A wafer etching uniformity improvement method based on plasma parameter closed-loop regulation

Abstract

The application discloses a wafer etching uniformity improvement method based on plasma parameter closed-loop regulation and control, and relates to the technical field of semiconductors.The method comprises the following steps: obtaining wafer information to be processed; and obtaining corresponding wafer etching history data based on wafer size data of the wafer information to be processed.The application realizes accurate calculation of theoretical etching depth by wafer etching history data cleaning, abnormal value elimination and load effect model construction, realizes quantitative characterization of wafer etching radial and angular fluctuation characteristics by polar coordinate mapping, Gaussian fitting and discrete Fourier transform extraction of radial and angular basic values, realizes adaptive and accurate division of wafer sub-regions by combining etching depth deviation threshold, pattern density and aspect ratio for partitioning, realizes uniformization compensation of wafer global etching depth by differentiating plasma characteristic parameters through sub-region etching attenuation coefficient correction, and effectively improves wafer etching uniformity.

Description

Technical Field

[0001] This invention relates to the field of semiconductor technology, specifically to a method for improving wafer etching uniformity based on closed-loop control of plasma parameters. Background Technology

[0002] Wafer etching is a core process in semiconductor chip manufacturing for achieving micro- and nano-pattern transfer. The uniformity of in-plane etching directly determines the etching depth, contour dimensions, and electrical parameter consistency of chips on the same wafer, making it a key indicator for improving chip yield, reducing production costs, and ensuring stable device performance. Plasma etching, with its high anisotropy, high precision, and high selectivity, has become the mainstream etching method in advanced processes. The spatial distribution of parameters such as electron density, electron temperature, ion flux, and ion energy directly determines the uniformity of etching rate and depth. Achieving precise control over wafer etching uniformity can effectively suppress over-etching, under-etching, and pattern distortion, adapting to the processing requirements of high aspect ratios and high-density patterns, and has significant engineering value for promoting the iteration of large-size wafers and advanced semiconductor processes.

[0003] Currently, wafer etching still faces several challenges, including the inability to accurately analyze etching patterns and divide the wafer based on them. Existing technologies often use fixed radial and angular intervals to divide regions without considering the radial fluctuation patterns and angular dominant frequency characteristics of historical etching data. This results in a mismatch between the partitioning and the actual etching deviation distribution. Furthermore, the impact of macroscopic pattern density and microscopic aspect ratio on etching depth is not comprehensively considered, making it impossible to accurately predict etching attenuation at different locations. In addition, the plasma parameter control is rigid, using globally uniform parameters without differentiated compensation and control for different sub-regions, all of which affect the stability and reliability of wafer etching quality. Summary of the Invention

[0004] To address the aforementioned technical problems, this paper provides a method for improving wafer etching uniformity based on closed-loop control of plasma parameters. This technical solution solves the problems mentioned in the background technology, such as the inability to accurately analyze etching patterns and accurately divide the wafer according to the etching patterns. Existing technologies mostly use fixed radial and angular intervals to divide regions without combining the radial fluctuation patterns and angular dominant frequency characteristics of historical etching data. The partitioning does not match the actual etching deviation distribution, and the influence of macroscopic pattern density and microscopic aspect ratio on etching depth is not comprehensively considered. It is also impossible to accurately predict etching attenuation at different locations. Furthermore, the plasma parameter control is rigid, using globally uniform parameters without differentiated compensation control for different sub-regions, which affects the stability and reliability of wafer etching quality.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for improving wafer etching uniformity based on closed-loop control of plasma parameters includes: Obtain information about the wafer to be processed, including wafer size data and target etching parameters; Based on the wafer size data of the wafer information to be processed, obtain the corresponding wafer etching history data; Based on historical wafer etching data, obtain the radial and angular baseline values; Based on the information of the wafer to be processed, the wafer to be processed is divided into regions using radial and angular base values ​​to obtain wafer sub-region information. Based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, the initial plasma characteristic parameters corresponding to each wafer sub-region are obtained. The plasma characteristic parameters include electron density, electron temperature, ion flux and ion energy. Based on historical wafer etching data, an etching depth influence model was constructed with plasma characteristic parameters as independent variables and etching depth as dependent variable. Based on the wafer sub-region information and initial plasma characteristic parameters, plasma characteristic parameters are obtained using the etching depth influence model. Based on plasma characteristic parameters, wafer etching is performed on each wafer sub-region.

[0006] Preferably, obtaining the radial and angular baseline values ​​based on wafer etching history data specifically includes: Based on the wafer size data of the wafer information to be processed, obtain the full-point etching history data of the wafer of the same specification; Based on the historical etching data of all points, after removing abnormal deviations, we obtained an effective etching depth dataset and a load effect model. Based on the effective etching depth dataset, a polar coordinate system is established with the wafer center as the origin. Each detection point is mapped to the radial distance and angular angle in polar coordinates to obtain the polar coordinate mapping data corresponding to each detection point. Based on the polar coordinate mapping data, with the wafer center as the origin, detection points are taken at equal intervals from the inside to the outside along the radial direction as radial feature points. The etching depth of each radial feature point is normalized to obtain the radial etching depth fluctuation sequence. A single-peak Gaussian fitting model is constructed based on the radial etching depth fluctuation sequence; Based on the single-peak Gaussian fitting model, the radial etching depth fluctuation sequence is fitted using the least squares method to obtain the fitted Gaussian curve. Based on the fitted Gaussian curve, the radial distance between two adjacent Gaussian peaks is extracted using a peak detection algorithm. The difference in radial distances between two adjacent Gaussian peaks is taken as the radial baseline value; Based on polar coordinate mapping data, corner feature points are extracted at equal intervals along the circumferential corner direction with the wafer center as the pole and a fixed angle as the step size to obtain the corner etching depth fluctuation sequence. Based on the angular etching depth fluctuation sequence, a frequency domain complex numerical sequence is obtained using discrete Fourier transform; Based on the frequency domain complex numerical sequence, the amplitude spectrum of each frequency domain component is calculated to obtain the amplitude spectrum data of the angular etching depth fluctuation. Based on the angular wave amplitude spectrum data, obtain the dominant frequency domain component with the largest amplitude; The ratio of the main frequency domain component to the total number of angular feature points is taken as the angular fluctuation main frequency, which represents the number of angular fluctuations per unit angle along the etching depth. The ratio of the total circumferential angle of the wafer to the dominant frequency of angular oscillation is used as the basic value for the angular direction.

