Multi-source cooperative high-precision exposure control system for photolithography machine

Abstract

The application discloses a photolithography machine multi-light-source cooperative high-precision exposure control system and relates to the technical field of photolithography machine control. The initial energy parameters of each light source are acquired, a coupling characteristic set of energy variation and light spot appearance variation is constructed, light intensity error, phase error and light spot distortion error are taken as inputs, a time sequence offset matrix between the light sources and a light source energy compensation matrix are respectively generated, a light source cooperative error model is constructed through matrix superposition and eigenvalue decomposition, exposure disturbance obtained by real-time monitoring of the photolithography machine is input into the error model, the final target exposure compensation parameters are obtained by using a quadratic optimization algorithm under a constraint condition after the preliminary compensation amount is obtained, the light source energy output, the light path phase and the pulse trigger time sequence are cooperatively adjusted according to the target exposure compensation parameters, the energy deviation, the phase drift and the time sequence mismatch can be simultaneously reduced, the exposure stability and the pattern shaping precision are significantly improved, and the application is suitable for high-resolution and multi-light-source photolithography equipment.

Description

Technical Field

[0001] This invention relates to the field of lithography machine control technology, specifically to a high-precision exposure control system for multi-source collaborative lithography machines. Background Technology

[0002] As advanced lithography processes evolve towards higher numerical apertures (NA), smaller linewidths, and more complex multiple exposure procedures, the light sources used in lithography machines have gradually evolved from single beams to multi-source superposition illumination modes. However, in actual working environments, energy fluctuations, phase shifts, and spot morphology changes among multiple light sources can exhibit significant inconsistencies on a nanosecond timescale, causing nonlinear perturbations in on-chip energy uniformity within tiny regions.

[0003] Especially under large field-of-view, high-scanning-speed exposure conditions, the differences in optical path lengths between different light sources accumulate due to ambient temperature gradients, micro-defects in optical components, and dynamic vibrations. This leads to unpredictable delays in the exposure timing between multiple light sources, causing a slight but continuous shift in the instantaneous exposure energy at the target illumination point. Although this shift is extremely small (typically less than 0.1%), it can cause cross-regional exposure dose metastability mismatches on the wafer, ultimately resulting in random walk of the critical dimension (CD), deterioration of the pattern edge roughness (LER), and even preventing some high-density patterns from being fully formed.

[0004] More seriously, during multi-source collaborative exposure, when a light source is affected by particle scattering, local degradation of the reflector, or thermal deformation of the aperture, its light intensity attenuation will exhibit a non-linear abrupt change. If this abrupt change cannot be detected and compensated in real time, it will further amplify the global exposure error, causing hundreds or thousands of micron-sized exposure dark bands to appear in local areas of the wafer, directly threatening process stability and chip yield. Summary of the Invention

[0005] The purpose of this invention is to provide a high-precision exposure control system for multi-light source collaboration in lithography machines, so as to overcome the shortcomings in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a multi-light source collaborative high-precision exposure control system for a lithography machine, comprising:

[0007] Parameter acquisition module: acquires the initial energy parameters of each light source in the current exposure cycle of the lithography machine, including three basic features: light intensity, phase, and spot morphology;

[0008] The coupling construction module collects the energy change sequence and spot morphology change sequence of each light source at different exposure times t1 to tn to obtain the energy change-morphology coupling feature set A=(A1,A2,…,Ai,…,An), where Ai is the combination of energy deviation and spot distortion of the light source at exposure time ti, and the coupling feature set corresponding to the exposure time tm with the maximum distortion is marked as the maximum distortion feature Am;

[0009] Light source coordination error modeling module: Based on the initial energy parameters and the maximum distortion feature Am, a light source coordination error model is established. The model includes the time offset matrix between light sources and the energy compensation matrix of light sources.

[0010] Compensation parameter calculation module: Input the exposure disturbance obtained by the lithography machine in real time into the light source cooperative error model to obtain the target exposure compensation parameters, which include three types of control quantities used to correct the relative energy, relative phase and exposure timing between light sources;

[0011] Control execution module: Adjusts the exposure control unit of each light source according to the target exposure compensation parameters, including energy adjustment, phase synchronization and timing reconstruction operations.

[0012] Preferably, the method for obtaining the energy change-morphology coupled feature set A includes:

[0013] During the exposure scanning process, a unified clock signal triggers the energy detection component and the spot imaging component, enabling them to synchronously acquire the instantaneous energy value and spot morphology image of the corresponding light source at each exposure moment.

[0014] Two-dimensional morphology reconstruction processing is performed on the spot morphology images acquired at each exposure time to extract the spot center drift, spot edge diffusion and spot ellipticity parameters, forming the corresponding spot morphology feature subset;

[0015] The instantaneous energy value and the subset of spot morphology features are paired and fused according to the exposure time to construct a coupled feature unit containing energy deviation parameters and spot distortion parameters;

[0016] All coupled feature units are sorted and cumulatively analyzed according to the exposure time sequence to form the energy change-morphology change coupled feature set A.

[0017] Preferably, the establishment of the light source cooperative error model based on the initial energy parameters and the maximum distortion feature Am includes:

[0018] The light intensity, phase, and spot morphology features in the initial energy parameters are compared with the corresponding energy deviation and spot distortion parameters in the maximum distortion feature Am to form the basic vector of light source error.

[0019] Based on the gradient changes in intensity deviation and phase deviation in the fundamental vector of light source error, a linear recursive operation is used to generate the time sequence offset matrix between light sources;

[0020] Based on the spot distortion parameters and energy deviation in the light source error fundamental vector, perform a per-light source energy mapping operation to construct the light source energy compensation matrix;

[0021] The time offset matrix between light sources and the energy compensation matrix of light sources are solved together by matrix superposition to form a collaborative error model of light sources.

