Optics: measuring and testing – Inspection of flaws or impurities – Surface condition
Reexamination Certificate
2002-01-15
2003-12-02
Font, Frank G. (Department: 2877)
Optics: measuring and testing
Inspection of flaws or impurities
Surface condition
C356S237500
Reexamination Certificate
active
06657714
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to optical inspection systems, and specifically to methods and systems for detecting defects on substrates.
BACKGROUND OF THE INVENTION
Dark-field systems are well known in the art of optical inspection, particularly for detection of defects on substrates such as semiconductor wafers. Optical signals generated in dark-field inspection are typically characterized by very high dynamic range. The signal depends both on the reflectance of the material in the spot under inspection (as a function of the complex index of refraction) and on spatial variations within the spot. In bright-field systems, the reflectance usually dominates, and the resulting variations in the collected signal are generally no more than about two orders of magnitude. In the case of dark-field detection, however, smooth surfaces lead to almost no collection signal, while surfaces with protruding features may scatter many orders of magnitude more.
Patterned wafers used in producing advanced integrated circuits typically contain regions whose dark-field scattering signals may vary by six orders of magnitude or more. Examples of this phenomenon include scattering variations between the following types of regions:
Scribe lines scatter light differently from pattern regions.
Memory blocks scatter differently from their associated I/O circuitry.
Cache memory of a microprocessor unit scatters differently from the logic area.
A given memory region may have a pitch that generates a strong diffraction lobe toward a detector used to collect the scattered radiation, while the lobes of another region with a different pitch may escape detection.
Bare patches (down to several microns in size) scatter far less than adjoining regions of dense pattern.
The sensitivity of an inspection tool can be optimized for these different regions by controlling both signal acquisition (e.g., laser power and detector gain) and signal processing parameters. It is very difficult, however, to vary the acquisition parameters on the fly, in the process of scanning a single wafer, without reducing the inspection throughput, because the scan speed must generally be reduced in order to avoid artifacts due to rapid changes in the acquisition parameters. Furthermore, defining different regions on the wafer for signal acquisition and processing is a cumbersome task. In the particular case of bare patches (which may be as little as several microns in size, corresponding to a few pixels of the inspection system) between regions of dense pattern, it is impractical to program the inspection system to change its acquisition and processing parameters per region. Although the inspection system may, in principle, be able to adaptively learn the signal processing parameters, it is nearly impossible for the acquisition parameters to adapt within the time span of several pixels. In typical operation of a modern inspection system, this time span is typically no more than tens to hundreds of nanoseconds.
There is therefore a need, particularly in dark-field inspection, for detection systems with a very large dynamic range, in order to collect meaningful signals from both the very dark and the bright areas of the substrate being inspected without on-the-fly adjustment. Achieving a dynamic range greater than 10 or 12 bits (1:1024 or 1:4096) is very difficult, however, with detection systems operating at high data rates (tens to hundreds of mega-samples per second). The dynamic range of the detector itself is limited by the ratio of the saturation power to the minimal detectable signal, typically governed by the noise level of the detector and amplification circuitry. A further limitation is imposed by the restricted bit range of fast analog-to-digital converters.
One possible solution to this problem is to apply non-linear amplification to the output signal from the detector, in order to emphasize the low-amplitude signal range. An approach of this sort is described by Wolf in U.S. Pat. No. 6,002,122, whose disclosure is incorporated herein by reference. The dark-field detector output in this case is processed by a logarithmic amplifier and gain correction mechanism. Although this approach may provide improved visibility of defects in a dark-field image of a substrate, it does nothing to address the fundamental limitation of the dynamic range of the detection system.
Multiple-exposure imaging systems are known in the art. For example, U.S. Pat. No. 4,647,975, to Alston et al., whose disclosure is incorporated herein by reference, describes an electronic imaging camera with expanded dynamic exposure range, based on implementing two succeeding exposure intervals with different exposure parameters. A combined image is then constructed by choosing between the electronic information signals sensed during the two exposure intervals. Typically, the camera is used to combine one exposure of a scene taken with ambient light and another taken under flash illumination. PCT patent publication WO 90/01845, to Ginosar et al., whose disclosure is also incorporated herein by reference, describes image pickup apparatus including multiple image sensors receiving an image at various exposure levels. The sensor outputs are combined to create a single image with widened dynamic range.
SUMMARY OF THE INVENTION
It is an object of some aspects of the present invention to provide improved systems and methods for optical inspection, and particularly for detecting defects on a substrate.
It is a further object of some aspects of the present invention to provide dark-field detection systems having enhanced dynamic range.
In preferred embodiments of the present invention, a dark-field optical detection system comprises an illumination source, such as a laser, which scans a substrate under inspection, and two or more detectors for capturing light scattered from the substrate. Preferably, one of the detectors is optimized for high sensitivity, while another is designed to have a high saturation level, typically at the expense of its sensitivity. The light scattered from each point on the substrate is split among the detectors, wherein the respective portions of the radiation directed to the different detectors are not necessarily equal. Preferably, the light is split among the detectors in a consistent manner that is independent of polarization and angle, using a common integrating sphere, for example, to feed all the detectors.
The output of each of the detectors is sampled by its own processing channel. The output of the high-sensitivity channel provides information regarding defects on the substrate in areas of low scattered intensity, while that of the high-saturation channel provides information regarding areas of high scattered intensity. The use of multiple parallel detectors thus allows shot-noise-limited detection in areas with very little available light, without losing the signal due to saturation in bright areas. In this manner, defects on the substrate are detected with far greater dynamic range than can be achieved using single-detector systems, as are known in the art.
In some preferred embodiments of the present invention, spatial filtering is applied to the light scattered from the substrate before the light impinges on the detectors. A key purpose of such filtering is to eliminate constructive interference lobes at certain angles due to repetitive patterns on the substrate, as well as blocking reflections from other bright features on the substrate. Such spatial filtering improves the sensitivity of the detectors to weak scattering signals from local defects. In some of these preferred embodiments, a single spatial filter is used to filter the scattered light that is collected by all the detectors. In other preferred embodiments, at least two of the detectors have their own, separate spatial filters. In this manner, a different filtering characteristic can be applied to the weaker scattered light impinging on the high-sensitivity detector from that applied to the strong scattered light impinging on the high-saturati
Almogy Gilad
Goldberg Boris
Naftali Ron
Applied Materials Inc.
Blakely & Sokoloff, Taylor & Zafman
Font Frank G.
Nguyen Sang H.
LandOfFree
Defect detection with enhanced dynamic range does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Defect detection with enhanced dynamic range, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Defect detection with enhanced dynamic range will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3168994