Photocopying – Projection printing and copying cameras – Methods
Reexamination Certificate
2003-03-11
2004-07-20
Kim, Peter B. (Department: 2851)
Photocopying
Projection printing and copying cameras
Methods
C382S144000, C438S016000
Reexamination Certificate
active
06765651
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to photolithography. More specifically, the present invention relates to the simulation of an image during semiconductor manufacturing.
BACKGROUND OF THE INVENTION
Photolithography is the process of transferring geometric shapes on a photographic mask to the surface of a silicon wafer; it falls under the more broad category of microlithography. A photographic mask (or “photomask”), typically a glass plate with a patterned emulsion of metal film on one side, is used in photolithography to create integrated circuits. Chromium (chrome) is typically used to produce the pattern on the photomask.
The successful manufacture of advanced sub-micron sized semiconductor devices requires accuracy in production of the photomask, and in the photolithography processes used to pattern the wafer. Photolithography processes for semiconductor manufacturing frequently use image simulation for predicting the outcome of the manufacturing process. Simulation allows an evaluation of the quality of the product before spending time and money producing the actual product. The simulation takes as input either the electronic, geometrical design of the circuit to be produced, or the observed photomask image made from that design. The output is either a representation of the image as formed on the resist on the wafer, the so-called “aerial image,” or a representation of the result after the wafer has been exposed and developed.
The current standard procedure, as implemented in products such as VSS by Numerical Technologies, Inc. and ProLith by KLA-Tencor, is to use the Hopkins Method for modeling the electric fields that create the final image on the wafer. The Hopkins method is described in the following references, which are incorporated by reference: the Kirchauer Thesis available at http://www.iue.tuwien.ac.at/publications/PhD%20Theses/kirchauer
ode62.html; Professor Neurcuther's work on UC Berkeley's “SPLAT” simulation programi, available at http//cuervo.ccs.berkeley.edu/Volcano/applications/Defect/directory.html; and A. K. Wong and A. R. Neureuther,
Rigorous Three
-
Dimensional Time
-
Domain Finite
-
Difference Electromagnetic Simulation for Photolithographic Applications
, IEEE Trans. Semicond. Manufact., 8(4):419-431, November 1995).
The Hopkins Method requires a large number of calculations, and therefore is quite slow. A faster technique for simulating an image would be highly desirable. Faster simulation is important anywhere simulation is used. In photomask defect detection and analysis it allows determination of defect severity on the resultant wafer at a rate similar to the speed of current mask inspection machines. This greatly reduces the number of false defects reported while increasing the available sensitivity of inspections. Reduced false defect reports decreases costs involved with 1) reviewing reported defects 2) repairing false defects 3) damage caused by repair, and 4) re-inspecting masks after repair. Increasing sensitivity allows using existing inspection machines for newer, smaller geometry chip designs.
Fast simulation is also important for chip design and photolithography process development. A faster simulation method would allow more iterations of a chip design to optimize feature placement and optical enhancement techniques. It would also allow more of a chip's logic to be simulated to verify correct operation in the finished product.
SUMMARY OF THE INVENTION
To achieve the foregoing, and in accordance with the purpose of the present invention, a fast method of simulating the results of imaging and wafer processing using conventional image processing techniques is disclosed. The present invention uses conventional image processing techniques to produce an improved result with less computation. A typical speed increase is 5000× compared to the Hopkins Method.
This method models two optical processes to produce an accurate simulation more quickly: blurring and edge diffraction. Blurring is introduced by the optical resolution of the projection lens. This is defined in optical texts as the Rayleigh resolution criterion: Res=/0.61&lgr;/NA, where &lgr; is the wavelength of light used in the microscope, and NA is the Numerical Aperture of the main microscope lens, a measure of the lens's diameter. NA is defined as NA=n/2f#, where n is the index of refraction of the glass, and f# is the ratio of the lens focal length to its diameter. Edge diffraction, as defined in elementary physics texts, causes opaque areas to appear larger in a microscope than if measured mechanically. The nature of this edge diffraction is that photons that graze close to the edge of an opaque area get diffracted away from the microscope objective lens, causing the opaque area to appear larger optically than it is physically. Convolution with a gaussian kernel simulates the blurring; erosion and dilation simulate the edge diffraction.
Convolution and deconvolution are known image processing techniques that can be performed by several methods, as described in
The Image Processing Handbook
, by John C. Russ, CRC Press, 1992, incorporated herein by reference.
Erosion is a known image processing technique and may be performed by replacing each pixel with the darkest of the nine pixels adjacent to it, including itself. Dilation is the opposite function, replacing each pixel with the brightest of the nine pixels adjacent to it, including itself. References to methods of performing erosion apply to dilation by replacing “minimum” by “maximum.” In the literature these are referred to as gray scale erosion and gray scale dilation in “The Image Processing Handbook” mentioned above.
In a first embodiment of the invention, the transmission optical source image of the photomask is deconvolved to remove optical blurring, which is then dilated to remove edge diffraction as described above. This produces a simulated physical image corresponding to a theoretical infinite resolution optical microscope. This intermediate simulated physical image is eroded, and then convolved according to the resolution of the stepper at the photomask plane. This convolution produces a simulated image projected onto the wafer, or a “simulated wafer aerial image.” This aerial image can then be further eroded to match the effects of resist and developing, producing a “simulated wafer resist image.” Optional thresholding may be performed on the simulated wafer resist image to produce a simulated processed wafer image. Thresholding is described in Russ mentioned above, incorporated by reference herein.
In a second embodiment of the invention useful in practice, several steps are combined. The microscope resolution is typically two to three times higher than the stepper resolution being simulated. This fact allows steps to be combined because deconvolution as a separate step is not required. Thus, the deconvolution step may be eliminated by reducing the amount of blurring used to produce the aerial image. In addition, the dilation and erosion steps used to produce the aerial image are combined into a single erosion.
Where a phase shift mask is involved, a complex convolution is used. A phase shift mask has areas where the glass substrate is thinned, usually by an amount that causes the light to be delayed by ½ wavelength, or a phase of 180 degrees. This phase shift of 180 degrees causes dark destructive interference at the edge between the shifted and unshifted areas. The interference causes edges to appear sharper on the wafer, and that allows for more focus and illumination error during printing while yielding good devices. This technique is explained in Kirchauer, cited above.
Complex convolution is the same as standard convolution except that the data (images and kernel) are complex numbers that represent magnitude and phase. The pixel values in the source image are converted from energy to voltage by taking the square root. At the end the pixel values are squared to convert voltage back to energy or magnitude. Basically, an ima
Dutta April
Fiekowsky Peter J.
Kube Paul R.
Beyer Weaver & Thomas LLP
Kim Peter B.
LandOfFree
Fast image simulation for photolithography does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Fast image simulation for photolithography, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Fast image simulation for photolithography will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3209872