[0022] Preferably, the step of generating the time-series offset matrix between light sources using linear recursive operations includes:

[0023] The intensity deviation and phase deviation in the light source error base vector are calculated in the order of exposure time to form the difference components between adjacent time moments, thus forming a changing gradient sequence;

[0024] The changing gradient sequence is input into the linear recursive calculation process, and a time-by-time superposition operation is performed according to the preset recursive coefficients to obtain the recursive cumulative amount representing the cumulative trend of the light source temporal drift.

[0025] Based on the magnitude and direction of change of the recursive cumulative amount, a timing correction offset is assigned to each light source so that the timing offset between adjacent light sources has a linear mapping relationship.

[0026] All timing correction offsets are arranged into a matrix structure according to the light source number to generate a timing offset matrix between light sources, which is used to represent the overall timing mismatch between the exposure times of multiple light sources.

[0027] Preferably, the construction of the light source energy compensation matrix includes:

[0028] The light spot center drift, light spot edge diffusion, and light spot ellipticity in the light source error base vector are used as distortion feature inputs. Each distortion feature is weighted according to a preset distortion weight to form a comprehensive light spot distortion quantity.

[0029] Based on the correspondence between the overall beam distortion and the energy deviation, the light source energy correction coefficient is generated according to the linear mapping function;

[0030] Using the light source energy correction coefficient as the core parameter, a per-light source energy mapping operation is performed on the energy deviation of each light source to obtain the energy compensation value used to restore the expected energy output.

[0031] All energy compensation values ​​are arranged in matrix form according to the light source number, and the resulting matrix is ​​used as the light source energy compensation matrix to represent the energy correction relationship between multiple light sources.

[0032] Preferably, the step of jointly solving the time-series offset matrix between light sources and the energy compensation matrix of the light sources using a matrix superposition relationship to form a coordinated error model of the light sources includes:

[0033] Normalization is performed on the time offset matrix between light sources and the energy compensation matrix of light sources, respectively.

[0034] The normalized inter-source time offset matrix and the source energy compensation matrix are superimposed on corresponding elements to obtain the initial error superposition matrix;

[0035] Eigenvalue decomposition is performed on the initial error superposition matrix, and the dominant components of the light source error are determined by extracting the principal eigenvalues ​​and principal eigenvectors.

[0036] The joint error matrix is ​​reconstructed based on the principal eigenvectors, and the joint error matrix is ​​used as a light source cooperative error model to describe the comprehensive error generated by multiple light sources in the actual exposure process.

[0037] Preferably, the step of inputting the exposure perturbation obtained from real-time monitoring of the lithography machine into the light source collaborative error model to obtain the target exposure compensation parameters includes:

[0038] The exposure disturbance data obtained from real-time monitoring is preprocessed, and the disturbance data is divided into three disturbance subsets according to the classification of light intensity disturbance, phase disturbance and spot morphology disturbance.

[0039] Perform matrix multiplication on each perturbation subset and the corresponding error term of the light source cooperative error model to obtain the perturbation response matrix;

[0040] Numerical solution is performed on the disturbance response matrix, and the preliminary compensation values ​​of light intensity compensation, phase compensation and timing compensation are calculated by element-wise summation.

[0041] Based on the magnitude and calculation order of the initial compensation values, a quadratic optimization algorithm under constraints is used to generate the final target exposure compensation parameters.

[0042] Preferably, the step of generating the final target exposure compensation parameters using a constrained quadratic optimization algorithm includes:

[0043] The initial compensation values ​​are arranged in order according to the classification of light intensity compensation, phase compensation and timing compensation, and a compensation initial value vector is constructed as the initial input of the secondary optimization algorithm.

[0044] Multidimensional constraints are set based on the maximum allowable energy adjustment range, maximum phase synchronization offset, and maximum timing adjustment range of the lithography machine, so that each constraint restricts the range of values ​​of the optimization variables in the form of inequalities.

[0045] Based on the initial compensation vector and the aforementioned constraints, a quadratic objective function is constructed, with the objective function aiming to minimize the sum of squares of the compensation amounts.

[0046] The quadratic objective function is solved iteratively, and the optimal solution that satisfies all constraints is obtained through a stepwise convergence method. The optimal solution is then used as the final target exposure compensation parameter.

[0047] Preferably, the step of adjusting the exposure control unit of each light source according to the target exposure compensation parameters includes:

[0048] The light intensity compensation, phase compensation, and timing compensation in the target exposure compensation parameters are respectively input into the exposure control command generation process of the corresponding light source;

[0049] The amplitude adjustment calculation of the light source driving current is performed based on the light intensity compensation amount of each light source.

[0050] Based on the phase compensation amount, the offset command is calculated for the optical path phase adjustment component of the light source;

[0051] The timing offset is calculated based on the timing compensation amount for the light source pulse trigger time.

[0052] Preferably, the center coordinates of the light spot are calculated based on the two-dimensional light intensity value. The center coordinates of the light spot are weighted by the light intensity values ​​of all pixels, and the spatial coordinates corresponding to the pixels are weighted and summed. The weighted summation result is divided by the sum of all light intensity values ​​to obtain the numerical representation of the center position of the light spot. The difference between this value and the preset reference coordinate point is defined as the center drift of the light spot.

[0053] Subsequently, the edge diffusion of the light spot is extracted. During the extraction process, the light intensity threshold is first screened. The threshold is set to 50% of the maximum light intensity value of the image. All pixels greater than the threshold are marked as the main area of ​​the light spot. The width of the light spot is calculated along the main diameter direction and the secondary diameter direction of the main area of ​​the light spot. The obtained width value is the edge diffusion of the light spot.

[0054] For the ellipticity parameter of the light spot, the least squares fitting is performed on the pixel coordinates of the main region of the light spot to form an ellipse equation. The lengths of the major axis and minor axis in the ellipse equation are directly obtained from the fitting parameters. The ellipticity of the light spot is obtained by dividing the length of the major axis by the length of the minor axis.

