JP2532926B2 - Alignment mark position detection method - Google Patents

Description

The present invention relates to a mask formed on a mask in a proximity aligner or stepper used in an exposure apparatus such as an X-ray exposure apparatus that prints a circuit pattern on a wafer. The present invention relates to a method of detecting an alignment mark including a mark and a wafer mark formed on a wafer, and obtaining a relative position between the mask and the wafer from the detected waveform with high accuracy.

[Prior Art] In general, two types of methods, a diffraction grating method using a Fresnel zone or the like and a pattern measuring method, are used as a method for obtaining the position of an alignment mark from a detected waveform with high accuracy. Among them, the diffraction grating method has a drawback that the detection accuracy is greatly influenced by the variation of the gap between the mask and the wafer, and the alignment mark having a step cannot be detected accurately. On the other hand, the pattern measurement method has a drawback that the accuracy depends on the shape of the pattern of the alignment mark to be detected, and thus the detection accuracy deteriorates due to the deterioration of the pattern shape of the alignment mark. Here, an example in which the diffraction grating method depends on the symmetry of the alignment mark is 19
IEEE TRANS Transaction on Electron Devices issued in April 1979 (IEEE TRANS
ACTION ON ELECTRON DEVICES) VOL.ED-2
6, No. 4 "Automatic alignment system
For Optical Projection Printing (Automatic Alignment System for Optical Project)
Ion Printing) ”by Gijs Bouwhuis and others. Also,
An example in which the pattern measuring method depends on the symmetry of the alignment mark is disclosed in JP-A-62-54918 and JP-A-61-199633.

The pattern measurement method is roughly classified into two methods, a contrast method and an edge detection method.

Among these, the conventional example of the contrast method is disclosed in Japanese Patent Laid-Open No. 61-2361.
No. 17 is disclosed. In this Japanese Patent Laid-Open No. 61-236117, a symmetric pattern matching process represented by the following equation is performed.

Or However, D (n) is a differential value, M is a comparison range, and n is a position.

The above expression is an application of the pattern matching processing in the normal image processing, and this Japanese Laid-Open Patent Publication No. 61-236117 utilizes the fact that the pattern of the alignment mark is line symmetric, and this is expressed in the above expression. I am incorporating it. That is, to find the center line of a line-symmetric figure, fold the figure in half,
It is equivalent to performing the operation to find the center line. The place of folding in half is of course an edge, but this uses an edge processed by the edge extraction method by differentiation which is a general image processing method.

As described above, in Japanese Patent Laid-Open No. 61-236117, the symmetry of the figure can be taken into account and the center line can be obtained by a relatively simple arithmetic expression. By using this symmetric pattern matching process to detect the alignment mark on the mask or the wafer by the proximity aligner or stepper, the relative positional relationship can be obtained from the center line of each alignment mark.

[Problems to be Solved by the Invention] However, the method using the symmetric pattern matching process described in Japanese Patent Laid-Open No. 61-236117 has the following drawbacks.

That is, this method presupposes that the figure is symmetrical. Ideally, the mask mark (chrome forming mark) on the mask and the wafer mark (resist forming mark or process forming mark) on the wafer are line symmetric, but in reality, the symmetry is broken by various factors. .

For example, in the case of a resist formation mark, the symmetry of the alignment mark is broken due to the irradiation angle of X-rays when exposing the alignment mark, the developing speed of the resist, etc. (see FIG. 2). Further, in the case of the process formation mark, the symmetry of the alignment mark is likely to be broken, and the alignment mark can be easily deformed due to thermal deformation (chemical deformation, deformation due to force) of the process or a thin film applied on the alignment mark. The symmetry of is broken (see Fig. 3).

As is clear from the above description, in the actual wafer mark on the wafer, it is extremely difficult to guarantee the symmetry in all layers (layers). Therefore, using the data of the alignment mark which is not symmetric with respect to the calculation assuming the symmetry causes a contradiction, and the calculation result includes an error. That is, as the symmetry of the alignment mark is broken, the error included in the calculation result is further increased.

