Photomask inspection apparatus and method with improved defect detection

• US 4,579,455 A
• Filed: 05/09/1983
• Issued: 04/01/1986
• Est. Priority Date: 05/09/1983
• Status: Expired due to Term

• Current Assignee: KLA Instruments, Corporation
• Original Assignee: KLA Instruments, Corporation
• Inventors: Levy, Kenneth, Wihl, Mark J., Buchholz, Steve, Broadbent, William H.
• Primary Examiner(s): Rosenberger, R. A.
• Assistant Examiner(s): Cooper, Crystal
• Application Number:
  US06/492,658
• Time in Patent Office: 1,058 Days
• Field of Search: 356/394, 356/398, 358/106, 358/101, 371/68
• US Class Current: 356/394
• CPC Class Codes: G03F 1/84 Inspecting
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/001 using an image reference approach

Abstract

Apparatus 20 for inspecting photomasks 26 and the like by comparison of duplicate die patterns, including improved defect detection. Two-dimensional pixel representations of two die patterns are formed, with pixels having values or black or white or shades or grey, depending upon the features of the die patterns. Defects in the die patterns are found by a defect detector circuit 60 at points of non-agreement between the pixel representations. Two window matrices 130 and 134 of adjacent pixels are defined for corresponding areas of the two die patterns. The center matrix of each window matrix is defined as a comparison matrix 132 and 136. An error value is computed for subsets of the window matrix by summing the squares of the differences between each of the pixel values of each subset and the corresponding pixel values of the opposite comparison matrix. If there is no defect and any misalignment between the two representations minimal, at least one error value will be less than a threshold error value. If none of the error values are less than the threshold error value, a defect is assumed. The magnitude of the threshold error value is automatically varied according to the number and type of edges within the window matrices to compensate for edge quantization errors of less than one pixel.

Claims

1. A method for detecting defects in a reticle, a metal mask, or other object such as a photomask having a plurality of duplicate two-dimensional patterns thereon, said method being operable for comparing two like patterns and for indicating defects wherever said patterns do not match, said method comprising the steps of:

forming a first representation of a first of said patterns and a second representation of a second of said patterns for comparison to said first representation, each of said patterns being represented by and each of said representations being composed of a two-dimensional array of uniformly spaced apart pixels having relative positions that correspond to spatial positions on the object being inspected, each pixel having a value that indicates the absence or presence of said pattern at the spatial position represented by that pixel;

forming a first comparison matrix of a group of adjacent pixels of said first representation, and forming a first window matrix of a group of adjacent pixels of said second representation, the pixels of said first comparison matrix representing an area of said first pattern that corresponds to a like area of said second pattern that is contained within the area represented by the pixels of said first window matrix, said first comparison matrix being smaller in extent that said first window matrix, said first window matrix having subsets of adjacent pixels that are equivalent in relative orientation to said first comparison matrix and equal in number to the number of unique positions that said first comparison matrix could be positioned within said first window matrix;

forming a second comparison matrix of a group of adjacent pixels of said second representation, and forming a second window matrix of a group of adjacent pixels of said first representation, the pixels of said second comparison matrix representing an area of said second pattern that corresponds to a like area of said first pattern that is contained within the area represented by the pixels of said second window matrix, said second comparison matrix being smaller in extent than said second window matrix, said second window matrix having subsets of adjacent pixels that are equivalent in relative orientation to said second comparison matrix and equal in number to the number of unique positions that said second comparison matrix could be positioned within said second window matrix;

computing first error values that are proportional to the magnitude of the difference between the pixel values of equivalently oriented pixels of said first comparison matrix and said subsets of said first window matrix;

computing second error values that are proportional to the magnitude of the difference between the pixel values of equivalently positioned pixels of said second window matrix and said subsets of said second window matrix;

comparing said first error values to a first error threshold value, and determining that there exists no defect in said first pattern within the area represented by the pixels of said first comparison metrix if at least one of said first error values is less than said first error threshold value; and

comparing said second error values to a second error threshold value, and determining that there exists no defect in said second pattern within the area represented by the pixels of said second comparison matrix if at least one of said second error values is less than said second error threshold value.

2. A method as recited in claim 1 additionally repeating said steps of forming first and second comparison matrices, computing first and second error values therefor, and comparing said error values to error threshold values until the entire area of said first and second representation has been covered.

3. A method as recited in claim 1 wherein said pixels are square in shape, and said first and second comparison and window matrices are also square in shape.

4. A method as recited in claim 3 wherein said first and second window matrices each have seven rows and seven columns of adjacent pixels and said first and second comparison matrices each have three rows and three columns of adjacent pixels.

5. A method as recited in claim 3 wherein said first comparison matrix is centered within said second window matrix, and said second comparison matrix is centered within said first window matrix.

6. A method as recited in claim 1 wherein said pixel values vary from a maximum value indicating the absence of said pattern at said spatial position represented by said pixel to a minimum value indicating the presence of said pattern at said spacial position.

7. A method as recited in claim 6 wherein said pixel values also include intermediate values indicating the presence of a pattern edge at said spatial position represented by said pixel.

8. A method as recited in claim 1 wherein said step of computing said first error values includes computing a first error value for each subset of said first window matrix, each subset having a first error value equal to the sum of the squares of the differences between the pixel value of each of the pixels of said first comparison matrix and the pixel value of the equivalently oriented pixel of said subset.