[0007] Preferably, the step of dividing the wafer to be processed into regions based on radial and angular baseline values, and obtaining wafer sub-region information, specifically includes: S100: Based on wafer etching requirements, obtain an etching depth deviation threshold, wherein the etching depth deviation threshold represents the maximum allowable etching depth deviation value; S200: Based on the wafer information to be processed and the wafer layout design, obtain the pattern structure information and target etching depth information; S300: Based on the pattern structure information, obtain the pattern density information and aspect ratio of the wafer to be processed, wherein the aspect ratio represents the ratio of the target etching depth to the opening width corresponding to the pattern structure. S400: Based on the pattern density information and aspect ratio, and based on the load effect model, obtain the predicted etching depth at any position of the wafer to be processed, and use the difference between the predicted etching depth and the target etching depth as the predicted etching depth deviation value. S500: Using the center of the wafer to be processed as the origin, traverse radially from the inside to the outside until the predicted etching depth deviation exceeds the etching depth deviation threshold, or the distance between the current position and the center reaches the radial baseline value, and take the current position as the radial reference point. S600: Using the radial reference point as the new origin, repeat step S500 until the entire wafer to be processed is traversed to obtain radial reference point information. S700: Based on the center of the wafer to be processed, and with the distance from the center to the radial reference point as the radius, the annular zone is divided, and the line connecting the center and the radial reference point is used as the radial reference line for each annular zone. S800: For each annular region, based on the radial reference line, traverse the annular regions on both sides of the radial reference line along the angular direction until the predicted etching depth deviation exceeds the etching depth deviation threshold, or the angle between the line connecting the current position and the center of the circle and the radial reference line reaches the angular reference value, and take the current position as the angular reference point. S900: Take the line connecting the center of the wafer to be processed and the angular reference point as the new radial reference line, repeat step S800 until the entire annular zone is traversed, divide the annular zone with the radial reference line, and obtain wafer sub-region information.

[0008] Preferably, the step of obtaining an effective etching depth dataset by removing abnormal deviations from the historical etching data of all points specifically includes: Based on the historical etching data of all points, the historical etching structure data corresponding to each detection point is obtained. The historical etching structure data includes historical opening width data and corresponding historical etching depth data. The etching structure includes trenches, through holes and contact holes. Based on historical etching structure data, the historical opening width is used as the basis to group the historical etching data of all points. Detection points with the same historical opening width are divided into the same group to obtain the historical etching data group of all points. Based on the historical etching data set of all points, obtain the historical etching depth, target etching depth and etching pattern density corresponding to each detection point in each set; Based on historical etching depth, target etching depth, and etching pattern density, the characteristic points of the loading effect are obtained; Based on the characteristic points of the load effect, a load effect model is constructed based on macroscopic and microscopic load effects; Based on the load effect model, the theoretical etching depth corresponding to each detection point is obtained; The standard deviation is obtained based on the residual value between the theoretical etching depth and the historical etching depth corresponding to each detection point; Based on standard deviation, The principle is to obtain the abnormal depth threshold; Based on the abnormal depth threshold, the historical etching data of all points is filtered, and the historical etching data of all points with an etching depth exceeding the abnormal depth threshold is removed to obtain the effective etching depth dataset. Specifically, the load effect model is as follows: In the formula, For theoretical etching depth, For the target etching depth, This is the macroeconomic load effect coefficient. To etch pattern density, This is the micro lag coefficient. This refers to the historical opening width.

[0009] Preferably, obtaining the loading effect feature points based on the historical etching depth, target etching depth, and etching pattern density specifically includes: Based on the macroscopic loading effect, data fitting is performed according to historical etching depth, target etching depth and etching pattern density to obtain the loading effect characteristic coefficients corresponding to each full-point etching historical data set. The mean value of the load effect characteristic coefficients corresponding to the historical data set of all etching points is used as the load effect coefficient. Based on the historical etching data of all points, the etching depth deviation coefficient is obtained based on the etching pattern density and loading effect coefficient corresponding to each detection point. The product of the target etching depth and the etching depth deviation coefficient corresponding to each detection point is used as the etching depth threshold corresponding to that detection point. Based on the etching depth threshold, the detection points are divided, and the detection points whose historical etching depth has not exceeded the etching depth threshold are classified as load effect feature points. Specifically, the etching depth deviation coefficient is: In the formula, This is the etching depth deviation coefficient. This is the load effect coefficient. Etching pattern density represents the area of ​​the etched region per unit area. The characteristic coefficient of the load effect. For the target etching depth, Indicates the first Historical etching depth at each detection point Indicates the first The etching pattern density at each detection point.

[0010] Preferably, the step of obtaining plasma characteristic parameters based on wafer sub-region information and initial plasma characteristic parameters, using an etching depth influence model, specifically includes: Based on the wafer sub-region information, obtain the pattern density, aspect ratio, and target etching depth of each wafer sub-region; Based on the load effect model, the etching depth attenuation coefficient corresponding to each wafer sub-region is obtained according to the pattern density and aspect ratio of each wafer sub-region. The etching depth attenuation coefficient represents the ratio of the predicted etching depth to the target etching depth. Based on the etching depth attenuation coefficient corresponding to each wafer sub-region, the plasma characteristic parameters are obtained by substituting them into the etching depth influence model.

