JPS6236638B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a pattern detection method and apparatus for automatically detecting the appearance of a fine pattern on an LSI wafer or the like.

For LSI wafers and very LSI ware, it is required that various patterns on the ware be arranged regularly. Patterns on clothing have been checked by checking each point on the clothing to be inspected and evaluating each point. However, this method has problems due to the increased number of checks and the need to determine whether the pattern is good or bad for each check.
Work efficiency was poor and processing time was long.

An object of the present invention is to provide a pattern detection method and apparatus that enable efficient detection of fine patterns.

The gist of the invention is as follows. LSI wear and masks have multiple identical patterns called chips. Therefore, when detecting the same pattern or evaluating the quality of a wafer or mask by checking the same pattern, instead of checking each pattern individually, Mutual (comparison) checking produces the effect that two checks can be done in one check. The present invention consists of such aspects,
For the same pattern on a wafer or mask, pattern detection is performed on both at the same time, and the detected patterns are compared with each other. If the two match, the pattern formation is good, and if they do not match, the pattern formation is defective. Comparative inspection. Defects are determined by Furthermore,
The present invention provides a novel pattern detection mechanism and a novel automatic control system for commanding the mechanism to accurately and simultaneously detect the same pattern on different chips during pattern detection. The present invention will be explained in detail below with reference to the drawings.

FIG. 1 is a diagram showing an embodiment of the detection system of the present invention. Sample 1 is LSI wear or a mask, and has a fine pattern on its surface. Two points 2A and 2B on the sample 1 are detection target points. The two incident lights A and B are irradiated onto the detection target points 2A and 2B via the half mirrors 3A and 7B and the objective lenses 3 and 7, and the detection target points 2A and 2B are
The reflected lights from A and 2B are magnified by objective lenses 3 and 7. The rotating polygon mirror 4 has a polygonal prism configuration,
Detection light from detection target points 2A, 2B obtained via objective lenses 3, 7, half mirrors 3A, 7B, prism 4A, and lenses 4B, 4C is received, reflected, and guided to optical fibers 5, 8. Rotating polygon mirror 4
rotates at a constant speed in the direction of the arrow. Therefore, the transmitted light from the lenses 4B and 4C is incident on the same mirror surface on the rotating polygon mirror 4 and is reflected, and as the rotating polygon mirror 4 rotates, a real image is scanned, and the detection target point 2A is , 2B can be moved. Photomares 6a, 6b,..., 6n correspond to the detection target point 2A, and photomargins 9a, 9b,..., 9n
corresponds to the detection target point 2B. The optical fiber 5 corresponds to the detection target point 2A, and the optical fiber 8 corresponds to the detection target point 2B. Photomaru 6a, 6b,..., 6
n receives light through each element of the optical fiber 5 and converts it into an electrical signal. Photomaru 9a, 9b,..., 9
n receives light through each element of the optical fiber 8 and converts it into an electrical signal. The comparator 10a inputs the photomultiplies 6a and 9a, which correspond to each other at the detection target point, and compares the two in magnitude. Hereinafter, the comparator 10b is the photomaru 6b and 9b,..., the comparator 10
n compares photomars 6n and 9n. Or game 11 has all comparators 10a, 10b,..., 10n
Take the output of as input.

Furthermore, the entire surface of the sample is inspected by placing the sample on an X-Y table and scanning the image with the rotating polygon mirror 4 and scanning the sample with the X-Y table. Here, the X-Y table is scanned by moving it continuously in the direction perpendicular to the image scanning by the rotating polygon mirror 4 and stepwise in the parallel direction. The step amount is made the same as the image scanning width of the rotating polygon mirror 4.

With the above configuration, the detection target points 2A and 2B move in synchronization with each other and are used for pattern search and detection. That is, the surface of the sample 1 is scanned, and the pattern of each detection target point is comparatively detected through the above-mentioned detection system. Each of the above detectors 10a, 10
b, . . . , 10n have the same configuration. FIG. 2 shows an embodiment of the comparator 10n. The comparator 10n includes a subtraction circuit 12, an absolute value circuit 13, an analog comparator 14, and a potentiometer 14A for setting the allowable deviation value ε. The subtraction circuit 12 is a resistor 12
A, 12B, 12D, 12E, and differential amplifier 1
Consists of 2C. The absolute value circuit 13 has a resistor 13A,
It consists of 13B, 13C, diodes 13E, 13F, and differential amplifiers 13D, 13G. Absolute value circuit 1
3 outputs the absolute value deviation |Vn|, and the analog comparator 14 compares the allowable deviation value ε with |Vn|, and generates a defect signal when |Vn|>ε.

