Automatic photomask and reticle inspection method and apparatus including improved defect detector and alignment sub-systems

• US 4,805,123 A
• Filed: 07/14/1986
• Issued: 02/14/1989
• Est. Priority Date: 07/14/1986
• Status: Expired due to Term

• Current Assignee: KLA Instruments, Corporation (KLA Corporation)
• Original Assignee: KLA Instruments, Corporation (KLA Corporation)
• Inventors: Young, Scott A., Lutzker, Matthew B., Hager, James J. Jr., Wihl, Tim S., Specht, Donald F.
• Primary Examiner(s): Gruber, Felix D.
• Assistant Examiner(s): Dixon, Joseph L.
• Application Number:
  US06/885,197
• Time in Patent Office: 946 Days
• Field of Search: 364/518, 364/559, 356/394, 382/8, 382/34, 382/48, 382/45, 382/46, 358/105, 358/106, 358/160, 358/166
• US Class Current: 382/144
• CPC Class Codes: G01N 21/95607 using a comparative method
• CPC Class Codes: G06T 2207/10056 Microscopic image
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/001 using an image reference approach
• CPC Class Codes: G06T 7/32 using correlation-based methods

Abstract

A photomask and reticle inspection method and apparatus wherein a selected surface area of an object is inspected and a first stream of data having signal values representing the image content of each pixel thereof is generated, a second stream of data having signal values representing the intended image content of each pixel of the first stream of data is generated, corresponding portions of the first and second streams of data are stored in memory, any misalignment between the stored portions of the first and second streams of data is detected, the misaligned first and second portions of data are then aligned using shifts of an integral number of pixels and/or subpixel interpolation to correct the detected misalignment therebetween, corresponding subportions of the stored and aligned first and second portions of data are then compared to detect difference therebetween, and upon detecting a difference exceeding a predetermined threshold, the presence of a defect at a particular pixel location on the inspected object is indicated.

Claims

1. A method of detecting defects in objects such as photomasks, reticles, wafers, printed circuit boards, and the like, comprising the steps of:

inspecting a selected surface area of an object and generating a first stream of data having signal values representing the image content of each pixel thereof;

generating a second stream of data having signal values representing the image content of each pixel of said second stream of data;

storing corresponding portions of the first and second streams of data in memory;

detecting with a resolution to a fraction of a pixel any misalignment between the stored portions of said first and second streams of data;

aligning the stored first and second portions of data using subpixel interpolation to correct any detected misalignment therebetween; and

comparing corresponding subportions of the stored and aligned first and second portions of data to detect differences therebetween, and upon detecting a difference, indicating the presence of a defect at a particular pixel location on the inspected object.

2. A method as recited in claim 1 wherein the aligning of the stored first and second portions of data is further accomplished by shifting at least one of said stored first and second portions of data an integral number of pixels.

3. A method as recited in claim 2 wherein the detection of misalignment of the stored first and second portions of data is accomplished using a sum of squared difference algorithm to dynamically determine any registration error therebetween.

4. A method as recited in claim 3 wherein the detection of misalignment of the stored first and second portions of data is accomplished by computing the sum of squared differences of the compared subportions of data shifted integral pixel amount relative to one another, finding the shift which minimizes the sum of squared differences, and interpolating between integer shifts to find the optimum subpixel shift.

5. A method as recited in claim 3 wherein said sum of squared difference algorithm includes the comparison of first data from a particular array of pixels in said first portion of data to second data from each of a plurality of iterations of a corresponding array of pixels from said second portion of data, each iteration being shifted a selected predetermined increment in orthogonal X and/or Y directions relative to said corresponding array of pixels to perform a sum of squared difference operation between said first data and each of said second data so as to dynamically determined which shifted position of said second data has the best correlations to said first data.

6. A method as recited in claim 5 wherein said particular array is an nxm array of pixel data from said first portion of data, and said plurality of iterations are arrays of shifted data derived from a corresponding nxm array of pixel data from said second portion of data, each said shifted array being shifted a different increment relative to said corresponding nxm array, and wherein the minimum value array the sum of squared differences computed from said particular array and said nxm shifted arrays indicates the X and Y shifts required for alignment of said first and second portions of data, said X and Y shifts being used to align the stored portions of data.

