An automatic alignment method for laser interferometry endpoint detection

Abstract

The present invention discloses an automatic alignment method for laser interference endpoint detection, comprising the following steps: illuminating the surface of a wafer, irradiating a laser beam to the surface of the wafer; receiving an output image S1 formed by the illuminated area of ​​the wafer surface, wherein the output image S1 includes a target image Sg formed by a target pattern on the wafer and a laser spot image L reflected by the wafer; calculating the geometric center coordinates of the target image Sg in the output image S1 and the centroid coordinates of the laser spot image L to obtain the pixel distance corresponding to the two coordinates; according to the pixel distance, controlling the laser beam to move toward the target pattern on the surface of the wafer, so that the centroid of the laser spot image coincides with the geometric center of the target image Sg on the output image S1, thereby achieving automatic alignment. The present invention can achieve automatic alignment of the laser beam and the target pattern in the process of laser interference endpoint detection, thereby improving the efficiency of etching endpoint detection.

Description

Technical Field

[0001] The invention relates to the technical field of semiconductor detection, and in particular to an automatic alignment method for laser interference endpoint detection. Background Art

[0002] In recent years, as the integration of circuits increases, the critical dimensions of graphics need to be reduced exponentially. The reduction in critical dimensions further increases the difficulty of aligning the beam of the interferometric endpoint detection device. In semiconductor etching, the laser emitted by the laser interferometric endpoint detection device needs to be aligned with the target pattern of the wafer sample to be etched in order to achieve the purpose of real-time and accurate etching endpoint detection. In this process, the corresponding algorithm is required to calculate the coordinates of the laser spot image and the target image coordinates received by the camera on the system optical path to achieve automatic alignment of the laser beam with the target pattern of the wafer sample to be etched. The automatic alignment method of laser interferometric endpoint detection saves a lot of time and labor costs for the semiconductor chip manufacturing process and improves the efficiency of the alignment operation.

[0003] In the prior art, the electric motion platform configured for the laser interferometric endpoint detection device has problems such as being too large in size, insufficient bidirectional positioning accuracy and repeated positioning accuracy, limitations in the image recognition algorithm caused by the low lens imaging resolution, and being unable to accurately calculate the target coordinates of small-size images or images with blurred contours.

[0004] Therefore, it is necessary to provide a new automatic alignment method for laser interferometric endpoint detection to solve the above problems existing in the prior art. Summary of the invention

[0005] The purpose of the present invention is to provide an automatic alignment method for laser interference endpoint detection, so as to enable automatic alignment of the laser beam and the target pattern during the laser interference endpoint detection process, thereby improving the etching endpoint detection efficiency.

[0006] To achieve the above object, the technical solution of the present invention is as follows:

[0007] An automatic alignment method for laser interferometric endpoint detection comprises the following steps:

[0008] The wafer surface is illuminated and the laser beam is directed onto the wafer surface;

[0009] Receiving an output image S1 formed by the illuminated area on the wafer surface, wherein the output image S1 includes a target image Sg formed by a target pattern on the wafer and a laser spot image L reflected by the wafer;

[0010] Calculate the geometric center coordinates of the target image Sg in the output image S1 and the centroid coordinates of the laser spot image L to obtain the pixel distance corresponding to the two coordinates;

[0011] According to the pixel distance, the laser beam is controlled to move toward the target pattern on the wafer surface so that the centroid of the laser spot image L coincides with the geometric center of the target image Sg, thereby achieving automatic alignment.

[0012] By adopting the above technical solution, since the output image S1 includes the target image Sg formed by the target pattern on the wafer and the laser spot image L reflected by the wafer, and there is a certain position deviation between the target image Sg and the laser spot image L reflected by the wafer on the output image S1, in order to make the detection result more accurate, it is necessary to align the target image Sg and the laser spot image L reflected by the wafer. In this process, the pixel distance corresponding to the coordinates of the two can be obtained through the pixel of the geometric center of the target image Sg and the pixel of the centroid of the laser spot image L reflected by the wafer. The pixel distance between the two can be used to control the movement of the laser beam to align the target image Sg and the laser spot image L reflected by the wafer.