Examples of measures for preventing the accuracy from being deteriorated due to the asymmetry of the alignment mark are disclosed in the above-mentioned JP-A-62-54918 and JP-A-61-199633. In JP-A-62-54918, the film thickness distribution of the photoresist coated on the wafer is measured, the wafer alignment pattern position detection error is determined from the measured film thickness distribution, and the alignment amount is determined based on the position detection error. Is being corrected. However, this method has a drawback that the processing becomes complicated. Further, in JP-A-61-199633,
The shape of the alignment mark itself does not easily cause a position detection error. However, this JP-A-61
The alignment mark disclosed in Japanese Patent Laid-Open No. 199633 has a drawback that its shape is very complicated.

[Means for Solving the Problems] The alignment mark position detection method according to the present invention is
An analog image signal obtained by imaging an alignment mark consisting of a wafer mark formed on a wafer and a mask mark formed on a mask with an image pickup device is converted into a digital image signal, and the converted digital image signal is processed. In the method for detecting the relative position between the wafer and the mask by performing the above, the wafer mark and the mask mark are a wafer mark differential signal obtained by first differentiating the wafer mark and a mask obtained by first differentiating the mask mark. The mark differential signal has a mark shape that maintains the similarity of being substantially overlapped with each other when horizontally moved, the digital image signal processing differentiates the digital image signal, and with respect to the differentiated signal, (Here, V '(j) is the differential value of the digital image signal,
R is a correlation section, P is a start point of the correlation section, and W is a similarity pattern matching process such as the distance between the center position of the mask mark and the center position of the wafer mark which are adjacent to each other. The relative position between the wafer and the mask is detected from the position where the result is maximum.

[Embodiment] An embodiment of the present invention will be described below with reference to the drawings.

 First, the principle of the present invention will be described.

The present invention has a feature that a part of the pattern matching method generally applied in the field of pattern measurement is modified. The flow of processing is a procedure for performing edge extraction of a pattern by operating a differential operator on an input image, and performing similarity pattern matching on the edge-extracted data. The definition of similarity pattern matching and its method will be gradually clarified by the following explanation. In the similarity pattern matching, an expression obtained by modifying a part of the known autocorrelation function is used.

In normal pattern matching or waveform data processing, the autocorrelation function is defined by the following equation.

Here, i = 0, 1, ..., N−1, w (i) is a finite number N of sample values, and Rx (i) is an autocorrelation function.

By applying the above equation to data or signal waveform for the purpose of matching, it is possible to obtain a correlation result of a pattern or signal waveform serving as a certain reference. As is clear from this equation, this calculation requires a considerable number of sum-of-products calculations, so the calculation takes time. For this reason, the autocorrelation function is suitable for batch processing, but cannot fully support real-time processing.

For example, when the above equation is used for alignment processing of an X-ray exposure apparatus, the long calculation time means that the control system including the X / Y stage also uses the alignment processing result as a feedback signal. It becomes a big obstacle. Furthermore, the long calculation time is the biggest cause of the decrease in the throughput of the entire apparatus, which may cause a fatal defect.

Therefore, in the present invention, in order to shorten the calculation result of the autocorrelation function represented by the above formula to several hundredths or several thousandths, the autocorrelation function is devised after devising the shape of the alignment mark. The equation was transformed as follows.

(1) Deviation of alignment mark shape An alignment mark shape that allows a general superposition function by an autocorrelation function in a data signal whose edges have been extracted by applying a first-order differential operator to the input image of the alignment mark. And

That is, the wafer mark formed on the wafer and the mask mark formed on the mask are moved in parallel by a wafer mark differential signal obtained by first-order differentiation of the wafer mark and a mask mark differentiated signal obtained by first-order differentiation of the mask mark. For example, the shape of the mark was such that the two overlapped with each other and maintained similarities.

For example, if the pattern is as shown in FIG. 4, a superimposing method using an autocorrelation function is possible. In FIG. 4, (a) is a top view, (b) is a side view, (c) is a raster compressed image V, and (d) is a raster compressed image 1.
Each of them represents the secondary differential value V '. As is apparent from FIG. 4 (d), if the first-order differential value V'of the raster compressed image V is moved in parallel to the left and right, the positions L1 and L2
So it almost overlaps with itself. Therefore, (L2 + L1) / 2
Thus, the relative position between the mask and the wafer can be measured.