7. A method as recited in claim 6, wherein n equals 5, m equals 480, and wherein said different increments are selected from the series -7, -6, -5 . . . 0 . . . 5, 6, 7.

8. A method as recited in claim 7 wherein the comparing of corresponding subportions of said aligned first and second portions of data is accomplished by correlating data corresponding to a particular pxp array of pixel data from said aligned first portion of data to a plurality of pxp arrays of shifted data derived from a corresponding pxp array of pixel data from said aligned second portion of data, each said shifted pxp array being shifted a different subpixel increment relative to said corresponding pxp array, and developing corresponding correlation values, and then comparing said correlation values to a predetermined threshold to detect the presence of a defect.

9. A method as recited in claim 8, wherein p equals 5, and wherein said different subpixel increments are m×

1/3 pixel where m is selected from the series -2, -1, 0, 1, 2.

10. A method as recited in claim 7, wherein the comparing of corresponding subportions of said aligned first and second portions of data is accomplished by an area subtraction algorithm in which pixel data representing a particular q×

q array of pixel data from said aligned first portion of data is added together and subtracted from sum of pixel data from a shifted q×

q array of pixel data from said second stored portion of data shifted a subpixel increment relative to a q×

q array corresponding to said particular q×

q array, and wherein the result of such subtraction is compared to a threshold, such that if said result exceeds said threshold, the presence of an error is indicated.

11. A method as recited in claim 10, wherein q equals 2, and wherein said subpixel increment is r×

1/16 pixel where r is selected from the series 0, 1, 2, . . . 15.

12. A method as recited in claim 1 wherein the detection of misalignment of the stored first and second portions of data is accomplished by computing the maximum of the cross correlation function and then performing a commensurate shift of at least one of said stored first and second portions of data to achieve alignment.

13. A method as recited in claim 1, wherein the stored first and second portions of data are aligned by shifting one portion with respect to the other in integral pixel shifts accomplished by displacing addresses in memory, shifts of less than one pixel being accomplished by subpixel interpolation.

14. A method as recited in claim 13 wherein the type of interpolation used is selected from the group consisting of;

(a) linear interpolation,

(b) s-curve interpolation,

(c) fitting a polynomial or other general function to several pixels, and

(d) sin x/x interpolation or an approximation thereof.

15. A method as recited in claim 1 wherein the comparing of corresponding subportions of said aligned first and second portions of data is accomplished by correlating of data corresponding to a particular pxp array of pixel data from said aligned first portion of data to a plurality of pxp arrays of shifted data derived from a corresponding pxp array of pixel data from said aligned second portion of data, each said shifted pxp array being shifted a different subpixel increment relative to said corresponding pxp array, and developing corresponding correlation values, and then comparing said correlating values to a predetermined threshold to detect the presence of a defect.

16. A method as recited in claim 15, wherein p equals 5, and wherein said different subpixel increments are m×

1/3 pixel where m is selected from the series -2, -1, 0, 1, 2.

17. A method as recited in claim 1, wherein the comparing of corresponding subportions of said aligned first and second portions of data is accomplished by an area subtraction algorithm in which pixel data representing a particular qxq array of pixel data from said aligned first portion of data is added together and subtracted from sum of pixel data from a corresponding qxq array of pixel data from said second stored portion of data shifted a subpixel increment relative to said corresponding qxq array, and wherein the result of such subtraction is compared to a threshold, such that if said result exceeds said threshold, the presence of an error is indicated.

18. A method as recited in claim 17, wherein q equals 2, and wherein said different subpixel increments are m×

1/16 pixel where r is selected from the series 0, 1, 2, . . . 15.

19. A method as recited in claim 1 and further comprising the step of:

performing interpolations in orthogonal X and Y directions on said first stream of data, before said portions thereof are stored in memory, to correct geometrical distortions of the image data, and

delaying said second stream of data for a period commensurate with the time required to correct said first stream of data.

20. A method as recited in claim 19, wherein said interpolations in said X and Y directions are made in response to adjustment parameters previously determined during a calibration operation.

21. A method as recited in claim 20, wherein said first stream of data is first interpolated in the X direction to correct for rotation of a sensor array and the results thereof are then interpolated in the Y direction to correct for said magnification and distortion errors.

