans the entire large aperture wavefront information along the I-shaped with a preset Hadamard measurement matrix; Step 2: Since the aperture of the sensor is small and the complete wavefront information cannot be detected,

[0053] When the sensor moving stage scans the entire large aperture wavefront information along the I-shaped with a preset Hadamard measurement matrix; Step 3: Through the multiple measured small aperture wavefront information, the adjustment errors caused by translation and tilt during the moving scan can be eliminated;

[0054] Step 4: Project the sampled data in the wavelet orthogonal basis domain to make the signal sparse in this domain, and at the same time the wavelet transform improves the sparsity of the data;

[0055] Step 5: Input the measurement signal, measurement matrix, orthogonal basis, and other parameters. Set the residual r. Perform an inner product of the residual and the observation matrix to find the best matching column and store it. Clear the matching column of the observation matrix and update the residual. When the residual is less than the threshold or the number of iterations has reached the upper limit, perform an inverse wavelet transform on the stored matrix and use least squares recovery to reconstruct the wavefront aberration matrix of the large numerical aperture lithography projection lens. This enables the compressed sensing algorithm to stitch together the collected small aperture wavefront information into the actual wavefront aberration information of the large numerical aperture lithography projection lens.

Claims

1. A wavefront abberation stitching device for large numerical aperture lithographic projection objective, characterized in that: The device comprises an illumination system (1), a mask (2), a mask table (3), a large numerical aperture lithography projection objective (4), a Shack-Hartmann sensor (5), a sensor motion table (6) and an interferometer (7); wherein: The illumination system (1) provides a laser beam for illumination, the mask (2) and the Shack-Hartmann sensor (5) are installed on the mask table (3) and the sensor motion table (6) with six degrees of freedom, the purpose is to facilitate the adjustment of measurement errors caused by translation and tilt during the test; the Shack-Hartmann sensor (5) is aligned with the actual image point by the sensor motion table (6) and scans the large numerical aperture lithography projection objective (4) wavefront information along the H-shaped scan and in the Hadamard measurement matrix form, the interferometer (7) measures the three-dimensional coordinates of the alignment position, the measurement error caused by the translation and tilt adjustment error is separated out by using the differential measurement, the signal sparsity is improved by the wavelet transform and the actual wavefront is reconstructed by the compressed sensing algorithm to realize the large numerical aperture wave aberration splicing and stitching technology.

2. The wavefront stitching device for a large numerical aperture lithographic projection objective according to claim 1, characterized in that: The Shack-Hartmann sensor (5) divides the large numerical aperture test wavefront into a sub-wavefront set by the microlens array, the offset of the spot centroid is collected by the CCD to calculate the wavefront tilt, and then the large numerical aperture wave aberration is reconstructed.

3. The wavefront stitching device for a large numerical aperture lithographic projection objective according to claim 1, characterized in that: After the sensor motion table (6) carries the Shack-Hartmann sensor (5) to align the actual image point, the large numerical aperture lithography projection objective (4) wavefront information is scanned along the H-shaped scan and in the Hadamard measurement matrix form, and the large numerical aperture lithography projection objective (4) wavefront information is stored in the form of Hadamard measurement matrix encoding.

4. The wavefront stitching device for a large numerical aperture lithographic projection objective according to claim 1, characterized in that: The signal sparsity is improved by using the wavelet transform to reduce the aliasing error, thereby improving the accuracy of the compressed sensing algorithm to reconstruct the large numerical aperture lithography projection objective (4) wavefront information.

5. A wavefront aberration stitching method for large numerical aperture lithographic projection objective, characterized in that: The steps include: Step 1: the light of the illumination system (1) passes through the mask pattern along the projection direction and is zoomed and imaged on the best focal plane by the large numerical aperture lithography projection objective (4), the sensor motion table (6) carries the Shack-Hartmann sensor (5) to find the actual imaging position of the mask pattern on the imaging plane; Step 2: the sensor motion table (6) scans the entire large aperture wavefront information along the H-shaped scan in the preset Hadamard measurement matrix; Step 3: by measuring a plurality of small aperture wavefront information, the adjustment error caused by translation and tilt during the motion scanning can be eliminated; Step 4: the wavelet orthogonal basis field projection is performed on the sampling data, so that the signal is sparse in the field, and the wavelet transform improves the sparsity of the data; Step 5: input the measurement signal, the measurement matrix and the orthogonal basis parameters, set the residual r, find the most matching column by doing the inner product of the residual and the observation matrix and store it, clear the observation matrix matching column, update the residual, when the residual is less than the threshold value or the iteration number has reached the upper limit, perform the inverse wavelet transform on the stored matrix and use the least square recovery to reconstruct the large numerical aperture lithography projection objective wave aberration matrix, thereby realizing the compressed sensing algorithm to splice and stitch the set of collected small aperture wavefront information into the actual wave aberration information of the large numerical aperture lithography projection objective (4).

Images