9. A method as recited in claim 1 wherein said step of computing said second error values includes computing a second error value for each subset of said second window matrix, each subset having a second error value equal to the sum of the squares of the differences between the pixel value of each of the pixels of said second comparison matrix and the pixel value of the equivalently oriented pixel of said subset.

10. A method as recited in claim 1 wherein said first and second error threshold values are variable and are a function of the number and type of edges or other feature of said first and second patterns respectively within said first and second window matrices.

11. A method as recited in claim 10 wherein said first error threshold value equals a minimum value when there are no pattern edges within said first window matrix, said first error threshold value equals a maximum value when there are two types of pattern edges within said first window matrix, and said first error threshold value equals an intermediate value when there is one type of pattern edge within said first window matrix, and wherein said second error threshold value equals said minimum value when there are no pattern edges within said second window matrix, said second error threshold value equals said maximum value when there are two types of pattern edges within said second window matrix, and said second error threshold value equals said intermediate value when there is one type of pattern edge within said second window matrix.

12. A method as recited in claim 1 wherein said patterns are opaque areas in an otherwise transparent object, and said step of forming said first representation of said first pattern and said second representation of said second pattern includes the steps of:

projecting images of portions of said first pattern onto a first one-dimensional array of uniformly spaced apart sensors and projecting images of portions of said second pattern onto a second one-dimensional array of uniformly spaced apart sensors, each of said sensors being operable for detecting opaque and transparent portions of said patterns and also operable for generating sensor signals indicative of the presence of an opaque or transparent portion of said pattern in the image projected thereupon;

recording said sensor signals respectively into first and second pixel memories;

moving said patterns in a direction perpendicular to the length of said sensors to project adjacent portions of said patterns onto said sensors;

recording said sensor signals that correspond to said adjacent portions of said patterns into said pixel memories; and

repeating the previous two steps of moving said patterns and recording said sensor signals for each adjacent portion of said patterns until complete representations of said patterns have been formed and recorded in said pixel memories.

13. A method as recited in claim 12 wherein:

said step of projecting images of portions of said patterns is accomplished by optical lens means;

said patterns are illuminated from the side opposite said sensors such that light passing through the transparent areas is projected onto said sensors, and light incident upon the opaque areas is blocked from reaching said sensors; and

said sensors are photodiodes that are responsive to the intensity of light incident thereupon.

14. A method as recited in claim 12 wherein said step of forming said first comparison and window matrices and of computing first error values occurs concurrently with said steps of moving said patterns and recording said sensor signals.

17. A method is recited in claim 10 wherein said first and second threshold values are selectable from sets of predetermined values and are selected by an operator input threshold select signal.

15. Inspection apparatus for locating and identifying defects in a photomask having multiple die patterns thereon, said apparatus comprising:

carriage means for supporting the photomask and for moving said photomask along an inspection path;

illumination means for illuminating portions of said photomask as it is moved along said inspection path;

optical means for focusing images of portions of two die patterns at a focal plane;

first and second photosensitive detector means disposed at said focal plane for generating first and second sensor signals indicative of the images of said die patterns incident thereon; and

signal processing means for processing said first and second sensor signals to locate and identify defects in said die patterns by finding differences between said first and second sensor signals, said signal processing means including;

first and second pixel memory means for storing said first and second sensor signals to form first and second pixel representations of said die patterns,

defect detection means for finding portions of said first and second representations that do not match sufficiently to indicate defects, said defect detection means having area subtraction means for computing error values that are proportional to the differences in the sensor signal magnitudes between corresponding portions of said first and second pixel representations, and having threshold comparison means for evaluating the magnitude of said error values to determine the existence of defects,

error threshold calculation means for determining threshold values used by said threshold comparison means in the determination of defects, said error threshold calculation means being operable for increasing defect detection sensitivity in areas not prove to false error detection, and for decreasing defect detection sensitivity in other areas to minimize the detection of misalignment errors as defects,

microprocessor means for controlling said signal processing means, and

input/output means for communication of defect information and inspection results.

16. Inspection apparatus for locating and identifying defects in an object such as a reticle, a photomask with multiple openings therein, or a photomask having multiple die patterns thereon, said apparatus comprising:

carriage means for supporting the object and for moving said object along an inspection path;

illumination means for illuminating a portion of said object as it is moved along said inspection path;

optical means for focusing an image of a portion of an object pattern at a focal plane;

first photosensitive detector means disposed at said focal plane for generating a first sensor signal indicative of the image of said object pattern incident thereon;

data base means having pattern data stored therein corresponding to the image patterns of said object, said pattern data being selectively output in the form of second sensor signals corresponding to said first sensor signals; and

signal processing means for processing said first and second signals to locate and identify defects in said object pattern by finding differences between said first and second sensor signals, said signal processing means including;

first and second pixel memory means for storing said first and second sensor signals to form first and second pixel representations of said object pattern and said pattern data,

defect detection means for finding portions of said first and second pixel representations that do not match sufficiently to indicate defects, said defect detection means having area subtraction means for computing error values that are proportional to the differences in the sensor signal magnitudes between corresponding portions of said first and second pixel representations, and having threshold comparison means for evaluating the magnitude of said error values to determine the existence of defects,

error threshold calculation means for determining threshold values used by said threshold comparison means in the determination of defects, said error threshold calculation means being operable for increasing defect detection sensitivity in areas not prove to false error detection, and for decreasing defect detection sensitivity in other areas to minimize the detection of misalignment errors as defects,

microprocessor means for controlling said signal processing means, and

input/output means for communication of defect information and inspection results.