22. A method as recited in claim 20, wherein said first stream of data is first interpolated in the Y direction to correct for magnification and distortion errors and the results thereof are then interpolated in the X direction to correct for rotation of a sensor array.

23. A method as recited in claim 1, wherein the said first and second surface areas of an object are inspected using a pair of electro-optical systems and said first and second streams of data respectively correspond thereto, and further comprising the steps of:

developing orthogonal X and Y interpolation data corresponding to the detection characteristics of said pair of electro-optical systems, where any combination of X and Y interpolation data uniquely defines joint interrelational detection characteristics of the electro-optical systems as they relate to optical and mechanical distortion of detected pixel data passing through said electro-optical systems;

using said X interpolation data to perform a first dynamic interpolation of data in said first stream of data before said portions thereof are stored in memory;

using said Y interpolation data to perform a second dynamic interpolation of the X interpolated data in said first stream before said portions thereof are stored in memory; and

delaying said second stream of data for a period of time such that the delayed second stream of data is time coincident with the X and Y interpolated first stream of data.

24. A method as recited in claim 1 and further comprising the steps of:

performing a critical dimension correction on said first and second streams of data using an interpolation scheme which adjusts the data in at least one of said data streams so that the dimensions of object features represented in the two streams are substantially equal.

25. A method as recited in claim 24, wherein said critical dimension correction is performed by measuring the percentage of corresponding large areas of image data covered by said object features in said first and second data streams, and measuring the respective perimeters of such object features, and then dividing the measured value of said area by the measured value of said perimeters to obtain a measure of the dimensional change required to cause corresponding data in both streams to match, and adjusting the data in at least one of said data streams to effect said match.

26. A method as recited in claim 25, wherein said adjustments are made by digitally remapping the pixel grey level data up or down in magnitude to dynamically cancel the effects of variations in linewidths due to process variations.

27. A method as recited in claim 26, wherein said digital remapping is accomplished using an S-curve interpolation algorithm in which the magnitude of said pixel grey level data is adjusted in a predetermined manner depending on where the corresponding pixel falls on an edge thereby effectively increasing or decreasing the linewidth.

28. A method as recited in claim 1 wherein the said inspecting step is accomplished using a pair of optical systems which generate said first and second streams of data, and wherein said first and second streams of data relate to corresponding surface areas of objects having supposedly identical features represented by data of a first signal level, background represented by data of a second signal level, and feature edges represented by data having signal levels in between said first level and said second level, comprising:

mapping the edge data of said first data stream;

mapping the edge data of said second data stream;

detecting the positional relationship of corresponding edge data in said first and second data streams; and

using the detected positional relationship to remap the edge data of at least one of said data streams so that edge data of the remapped data stream is more nearly positionally coincident with the corresponding edge data of the other data stream.

29. An Apparatus for inspecting objects such as photomasks and the like for defects, comprising:

means for inspecting a selected surface area of an object and for generating a first stream of data having signal values representing the image content of each pixel thereof;

means for generating a second stream of data having signal values representing the image content of each pixel of said second stream of data;

memory means for storing corresponding portions of said first and second streams of data;

detector means for detecting with resolution to a fraction of a pixel any misalignment between the stored first and second portions of data;

alignment means using subpixel interpolation to correct any detected misalignment in the stored first and second portions of data; and

detector means for comparing corresponding subportions of the aligned first and second portions of data to detect any difference therebetween, and upon detecting such difference, for indicating the presence of a defect at a particular pixel location on the inspected object.

30. Apparatus as recited in claim 29 wherein said alignment means includes logic circuitry for implementing a sum of squared difference algorithm to dynamically determine any registration error between the stored firs and second subportions of data.

31. Apparatus as recited in claim 30 wherein said logic circuitry includes

means for generating sum of squared difference data relating to a particular nxm array of pixel data from said first portion of data and a corresponding nxm array of pixel data from said second portion of data, and for further generating sum of squared difference data relating to said particular nxm array and a plurality of shifted iterations of said corresponding nxm array, each said shifted nxm array being shifted a pixel integer increment in X and/or Y directions relative to said corresponding nxm array, and

wherein said alignment means further includes processor means for using said sum of squared difference data to correct said detected misalignment.

