JPS6327749B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention provides a mask alignment apparatus for aligning the relative positional relationship between a wafer for a semiconductor integrated circuit and a mask corresponding thereto.
The present invention relates to an image signal binarization threshold setting method for setting a binarization threshold for an image signal for detecting alignment marks provided on a wafer.

[Prior art]

First, a conventional example will be explained according to the drawings.

FIG. 1 is a configuration diagram of an optical system of a general mask alignment apparatus, and FIG. 2 is an image signal waveform diagram showing the principle of detection of alignment marks on a mask and wafer.

The detection light 15 is reflected and irradiated onto the alignment marks 12 and 13 provided on the mask 10 and the wafer 11 by a half mirror 17, and the reflected light is magnified by an objective lens 16 to obtain the optical image 18. , this is scanned and detected by a photoelectric element 19 such as an image sensor and converted into an image signal.

Here, for example, the wafer 1 is placed in the center of the alignment mark images 12A, 12B (distance 1) on the mask 10.
The table 14 is moved so that the No. 1 alignment mark image 13A comes, and the relative positions of the mask 10 and the wafer 11 are aligned.

In this case, the image signal of the optical image (photoelectric element 1
9 output) From the waveform, each alignment mark 12,1
In order to obtain the position information of 3, it is necessary to find the intersection of the position detection signals P ₁ , P ₃ and P ₂ of each of the alignment mark images 12A, 12B and 13A and the threshold level for detecting them. Must be.

However, the image signal generally has a waveform that tends to have "undulations" (level fluctuations in the baseline) over the entire image signal due to unevenness in illuminance of the detection illumination.

Therefore, as shown in Fig. 2a, if a horizontal (constant value) threshold level H is set for the image signal J (baseline K), there will be intersections A and B at the periphery, and Position detection signal is interfered with
It is not possible to accurately capture P ₁ , P ₂ , and P ₃ .

Therefore, as shown in FIG. 2b, the position detection signals P ₁ , P ₂ , and P ₃ must be captured by setting each threshold level I to follow the baseline K.

On the other hand, the conventional method, for example, performs predetermined sampling on the image signal J,
The average value and variance of the sampled values are calculated within each interval by dividing the predetermined number of samples, and in the interval where the variance is small, the average value itself is used as the threshold for that interval, and the variance value is For a large section, the threshold value was determined by performing linear correction based on the average value of both sections with small variance values sandwiching the section.

In this conventional method, constants obtained from experience, experiments, etc. must be used to determine the size of the variance value, so for example, constants must be set each time according to the smoothness of the wafer surface in various processes. Constants had to be changed, which was complicated and inefficient, and the reliability was not sufficient.

[Purpose of the invention]

An object of the present invention is to provide an efficient image signal binarization threshold setting method that eliminates the drawbacks of the above-mentioned conventional techniques and can ensure binarization even if there are level fluctuations in the baseline of the image signal. Our goal is to provide the following.

[Summary of the invention]

The configuration of the image signal binarization threshold setting method according to the present invention is based on a baseline of an image signal obtained by scanning each alignment mark provided on a wafer for a semiconductor integrated circuit and a mask corresponding thereto. In a mask alignment device that has a function of identifying each alignment mark using a binarization threshold set to follow the above-mentioned wafer and mask, and aligning the relative positional relationship between the wafer and the mask, The image signal of each alignment mark is sampled at a predetermined period, and the sampled values are sequentially divided into threshold unit intervals for each predetermined number of continuous marks, and the average value and variance value are calculated for each sampling value within each threshold unit interval. , and use the maximum value of the sampling variance values in the threshold unit section where no image signal of each alignment mark exists as a reference,
For one or more consecutive threshold unit intervals with a sampling variance value exceeding this, each sampling average value is interpolated and corrected from the sampling average values of both threshold unit intervals sandwiching it, and a series of sampling average values using this is interpolated and corrected. Based on this, a binarization threshold that follows the baseline of the image signal is set.