[0013] Optionally, calculating the geometric center coordinates of the target image Sg includes the following steps:

[0014] Intercepting a target image Sg from the output image S1 to obtain a template image g1;

[0015] A local area H is intercepted from the output image S1, and the position of the target image Sg in the output image S1 is calculated according to the correlation coefficient between the local area H and the template image g1.

[0016] By adopting the above technical solution, before detection, it is necessary to set a template image g1 with the same shape as the target image Sg. During the detection process, a local area H with the same size as the target image Sg is scanned and intercepted in the camera's output image S1, and the local area H is scanned line by line on the output image S1, so that the scanning result position can best fit the position of the target image Sg.

[0017] Optionally, scanning the target image Sg includes the following steps:

[0018] When the target image Sg and the template image g1 have the same angle;

[0019] Setting the size of the local area H to m*n, and sequentially scanning the local area H in the output image S1;

[0020] The correlation coefficient between each local area H and the template image g1 is calculated according to the normalized correlation coefficient formula to obtain a correlation coefficient matrix ;

[0021] According to the correlation coefficient matrix It is determined whether the local area H in the output image S1 is the same as the template image g1. If they are the same, the position of the local area H is the position of the target image Sg.

[0022] By adopting the above technical solution, when the target image Sg and the template image g1 have the same angle, the m*n area with the same size as the template image g1 is directly scanned in the output image S1, and the correlation coefficient matrix is ​​calculated , where according to the correlation coefficient matrix The value of can determine whether the local area in the output image S1 is the same as the template image g1. When the local area in the output image S1 is the same as the template image g1, the local area in the output image S1 is the position of the target image Sg.

[0023] Optionally, scanning the target image Sg includes the following steps:

[0024] When the target image Sg and the template image g1 have different angles;

[0025] Set the initial template image to g1;

[0026] The initial template image g1 is rotated by using the forward mapping method to obtain the rotated template image g2, and the template image g3 is obtained by using the reverse mapping method and the bilinear interpolation algorithm;

[0027] Eliminate the portion of the edge of the template image g3 where the grayscale value changes due to the rotation and obtain the template image g4, so that the grayscale value in the rotated template image g4 is consistent with the grayscale value in the template image g3 before the rotation;

[0028] The template images g4 at multiple different angles form a template library;

[0029] The local area H intercepted in the output image S1 is used to calculate the position of the target image Sg in the output image S1 according to the correlation coefficient between the local area H and a plurality of template images g4 in the template library.

[0030] By adopting the above technical solution, multiple rotated template images g4 are set, and the angles of the multiple rotated template images g4 are different, so as to form a template library. In order to make the detection result more accurate, after scanning, it is necessary to calculate the correlation coefficient between the scanned image and the multiple template images g4 in the template library. In the process of setting up the template library, it is necessary to rotate the template image, and the initial template image g1 is rotated by the forward mapping method to obtain the rotated template image g2. When the position after rotation contains a decimal point, it is directly rounded to cause a hole in the template image g2, and the hole is filled by the reverse mapping method and the bilinear interpolation algorithm to obtain the template image g3. In addition, the edge part of the rotated template image g3 where the gray value changes is removed. Since the detection process is scanned according to the gray value, the scanning result will be more accurate after removing the edge part of the rotated pattern where the gray value changes.

[0031] Optionally, the step of scanning the target image Sg according to the template image g4 includes:

[0032] Setting the size of the local area H to m*n, and sequentially scanning the local area H in the output image S1;

[0033] The correlation coefficient between each of the output images S1 and each of the template images g4 in the template library is calculated according to the normalized correlation coefficient formula to obtain a correlation coefficient matrix ;

[0034] According to the correlation coefficient matrix It is determined whether the local area in the output image S1 is the same as the template image g4. If they are the same, the position of the local area H is the position of the target image Sg.