[0011] Furthermore, a wafer etching uniformity improvement system based on closed-loop control of plasma parameters is proposed to achieve the uniformity improvement method described above, including: The main control module is used to construct an etching depth influence model based on historical wafer etching data, with plasma characteristic parameters as independent variables and etching depth as dependent variable. Based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, the module obtains the initial plasma characteristic parameters corresponding to each wafer sub-region. Based on the wafer sub-region information and the initial plasma characteristic parameters, the module obtains the plasma characteristic parameters based on the etching depth influence model. The information acquisition module is used to acquire wafer information to be processed, including wafer size data and target etching parameter information. Based on the wafer size data of the wafer information to be processed, the module acquires corresponding historical wafer etching data. The region division module is used to obtain radial and angular base values ​​based on wafer etching history data, and to divide the wafer to be processed into regions based on the radial and angular base values ​​according to the wafer information, thereby obtaining wafer sub-region information.

[0012] Optionally, the main control module specifically includes: The control unit is used to obtain the initial plasma characteristic parameters corresponding to each wafer sub-region based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, and to obtain the plasma characteristic parameters based on the etching depth influence model based on the wafer sub-region information and the initial plasma characteristic parameters. An information receiving unit interacts with an information acquisition module and a region division module to receive data and transmit it to a control unit. The model building unit is used to build an etching depth influence model based on historical wafer etching data, with plasma characteristic parameters as independent variables and etching depth as dependent variable.

[0013] Optionally, the information acquisition module specifically includes: The first acquisition unit is used to acquire wafer information to be processed, the wafer information to be processed including wafer size data and target etching parameter information; The second acquisition unit is used to acquire corresponding wafer etching history data based on the wafer size data of the wafer information to be processed.

[0014] Optionally, the region division module specifically includes: The data analysis unit is used to obtain radial and angular baseline values ​​based on wafer etching history data. The region division unit is used to divide the wafer to be processed into regions based on radial and angular base values ​​according to the wafer information to be processed, and to obtain wafer sub-region information.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes a method for improving wafer etching uniformity based on closed-loop control of plasma parameters. By cleaning historical wafer etching data, outliers are removed and a loading effect model is constructed, enabling accurate calculation of the theoretical etching depth. Radial and angular baseline values ​​are extracted through polar coordinate mapping, Gaussian fitting, and discrete Fourier transform, achieving quantitative characterization of the radial and angular fluctuation characteristics of wafer etching. By combining etching depth deviation threshold, pattern density, and aspect ratio for partitioning, adaptive and accurate division of wafer sub-regions is achieved. By differentiating the etching attenuation coefficient of sub-regions to correct plasma characteristic parameters, uniform compensation of the etching depth across the entire wafer is achieved, effectively improving the uniformity of wafer etching. Attached Figure Description

[0016] Figure 1 This is a flowchart of a method for improving wafer etching uniformity based on closed-loop control of plasma parameters proposed in this invention. Figure 2 This is a flowchart of the radial base value acquisition process in this invention; Figure 3 This is a flowchart of the process for obtaining the basic angle value in this invention; Figure 4 This is a flowchart of the wafer sub-region information acquisition process in this invention; Figure 5 This is a block diagram of a wafer etching uniformity improvement system based on closed-loop control of plasma parameters proposed in this invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 - Figure 4 As shown in the figure, a method for improving wafer etching uniformity based on closed-loop control of plasma parameters in an embodiment of the present invention includes: Obtain information about the wafer to be processed, including wafer size data and target etching parameters; Based on the wafer size data of the wafer information to be processed, obtain the corresponding wafer etching history data; Based on historical wafer etching data, obtain the radial and angular baseline values; Specifically, based on historical wafer etching data, radial and angular baseline values ​​are obtained, including: Based on the wafer size data of the wafer information to be processed, obtain the full-point etching history data of the wafer of the same specification; Based on the historical etching data of all points, after removing abnormal deviations, we obtained an effective etching depth dataset and a load effect model. Based on the effective etching depth dataset, a polar coordinate system is established with the wafer center as the origin. Each detection point is mapped to the radial distance and angular angle in polar coordinates to obtain the polar coordinate mapping data corresponding to each detection point. Based on the polar coordinate mapping data, with the wafer center as the origin, detection points are taken at equal intervals from the inside to the outside along the radial direction as radial feature points. The etching depth of each radial feature point is normalized to obtain the radial etching depth fluctuation sequence. A single-peak Gaussian fitting model is constructed based on the radial etching depth fluctuation sequence; Based on the single-peak Gaussian fitting model, the radial etching depth fluctuation sequence is fitted using the least squares method to obtain the fitted Gaussian curve. Based on the fitted Gaussian curve, the radial distance between two adjacent Gaussian peaks is extracted using a peak detection algorithm. The difference in radial distances between two adjacent Gaussian peaks is taken as the radial baseline value; Based on polar coordinate mapping data, corner feature points are extracted at equal intervals along the circumferential corner direction with the wafer center as the pole and a fixed angle as the step size to obtain the corner etching depth fluctuation sequence. Based on the angular etching depth fluctuation sequence, a frequency domain complex numerical sequence is obtained using discrete Fourier transform; Based on the frequency domain complex numerical sequence, the amplitude spectrum of each frequency domain component is calculated to obtain the amplitude spectrum data of the angular etching depth fluctuation. Based on the angular wave amplitude spectrum data, obtain the dominant frequency domain component with the largest amplitude; The ratio of the main frequency domain component to the total number of angular feature points is taken as the angular fluctuation main frequency, which represents the number of angular fluctuations per unit angle along the etching depth. The ratio of the total circumferential angle of the wafer to the dominant frequency of angular oscillation is used as the basic value for the angular direction.