The operation is as follows. A subtraction circuit 12 inputting the outputs V _6o and V _9o of the photomultipliers 6n and 9n calculates the difference between the two, and an absolute value circuit 13 at the next stage obtains the absolute value of the deviation |Vn|. This absolute value |Vn| is determined by the allowable deviation value ε and the analog comparator 1 as described above.
Take the comparison at 4, and if |Vn|≦ε, 2 points 2
It turns out that the patterns in A and 2B match,
If |Vn|>ε, the patterns up to the two points 2A and 2B do not match, and a defect signal is output. Further, the output |V _o | of the absolute value circuit 13 is used for other processing. Note that as the allowable deviation value ε becomes smaller, even minute defects become undetectable.

There are many types of LSI wafers and masks, and the chip sizes vary depending on the type. Therefore, when inspecting samples with different chip sizes, it is necessary to adjust the distance between the objective lenses 3 and 7 to detect the same pattern portion.

When testing samples with different chip sizes,
Accurately adjust the distance between the objective lenses, but since moving the objective lens also moves the optical path of the optical system, adjust the optical components in the optical path from the objective lens to the optical fiber and the light receiving part of the optical fiber accordingly. There is a need. By accurately adjusting the position of the objective lens gap, optical path components, and optical fiber receiver, the relative error (displacement) between the two pattern detection positions is reduced, improving the accuracy of defect determination and making it possible to inspect even the smallest defects. It becomes possible. In other words, it is a necessary condition for defect determination that there is no deviation between the detection patterns captured by the two optical fibers.

The chip on the sample may be misaligned depending on its location. Even if the pattern for each chip is the same, if the chips are not aligned accurately in XY, a normal pattern will be mistakenly determined to be defective. Therefore, it is necessary to adjust the optical system before the test moves to another value chip, which makes continuous automatic testing impossible. In order to perform continuous automatic inspection, it is necessary to constantly check the position between the two detection patterns while performing defect determination, and adjust the detection pattern capture position so that there is no difference between the two detection patterns.

Normally, the amount of deviation between chips is minute, so there is no need to adjust the position of the objective lens and optical path components for the same sample, and only the optical fiber receiver can be adjusted.
The amount of deviation between the two detection patterns can be eliminated. Therefore, the key point in defect determination is how accurately the position of the optical fiber light receiving section is adjusted.

In addition, if one optical path (for example, the left side in Figure 1) is fixed and the other optical path (for example, the right side) is made movable, the detection pattern capture position can be adjusted to eliminate the amount of deviation between the two detection patterns. Adjustment is possible. In an embodiment of the invention, the first
Although the left optical path in the figure is fixed and the right optical path is movable, the same effect can be obtained even if the fixed side and movable side are reversed.

When inspecting different types of samples, if you adjust the objective lens 3, lens 4B, and the light receiving part of the optical fiber 5 before the test, you can achieve precision by just finely adjusting the light receiving part of the optical fiber 5 during the inspection. Defects can be determined.

In the present invention, the light receiving section of the optical fiber 5 is attached to the X-Y table, and a positioning control device is provided. The alignment control device takes in the output of the pattern check comparator while determining defects, processes it, calculates the amount of positional deviation, and then drives the X-Y table based on this calculated amount to eliminate the amount of positional deviation. control as follows.

Next, an example of the X-Y table is shown in FIG.
The X-Y table consists of a fixed frame 17A and a movable frame 18A. Optical fiber light receiving section 16
Guide rails 18E, 18 via sliding bearings
Supported by F. Guide rail 18E, 18
F is arranged in parallel on the movable frame 18A, and both ends of each are attached to the movable frame 18A with the pedestal 1.
It is fixed via 8D. The feed screw 18B is arranged parallel to the guide rails 18E and 18F, and is held by a bearing 18C to prevent movement of the feed screw. A nut is disposed on the optical fiber light receiving portion 16 and engages with the feed screw 18B.

The movable frame 18A is supported by guide rails 17D and 17G via sliding bearings. The guide rails 17D and 17G are arranged in parallel on the fixed frame 17A, and both ends of each guide rail are fixed to the fixed frame 17A via a pedestal 17E. Guide rails 17D and 17G are in an angular relationship perpendicular to guide rails 18E and 18F. The feed screw 17B is connected to the guide rail 17D, 1
7G and is held by a bearing 17C to prevent the feed screw from moving. The movable frame 18A is connected to the feed screw 17 via a nut 17F.
It is engaged with B.