32. Apparatus as recited in claim 31 wherein said processor means includes means for determining the pixel shift which minimizes the sum of squared differences, and means for interpolating between integer shifts to find the optimum subpixel shifts.

33. Apparatus as recited in claim 32 wherein said alignment means causes the stored first and second portions of data to be aligned by shifting one portion with respect to the other an amount determined by said alignment means, integral pixel shifts being accomplished by displacing addresses in memory means, and shifts of less than one pixel in X and Y being accomplished by interpolation.

34. Apparatus as recited in claim 33 wherein said alignment means includes interpolation means for implementing interpolation of a type selected from the group including;

(a) linear interpolation,

(b) s-curve interpolation,

(c) fitting a polynomial or other general function to several pixels, and

(d) sin x/x interpolation or an approximation thereof.

35. Apparatus as recited in claim 31 wherein n equals 5, m=480 and wherein said integer shifts are selected from the series -7, -6, -5 . . . 0 . . . 5, 6, 7.

36. Apparatus as recited in claim 31, wherein said detector means includes

means for generating pixel data corresponding to a particular pxp array of said first portion of data and a corresponding pxp array of pixel data from said second portion of data, and for further generating sum of squared difference data relating to said particular pxp array and a plurality of shifted iterations of said corresponding pxp array, each said shifted pxp array being shifted a pixel integer increment in X and/or Y directions relative to said corresponding p x p array, and

means for comparing said sum of squared difference data to a selected threshold value and for indicating a defect if all of said sum of squared difference data exceeds said threshold.

37. Apparatus as recited in claim 36, wherein p equals 5, and wherein said different subpixel increments are m×

1/3 pixel where m is selected from the series -2, -1, 0, 1, 2.

38. Apparatus as recited in claim 31 wherein said detector means includes

logic circuitry for implementing an area subtraction algorithm in which pixel data representing a particular q×

q array of pixel data from said aligned first portion of data is added together and subtracted from sum of pixel data from a shifted q×

q array of pixel data from said second stored portion of data shifted a subpixel increment relative to a q×

q array corresponding to said particular q×

q array and

means for comparing the result of such subtraction to a threshold, such that if said result exceeds said threshold, the presence of a defect is indicated.

39. Apparatus as recited in claim 38, wherein q equals 2, and wherein said subpixel increment is r×

1/16 pixel where r is selected from the series 0, 1, 2, . . . 15.

40. Apparatus as recited in claim 29, wherein said detector means includes

means for generating pixel data corresponding to a particular p×

p array of said first portion of data and a corresponding p×

p array of pixel data from said second portion of data, and for further generating sum of squared difference data relating to said particular p×

p array and a plurality of shifted iterations of said corresponding p×

p array, each said shifted p×

p array being shifted an integer pixel increment in X and/or Y directions relative to aid corresponding p×

p array, and

means for comparing said sum of squared difference data to a selected threshold value and for indicating a defect if all of said sum of squared difference data exceed said threshold.

41. Apparatus as recited in claim 29 wherein said detector means includes

logic circuitry for implementing an area subtraction algorithm in which pixel data representing a particular q×

q array of pixel data from said aligned first portion of data is added together and subtracted from sum of pixel data from a shifted q×

q array of pixel data from said second stored portion of data shifted a subpixel increment relative to a q×

q array corresponding to said particular q×

q array and

means for comparing the result of such subtraction to a threshold, such that if said result exceeds said threshold, the presence of an error is indicated.

42. Apparatus as recited in claim 29 wherein at least said means for inspecting includes an optical system and associated photodetector means for generating said first stream of data, and further comprising:

means for performing interpolations in orthogonal X and Y directions on said first stream of data before said portions thereof are stored in said memory means, said interpolations being operative to correct for opto-mechanical distortion of said data; and

means for delaying said second stream of data for a period commensurate with the time required to correct said first stream of data.

43. Apparatus as recited in claim 29 wherein said means for generating the second stream of data having signal values representing the image content of each pixel thereof is a data processing system designed to produce from a database ideal image data against which the first stream of data can be compared.

