Inspection method and apparatus for the inspection of either random or repeating patterns

• US 20040202361A1
• Filed: 02/13/2003
• Published: 10/14/2004
• Est. Priority Date: 09/30/1993
• Status: Active Grant

• Current Assignee: Bin-Ming Ben Tsai, David M.W. Evans, Jason Z. Lin
• Original Assignee: Bin-Ming Ben Tsai, David M.W. Evans, Jason Z. Lin
• Inventors: Evans, David M.W., Lin, Jason Z., Tsai, Bin-Ming Ben
• Granted Patent: US 7,218,768 B2
• Time in Patent Office: Days
• US Class Current: 382/141
• CPC Class Codes: G03F 7/7065 Defects, e.g. optical inspection of patterned layer for defects
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 2207/20056 Discrete and fast Fourier transform, [DFT, FFT]
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/0004 Industrial image inspection
• CPC Class Codes: G06T 7/37 using transform domain methods
• CPC Class Codes: G06V 10/431 Frequency domain transformation; Autocorrelation
• CPC Class Codes: G06V 10/751 Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching

Abstract

The present invention is a hybrid technique for finding defects on digitized device images using a combination of spatial domain and frequency domain techniques. The two dimensional spectra of two images are found using Fourier like transforms. Any strong harmonics in the spectra are removed, using the same spectral filter on both spectra. The images are then aligned, transformed back to the spatial domain, and subtracted. The resulting spectrally-filtered difference image is thresholded and analyzed for defects. Use of the hybrid technique of the present invention to process digitized images results in the highest-performance and most flexible defect detection system. It is the best performer on both array and random devices, and it can cope with problems such as shading variations and the dark-bright problem that no other technique can address. The hybrid technique of the present invention also uses frequency domain techniques to align the images with fewer errors than spatial domain techniques of similar or lesser complexity. Further, the relative offsets of the pairs of images are determined by frequency domain techniques—and this method may be the most accurate and the least expensive.

Claims

1. In an inspection system for detecting defects in devices of any type having either random, repeating or mixed patterns within the same device, said inspection system includes first and second fields of view each positioned along a different optic axis, the first field of view creates a spatial domain pixel map of a portion of a first substrate pattern;

and said second field of view provides a spatial domain pixel map against which said first substrate pattern is to be checked from either a pre-stored pixel map of the desired pattern, a scaled drawing of the desired pattern, a repeating pattern within the same device on said first substrate as in said first field of view, a pattern within a different device on said first substrate as in said first field of view, or a second substrate pattern;

a method for detecting defects in said first substrate pattern in said first field of view comprising the steps of;

a. transforming said pixel maps from said first and second fields of view using a two dimensional Fourier like transform to convert those pixel maps from the spatial domain to the frequency domain;

b. identifying those frequencies in both dimensions having amplitude maxima in the frequency domain of step a. with the repeating pattern portions of each of said pixel maps from said first and second fields of view occurring at those frequencies in the frequency domain;

c. entering zero values for the amplitudes at the frequencies in both dimensions identified in step b. to remove the amplitude at the harmonic frequency values of the repeating portions of said first and second fields of view; and

d. inverse transforming the results from step c. using the inverse of the two dimensional Fourier like transformation of step a. to obtain modified spatial domain pixel maps of said first and second fields of view with the repeating patterns removed.

2. A method as in claim 1 further comprising the steps of:

e. conjugating one of said frequency domain maps of said first and second fields of view from step a.;

f. multiplying the conjugated frequency domain map from step e. with the other frequency domain map of said first and second fields of view in step a. that was not conjugated in step e.;

g. inverse transforming the result of step f. using the inverse of the two dimensional Fourier like transformation of step a. to create the correlation surface of the patterns of said first and second fields of view that is a function of the spatial offset between those patterns;

h. identifying the peaks of the correlation surface of step g. along both the x and y axis, the location of those peaks representing the offset, Δ

x and Δ

y, between the first and second field of view patterns; and

before performing step d.;

i. trigonometrically altering the two dimensional sinusoids of the frequency domain map of the pattern of one of the first and second fields of view from step c. using the x and y axis offset values of step h.

