Image analysis – Applications – Manufacturing or product inspection
Reexamination Certificate
1998-07-27
2002-06-04
Boudreau, Leo (Department: 2721)
Image analysis
Applications
Manufacturing or product inspection
C382S256000
Reexamination Certificate
active
06400838
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a technique of inspecting patterns such as photomask patterns formed on a sample used to fabricate semiconductor devices or liquid-crystal display (LCD) devices, or patterns formed on a sample such as a semiconductor or LCD device, to see if the patterns are sound without faults. In particular, the present invention relates to pattern inspection equipment, a pattern inspection method, and a computer-readable storage medium storing a pattern inspection program, for realizing such a technique.
2. Description of the Related Art
Large-scale integrated circuits (LSIs) are manufactured with the use of photolithography and photomasks. The yields of LSIs becomes poor if patterns of the photomasks have faults. Then, various kinds of equipment have been developed to inspect the faults in the photomask patterns.
Photomask patterns are formed by depositing a metal film such as a chrome film on a glass substrate, coating the metal film with photoresist, exposing, developing, and baking them, and etching the metal film with the photoresist as a mask. The etching may be wet etching, dry etching, or any other. According to the etching conditions, the undercuts, sidewall angles, anisotropy, pattern transformation of the metal film vary. The photomask patterns may have rounded corners depending on the photoresist exposing conditions and the metal film etching conditions. Then, an optical image taken from the actual photomask patterns generated on the glass substrate does not correctly match with designed patterns. Namely, the sizes, line widths, and edge positions of the actual photomask patterns slightly differ from those of the designed patterns serving as a reference to inspect the actual photomask patterns because the actual photomask patterns involve edge displacement and corner roundness while the designed patterns involve none of them.
As a result, the pattern inspection equipment frequently determines that the photomask patterns are defective even if they are substantially homothetic transformation of the designed patterns. And, if designed patterns are used as they are to inspect actual patterns, they will lead to erroneous results determining that the actual patterns have rounded corners and are defective, even if they are substantially same geometry as the designed patterns.
FIG. 1
shows pattern inspection equipment according to a prior art to cope with the situation that the actual photomask patterns can have slightly displaced edges. This equipment develops design data into pattern data and modifies the pattern data to adjust it to match with actual photomask patterns. The size-modified pattern data is used as reference data, which is compared with measured pattern data taken from the actual photomask patterns. A light source
2
emits light to irradiate a photomask
1
having patterns delineated according to the design data. The light passes through the photomask
1
and an object lens
4
and forms an optical image of the photomask patterns on a sensor
3
. A sensor circuit
5
measures the image, digitizes it into measured pattern data, and transfers it to a fault decision circuit
6
. Even if the actual photomask patterns have sharp edges, the measured pattern data does not provide perfect rectangular waveforms because the measured pattern data involves a blur produced by the optical observation system. Namely, the measured pattern data involves blurred edges.
On the other hand, a CPU
9
in a host computer system transfers the design data from a data memory to a binary pattern development circuit
8
, which develops the design data into binary data made of “1s” and “0s”. The binary data is sent to a size modification circuit
7
, which adjusts the binary data to match with changes in the photomask patterns caused by mask manufacturing processes. The adjusted pattern data is sent to the fault decision circuit
6
. A front stage of the fault decision circuit
6
carries out a convolution process on the resized pattern data, to correct it for the blur of the measured pattern data caused by the observation optical system. A rear stage of the fault decision circuit
6
compares the adjusted, convoluted pattern data with the measured pattern data, to see if the photomask patterns have faults.
The size modification process carried out by the size modification circuit
7
will be explained. The design data used to delineate the photomask patterns is developed by the binary pattern development circuit
8
into binary data consisting of “1s” and “0s”.
FIG. 2A
shows an array
15
of “n×n” pixels in the binary data. The size modification circuit
7
calculates an OR of “1s” and “0s” in peripheral pixels
21
of the array
15
. According to the OR, the size modification circuit
7
corrects a value at central pixel
22
of the array
15
to 1 or 0, thereby adjusting the sizes and line widths of the binary data to match with those of the actual photomask patterns. The fault decision circuit
6
compares the corrected pattern data with the measured pattern data, to see if the photomask patterns have faults.
The size of a sensor element (sensor pixel) of the sensor
3
is usually equal to the size of a pixel of the pattern data, and the size modification process is carried out pixel by pixel. As a result, the adjustment of design data to match with actual photomask patterns is restricted by the pixel size, and if a pattern edge is displaced by half a pixel size, the pattern inspection equipment will detect it as a fault. To avoid this type of erroneous detection, the equipment of the prior art must lower a threshold to determine a fault. This, however, may result in overlooking actual faults that must be detected.
To correctly adjust design data to measured data that contains slightly displaced edges, there is an idea of developing the design data into pattern data consisting of pixels each being smaller than a sensor pixel. This idea, however, increases the quantity of the pattern data and needs highspeed data processing circuits to elongate a processing time and increase inspection costs.
To cope with the situation that actual photomask pattern has the roundness of corners, the prior art develops design data into binary data and rounds corners in the binary data to approximate the actual photomask patterns. The binary data is made of “1s” and “0s” and is scanned with a pixel area
20
of FIG.
2
B. The pixel area
20
has a central pixel
22
and a radius “r.” The value of the central pixel
22
is set to “1” or “0” according to a majority of “1s” and “0s” in the pixel area
20
. By determining the value of the central pixel of each pixel area according to a majority, the prior art adjusts corners of the design data to match with those of the actual photomask patterns. The corner-rounded design data serves as reference data. The fault decision circuit
6
compares the reference data with measured pattern data taken from the actual photomask patterns, to see if the photomask patterns have faults. The measured pattern data is obtained by irradiating the photomask patterns with light and by detecting light transmitted through the photomask patterns with a sensor.
The prior art hardly determines a value representing a pattern corner if the size of a sensor pixel is equal to the size of a pattern pixel. More precisely, if an edge of a pattern image crosses the center of a sensor pixel and covers a half of the sensor pixel, the prior art is unable to determine whether it is “1” or “0”. Namely, the rounding process that is carried out pixel by pixel is incapable of precisely expressing actual photomask patterns. Eventually, the prior art must lower a fault detecting threshold to avoid erroneous detection at the corners of photomask patterns. This leads to overlooking fatal faults that must be detected.
To solve this problem, design data may be developed into binary data whose pixel size is smaller than a sensor pixel, and the binary data is adjusted to match with the roundness of corners of actual phot
Boudreau Leo
Finnegan Henderson Farabow Garrett & Dunner L.L.P.
Kabushiki Kaisha Toshiba
Werner Brian P.
LandOfFree
Pattern inspection equipment, pattern inspection 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 Pattern inspection equipment, pattern inspection method, and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Pattern inspection equipment, pattern inspection method, and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2926899