44. Apparatus as recited in claim 29 wherein at least one of said means for inspecting and said means for generating includes an optical system and associated photodetection means for generating said streams of data, and further comprising:

means for developing orthogonal X and Y interpolation data corresponding to the detection characteristics of said optical system, where any combination of X and Y interpolation data uniquely defines the optical characteristics of the optical system as they relate to optical and mechanical distortion of detected pixel data obtained from corresponding pixel areas of said object;

means for using said X interpolation data to perform a first dynamic interpolation of data in said first stream of data before said portion thereof is stored in said memory means;

means for using said Y interpolation data to perform a second dynamic interpolation of the X interpolated data in said first stream before said portion thereof is stored in said memory means; and

means for delaying said second stream of data for a period of time such that the delayed second stream of data is time coincident with the X and Y interpolated first stream of data.

45. Apparatus as recited in claim 29 and further comprising:

means for performing a critical dimension correction on said first and second streams of data by adjusting the data in at least one of said data streams so that the dimensions of object features represented in the two streams are substantially equal.

46. Apparatus as recited in claim 45, wherein said means for performing a critical dimension correction includes means for measuring the percentage of corresponding large areas of image data covered by said object features in said first and second data streams, and for measuring the respective perimeteres of such object features, and then dividing the measured value of said area by the measured values of said perimeters to obtain a measure of the dimensional change required to cause data in both streams to match, and means for adjusting the data in at least one of said data streams to effect said match.

47. Apparatus as recited in claim 46 wherein said means for adjusting includes means for digitally remapping up or down in magnitude the pixel grey level data representing said edges to dynamically cancel the effects of variations in linewidths due to process variations.

48. Apparatus as recited in claim 47 wherein said means for remapping uses an S-curve interpolation algorithm in which the magnitude of said pixel grey level data is adjusted in a predetermined manner depending on where a particular pixel falls on an edge, thereby effectively increasing or decreasing the linewidth.

49. Apparatus as recited in claim 46 wherein the detected positional relationship is used to remap the corresponding edge data of each data stream so that each is more nearly positionally coincident with the remapped edge data of the other.

50. Apparatus as recited in claim 49 wherein the remapping operation is accomplished by means of an S-curve algorithm in which the signal level of edge data corresponding to each pixel of an edge transition is adjusted in magnitude by a predetermined increment depending on the position of the pixel in the edge transition and the direction of the remapping operation.

51. A method of detecting defects in objects such as photomasks, reticles, wafers, printed circuit boards, and the like, comprising the steps of:

inspecting a selected surface area of an object and generating a first stream of data having signal values representing the image content of each pixel thereof;

generating a second stream of data having signal values representing the image content of each pixel of said second stream of data;

storing corresponding portions of the first and second streams of data in memory;

detecting with a resolution to an integer pixel any misalignment between the stored portions of said first and second streams of data;

aligning the stored first and second portions of data to the closest integer pixel to correct any detected misalignment therebetween; and

comparing corresponding subportions of the stored and aligned first and second portions of data to detect differences therebetween by correlating data corresponding to a particular pxp array of pixel data from said aligned first portion of data to a plurality of p×

p arrays of shifted data derived from a corresponding p×

p array of pixel data from said aligned second portion of data, each said shifted pxp array being shifted a different subpixel increment relative to said corresponding p×

p array and developing corresponding correlation values, and then comparing said correlation values to a predetermined threshold to detect the presence of a defect.

52. A method of detecting defects in objects such as photomasks, reticles, wafers, printed circuit boards, and the like, comprising the steps of:

using a first electro-optical system to inspect a first selected surface area of an object and to generate a first stream of data having signals values representing the image content of each pixel thereof;

using a second electro-optical system to inspect a second surface area of said object and to generate a second stream of data having signal values representing the image content of each pixel of said second stream of data;

storing corresponding portions of the first and second streams of data in memory;

detecting with a resolution to a fraction of a pixel any misalignment between the stored portions of said first and second streams of data;

aligning the stored first and second portions of data using subpixel interpolation to correct any detected misalignment therebetween; and

comparing corresponding subportions of the stored and aligned first and second portions of data to detect differences therebetween by correlating data corresponding to a particular p×

p array of pixel data from said aligned first portion of data to a plurality of p×

p arrays of shifted data derived from a corresponding p×

p array of pixel data from said aligned second portion of data, each said shifted p×

p array being shifted a different subpixel increment relative to said corresponding p×

p array and developing corresponding correlation values, and then comparing said correlation values to a predetermined threshold to detect the presence of a defect.