[0035] By adopting the above technical solution, when the target image Sg and the template image g1 have different angles, the template image g1 is rotated to form a template library with different rotation angles, and the m*n area with the same size as the template image g4 is scanned in the output image S1, and the correlation coefficient matrix is ​​calculated. , where according to the correlation coefficient matrix The value of can determine whether the local area in the output image S1 is the same as the template image g4. When the local area in the output image S1 is the same as the template image g4, the local area in the output image S1 is the position of the target image Sg.

[0036] Optionally, the coordinates of the upper left corner of the target image Sg in the output image S1 are (Xsg, Ysg);

[0037] The geometric center coordinates of the target image Sg are (Xsg+m / 2, Ysg+n / 2).

[0038] By adopting the above technical solution, the coordinates of the upper left corner of the target image Sg in the output image S1 can be obtained according to calculation. At the same time, since the target image Sg is a rectangle and the length and width of the target image Sg are known, the geometric center coordinates of the target image Sg can be calculated.

[0039] Optionally, calculating the centroid coordinates of the laser spot image L comprises the following steps:

[0040] Scanning pixels of the output image S1;

[0041] Perform mean filtering on the output image S1 and select multiple pixels with the largest grayscale values;

[0042] The pixel grayscale values ​​in the neighborhood of the pixel point are counted to calculate the maximum mean value. The central area of ​​the laser spot image L is located near the pixel point corresponding to the maximum mean value.

[0043] By adopting the above technical solution, after the camera receives the laser beam, in the output image S1, the gray value of the laser beam is relatively large, so the laser spot image L can be located by the gray value. After the positioning is completed, the centroid position of the laser spot image L is calculated according to the laser spot image L; at the same time, due to its high brightness and concentration, the laser spot produces a higher gray value on the camera photosensitive element; this means that the pixel gray value of the laser spot area will be relatively high, while the gray value of the background or noise will be relatively low; when the gray histogram of the image is counted, the pixels in the area where the laser spot is located will be at a high gray value. The gray value interval forms a peak value. By finding this peak value, the approximate position of the laser spot can be determined. By screening out multiple pixels with the largest gray value, the brightest area in the laser spot is actually selected. These pixels are located very close to or at the center of the laser spot. By counting the pixels in the neighborhood of each base point pixel and calculating the gray mean of these neighborhood pixels, the most stable and concentrated area in the laser spot is actually found. The base point corresponding to the maximum value of the mean value means that the gray value near this point changes the least, which is closest to the actual shape and position of the laser spot.

[0044] Optionally, calculating the centroid coordinates of the laser spot image L includes the following steps:

[0045] Calculate the gray value of the pixel in the neighborhood corresponding to the maximum mean value to decay to its gray value e -2 The pixel coordinates of

[0046] will decay to its gray value e -2 The pixel coordinates are marked as (m i , n j ), i, j are natural numbers;

[0047] The corresponding gray value is recorded as e i, j , i, j are natural numbers;

[0048] According to the centroid calculation method, the horizontal axis coordinate m of the centroid of the laser spot image L is obtained. α and the vertical axis coordinate n β .

[0049] By adopting the above technical solution, the photosensitive element of the camera is composed of a pixel array. The laser spot may span multiple pixels during imaging, and each pixel can only record one grayscale value; therefore, the true center of the laser spot may be located between pixels, rather than exactly at the center of a certain pixel; in addition, the laser spot may not be a perfect circle, and its shape and size may vary due to the influence of the optical system; this results in the center of the spot not being a simple pixel position with the maximum grayscale value; in addition, noise and other stray light in the image may interfere with the precise positioning of the center of the laser spot; even if grayscale screening is performed, there may be slight errors. In order to reduce these errors, a centroid algorithm is used to more accurately calculate the center of the laser spot; the centroid algorithm determines the center of the spot by calculating the weighted average of all pixel positions and grayscale values ​​that meet specific grayscale threshold conditions, thereby more accurately reflecting the true position of the laser spot. By calculating the centroid, the errors caused by pixel discreteness, spot shape changes and noise can be effectively reduced, and the accurate detection of the center of the laser spot can be achieved.

