JP2692890B2 - Positional relationship detector - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial application] The present invention relates to an apparatus for detecting a relative positional relationship between two objects with high accuracy, for example, for detecting a positional deviation between a mask and a wafer in a semiconductor printing apparatus. It concerns the detection of intervals.

[Conventional technology]

2. Description of the Related Art Conventionally, in an exposure apparatus for manufacturing a semiconductor, relative positioning of a mask and a wafer has been an important factor for improving performance. Particularly in the recent alignment and gear setting in an exposure apparatus, one having a positioning precision (gear setting precision) of, for example, submicron or less is required for high integration of semiconductor elements.

In many alignment apparatuses, a so-called alignment pattern for alignment is provided on a mask and a wafer surface, and both alignments are performed using positional information obtained from the alignment patterns. As an alignment method at this time, for example, the amount of deviation between the two alignment patterns is detected by performing image processing, or alignment is proposed as proposed in U.S. Pat. No. 4,037,969 or Japanese Patent Application Laid-Open No. 56-157033. This is performed by using a zone plate as a pattern, irradiating the zone plate with a light beam, and detecting the position of a light-converging point on a predetermined surface of the light beam emitted from the zone plate.

Generally, an alignment method using a zone plate has a feature that relatively high-precision alignment can be performed without being affected by a defect in an alignment pattern, as compared with a method using a simple alignment pattern.

FIG. 7 is a schematic view of a conventional positioning device using a zone plate.

In the same figure, a parallel light beam emitted from a light source 72 passes through a half mirror 74 and is condensed at a converging point 78 by a converging lens 76, and is then placed on a mask alignment pattern 68a on a mask 68 and a support 62. The wafer alignment pattern 60a on the surface of the wafer 60 is irradiated. These alignment patterns 68a and 60a are constituted by reflection type zone plates,
Condensing points (light spots) are formed on a plane orthogonal to the optical axis including the converging points 78. At this time, the amount of shift of the condensing point position on the plane is detected by guiding the light onto the detection surface 82 by the condensing lenses 76 and 80.

Then, based on the output signal from the detector 82, the control circuit 84
Drives the drive circuit 64 to perform relative positioning between the mask 68 and the wafer 60.

FIG. 8 shows the mask alignment pattern shown in FIG.
FIG. 8 is an explanatory diagram showing an image forming relationship of a light beam from 68a and a wafer alignment pattern 60a.

In the figure, a part of the luminous flux diverging from the converging point 78 is diffracted from the mask alignment pattern 68a to form a converging point 78a indicating the mask position near the converging point 78.
Further, some other light beams pass through the mask 68 as zero-order transmission light and enter the wafer alignment pattern 60a on the wafer 60 without changing the wavefront. At this time, after the light beam is diffracted by the wafer alignment pattern 60a, the light beam is transmitted again through the mask 68 as zero-order transmission light, condensed in the vicinity of the converging point 78, and forms a converging point 78b representing the wafer position. In the figure, when the light beam diffracted by the wafer 60 forms a converging point, the mask 68 acts as a simple transparent state.

The position of the focal point 78b by the wafer alignment pattern 60a formed in this way is determined by the amount of the deviation along a plane orthogonal to the optical axis including the focal point 78 according to the deviation Δσ of the wafer 60 with respect to the mask 68. The deviation amount Δ of the amount corresponding to Δσ
σ '. Δσ was determined by measuring this Δσ ′ with reference to an absolute coordinate system provided on the sensor.

FIG. 9 is a schematic diagram of a distance measuring device proposed in Japanese Patent Laid-Open No. 61-111402. In the figure, a mask M as a first object and a wafer W as a second object are arranged so as to face each other, and a laser beam is focused on a point Ps between the mask M and the wafer W by a lens L1.

At this time, the light flux is reflected on the surface of the mask M and on the surface of the wafer W, and is focused and projected onto the points Pw and P _M on the surface of the screen SC via the lens L2. The distance between the mask M and the wafer W is measured by detecting the distance between the light converging points Pw and P _M on the screen S surface.