The optical fiber 5 is inserted and fixed in the center of the optical fiber light receiving section 16 in parallel to the guide rails 18E and 18F. The position of the light receiving end 15 of the optical fiber 5 is adjusted so that the light reflected from the rotating polygon mirror 4 can be accurately received. The other end of the optical fiber 15
A is derived and connected to an external photomultiple.
The optical fiber light receiving section 16 is driven by a pulse motor 18 provided at one end of the feed screw 18B.
The feed screw 17B is configured to be movable in the Y direction by a pulse motor 17 provided at one end of the feed screw 17B.

FIG. 4 shows a drive circuit for driving the pulse motors 17 and 18. The position command circuit 21 operates in the x direction and the y direction.
The amount of movement in the direction (position adjustment signal) is output as a command value and sent to the drive circuits 19 and 20. The drive circuits 19 and 20 convert input command values into drive amounts for the pulse motors 17 and 18, and drive the pulse motors 17 and 18. Pulse motor 1
7 and 18 drive the feed screws 17B and 18B to bring the optical fiber light receiving section 16 to a predetermined position.
The position command circuit 21 has a function of inputting a position command manually given from the outside, and a function of inputting an error amount based on a position detection value during the movement process (so-called feedback control function). , a function to calculate the amount of displacement in each of x and y based on the captured values, and a function to output the calculated values. The error amount based on the position detection value of the position detection function in the feedback control function mentioned above is the amount of error based on the position detection value of the position detection function in the feedback control function.
This value is obtained by checking the output of the comparator shown in the figure, and will be explained in detail below.

FIG. 5 is a diagram showing the relationship between the arrangement of the optical fiber light receiving ends 15 and the scanning direction of the real image by the rotating polygon mirror. The direction in which the optical fibers are arranged is defined as x, and the direction perpendicular to this direction is defined as y. Furthermore, the scanning direction of the real image by the rotating polygon mirror is made in the direction of the arrow. Therefore, the y direction is the scanning direction of the real image by the rotating polygon mirror.

FIGS. 6 and 7 show a positional deviation calculation circuit used for alignment. Figure 8 is a time chart. The positional deviation calculation circuit in Fig. 6 is for calculating positional deviation in the y direction, and the positional deviation calculation circuit in Fig. 7 is for calculating the positional deviation in the y direction.
It is used to calculate the positional deviation in the direction, and both output the error amount based on the position detection value described above.

The positional deviation calculation circuit in FIG. 6 includes an addition circuit 23,
Sample hold circuit 24, AD converter 25, AND gate 26, addition circuit 27, latch circuit 2
8, 29, and a comparison circuit 30. Countable circuit 23
Absolute value input to |Va|, |Vb|,..., |Vn|
are the comparator outputs of the corresponding photomal outputs, and looking at the correspondence with Figure 2, |Va| is V ₆₁ and
| _Vb | is the absolute value output of the difference between V ₆₂ and V ₉₂ , ..., |Vn| is the absolute value output of the difference between V _6o and V _9o . The adder circuit 23 has |Va|, |Vb
Add |, ..., |Vn| to find the amount of positional deviation in the x direction. The sample hold circuit 24 samples and holds the output of the adder circuit 23, and the sample value is
An AD converter 25 performs AD conversion. AD converter 25
AD conversion period and AD conversion time are AND gate 2
It is determined by the output of 6. As shown in FIG. 8, the effective range, which is one input to the AND gate 26, is given by a specific time span within one scanning section of the rotating polygon mirror and is determined by the field of view of the objective lens. It's decided. Due to the clock pulse during this interval
AD conversion is performed. The adder circuit 27 adds up all the AD conversion outputs within the above-mentioned effective range within one scanning section, and averages the amount of x-direction positional deviation within the effective range. The sampling period of the AD converter 25 is determined by the diameter of the optical fiber light receiving surface and the real image scanning speed by the rotating polygon mirror. Now, the diameter of the optical fiber receiving surface is D (μm), and the real image scanning speed is v.
(m/s), a clock pulse of f=v/D (MPPS) is required.

Added data before one scan of the rotating polygon mirror is stored in the latch circuit 29, and the driving direction of the pulse motor is also stored. The addition data (δxn) obtained at the end of the effective range and the latch circuit 29
The output (added data δxn _-1 from one scan before is compared with the comparator circuit. As a result, if δxn < δxn _-1 , the positional deviation is small, so the pulse is sent via the position command circuit 21 in Fig. 4). The motor 17 is driven by a certain amount in the same direction as the direction in which it was driven one scan ago.δ
If xn>δxn _-1 , it is assumed that the positional deviation has become even larger, and the pulse motor 17 is driven by a certain amount in the opposite direction via the position command circuit 21 shown in FIG.
In FIG. 6, the outputs CW and CCW of the comparator circuit 30 respectively indicate the comparison results. For example, CW is δxn>δxn _-1
CCW is an output that becomes "1" when δxn<δxn _-1 . In addition, CW may be a signal indicating the polarity of (δxn - δxn _-1 ), and CCW may be a deviation amount (δxn - δxn _-1 ). Either one is selected depending on the internal configuration of the position command circuit 21. CW and CCW are the error amounts based on the position detection value described above.