[0019] In this solution, historical etching data for all points on a wafer of the same specification is obtained based on the size of the wafer to be processed. This ensures that the historical data perfectly matches the physical specifications of the current processing object, providing a highly adaptable raw data source for subsequent fluctuation feature extraction. By removing abnormal deviation values ​​from the historical data for all points, a high-purity effective etching depth dataset is obtained. Simultaneously, a loading effect model is constructed to quantify the impact of pattern density and aperture size on etching depth, solving the problems of interference values ​​in the raw data and the lack of a quantified model for the loading effect. This ensures the accuracy of subsequent fluctuation analysis. A polar coordinate system is established with the wafer center as the origin, transforming the detection points into radial distance and angular angle data, perfectly fitting the wafer. The circular physical structure solves the problem that it is difficult to accurately analyze radial and angular etching fluctuations in the rectangular coordinate system. It realizes the two-dimensional standardized decomposition of wafer etching features, performs single-peak Gaussian fitting on the radial etching depth fluctuation sequence, calculates the radial base value through the peak spacing, accurately characterizes the natural fluctuation law of etching depth along the wafer radial direction, and replaces the fixed interval value method. It allows the radial partition to be based on the actual processing fluctuation characteristics. The discrete Fourier transform is used to convert the angular fluctuation sequence into frequency domain data, extract the main frequency component to calculate the main frequency of angular fluctuation and the angular base value, realize the quantitative characterization of angular periodic etching fluctuations, and solve the problems of no data basis and unreasonable division of angular partitions. In this embodiment, when taking detection points at equal intervals from the inside to the outside along the radial direction, the interval is 1000μm; when extracting angular feature points at equal intervals along the circumferential angle, the interval is 1°. The Gaussian fitting model is as follows: In the formula, This represents the model output value. Gaussian amplitude, Gaussian mean, Gaussian standard deviation, This represents the baseline offset of the fluctuation. Based on the radial etching depth fluctuation sequence and Gaussian fitting model, a least-squares fitting objective function is constructed: In the formula, To fit the objective function of residual sum of squares, For the first Historical etching depth of each radial feature point For the first The radial position of each radial feature point This represents the total number of radial feature points; The objective function To each Find the first-order partial derivatives and set them equal to 0 to obtain the nonlinear overdetermined system of equations: Gaussian amplitude Find the partial derivative: Gaussian mean Find the partial derivative: Gaussian standard deviation Find the partial derivative: For fluctuating baseline offset Find the partial derivative: To minimize the objective function To achieve the goal, the parameters are iteratively updated using the LM algorithm; When the change in the sum of squared residuals between two consecutive iterations is less than the preset convergence threshold ( When the iteration converges, it is determined that the iteration has converged. The Discrete Fourier Transform is specifically as follows: In the formula, It is a complex numerical sequence in the frequency domain. This represents the frequency domain component index. The imaginary unit, Indicates the first The angle of each angle towards the feature point Indicates the first Historical etching depth of each angular feature point This represents the total number of angular feature points. Amplitude spectrum of frequency domain components: Specifically, based on the historical etching data of all points, after removing outlier values, an effective etching depth dataset is obtained, which includes: Based on the historical etching data of all points, the historical etching structure data corresponding to each detection point is obtained. The historical etching structure data includes historical opening width data and corresponding historical etching depth data. The etching structure includes trenches, through holes and contact holes. Based on historical etching structure data, the historical opening width is used as the basis to group the historical etching data of all points. Detection points with the same historical opening width are divided into the same group to obtain the historical etching data group of all points. Based on the historical etching data set of all points, obtain the historical etching depth, target etching depth and etching pattern density corresponding to each detection point in each set; Based on historical etching depth, target etching depth, and etching pattern density, the characteristic points of the loading effect are obtained; Based on the characteristic points of the load effect, a load effect model is constructed based on macroscopic and microscopic load effects; Based on the load effect model, the theoretical etching depth corresponding to each detection point is obtained; The standard deviation is obtained based on the residual value between the theoretical etching depth and the historical etching depth corresponding to each detection point; Based on standard deviation, The principle is to obtain the abnormal depth threshold; Based on the abnormal depth threshold, the historical etching data of all points is filtered, and the historical etching data of all points with an etching depth exceeding the abnormal depth threshold is removed to obtain the effective etching depth dataset. Specifically, the load effect model is as follows: In the formula, For theoretical etching depth, For the target etching depth, This is the macroeconomic load effect coefficient. To etch pattern density, This is the micro lag coefficient. This refers to the historical opening width.

[0020] In this scheme, historical etching data at all points are categorized based on historical aperture width, accurately distinguishing the etching load differences of different etching widths. This provides accurate and consistent basic data for load effect modeling. Multi-dimensional features such as historical etching depth, target etching depth, and etching pattern density are extracted for each group of detection points, comprehensively covering core influencing factors of the load effect, such as macroscopic pattern density and microscopic structure size. This provides sufficient and comprehensive feature support for the construction of the load effect model. Based on macroscopic and microscopic load effects, a dedicated load effect model is constructed, achieving accurate prediction of theoretical etching depth under different etching structures, different pattern densities, and different aperture sizes. This provides a scientific and accurate theoretical benchmark for outlier removal. By calculating the residual and standard deviation between theoretical and historical etching depths, an outlier depth threshold is set based on the 3σ statistical principle, automatically removing outlier data exceeding the threshold. This ensures the high purity and high reliability of the effective etching depth dataset, preventing interference from outlier data on subsequent fluctuation analysis and zonal control from the source.