3. A method as in claim 1 further comprising the steps of:

j. subtracting one of said spatial domain pixel maps of step d. from the other of said spatial domain pixel maps of step d. to form a composite spatial domain pixel map that contains only the combined noise and defects of said two spatial domain pixel maps;

k. filtering, individually, each of the modified spatial domain pixel domain maps of steps d. and composite spatial domain map of step j. using a threshold amplitude that is selected to be above the expected noise level and below the expected defect level, to remove defect shadows introduced by the inverse transformation step d. where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, each resulting further modified spatial domain pixel pattern including the random and defects for each of said first and second fields of view and only defects for said composite spatial pixel pattern of step j.; and

l. comparing the three further modified spatial domain pixel patterns from step k. to determine the defects present in said first field of view original pixel image.

4. A method as in claim 3 wherein in step k. the threshold amplitude is selected by an exponential thresholding technique.

5. A method as in claim 3 wherein step l. includes the steps of:

m. comparing said further modified spatial domain pixel patterns of said first and second fields of view to determine the number of amplitude peaks of each of said patterns; and

n. identifying the location of defects as those occurring at the amplitude peaks in said further modified composite spatial domain pixel map from step k. if the number of peaks identified for both said first and second fields of view in step m. are both large numbers indicating that there are numerous random pattern portions in the original first and second fields of view.

6. A method as in claim 3 wherein step l. includes the steps of:

o. comparing said further modified spatial domain pixel patterns of said first and second fields of view to determine the number of amplitude peaks of each of said patterns; and

p. identifying the location of defects as those occurring at the amplitude peaks in said further modified spatial domain pixel map of said first field of view from step k. if the number of peaks identified for both said first and second fields of view in step o. are both small numbers indicating that there are few or no random pattern portions in the original first and second fields of view.

7. A method as in claim 2 further comprising the steps of:

q. subtracting one of said spatial domain pixel maps of step d. from the other of said spatial domain pixel maps of step d. to form a composite spatial domain pixel map that contains only the combined noise and defects of said two spatial domain pixel maps;

r. filtering, individually, each of the modified spatial domain pixel domain maps of steps d. and composite spatial domain map of step q. using a threshold amplitude that is selected to be above the expected noise level and below the expected defect level, to remove defect shadows introduced by the inverse transformation step d. where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, each resulting further modified spatial domain pixel pattern including the random and defects for each of said first and second fields of view and only defects for said composite spatial pixel pattern of step q.; and

s. comparing the three further modified spatial domain pixel patterns from step r. to determine the defects present in said first field of view original pixel image.

8. A method as in claim 7 wherein in step r. the threshold amplitude is selected by an exponential thresholding technique.

9. A method as in claim 7 wherein step s. includes the steps of:

t. comparing said further modified spatial domain pixel patterns of said first and second fields of view to determine the number of amplitude peaks of each of said patterns; and

u. identifying the location of defects as those occurring at the amplitude peaks in said further modified composite spatial domain pixel map from step r. if the number of peaks identified for both said first and second fields of view in step t. are both large numbers indicating that there are numerous random pattern portions in the original first and second fields of view.

10. A method as in claim 7 wherein step s. includes the steps of:

v. comparing said further modified spatial domain pixel patterns of said first and second fields of view to determine the number of amplitude peaks of each of said patterns; and

w. identifying the location of defects as those occurring at the amplitude peaks in said further modified spatial domain pixel map of said first field of view from step r. if the number of peaks identified for both said first and second fields of view in step v. are both small numbers indicating that there are few or no random pattern portions in the original first and second fields of view.

