A method and system for compensating and correcting mask plate in three-dimensional space in Z direction

Abstract

The application provides a Z-direction adjustable mask three-dimensional space compensation correction method and system, and relates to the technical field of semiconductor lithography process. The method comprises the following steps: obtaining initial imaging deviation sequences of each depth layer of a three-dimensional memory step structure; extracting corresponding axial position coordinates of each depth layer according to the initial imaging deviation sequences to obtain a depth layer imaging deviation distribution data set; calculating an axial deviation gradient rate of adjacent depth layer imaging deviations according to the depth layer imaging deviation distribution data set; and quantifying the axial deviation gradient rate layer by layer to obtain a depth direction imaging deviation gradient set. The application can compensate and correct the optical proximity effect in the three-dimensional space through a mask Z-direction partition dynamic adjustable mode, and improve the imaging uniformity of the word line contact window of each depth layer of the step structure.

Description

Technical Field

[0001] This invention relates to the field of semiconductor photolithography technology, and in particular to a Z-axis adjustable mask three-dimensional space compensation and correction method and system. Background Technology

[0002] As the number of stacked layers in 3D memory continues to evolve from over one hundred to over two hundred, the word line contact windows in the stepped structure need to be formed layer by layer at different depths through multiple photolithography-etching cycles. Since the total height of the stepped morphology stack can reach several micrometers, while the effective depth of focus of immersion lithography equipment is only within the range of hundreds of nanometers, there is a certain difference in the actual defocusing degree between the top and bottom regions, resulting in a gradually amplified distribution of photolithographic imaging deviations along the depth direction between layers at different depths.

[0003] Currently, the industry mainly uses traditional two-dimensional optical proximity effect correction, which improves imaging quality by adjusting the size of the pattern within the mask plane. Some also use mask effect analysis models based on three-dimensional electromagnetic field simulation, incorporating the influence of the depth direction into the compensation consideration. However, most of these techniques have the following drawbacks: traditional two-dimensional correction only acts on the mask plane dimension, does not distinguish the differences in imaging conditions between different depth layers, and is difficult to apply differentiated compensation to different depth positions; while three-dimensional simulation methods have the ability to analyze the depth dimension, they consume a lot of computational resources and are time-consuming, and are usually only suitable for pre-static compensation in the design stage. For example, in actual production line operation, when the photoresist thickness fluctuates or the etching parameters drift, the preset static compensation amount is difficult to dynamically adjust accordingly.

[0004] This results in insufficient consistency of the critical dimensions of the word line contact windows along the depth direction, which in turn manifests as a large dispersion of the contact resistance of word lines at different depths. This affects the uniformity of the read and write operation characteristics of the memory cells and restricts the yield level of products with higher stacking layers under mass production conditions. Summary of the Invention

[0005] This invention provides a Z-axis adjustable mask for compensation and correction in three-dimensional space. It can compensate and correct the optical proximity effect in three-dimensional space by dynamically adjusting the Z-axis partition of the mask, thereby improving the imaging uniformity of the word line contact window at each depth layer of the stepped structure.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: In a first aspect, a Z-axis adjustable mask 3D space compensation and correction method is provided, the method comprising: Step 1: Obtain the initial imaging deviation sequence of each depth layer of the three-dimensional memory stepped structure, and extract the axial position coordinates corresponding to each depth layer based on the initial imaging deviation sequence to obtain the depth layer imaging deviation distribution dataset. Step 2: Based on the depth layer imaging deviation distribution dataset, calculate the axial deviation rate of adjacent depth layer imaging deviations, and quantize the axial deviation rate layer by layer to obtain the depth direction imaging deviation gradient set. Step 3: Based on the depth direction imaging deviation gradient set, accumulate the axial deviation rate of each depth layer along the axial direction to derive the target compensation displacement corresponding to each mask compensation partition; bind the target compensation displacement to partition to obtain the Z-axis target compensation displacement matrix of each compensation partition of the mask. Step 4: Based on the Z-axis target compensation displacement matrix of each compensation zone of the mask, the physical drive stroke of each compensation zone is obtained, and a mask zone drive control instruction set is generated. Step 5: Based on the mask partition drive control instruction set, drive each compensation partition to perform differentiated physical displacement adjustment along the Z direction to obtain the three-dimensional spatial physical state of the mask after Z-direction partition compensation. Step 6: Based on the three-dimensional spatial physical state of the mask after Z-axis partition compensation, calculate the actual focal plane position corresponding to each depth layer to obtain the dynamic focal plane distribution map of the stepped structure. Step 7: Based on the dynamic focal plane distribution map of the stepped structure, perform differentiated exposure correction on each depth layer to obtain the three-dimensional spatial compensation correction result of the word line contact window.

[0007] Secondly, a Z-axis adjustable mask 3D spatial compensation and correction system includes: The imaging deviation distribution construction module is used to obtain the initial imaging deviation sequence of each depth layer of the three-dimensional memory ladder structure, extract the axial position coordinates corresponding to each depth layer, and obtain the depth layer imaging deviation distribution dataset. The deviation gradient calculation module is used to calculate the axial deviation rate of imaging deviation between adjacent depth layers based on the depth layer imaging deviation distribution dataset, and to quantize the axial deviation rate layer by layer to obtain the depth direction imaging deviation gradient set. The cumulative compensation matrix construction module is used to accumulate the axial deviation rate of each depth layer along the axial direction based on the depth direction imaging deviation gradient set, derive the target compensation displacement corresponding to each mask compensation partition, bind the target compensation displacement to partition, and obtain the Z-direction target compensation displacement matrix of each mask compensation partition. The drive instruction generation module is used to convert the physical drive stroke of each compensation partition into the Z-axis target compensation displacement matrix of each compensation partition of the mask, and generate a mask partition drive control instruction set. The partition drive execution module is used to drive each compensation partition to perform differentiated physical displacement adjustment along the Z direction according to the mask partition drive control instruction set, so as to obtain the three-dimensional physical state of the mask after Z-direction partition compensation. The focal plane distribution construction module is used to calculate the actual focal plane position corresponding to each depth layer based on the three-dimensional spatial physical state of the mask after Z-axis partitioning compensation, and obtain the dynamic focal plane distribution map of the stepped structure. The differential exposure correction module is used to perform differential exposure correction on each depth layer according to the dynamic focal plane distribution map of the stepped structure, so as to obtain the three-dimensional space compensation correction result of the word line contact window.

[0008] The above-described solution of the present invention has at least the following beneficial effects: By acquiring imaging deviation distribution data at each depth layer and calculating the axial deviation rate in the depth direction, the target compensation displacement is derived by accumulating the rate along the axial direction. No three-dimensional electromagnetic field simulation is required throughout the process, which reduces the dependence of compensation calculation on simulation models and computing resources, thereby improving the efficiency of obtaining compensation amount and making it easier to implement in a production line environment.

[0009] By partitioning, merging, and aggregating the compensation sequence accumulated along the axial direction, the compensation requirements of the continuous depth domain are bound to the discrete physical partitions of the mask, enabling different depth layers to obtain their own independent and differentiated compensation amounts, thus providing a path for optical proximity effect compensation correction in three-dimensional space.

[0010] By converting the target compensation displacement into physical drive stroke and generating a partition drive control instruction set, the compensation partitions of the drive mask are driven to perform differentiated displacement adjustments along the Z direction. Then, based on the physical state of the mask after displacement, the dynamic focal plane distribution map is calculated. The actual focal plane position of each depth layer is verified before exposure. Then, differentiated exposure correction is performed based on the dynamic focal plane distribution map, forming a closed-loop verification and correction link of drive execution, focal plane verification, and exposure correction. This helps to improve the uniformity of key dimensions of word line contact windows in each depth layer and the reliability of compensation correction results. Attached Figure Description

[0011] Figure 1 This is a schematic flowchart of a Z-axis adjustable mask three-dimensional space compensation and correction method provided by an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of a Z-axis adjustable mask three-dimensional space compensation and correction system provided by an embodiment of the present invention.

[0013] Figure 3 This is a schematic diagram of the process of generating a mask partition drive control instruction set by converting the Z-axis target compensation displacement matrix of each compensation partition of the mask into the physical drive stroke of each compensation partition. Detailed Implementation

[0014] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0015] like Figure 1 As shown, an embodiment of the present invention proposes a Z-axis adjustable mask three-dimensional space compensation and correction method, the method comprising the following steps: Step 1: Obtain the initial imaging deviation sequence of each depth layer of the three-dimensional memory stepped structure, and extract the axial position coordinates corresponding to each depth layer based on the initial imaging deviation sequence to obtain the depth layer imaging deviation distribution dataset. Step 2: Based on the depth layer imaging deviation distribution dataset, calculate the axial deviation rate of adjacent depth layer imaging deviations, and quantize the axial deviation rate layer by layer to obtain the depth direction imaging deviation gradient set. Step 3: Based on the depth direction imaging deviation gradient set, accumulate the axial deviation rate of each depth layer along the axial direction to derive the target compensation displacement corresponding to each mask compensation partition; bind the target compensation displacement to partition to obtain the Z-axis target compensation displacement matrix of each compensation partition of the mask. Step 4: Based on the Z-axis target compensation displacement matrix of each compensation zone of the mask, the physical drive stroke of each compensation zone is obtained, and a mask zone drive control instruction set is generated. Step 5: Based on the mask partition drive control instruction set, drive each compensation partition to perform differentiated physical displacement adjustment along the Z direction to obtain the three-dimensional spatial physical state of the mask after Z-direction partition compensation. Step 6: Based on the three-dimensional spatial physical state of the mask after Z-axis partition compensation, calculate the actual focal plane position corresponding to each depth layer to obtain the dynamic focal plane distribution map of the stepped structure. Step 7: Based on the dynamic focal plane distribution map of the stepped structure, perform differentiated exposure correction on each depth layer to obtain the three-dimensional spatial compensation correction result of the word line contact window.