[0021] Specifically, based on historical etching depth, target etching depth, and etching pattern density, loading effect feature points are obtained, including: Based on the macroscopic loading effect, data fitting is performed according to historical etching depth, target etching depth and etching pattern density to obtain the loading effect characteristic coefficients corresponding to each full-point etching historical data set. The mean value of the load effect characteristic coefficients corresponding to the historical data set of all etching points is used as the load effect coefficient. Based on the historical etching data of all points, the etching depth deviation coefficient is obtained based on the etching pattern density and loading effect coefficient corresponding to each detection point. The product of the target etching depth and the etching depth deviation coefficient corresponding to each detection point is used as the etching depth threshold corresponding to that detection point. Based on the etching depth threshold, the detection points are divided, and the detection points whose historical etching depth has not exceeded the etching depth threshold are classified as load effect feature points. Specifically, the etching depth deviation coefficient is: In the formula, This is the etching depth deviation coefficient. This is the load effect coefficient. Etching pattern density represents the area of ​​the etched region per unit area. The characteristic coefficient of the load effect. For the target etching depth, Indicates the first Historical etching depth at each detection point Indicates the first The etching pattern density at each detection point.

[0022] In this scheme, for etching data sets with different aperture widths, the historical etching depth, target etching depth, and etching pattern density are fitted based on the principle of macroscopic loading effect. The abstract macroscopic loading effect of "pattern density suppressing etching rate" is transformed into quantifiable loading effect characteristic coefficients, laying the core mathematical foundation for the accurate characterization of the loading effect. The average of the loading effect characteristic coefficients of each data set is used as the global loading effect coefficient. This not only fully adapts to the loading characteristics of different aperture widths and different etching structures (trenches, vias, contact holes), but also achieves unified calibration of the loading effect coefficients across the entire wafer, avoiding calculation deviations caused by differences in coefficients between groups, and significantly improving the overall performance. The stability and reliability of the loading effect at different locations on the wafer are characterized by constructing an etching depth deviation coefficient model, which directly establishes a quantitative correlation between etching pattern density and etching depth attenuation. This model can accurately calculate the etching depth deviation caused by uneven pattern density distribution at each detection point, providing a precise quantitative basis for subsequent threshold calculation and feature point selection. Using "target etching depth × etching depth deviation coefficient" as the threshold, detection points whose historical etching depth conforms to the loading effect law are automatically selected as loading effect feature points. This accurately extracts effective samples affected by the loading effect and eliminates interference data without loading effect correlation, providing high-quality and highly relevant feature samples for subsequent loading effect model construction.

[0023] Based on the information of the wafer to be processed, the wafer to be processed is divided into regions using radial and angular base values ​​to obtain wafer sub-region information. Specifically, based on the information of the wafer to be processed, the wafer is divided into regions using radial and angular baseline values ​​to obtain wafer sub-region information, including: S100: Based on wafer etching requirements, obtain an etching depth deviation threshold, wherein the etching depth deviation threshold represents the maximum allowable etching depth deviation value; S200: Based on the wafer information to be processed and the wafer layout design, obtain the pattern structure information and target etching depth information; S300: Based on the pattern structure information, obtain the pattern density information and aspect ratio of the wafer to be processed, wherein the aspect ratio represents the ratio of the target etching depth to the opening width corresponding to the pattern structure. S400: Based on the pattern density information and aspect ratio, and based on the load effect model, obtain the predicted etching depth at any position of the wafer to be processed, and use the difference between the predicted etching depth and the target etching depth as the predicted etching depth deviation value. S500: Using the center of the wafer to be processed as the origin, traverse radially from the inside to the outside until the predicted etching depth deviation exceeds the etching depth deviation threshold, or the distance between the current position and the center reaches the radial baseline value, and take the current position as the radial reference point. S600: Using the radial reference point as the new origin, repeat step S500 until the entire wafer to be processed is traversed to obtain radial reference point information. S700: Based on the center of the wafer to be processed, and with the distance from the center to the radial reference point as the radius, the annular zone is divided, and the line connecting the center and the radial reference point is used as the radial reference line for each annular zone. S800: For each annular region, based on the radial reference line, traverse the annular regions on both sides of the radial reference line along the angular direction until the predicted etching depth deviation exceeds the etching depth deviation threshold, or the angle between the line connecting the current position and the center of the circle and the radial reference line reaches the angular reference value, and take the current position as the angular reference point. S900: Take the line connecting the center of the wafer to be processed and the angular reference point as the new radial reference line, repeat step S800 until the entire annular zone is traversed, divide the annular zone with the radial reference line, and obtain wafer sub-region information.

[0024] In this solution, the etching depth deviation threshold is combined with the radial and angular baseline values ​​as the partitioning termination condition. This not only strictly conforms to the allowable deviation range of the wafer etching process but also fits the radial and angular fluctuation patterns of the wafer etching depth, achieving a scientific and adaptive region division. Key graphic parameters such as pattern density and aspect ratio are introduced during the partitioning process, and the predicted etching depth deviation at any position is quantitatively calculated based on the load effect model. This ensures that the region division fully conforms to the actual etching characteristics of the wafer. A two-layer partitioning logic is adopted, first radially dividing the annular region and then angularly dividing the sector sub-regions. This perfectly adapts to the geometric structure characteristics of circular wafers, achieving a fine sub-region division with no overlap or omissions across the entire wafer to be processed. It automatically traverses along the radial and angular directions, terminating when either the deviation exceeds the threshold or reaches the baseline value. This balances partitioning accuracy and execution efficiency, avoiding control redundancy and blind spots caused by over-partitioning or under-partitioning.

[0025] Based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, the initial plasma characteristic parameters corresponding to each wafer sub-region are obtained. The plasma characteristic parameters include electron density, electron temperature, ion flux and ion energy. Based on historical wafer etching data, an etching depth influence model was constructed with plasma characteristic parameters as independent variables and etching depth as dependent variable. Based on the wafer sub-region information and initial plasma characteristic parameters, plasma characteristic parameters are obtained using the etching depth influence model. Specifically, based on wafer sub-region information and initial plasma characteristic parameters, and using the etching depth influence model as a foundation, plasma characteristic parameters are obtained, including: Based on the wafer sub-region information, obtain the pattern density, aspect ratio, and target etching depth of each wafer sub-region; Based on the load effect model, the etching depth attenuation coefficient corresponding to each wafer sub-region is obtained according to the pattern density and aspect ratio of each wafer sub-region. The etching depth attenuation coefficient represents the ratio of the predicted etching depth to the target etching depth. Based on the etching depth attenuation coefficient corresponding to each wafer sub-region, the plasma characteristic parameters are obtained by substituting them into the etching depth influence model.

