Search

CN-122023246-A - Wafer gumming and developing real-time state detection method

CN122023246ACN 122023246 ACN122023246 ACN 122023246ACN-122023246-A

Abstract

The application provides a method for detecting the real-time state of gluing and developing of a wafer, which comprises the steps of acquiring a wafer edge image from which glue is removed by an image acquisition device when a gluing and developing device executes a edging process on the wafer on a sucker of a gluing cavity of the wafer, determining the fusion gradient amplitude of each pixel of the wafer edge image by an edge positioning algorithm to obtain a gradient amplitude image, determining the outer contour of the wafer and the inner contour of a washing edge by using contour searching based on the gradient amplitude image to determine an edge width value and an error value, and acquiring the wafer surface image by the image acquisition device when the gluing process is finished and detecting defects by an image codec based on a convolutional neural network. Therefore, the production efficiency is quickened and the production cost is reduced by carrying out edge width detection of the edge removing process and defect detection of the gluing process on the wafer in the gluing cavity.

Inventors

  • ZHANG QI
  • FU YOUYIN
  • YANG TAO

Assignees

  • 芯传科半导体科技(江苏)有限公司

Dates

Publication Date
20260512
Application Date
20251223

Claims (10)

  1. 1. A method for detecting the glue spreading and developing real-time state of a wafer comprises the following steps: when the gumming developing device executes a deburring process on the wafer positioned on the sucker of the gumming cavity, the image acquisition device acquires the wafer edge image from which the gum is removed; Determining the fusion gradient amplitude of each pixel of the wafer edge image through an edge positioning algorithm to obtain a gradient amplitude image; determining a wafer outer contour and a line inner contour using contour lookup based on the gradient magnitude image to determine edge width values and error values, and And acquiring a wafer surface image by the image acquisition equipment at the end of the gluing process, and performing defect detection by an image coder-decoder based on a convolutional neural network.
  2. 2. The method for detecting a real-time state of photoresist development of a wafer according to claim 1, wherein a light source is spaced apart from the wafer by 95mm in a direction perpendicular to a moving direction of the wafer, an optical axis of the light source makes an angle of 60 ° with a surface of the wafer, and a thickness of the light source in a direction of the optical axis thereof is 30mm.
  3. 3. The method for detecting a real-time state of a photoresist coating development of a wafer according to claim 2, wherein an optical axis of the image pickup device and an optical axis of the light source vertically intersect at a surface of the wafer, a distance between an incident surface of the image pickup device and the wafer along an optical axis direction thereof is 115mm, and a distance between the incident surface of the image pickup device and the optical axis of the light source in a direction parallel to the surface of the wafer is 100mm.
  4. 4. A glue develop real-time status detection method of a wafer according to claim 3, wherein the distance from the end of the image acquisition device opposite to the light incident surface thereof to the surface of the wafer in the optical axis direction thereof is 280mm, and the distance of the projection of the surface of the wafer onto the plane on which the surface of the wafer is located is 310mm with respect to the most distal end of the wafer.
  5. 5. The method for detecting a photoresist development real-time state of a wafer according to claim 1, wherein determining a fusion gradient magnitude of each pixel of the wafer edge image by an edge localization algorithm to obtain a gradient magnitude image comprises: Based on each pixel in the wafer edge image Horizontal gradient of (2) Vertical gradient Gradient in 45 degree direction And a 135 deg. directional gradient Determining fusion gradient magnitude To obtain the gradient magnitude image, wherein: Wherein the method comprises the steps of Is a pixel Is a gray value of (a).
  6. 6. The method for detecting a photoresist development real-time state of a wafer according to claim 5, wherein determining a fusion gradient magnitude of each pixel of the wafer edge image by an edge localization algorithm to obtain a gradient magnitude image comprises: Performing iterative weighted filtering on the wafer edge image through a weight function constructed based on the gradient amplitude, wherein the weight function is expressed as: Wherein the method comprises the steps of And (1) Pixel values for multiple iterations Expressed as: Wherein the method comprises the steps of Is the first The weight function of the next iteration.
  7. 7. The method of wafer glue develop real time status detection of claim 1, wherein determining the wafer outer contour and the edge wash inner contour using contour lookup based on the gradient magnitude image to determine the edge width value and the error value comprises: performing non-maximum suppression and dual-threshold edge connection on the gradient amplitude image to obtain a binarized edge image, wherein edge pixel points in the binarized edge image are 1, and non-edge points are 0; Determining the outer contour of the wafer and the inner contour of the edge washing line in the binarized edge image by using contour searching; fitting the wafer outer contour and the edge washing line inner contour by adopting a least square method to determine the wafer center and radius and the edge washing line center and radius, and The error value is determined based on a Euclidean distance between the wafer center and the edge wash line center, and an edge width value is determined based on a difference between the wafer radius and the edge wash line radius.
  8. 8. The method for detecting a photoresist-coated developing real-time state of a wafer according to claim 7, wherein determining the wafer outer contour and the line inner contour using contour search in the binarized edge image comprises: Determining the center and radius of the wafer by least square fitting to the wafer outer contour, uniformly taking a plurality of points on the inner contour of the wafer along the radial direction of the wafer, and calculating a plurality of distances from the wafer outer contour along the radial direction of the wafer, and Calculating an average of the plurality of distances to obtain an edge width value, calculating a difference between a maximum distance and a minimum distance to obtain a width uniformity error, and calculating a difference between the edge width value and a target width to obtain an error value.
  9. 9. The wafer glue develop real-time status detection method of claim 1, wherein prior to defect detection by a convolutional neural network based image codec, comprising: And cutting the wafer surface image by adopting a sliding window, scaling the cut image to a preset size, converting the image into a gray scale image, and normalizing the pixel value of the gray scale image from 0-255 to 0-1.
  10. 10. The wafer glue develop real-time status detection method of claim 1, wherein the defect detection by the convolutional neural network based image codec comprises: carrying out 3-layer convolution and pooling operation on the preprocessed wafer surface image through an encoder of an image coder-decoder to obtain an image potential vector; subjecting the image potential vector to a 3-layer deconvolution and upsampling operation by a decoder of an image codec to obtain a reconstructed image, and And determining defects based on the reconstructed image and the residual image of the processed wafer surface image.