[0016] In this embodiment of the invention, starting from the actual imaging deviation measurement data of each depth layer, the progressive features of the imaging deviation are extracted layer by layer along the depth direction, and then the compensation displacement required for each partition is obtained by axially accumulating the progressive features. This processing path replaces three-dimensional electromagnetic field simulation with measurement-driven deviation analysis, which reduces the occupation of complex simulation models and a large amount of computing resources in the compensation calculation link, and is conducive to quickly obtaining compensation solutions in the production line environment.

[0017] The binding of the compensation displacement amount with the physical partition of the mask allows regions located in different partitions within the same plane of the mask to obtain their own independent Z-axis position adjustment amount before exposure. After each partition is driven to its target position, the actual landing point of the focus at different depth layers of the stepped structure is no longer in the same plane, but is spatially distributed along the depth direction in a way that is inversely adapted to the degree of defocus of each layer, thus creating conditions for subsequent differential exposure of each depth layer.

[0018] After the drive is executed, the actual position of the current focal plane in each depth layer is deduced from the actual displacement state of each partition of the mask. Then, the exposure parameter correction amount corresponding to each layer is generated and the exposure is performed. This order of verification before exposure ensures that the focal plane information on which the exposure correction is based is the real state after the drive is executed rather than the preset value. This helps to reduce the compensation error caused by the deviation of the drive execution and improve the final uniformity of the key dimensions of the word line contact window in each depth layer.

[0019] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Perform imaging acquisition on each depth layer of the three-dimensional memory stepped structure to acquire measurement image data of the word line contact windows at each depth layer and the corresponding axial position identification information, obtaining raw measurement image data of each depth layer with position marks. Specifically, this includes: performing imaging acquisition on each depth layer of the three-dimensional memory stepped structure. The imaging acquisition equipment used is a critical dimension scanning electron microscope (CD-SEM), which can perform high-resolution imaging of the pattern at a specified location on the wafer surface and acquire the grayscale contour information of the pattern edge. During acquisition, the word line contact window area of ​​each depth layer is scanned sequentially according to the direction from the top layer to the bottom layer of the stepped structure. Before each scan, the positioning system of the wafer stage provides the position identification information of the current scan position in the axial direction of the stepped structure. This position identification information adopts a combination of layer number and axial coordinate encoding, that is, the layer number is used to represent the current scan position in the axial direction of the stepped structure. Layer numbering of depth layer Theoretical relationship between this layer and the wafer coordinate system Z Pack the coordinate values ​​into a single location metadata record (where For natural numbers, counting from the top level, =1 corresponds to the top depth layer); after scanning, the measurement image data of this depth layer is bound and stored with the corresponding axial position identification information, thereby obtaining the original measurement image data of each depth layer with position markings.

[0020] It is important to note that the correspondence between the layer number in the axial position identification information and the actual physical depth needs to be established beforehand by calibrating the cross-section of the stepped structure using transmission electron microscopy (TEM) or atomic force microscopy (AFM). The specific calibration process is as follows: Select a test wafer with the stepped structure etched but not yet filled with word line contact windows, prepare a cross-sectional sample along the longitudinal direction of the stepped structure, and use TEM to perform high-resolution imaging of this cross-section to obtain a continuous cross-sectional image containing all depth layers. From the TEM cross-sectional image, starting from the top surface, measure the actual physical distance of each layer relative to the top surface along the axial direction. This distance is the position coordinate of each depth layer. (Units are uniformly in nm); Simultaneously record the layer number of each depth layer in the manufacturing process. This establishes a system based on layer numbering. to physical location coordinates The discrete mapping relationship table; if AFM is used for calibration, the AFM probe is used to perform line scanning on the surface of each step of the stepped structure, and the height difference of each step relative to the top surface is measured to obtain the result. After calibration, the mapping table is stored in the measurement data processing unit. When step 1.1 is executed, the layer number obtained by CD-SEM can be quickly converted into the actual physical location coordinates of the layer at that depth by looking up the table. The mapping relationship should be periodically calibrated and updated with the periodic monitoring data of the process parameters (such as the deposition thickness of each layer) of the wafer batch to ensure the accuracy of the location coordinates.

[0021] Step 1.2: Based on the original measurement image data with position markers, extract the actual imaging feature parameters of the word line contact window at each depth layer to obtain the set of actual imaging feature parameters with position markers for each depth layer. Specifically, this includes: based on the obtained original measurement image data with position markers, extract the actual imaging feature parameters of the word line contact window for each depth layer measurement image. The actual imaging feature parameters refer to physical quantities that can characterize the imaging quality of the word line contact window. In this embodiment, at least the following are included: the key dimension value of the word line contact window in the X direction, the key dimension value in the Y direction, the ellipticity or roundness of the contact window outline, and the offset of the contact window edge position relative to the design position.

[0022] The specific extraction process is as follows: First, the grayscale image obtained by CD-SEM is preprocessed, including grayscale normalization and noise suppression, to eliminate the influence of uneven illumination and electronic noise on the edge determination accuracy during the imaging process; then, grayscale thresholding is performed on the preprocessed image: a threshold between the background grayscale and the graphic grayscale is selected, and pixels with grayscale values ​​higher than the threshold are determined to belong to the inner region of the contact window, and pixels with grayscale values ​​lower than the threshold are determined to belong to the outer region of the contact window, thereby obtaining the binary contour boundary of the contact window; the threshold selection can adopt the Otsu automatic thresholding method to maximize the inter-class variance between the background region and the graphic region, thereby obtaining the optimal segmentation threshold; after determining the edge contour of the contact window, the edge coordinates are extracted point by point along the contour. Specifically, with the centroid of the contour as the reference origin, each sampling point is extracted along the contour at a preset angle step (e.g., one sampling point every 1° or 2°). coordinate pairs, where The sampling point number represents the total number of samples collected. Each sampling point; based on the edge coordinate sequence, each imaging feature parameter is calculated item by item.

[0023] Ellipticity of the contact window outline and roundness The calculation method is as follows: for the extracted set of edge coordinate points, the least squares ellipse fitting method is used to determine an ellipse that best approximates the contour, and the length of the major semi-axis of the ellipse is obtained. and the length of the short half-axis (Units are all in nm); Ellipticity Defined as the ratio of the difference between the major and minor semi-axis to the major semi-axis, i.e. , It is a dimensionless quantity, with a value range between 0 and 1. The closer the outline is to 0, the closer it is to a perfect circle. The closer to 1, the flatter the outline; roundness Defined as the normalized relation between the area enclosed by the contour and its perimeter, i.e. ,in The area of ​​the region enclosed by the outline edge (unit) ), The perimeter of the outline edge (unit: ), It is a dimensionless quantity, with a value range between 0 and 1. The closer the value is to 1, the closer the outline is to a perfect circle, and the larger the area. The perimeter can be calculated from the edge coordinate sequence using the polygon face accumulation formula; It can be obtained by accumulating the Euclidean distance between adjacent sampling points segment by segment.

[0024] Among them, when the contact window contour is severely distorted, locally broken, or measurement noise causes the calculated result of the fitted ellipse or roundness to exceed the normal range of [0,1] (for example) If the calculated value is less than 0 or greater than 1, the value should be truncated to [0, 1]. , The nearest boundary value of 1] is set to 0 when it is less than 0 and 1 when it is greater than 1, and the data point is marked as an outlier. When synthesizing the feature deviation in the subsequent process, the ellipticity or roundness component corresponding to the outlier can be removed or given a lower weight coefficient to ensure the robustness of the feature deviation extraction.

[0025] The key dimension value in the X direction is defined as follows: In the binary contour of the contact window, scanning from left to right along the X-axis, the pixel span length between the left and right edges of the window is converted into a physical length value (unit: physical length) using a pixel size calibration coefficient. The key dimension values ​​in the Y direction are defined similarly. The physical span between the lower and upper edges of the scanned window is determined along the Y-axis. The offset of the contact window edge position relative to the design position is defined as follows: The extracted window edge coordinates are matched point-by-point with the design edge coordinates of the window at that depth layer in the mask design data. The differences between the actual centroid position and the design centroid position in the X and Y directions are calculated. These two differences are the X-axis offset and the Y-axis offset, respectively (unit: X). After extraction, the actual imaging feature parameters are saved together with the original axial position identification information to obtain the actual imaging feature parameter set with position markings for each depth layer; each record in this feature parameter set contains the layer number of that depth layer. Axial position coordinates The values ​​of the above-mentioned characteristic parameters.

[0026] Step 1.3: Based on the actual imaging feature parameter set with position markers, compare the actual imaging feature parameters of each depth layer with the corresponding design target feature parameters of each depth layer layer by layer to obtain the layer-by-layer comparison difference results of each depth layer; based on the layer-by-layer comparison difference results, extract the feature deviation amount of each depth layer to obtain the initial imaging deviation value with position markers of each depth layer. Specifically, the design target feature parameters are derived from the design data of the mask, that is, the ideal key dimensions, ideal contour shapes and ideal edge positions of the contact windows of the word lines of each depth layer in the design graphic; before formal mass production, the graphic feature parameters corresponding to each depth layer in the mask design data can be pre-extracted and stored as a design target feature parameter library for reference during comparison.