11. A method as in claim 1:

preceding step d. further comprising the step of;

x. subtracting one of said frequency domain pixel maps of step c. from the other of said frequency domain pixel maps of step c. to form a composite frequency domain pixel map that contains only the combined noise and defects of-said two frequency domain pixel maps; and

the method of claim 1 further comprising the steps of;

y. inverse transforming the results from step x. using the inverse of the two dimensional Fourier like transformation of step a. to obtain a modified spatial domain pixel map of said composite pixel map of step x.;

z. filtering, individually, each of the modified spatial domain pixel domain maps of steps d. and composite spatial domain map of step y. using a threshold amplitude that is selected to be above the expected noise level and below the expected defect level, to remove defect shadows introduced by the inverse transformation step d. where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, each resulting further modified spatial domain pixel pattern including the random and defects for each of said first and second fields of view and only defects for said composite spatial pixel pattern of step x.; and

aa. comparing the three further modified spatial domain pixel patterns from step z. to determine the defects present in said first field of view original pixel image.

12. A method as in claim 11 wherein in step z. the threshold amplitude is selected by an exponential thresholding technique.

13. A method as in claim 11 wherein step aa. includes the steps of:

ab. comparing said further modified spatial domain pixel patterns of said first and second fields of view to determine the number of amplitude peaks of each of said patterns; and

ac. identifying the location of defects as those occurring at the amplitude peaks in said further modified composite spatial domain pixel map from step z. if the number of peaks identified for both said first and second fields of view in step ab. are both large numbers indicating that there are numerous random pattern portions in the original first and second fields of view.

14. A method as in claim 11 wherein step aa. includes the steps of:

ad. comparing said further modified spatial domain pixel patterns of said first and second fields of view to determine the number of amplitude peaks of each of said patterns; and

ae. identifying the location of defects as those occurring at the amplitude peaks in said further modified spatial domain pixel map of said first field of view from step z. if the number of peaks identified for both said first and second fields of view in step z. are both small numbers indicating that there are few or no random pattern portions in the original first and second fields of view.

15. A method as in claim 1 wherein said two dimensional Fourier like transform of step a. and the inverse transform of step d. are digital transforms.

16. A method as in claim 1 preceding step d. further including the step of:

af. low pass filtering, individually, each of said first and second fields of view from step c. along both the f_x and f_y axis at frequencies below the lowest frequencies at which zero values were entered in step c. to remove the harmonic echoes of the random patterns, defects and noise that will occur at values of x and y in the spatial domain below the locations of the first location of the repeating pattern removed in step c. when step d. is performed.

17. A method as in claim 1 preceding step d. further including the step of:

ag. filtering, individually, each of said first and second fields of view from step c. along both the f_x and f_y axis for compensating the system modular transfer function.

18. In an inspection system for detecting defects in devices of any type having either random, repeating or mixed patterns within the same device, said inspection system includes first and second fields of view each positioned along a different optic axis, the first field of view creates a spatial domain pixel map of a portion of a first substrate pattern;

and said second field of view provides a spatial domain pixel map against which said first substrate pattern is to be checked from either a pre-stored pixel map of the desired pattern, a scaled drawing of the desired pattern, a repeating pattern within the same device on said first substrate as in said first field of view, a pattern within a different device on said first substrate as in said first field of view, or a second substrate pattern;

a method for detecting defects in said first substrate pattern in said first field of view comprising the steps of;

a. individually identifying and removing the repeating pattern portion of each of said first and second spatial domain pixel maps in the frequency domain to create modified first and second spatial domain pixel maps;

b. individually identifying and removing the random pattern portion of each of said modified first and second spatial domain pixel maps in the spatial domain to create further modified first and second spatial domain pixel maps; and

c. individually and collectively examining said further modified first and second spatial domain pixel maps in the spatial domain.