Another example is shown in FIG. In FIG. 5, (a) is a top view, (b) is a side view, (c) is a raster compressed image V, and (d) is a primary differential value V ′ of the raster compressed image V, respectively. In the example of FIG. 5, as in the case of FIG. 4, the relative position of the mask and the wafer can be measured by (L2 + L1) / 2.

In the examples of FIGS. 4 and 5, if the wafer mark and the mask mark are translated in parallel, the wafer mark differential signal obtained by the first-order differentiation of the wafer mark and the mask mark differentiated signal obtained by the first-order differentiation of the mask mark are almost the same. Although the mark shapes are maintained so as to overlap each other, the wafer mark and the mask mark have a wafer mark differential signal which is the first derivative of the wafer mark and a mask mark differential inversion signal which is the first derivative of the mask mark and the polarity of which is inverted. If the marks are moved in parallel, the marks may be shaped so as to overlap each other and maintain the similarities.

(2) Formula modification of autocorrelation function In the present invention, in order to modify the formula of the autocorrelation function, attention is paid to the feature of the alignment mark used in the exposure apparatus, and an analog image obtained by imaging the alignment mark with an imaging device is used. The signal (edge pattern) obtained by operating the first-order differential operator on the image signal (input image) is transformed based on the following three points different from the normal signal waveform.

[1] The distance W between two edge patterns for pattern matching is basically a constant value, and the variation of the distance W due to the process is extremely small.

[2] The range for pattern matching may be set only near the edge pattern.

[3] Mask at detection coordinates (coordinates on TV camera) /
The position of the alignment mark on the wafer may be considered to be almost at a fixed position due to coarse alignment (coarse alignment) in the alignment procedure.

Based on these features [1], [2], and [3], the process of formula transformation of the autocorrelation function will be described below.

The autocorrelation function for discrete values is expressed by the following equation as described above.

However, i = 0, 1, ..., N−1.

Here, the normalization coefficient 1 / N is omitted to simplify the expansion.

Equation (2) is used for the data shown in FIG. In this figure, the horizontal axis represents the pattern distance. The image input uses an image captured by a two-dimensional camera, and the discrete distance is represented by the number of pixels (j) on the imaging tube surface of the two-dimensional camera. The vertical axis represents the result of converting the analog image signal captured by the two-dimensional camera into a digital image signal and applying a first-order differential operator to the digital image signal obtained by this conversion. ), It is represented by V '. Therefore, in FIG.
Expression (2) can be replaced with the following expression.

From the feature [1], the distance W between the wafer mark and the mask mark is basically a constant value, and in the formula (3), the width W corresponds to the delay time i.

Expression (4) finds the correlation with respect to the distance W between the left and right marks in the entire area of j in FIG.

However, from the feature [2], the pattern for obtaining the correlation (superposition) may be set only for the portion where the edge pattern exists, and as is clear from FIG. 6, the range where the mask mark and the wafer mark exist. Is finite.

Further, from the feature [3], P is given to the position where the edge pattern exists as the starting point of the correlation section from the rough position detection performed in advance.

In FIG. 6, assuming that the range in which the wafer mark exists is a correlation section R and the starting point of the correlation section is P, the range of correlation is a range from j = P to j = P + R in FIG. Limited. This is shown in FIG.

 From this, the equation (4) is modified as follows.

The expression (5) is defined as a function of the width W, but from the feature [2] of the alignment mark, the correlation section R
Also, the start point P of the correlation section is a parameter of A. Therefore, if the correlation section R and the start point P of the correlation section are also parameters, the equation (5) is converted as follows.

From the above expansion, the formula (2) of the autocorrelation function could finally obtain the formula (6) by considering the features [1], [2] and [3] of the alignment mark. (6)
The expression can be regarded as a basic definition expression of similarity pattern matching, and the following six expressions can be considered as expansion expressions thereof.

In other words, A (W, P, R) does not have W, P, R as parameters at the same time,
It is an equation for similarity pattern matching having one or a combination thereof as a parameter.

However, as a result of the inventor's experiment, the function A (P, W,
It became clear that R) changes rapidly with changes in W. That is, the starting point P of the remaining correlation and the correlation section R
For, it was found that A has almost no effect on the value of W having the maximum value.