[0027] The layer-by-layer comparison operation method is as follows: for each depth layer ( i =1 , 2 M), by subtracting each item in the actual imaging feature parameters of the layer from the corresponding item in the design target feature parameters of the layer, a deviation vector is obtained, which is the layer-by-layer comparison difference result of the depth layer; then, based on the layer-by-layer comparison difference result, the feature deviation is extracted as the initial imaging deviation value. The characteristic deviation can be a weighted composite value of the components in the deviation vector, or the component with the most significant impact on imaging quality (such as critical size deviation) can be directly selected as the representative value. In this embodiment, the initial imaging deviation value is... The critical dimension deviation (the larger of the critical dimension deviations in the X and Y directions, or the root mean square value of both) is preferentially selected as representative, in nm, while other sub-deviation data are retained for use in subsequent steps when necessary; initial imaging deviation values ​​for each depth layer. By maintaining association with its axial position identifier, initial imaging deviation values ​​with position markers are obtained for each depth layer.

[0028] As a supplement, in terms of ellipticity Roundness When weighting and synthesizing dimensionless characteristic deviations and other characteristic deviations with physical dimensions, such as critical dimension deviations (in nm), each sub-item deviation should be normalized beforehand. The specific normalization method is as follows: for each sub-item deviation, divide its value by its corresponding process tolerance range (or historical statistical standard deviation) to convert each sub-item into a dimensionless normalized deviation value, and then perform weighted summation. After normalization, the interference of different dimensions and different numerical scales on the synthesis results is eliminated, ensuring the physical rationality of the weighted synthesis.

[0029] Step 1.4: Based on the initial imaging deviation values ​​with position markers, read the axial position identification information associated with each deviation value, convert the axial position identification information into position coordinates of each depth layer in the axial direction of the stepped structure, and associate the initial imaging deviation value of each depth layer with the converted position coordinates to obtain a depth layer imaging deviation distribution dataset, specifically including: the layer number in the axial position identification information. It needs to be converted into the actual physical position coordinates of this depth layer in the axial direction of the stepped structure. The mapping relationship used for the conversion has been pre-established in step 1.1 through TEM or AFM section calibration and stored in the measurement data processing unit in the form of a layer number-position coordinate mapping table; during the actual conversion, the layer number in each record is used as the basis for the conversion. Use the index to find the corresponding entry in the mapping table. The value represents the conversion from location identifier to location coordinates.

[0030] After completing the position coordinate transformation, each depth layer ( Initial imaging deviation value (=1 to M) The position coordinates obtained after transformation To establish a one-to-one correspondence, that is... The data is organized in a specific format; the offset-coordinate data of all M depth layers are arranged according to... The data, arranged in order from 1 to M, constitutes a depth layer imaging deviation distribution dataset. This dataset provides basic distribution information on the magnitude of imaging deviation at various locations along the depth direction for subsequent steps. It should be noted that, due to the different physical locations and defocusing conditions of each depth layer within the stepped structure, this distribution dataset contains... The overall distribution trend is usually as follows: near the top layer area Smaller values ​​(absolute values), and closer to the bottom layer region A large value (absolute value) indicates that the imaging deviation gradually increases along the depth direction. This distribution characteristic is the basis for calculating the axial deviation rate in step 2 to quantify the change in depth deviation.

[0031] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1: Based on the depth layer imaging deviation distribution dataset, sort the position coordinates corresponding to each depth layer along the axial direction to obtain the deviation-coordinate correspondence sequence sorted along the axial direction; extract the initial imaging deviation values ​​arranged in order from the deviation-coordinate correspondence sequence sorted along the axial direction to obtain the imaging deviation sequence arranged by the depth direction. Specifically, this includes: based on the obtained depth layer imaging deviation distribution dataset, sorting the position coordinates corresponding to each depth layer... Sort along the axial direction (from top to bottom); since the dataset output in step 1.4 is already numbered by layer. From 1 to The natural order of arrangement, and the position coordinates Follow Monotonically increasing ( Therefore, the order of each depth layer remains unchanged after sorting; the sorting operation mainly serves as a verification and confirmation function. After sorting, a deviation-coordinate correspondence sequence is obtained along the axial direction. The first [layer] in this sequence... The record is From this sequence, the initial imaging bias values ​​for each depth layer are extracted. Arranged in order as follows This yields an imaging bias sequence arranged along the depth direction (from top to bottom).

[0032] Step 2.2: Based on the imaging deviation sequence arranged along the depth direction, group two adjacent depth layers in the sequence into a set of adjacent depth layer pairs; extract the imaging deviation values ​​of each depth layer in each pair of adjacent depth layers in the set, and simultaneously extract the associated position coordinates of each depth layer; determine the axial distance between the two depth layers based on the position coordinates; calculate the variation range of the imaging deviation values ​​along the axial distance in each group to obtain the axial deviation rate of each group; collect the axial deviation rates of each group to obtain the set of axial deviation rates between adjacent layers, specifically including: based on the imaging deviation sequence arranged along the depth direction obtained in Step 2.1. The two depth layers that are adjacent in position in the sequence are grouped together; the grouping method is the first... layer( ) and the layer( ) is classified as Group 1, No. layer( ) and the layer( ) is assigned to group 2, and so on, resulting in a total of The data is divided into groups; each group constitutes a set of paired groups for adjacent depth layers; for each paired group of adjacent depth layers, the imaging deviation values ​​of the two depth layers in that group are extracted. and Simultaneously extract the associated position coordinates of the two depth layers. and The axial spacing between the two depth layers is determined based on their respective associated position coordinates. ,in (Unit is nm, sorted from top to bottom) ).

[0033] The variation range of the imaging deviation value along the axial spacing of each group is calculated to obtain the corresponding axial deviation rate of each group. ; Axial deviation rate The calculation formula is as follows:

[0034] In the formula Values ​​range from 1 to (Total) (Groups) For the first Initial imaging deviation value of the depth layer (unit: nm); For the first Initial imaging deviation value of the depth layer (unit: nm); and The first and the Axial position coordinates of the depth layer (unit: nm); As a dimensionless quantity, its absolute value characterizes the degree of drastic change in imaging deviation with depth within the span between two adjacent layers: the larger the absolute value, the more drastic the deviation change within that range, and the more urgent the compensation correction required. The sign of the image reflects the trend of image deviation along the depth direction: A positive value indicates that the deviation increases from the upper layer to the lower layer. A negative value indicates that the deviation decreases from the upper layer to the lower layer; for adjacent depth layers paired and grouped into sets... The above calculation operation is performed on each group one by one, generating the corresponding axial deviation rate of change for each group. Collect all After obtaining the value, the set of axial deviation rate variation between adjacent layers is obtained { }

[0035] Step 2.3: Based on the set of axial deviation gradient rates between adjacent layers, allocate the corresponding axial deviation gradient rates of each group to the depth layers associated with each group, so that each depth layer has a corresponding axial deviation gradient rate value, thus obtaining the depth direction imaging deviation gradient set. Specifically, this includes: based on the set of axial deviation gradient rates between adjacent layers obtained in Step 2.2, allocating the corresponding axial deviation gradient rates of each group to the depth layers associated with each group, so that each depth layer ( to Each of them has a corresponding axial deviation rate of change value. The basic rule for allocation is that for the first... Layer and First The rate of change corresponding to the grouping of layers The two layers associated with it are the first... Layer and first The rate of change is shared by these two layers; for the depth layer located in the middle ( to Since each layer belongs to both the preceding and following groups, the rate of change value of that layer... Take the arithmetic mean of the two sets of rate of change, i.e. .

[0036] For the top-level depth layer ( ), which belongs to only the first group, therefore For the deepest layers ( ), which belongs only to the first There are groups, therefore The above-mentioned method of averaging the gradient rates across multiple layers can smooth out local jumps in the gradient rates caused by measurement noise in individual layers, making the overall gradient distribution more continuous and stable. Even if the interlayer spacing between certain depth layers in the stepped structure is not completely equal in all locations (for example, the single-layer thickness in the top and bottom regions may differ slightly due to process variations), the above allocation rule still applies because the gradient rate... It has already passed Normalization was performed, and the differences in interlayer spacing have been addressed. The axial deviation rate values ​​for all depth layers were eliminated during the calculation; after allocation, the axial deviation rate values ​​for all depth layers were... Arranged in depth order, forming a set of depth-direction imaging bias gradients { Each of the sets The local deviation rate of change at this depth layer is the basic unit for the axial accumulation calculation in step 3.

[0037] As a supplement, the above The mean distribution method and the direct use of inter-layer variation rate There is an intentional trade-off between performing accumulation; if directly using... To accumulate, that is Then the cumulative result will exactly satisfy (when (Time); the path is mathematically precise, but it is extremely sensitive to single-layer measurement noise: any layer of Measurement errors will be directly transmitted to the rate of change of the corresponding interval of that layer. Then, by accumulating and amplifying, the effect is amplified on all subsequent layers. Value; the value used in this embodiment The mean smoothing strategy disperses the impact of single-layer measurement noise between adjacent layers by arithmetically averaging the rates of change of two consecutive sets, thereby introducing an order of magnitude... The local approximation error (where At the cost of interlayer spacing (typically on the order of tens of nanometers), robustness of the overall gradient sequence to measurement noise is achieved. Under typical conditions where the interlayer spacing in the stepped structure is sufficiently small and the imaging deviation exhibits a smooth, monotonically varying trend along the depth direction, this approximation error is far smaller than the measurement noise level. Therefore, it is used... It is a more stable choice in engineering.