19. In an inspection system for detecting defects in devices of any type having either random, repeating or mixed patterns within the same device, said inspection system includes first and second fields of view each positioned along a different optic axis, the first field of view creates a first spatial domain pixel map of a portion of a first substrate pattern;

and said second field of view provides a second spatial domain pixel map against which said first substrate pattern is to be checked from either a pre-stored pixel map of the desired pattern, a scaled drawing of the desired pattern, a repeating pattern within the same device on said first substrate as in said first field of view, a pattern within a different device on said first substrate as in said first field of view, or a second substrate pattern;

a method for detecting defects in said first substrate pattern in said first field of view comprising the steps of;

a. generating a residual spatial domain pixel map by subtracting one of said first and second spatial domain pixel maps from the other of said first and second spatial domain pixel maps;

b. determining the pixel gray level histogram of said residual spatial domain pixel map of step a.;

c. fitting the histogram of step b. to an exponential distribution function;

d. determining a threshold amplitude that is above the expected noise level and below the expected defect level from said exponential distribution function of step c.; and

e. thresholding said residual spatial domain pixel map of step a. with the determined threshold from step d. to identify all of those values that exceed said threshold as defects in said first spacial domain pixel map.

20. In an inspection system for detecting defects in devices of any type having either random, repeating or mixed patterns within the same device, said inspection system includes first and second fields of view each positioned along a different optic axis, the first field of view creates a first spatial domain pixel map of a portion of a first substrate pattern;

and said second field of view provides a second spatial domain pixel map against which said first substrate pattern is to be checked from either a pre-stored pixel map of the desired pattern, a scaled drawing of the desired pattern, a repeating pattern within the same device on said first substrate as in said first field of view, a pattern within a different device on said first substrate as in said first field of view, or a second substrate pattern;

a method for detecting defects in said first substrate pattern in said first field of view comprising the steps of;

a. generating a residual spatial domain pixel map by removing the repeating pattern from said first spatial domain pixel map;

b. determining the pixel gray level histogram of said residual spatial domain pixel map of step a.;

c. fitting the histogram of step b. to an exponential distribution function;

d. determining a threshold amplitude that is above the expected noise level and below the expected defect level from said exponential distribution function of step c.; and

e. thresholding said residual spatial domain pixel map of step a. with the determined threshold from step d. to identify all of those values that exceed said threshold as defects in said first spacial domain pixel map.

21. An inspection system for detecting defects in devices of any type having either random, repeating or mixed patterns within the same device comprising:

a processor for controlling the operation of said inspection system;

first and second fields of view means each under the control of said processor and each positioned along a different optic axis, the first field of view for creating a first spatial domain pixel map of a first substrate pattern, and said second field of view for providing a second spatial domain pixel map against which said first substrate pattern is to be checked from either a pre-stored pixel map of the desired pattern, a scaled drawing of the desired pattern, a repeating pattern within the same device on said first substrate as in said first field of view, a pattern within a different device on said first substrate as in said first field of view, or a second substrate pattern;

a transform system for individually identifying and removing repeating patterns from each of said first and second spatial domain pixel maps, said transform system includes;

transform means for individually transforming said first and second spatial domain pixel maps into first and second frequency domain pixel maps using a two dimensional Fourier like transform;

harmonic trimming means for individually determining the frequencies of said first and second frequency domain pixel maps in both dimensions having amplitude maxima and individually zeroing out the values at the frequencies at which the maxima values are detected to create modified first and second frequency domain pixel maps; and

first inverse transformation means for transforming said modified first and second frequency domain pixel maps using the inverse of said two dimensional Fourier like transformation to obtain into modified first and second spatial domain pixel maps without the repeating patterns of said first and second spatial domain pixel maps.

22. An inspection system as in claim 21 wherein said transform system further includes:

conjugation means for conjugating one of said first and second frequency domain pixel maps;

multiplication means for multiplying said conjugated one of first and second frequency domain pixel maps with the other of said first and second frequency domain pixel maps to create a multiplied frequency domain pixel map;

second inverse transformation means for inverse transforming said multiplied frequency domain pixel map using the inverse of said two dimensional Fourier like transform to create a correlation surface of the patterns of said first and second spatial domain pixel maps, said correlation surface being a function of the spatial offset between said first and second spatial domain pixel maps;

peak identification means for identifying the peaks of said correlation surface along both the x and y axis with the location of those peaks representing the offset, Δ

x and Δ

y, between said first and second spatial domain pixel maps; and

alignment means for trigonometrically altering the two dimensional sinusoids of one of said modified first and second frequency domain pixel maps using the Δ

x and Δ

y axis offset values identified by said peak identification means.

