On-the-fly automatic defect classification for substrates...

Details On-the-fly automatic defect classification for substrates... On-the-fly automatic defect classification for substrates...

1998-06-25

2001-07-03

Pham, Hoa Q. (Department: 2877)

Optics: measuring and testing

Inspection of flaws or impurities

Surface condition

C356S237400, C356S237500

Reexamination Certificate

active

06256093

ABSTRACT:

FIELD OF THE INVENTION
The present invention is in the general field of detecting and classifying defects in wafers during and/or after manufacture.
BACKGROUND OF THE INVENTION
As is well known, considerable effort is spent in academia and industry towards developing techniques enabling the detection of defects in wafers before their cleavage into distinct dies, capsulation and subsequent distribution to the marketplace in the form of chips. Preliminary defect detection inter alia improves the “yield” in terms of discarding defective wafers and consequently increasing the percentage of fault-free chips delivered from the production line.
In recent years, defect detection has been improved by techniques for classifying defects into distinct defect types. These techniques not only allow identification of defective wafers so as to increase the yield, but also to provide some information on the cause of the defects. The production stage can then be reconfigured or modified in order to produce a better wafer series.
The improved sensitivity accomplished by classifying defects is illustrated graphically in
FIG. 1
(
10
), which shows the number of defects (Z) of various types (Y) for four sequentially produced wafers (X). Whilst the total count (
12
) is essentially the same for the four wafers, the count for defect type D (
14
) dramatically increased over time. This allows the cause of the defect to be identified and corrected.
A rudimentary defect classification method involves a person visually inspecting wafers for defects and classifying the defects according to predetermined criteria. This manual procedure is not only slow but also error prone, since it relies on the inspector's professional skills which obviously differ from one inspector to another.
Automatic Defect Classification (ADC) techniques cope in many respects with the shortcomings of the manual procedures. Thus, for example.
KLA of Santa Clara, Calif., markets a software package, called Impact, as an add-on option to its inspection-review systems, such as the Indy 2230. The software applies certain algorithm to the defect image taken by a CCD camera to attempt and classify the defect. Notably, the Impact algorithm can be operated only when the system is in the review mode, and is inoperable when the system is in the inspection mode. Thus, in order to classify the defects, the system first has to scan the entire wafer in the inspection mode, switch to the review mode and re-visit the suspected sites identified during the inspection mode, take a magnified image of the defect and a reference site, and only then apply the ADC algorithm to classify the defect. The latter mode of operation is generally known in the industry as “re-visit ADC”. Also notably, the system uses only a single image of the defect from a single perspective.
Whilst the CCD image based analysis is more accurate and reproducible than the manual procedure (i.e. virtually identical results ate obtained by repeated analysis of the same wafer), it still has some major shortcomings, e.g. it is time consuming. It is important to note in this context that a prolonged wafer inspection session adversely affects the entire production line throughout. Unduly slowing down the production line throughput for wafer inspection is, in many cases, commercially infeasible from a cost perspective, considering the high cost of the chip fabrication process.
There is accordingly, a need in the art to provide for an ADC technique that will provide a relatively accurate and reproducible defect classification, and that at the same time, will substantially reduce the duration of the defect classification stage of hitherto known devices. To this end, in the present invention, defect classification is substantially incorporated into the wafer inspection phase so as to constitute an on-the-fly ADC, i.e. the ADC is inspected as the wafer is scanned for defects.
Throughout this specification, a use is made of the convention that “inspection” refers to the process wherein a substrate is scanned to identify locations suspected of having defects thereon, whilst “review” refers to the process wherein the suspected locations are revisited to confirm/refute the presence of a defect in the suspect location and investigate the defect should such indeed exists, all as known per se.
SUMMARY OF THE INVENTION
The present invention is based on the finding that by illuminating a wafer and collecting the scattered light by at least two detectors it is possible not only to detect defects in the wafer but also to classify the defects into distinct types by analyzing the attributes of the collected light. By this approach, an “on the fly” ADC is accomplished. According to the invention, dark field type detectors are utilized for collecting the scattered light.
Wafer has substantially mirror-like surface and, accordingly, when the illuminating beam is incident on a fault-free region of the inspected wafer it is reflected at expected direction (in compliance with the SNELL law). Thus, by way of example, if the angle of the incident beam is normal to the wafer's surface, the expected direction of the scattered light is also normal to the wafer surface (i.e. the angle between the illuminating beam and the scattered beam is substantially zero). By way of another example if the angle of the incident beam is in 45 relative to the wafer's surface, the expected direction of the scattered light is 135 relative to the wafer surface (i.e. the angle between the illuminating beam and the scattered beam is substantially 90).
Dark field detector is placed such that it is adapted to detect light scattered in direction substantially different from the specified expected direction. Accordingly, when the beam is incident on a fault-free region of the inspected wafer a dark field detector senses low (if any) energy. If, on the other hand, high energy is sensed by a dark field detector, this indicated, that the illuminating beam is not incident on a mirror like surface but rather on an irregular geometry being a defect of a given type. In contrast for bright field detectors an attenuated energy indicates that the illuminating beam is incident on an irregular geometry, whereas high energy indicates that the illuminating beam is not incident on a mirror like surface.
Since dark field detectors respond predominantly to defects (unlike the so called “bright field detectors”) they virtually filter out “noise” and respond to “signal” (signal stands for “defect”), and substantially the whole process of defect detection and classification is not only simplified but is also more accurate.
Accordingly, the present invention provides for a method for on-the-fly automatic defect classification (ADC) in a scanned wafer, comprising:
(a) providing at least two spaced apart detectors;
(b) illuminating the scanned wafer so as to generate an illuminating spot incident on the wafer;
(c) collecting light scattered from the spot by the at least two spaced apart detectors; and
(d) analyzing said collected light so as to detect defects in said wafer and classifying said defects into distinct defect types.
In the context of the invention, when referring to the location of detectors, this does not necessarily imply that all the constituents of the detector are placed at the prescribed location, but rather that at least the light sensing or collecting part of the detector is placed at the specified location. Thus, for example, by one embodiment, the light receiving edge of a bundle of optic fibers (forming part of a detector) are placed in the vicinity of the wafer whereas the rest of the detector's components are coupled to the opposite remote edge of the bundle.
By one embodiment the illuminating light is incident on the wafer at an angle substantially normal to the wafer surface. According to this embodiment the optical axis of each dark field detector is at a grazing angle to the surface of the inspected wafer.
By another embodiment, the illuminating light is incident on the wafer at a grazing angle to the wafer surfac

Affiliated with

Also associated with

Rating

On-the-fly automatic defect classification for substrates... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with On-the-fly automatic defect classification for substrates..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and On-the-fly automatic defect classification for substrates... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2475858