[Problems to be solved by the invention]

In the above-mentioned conventional example, a laser beam having good directivity is used as the light flux. However, since the laser light has a large coherence at the same time, the light intensity distribution or the cross-sectional shape of the light beam incident on the sensor due to multiple interference between the mask and wafer, interference with the speckle scattered light from the mask and wafer surface, etc. It is easy to change from position to position. If the sensor incident position is constant, the noise as described above is often fixed to the spatial device, so it does not change over time, so even if the light intensity distribution is averaged over time, this noise should be canceled. I can't. Therefore, in order to detect the movement of the focal point of the light flux on the sensor, a reference point based on the cross-sectional shape (for example, the center of the circular light flux)
If you try to detect the movement of a reference point (for example, the point where the light intensity has a peak) based on the light intensity or the light intensity, the cross-sectional shape and the light intensity will change. There is a risk that the displacement does not match the movement amount of the light spot, and an error may occur in the position shift and the interval detection.

In view of the above-mentioned drawbacks of the conventional example, the present application provides a positional relationship detection device that enables highly accurate detection when detecting a positional shift or a gap between two objects by moving a light beam having a large coherence such as a laser beam. The purpose is to do.

[Means for solving the problem]

According to the present invention, the light intensity distribution of the cross section of the light flux on the sensor is arithmetically processed to obtain a predetermined position that is unlikely to cause a position change due to the influence of the noise and the like, and by detecting the movement of this position, the first object is detected. By obtaining the relative positional relationship with the second object, accurate relative position detection with less error is realized.

In one embodiment described later in the present application, the light intensity distribution in the converging point on the sensor is converted into an electric signal, and the light intensity signal at each point in the converging point is amplified by a different constant multiple depending on the signal (for example, Squared amplification), the positions of the respective points within the condensing point are averaged for the amplified signal, and the movement of the light flux is detected by obtaining the movement of this average position with the obtained average position as a reference. . Even if noise of the same magnitude is generated by performing square amplification, the signal in the low light intensity part, which has a large influence of noise (low S / N ratio), is reduced, and the influence of noise is small (S / N ratio Since it is possible to highly amplify the signal of the (high) high-intensity part, the position obtained by averaging the positions of each point for this signal is unlikely to cause position fluctuation within the condensing point due to noise, and this position is Accurately follow the movement of.

〔Example〕

 The draft of the present invention will be described.

As a comparative example, an example will be given in which the position of each point is averaged with respect to the light intensity of each point (hereinafter referred to as the center of gravity of the light flux).

The barycentric position (vector) of the light flux is expressed by the following equation.

Where i: Natural number Si: Light intensity at i-th position ▲ ▼: Position vector at i-th position This method is used for the above-mentioned focal point detection type position shift detection and interval detection where the light intensity signal is locally concentrated. Consider the case of applying. In this case, a large light intensity signal is concentrated around the center of gravity, and the light intensity signal becomes smaller as the distance from the center of gravity increases. On the other hand, assuming that noise is generated at all locations on the sensor with the same probability and with the same magnitude, the original signal is large even if this noise occurs in the narrow range around the center of gravity position. Since the / N ratio is sufficiently large and the contribution amount of noise on the right side of the equation (1) is also sufficiently small, the amount of fluctuation of the barycentric position of the light flux due to the noise generated here is small. However, noise is likely to occur over the entire surface of the sensor, so the noise rate is higher in the wide low-signal area far from the center of gravity than in the narrow area around the center of gravity. The generated noise has a low S / N ratio because the original signal is small, and therefore gives a large change to the right-hand side of equation (1), and the amount of fluctuation of the barycentric position of the light flux due to this noise is large. If this barycentric position is used as the reference position of the light condensing point and the position shift and the distance are detected from the displacement amount of this reference position, noise is likely to occur and a detection error is likely to occur.

Therefore, the present application amplifies the locally concentrated light intensity signal more than the signal in a wide low signal region, and averages each position of this signal to obtain the reference position. This makes it difficult for the reference position to change due to noise in the low level region.