[0038] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Based on the depth-direction imaging deviation gradient set, starting from the depth layer at one end of the stepped structure's axial direction, accumulate the axial deviation rate of each depth layer along the axial direction to obtain the axial cumulative compensation sequence. Specifically, this includes: based on the obtained depth-direction imaging deviation gradient set... The depth layer at one end of the stepped structure (in this embodiment, the top depth layer) Starting from the first point, the axial deviation rate of each depth layer is accumulated layer by layer along the axial direction (from the top layer to the bottom layer). The axial cumulative compensation amount corresponding to each depth layer is obtained. ; Accumulated compensation amount of the starting depth layer Setting it to 0 indicates that the layer corresponds to the nominal focal plane position and requires no additional compensation.

[0039] for The depth of the layer, the cumulative compensation amount In the formula For the first Cumulative compensation amount for depth layers; For the first Layer and First Axial spacing between layers (unit: nm); the physical meaning of the above recursive formula is the first... The cumulative compensation of a layer equals the cumulative compensation of the previous layer plus the contribution of the current layer's rate of change within the current interlayer spacing; rate of change. Multiplying the dimensionless rate of change by the interlayer spacing converts it into a compensation increment with length dimensions. This increment is then accumulated layer by layer to obtain the total cumulative compensation required for each depth layer. The cumulative compensation value for each depth layer is calculated using the above recursive formula. according to From 1 to The sequential arrangement yields the axial cumulative compensation sequence { }; each of the sequences It represents the sum from the starting point to the th The cumulative effect of the interlayer deviation variation in the compensation domain has the same physical dimension as the displacement dimension (nm), which can directly guide the Z-axis displacement adjustment of the mask.

[0040] Furthermore, regarding Mean smoothing pair The impact of cumulative results on accuracy can be quantified and evaluated as follows. With the true rate of change of the interval The local deviation between them is approximately The compensation increment error introduced in single-step accumulation is approximately Since the variation of imaging deviation along the depth direction is smooth and continuous, the difference in the rate of change between adjacent layers... Typically much smaller than the absolute value of the rate of change itself, and the interlayer spacing The error is only on the order of tens of nanometers, so the single-step error is usually on the sub-nanometer order; After layer accumulation, the total amount of accumulated error is on the order of... In the worst-case scenario, it is approximately one-step error. However, because the signs of the errors at each step tend to cancel each other out rather than accumulate unidirectionally under a smooth distribution, the actual cumulative error is much smaller than the theoretical worst-case value; in summary, The cumulative error introduced by smoothing is in the nanometer or sub-nanometer range under typical operating conditions, which is significantly lower than that caused by measurement noise and process fluctuations. The resulting uncertainty necessitates adjustments to the target displacement for subsequent partition compensation. The accuracy of the calculation does not have a substantial impact.

[0041] In the above accumulation formula, Assigned to the first in step 2.3 The axial deviation rate of change of depth layers, in its physical sense, is an estimate of the local rate of change of imaging deviation in the depth direction, i.e., the rate of change between adjacent layers. and To estimate the mean of the first The instantaneous rate of change at the layer; given a sufficiently small interlayer spacing, multiplying this local rate of change estimate by the interlayer spacing to approximate the deviation increment within that small interval is the standard approximation method for infinitesimal element accumulation; simultaneously at the accumulation starting point Under this setting, the entire accumulation chain The cumulative result is approximately equal to the first Initial imaging bias value of depth layer (Right now Because of the top floor (usually close to 0), therefore the cumulative compensation amount The physical dimensions and numerical range are consistent with the displacement required for direct compensation, and can be directly used as the target reference quantity for subsequent zonal compensation.

[0042] Step 3.2: Based on the axial cumulative compensation sequence, compare the differences in cumulative compensation between adjacent depth layers, merge adjacent depth layers with differences within a preset range into the same compensation partition, determine the depth layer range included in each compensation partition, and obtain the mask compensation partitioning scheme. Specifically, this includes: based on the obtained axial cumulative compensation sequence... The method compares the degree of difference in cumulative compensation between adjacent depth layers, groups adjacent depth layers with differences within a preset range into the same compensation partition, and determines the range of depth layers included in each compensation partition; the degree of difference is measured by the absolute value of the difference in cumulative compensation between two adjacent layers, i.e. Preset range The determination method is as follows: based on the depth-of-focus window value of the current photolithography process. (Unit: nm) and the reduction magnification Ώ of the projection objective (dimensionless, typically ranging from 4 to 5), are determined by the formula... The calculation shows that, among which This is the allowable utilization coefficient of the depth-of-focus window (dimensionless, recommended value range is 0.10 to 0.25).

[0043] The physical basis of this formula is: Z-axis displacement of the mask plate side. After the magnification is transferred through the projection lens, the focal plane offset caused on the wafer side is approximately The focal plane offset is limited to not exceeding the depth of field window. Permissible ratio This ensures that the residual focal plane deviation of each depth layer belonging to the same partition is within the process-tolerable range after unified compensation; as a specific example, if , , ,but (Mask side); The value can be adjusted according to the actual width of the process window: the wider the process window, the better. The larger the possible value, the fewer the number of partitions; the narrower the process window. The smaller the value, the more partitions are needed.

[0044] The specific execution rules for partition merging are as follows: starting from the top level ( Starting from this point, the first layer is incorporated into the first partition (marked as partition). Record the starting level index of this partition. Traverse downwards layer by layer, for the current layer and the next layer ( From 1 to ),like Then the first The layer is assigned to the current partition; if Then in the first Layer and First Divide the layers into partition boundaries and record the index of the terminating layer of the current partition. and open a new partition ( Incremental, new partition starting level index ); after traversing all After layering, we obtain One compensation zone ( ), each partition ( It has a clearly defined depth range. The number of depth layers within a partition The depth range information of all partitions is summarized to obtain the mask compensation partitioning scheme.

[0045] Step 3.3: Based on the mask compensation partitioning scheme and the axial cumulative compensation sequence, for each compensation partition, the cumulative compensation amounts corresponding to each depth layer within the partition are aggregated to obtain the partition aggregated compensation amount; the partition aggregated compensation amount is used as the target compensation displacement amount for the corresponding partition to obtain the partition target compensation displacement amount for each compensation partition. Specifically, this includes: based on the obtained mask compensation partitioning scheme and the obtained axial cumulative compensation sequence... For each compensation partition For each depth layer contained in this partition (from the 1st to the 2nd depth layer) Layer to the first The cumulative compensation amount corresponding to the layer) Perform aggregation processing to obtain the partition aggregation compensation amount, and use the aggregation result as the partition's compensation amount. Target compensation displacement Aggregation processing can employ an arithmetic average strategy, the calculation formula of which is as follows: In the formula For the first The number of depth layers contained within a partition; and The first The partition contains a start layer index and an end layer index; to The first one obtained in step 3.1 The cumulative compensation value of each depth layer within the partition; the physical meaning of the arithmetic mean strategy is to use the arithmetic mean of the cumulative compensation of each layer within the partition as the unified compensation representative value of the partition, so that the residual deviation of each layer from its own optimal focal plane after unified compensation is minimized overall.

[0046] Besides the arithmetic mean, other aggregation strategies can be selected to adapt to different process requirements. Weighted average strategy: The weight coefficients of each layer are set according to the functional importance of each depth layer within the partition in the stepped structure (each weight coefficient corresponds to the [number]th [layer] in that partition). Layer to the first Each layer in the hierarchy has a weight coefficient summed to 1. The weight values ​​can be calibrated using electrical test data on the impact of word line contact resistance on the read / write performance of memory cells. After calibration... Median strategy: Take all values ​​within the partition. The median value is used as This method is suitable for robust compensation scenarios where there are abnormal measurement values ​​in individual layers within a partition. This embodiment uses the arithmetic mean aggregation strategy by default.

[0047] The specific calibration method for the weighting coefficients of each depth layer in the aforementioned weighted average strategy is as follows: First, select a batch of test wafers that have completed all process flows, and calibrate the weighting coefficients of each depth layer ( Contact resistance was measured at the contact windows of each word line, and the contact resistance values ​​at each depth layer were obtained using the four-probe method or the Kelvin structure method. Further calculations of all The statistical dispersion index of the contact resistance value at each depth layer can be selected by the deviation of the resistance value of each layer from the mean resistance of all layers. absolute deviation or the standard deviation of the resistance values ​​of each layer (Based on statistics of multi-point measurements within the same depth layer); the greater the dispersion of each layer, the more significant the contribution of the imaging deviation of that layer to the fluctuation of the device's electrical performance, and the higher the weight should be given in the compensation.

[0048] Finally, the dispersion indices of each depth layer are summed and normalized: Let ;in, For the first j Within the partition, starting from the initial layer To the termination layer Standard deviation of resistance at each depth layer; denominator Indicates the first j The sum of the standard deviations of resistance across all depth layers within the partition; ensuring the sum of the weights of all layers within the partition is 1; after calibration, the weight coefficients... ( and that is Under a specific index value, all three share the same base character, and their meaning is always the weight coefficient. The subscript only indicates which depth layer the weight is. The weight is stored in the compensation calculation unit along with the process formula and can be called during each batch of mass production. If the subsequent process conditions change, such as changing the photoresist type or adjusting the deposition process parameters, the weight coefficient should be recalibrated to ensure that it reflects the current electrical characteristics of the device.