As for P, unless a very inappropriate value is substituted,
The value of W for which A has the maximum value does not change, and the correlation interval R should be a variable because the thickness of the edge changes depending on the type of process mark, but in reality it is considered to be a constant value. I can hardly do it.

By the way, according to the experiment, R = 32 (constant value) was used for all process marks, but the target detection accuracy was about 0.01 μm for all patterns.

From the above, the expression having a substantial meaning among the expanded expressions (6-1) to (6-6) is the expression (6-3), which is the basic expression.

Further, from the above, R can be considered as a constant (6−
Partly transforming 3), Is obtained.

FIG. 8 shows the result obtained by the experiment using the equation (7).

In FIG. 8, the horizontal axis shows the width W, the depth axis shows the inter-contact point P of the correlation, and the vertical axis shows the function A (P, W). It can be seen from FIG. 8 that the peak position of the function A (P, W) is constant with the width W = 204 regardless of P.

Further, in FIG. 8, the height of the peak of the function A (P, W) changes depending on the value of R and affects the interpolation accuracy, but the experimental results show that the detection on the 0.01 μm order is significantly affected. It was confirmed that there was not.

Next, the number of calculations is compared between the autocorrelation functions (2) and (7).

The number of data is assumed to be 1000. In the case of the autocorrelation function (2), since the number of integrations is 999 and the number of summations is 998 for each data, (999+
998) × 1000 = 1997000 times of operations. On the other hand, the number of calculations of the equation (7) is calculated based on the following assumption.

In the equation (7), when a certain process formation mark is considered, the scale per pixel is 0.1 μm, the starting point P is 100, the correlation section R = 30 (3 μm), and the expected pattern width W is WMIN = 200 (20 μm), WMAX = 210 (21 μ
m) and substitute it into the equation (7).

However, in the case of W = 200,201, ..., 209,210 (8), since the number of times of integration is 30 and the number of times of summation is 29 for each W, the total number of operations is (30 + 29) × 11.
= 649 times. Therefore, it is reduced to about 1/3077.
If it is actually this number, digital signal processing (DSP)
Even with a filter, it is possible to calculate in less than 1ms. In addition, for the multiply-accumulate operation (MAC: mulu
The speed can be further increased by using a tiplier-accumulatror) chip.

As described above, by applying the autocorrelation function only to the alignment mark, the number of calculations can be extremely reduced.

Further, in the present invention, the symmetry of the pattern is not presupposed from the expression (7), but depends on the similarity between the mask mark and the wafer mark.

That is, according to the method of the present invention, the equations (6-3) and (7) are irrespective of the symmetry of the mask and the wafer alignment mark,
It is called similarity pattern matching because it is a pattern matching method using similarity.

Next, in FIG. 4, the processing result by the equations (6-3) and (7) is shown in FIG. In Fig. 9, A (W, P)
W of the first peak is designated as W1, and that of the second peak is designated as W2. Since the true peak is present in the middle of W1 and W2, by interpolation operation, obtains the value of W indicating the true maximum value, it puts the W _1-2. As a result, L1 = W _1-2 is obtained.

Next, the same process is performed for L2, and if the obtained W is W _3-4 , L2 = W _3-4 is obtained. From this, the relative position of the mask mark and the wafer mark is obtained from (L1 + L2) / 2.

 Next, examples of the present invention will be described.

In FIG. 1 and FIG. 4, in the present embodiment, the wafer mark and the mask mark are moved in parallel by the wafer mark differential signal obtained by the first-order differentiation of the wafer mark and the mask mark differential signal obtained by the first-order differentiation of the mask mark. For example, it has a mark shape in which the similarities are substantially overlapped with each other.

This alignment mark is a two-dimensional camera (imaging device)
An image is captured (not shown) to obtain an analog image signal. The analog image signal is an analog-digital converter (not shown)
Is converted into a digital image signal. Here, the number of gradations for digitization is 6 bits or more, preferably 8
A bit desirable. By synchronously integrating this digital image signal, noise is removed, the S / N ratio is increased, and at the same time, data compression is performed. Here, the number of times of integration may be any number. Next, the noise-removed and data-compressed digital image signal is differentiated to extract the edge (outline) of the alignment mark. Similarity pattern matching processing represented by equation (6-3) or equation (7) is performed on the differentiated data.