23. An inspection system as in claim 21 further comprises:

subtraction means for subtracting one of said modified first and second spatial domain pixel maps from the other of said modified first and second spatial domain pixel maps to form a composite modified spatial domain pixel map that contains only the combined noise and defects of both of said modified first and second spatial domain pixel maps;

threshold means for individually filtering each of said modified first and second spatial domain pixel maps and said composite spatial domain pixel map using a threshold amplitude that is selected to be above the expected noise level and below the expected defect level, to remove defect shadows introduced by said first inverse transformation means where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, each resulting further modified first and second spatial domain pixel maps including the random and defects for each of said first and second spatial domain pixel maps and said modified composite spatial domain pixel map containing only defects for said composite spatial domina pixel map; and

defect determination means for comparing all three of said further modified first and second spatial domain pixel maps and said modified composite spatial domain pixel map to determine the defects present in said first spatial domain pixel map.

24. An inspection system as in claim 23 wherein in the threshold means the threshold amplitude is selected by an exponential thresholding technique.

25. An inspection system as in claim 23 wherein said defect determination means includes:

comparison means for comparing all three of said further modified first and second spatial domain pixel maps and said modified composite spatial domain pixel map to determine the number of amplitude peaks of each of said pixel maps; and

identification means for identifying the location of defects as those occurring at the amplitude peaks in said modified composite spatial domain pixel map as identified by said threshold means if the number of peaks identified for both of said further modified first and second spatial domain pixel maps by said comparison means are both large numbers indicating that there are numerous random pattern portions in said first and second spatial domain pixel maps.

26. An inspection system as in claim 23 wherein defect determination means includes:

comparison means for comparing all three of said further modified first and second spatial domain pixel maps and said modified composite spatial domain pixel map to determine the number of amplitude peaks of each of said pixel maps; and

identification means for identifying the location of defects as those occurring at the amplitude peaks in said further modified first and second spatial domain pixel maps if the number of peaks identified for both of said further modified first and second spatial domain pixel maps are both small numbers indicating that there are few or no random pattern portions in the original first and second fields of view.

27. An inspection system as in claim 21:

subtraction means for subtracting one of said modified first and second frequency domain pixel maps from the other of said modified first and second frequency domain pixel maps to form a composite frequency domain pixel map that contains only the combined noise and defects of both of said modified first and second frequency domain pixel maps; and

second inverse transform means for transforming said composite frequency domain pixel maps using the inverse of the two dimensional Fourier like transform used by said transform means to a composite spatial domain pixel map;

threshold means for individually filtering each of said modified first and second spatial domain pixel maps and said composite spatial domain pixel map using a threshold amplitude that is selected to be above the expected noise level and below the expected defect level, to remove defect shadows introduced by said first inverse transformation means where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, each resulting further modified first and second spatial domain pixel maps including the random and defects for each of said first and second spatial domain pixel maps and said modified composite spatial domain pixel map containing only defects for said composite spatial domain pixel map; and

defect determination means for comparing all three of said further modified first and second spatial domain pixel maps and said modified composite spatial domain pixel map to determine the defects present in said first spatial domain pixel map.

28. An inspection system as in claim 27 wherein in the threshold means the threshold amplitude is selected by an exponential thresholding technique.

29. An inspection system as in claim 23 wherein said defect determination means includes:

comparison means for comparing all three of said further modified first and second spatial domain pixel maps and said modified composite spatial domain pixel map to determine the number of amplitude peaks of each of said pixel maps; and

identification means for identifying the location of defects as those occurring at the amplitude peaks in said modified composite spatial domain pixel map as identified by said threshold means if the number of peaks identified for both of said further modified first and second spatial domain pixel maps by said comparison means are both large numbers indicating that there are numerous random pattern portions in said first and second spatial domain pixel maps.