[0055] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0056] 1. This invention constructs a fundamental vector of light source errors, a temporal offset matrix between light sources, and a light source energy compensation matrix. Based on real-time exposure perturbations, it performs multi-dimensional error coupling solutions, enabling unified modeling and coordinated compensation of light intensity errors, phase errors, and spot morphology distortions during exposure. This method can correct the energy output state, phase synchronization state, and exposure timing relationship of multiple light sources in real time within the exposure cycle, significantly improving the exposure consistency between multiple light sources and reducing the superposition errors caused by optical path drift, energy fluctuations, and timing mismatches. Especially under conditions of high numerical aperture, high scanning speed, and multiple exposures, this invention can capture and compensate for nanosecond-level timing offsets and sub-milliwatt-level energy errors, achieving exposure stability that is difficult to achieve with existing technologies.

[0057] 2. This invention establishes a quadratic optimization algorithm under constraints to achieve optimal solutions for light intensity compensation, phase compensation, and timing compensation while meeting hardware limits and exposure stability requirements. This enables adaptive adjustment of the exposure compensation strategy and minimizes error distribution. Through this optimization process, the compensation parameters not only avoid the accumulation of reverse errors caused by over-adjustment but also ensure that the energy, phase, and timing of multiple light source outputs achieve a coordinated and consistent compensation effect globally. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0059] Figure 1 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] For examples, please refer to Figure 1 As shown, the multi-light source collaborative high-precision exposure control system for the lithography machine described in this embodiment includes:

[0062] Parameter acquisition module: acquires the initial energy parameters of each light source in the current exposure cycle of the lithography machine, including three basic features: light intensity, phase, and spot morphology.

[0063] The parameter acquisition module of this invention is used to scan the parameters of all light sources to be exposed in the system before the lithography machine enters the exposure cycle, in order to obtain three basic characteristic values: light intensity, phase, and spot morphology. Since multiple light sources in advanced lithography systems are usually distributed in independent optical paths, and the light sources are superimposed into the final illumination field through complex optical elements, if the initial light field cannot be accurately quantified, the subsequent light source coordination error model will lack a reliable benchmark, thus causing the entire exposure compensation chain to fail. The parameter acquisition module includes the following sub-modules:

[0064] The light intensity detection unit consists of a high-speed photodiode array, a power sampling mirror, and an energy conversion amplifier. The power sampling mirror separates 0.1% to 0.3% of the energy of each light source, so that the detection does not affect the exposure of the main optical path. The sampling frequency of the photodiode array reaches 1 MHz, which can capture the fluctuation of the light source energy in the microsecond range. The signal is converted into a digital energy sequence by a low-noise current amplifier, which is used to form the initial light intensity parameter φ0(I).

[0065] Phase measurement unit: It consists of a micro-interference phase detector, a collimation module and a high-stability reference optical path; each light source is introduced into the interferometer arm through a beam splitter and superimposed with the reference beam to generate interference fringes; the fringe displacement is captured by a high-speed CMOS imaging chip and converted into a phase difference Δφ; the initial phase parameter φ0(φ) is obtained by repeating the calculation for each light source.

[0066] The spot morphology acquisition unit consists of an online spot imaging lens group, a spot CCD, and a morphology reconstruction processor. The imaging lens projects the actual spot of the light source onto the CCD in a scaled-down manner. The reconstruction processor performs a two-dimensional Fourier transform on the spot to extract the spot energy distribution, half-width at half-maximum (FWHM), ellipticity, and center drift, generating the initial spot morphology parameters φ0(S).

[0067] It should be noted that, firstly, the controller sequentially illuminates each light source, stabilizing its output beam for 5-10ms before entering the parameter acquisition window. Furthermore, the three sub-modules start acquisition under a unified clock synchronization, ensuring that the light intensity, phase, and spot morphology correspond to the same light source state. The synchronization error is less than 20ns to avoid modeling errors caused by time shifts between different physical characteristics. The acquisition module performs an automatic baseline calibration, including dark current compensation, background scattering subtraction, and optical path temperature drift correction, ensuring that the parameters accurately reflect the light source output, rather than being caused by environmental noise.

[0068] Finally, φ0 = {I0, φ0, S0} is obtained, which correspond to the three initial features of light intensity, phase, and spot morphology, respectively.

[0069] The coupling construction module collects the energy change sequence and spot morphology change sequence of each light source at different exposure times t1 to tn to obtain the energy change-morphology coupling feature set A=(A1,A2,…,Ai,…,An), where Ai is the combination of energy deviation and spot distortion of the light source at exposure time ti, and the coupling feature set corresponding to the exposure time tm with the maximum distortion is marked as the maximum distortion feature Am.

[0070] During exposure scanning, a unified clock signal is generated by the main timing control unit. This unified clock signal is transmitted along the internal timing circuit of the lithography machine in 10-nanosecond increments. Upon receiving this unified clock signal, the energy detection component causes the photoelectric conversion element to enter a short-time sampling window with a length of 5 microseconds. Within the entire sampling window, the energy output from the light source is continuously integrated to obtain the instantaneous energy value at the corresponding exposure moment.

[0071] Simultaneously, the spot imaging component initiates the exposure process under the trigger of the same unified clock signal, enabling the imaging sensor to capture an image of the spot shape within a 2-microsecond shutter time. This image is projected onto the photosensitive pixel matrix in scale by the imaging lens, and the light intensity received by different pixels is converted into a two-dimensional light intensity array, with each position in the array corresponding to a value.

[0072] To ensure that the energy data and the spot morphology image belong to the same light source output state, a unified clock signal not only triggers synchronous sampling of both but also marks the sampling number. Each trigger generates a number, which is written with the same number in both the energy data and the image data, facilitating pairing according to exposure time in subsequent steps.

[0073] The sampling times are set at equal intervals from t1 to tn, with the interval being a fixed value obtained by dividing the exposure path length by the number of exposure samplings, so that the sampling positions uniformly cover the entire exposure trajectory.

[0074] After acquisition, the light spot morphology image first enters the preprocessing process. During the preprocessing process, noise correction is performed on the two-dimensional light intensity array. Noise correction includes a dark level subtraction operation, which subtracts the dark level reference value from the light intensity value of each imaging pixel.