[0049] The partition aggregation compensation value obtained after aggregation processing is used as the corresponding partition. Target compensation displacement The target compensation displacement of each compensation zone is obtained. Finally, the partition identifier (i.e., partition number) of each compensation partition will be determined. ) and the corresponding target compensation displacement Organize in a matrix manner and build a The correspondence matrix with 2 rows and 2 columns: the matrix's first... The first column of the row is the partition identifier. The second column represents the target compensation displacement for this partition. This yields the Z-axis target compensation displacement matrix for each compensation zone of the mask plate; this matrix provides structured connection data for the subsequent step 4, which transitions from the compensation calculation domain to the drive execution domain.

[0050] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Based on the Z-axis target compensation displacement matrix of each compensation partition of the mask, read the target compensation displacement associated with the partition identifier partition by partition to obtain the target displacement set of the partitions to be converted. Specifically, this includes: based on the Z-axis target compensation displacement matrix of each compensation partition of the mask obtained in Step 3.3, reading the partition identifier of each partition in the matrix partition by partition. Associated target compensation displacement Read operations are performed according to the partition identifier. From 1 to The process is executed sequentially, extracting the partition identifier and corresponding target compensation displacement from each record to obtain the set of target displacements for the partitions to be converted. .

[0051] Step 4.2: Based on the target displacement set of the partitions to be converted, extract the corresponding drive mechanism calibration parameters for each compensation partition; based on the drive mechanism calibration parameters, convert the partition target displacement into physical drive stroke values ​​to obtain the physical drive stroke parameter set for each compensation partition. Specifically, this includes: based on the obtained target displacement set of the partitions to be converted, extract the corresponding drive mechanism calibration parameters for each compensation partition. Extract the calibration parameters of the drive mechanism configured for the corresponding partition; the calibration parameters of the drive mechanism refer to the conversion relationship parameters between the input control quantity and the actual output displacement quantity of the Z-axis drive actuator of each partition of the mask; in this embodiment, the Z-axis drive actuator of each partition adopts a piezoelectric ceramic actuator (or MEMS actuator), whose input is the drive voltage and output is the Z-axis physical displacement of the mask of that partition.

[0052] The calibration parameters for the drive mechanism are obtained as follows: Before the mask is put into use on the production line, the drive actuators of each zone are calibrated offline; the calibration process is as follows: for the first... The partition's drive actuator applies a set of known incremental drive voltage values, totaling... The calibration sampling point, at the _ , on the _ Apply driving voltage value to each sampling point ( (Unit: millivolts or volts), and simultaneously, a laser interferometer or a capacitive displacement sensor is used to measure the actual Z-axis displacement of the mask surface in this partition in real time, and record it as... (Unit: nm); thus obtained Group input-output discrete corresponding data A linear fit is performed on this correspondence, and the linear fit model is as follows: The least squares method is used to solve for the fitting coefficients. and :

[0053] In the formula Indicates to From 1 to Summation operation: For all The sum of all driving voltage values For all The sum of all measured displacement values It is the sum of the products of each voltage value and its corresponding displacement value; The sum of the squares of all voltage values, zero-point offset ; The unit is nm / mV (or nm / V), and its typical value range is 0.5 nm / mV to 20 nm / mV (the specific value depends on the actuator stroke and drive voltage range). The unit is nm, and its typical absolute value range is usually 0 nm to 50 nm; if the calibration data exhibits obvious nonlinear characteristics (for example, the piezoelectric ceramic actuator has a hysteresis effect under open-loop conditions), a quadratic polynomial fitting can be used instead. To improve conversion accuracy, the least squares polynomial fitting method can be used to solve for the three coefficients. 、 and .

[0054] After calibration is completed, each partition will be... Calibration parameters ( and ,or , and This information is pre-stored in the configuration register of the drive control unit; during the actual execution of step 4.2, for the partition... Read the corresponding calibration coefficient from the configuration register. and The target compensation displacement is calculated based on this calibration coefficient. Convert to physical drive input value Physical drive input value The calculation formula (based on the linear calibration model) is as follows: In the formula For the first The physical drive input values ​​required by the partition drive actuator (in units and calibrated) Consistent (e.g., mV); through the above formula, the target compensation displacement is converted from the Z-axis displacement domain of the mask plate to the input domain of the drive actuator, completing the connection and conversion from the compensation calculation domain to the physical drive domain; each compensation partition Partition identifier and corresponding physical drive input value By associating the storage, we obtain the set of physical drive travel parameters for each compensation partition.

[0055] Step 4.3: Based on the physical drive stroke parameter set of each compensation partition, encode and encapsulate the partition identifier and corresponding physical drive stroke parameters according to a preset instruction format to generate a mask partition drive control instruction set. Specifically, the preset instruction format is a standardized instruction frame structure defined by the drive control system, which includes at least the following fields: instruction header (identifying the start and type of the instruction), partition address (specifying the partition drive execution unit number to which the instruction is sent, and the partition identifier). One-to-one correspondence), physical drive input values (The drive voltage data to be executed in this partition), a check field (used for integrity verification of command transmission, which can use cyclic redundancy check CRC-8), and a command tail (marking the end of the command); the encoding and encapsulation operation is performed by the command generation submodule in the drive control unit: reading each partition one by one from the physical drive stroke parameter set. The data is filled field by field according to the above field structure to generate a complete control command frame; all of them are then processed. The control instruction frames of each partition are sorted by partition address and then packaged to generate a mask partition drive control instruction set.

[0056] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Based on the mask partition drive control instruction set, parse the drive control instructions corresponding to each compensation partition to obtain an independent drive instruction sequence for each compensation partition. Specifically, this includes: according to the obtained mask partition drive control instruction set, the instruction parsing submodule parses the drive control instructions corresponding to each compensation partition frame by frame; during parsing, according to the preset instruction format, the instruction header, instruction tail, and check field are extracted from each frame to extract the partition address and physical drive input value. The integrity of the transmission is verified using a validation field. If the verification fails, a retransmission request is initiated for that frame. Each partition address that passes verification is then matched with its corresponding... By merging the records, an independent drive instruction sequence for each compensation partition is obtained. Each item in this sequence contains the partition's instruction sequence. The address and the drive input value to be applied .

[0057] Step 5.2: Based on the independent drive instruction sequence for each compensation partition, distribute the drive control instructions for each compensation partition to the corresponding drive execution unit; each drive execution unit, according to the received drive control instructions, drives the corresponding partition to perform differentiated physical displacement along the Z direction, thereby obtaining the actual displacement state of each compensation partition. Specifically, this includes: based on the obtained independent drive instruction sequence for each compensation partition, distributing the drive control instructions for each compensation partition to the corresponding drive execution unit. The drive control instructions are distributed to the corresponding drive execution units of the partitions via a multi-channel parallel bus; during distribution, the instruction distribution submodule selects the appropriate instruction from the partition address. The value is written to the output register of the corresponding channel, and each channel's output register is connected to the input terminal of the drive amplifier circuit of the corresponding partition's piezoelectric ceramic actuator; after receiving the drive command, the drive execution unit of each partition uses the drive amplifier to... The driving voltage is converted into a power stage and applied to both ends of the actuator. The actuator responds to the driving voltage by generating a Z-axis displacement, which drives the corresponding partition of the mask to complete the physical movement of the specified stroke.

[0058] To ensure that the actual displacement of each zone matches the target value To ensure consistency, each partition's drive execution unit incorporates a closed-loop feedback control system. Within this system, capacitive gap sensors integrated beneath the mask of each partition monitor the actual Z-axis displacement of that partition in real time. Represented (unit: nm); the measured displacement value is transmitted by the sensor. Feedback is sent to the closed-loop feedback controller in the drive control unit; the closed-loop feedback controller will then transmit the measured displacement value. With target value Perform real-time comparison to obtain the first Instantaneous deviation of the partition (unit: nm), where Based on instantaneous deviation The cumulative amount of deviation over time and the rate of change of deviation over time are used to calculate the feedback control correction. and based on The driving voltage applied to the actuator is adjusted in real time; wherein, In the formula Indicates deviation amount The cumulative amount over time reflects the extent to which the deviation persists over time. Indicates deviation amount The rate of change over time is used to reflect the current trend (increasing or decreasing) of the deviation and how fast it changes; For the first The feedback control correction amount for the partition, in mV, is used to correct the drive voltage of the partition actuator in real time. This is the bias proportional gain coefficient (unit: mV / nm). This is the cumulative gain coefficient for the deviation (unit: mV / (nm∙s)). This is the gain coefficient for the rate of change of deviation (unit: mV∙s / nm).

[0059] The above three gain coefficients , and The typical range of values ​​is: Pick to , Pick to , Pick to The specific tuning method for the three is as follows: First, set the closed-loop feedback controller to pure proportional control mode (i.e., set...). and ), for the first Apply step target displacement command to the partition Gradually increase The values ​​of the proportional gain are determined and the measured displacement response curve fed back by the capacitive gap sensor is recorded; when the measured displacement response exhibits constant amplitude oscillation, the critical value of the proportional gain at this point is recorded. and oscillation period Based on and Calculate the tuning values ​​for the three gain coefficients: Pick to The values ​​between Pick to The values ​​between Pick to The values ​​are between; after tuning, the system's response performance is verified by applying step target displacement commands of different amplitudes.

