Image analysis – Image transformation or preprocessing – Fourier transform
Reexamination Certificate
2001-02-21
2004-05-04
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Image transformation or preprocessing
Fourier transform
C382S260000, C382S275000, C382S145000, C382S257000
Reexamination Certificate
active
06731824
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to optical detection systems for failure analysis of devices. More particularly, the present invention relates to a spatial filtering tool for integrated circuit or semiconductor wafer failure analysis for use in a commercial or laboratory fabrication environment.
2. Description of the Prior Art
Many devices, such as integrated circuits or semiconductor wafers, have structural patterns which can be detected by optical inspection systems in both a spatial domain and a frequency domain. In the prior art, optical inspection of such devices has been accomplished using several methods.
One such method employs an optical bench and visual inspection of the device thereon. The length of time required for such inspections renders this method impractical in an industrial environment. In addition, this type of inspection is prone to human error and, therefore, has limited application to a laboratory environment.
Other known optical inspection methods employ general image processing software suites and even hardware inspection systems, but these methods are incompatible with a commercial fabrication environment or a laboratory fabrication environment because they do not offer an optimized, automatic solution. To use these methods, spatial filtering must be “hand-crafted” to the device being inspected. Therefore, these methods are more suited to research because their use requires a high degree of customization and expertise.
Another known optical inspection method uses inline inspection tools. Such existing inline tools require expensive hardware, and lack the flexibility required in a failure analysis environment implemented in a laboratory.
Yet another technique uses conventional Scanning Electron Microscopy (SEM) images provided by a finely focused electron beam scanned across the surface of a sample and then inspected visually for defects (irregularities) by engineers or technicians in a failure analysis (FA) laboratory. While this approach is generally acceptable for both commercial and laboratory environments, small irregularities can be difficult to find.
Various refinements of SEM methods have been attempted to make detection of small irregularities easier. Spatial filtering has been applied to an SEM image of an object by using slides of SEM micrographs and a laser optical bench. By manually constructing a mask for the diffraction peaks formed in the optical path, the masked, reconstructed image primarily comprises defects. With the majority of the acceptable area masked, this approach facilitates the detection and study of just the defective areas of a circuit.
U.S. Pat. No. 6,021,214 to Evans et al. discloses a hybrid technique for finding defects using digitized images of a device and a combination of spatial domain and frequency domain techniques. Evans et al. teach obtaining two dimensional spectra of images of two samples of the device using Fourier-like transforms, removing any strong harmonics with a spectral filter, aligning the images and transforming them back to the spatial domain where they are then subtracted and the difference image analyzed for defects. Frequency domain techniques are employed to align the images.
In addition, U.S. Pat. No. 5,506,676 to Hendler et al. discloses defect detection using Fourier optics and a spatial separator. The Fourier transform components of an ideal pattern are compared to the Fourier transform components of a measured pattern, and differences in relative intensities of the spatial components indicate defects.
The contents of each of the above-discussed references are are hereby incorporated by reference.
The power spectrum of the image of a periodic circuit comprises a regular array of bright spots, which correspond to the periodic features in the circuit, and a darker background, which contains most of the information due to the presence of irregular defects. This power spectrum of a regular array of bright spots is an ideal target for masking techniques. In the prior art, two approaches are employed to define which spatial frequencies in the power spectrum should be masked out.
The first approach to masking spatial frequencies is referred to as “manual mask definition,” which involves manually defining the bright spots of a circuit image in the power spectrum. This approach is extremely time consuming, because it requires manual intervention. It is also prone to human error.
The second approach to masking uses Fourier analysis, whereby a periodic waveform is deconstructed into a series of sine waves or cosine waves, the resultant transform is masked for regular features, and then the masked transform is subjected to an inverse Fourier transform to obtain an image which allows visualization of any defects. This second approach to masking spatial frequencies is referred to as “using a golden circuit area” whereby the Fourier transform of a perfect region of the circuit is subtracted from a Fourier transform of a test area. Then, an inverse Fourier transform is performed on the difference, and because of the subtraction, only the defects are reconstructed. This second approach has the disadvantage of requiring that a suitable “golden circuit area” (i.e., flawless area) needs to be defined.
Thus, the prior art lacks a quick method to detect and visualize irregularities in SEM images of regular integrated circuits, that is both fully automatic and fine-tunable.
SUMMARY OF THE INVENTION
The present invention overcomes the above-noted problems of the prior art by providing such a quick method for visualization of irregularities which may be defects, in the images of regular integrated circuits. The method of the present invention is both fully automatic, and fine-tunable. To this end, according to the present invention, there is provided a spatial filtering method comprising the steps of:
(a) acquiring a digital image of a two-dimensional array of pixel data from an original image;
(b) calculating a first Fourier Transform of the acquired image data to generate a two-dimensional array of complex data from the equation:
F
(
u
,
v
)
=
1
/
N
∑
x
=
0
N
-
1
∑
y
=
0
N
-
1
f
(
x
,
y
)
ⅇ
-
2
πj
(
ux
+
vy
)
;
(c) calculating a power spectrum for the first Fourier Transform calculated in step (b) to provide a real function representing a weighting of each spatial frequency in the original image from the equation:
P
(
u,v
)=log (1
+F
(
u,v
)
F
*(
u,v
));
(d) generating a mask from the power spectrum calculated in step (c) to apply to the data resulting from step (b) for suppressing regular structures of the original image and undesirable artifacts introduced by the acquiring of the digital image in step (a);
(e) dilating the mask generated in step (d) by extending masking spots generated therein to increase the suppression of regular structures of the original image;
(f) in order to remove periodic data, applying the mask dilated in step (e) to the first Fourier transform calculated in step (b) which results in a second Fourier transform;
(g) calculating an inverse Fourier transform of the second Fourier transform of the digital image masked in step (f) to obtain a spatially filtered image or defect image from the equation:
f
(
x
,
y
)
=
1
/
N
∑
u
=
0
N
-
1
∑
v
=
0
N
-
1
F
(
u
,
v
)
ⅇ
+
2
πj
(
ux
+
vy
)
;
(h) scaling the spatially filtered image to greyscale; and
(i) providing a visual representation of the greyscaled image obtained in step (h).
In a first aspect of the present invention, the mask generation in step (d) is performed by calculating a brightness intensity threshold in the power spectrum which includes the brightest 10% of the pixels. These pixels then form the initial mask of detected bright spots (0)'s and background (1)'s.
In a second aspect of the present invention, the dilating of the mask in step (e) may include extending the mask to maximize suppression of X and Y
Chawan Sheela C
Mehta Bhavesh M.
PROMos Technologies, Inc.
Stevens Davis Miller & Mosher LLP
LandOfFree
Spatial filtering method for failure analysis of a device does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Spatial filtering method for failure analysis of a device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Spatial filtering method for failure analysis of a device will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3246468