[0075] After preprocessing, the coordinates of the light spot center are calculated based on the two-dimensional light intensity values. The light spot center coordinates are obtained by weighting the spatial coordinates of each pixel using the light intensity values ​​as weights, and then dividing the weighted sum by the total light intensity values. The difference between this value and the preset reference coordinate point is defined as the light spot center drift.

[0076] The diffusion at the edge of the light spot is then extracted. The extraction process begins with a light intensity threshold screening, set to 50% of the maximum light intensity value in the image. All pixels exceeding this threshold are marked as the main area of ​​the light spot. The width of the light spot is calculated along both the primary and secondary diameter directions of the main area, and the resulting width value represents the diffusion at the edge of the light spot.

[0077] For the ellipticity parameter of the light spot, the least squares fitting is performed on the pixel coordinates of the main region of the light spot to form an elliptical equation. The lengths of the major axis and minor axis in the elliptical equation are directly obtained from the fitting parameters. The ellipticity of the light spot is obtained by dividing the length of the major axis by the length of the minor axis. The larger the value, the more obvious the distortion.

[0078] The amount of center drift, edge diffusion, and ellipticity of the light spot are combined in a fixed order to form a subset of light spot morphology features, which corresponds to a unique exposure time.

[0079] The instantaneous energy value is represented as a single numerical value after sampling. The difference between this value and the initial energy value is calculated to obtain the energy deviation. The energy deviation is obtained by subtracting the initial energy value from the instantaneous energy value, and the direction of deviation is indicated by positive or negative numbers.

[0080] The subset of spot morphology features and the energy deviation are mapped one-to-one according to the sampling number. The energy deviation is placed in the first position of the coupled feature unit, and the spot center drift, spot edge spread, and spot ellipticity are placed in the second to fourth positions in sequence, so that the four data items are arranged consecutively. These four data items constitute a single coupled feature unit, denoted as Ai, which represents the overall deviation state at the corresponding exposure time.

[0081] All coupled feature units acquired from exposure times t1 to tn are arranged in the exposure order to obtain the energy change-morphology coupled feature set A=(A1,A2,…,Ai,…,An), where Ai is the combination of energy deviation and spot distortion of the light source at exposure time ti. The sorted coupled feature units are then used in the cumulative analysis process.

[0082] The cumulative analysis process first performs a time-by-time accumulation operation on the energy deviation, so that the cumulative deviation at the k-th position is equal to the sum of the energy deviations from A1 to Ak. Then, the same accumulation operation is performed on the spot distortion parameters, so that the cumulative values ​​of spot center drift, spot edge diffusion, and spot ellipticity form a cumulative distortion sequence in sequence.

[0083] The cumulative results are used to determine the growth trend of distortion in the exposure path and to filter the exposure time tm with the maximum distortion. The exposure time with the maximum distortion is defined as the position where the cumulative distortion parameter reaches its maximum value among all time-consuming nodes. The coupled feature unit corresponding to this position is recorded as the maximum distortion feature, denoted as Am. This feature is used in subsequent error model construction steps, enabling the error model to use the state with the most severe distortion as the basis for modeling, thereby improving the model's adaptability to extreme perturbations.

[0084] Light source coordination error modeling module: Based on the initial energy parameters and the maximum distortion feature Am, a light source coordination error model is established. The model includes the time offset matrix between light sources and the energy compensation matrix of light sources.

[0085] After obtaining the initial energy parameters and maximum distortion features, the light intensity, phase, and spot morphology features in the initial energy parameters of the light source are compared with the energy deviation and spot distortion parameters in the maximum distortion features. The difference calculation is performed in a corresponding manner, where the light intensity error is obtained by subtracting the energy deviation value in the maximum distortion features from the initial light intensity value; the phase error is obtained by subtracting the phase shift value in the maximum distortion features from the initial phase value; and the spot center drift, spot edge spread, and spot ellipticity are obtained by subtracting the corresponding morphology distortion values ​​of the maximum distortion features from the initial morphology features.

[0086] The above difference calculation yields multiple error quantities, which are arranged in the order of light intensity error, phase error, and spot distortion-related error, forming the fundamental error vector of the light source. This fundamental error vector is used in subsequent steps of generating the time-series offset matrix and constructing the energy compensation matrix.

[0087] A temporal drift model is established based on the changing trends of light intensity deviation and phase deviation in the fundamental vector of light source error. To this end, differential operations are performed on the light intensity deviation and phase deviation in the order of exposure time. This differential operation is performed by subtracting the error of the previous time from the error of the next time step, resulting in a gradient sequence that reflects the rates of change of light intensity and phase.

[0088] The changing gradient sequence is then input into the linear recursive calculation process. The linear recursive calculation uses preset recursive coefficients to perform time-by-time recursive superposition operations. These recursive coefficients are fixed values, for example, between 0.1 and 0.9. Through recursion, the error accumulates over time to form a recursive cumulative amount. This recursive cumulative amount characterizes the temporal drift accumulation trend of the light source in the exposure path.

[0089] Based on the magnitude and direction of change of the recursive cumulative amount, a timing correction offset is assigned to each light source so that the offset can counteract the timing drift trend. The timing correction offset is calculated from the recursive cumulative amount using a linear mapping method, so that the timing offset between adjacent light sources has a linear correspondence.

[0090] Finally, the timing correction offsets of all light sources are arranged into a matrix structure according to the light source number, and this matrix is ​​used as the timing offset matrix between light sources to represent the overall timing mismatch between the exposure times of multiple light sources.

[0091] Based on the spot center drift, spot edge spread, and spot ellipticity in the fundamental vector of light source error, the three types of spot distortion parameters are first weighted according to preset distortion weights to form a comprehensive spot distortion value. The weighting operation is performed by summing the products of the distortion weights and the corresponding distortion parameters. The distortion weights are set according to the degree of influence of distortion on exposure quality; for example, the weight for spot center drift is higher than the weight for spot ellipticity.

[0092] Subsequently, based on the mapping relationship between the overall beam distortion and the energy deviation, a linear mapping function is used to generate the light source energy correction coefficient. The linear mapping function automatically outputs a higher energy correction coefficient when the beam distortion worsens, thus increasing the compensation amount with the degree of distortion.