The reference position (vector) 'at this time can be expressed as follows.

Here, Ai, αi: The specific coefficient Ai, αi at the i-th position is set so as to amplify the signal Si in the signal concentrating portion to a greater extent and not to amplify or reduce the signal in the low signal region so much. As such setting, Ai = 1
There is αi = 2, ie the square amplification of the signal Si. A specific example of this setting will be described below.

FIG. 1 shows an embodiment in which the present invention is applied to a positioning device.
S is a laser light source that is a light source that emits light with good directivity, and L is a lens that collimates the laser light. The collimated laser beam illuminates the physical optical element A1 which is an alignment mark on the object Q1 (eg mask) to be aligned. The physical optical element A1 is, for example, a Fresnel zone plate that functions as a lens that focuses the emitted light beam. A1 on alignment object Q2 (eg wafer)
The physical optical element A2, which is also an alignment mark, has a lens action of condensing the light beam incident to the condensing point P by the physical optical element A1 to a point on the image sensor IS, for example, a Fresnel zone plate. Is. The physical optics A1 and A2 focus the light flux at the center point P ₀ of the image sensor IS when the objects Q1 and Q2 are not displaced, but when the object Q2 is displaced vertically in the drawing. A light beam is condensed at a point P ′ on the sensor IS, which is displaced from the point P ₀ by an amount obtained by multiplying the displacement amount by β in the displacement direction of the objects Q1 and Q2. Therefore, by detecting the position shift amount and direction of the condensing point P'by the image sensor IP, the position shift between the objects Q1 and Q2 can be detected. An example of the physical optical elements A1 and A2 is a Fresnel zone plate as shown in FIG. The black portion in the figure is the light shielding grating portion. Such a Fresnel zone plate has the following shape.

First, the mark A1 for the mask Q1 is designed so that a parallel light beam of a predetermined beam system enters at a predetermined angle and is condensed at a predetermined position. Generally, the pattern of the Fresnel zone plate is an interference fringe pattern on the lens surface when the coherent light sources are placed at the light source (object point) and the image point, respectively.

Now, the coordinate system on the first object Q1 plane is determined. Here, the origin is located in the center of the Fresnel zone plate A1, the x axis is in the direction of displacement detection (vertical direction in the drawing), the y axis is orthogonal to the x axis on the object Q1 surface, and the z axis is the normal to the object Q1 surface. To take. A parallel light beam that is incident at an angle α with respect to the normal to the object Q1 and has its projection component orthogonal to the x-axis direction is transmitted and diffracted through the mark on the object Q1, and then the converging point P (x ₁ , y ₁ , z ₁ ) The equation of the curve group of the grating lens which forms an image at the position of is, when the Fresnel zone plate has the same focusing power in the x and y directions, represents the contour position of the grating by x and y. Given by Where λ is the wavelength of the light beam and m is an integer.

The chief ray is incident at an angle α, passes through the origin on the surface of the object Q1,
Assuming that the light ray reaches the converging point P (x ₁ , y ₁ , z ₁ ), the right side of the equation (3) indicates that the optical path length m / 2 times the wavelength is longer (shorter) than the principal ray depending on the value of m. The left side represents the difference in the optical path lengths of the rays that reach the point (x ₁ , y ₁ , z ₁ ) through the point (x, y, 0) on the object Q1 with respect to the optical path of the principal ray.

On the other hand, the Fresnel zone plate A2 on the object Q2 places the spherical wave emitted from a predetermined point light source at a predetermined position (on the sensor IS surface).
It is designed to focus light on. Point light source is object Q1 and object
Letting the gear of Q2 be g, it is represented by (x ₁ , y ₁ , z ₁ −g) and is the position of the image formation point by the Fresnel zone Q1 (y is a variable). It is assumed that the alignment of the object Q1 and the object Q2 is performed in the X-axis direction, and the point P on the sensor IS surface when the alignment is completed.
If the light beam is focused at the position of ₀ (x ₂ , y ₂ , z ₂ ),
The equation of the curves of the contour of the grating of the Fresnel zone plate A2 on the object Q2 is in the coordinate system defined earlier. It is expressed as

In the equation (4), the object Q2 is at z = -g, and the chief ray is the object Q1.
On the surface of the object Q2 as a ray passing through the origin on the surface and the point (0,0, -g) on the surface of the object Q2, and the point P ₀ (x ₂ , y ₂ , z ₂ ) on the surface of the sensor IS It is an equation that satisfies the condition that the difference in optical path length between the ray passing through the gray tint (x, y, −g) and the principal ray is an integral multiple of half wavelength.