[0050] Optionally, the pixel distance between the centroid of the laser spot image L and the geometric center of the target image Sg in the X-axis direction is obtained by subtracting the centroid of the laser spot image L from the centroid of the target image Sg. -(Xsg+m / 2));

[0051] Subtract the vertical axis coordinate of the geometric center of the target image Sg from the vertical axis coordinate of the laser spot image L to obtain the pixel distance between the centroid of the laser spot image L and the geometric center of the target image Sg in the Y-axis direction: ( -(Ysg+n / 2));

[0052] The pixel distance is divided by the system optical path magnification to obtain the moving distance of the laser beam, and the moving direction of the laser beam is set according to the positive and negative values ​​of the calculation result to control the center of mass of the laser beam to move to the geometric center of the target pattern.

[0053] By adopting the above technical solution, when the centroid coordinates of the laser spot image L and the geometric center coordinates of the target image Sg are known, the pixel distance between the two can be calculated, so that the moving direction of the laser beam can be set according to the positive and negative values ​​of the calculation results, and the centroid of the laser beam can be controlled to move to the geometric center of the target pattern.

[0054] Optionally, when the correlation coefficient matrix When it is close to 1, the local area H in the output image S1 is the position of the target image Sg.

[0055] By adopting the above technical solution, according to the normalization coefficient formula, when the correlation coefficient matrix The absolute value of is close to 1, indicating that there is a trend of complete positive correlation or complete negative correlation between the two images. Therefore, the gray value change trend of the corresponding sub-block in the template image and the required image at each pixel position is consistent; that is, the two image blocks are very similar in shape, grayscale distribution and structure, and can be identified as the same image.

[0056] The beneficial effects of the automatic alignment method for laser interferometric endpoint detection provided by the present invention are:

[0057] 1. By calculating the pixel distance between the centroid coordinates of the laser spot image L and the geometric center coordinates of the target image Sg, the laser beam is controlled to move toward the target pattern on the wafer surface, so that the centroid of the laser spot image L coincides with the geometric center of the target image Sg on the output image S1, thereby achieving automatic alignment.

[0058] 2. This method uses the centroid algorithm to calculate the centroid of the laser spot, which can effectively reduce the errors caused by pixel array, spot shape changes and noise, thereby realizing the accurate detection of the centroid of the laser spot and improving the anti-interference ability and automatic detection accuracy of the system;

[0059] 3. Through the automatic alignment method, the laser beam can be quickly aligned to the target area, reducing the time and error of manual operation, improving the detection efficiency, and saving a lot of time and labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of the steps of an automatic alignment method for laser interferometric endpoint detection according to an embodiment of the present invention;

[0061] Figure 2 Schematic diagram of the positions of the output image S1, the laser spot image L, and the target image Sg on the output image S1 according to an embodiment of the present invention;

[0062] Figure 3 is a schematic diagram of the grayscale of the template image g1 and the angle before rotation according to an embodiment of the present invention;

[0063] Figure 4 Schematic diagram of the output image S1 and the scanning process according to an embodiment of the present invention;

[0064] Figure 5Schematic diagram of the positions of the output image S1 and the laser spot image L after alignment with the target image Sg according to an embodiment of the present invention;

[0065] Figure 6 is a schematic diagram of the grayscale of the template image g1 and a certain angle after rotation according to an embodiment of the present invention;

[0066] Figure 7 is a schematic diagram of the grayscale and a certain rotation angle of the template image g2 according to an embodiment of the present invention;

