Radiant energy – Photocells; circuits and apparatus – With circuit for evaluating a web – strand – strip – or sheet
Reexamination Certificate
2001-11-20
2003-12-02
Le, Que T. (Department: 2878)
Radiant energy
Photocells; circuits and apparatus
With circuit for evaluating a web, strand, strip, or sheet
C250S208100, C250S306000, C250S201300
Reexamination Certificate
active
06657221
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to an apparatus and method for automatically observing or classifying defects caused on a semiconductor wafer in a process for producing a semiconductor product.
In a process for producing a semiconductor product, in order to ensure high producing yield, it is necessary to be found various defects caused in the production process and to be taken measures for generation of the various defects, at an early stage. This is typically performed in the following steps.
(1) Detecting the locations of the caused defects or the deposited foreign particles by that a semiconductor wafer to be inspected is inspected by a wafer visual inspection unit or a wafer foreign particle unit.
(2) Classifying the detected defects under each of generating causes by observing (this is called a review) them. This review operation typically uses a review-specific unit having a microscope for observing defect parts at high magnification. Other unit with a review function, e.g., a visual inspection unit may be also used.
(3) Solving measures for each of the causes are taken.
In the case that the number of defects detected by the inspection unit is very high, the review operation needs enormous efforts. Therefore, in recent years, there has been developed actively a review unit having an Automatic Defect Review function for automatically imaging and collecting images of defect parts and an Automatic Defect Classification function for automatically classifying the collected images. Japanese Opened Patent Publication No. Hei 10-135288 discloses a review unit having an Automatic Defect Review function and an Automatic Defect Classification function and using a scanning microscope for an imaging system of it, and a production system.
FIG. 2
shows one example of a prior art ADR processing flow. First, a wafer to be inspected is placed onto the stage of a review unit (S
21
) and inspection data as results inspected by the inspection unit is read into the review unit from the database (S
22
). Then, the operator selects and specifies the defect targeted for ADR from the inspection results obtained from the inspection unit (S
23
). When the throughput of the ADR is high and the defected defect data are small in number, all the defects can be subject to ADR.
The review unit selects one from the specified defects and moves the stage so that the selected defect is positioned in the center of the visual field of an observation system. Thereafter, optimal focus setting is performed and an image of the selected defect is imaged by the observation system (S
24
). This image is called a defect image. The imaged defect image is stored into a recording medium (e.g., a magnetic disk) in the review unit.
Next, while the stage is moved, an image of the same part of the chip adjacent to the semiconductor chip in which the defect part in the wafer exists, is imaged (S
25
). This part is formed with the same pattern as that of the defect part. This image is called a reference image to the defect image. The reference image is also stored into the recording medium in the review unit. At the completion of imaging of the reference image, the defect image and the reference image of the next defect are imaged, as described above. These processes are repeated for all the defects to be subject to ADR, and are then terminated (S
26
).
FIG. 3
shows one example of a prior art ADC processing flow. ADC is a process for automatically deciding a category of the defect by using the defect image and the reference image acquired by ADR. First, the defect part is specified from the defect image and the reference image (S
31
). Specifically, a differential image is generated by being differentially operated between the defect image and the reference image. As a result, only the part in which the defect image and the reference image are different from each other appears in the differential image, which exhibits the defect part. The feature amounts of the defect are calculated by using the differential image, the defect image and the reference image (S
32
). The feature amounts quantitatively and numerically express the size of the defect, the shape of the defect, and the contrast on the image of the defect. An Automatic Defect Classification process for deciding the defect category is performed by using the feature amount data (S
33
).
The prior art ADR and ADC shown in
FIGS. 2 and 3
are disclosed in Japanese Opened Patent Publication No. Hei 10-135288. In the prior art, the defect image or the reference image is imaged after the stage is stopped once. The imaging of one imaged part consists of three steps for: (1) moving of the stage to the imaged part, (2) stopping of the stage, and (3) imaging of an image.
When the stage is stopped for imaging, stage control and beam control during imaging can be simplified. On the other hand, the stage must be stopped completely. As the stage has some weight, if a stop command is issued from the control unit of the stage, the stage will not be stopped soon and the time to stop the stage completely is required to some extent. As the stage is moved slightly while the time to stop the stage elapses, if a review image is imaged when the stage is moved slightly, blurring or flow is caused in the image. As the result, the image quality needed for review cannot be obtained. For this reason, it must be waited to start imaging the review image until the time to stop the stage elapses after the command to stop the stage is issued. As this time is longer than the time needed for imaging, the prior art imaging method cannot acquire an image fast.
Here, consider the limit of throughput on the prior art. For simplification, chips produced in a wafer have a 15 mm pitch, and the number of defects per chip is 1. In other words, assume that the interval between defects and the interval between the defect part and the reference part are about 15 mm. The stage moving velocity is assumed to be 50 [mm/sec]. In this case, the time to move a distance of 15 mm is 15/50=0.3 [sec]. Actually, the stage moving needs acceleration or deceleration, and this time must be considered. However, this time is omitted here. The waiting time from stopping of the stage to starting of imaging is 0.2 [sec] as an experience value.
For imaging, a beam needs scanning in two dimensions. If an image of 512×512 pixels as one frame is acquired by scanning at 100 MHz, that is, 10 [n sec/pixel] for one pixel, it need 512×512×10 n[sec]=about 3 [m sec]. As a scanning electron microscope causes much noise in a detected signal, frame addition is typically performed to acquire a high-quality image. The number of the frame additions is assumed to be 16. In the operation of frame addition, a plurality of images of the same part are imaged, so that an average gray-scale value of the same pixel over the plurality of images is obtained as a pixel value of the same pixel, thereby acquiring an image reducing the influence of the noise.
As the number of frames is increased, the image quality is enhanced, but long time is required accordingly. The number of frames is set by considering image quality to be acquired. When 16 frame additions are performed, an electric current value of an irradiation beam is, for example, 200 [pA]. A signal amount n for irradiation to one pixel is n=(200 [pA]×10 [nsec]×16)/1.6×10
−19
=200 (one electron is 1.6×10
−19
[coulomb]. Since a noise of the signal n by statistical fluctuation is &Dgr;n=n
½
, in this case, &Dgr;n=14.1. As an index to quantify the image quality, a standard deviation &sgr; of noise variation to a signal is used to assume that the fluctuation amount &Dgr;n of the signal is 3&sgr;. From 3&sgr;=14.1, &sgr;=4.7 is determined. With this value, it is found from experience that the image quality needed for review can be ensured.
In imaging, other than the beam
Aoki Kazuo
Hiroi Takashi
Nakagaki Ryo
Noguchi Minori
Takagi Yuji
Le Que T.
Townsend and Townsend & Crew LLP
LandOfFree
Image classification method, observation method, and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image classification method, observation method, and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image classification method, observation method, and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3152489