Note that the maximum value of the sampling variance value of the threshold unit interval in which the image signal of each of the above alignment marks does not exist is the minimum number of threshold unit intervals that can include the image signal of each of the above alignment marks to cover the entire image interval. The average value of the sampling variance value is calculated for each divided section, and the maximum sampling variance value in the divided section where the average value is the minimum is used.

[Embodiments of the invention]

Embodiments of the present invention will be described below based on the drawings.

FIG. 3 is a block diagram of an embodiment of the binarization threshold setting section of the mask alignment apparatus using the image signal binarization threshold setting method according to the present invention, and FIG.
It is a waveform diagram of the main part.

Here, 1 is a photoelectric device (CCD) related to image signal generation, 2 is a sample hold circuit (SH), and 3
is a clock generator (CLOCK), 4 is a timing circuit (TIMING), 5 is an A/D converter (A/D),
6 is a processing circuit (COMP), 7 is a register (REG),
8 is a D/A converter (D/A), and 9 is an output terminal (OUT).

The photoelectric element 1 generates an image signal v according to its optical input.
Output. An example of the waveform is shown in FIG. 4a.

The sample and hold circuit 2 is connected to the clock generator 3.
The image signal v is sampled and held sequentially according to the timing of the sampling clock from , and each sampling value Vi (i=1, 2, . . . ) is obtained. An example of the waveform is shown in FIG. 4b.

Here, the period τ of the sampling clock may be less than or equal to a value that allows the sampling value Vi to approximate the image signal v to a degree that does not cause any practical problems.

On the other hand, the timing circuit 4 generates a timing pulse synchronized with the sampling clock, and starts the A/D converter 5 with this timing pulse.

As a result, the A/D converter 5 takes in the sampling value Vi and converts it into a corresponding digital value V.
(i) and sequentially input and store it in the processing circuit 6.

The processing circuit 6 calculates and stores the average value (j) and variance value E(j) of the digital value V(i) for each period Nτ for a predetermined number N consecutive in the storage order using the following equation. .

(j)={ _N 〓 ^k=1 V(j・N+k)}/N…(1) Examples of these calculation results are shown in FIGS. 4c and 4d.

Here, the period Nτ for the predetermined number N of digital values V(i) will be referred to as a threshold unit interval since various calculation processes regarding the binarization threshold will be performed hereafter as a unit. The length of this threshold unit interval Nτ is such that the image signal of the alignment mark can be identified even if the corresponding binarization threshold is constant within the interval. The following are sufficient.

Next, the processing circuit 6 converts all threshold unit sections (all image signal sections) into sections L1, L2, Divide into L3, L4, ... and calculate the average value of each sampling variance E(j),
Determine the interval (for example, L1) that shows the minimum value,
The maximum value E of the sampling variance value E(j) within the interval L1 is determined.

This value E indicates the maximum value of the sampling variance value E(j) in the section (for example, L1) that does not include the alignment mark detection signal, and the magnitude of each sampling variance value E(j) is determined. This is the reference value. That is, the maximum value of the sampling variance value E(j) among the threshold unit sections that do not include the detection signal of the alignment mark is used as the reference.

Each sampling variance value E(j) is compared with the above reference value E, and the threshold unit interval exceeding this value (for example, those in the intervals L2 and L3 in Fig. 4d)
Interpolate and modify as shown below.

For example, m in the threshold unit interval p~(p+m-1)
In each interval, each sampling variance value E
(p) to E(p+m-1) are consecutive to the above reference value E
When it exceeds, the sampling average value (p-1), (p+m) of both threshold unit intervals (p-1), (p+m) sandwiching this continuous interval is used as the reference, and the sampling average value (p) between the two threshold unit intervals (p-1), (p+m) is the standard. ~(p+m
-1) is interpolated in a stepwise manner using the following equation to obtain the respective correction values W, p to W (p+m-1).