Returning to FIG. 1, the light flux condensed on the image sensor IS is photoelectrically converted, converted into a digital signal by the A / D converter AD, and the light flux condensed on the sensor IS is converted by the microprocessor MP. The aforementioned reference position is detected. That is, the light intensity detection signal of each pixel on the image sensor IS is squared, and each pixel position is averaged for the obtained signal, that is, for the square of the light intensity to obtain an average position.

The control system CT commands the operation amount of the actuator AC through the driver DR to move the object Q2 to be aligned by a predetermined amount based on the information from the microprocessor MP.

 Next, the operation will be described.

It is assumed that the optical axis of the physical optical element A2 on the object Q2 to be aligned is decentered by δ while being parallel to the optical axis 0 formed by the physical optical element A1 (this coincides with the optical axis of the lens L). The position of the focal point P ′ on the image sensor IS is displaced from P ₀ by β · δ. The light intensity detection signal of the spot on the image sensor is from the spot-shaped profile of the original condensing point P as well as the high-order diffraction image of the Fraunhofer by the shape of the openings of the physical optical elements A1 and A2 from the objects 1 and 2. Scattered light Spectral light is formed by combining electrical noise.

Regarding the time-varying component of the electrical noise of,
This can be removed by averaging the data obtained within a certain time and using this as detection data. It is possible to remove temporally fixed components of electrical noise by obtaining them in advance and correcting them by providing an offset to the output signal.

Further, on the sensor IS, for example, the higher-order component of the Founfer-Four-Diffraction image of the aperture of the physical optical element having the shape as shown in FIG. 3 is generated centering on the 0th-order light of the spot. The effect of this will be explained. Due to the finite number of image sensor IS, it is out of the sensor IS from the higher order with the Fraunhofer diffraction pattern. However, when the alignment objects Q1 and Q2 are eccentric by δ, the spots on the sensor IS also deviate from the center, so the high-order component that was on this edge moved to the sensor IS, or it was on the sensor IS before. When the high-order components are projected from the sensor IS and the respective positions of the light intensity of the entire sensor IS are averaged, an error occurs at the reference position of the detected light flux.

A process for removing an error, especially an error due to the influence of this Fraunhofer diffraction image is shown in a flow chart of FIG. However, the process of FIG. 4 is a process in the case where the image sensor IS is a linear image sensor in which pixels are arranged only in the positional deviation detection direction (x direction). In the figure, N: metal prime number, i: number from one end of pixel, l: sensor pixel interval, Si: average light intensity at i-th pixel, g ₁ : scattered light noise component at i-th pixel. From the signal from each pixel, obtain the average value Si of the light intensity within a fixed time, and use this Si However, when Ti = Si-gi Si-gi ≧ 0, Ti = 0 Si-gi <0 is calculated and the reference position M is obtained. Here, gi is the object Q1 in FIG.
The sensor output data due to the scattered light component when only Q2 is stored in the memory and calculated by the sum of the data. This becomes scattered light data. In this case, even if gi is a constant value (average value of all scattered light noise components), the S / N ratio is increased as a whole, so that there is an effect of removing noise components. Further, it is considered that when gi is all 0, only the Fraunhofer diffraction image is generated on the sensor.

