Radiant energy – With charged particle beam deflection or focussing – Magnetic lens
Reexamination Certificate
2000-02-26
2002-11-05
Berman, Jack (Department: 2881)
Radiant energy
With charged particle beam deflection or focussing
Magnetic lens
Reexamination Certificate
active
06476398
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to automation of charged-particle-beam system operations, in particular focusing, contrast setting and astigmatism corrections.
2. The Prior Art
As in optical microscopes, focus, contrast and astigmatism correction are fundamental to imaging in charged-particle-beam systems such as scanning-electron-microscope (SEM) systems and focused-ion-beam (FIB) systems. As the semiconductor diagnostics industry develops, the microscope systems are automated, and so are the focus, contrast and astigmatism correction. However, it is a lot more difficult to perform these beam automation tasks in charged-particle-beam systems than in other optical systems. Since the early 1970's, a number of attempts have been made to achieve auto-focus and auto-correction of astigmatism. None of these has proven adequate in terms of reliability, speed, accuracy and repeatability. Better solutions are therefore needed.
Difficulties in charged-particle-beam system automation have many causes. Typically these include:
focus variation in run time due to column contamination, current drift, inspected sample navigation, etc.
imaging interdependence between focus, x-astigmatism and y-astigmatism, lens alignment, etc., although the settings are independent
poor image contrast, including saturation and low contrast
contrast variation at various focus positions
sample feature orientation
feature density over the image
image rotation as focus varies (the rotation angle could be up to 90 degrees), defined herein as “focus-induced-image-rotation” (“FIIR”)
image rotation center moves as focus varies
image size or magnification change as focus varies
image deformation including changes in aspect ratio and orthogonality
rolling wave noise, shot noise, synchronization noise, and other kinds of uneven patterned noise
poor electron gun alignment
Of all the factors in the above, FIIR, poor contrast, sample feature orientation variation, magnification change, and deformation are the worst obstacles to achieving good-quality beam automation. All the existing techniques mentioned below are vulnerable to these obstacles. As a result, measurements performed with the existing techniques are often not reliable and repeatable.
Auto-focus, auto-correction of astigmatism, and auto-contrast have long been issues existing in charged-particle-beam systems. Prior-art techniques used to address these issues include one-dimensional fast-Fourier transform (1-D FFT), two-dimensional fast-Fourier transform (2-D FFT), power spectrum analysis, maximizing the intensity gradient, profile analysis, polar line gradient, variance of differences, etc. These methods fall into the following categories: FFT, power spectrum, maximum gradient, and variance of differences.
Weak points of the prior techniques are in general:
The ways of measuring focus goodness and astigmatism are in many situations unstable.
The search paths are not optimized
FIG. 1
shows one general procedure of auto-focus in these prior methods. An image with good edges is selected (step
105
). The focus is set at a starting position (
110
). The image is acquired (step
115
). The image is processed to determine a value representing a goodness index (GI) of focus (step
120
). The focus is then changed by one step (step
125
). Steps
115
-
125
are repeated for N times (step
130
). After N repetitions, the largest index of focus value is found through comparison (step
135
). Finally, the charged-particle-beam system is set at the corresponding focus (step
140
).
Another procedure of auto-focus is shown in FIG.
2
. In its data processing, Fast-Fourier transform (FFT) method is used as in step
120
. The procedure starts by reading the current focus position F (step
205
). Using an offset &Dgr;F, the system focus is set at F+&Dgr;F (step
210
). An image is captured, then processed using FFT (step
215
). The image is thresholded using pre-set criteria. Count the pixel numbers above the criteria as P, P=P(F+&Dgr;F) (step
220
). Repeat the above to get P=P(F−&Dgr;F) (steps
225
-
235
). Compare P(F+&Dgr;F) and P(F−&Dgr;F). If P(F+&Dgr;F)>P(F−&Dgr;F), then F is increased to F+&Dgr;F. Otherwise, F is decreased to F−&Dgr;F (Step
245
). Repeat the procedure and in the same time reduce the step size &dgr;F proportionally to the ratio of |P(F+&Dgr;F)−P(F−&Dgr;F)|/|P(F+&Dgr;F)−P(F−&Dgr;F)| until the two FFTs are similar, that is, P(F+&Dgr;F)≡P(F−&Dgr;F). The F is then finally chosen as the best focus position (Step
250
).
Techniques for FFT data processing are discussed, for example, in National Univ. of Singapore, Scanning, Vol. 19,553 (1997), and in National Institute of Standards and Technology, SPIE Vol. 2725, p. 504 (1996).
Disadvantages of FFT processing include:
Too slow
Very much image feature shape and feature density dependent
Very vulnerable to FIIR
Very vulnerable to magnification change during focus
The search time and accuracy heavily depends on the shape of the curve of P value versus focus. If with a nearly flat top, the search will either be extremely slow or stop at an inaccurate position. If with a sharp top and the curve is not symmetric about the peak, the search will end at a position far from the ideal.
Power spectrum analysis methods determine resolution of a SEM from the width of the power spectrum, and detect astigmatism from the asymmetry of the power spectrum. Disadvantages include:
Too slow
Sensitive to noise
Maximum-gradient method measures gradient sum over image lines, and maximizes the sum while adjusting focus.
FIG. 3
shows a prior art maximum-gradient method used to process the data as at step
120
. A region of interest (ROI) is chosen with random features or good edges (step
305
). Gradient is calculated either by each pixel or along certain directions (step
310
). Total gradient is taken as a measurement of focus sharpness (step
315
). Astigmatism correction is achieved using a similar method.
Gradient sum data processing techniques are discussed, for example, in W. J. Tee et al., J. PHYS. E: SCI. INSTRUM. 12, 35 (1979).
Disadvantages of maximum-gradient methods include:
Very sensitive to noise
Sensitive to sample rotation while focusing
The variance of differences method finds the difference of two scan lines in two images each acquired at a different focus or stigmator setting, and looks for the highest variance of the difference. This method is better than the gradient methods in that it is insensitive to tilted-plan background noise. However, it also suffers from all other kinds of noise and image rotation, the FIIR.
All the above techniques have more difficulties and failures if working under coarse adjustment of focus and astigmatism correction. The reason is that these techniques do not take into account the factors of image rotation and magnification change during focus. Besides, all these methods are not coupled with auto-contrast. Without proper contrast during runt time, auto-focus and auto-astigmatism-correction will not run properly. In many situations, the system needs to run coarse adjustment, even during real time test. In general, the existing techniques lack high reliability, accuracy and repeatability at high speed.
SUMMARY OF THE INVENTION
Embodiments in accordance with the invention provide respectively for auto-focus, auto-contrast, and auto-correction of astigmatism in both x and y directions independent of sample feature orientation and image deformation, insensitive to various kinds of noise, and insensitive to image magnification change caused by focus. Poor image contrast is handled by the auto-contrast capability. Thus, embodiments in accordance with the invention can achieve high reliability and repeatability at relatively high speed.
Embodiments in accordance with the invention provide methods for setting parameters of a focused-particle-beam column (e.g., of a SEM or FIB system) to acquire a f
Barnard Richard
Xu Li
Applied Materials Inc.
Berman Jack
Riter Bruce D.
Smith Johnnie L
LandOfFree
Beam automation in charged-particle-beam systems does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Beam automation in charged-particle-beam systems, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Beam automation in charged-particle-beam systems will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2929686