[0093] Using the generated energy correction coefficient as the core parameter, a per-source energy mapping operation is performed on the energy deviation of each light source, mapping each energy deviation to an energy compensation value used to restore the expected energy output. The energy compensation value is calculated by multiplying the energy deviation by the energy correction coefficient.

[0094] Finally, all energy compensation values ​​are arranged in order of light source number to form a matrix structure, which serves as the light source energy compensation matrix to represent the energy correction relationship between multiple light sources.

[0095] The time-series offset matrix between light sources and the energy compensation matrix contain temporal and energy errors, respectively, and their numerical ranges differ. To avoid inconsistencies in dimensions during joint solution, normalization is first performed on both matrices. Normalization uses the largest absolute value in the matrix as a reference, dividing all matrix elements by this reference value to make all matrix elements zero to one.

[0096] After normalization, the two matrices are superimposed element-wise to obtain the initial error superposition matrix. This initial error superposition matrix simultaneously expresses both the temporal drift and energy deviation of the light source, providing a data foundation for the next step of feature extraction.

[0097] To identify the primary sources of error, eigenvalue decomposition is performed on the initial error superposition matrix. Eigenvalue decomposition extracts principal eigenvalues ​​and principal eigenvectors from the initial error superposition matrix, ensuring that the direction of the principal eigenvectors characterizes the main deformation patterns of error accumulation. The magnitude of the principal eigenvalues ​​is used to determine the error intensity.

[0098] Based on the direction of the principal eigenvector, the elements of the initial error superposition matrix are remapped into an error space based on the principal eigenvector, enabling the reconstructed matrix to comprehensively reflect the combined error coupling behavior among multiple light sources. This reconstructed matrix serves as a collaborative error model for light sources, expressing the dominant causes and coupling trends of multi-light source errors during actual exposure.

[0099] The following section provides a detailed explanation of the entire process of establishing a light source cooperative error model based on initial energy parameters and the maximum distortion feature Am, using specific exposure scenarios as examples.

[0100] For example, during a single exposure cycle, the initial energy parameters of multiple light sources are acquired, including initial light intensity, initial phase, spot center position, spot edge spread, and spot ellipticity. Assume that the initial light intensity of a certain light source is 25.0 mW, the initial phase is 1.2 radians, the spot center coordinates are (64, 64), the spot edge spread is 38 pixels, and the ellipticity is 1.10.

[0101] The maximum distortion feature Am was detected during the exposure process, with an energy deviation of -0.4 mW, a phase shift of 0.3 radians, a spot center drift of (2, 1.5), a spot edge spread shift of 8 pixels, and an ellipticity shift of 0.07.

[0102] Subtracting the initial parameters from the maximum distortion feature term by term yields the basic vector of light source errors. For example, the light intensity error is 25.0 - (25.0 - 0.4) forming 0.4 mW; the phase error is 1.2 - (1.2 + 0.3) forming -0.3 radians; the spot center drift error is -2 and -1.5; the spot edge diffusion error is -8; and the ellipticity error is -0.07. These error values ​​are arranged in a fixed order to form the basic vector of light source errors, which is used for subsequent error modeling.

[0103] The light intensity error and phase error in the fundamental vector of light source error are subjected to adjacent difference operations in the order of exposure time. The gradient value sequence is obtained by subtracting the error value of the previous time from the error value of the next time step, so that the gradient can reflect the rate of change of light intensity and the rate of change of phase.

[0104] For example, if the light intensity error between two adjacent moments is 0.25 mW and 0.40 mW, then the gradient is 0.15 mW. If the phase error is -0.20 and -0.30 between adjacent moments, then the gradient is -0.10 radians. The gradients of all exposure moments are arranged sequentially to form a gradient variation sequence.

[0105] The changing gradient sequence is input into the linear recursive calculation process. The linear recursion uses a preset recursion coefficient, such as 0.6. The recursive process multiplies the current gradient by the recursion coefficient and then adds it to the accumulated value from the previous time step to obtain the current accumulated value. The larger the accumulated value, the more pronounced the temporal drift trend.

[0106] Based on the positive or negative direction and magnitude of the recursive cumulative amount, a corresponding timing correction offset is calculated for each light source. For example, if the recursive cumulative amount is positive, the offset is a positive correction, and vice versa. The offset magnitude is calculated from the recursive cumulative amount in a linear proportion, i.e., offset = recursive cumulative amount multiplied by the timing mapping coefficient (e.g., 2 nanoseconds).

[0107] The timing correction offsets of all light sources are arranged into a matrix according to the light source number, so that it can represent the overall timing mismatch relationship between multiple light sources.

[0108] The calculation of the overall beam distortion includes: weighting the beam center drift, beam edge spread, and ellipticity according to preset distortion weights, for example: center drift weight: 0.5; edge spread weight: 0.3; ellipticity weight: 0.2; and summing the weighted values ​​to obtain the overall beam distortion. For example: overall beam distortion = 0.5 × 2 + 0.3 × 8 + 0.2 × 0.07; the entire calculation yields the overall distortion value.

[0109] Based on the linear mapping relationship between the overall beam distortion and the energy deviation, a correction coefficient is generated using a mapping function that increases the compensation coefficient as the distortion increases. For example, the correction coefficient = overall distortion multiplied by 0.1.

[0110] The energy deviation of each light source is multiplied by the correction coefficient to obtain the energy compensation value, which can restore the energy compensation value to the expected energy output range.

[0111] All energy compensation values ​​are arranged according to the light source number to form a light source energy compensation matrix, which can reflect the energy correction requirements between light sources.

[0112] The time offset matrix between light sources and the energy compensation matrix of light sources are normalized based on their respective maximum absolute values, so that all elements in the matrix are values ​​between zero and one.

[0113] Adding corresponding elements of the two normalized matrices yields the initial error superposition matrix. This matrix contains the coupling between energy bias and timing offset.