[0067] Figure 8 is a schematic diagram of the grayscale and a certain rotation angle of the template image g3 according to an embodiment of the present invention;

[0068] Fig. 9 is a schematic diagram of the grayscale and a certain rotation angle of the template image g4 according to an embodiment of the present invention;

[0069] Fig.10 Schematic diagram of the positions of the output image S1 and the target image Sg corresponding to the laser spot image L and the template image g4 after rotation on the output image S1 according to an embodiment of the present invention;

[0070] Fig.11 Schematic diagram of the scanning process of the output image S1 and the target image Sg corresponding to the rotated template image g4 according to the embodiment of the present invention;

[0071] Fig.12 It is a schematic diagram of the positions of the output image S1 and the laser spot image L after the template image g4 is rotated and aligned with the corresponding target image Sg according to the embodiment of the present invention;

[0072] Fig.13 Schematic diagram of the actual position of the laser spot image L and the target image Sg before alignment in the output image S1 according to the embodiment of the present invention;

[0073] Fig.14 Schematic diagram of the actual position of the laser spot image L and the target image Sg after alignment in the output image S1 before the process according to the embodiment of the present invention;

[0074] Fig.15 is a grayscale histogram of an embodiment of the present invention, wherein the abscissa is the grayscale value and the ordinate is the number of pixels of the grayscale value;

[0075] Fig.16 is a schematic diagram of an application scenario of an embodiment of the present invention;

[0076] Fig.17 Schematic diagram of the etching groove and depth during the small aperture ratio etching process according to an embodiment of the present invention. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Unless otherwise defined, the technical terms or scientific terms used herein should be understood by people with general skills in the field to which the present invention belongs. "Including" and similar words used in this article mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

[0078] The specific implementation modes of the present invention are further described in detail below in conjunction with the accompanying drawings.

[0079] refer to Figure 1-17 An embodiment of the present invention provides an automatic alignment method for laser interferometric endpoint detection, comprising the following steps:

[0080] S100, illuminating the wafer surface, and irradiating the laser beam onto the wafer surface.

[0081] Specifically, in S100, the surface of the wafer is illuminated by an LED lamp, and a laser beam is irradiated onto the surface of the wafer by a laser emitter.

[0082] S200, receiving an output image S1 formed by the illuminated area on the wafer surface, wherein the output image S1 includes a target image Sg formed by a target pattern on the wafer and a laser spot image L reflected by the wafer.

[0083] Specifically, in S200, the camera is set at a position relative to the wafer so that the illuminated area on the wafer surface can be reflected to the photosensitive surface of the camera. The camera outputs the received image to the host computer to form an output image S1, wherein the output image S1 includes a target image Sg formed by a target pattern on the wafer and a laser spot image L reflected by the wafer.

[0084] S300, calculating the geometric center coordinates of the target image Sg and the centroid coordinates of the laser spot image L in the output image S1, and obtaining the pixel distance corresponding to the two coordinates.

[0085] Specifically, in S300, the geometric center coordinates of the target image Sg in the camera's output image S1 are calculated, the centroid coordinates of the laser spot image L in the camera's output image S1 are calculated, and the corresponding pixel distance between the geometric center coordinates of the target image Sg and the centroid coordinates of the laser spot image L is obtained.

[0086] S400, according to the pixel distance, the laser beam is controlled to move toward the target pattern on the wafer surface, so that the centroid of the laser spot image L coincides with the geometric center of the target image Sg, thereby achieving automatic alignment.

[0087] Specifically, in S400, the laser beam is controlled to move toward the target pattern on the wafer surface according to the pixel distance between the two coordinates, so that the centroid of the laser spot image L coincides with the geometric center of the target image Sg, thereby achieving automatic alignment.