[0060] In practical digital control systems, the above formula for calculating the feedback control correction in the continuous-time domain needs to be discretized to adapt to the periodic sampling execution mode of the digital controller. This embodiment uses the backward difference method for discretization approximation: let the sampling period of the digital controller be... (Suggested value range is) to The default value is ), by sampling point number Indicates the first The first control cycle, then the first control cycle within that cycle The instantaneous deviation of the partition is denoted as The discretized feedback control correction takes the following form: the proportional term remains constant. The integral term is approximated by the summation. The differential term is approximated by the backward difference as follows: To prevent integral saturation due to long-term deviation accumulation, when the calculated correction exceeds the actuator's allowable drive voltage range, the accumulation of the integral term is paused, and only the contributions of the proportional and derivative terms are retained until the deviation reaches its maximum value. Once the value falls back to the normal range, the accumulation of integral terms will resume.

[0061] Closed-loop feedback controller according to The actuator's drive voltage is adjusted in real time until the measured displacement value is reached. Enter target value Within the allowable error band; after each partition is driven, the capacitive gap sensor records and reports the actual final displacement value of that partition and whether it has reached the allowable error band flag, thus obtaining the compensation value for each partition. The actual displacement state of the partition; the allowable error band can be set as the target value. absolute value and not less than (Take the larger of the two values) to ensure that the compensation accuracy meets the process requirements.

[0062] Step 5.3: Based on the actual displacement state of each compensation zone, collect all spatial position information after the displacement of all compensation zones is completed to obtain the three-dimensional physical state of the mask after Z-axis partition compensation. Specifically, this includes: based on the obtained actual displacement state of each compensation zone, collecting all... The spatial position information after the displacement of each compensated partition is completed; during the aggregation operation, the actual Z-axis displacement value reported by each partition is combined with the predefined two-dimensional coordinate position of each partition in the mask plane (the center coordinates or boundary coordinates of each partition on the XY plane of the mask) to form a set of three-dimensional spatial position information: each partition With (X coordinate_) j, Y coordinate j, The triplet representing the actual Z-axis displacement value describes its spatial position in the global coordinate system of the mask; all After summing the triplet data of each partition, the three-dimensional physical state of the mask after Z-axis partition compensation is obtained. This state fully describes the actual three-dimensional geometry of the mask after completing the differential displacement adjustment of all partitions, and is the direct input for back-calculating the focal plane position of each depth layer in step 6.

[0063] In a preferred embodiment of the present invention, step 6 above may include: Step 6.1: Based on the three-dimensional physical state of the mask after Z-axis partition compensation, extract the actual Z-axis position of each compensated partition after displacement, and obtain the actual Z-axis position record of each compensated partition. Specifically, this includes: extracting the actual Z-axis position value after displacement of each partition based on the obtained three-dimensional physical state of the mask after Z-axis partition compensation; since the physical state data in step 5.3 already contains the values ​​of each partition... j The actual Z-axis displacement value is directly obtained from the state data according to the partition identifier in this step. jRead the corresponding Z-axis displacement values ​​one by one to obtain the actual Z-axis position record of each compensation zone; each entry in this record contains a zone identifier. j And the actual Z-axis displacement (in nm) completed in this partition.

[0064] Step 6.2: Based on the actual Z-axis position records of each compensation partition, for each compensation partition, determine the focal plane position of the corresponding depth layer from the actual Z-axis position of the partition, and obtain the focal plane position data of each depth layer. Specifically, this includes: based on the obtained actual Z-axis position records of each compensation partition, for each compensation partition... j The depth range corresponding to the partition is determined by the actual Z-axis position of the partition (i.e., the depth layer range of the partition). Layer to the first The focal plane position of the mask (layer); the correspondence between the Z-axis displacement of the mask partition and the focal plane position on the wafer surface depends on the optical characteristics of the projection optical path and needs to be established in advance through system-level focal plane calibration; the specific process of system-level focal plane calibration is as follows: on the photolithography equipment, a test wafer with a flat surface is selected and coated with the same photoresist as in the actual production; the Z-axis displacement actuators of each partition of the mask are sequentially adjusted to a series of known calibration displacements, and a set of test patterns on the test wafer is exposed at each calibration displacement; after exposure, the test wafer is developed and CD-SEM measured to obtain the key dimensions of the test patterns at each calibration displacement.

[0065] Furthermore, under each calibration displacement, the nominal focal plane setting during exposure is changed so that the nominal focal plane is scanned and exposed near the designed focal plane of the test pattern with a preset small step size (10nm to 30nm is recommended). This yields a set of discrete data points of nominal focal plane offset versus critical dimension, i.e., the Bossung curve. On the Bossung curve, the nominal focal plane offset corresponding to the critical dimension reaching its extreme value (maximum value for dark field line patterns, minimum value for bright field contact hole patterns) is the optimal focal plane position offset of the wafer surface corresponding to that partition of the mask under the current calibration displacement. ; calibrate displacement Extracted from the vertices of the Bossung curve By recording the one-to-one correspondence, a map of Z-axis displacements can be established point by point, partitioned by the mask plate. For the input index, use the corresponding wafer surface focal plane position offset. A calibration table for the Z-axis displacement-focal plane offset of the output values; the data coverage of the calibration table should be at least the full range of possible Z-axis travel for each partition of the mask (from 0 nm to ... This ensures that any possible displacement during actual use will not exceed the coverage boundary of the calibration table.

[0066] It should be noted that this calibration table establishes a causal relationship between the physical displacement of the mask partition and the focal plane offset of the wafer surface. After the mask undergoes physical displacement in the Z direction, the light beam passing through the mask pattern produces a change in optical path difference at the entrance pupil of the projection lens. This optical path difference is transmitted to the wafer surface through the reduction ratio of the projection lens (e.g., 4:1) and the aberration characteristics of the optical system, which manifests as a corresponding shift in the optimal focal plane position of the wafer surface. Since this transmission process is affected by multiple factors such as the numerical aperture of the projection lens, the illumination mode, and the characteristics of the mask pattern, the calibration table can directly establish an empirical mapping relationship between input and output using measured data, avoiding the difficulty of establishing a complex optical analytical model.

[0067] In the actual execution of step 6.2, for each compensation partition j Using the Z-axis displacement recorded in that partition as an index, the calibration table is looked up to obtain the depth layer range corresponding to that partition (the first...). Layer to the first The focal plane position data of each depth layer in the layer; if the value of the partition displacement falls between two adjacent sampling points in the calibration table, the corresponding focal plane position value can be obtained by linear interpolation; if a compensation partition corresponds to multiple depth layers, the focal plane position of all depth layers in the current partition under the Z-direction displacement of the current partition is assigned the same lookup (or interpolation) result value, because under the same mask partition Z-direction displacement condition, the optical transmission path of each depth layer within the coverage of the partition is approximately consistent.

[0068] Step 6.3: Based on the focal plane position data of each depth layer, associate and organize the axial position coordinates of the depth layers with the corresponding focal plane positions to obtain a dynamic focal plane distribution map of the stepped structure. Specifically, this includes: based on the obtained focal plane position data of each depth layer, associating and organizing the axial position coordinates of the depth layers with the corresponding focal plane positions to obtain a dynamic focal plane distribution map of the stepped structure. The corresponding focal plane position is associated with the organization; the association organization method is: based on the axial position coordinates of the depth layer. A two-dimensional correspondence distribution map is constructed, with the focal plane position of each depth layer as the first dimension and the focal plane position of each depth layer as the second dimension. In this distribution map, the position coordinates of each depth layer are arranged horizontally. Mark the focal plane position of the corresponding depth layer in the current mask partition displacement state along the vertical direction.

[0069] This distribution map is called a dynamic focal plane distribution map because the focal plane positions corresponding to each depth layer are not on the same plane. Due to the differentiated Z-axis displacement adjustment performed by different partitions of the mask, the focal plane offset received by the depth layers covered by each partition is different. The different depth layers on the entire stepped structure present a spatially non-uniformly distributed focal plane shape. This non-uniform distribution feature is fundamentally different from the uniform planar distribution in the traditional overall focusing method where all depth layers share a single focal plane. At the same time, the specific shape of this distribution map also depends on the physical state achieved by the actual driving of each partition in step 5: if the actual displacement of each partition deviates from the command target due to process fluctuations or sensor calibration differences, the shape of the distribution map will change accordingly. Step 6.2 uses the actual Z-axis displacement value of each partition (rather than the command target value) to look up a table to ensure that the distribution map reflects the real focal plane state after driving. The stepped structure dynamic focal plane distribution map will be directly sent to step 7 as the basis for differentiated exposure correction of each depth layer.

[0070] The physical mechanism by which the Z-axis differential displacement of the mask partition achieves the differential distribution of the focal plane at each depth layer can be further explained as follows: In the three-dimensional memory stepped structure, the word line contact windows of different depth layers are not vertically stacked at the same lateral position on the wafer surface: each step of the stepped structure unfolds horizontally in sequence along a specific direction (e.g., the X or Y direction) within the wafer plane, i.e., the... i The lateral coordinate range of the depth layer within the wafer plane and the first i The lateral coordinate intervals of the +1 depth layers are adjacent to each other rather than overlapping; under this structural feature, there is a spatial correspondence between each depth layer and each partition of the mask: the exposure area projected by a certain compensation partition of the mask exactly covers the step area corresponding to one or a group of specific depth layers on the wafer plane.

[0071] Therefore, when different partitions of the mask are driven to different Z-axis positions, the focal plane position of the wafer plane region corresponding to each partition is differentially shifted. Different wafer plane regions correspond to different depth layers, thus achieving a differential distribution of focal plane positions for each depth layer. This mechanism does not apply selective focusing to different depth layers at the same vertical axis position through the optical system, but utilizes the geometric feature of the stepped structure to flatten the depth dimension into the lateral dimension. Through the lateral-depth mapping chain of mask partition Z-axis displacement, wafer lateral region focal plane differentiation, and focal plane differentiation for each depth layer, a differential compensation effect in three-dimensional space is indirectly achieved.