30. An inspection system as in claim 27 wherein defect determination means includes:

comparison means for comparing all three of said further modified first and second spatial domain pixel maps and said modified composite spatial domain pixel map to determine the number of amplitude peaks of each of said pixel maps; and

identification means for identifying the location of defects as those occurring at the amplitude peaks in said further modified first and second spatial domain pixel maps if the number of peaks identified for both of said further modified first and second spatial domain pixel maps are both small numbers indicating that there are few or no random pattern portions in the original first and second fields of view.

31. An inspection system for detecting defects in devices of any type having either random, repeating or mixed patterns within the same device comprising:

a processor for controlling the operation of said inspection system;

first and second fields of view means each under the control of said processor and each positioned along a different optic axis, the first field of view for creating a first spatial domain pixel map of a first substrate pattern, and said second field of view for providing a second spatial domain pixel map against which said first substrate pattern is to be checked from either a pre-stored pixel map of the desired pattern, a scaled drawing of the desired pattern, a repeating pattern within the same device on said first substrate as in said first field of view, a pattern within a different device on said first substrate as in said first field of view, or a second substrate pattern;

repeating pattern identification means for individually identifying and removing the repeating pattern portion of each of said first and second spatial domain pixel maps in the frequency domain to create modified first and second spatial domain pixel maps;

random pattern identification means for individually identifying and removing the random pattern portion of each of said modified first and second spatial domain pixel maps in the spatial domain to create further modified first and second spatial domain pixel maps; and

defect identification means for individually and collectively examining said further modified first and second spatial domain pixel maps in the spatial domain.

32. An inspection system for detecting defects in devices of any type having either random, repeating or mixed patterns within the same device comprising:

a processor for controlling the operation of said inspection system;

first and second fields of view means each under the control of said processor and each positioned along a different optic axis, the first field of view for creating a first spatial domain pixel map of a first substrate pattern, and said second field of view for providing a second spatial domain pixel map against which said first substrate pattern is to be checked from either a pre-stored pixel map of the desired pattern, a scaled drawing of the desired pattern, a repeating pattern within the same device on said first substrate as in said first field of view, a pattern within a different device on said first substrate as in said first field of view, or a second substrate pattern;

generation means for generating a residual spatial domain pixel map by subtracting one of said first and second spatial domain pixel maps from the other of said first and second spatial domain pixel maps;

histogram means for determining the pixel gray level histogram of said residual spatial domain pixel map of said generation means;

fitting means for fitting the histogram of the histogram means to an exponential distribution function;

selection means for determining a threshold amplitude that is above the expected noise level and below the expected defect level from said exponential distribution function of said fitting means; and

thresholding means for thresholding said residual spatial domain pixel map with the determined threshold to identify all of those values that exceed said threshold as defects in said first spacial domain pixel map.

33. An inspection system for detecting defects in devices of any type having either random, repeating or mixed patterns within the same device comprising:

a processor for controlling the operation of said inspection system;

first and second fields of view means each under the control of said processor and each positioned along a different optic axis, the first field of view for creating a first spatial domain pixel map of a first substrate pattern, and said second field of view for providing a second spatial domain pixel map against which said first substrate pattern is to be checked from either a pre-stored pixel map of the desired pattern, a scaled drawing of the desired pattern, a repeating pattern within the same device on said first substrate as in said first field of view, a pattern within a different device on said first substrate as in said first field of view, or a second substrate pattern;

generation means for generating a residual spatial domain pixel map by removing the repeating pattern from said first spatial domain pixel map;

histogram means for determining the pixel gray level histogram of said residual spatial domain pixel map of said generation means;

fitting means for fitting the histogram of the histogram means to an exponential distribution function;

selection means for determining a threshold amplitude that is above the expected noise level and below the expected defect level from said exponential distribution function of said fitting means; and

thresholding means for thresholding said residual spatial domain pixel map with the determined threshold to identify all of those values that exceed said threshold as defects in said first spacial domain pixel map.