Description

Wafer gumming and developing real-time state detection method Technical Field The application relates to the technical field of semiconductor wafer processing, in particular to a method for detecting a glue spreading and developing real-time state of a wafer. Background In the wafer processing process in the semiconductor industry, a photoetching machine is matched with a photoresist coating and developing device, the photoresist coating and developing device firstly uniformly coats a layer of photoresist on the surface of a clean silicon wafer, then the photoetching machine exposes the photoresist, and finally the photoresist coating and developing device removes soluble parts in the photoresist after exposure to form a three-dimensional pattern structure. When the traditional glue coating developing equipment is in normal production, the process is debugged, and after various equipment indexes and process parameters are achieved, the equipment is put into normal operation, and the output of default equipment is good. Or after a batch of products is finished, performing point inspection or full inspection, such as defect detection or overlay accuracy measurement, wherein the mode obviously has feedback hysteresis of process problems. It is therefore desirable to provide a wafer gumming development real-time status detection scheme. Disclosure of Invention The embodiment of the application provides a method for detecting the real-time state of the photoresist coating development of a wafer, which accelerates the production efficiency and reduces the production cost by detecting the edge width of a photoresist coating process and detecting the defects of the photoresist coating process in a photoresist coating cavity. According to one aspect of the application, a method for detecting the real-time state of the gluing and developing of a wafer is provided, which comprises the steps of acquiring a wafer edge image from which glue is removed by an image acquisition device when a gluing and developing device executes a deburring process on a wafer on a sucker of a gluing cavity of the wafer, determining a fusion gradient amplitude value of each pixel of the wafer edge image through an edge positioning algorithm to obtain a gradient amplitude image, determining an outer contour of the wafer and an inner contour of a washing edge by using contour searching based on the gradient amplitude image to determine an edge width value and an error value, and acquiring a wafer surface image by the image acquisition device when the gluing process is finished and performing defect detection through an image codec based on a convolutional neural network. In the method for detecting the real-time state of the photoresist coating development of the wafer, the distance between the light source and the wafer is 95mm in the direction perpendicular to the moving direction of the wafer, the optical axis of the light source forms an angle of 60 degrees with the surface of the wafer, and the thickness of the light source in the direction of the optical axis is 30mm. In the method for detecting the glue spreading and developing real-time state of the wafer, the optical axis of the image acquisition device and the optical axis of the light source vertically intersect on the surface of the wafer, the distance between the light incident surface of the image acquisition device and the wafer along the direction of the optical axis of the light incident surface is 115mm, and the distance between the light incident surface of the image acquisition device and the optical axis of the light source along the direction parallel to the surface of the wafer is 100mm. In the method for detecting the real-time state of the photoresist coating development of the wafer, the distance between the end of the image acquisition device opposite to the light incident surface of the image acquisition device and the surface of the wafer in the optical axis direction of the image acquisition device is 280mm, and the distance between the projection of the surface of the wafer on the plane of the surface of the wafer and the farthest end of the wafer is 310mm. In the method for detecting the real-time state of the photoresist coating development of the wafer, determining the fusion gradient amplitude of each pixel of the wafer edge image by an edge positioning algorithm to obtain a gradient amplitude image comprises the following steps of based on each pixel in the wafer edge imageHorizontal gradient of (2)Vertical gradientGradient in 45 degree directionAnd a 135 deg. directional gradientDetermining fusion gradient magnitudeTo obtain the gradient magnitude image, wherein: Wherein the method comprises the steps of Is a pixelIs a gray value of (a). In the method for detecting the real-time state of the photoresist coating development of the wafer, determining the fusion gradient amplitude of each pixel of the wafer edge image through an edge positioning algorithm to o