FIG. 5 shows a graph of simulation results when it is assumed that the spot shape is a function of sinC ² of 200 μ and the reference position is detected by a linear image sensor of pitch 30 μm × 40 pieces. This graph shows the normal center of gravity detection when the reference position is obtained by performing square calculation processing with this device (dotted line in the figure) when the positional deviation amount β / δ of the spot on the line sensor is taken as the horizontal axis. It shows the error from the reference position movement amount β / δ in the case of performing (solid line in the figure).
As shown in FIG. 5, for example, the spot displacement detection error is 0.
It can be seen that the displacement amount of the spot is 1 μ or less, which is 20 μ in the center-of-gravity detection method, while it is 440 μ, which is about 20 times that in the detection method using the square calculation. From this, it is proved that the positional relationship detecting apparatus according to the present embodiment enables highly accurate positional relationship detection over a wide range.

When the image sensor IS is a three-dimensional image sensor, the square calculation process may be performed with Ai = 1, αi = 2 in the equation (2).

FIG. 6 shows a second embodiment of the present invention. The parts different from the embodiment in FIG. 1, that is, the image sensor IS, A / D
This is another embodiment in which the parts of the converter AD and the microprocessor MP are changed, and the other parts are the same as in FIG. 1 and therefore omitted in FIG.

In the figure, GG is a gi wave generator, ADD is an adder, GA is a gamma amplifier, IA ₁ and IA ₂ are integrators, MPX is a multiplier, LG is a sawtooth wave generator, and DEV is a divider.

The second embodiment is largely different from the first embodiment in that the digital signal is processed, whereas the analog signal is processed.

In Fig. 6, the luminous flux condensed on the image sensor IS is photoelectrically converted into Si, which is the sensor output (analog signal).
Put out. The output Si of the image sensor IS does not undergo A / D conversion,
As it is, the analog signal gi of the scattered light noise component corresponding to the pixel number i of the image sensor output by the gi wave generator GG is subtracted by the adder ADD, and Ti (= Si-g
i) becomes an analog signal. The signal Ti is further squared by the gamma amplifier GA, for example, and multiplied by the output signal of the sawtooth wave generator LG corresponding to the pixel number of the image sensor. Further, it is integrated by an integrator IA ₁ that integrates in synchronization with the scanning of the image sensor IS (= I _1i ). At the same time, the output of the gamma amplifier is integrated by the integrator IA ₂ (I _2i ). The output of each integrator is calculated as I _1i / I _2i by the divider DEV, and the output of the divider DEV showing the average position obtained by averaging the positions of the respective points with respect to the square of the light intensity is controlled. It is sent to the circuit CT.

Regarding operation, there is a difference between digital and analog, but first
The description is omitted because it is similar to the embodiment.

〔The invention's effect〕

As described above, by obtaining a predetermined position in the light flux cross section that is unlikely to cause position variation due to the influence of noise, etc., and by detecting the movement of this position, the relative positional relationship between the first object and the second object is obtained. , It became possible to detect the positional relationship with less error.

[Brief description of the drawings]

FIG. 1 is a block diagram of a positioning device embodying the present invention, and FIG.
The figure is an example of the physical optical element used in FIG. 1, FIG. 3 is an example of the spot shape on the sensor, FIG. 4 is the processing flow of the apparatus of FIG. 1, and FIG. 5 is the apparatus of FIG. Fig. 6 is an error comparison graph of the reference position detection method and the center of gravity detection method, Fig. 6 is a partial configuration diagram of another embodiment of the present invention, and Figs. 7, 8 and 9 are explanatory views of a conventional example. In the figure, S ... Light source A1, A2 ... Physical optical element IS ... Image sensor MP ... Microprocessor

Claims (1)

(57) [Claims] 1. A light source means for irradiating a light flux to a first and a second object whose relative positional relationship is to be detected, and at least one of a first and a second object illuminated with the light flux by the light source means. The light detecting means for receiving the light flux on the predetermined surface and detecting the light intensity at each point on the predetermined surface of the light flux, and the position of each point on the predetermined surface of the light flux for the light detecting means. An average position detecting means for obtaining an averaged position for values obtained by multiplying the value of the light intensity of each point detected by a constant different according to the value of the light intensity, and the average position obtained by the average position detecting means. Means for detecting a relative positional relationship between the first and second objects based on the positional relationship detecting means.