When the driving of the pulse motor 17 is completed, the newly obtained addition data is stored in the latch 29. Thereby, the real image detection position is adjusted for each scan, and if a deviation occurs, the pulse motor 17 is driven to correct it. FIG. 8 also shows the drive timing of the pulse motor 17 and the latch pulse timing of the latch circuit 29.

The positional deviation calculation circuit in FIG. 7 includes a sample hold circuit 31, an AD converter 32, and an AND gate 3.
3. It consists of an adder circuit 34, latch circuits 35 and 36, and a comparison circuit 37. The operation timing is substantially the same as in FIG. The absolute value |Va| of the difference between the corresponding photomultiplies is selected (any value is acceptable) and formed in the sample-and-hold circuit 31. The sample and hold circuit 31 samples and holds |Va| and sends it to the AD converter 32.

The AD converter 32 performs AD conversion under the control of an AND gate 33, adds an addition circuit 34, and causes a latch circuit 35 to latch. Latch circuit 3
The latch output of No. 5 is compared with the latch output of the previous scan, which is latched by the latch circuit 36, in the comparison circuit 37, and the δyn and δ of the previous scan and the current scan are compared.
Compare with yn _-1 . The comparison results are output as CW and CCW, input to the position command circuit 21, and drive the pulse motor 18. The output of the comparison circuit 37 becomes the error amount, and depending on the polarity, the pulse motor 18
is driven. This performs alignment in the y direction. In this embodiment, a photomultiplier was used as the photoelectric conversion means, but a photoelectronic device such as a photodiode may also be used.

According to the present invention, it was possible to efficiently automatically scan patterns on LSI wafers and masks. Furthermore, since the real image detection position is changed so as to eliminate the amount of positional deviation, it becomes possible to detect minute defects that could not be detected conventionally.

[Brief explanation of the drawing]

FIG. 1 is an embodiment of the pattern detection system of the present invention.
Fig. 2 is an embodiment of a comparison circuit of the present invention, Fig. 3 is an embodiment of an X-Y table of the invention, Fig. 4 is an embodiment of a pulse motor drive circuit, and Fig. 5 is an embodiment of an optical fiber. Diagram showing the relationship between the light receiving surface and the scanning direction, No. 6
FIG. 7 is an embodiment of a positional deviation calculation circuit in the x direction of the present invention, FIG. 7 is an embodiment of a positional deviation calculation circuit in the y direction of the invention, and FIG. 8 is a time chart. 1... Sample, 2A, 2B... Detection point, 4... Rotating polygon mirror, 5, 8... Optical fiber, 6a to 6n,
9a to 9n...Photomaru, 10a to 10n...Comparison circuit, 16...Optical fiber light receiving section, 17, 18
...Pulse motor.

Claims (1)

[Claims] 1. Irradiating light onto patterns at corresponding relative positions on the same pattern of at least two sections at different positions on a sample surface where sections having the same pattern are repeatedly arranged on the surface. a first and second imaging optical system that forms an image of the reflected light from the two patterns, and a first, which is imaged by the first and second imaging optical systems A rotating polygon mirror for scanning a second image, and a light receiving surface arranged in a direction perpendicular to the direction scanned by the rotating polygon mirror, each of the above-mentioned,
A first device that receives and guides the first and second images;
A second optical fiber group, and first and second optical fiber groups, which are provided corresponding to the first and second optical fiber groups, capture the detection image obtained from each optical fiber, and convert it into an electrical signal independently. a photomultiplier group and a comparison means for comparing electrical signals obtained from the photomultipliers in a corresponding relationship between the first and second photomultiplier groups to determine a defect in the pattern; The surface is attached to an X-Y table whose position can be adjusted, and a drive system is provided to finely drive the X-Y table in the X-axis and Y-axis directions,
Taking in a signal indicating pattern positional deviation from the comparison means to obtain a position adjustment signal for positional deviation correction;
A pattern inspection device comprising circuit means for applying the position adjustment signal to the drive system in order to eliminate the positional deviation.