[0088] In some embodiments, the laser beam is emitted by a laser, and the laser, LED light, motion platform, and host computer lighting equipment together constitute a detection system. The movement of each component or device is controlled by operating the host computer to send instructions. The equipment of the detection system is existing technology, and its connection method will not be described here.

[0089] Among them, the pictures received by the camera are as follows Figure 2 As shown, the large rectangular area is the range of the camera's photosensitive surface F, that is, Figure 2 As shown in the photosensitive surface F in the figure, L represents the laser spot image L obtained by the system imaging on the camera after being reflected by the surface of the wafer to be etched, and the target area of ​​the sample to be etched is the target image Sg. It is necessary to quickly move the laser spot image L to the target image Sg area, so it is necessary to calculate the geometric center coordinates of the laser spot image L and the target image Sg.

[0090] In this embodiment, the camera parameters are: camera resolution 3088x2064, pixel size 2.4umx2.4um, sensitivity 425mV 1 / 30s (capable of detecting and responding to signal changes within 1 / 30 second, and this change triggers a voltage change of 425 millivolts), maximum gain (multiple) 32, exposure time 0.025~3612ms.

[0091] The following steps are included when calculating the geometric center coordinates of the target image Sg:

[0092] First, the template image g1 is set in the host computer so that the template image g1 has the same shape as the target image Sg; in some embodiments, the host computer is a control system, which can be a computer host for sending or receiving instructions, or other devices that can send or receive instructions.

[0093] Secondly, a local area H is cut out from the output image S1, and the position of the target image Sg in the output image S1 is calculated according to the correlation coefficient between the local area H and the template image g1.

[0094] In this process, the angle between the set target image Sg and the template image g1 may be the same or different. Different methods are used for calculation when the angle between the two is the same or different. For details, refer to the following contents.

[0095] Specifically, when the angles between the target image Sg and the template image g1 are the same, the size of the local area H is set to m*n, where m is the long side of the local area H, and n is the short side of the local area H; specifically, the size of the target image Sg is m*n, and the size of the template image g1 is m*n, that is, the target image Sg, the template image g1 and the local area H are the same size.

[0096] In the output image S1 of the camera, the local areas H are scanned sequentially.

[0097] In some embodiments, reference Figure 2 , Figure 3 , Figure 4 and Figure 5 , the size of the output image S1 is M*N, and the local area H is scanned starting from the upper left corner of the output image S1, that is, the coordinates of the upper left corner of the local area H coincide with the coordinates of the upper left corner of the output image S1 as the initial point, and the scanning begins. During the scanning process, it is first scanned in the horizontal direction, that is, (1,1) to (M,1), and then it turns to (1,2) and scans to (M,2) after the scanning is completed, and finally it turns to (1,N) and scans to (M,N); specifically, in this process, each scanning interval is one pixel.

[0098] And calculate the correlation coefficient between each local area H and the template image g1 according to the normalized correlation coefficient formula to obtain the correlation coefficient matrix ;

[0099] According to the correlation coefficient matrix It is determined whether the local area H in the output image S1 is the same as the template image g1. If they are the same, the position of the local area H is the position of the target image Sg.

[0100] In some embodiments, when it is determined whether the scanned local area H is the same as the template image g1, the scanning is stopped.

[0101] Correlation coefficient matrix Processing is performed to determine whether there is a part in S that is highly correlated with g.

[0102] Specifically, the correlation coefficient matrix Defined as:

[0103]

[0104] In the formula for The covariance of and g, for The variance of g is D, and the grayscale mean of g is ,image The grayscale mean is , the calculation integration is:

[0105]

[0106] The correlation coefficient range of the above formula is:

[0107] ≤1

[0108] In summary, to determine whether there is a local area H in the output image S1 that is similar to the template image g1 or has a high similarity, use The results are used to judge when When the value is close to 1, the local area H in the output image S1 is very similar to the template image g1. Then, a reasonable threshold is set to determine whether the local area H analyzed in the output image S1 is the same as the template image g1.