[0114] The initial error superposition matrix is ​​subjected to eigenvalue decomposition. By extracting the largest eigenvalue and its corresponding eigenvector, the dominant causes of the error and the direction of error growth are determined.

[0115] The initial error superposition matrix is ​​remapped based on the principal eigenvectors, so that the reconstructed matrix can centrally represent the coupling mode of multi-source errors. This reconstructed matrix is ​​used as the final source cooperative error model.

[0116] The compensation parameter calculation module inputs the exposure disturbances obtained from the real-time monitoring of the lithography machine into the light source cooperative error model to obtain the target exposure compensation parameters, which include three types of control quantities used to correct the relative energy, relative phase, and exposure timing between light sources.

[0117] During the exposure process, the lithography machine monitors changes in light intensity, phase fluctuations, and spot morphology disturbances in real time, and this monitoring data is recorded in the form of a continuous time series. In order to ensure that these disturbances accurately correspond to the error dimensions in the light source cooperative error model, a preprocessing step is required for the disturbance data.

[0118] During preprocessing, noise reduction is first performed on the monitoring data. This is achieved by collecting baseline noise before exposure and subtracting it point by point from the real-time perturbation sequence. The processed perturbation data is then categorized into intensity perturbation, phase perturbation, and spot morphology perturbation. Intensity perturbation corresponds to the instantaneous shift in the energy output of the light source, phase perturbation corresponds to the change in the coherence state of the optical path, and spot morphology perturbation includes the spot center drift, the spot edge spread, and the change in ellipticity.

[0119] After splitting, three perturbation subsets are formed, which can directly correspond to the three types of error inputs in the light source cooperative error model.

[0120] Each perturbation subset is input into the collaborative error model of the light source, and matrix multiplication is performed with the corresponding error term in the model. The matrix multiplication operation is performed by multiplying each value in the perturbation subset with the corresponding value in the error model and then summing the results, so that the resulting matrix can reflect the propagation of the perturbation in the error model.

[0121] For example, the matrix obtained by multiplying the subset of light intensity perturbations with the energy error term of the error model contains matrix elements representing the contribution of light intensity perturbations to exposure error. Similarly, the matrix products of phase perturbations and topography perturbations yield response matrices characterizing the propagation of phase error and topography error, respectively. These matrices are combined to form the perturbation response matrix, which is used to describe the cumulative effect of perturbations in the error model.

[0122] The disturbance response matrix is ​​numerically solved by summing element by element, so that the summed value can be used as the initial compensation value for light intensity compensation, phase compensation and timing compensation.

[0123] The element-wise summation step obtains three types of compensation quantities by summing all elements in each row and column of the matrix: light intensity compensation quantity corresponds to the matrix summation result related to energy, phase compensation quantity corresponds to the matrix summation result related to phase, and timing compensation quantity corresponds to the matrix summation result related to timing.

[0124] The above three types of compensation quantities are arranged in their calculation order to form a preliminary compensation value sequence, providing initial input values ​​for subsequent optimization calculations.

[0125] Based on the initial compensation value, a quadratic optimization algorithm under constraints is executed, specifically including the following steps:

[0126] The intensity compensation, phase compensation, and timing compensation are arranged in a fixed order to form an initial compensation vector. This initial compensation vector serves as the initial input to the quadratic optimization algorithm, used to determine the starting reference point for the optimization variables.

[0127] Based on the maximum energy adjustment range, maximum phase synchronization offset, and maximum timing adjustment range allowed by the lithography machine hardware, the allowable range for each type of compensation is set as upper and lower limits. These limits are expressed as inequality conditions, which are used to restrict the value range of the optimization variables. For example, the light intensity compensation amount must not exceed the preset maximum energy adjustment range, the phase compensation amount must not exceed the adjustment capability of the phase synchronization device, and the timing compensation amount must not exceed the maximum timing offset allowed by the light source driver. These inequalities are uniformly used as constraints for the quadratic optimization algorithm.

[0128] To achieve a balance between minimizing the compensation amount and maintaining exposure stability, a quadratic objective function is constructed based on the initial compensation vector and constraints. This objective function uses the sum of squares of the three types of compensation amounts as the optimization objective, causing the optimization algorithm to tend towards a smaller and more convergent compensation amount.

[0129] The quadratic objective function is formed by squaring the three types of compensation quantities and then adding them together, so that the minimum value of the objective function represents the optimal compensation strategy.

[0130] An iterative optimization process is performed on the quadratic objective function. The optimization solution adopts a stepwise convergence method. In each iteration, the compensation variable is adjusted according to the gradient direction of the objective function, so that the compensation variable gradually approaches the optimal solution that satisfies the constraints.

[0131] When the iteration results satisfy all inequality constraints and the objective function decreases below the convergence threshold, the optimization stops, and the final compensation amount is used as the target exposure compensation parameter.

[0132] The target exposure compensation parameters include three types of control quantities used to correct relative energy changes, relative phase changes, and exposure timing shifts between light sources, which are used to compensate for exposure errors in real time during actual exposure.

[0133] This invention, by jointly solving the real-time exposure perturbation and the light source cooperative error model, enables simultaneous identification and quantitative compensation of light intensity changes, phase drift, and spot morphology distortion during the exposure process. By constructing a fundamental vector of light source errors, this invention can express the coupling deviation of the light source in the dimensions of energy, phase, and spot morphology in a single vector form, giving error analysis a unified expressive capability and avoiding the error dispersion problem caused by existing methods that rely on processing multiple path error parameters separately.

[0134] This invention employs linear recursive calculations to generate a temporal offset matrix between light sources, allowing exposure timing drift to be fully represented in a cumulative trend manner. This enables the capture of sub-millisecond-scale temporal drift perturbations, thereby significantly improving timing correction accuracy. Simultaneously, by deriving energy correction coefficients through a comprehensive measure of spot distortion, a clear mathematical mapping relationship is established between spot morphology distortion and energy compensation, fundamentally solving the problem that traditional energy compensation methods cannot adaptively adjust the compensation amplitude according to the degree of distortion.