As a result of the processing, the maximum value position is obtained, and then the inter-mark distances L1 and L2 are obtained more accurately by interpolation processing, and the relative position between the mask mark and the wafer mark is obtained by (L1 + L2) / 2.

Needless to say, the shape of the alignment mark is not limited to this embodiment, as long as the data obtained by processing the input image with the differential operator maintains the similarity.
Any shape is acceptable.

This method is called "position detection device for two objects separated by a minute distance"
An alignment system in an exposure apparatus such as an X-ray exposure apparatus can be configured by combining the processing apparatus with the above processing apparatus. This alignment system enables to set an arbitrary gap and to realize a highly accurate and high throughput alignment system with extremely small process dependency.

Further, the applicable range of the present invention is not limited to the exposure apparatus, but can be applied to those in which the shape and distance of the pattern are estimated in advance in other application fields of pattern recognition. in this case,
By giving appropriate parameters in the similarity pattern matching formula, the calculation time can be shortened from several hundredths to several thousandths as compared with the case of using the conventional autocorrelation function or cross-correlation function. .

EFFECTS OF THE INVENTION As is apparent from the above description, according to the present invention, a wafer mark differential signal obtained by first differentiating a wafer mark and a mask mark and a mask mark differential signal obtained by first differentiating a wafer mark. If and move horizontally, the marks will have overlapping shapes and the similarity will be maintained, and by performing similarity pattern matching on the signal obtained by differentiating the image input signal, the relative position of the wafer mark and mask mark can be processed with high accuracy and with ease. There is an effect that can be recognized in.
Therefore, throughput can be improved, which can greatly contribute to simplification of the production process and improvement of production efficiency.

[Brief description of drawings]

FIG. 1 is a diagram for explaining a method of detecting the position of an alignment mark according to an embodiment of the present invention, FIG. 2 is a diagram for explaining the breaking of the symmetry of a conventional resist formation mark, and FIG. 3 is a conventional diagram. For explaining the breakage of the symmetry of the process forming mark, and FIGS. 4 and 5 are views for explaining that the alignment marks of the present invention can be superposed by an autocorrelation function. FIG. 6 is a diagram showing the first-order differential value V ′ of the image signal of the alignment mark and the distance W between the mask mark and the wafer mark in FIG. 5, and FIG. 7 is a graph showing the start point P of the correlation section in FIG. A diagram showing a correlation section R,
FIG. 8 is a diagram showing an experimental result of the autocorrelation function A (P, W) according to the present invention, and FIG. 9 is a diagram showing an example of a result of the similarity pattern matching according to the present invention.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (56) Reference JP-A 61-104203 (JP, A) JP-A 62-213124 (JP, A) JP-A 60-209104 (JP, A) JP-A 61- 126407 (JP, A) JP 63-274144 (JP, A) JP 63-3244 (JP, B2) JP 3-7242 (JP, B2) JP 5-19297 (JP, B2) Japanese Patent Publication 6-63729 (JP, B2)

Claims (2)

(57) [Claims] 1. An analog image signal obtained by picking up an alignment mark composed of a wafer mark formed on a wafer and a mask mark formed on a mask with an image pickup device is converted into a digital image signal, and this converted image signal is converted. A method of detecting a relative position between the wafer and the mask by processing a digital image signal, wherein the wafer mark and the mask mark are the wafer mark differential signal obtained by first-order differentiating the wafer mark and the mask mark. The mask mark differential signal obtained by the first-order differentiation has a mark shape that substantially overlaps with each other when horizontally moved, and the digital image signal processing differentiates the digital image signal. This is a process for performing similarity pattern matching. From the position where this process result is maximum, Position detecting method of the alignment marks and detecting the Fine a relative position of the mask. 2. The similarity pattern matching process, (Here, V '(j) is the differential value of the digital image signal,
2. The alignment mark according to claim 1, wherein R is a correlation section, P is a start point of the correlation section, and W is a distance between a center position of the mask mark and a center position of the wafer mark which are adjacent to each other. Position detection method.