[0026] This solution extracts core structural parameters such as pattern density and aspect ratio of each wafer sub-region from wafer sub-region information. Combined with a loading effect model, it calculates the corresponding etching depth attenuation coefficient, achieving precise quantification of etching attenuation levels in different wafer sub-regions. This provides individualized attenuation basis for subsequent parameter control. Substituting the etching depth attenuation coefficient into the etching depth influence model, the required plasma characteristic parameters (such as ion energy, flux, and RF power) for each wafer sub-region are directly derived, constructing a closed-loop quantitative link of "pattern loading → etching attenuation → process parameters." This solves the problem of traditional plasma parameter adjustment relying on experience and lacking precise data support, ensuring a high degree of match between parameter configuration and etching attenuation requirements. It calculates and configures exclusive plasma characteristic parameters for each wafer sub-region, replacing the traditional coarse-grained control method of globally uniform parameters. This achieves precise compensation for etching depth attenuation regions and parameter optimization for normal regions, comprehensively improving the uniformity of etching depth across the entire wafer from the process parameter level.

[0027] In this embodiment, an etching depth influence model is constructed based on the classical ion-assisted etching (IAE) model: In the formula, For etching depth, For etching time, For electron density, It is a proportionality constant. To influence the index; Understandably, at the macroscopic level, electron density determines the efficiency of the generation of reactive free radicals in the plasma, which is the source of etching reactants, i.e., the raw materials for etching. Substituting historical data into the etching depth influence model yields the proportionality constant. and impact index ; Substituting the etching depth attenuation coefficient into the etching depth influence model, we obtain: In the formula, The corrected electron density, The initial electron density, The pattern density of the wafer sub-region; Simplifying the above formula, we get: For ion energy, pattern density does not affect ion energy. Pattern density only changes the amount of ion flux, not the bombardment energy of a single ion. Ion energy determines the vertical penetration capability of ions and directly solves the hysteresis of micro-RIE (insufficient bottom etching at high aspect ratio). Ion energy correction only needs to solve the insufficient bottom etching caused by high aspect ratio and is unrelated to pattern density. Therefore, it is only corrected according to aspect ratio. Etching rate ,Right now: In the formula, Etching rate, For ion energy, This is the etching threshold energy. Below this energy, ions cannot effectively break chemical bonds or sputter atoms, and etching hardly occurs. This refers to the ion energy index. For constant terms; The etching threshold energy is specifically: In the formula, The relative atomic mass of the sputtered ions. The relative atomic mass of the target atom; Acquire historical etching rate data and corresponding ion energy data, and substitute them into the fitting to obtain the ion energy index; Based on the ion energy index, we obtain: In the formula, This is the corrected ion energy. The initial ion energy; After simplification, we get: Regarding electron temperature, it regulates the efficiency of free radical generation. ,Right now: In the formula, This is the gas dissociation energy. For electron temperature; Electron temperature determines the average energy of ions, directly controlling sidewall etching damage. Aspect ratio does not affect electron temperature; it only changes whether ions can reach the bottom, without altering the electron temperature of the plasma itself, and does not cause sidewall damage. In the formula, The corrected electron temperature. This is the initial electron temperature; Ion flux is determined by electron density and electron temperature and does not require separate adjustment. Specifically, ion flux is as follows: In the formula, For electron density, For electron temperature, For ion mass, is the Boltzmann constant.

[0028] Based on plasma characteristic parameters, wafer etching is performed on each wafer sub-region.

[0029] Reference Figure 5 As shown, further, combining the above-mentioned method for improving wafer etching uniformity based on closed-loop control of plasma parameters, a wafer etching uniformity improvement system based on closed-loop control of plasma parameters is proposed, including: The main control module is used to construct an etching depth influence model based on historical wafer etching data, with plasma characteristic parameters as independent variables and etching depth as dependent variable. Based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, the module obtains the initial plasma characteristic parameters corresponding to each wafer sub-region. Based on the wafer sub-region information and the initial plasma characteristic parameters, the module obtains the plasma characteristic parameters based on the etching depth influence model. The information acquisition module is used to acquire wafer information to be processed, including wafer size data and target etching parameter information. Based on the wafer size data of the wafer information to be processed, the module acquires corresponding historical wafer etching data. The region division module is used to obtain radial and angular base values ​​based on wafer etching history data, and to divide the wafer to be processed into regions based on the radial and angular base values ​​according to the wafer information, thereby obtaining wafer sub-region information.

[0030] The main control module specifically includes: The control unit is used to obtain the initial plasma characteristic parameters corresponding to each wafer sub-region based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, and to obtain the plasma characteristic parameters based on the etching depth influence model based on the wafer sub-region information and the initial plasma characteristic parameters. An information receiving unit interacts with an information acquisition module and a region division module to receive data and transmit it to a control unit. The model building unit is used to build an etching depth influence model based on historical wafer etching data, with plasma characteristic parameters as independent variables and etching depth as dependent variable.

[0031] The information acquisition module specifically includes: The first acquisition unit is used to acquire wafer information to be processed, the wafer information to be processed including wafer size data and target etching parameter information; The second acquisition unit is used to acquire corresponding wafer etching history data based on the wafer size data of the wafer information to be processed.

[0032] The region division module specifically includes: The data analysis unit is used to obtain radial and angular baseline values ​​based on wafer etching history data. The region division unit is used to divide the wafer to be processed into regions based on radial and angular base values ​​according to the wafer information to be processed, and to obtain wafer sub-region information.