[0135] Control execution module: Adjusts the exposure control unit of each light source according to the target exposure compensation parameters, including energy adjustment, phase synchronization and timing reconstruction operations.

[0136] After obtaining the target exposure compensation parameters, the intensity compensation, phase compensation, and timing compensation values ​​from these parameters are input into the exposure control command generation process one by one according to the light source number. This command generation process uses the compensation values ​​as input variables and, based on the operating characteristics of the exposure control hardware, converts each compensation value into a numerical control command.

[0137] Among them, the light intensity compensation amount is used to generate the corresponding drive current amplitude adjustment command, so that the command expresses the magnitude by which the drive current should be increased or decreased in numerical form; the phase compensation amount is used to generate the optical path phase offset command, which uses the numerical value of changing the optical path length as the output value; the timing compensation amount is used to generate the trigger time offset command, so that the command expresses the degree to which the light source lighting action should be advanced or delayed in nanosecond or picosecond time offset.

[0138] Through the above instruction generation process, the three types of control variables in the target exposure compensation parameters are all converted into control instructions that can be directly executed by the light source driving device, laying the foundation for the implementation of subsequent adjustment actions.

[0139] After generating the light intensity adjustment control command, the amplitude adjustment of the light source drive current is calculated based on the light intensity compensation amount. The output energy of the light source has a monotonic relationship with the amplitude of the drive current; therefore, a linear proportional relationship can be used for current adjustment. Specifically, the light intensity compensation amount is divided by the light source energy-to-current response coefficient to obtain the adjustment range of the drive current. For example, when the light intensity compensation amount is positive, the adjustment range of the drive current is positive, indicating that the drive current needs to be increased; when the compensation amount is negative, the adjustment range is negative, indicating that the drive current needs to be decreased. By transmitting the adjusted drive current to the light source driver, the output energy of the light source reaches the energy level corresponding to the target light intensity compensation amount, thereby achieving precise control of the exposure energy.

[0140] Phase compensation is used to compensate for phase shifts in the optical path caused by temperature changes, optical element drift, or scanning vibrations. To ensure that the output phase of the light source matches the target phase compensation, the phase compensation is converted into a change in optical path length. Since the phase change is linearly proportional to the optical path length, the optical path length shift can be obtained by multiplying the phase compensation by the optical path conversion factor.

[0141] The obtained optical path length offset value is used as input to perform a phase adjustment operation, causing the adjustable reflector or adjustable refractive element in the optical path to move a corresponding small distance (e.g., tens to hundreds of nanometers) along the optical path direction. This physical displacement changes the effective length of the optical path, thereby changing the phase of the output light to match the target phase compensation amount, achieving precise phase synchronization control.

[0142] Multi-source coordinated exposure requires the light sources to be illuminated synchronously within a very short time. To ensure that the timing between the light sources meets the synchronization relationship after target compensation, the timing compensation amount is used as an input variable to calculate the offset of the light source pulse trigger time.

[0143] Timing offset calculation directly uses the timing compensation amount as the trigger time adjustment value, causing the light source driver to advance or delay the trigger action by a time offset on the order of nanoseconds or picoseconds when performing the pulse trigger action. For example, when the timing compensation amount is positive, the light source pulse trigger time is delayed by the corresponding value; when the timing compensation amount is negative, the trigger time is advanced by the corresponding value.

[0144] By fine-tuning the trigger time of the light source pulses as described above, multiple light sources can achieve higher-precision timing synchronization at the moment of exposure, reducing the exposure energy superposition error caused by inconsistent trigger times, thereby significantly improving the stability of collaborative exposure and the consistency of pattern forming.

[0145] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A high-precision exposure control system for multi-light source collaboration in a lithography machine, characterized in that: include: Parameter acquisition module: acquires the initial energy parameters of each light source in the current exposure cycle of the lithography machine, including three basic features: light intensity, phase, and spot morphology; The coupling construction module collects the energy change sequence and spot morphology change sequence of each light source at different exposure times t1 to tn to obtain the energy change-morphology coupling feature set A=(A1,A2,…,Ai,…,An), where Ai is the combination of energy deviation and spot distortion of the light source at exposure time ti, and the coupling feature set corresponding to the exposure time tm with the maximum distortion is marked as the maximum distortion feature Am; Light source coordination error modeling module: Based on the initial energy parameters and the maximum distortion feature Am, a light source coordination error model is established. The model includes the time offset matrix between light sources and the energy compensation matrix of light sources. Compensation parameter calculation module: Input the exposure disturbance obtained by the lithography machine in real time into the light source cooperative error model to obtain the target exposure compensation parameters, which include three types of control quantities used to correct the relative energy, relative phase and exposure timing between light sources; Control execution module: Adjusts the exposure control unit of each light source according to the target exposure compensation parameters, including energy adjustment, phase synchronization and timing reconstruction operations.

2. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 1, characterized in that: Methods for obtaining the energy change-morphology coupled feature set A include: During the exposure scanning process, a unified clock signal triggers the energy detection component and the spot imaging component, enabling them to synchronously acquire the instantaneous energy value and spot morphology image of the corresponding light source at each exposure moment. Two-dimensional morphology reconstruction processing is performed on the spot morphology images acquired at each exposure time to extract the spot center drift, spot edge diffusion and spot ellipticity parameters, forming the corresponding spot morphology feature subset; The instantaneous energy value and the subset of spot morphology features are paired and fused according to the exposure time to construct a coupled feature unit containing energy deviation parameters and spot distortion parameters; All coupled feature units are sorted and cumulatively analyzed according to the exposure time sequence to form the energy change-morphology change coupled feature set A.

3. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 1, characterized in that: The light source cooperative error model established based on the initial energy parameters and the maximum distortion feature Am includes: The light intensity, phase, and spot morphology features in the initial energy parameters are compared with the corresponding energy deviation and spot distortion parameters in the maximum distortion feature Am to form the basic vector of light source error. Based on the gradient changes in intensity deviation and phase deviation in the fundamental vector of light source error, a linear recursive operation is used to generate the time sequence offset matrix between light sources; Based on the spot distortion parameters and energy deviation in the light source error fundamental vector, perform a per-light source energy mapping operation to construct the light source energy compensation matrix; The time offset matrix between light sources and the energy compensation matrix of light sources are solved together by matrix superposition to form a collaborative error model of light sources.

4. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 3, characterized in that: The step of generating the time offset matrix between light sources using linear recursive calculation includes: The intensity deviation and phase deviation in the light source error base vector are calculated in the order of exposure time to form the difference components between adjacent time moments, thus forming a changing gradient sequence; The changing gradient sequence is input into the linear recursive calculation process, and a time-by-time superposition operation is performed according to the preset recursive coefficients to obtain the recursive cumulative amount representing the cumulative trend of the light source temporal drift. Based on the magnitude and direction of change of the recursive cumulative amount, a timing correction offset is assigned to each light source so that the timing offset between adjacent light sources has a linear mapping relationship. All timing correction offsets are arranged into a matrix structure according to the light source number to generate a timing offset matrix between light sources, which is used to represent the overall timing mismatch between the exposure times of multiple light sources.

5. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 3, characterized in that: The construction of the light source energy compensation matrix includes: The light spot center drift, light spot edge diffusion, and light spot ellipticity in the light source error base vector are used as distortion feature inputs. Each distortion feature is weighted according to a preset distortion weight to form a comprehensive light spot distortion quantity. Based on the correspondence between the overall beam distortion and the energy deviation, the light source energy correction coefficient is generated according to the linear mapping function; Using the light source energy correction coefficient as the core parameter, a per-light source energy mapping operation is performed on the energy deviation of each light source to obtain the energy compensation value used to restore the expected energy output. All energy compensation values ​​are arranged in matrix form according to the light source number, and the resulting matrix is ​​used as the light source energy compensation matrix to represent the energy correction relationship between multiple light sources.

6. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 3, characterized in that: The step of jointly solving the time-series offset matrix between light sources and the energy compensation matrix of the light sources using a matrix superposition relationship to form a collaborative error model of the light sources includes: Normalization is performed on the time offset matrix between light sources and the energy compensation matrix of light sources, respectively. The normalized inter-source time offset matrix and the source energy compensation matrix are superimposed on corresponding elements to obtain the initial error superposition matrix; Eigenvalue decomposition is performed on the initial error superposition matrix, and the dominant components of the light source error are determined by extracting the principal eigenvalues ​​and principal eigenvectors. The joint error matrix is ​​reconstructed based on the principal eigenvectors, and the joint error matrix is ​​used as a light source cooperative error model to describe the comprehensive error generated by multiple light sources in the actual exposure process.

7. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 1, characterized in that: The method of inputting exposure perturbation obtained from real-time monitoring of the lithography machine into the collaborative error model of the light source to obtain target exposure compensation parameters includes: The exposure disturbance data obtained from real-time monitoring is preprocessed, and the disturbance data is divided into three disturbance subsets according to the classification of light intensity disturbance, phase disturbance and spot morphology disturbance. Perform matrix multiplication on each perturbation subset and the corresponding error term of the light source cooperative error model to obtain the perturbation response matrix; Numerical solution is performed on the disturbance response matrix, and the preliminary compensation values ​​of light intensity compensation, phase compensation and timing compensation are calculated by element-wise summation. Based on the magnitude and calculation order of the initial compensation values, a quadratic optimization algorithm under constraints is used to generate the final target exposure compensation parameters.

8. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 7, characterized in that: The method of generating the final target exposure compensation parameters using a constrained quadratic optimization algorithm includes: The initial compensation values ​​are arranged in order according to the classification of light intensity compensation, phase compensation and timing compensation, and a compensation initial value vector is constructed as the initial input of the secondary optimization algorithm. Multidimensional constraints are set based on the maximum allowable energy adjustment range, maximum phase synchronization offset, and maximum timing adjustment range of the lithography machine, so that each constraint restricts the range of values ​​of the optimization variables in the form of inequalities. Based on the initial compensation vector and the aforementioned constraints, a quadratic objective function is constructed, with the objective function aiming to minimize the sum of squares of the compensation amounts. The quadratic objective function is solved iteratively, and the optimal solution that satisfies all constraints is obtained through a stepwise convergence method. The optimal solution is then used as the final target exposure compensation parameter.

9. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 1, characterized in that: The step of adjusting the exposure control unit of each light source according to the target exposure compensation parameters includes: The light intensity compensation, phase compensation, and timing compensation in the target exposure compensation parameters are respectively input into the exposure control command generation process of the corresponding light source; The amplitude adjustment calculation of the light source driving current is performed based on the light intensity compensation amount of each light source. Based on the phase compensation amount, the offset command is calculated for the optical path phase adjustment component of the light source; The timing offset is calculated based on the timing compensation amount for the light source pulse trigger time.

10. The multi-light source collaborative high-precision exposure control system for lithography machines according to claim 2, characterized in that: The center coordinates of the light spot are calculated based on the two-dimensional light intensity value. The center coordinates of the light spot are weighted by the light intensity values ​​of all pixels. The corresponding spatial coordinates of the pixels are weighted and summed. The weighted sum is divided by the sum of all light intensity values ​​to obtain the numerical representation of the center position of the light spot. The difference between this value and the preset reference coordinate point is defined as the center drift of the light spot. Subsequently, the edge diffusion of the light spot is extracted. During the extraction process, the light intensity threshold is first screened. The threshold is set to 50% of the maximum light intensity value of the image. All pixels greater than the threshold are marked as the main area of ​​the light spot. The width of the light spot is calculated along the main diameter direction and the secondary diameter direction of the main area of ​​the light spot. The obtained width value is the edge diffusion of the light spot. For the ellipticity parameter of the light spot, the least squares fitting is performed on the pixel coordinates of the main region of the light spot to form an ellipse equation. The lengths of the major axis and minor axis in the ellipse equation are directly obtained from the fitting parameters. The ellipticity of the light spot is obtained by dividing the length of the major axis by the length of the minor axis.

Images