[0072] In a preferred embodiment of the present invention, step 7 above may include: Step 7.1: Based on the stepped structure dynamic focal plane distribution map, read the focal plane position associated with each depth layer to obtain the exposure reference focal plane position data for each depth layer. Specifically, this includes: based on the obtained stepped structure dynamic focal plane distribution map, reading the focal plane position associated with each depth layer (…). i =1 to M) Read the focal plane position associated with each depth layer; read operation according to i Execute sequentially from 1 to M, and assign the axial position coordinates of each depth layer in the distribution map. Each entry is extracted along with its corresponding focal plane position value to obtain the exposure reference focal plane position data for each depth layer; each entry in this data contains a depth layer index. i The location of the exposure reference focal plane corresponding to this layer under the current physical state of the mask is the direct basis for correcting the exposure parameters of each depth layer.

[0073] Step 7.2: Based on the exposure reference focal plane position data for each depth layer, determine the exposure parameter correction amount for each depth layer based on the focal plane position to obtain a differentiated exposure parameter set for each depth layer. Specifically, the exposure parameter correction amount mainly includes two types: the correction ratio of the exposure dose and the offset of the focal position. The determination method is to establish an empirical mapping table of focal plane offset and exposure parameter correction amount in advance through exposure experiments on test wafers. The test experiment process is as follows: under fixed exposure dose conditions, a set of test patterns on the test wafer are exposed under multiple different focal plane offsets, and the key dimensions of the patterns obtained under each focal plane offset are measured. The optimal focal plane offset that makes the key dimensions closest to the design target value is found, and the additional exposure dose adjustment ratio that needs to be applied under the focal plane offset is recorded. The focal plane offset, the corresponding dose correction ratio, and the focal offset are organized into a correction amount mapping table. The typical value range of the dose correction ratio is 0.80 to 1.20 (i.e., 80% to 120% of the standard dose), and the typical value range of the focal offset is -300nm to +300nm.

[0074] During step 7.2, for the depth layer The focal plane position value of this layer provided in step 7.1 Use the index to look up the corresponding exposure parameter correction in the correction mapping table; if no exact match is found in the mapping table... The value exists, but there are two adjacent calibrated focal plane position values. and (satisfy Their respective exposure parameter corrections are as follows: and (This could be either a dose correction ratio or a focus offset), in which case linear interpolation would be used to determine it. Corresponding correction amount ;in, In the formula For the current depth layer The focal plane position value (in nm); and These are respectively below and above in the mapping table. The two most recent calibrated focal plane position values ​​(in nm); if Located outside the mapping table calibration range (i.e. Minimum calibration value in the mapping table or If the maximum calibration value in the mapping table is used, then the distance in the mapping table is taken. The correction at the most recent boundary calibration value is used as That is, when extending outwards, no extrapolation is performed; the boundary value is directly taken to ensure that the correction amount does not deviate from the calibration and verification range; all The exposure parameter corrections for each depth layer are summarized to obtain a set of differentiated exposure parameters for each depth layer.

[0075] Step 7.3: Based on the differentiated exposure parameter set for each depth layer, load the corresponding exposure parameters for each depth layer into the exposure control unit. The exposure control unit then performs differentiated exposure on the word line contact windows of each depth layer according to the corresponding exposure parameters, obtaining the three-dimensional spatial compensation and correction results for the word line contact windows. Specifically, this includes: based on the obtained differentiated exposure parameter set for each depth layer, loading the exposure parameters for each depth layer into the exposure control unit. i The corresponding exposure parameters (dose correction ratio and focus offset) are loaded into the exposure control unit of the lithography equipment; the loading operation is specifically implemented by indexing the differentiated exposure parameters of each depth layer by layer. i The data is sequentially written to the corresponding data fields in the exposure process formulation file; each record in the formulation file contains a depth layer identifier. i The lateral coordinate range of the layer in the wafer plane (i.e., the boundary of the layer step on the wafer XY plane), dose correction ratio, and focus offset.

[0076] During scanning exposure, the exposure control unit determines the depth layer to which the current exposure area belongs based on the current lateral position (XY coordinates) of the wafer stage. i In a stepped structure, different depth layers correspond to different lateral positions within the wafer plane (each step is arranged sequentially in the X or Y direction). Therefore, the depth layer can be automatically identified through real-time position feedback from the wafer stage; the current depth layer is identified. i Then, the exposure control unit reads the dose correction ratio and focus offset of the layer from the formula file, and adjusts the exposure dose output of the light source (multiplying the standard dose by the dose correction ratio) and the focal plane position of the projection lens (adding the focus offset on the nominal focal plane position) in real time.

[0077] The above adjustments are automatically switched layer by layer during the scanning exposure process: when the wafer stage moves so that the exposure area changes from the first layer to the second layer...i The horizontal region corresponding to the layer enters the first i When the horizontal region corresponding to layer +1 is reached, the exposure control unit automatically switches to the next layer. i The differentiated exposure parameters of the +1 layer ensure that the word line contact window of each depth layer is exposed and imaged under optimal conditions after focal plane verification and parameter correction. After the exposure of all depth layers is completed, the word line contact window of each layer is imaged under its own differentiated compensation and correction focal plane and dose conditions. The critical size difference between each depth layer is significantly reduced compared with that before compensation and correction, and the imaging uniformity of the word line contact window along the depth direction is improved. Finally, the three-dimensional spatial compensation and correction result of the word line contact window is obtained. This result can be directly verified by subsequent CD-SEM remeasurement: the critical size of the word line contact window of each depth layer is measured again on the compensated and corrected wafer to verify whether the consistency of the critical size between each layer meets the process specification requirements.

[0078] like Figure 2 As shown, embodiments of the present invention also provide a Z-axis adjustable mask three-dimensional space compensation and correction system, comprising: The imaging deviation distribution construction module is used to obtain the initial imaging deviation sequence of each depth layer of the three-dimensional memory ladder structure, extract the axial position coordinates corresponding to each depth layer, and obtain the depth layer imaging deviation distribution dataset. The deviation gradient calculation module is used to calculate the axial deviation rate of imaging deviation between adjacent depth layers based on the depth layer imaging deviation distribution dataset, and to quantize the axial deviation rate layer by layer to obtain the depth direction imaging deviation gradient set. The cumulative compensation matrix construction module is used to accumulate the axial deviation rate of each depth layer along the axial direction based on the depth direction imaging deviation gradient set, derive the target compensation displacement corresponding to each mask compensation partition, bind the target compensation displacement to partition, and obtain the Z-direction target compensation displacement matrix of each mask compensation partition. The drive instruction generation module is used to convert the physical drive stroke of each compensation partition into the Z-axis target compensation displacement matrix of each compensation partition of the mask, and generate a mask partition drive control instruction set. The partition drive execution module is used to drive each compensation partition to perform differentiated physical displacement adjustment along the Z direction according to the mask partition drive control instruction set, so as to obtain the three-dimensional physical state of the mask after Z-direction partition compensation. The focal plane distribution construction module is used to calculate the actual focal plane position corresponding to each depth layer based on the three-dimensional spatial physical state of the mask after Z-axis partitioning compensation, and obtain the dynamic focal plane distribution map of the stepped structure. The differential exposure correction module is used to perform differential exposure correction on each depth layer according to the dynamic focal plane distribution map of the stepped structure, so as to obtain the three-dimensional space compensation correction result of the word line contact window.

[0079] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0080] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for compensating and correcting a mask plate in a three-dimensional space in a Z direction, characterized in that, The method includes: The initial imaging deviation sequence of each depth layer of the three-dimensional memory stepped structure is obtained, and the axial position coordinates corresponding to each depth layer are extracted according to the initial imaging deviation sequence to obtain the depth layer imaging deviation distribution dataset. Based on the depth layer imaging deviation distribution dataset, the axial deviation rate of adjacent depth layer imaging deviations is calculated, and the axial deviation rate is quantized layer by layer to obtain the depth direction imaging deviation gradient set. Based on the depth-direction imaging deviation gradient set, the axial deviation rate of each depth layer is accumulated along the axial direction to derive the target compensation displacement corresponding to each mask compensation partition; the target compensation displacement is partitioned and bound to obtain the Z-direction target compensation displacement matrix of each mask compensation partition; Based on the Z-axis target compensation displacement matrix of each compensation zone of the mask, the physical drive stroke of each compensation zone is obtained, and a mask zone drive control instruction set is generated. Based on the mask partition drive control instruction set, each compensation partition is driven to perform differentiated physical displacement adjustment along the Z direction to obtain the three-dimensional spatial physical state of the mask after Z-direction partition compensation. Based on the three-dimensional spatial physical state of the mask after Z-axis partitioning compensation, the actual focal plane position corresponding to each depth layer is calculated to obtain the dynamic focal plane distribution map of the stepped structure. Based on the dynamic focal plane distribution map of the stepped structure, differential exposure correction is performed on each depth layer to obtain the three-dimensional spatial compensation correction result of the word line contact window.