[0033] In summary, the advantages of this invention are as follows: by cleaning historical wafer etching data, removing outliers, and constructing a load effect model, accurate calculation of theoretical etching depth is achieved; by extracting radial and angular baseline values ​​through polar coordinate mapping, Gaussian fitting, and discrete Fourier transform, quantitative characterization of the radial and angular fluctuation characteristics of wafer etching is achieved; by combining etching depth deviation threshold, pattern density, and aspect ratio for partitioning, adaptive and accurate division of wafer sub-regions is achieved; and by differentiating the etching attenuation coefficient of sub-regions to correct plasma characteristic parameters, uniform compensation of the global etching depth of the wafer is achieved, effectively improving the uniformity of wafer etching.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for improving wafer etching uniformity based on closed-loop control of plasma parameters, characterized in that, include: Obtain information about the wafer to be processed, including wafer size data and target etching parameters; Based on the wafer size data of the wafer information to be processed, obtain the corresponding wafer etching history data; Based on historical wafer etching data, obtain the radial and angular baseline values; Based on the information of the wafer to be processed, the wafer to be processed is divided into regions using radial and angular base values ​​to obtain wafer sub-region information. Based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, the initial plasma characteristic parameters corresponding to each wafer sub-region are obtained. The plasma characteristic parameters include electron density, electron temperature, ion flux and ion energy. Based on historical wafer etching data, an etching depth influence model was constructed with plasma characteristic parameters as independent variables and etching depth as dependent variable. Based on the wafer sub-region information and initial plasma characteristic parameters, plasma characteristic parameters are obtained using the etching depth influence model. Based on plasma characteristic parameters, wafer etching is performed on each wafer sub-region; The process of obtaining radial and angular baseline values ​​based on wafer etching history data specifically includes: Based on the wafer size data of the wafer information to be processed, obtain the full-point etching history data of the wafer of the same specification; Based on the historical etching data of all points, after removing abnormal deviations, we obtained an effective etching depth dataset and a load effect model. Based on the effective etching depth dataset, a polar coordinate system is established with the wafer center as the origin. Each detection point is mapped to the radial distance and angular angle in polar coordinates to obtain the polar coordinate mapping data corresponding to each detection point. Based on the polar coordinate mapping data, with the wafer center as the origin, detection points are taken at equal intervals from the inside to the outside along the radial direction as radial feature points. The etching depth of each radial feature point is normalized to obtain the radial etching depth fluctuation sequence. A single-peak Gaussian fitting model is constructed based on the radial etching depth fluctuation sequence; Based on the single-peak Gaussian fitting model, the radial etching depth fluctuation sequence is fitted using the least squares method to obtain the fitted Gaussian curve. Based on the fitted Gaussian curve, the radial distance between two adjacent Gaussian peaks is extracted using a peak detection algorithm. The difference in radial distances between two adjacent Gaussian peaks is taken as the radial baseline value; Based on polar coordinate mapping data, corner feature points are extracted at equal intervals along the circumferential corner direction with the wafer center as the pole and a fixed angle as the step size to obtain the corner etching depth fluctuation sequence. Based on the angular etching depth fluctuation sequence, a frequency domain complex numerical sequence is obtained using discrete Fourier transform; Based on the frequency domain complex numerical sequence, the amplitude spectrum of each frequency domain component is calculated to obtain the amplitude spectrum data of the angular etching depth fluctuation. Based on the angular wave amplitude spectrum data, obtain the dominant frequency domain component with the largest amplitude; The ratio of the main frequency domain component to the total number of angular feature points is taken as the angular fluctuation main frequency, which represents the number of angular fluctuations per unit angle along the etching depth. The ratio of the total circumferential angle of the wafer to the dominant frequency of angular oscillation is used as the basic value for the angular direction.

2. The method for improving wafer etching uniformity based on closed-loop control of plasma parameters according to claim 1, characterized in that, The step of dividing the wafer to be processed into regions based on radial and angular baseline values, and obtaining wafer sub-region information, specifically includes: S100: Based on wafer etching requirements, obtain an etching depth deviation threshold, wherein the etching depth deviation threshold represents the maximum allowable etching depth deviation value; S200: Based on the wafer information to be processed and the wafer layout design, obtain the pattern structure information and target etching depth information; S300: Based on the pattern structure information, obtain the pattern density information and aspect ratio of the wafer to be processed, wherein the aspect ratio represents the ratio of the target etching depth to the opening width corresponding to the pattern structure. S400: Based on the pattern density information and aspect ratio, and using the loading effect model, obtain the predicted etching depth at any position on the wafer to be processed, and use the difference between the predicted etching depth and the target etching depth as the predicted etching depth deviation value. S500: Using the center of the wafer to be processed as the origin, traverse radially from the inside to the outside until the predicted etching depth deviation exceeds the etching depth deviation threshold, or the distance between the current position and the center reaches the radial baseline value, and take the current position as the radial reference point. S600: Using the radial reference point as the new origin, repeat step S500 until the entire wafer to be processed is traversed to obtain radial reference point information. S700: Based on the center of the wafer to be processed, and with the distance from the center to the radial reference point as the radius, the annular zone is divided, and the line connecting the center and the radial reference point is used as the radial reference line for each annular zone. S800: For each annular region, based on the radial baseline, traverse the annular regions on both sides of the radial baseline along the angular direction until the predicted etching depth deviation exceeds the etching depth deviation threshold, or the angle between the line connecting the current position and the center of the circle and the radial baseline reaches the angular baseline value, and take the current position as the angular reference point. S900: Take the line connecting the center of the wafer to be processed and the angular reference point as the new radial reference line, repeat step S800 until the entire annular zone is traversed, divide the annular zone with the radial reference line, and obtain wafer sub-region information.