W(p+n-1)=(p-1)+ {(p+m)-(p-1)}n/(m+
1) ...(3) (n = 1, 2, ..., m) However, as an exception, if the start end of the continuous section above coincides with the start end of the full-image signal section, or similarly the ends coincide with each other, then All correction values of the sampling average values of continuous intervals are interpolated and corrected so as to be replaced with the sampling average values of the first threshold unit interval where the continuous interval ends or the threshold unit interval immediately before the start of the continuous interval.

When such interpolation/correction is performed, the series of sampling average values (j) becomes a series of corrected values W(j), as shown in Figure 4e, which is exactly the baseline of the image signal v. It will be followed.

Therefore, as shown in FIG.
By setting the value threshold T(j), the intersection point of the already stored sampling value Vi and the detection signal of the alignment mark provided on the mask
C1, C2, C5, C6, and the intersection point C3 with the detection signal of the alignment mark provided on the wafer,
C4 can be detected and its binary signal w can be obtained.

Incidentally, when it is necessary to output the above series of modified values W(j) as an analog signal, the data stored in the processing circuit 6 is sequentially set in the register 7 in synchronization with the clock from the clock generator 3. D/
The analog threshold signal may be converted into an analog signal by the A converter 7 and sent from the output terminal 9.

In this way, it is possible to always follow the baseline of the image signal and set an appropriate binarization threshold. It is possible to obtain a binarized signal in which the image signal of the service mark is accurately identified. Note that when handling discrete (staircase) signals such as image signals from a CCD image sensor, it is possible to eliminate the need for sampling, making it possible to realize the apparatus economically.

〔Effect of the invention〕

As described above in detail, according to the present invention, even if there is a level fluctuation in the baseline of an image signal, it is possible to ensure the binarization without the need for special processing such as changing constant settings. Because it is possible to
A remarkable effect can be obtained in improving the efficiency and reliability of the mask alignment device.

[Brief explanation of drawings]

Fig. 1 is a diagram of the optical system configuration of a general mask alignment device, Fig. 2 is an image signal waveform diagram showing the principle of detection of alignment marks on the mask and wafer, and Fig. 3 is a diagram of the optical system according to the present invention. Mask alignment device using image signal binarization threshold setting method 2
FIG. 4 is a block diagram of an embodiment of the valuation threshold setting section, and is a waveform diagram of its main parts. DESCRIPTION OF SYMBOLS 1... Optical element, 2... Sample hold circuit, 3... Clock generator, 4... Timing circuit, 5... A/D converter, 6... Processing circuit, 7...
...Register, 8...D/A converter, 9...Output terminal.

Claims (1)

[Claims] 1. Binary values set to follow the baseline of an image signal obtained by scanning each alignment mark provided on a wafer for a semiconductor integrated circuit and a mask corresponding thereto. In a mask alignment device that has a function of identifying each of the alignment marks using a threshold value and aligning the relative positional relationship between the wafer and the mask, images of each alignment mark provided on the wafer and the mask are used. The signal is sampled at a predetermined period, the sampled values are sequentially divided into threshold unit sections for each predetermined number of consecutive samples, the average value and the variance value are determined for each sampling value within each threshold unit section, and the average value and variance value are calculated for each of the above-mentioned alignments. Based on the maximum value of the sampling variance values in the threshold unit interval where there is no mark detection signal, for one or more consecutive threshold unit intervals with sampling variance values exceeding this value, the sampling average of both threshold unit intervals that sandwich it. An image characterized in that each sampling average value is interpolated and corrected from the values, and a binarization threshold that follows the baseline of the image signal is set based on a series of sampling average values using this. Signal binarization threshold setting method. 2. In what is stated in claim 1,
The maximum value of the sampling variance value of the threshold unit section in which there is no detection signal of each alignment mark is determined by dividing the entire image section into the minimum number of threshold unit sections that can include the image signal of each alignment mark, and An image signal binarization threshold setting method that calculates the average value of sampling variance values for each divided section, and allocates the maximum sampling variance value in the divided section where the average value is the minimum.