2. The Z-adjustable mask plate three-dimensional space compensation correction method according to claim 1, wherein, The initial imaging deviation sequence of each depth layer of the three-dimensional memory stepped structure is obtained. Based on the initial imaging deviation sequence, the axial position coordinates corresponding to each depth layer are extracted to obtain the depth layer imaging deviation distribution dataset, including: Imaging acquisition is performed on each depth layer of the three-dimensional memory ladder structure to obtain the measurement image data of the word line contact window of each depth layer and the corresponding axial position identification information, so as to obtain the original measurement image data of each depth layer with position markings. Based on the original measurement image data with position markers, the actual imaging feature parameters of the word line contact window of each depth layer are extracted to obtain the actual imaging feature parameter set with position markers for each depth layer. Based on the actual imaging feature parameter set with location markers, the actual imaging feature parameters of each depth layer are compared with the design target feature parameters corresponding to each depth layer layer by layer to obtain the comparison difference results of each depth layer layer by layer; based on the comparison difference results of each depth layer layer, the feature deviation amount of each depth layer is extracted to obtain the initial imaging deviation value with location markers of each depth layer. Based on the initial imaging deviation value with position markers, the axial position identification information associated with each deviation value is read, and the axial position identification information is converted into position coordinates of each depth layer in the axial direction of the stepped structure. The initial imaging deviation value of each depth layer is associated with the converted position coordinates one by one to obtain the depth layer imaging deviation distribution dataset.

3. The Z-adjustable mask plate three-dimensional space compensation correction method according to claim 2, wherein, Based on the depth layer imaging deviation distribution dataset, the axial deviation rate of adjacent depth layers is calculated, and the axial deviation rate is quantized layer by layer to obtain a set of depth-direction imaging deviation gradients, including: Based on the depth layer imaging deviation distribution dataset, the position coordinates corresponding to each depth layer are sorted along the axial direction to obtain the deviation-coordinate correspondence sequence after sorting along the axial direction; the initial imaging deviation values ​​arranged in order in the deviation-coordinate correspondence sequence after sorting along the axial direction are extracted to obtain the imaging deviation sequence arranged in the depth direction. Based on the imaging deviation sequence arranged along the depth direction, two adjacent depth layers in the sequence are grouped together to obtain a set of adjacent depth layer pairings. The imaging deviation values ​​of the two depth layers in each pairing group are extracted, and the associated position coordinates of the two depth layers are extracted simultaneously. The axial spacing between the two depth layers is determined based on the position coordinates. The variation range of the imaging deviation values ​​along the axial spacing in each group is calculated to obtain the axial deviation rate of each group. The axial deviation rates of each group are collected to obtain a set of axial deviation rates between adjacent layers. Based on the set of axial deviation gradient rates between adjacent layers, the corresponding axial deviation gradient rates of each group are assigned to the depth layers associated with each group, so that each depth layer has a corresponding axial deviation gradient rate value, thus obtaining the set of imaging deviation gradients in the depth direction.

4. The Z-adjustable mask plate three-dimensional space compensation correction method according to claim 3, wherein, Based on the depth-direction imaging deviation gradient set, the axial deviation rate of each depth layer is accumulated along the axial direction to derive the target compensation displacement corresponding to each mask compensation partition. The target compensation displacement is partitioned and bound to obtain the Z-axis target compensation displacement matrix of each compensation partition of the mask, including: Based on the depth-direction imaging deviation gradient set, taking the depth layer at one end of the stepped structure as the accumulation starting point, the axial deviation rate of each depth layer is accumulated layer by layer along the axial direction to obtain the axial cumulative compensation sequence. Based on the axial cumulative compensation sequence, the degree of difference in cumulative compensation between adjacent depth layers is compared. Adjacent depth layers with differences within a preset range are merged into the same compensation partition, and the depth layer range included in each compensation partition is determined to obtain the mask compensation partition division scheme. Based on the mask compensation partitioning scheme and the axial cumulative compensation sequence, for each compensation partition, the cumulative compensation amount corresponding to each depth layer in the partition is aggregated to obtain the partition aggregated compensation amount; the partition aggregated compensation amount is used as the target compensation displacement amount of the corresponding partition to obtain the partition target compensation displacement amount of each compensation partition. Based on the target compensation displacement of each compensation zone, the zone identifier and the corresponding target compensation displacement are organized into a matrix to obtain the Z-axis target compensation displacement matrix of each compensation zone of the mask.

5. The Z-adjustable mask plate three-dimensional space compensation correction method according to claim 4, wherein, Based on the Z-axis target compensation displacement matrix of each compensation zone of the mask, the physical drive stroke of each compensation zone is obtained, and a mask zone drive control instruction set is generated, including: Based on the Z-axis target compensation displacement matrix of each compensation zone of the mask, the target compensation displacement associated with the zone identifier is read zone by zone to obtain the target displacement set of the zone to be converted. Based on the target displacement set of the partitions to be converted, for each compensation partition, the corresponding drive mechanism calibration parameters are extracted; based on the drive mechanism calibration parameters, the partition target displacement is converted into physical drive stroke values ​​to obtain the physical drive stroke parameter set of each compensation partition. Based on the physical drive stroke parameter set of each compensation zone, the zone identifier and the corresponding physical drive stroke parameter are encoded and encapsulated according to a preset instruction format to generate a mask zone drive control instruction set.

6. The Z-adjustable mask plate three-dimensional space compensation correction method according to claim 5, wherein, Based on the mask partition drive control instruction set, each compensation partition is driven to perform differentiated physical displacement adjustment along the Z-axis, resulting in the three-dimensional spatial physical state of the mask after Z-axis partition compensation, including: Based on the mask partition drive control instruction set, the drive control instructions corresponding to each compensation partition are parsed to obtain the independent drive instruction sequence for each compensation partition. Based on the independent drive instruction sequence of each compensation partition, the drive control instructions of each compensation partition are distributed to the drive execution unit of the corresponding partition; each drive execution unit drives the corresponding partition to perform differentiated physical displacement along the Z direction according to the received drive control instructions, so as to obtain the actual displacement state of each compensation partition. Based on the actual displacement state of each compensation zone, the spatial position information of all compensation zones after displacement is completed is collected to obtain the three-dimensional physical state of the mask after Z-axis zone compensation.

7. The Z-adjustable mask plate three-dimensional space compensation correction method according to claim 6, wherein, Based on the three-dimensional physical state of the mask after Z-axis partitioning compensation, the actual focal plane positions corresponding to each depth layer are calculated, resulting in a dynamic focal plane distribution map of the stepped structure, including: Based on the three-dimensional physical state of the mask after Z-axis partition compensation, the actual Z-axis position after the displacement of each compensation partition is completed is extracted, and the actual Z-axis position record of each compensation partition is obtained. Based on the actual Z-axis position records of each compensation partition, for each compensation partition, the focal plane position of the corresponding depth layer is determined by the actual Z-axis position of the partition, and the focal plane position data of each depth layer is obtained. Based on the focal plane position data of each depth layer, the axial position coordinates of the depth layer are associated with the corresponding focal plane positions to obtain a dynamic focal plane distribution map of the stepped structure.

8. The Z-adjustable mask plate three-dimensional space compensation correction method according to claim 7, wherein, Based on the dynamic focal plane distribution map of the stepped structure, differential exposure correction is performed on each depth layer to obtain the three-dimensional spatial compensation correction results for the word line contact window, including: Based on the dynamic focal plane distribution map of the stepped structure, the focal plane position associated with each depth layer is read layer by layer to obtain the exposure reference focal plane position data for each depth layer. Based on the exposure reference focal plane position data for each depth layer, the exposure parameter correction amount is determined by the focal plane position for each depth layer, resulting in a differentiated exposure parameter set for each depth layer. Based on the differentiated exposure parameter set for each depth layer, the exposure parameters corresponding to each depth layer are loaded into the exposure control unit. The exposure control unit then performs differentiated exposure on the word line contact window of each depth layer according to the corresponding exposure parameters, thereby obtaining the three-dimensional spatial compensation and correction result of the word line contact window.

9. A Z-axis adjustable mask three-dimensional space compensation and correction system, the system implementing the method as described in any one of claims 1 to 8, characterized in that, include: The imaging deviation distribution construction module is used to obtain the initial imaging deviation sequence of each depth layer of the three-dimensional memory ladder structure, extract the axial position coordinates corresponding to each depth layer, and obtain the depth layer imaging deviation distribution dataset. The deviation gradient calculation module is used to calculate the axial deviation rate of imaging deviation between adjacent depth layers based on the depth layer imaging deviation distribution dataset, and to quantize the axial deviation rate layer by layer to obtain the depth direction imaging deviation gradient set. The cumulative compensation matrix construction module is used to accumulate the axial deviation rate of each depth layer along the axial direction based on the depth direction imaging deviation gradient set, derive the target compensation displacement corresponding to each mask compensation partition, bind the target compensation displacement to partition, and obtain the Z-direction target compensation displacement matrix of each mask compensation partition. The drive instruction generation module is used to convert the physical drive stroke of each compensation partition into the Z-axis target compensation displacement matrix of each compensation partition of the mask, and generate a mask partition drive control instruction set. The partition drive execution module is used to drive each compensation partition to perform differentiated physical displacement adjustment along the Z direction according to the mask partition drive control instruction set, so as to obtain the three-dimensional physical state of the mask after Z-direction partition compensation. The focal plane distribution construction module is used to calculate the actual focal plane position corresponding to each depth layer based on the three-dimensional spatial physical state of the mask after Z-axis partitioning compensation, and obtain the dynamic focal plane distribution map of the stepped structure. The differential exposure correction module is used to perform differential exposure correction on each depth layer according to the dynamic focal plane distribution map of the stepped structure, so as to obtain the three-dimensional space compensation correction result of the word line contact window.

Images