3. The method for improving wafer etching uniformity based on closed-loop control of plasma parameters according to claim 2, characterized in that, The effective etching depth dataset obtained by removing outlier values ​​from the historical etching data of all points specifically includes: Based on the historical etching data of all points, the historical etching structure data corresponding to each detection point is obtained. The historical etching structure data includes historical opening width data and corresponding historical etching depth data. The etching structure includes trenches, through holes and contact holes. Based on historical etching structure data, the historical opening width is used as the basis to group the historical etching data of all points. Detection points with the same historical opening width are divided into the same group to obtain the historical etching data group of all points. Based on the historical etching data set of all points, obtain the historical etching depth, target etching depth and etching pattern density corresponding to each detection point in each set; Based on historical etching depth, target etching depth, and etching pattern density, the characteristic points of the loading effect are obtained; Based on the characteristic points of the load effect, a load effect model is constructed based on macroscopic and microscopic load effects; Based on the load effect model, the theoretical etching depth corresponding to each detection point is obtained; The standard deviation is obtained based on the residual value between the theoretical etching depth and the historical etching depth corresponding to each detection point; Based on standard deviation, The principle is to obtain the abnormal depth threshold; Based on the abnormal depth threshold, the historical etching data of all points is filtered, and the historical etching data of all points with an etching depth exceeding the abnormal depth threshold is removed to obtain the effective etching depth dataset. Specifically, the load effect model is as follows: In the formula, For theoretical etching depth, For the target etching depth, This is the macroeconomic load effect coefficient. To etch pattern density, This is the micro lag coefficient. This refers to the historical opening width.

4. The method for improving wafer etching uniformity based on closed-loop control of plasma parameters according to claim 3, characterized in that, The process of obtaining load effect feature points based on historical etching depth, target etching depth, and etching pattern density specifically includes: Based on the macroscopic loading effect, data fitting is performed according to historical etching depth, target etching depth and etching pattern density to obtain the loading effect characteristic coefficients corresponding to each full-point etching historical data set. The mean value of the load effect characteristic coefficients corresponding to the historical data set of all etching points is used as the load effect coefficient. Based on the historical etching data of all points, the etching depth deviation coefficient is obtained based on the etching pattern density and loading effect coefficient corresponding to each detection point. The product of the target etching depth and the etching depth deviation coefficient corresponding to each detection point is used as the etching depth threshold corresponding to that detection point. Based on the etching depth threshold, the detection points are divided, and the detection points whose historical etching depth has not exceeded the etching depth threshold are classified as load effect feature points. Specifically, the etching depth deviation coefficient is: In the formula, This is the etching depth deviation coefficient. This is the load effect coefficient. Etching pattern density represents the area of ​​the etched region per unit area. The characteristic coefficient of the load effect. For the target etching depth, Indicates the first Historical etching depth at each detection point Indicates the first The etching pattern density at each detection point.

5. The method for improving wafer etching uniformity based on closed-loop control of plasma parameters according to claim 1, characterized in that, The process of obtaining plasma characteristic parameters based on wafer sub-region information and initial plasma characteristic parameters, using an etching depth influence model, specifically includes: Based on the wafer sub-region information, obtain the pattern density, aspect ratio, and target etching depth of each wafer sub-region; Based on the load effect model, the etching depth attenuation coefficient corresponding to each wafer sub-region is obtained according to the pattern density and aspect ratio of each wafer sub-region. The etching depth attenuation coefficient represents the ratio of the predicted etching depth to the target etching depth. Based on the etching depth attenuation coefficient corresponding to each wafer sub-region, the plasma characteristic parameters are obtained by substituting them into the etching depth influence model.

6. A wafer etching uniformity improvement system based on closed-loop control of plasma parameters, used to implement the uniformity improvement method as described in any one of claims 1-5, characterized in that, include: The main control module is used to construct an etching depth influence model based on historical wafer etching data, with plasma characteristic parameters as independent variables and etching depth as dependent variable. Based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, the module obtains the initial plasma characteristic parameters corresponding to each wafer sub-region. Based on the wafer sub-region information and the initial plasma characteristic parameters, the module obtains the plasma characteristic parameters based on the etching depth influence model. The information acquisition module is used to acquire wafer information to be processed, including wafer size data and target etching parameter information. Based on the wafer size data of the wafer information to be processed, the module acquires corresponding historical wafer etching data. The region division module is used to obtain radial and angular base values ​​based on wafer etching history data, and to divide the wafer to be processed into regions based on the radial and angular base values ​​according to the wafer information, thereby obtaining wafer sub-region information.

7. The wafer etching uniformity improvement system based on closed-loop control of plasma parameters according to claim 6, characterized in that, The main control module specifically includes: The control unit is used to obtain the initial plasma characteristic parameters corresponding to each wafer sub-region based on the target etching parameter information and wafer sub-region information in the wafer information to be processed, and to obtain the plasma characteristic parameters based on the etching depth influence model based on the wafer sub-region information and the initial plasma characteristic parameters. An information receiving unit interacts with an information acquisition module and a region division module to receive data and transmit it to a control unit. The model building unit is used to build an etching depth influence model based on historical wafer etching data, with plasma characteristic parameters as independent variables and etching depth as dependent variable.

8. The wafer etching uniformity improvement system based on closed-loop control of plasma parameters according to claim 6, characterized in that, The information acquisition module specifically includes: The first acquisition unit is used to acquire wafer information to be processed, the wafer information to be processed including wafer size data and target etching parameter information; The second acquisition unit is used to acquire corresponding wafer etching history data based on the wafer size data of the wafer information to be processed.

9. The wafer etching uniformity improvement system based on closed-loop control of plasma parameters according to claim 6, characterized in that, The region division module specifically includes: The data analysis unit is used to obtain radial and angular baseline values ​​based on wafer etching history data. The region division unit is used to divide the wafer to be processed into regions based on radial and angular base values ​​according to the wafer information to be processed, and to obtain wafer sub-region information.

Images