[0109] In some embodiments, the absolute value of the correlation coefficient range is manually judged. If the error is acceptable, the threshold value is set to 0.5. If the error deviation is unacceptable, 0.5 can be increased to 0.7, and the error is continued to be observed. If the error is acceptable, the threshold value is set to 0.7. If the error deviation is unacceptable, the threshold value is continued to be increased until the error is within an acceptable range.

[0110] When the angles between the target image Sg and the template image g1 are different, a template library is first established. Specifically, the forward mapping method is first used to rotate the initial template image g1 to obtain the rotated template image g2, and the template image g3 is obtained by the reverse mapping method and the bilinear interpolation algorithm; the portion of the edge of the template image g3 where the grayscale value changes due to the rotation is eliminated to obtain the template image g4, so that the grayscale value in the rotated template image g4 is consistent with the grayscale value in the template image g1 before rotation; a plurality of template images g4 at different angles form a template library.

[0111] A local area H is intercepted in the output image S1, and the position of the target image Sg in the output image S1 is calculated according to the correlation coefficient between the local area H and a plurality of template images g4 in the template library.

[0112] In some embodiments, the template image g4 obtained from each rotation is collected into a template library, and in the subsequent calculation process, comparison can be directly performed based on the content of the template library.

[0113] In the forward mapping, the corresponding position of each pixel in the output image in the original image is first determined, and then the pixel value is obtained from this position; if the mapped coordinates are not integers, an interpolation algorithm is usually required to estimate the pixel value. In the case of rotating the template image g1, the new position of each pixel in the template image g1 in the rotated template image g2 is calculated by the forward mapping method; the reverse mapping is opposite to the forward mapping, it first determines the corresponding position of each pixel position in the output image in the original image, and then obtains the pixel value from this position; in the reverse mapping, for each pixel in the output image, its coordinates in the original image are reversely calculated, and then the interpolation algorithm is used to obtain the pixel value; the reverse mapping method calculates the original position of each pixel in the template image g2 in the template image g1 before rotation, and uses the interpolation algorithm to obtain the template image g3; the bilinear difference algorithm is used to estimate the pixel value of non-integer coordinate points in the image; specifically, the bilinear difference first performs linear interpolation on two adjacent pixels in the horizontal direction, and then uses these two interpolation results to perform a second linear interpolation in the vertical direction, and finally obtains the pixel value of the non-integer coordinate point.

[0114] Reference Figure 3 , Figure 6 , Figure 7 , Figure 8 and Fig. 9 ,in Figure 3 is a schematic diagram of the grayscale and angle of the template image g1, that is, the schematic diagram before rotation. Figure 6 is a schematic diagram of the grayscale of the template image g1 and the angle after rotation. Figure 7 is a schematic diagram of the grayscale and angle of the template image g2, that is, the template image g2 is obtained by applying the forward mapping method to the template image g1. Figure 8 is a schematic diagram of the grayscale and angle of the template image g3, that is, the template image g3 is obtained by the reverse mapping method and the bilinear difference algorithm for the template image g2. Fig. 9 is a schematic diagram of the grayscale and angle of the template image g4, that is, the template image g4 is obtained by removing the grayscale value change of the edge of the template image g3 caused by rotation; as can be seen from the above, the grayscale gradient appears at the edge of the image after rotation, and it cannot keep the grayscale consistent with the original image. When the gradient part is used as the target pattern template and matched with the camera output image S1, it will result in The value is low and deviates from 1, which makes the error larger. Removing the edge from the calculation will improve the matching accuracy. After the target pattern template g1 is rotated, a fixed number of edge pixels can be removed to ensure that the position of the target image Sg is accurately calculated.

[0115] The step of scanning the target image Sg according to the template image g4 comprises:

[0116] The size of the local area H is set to m*n, where m is the long side of the local area H, and n is the short side of the local area H; specifically, the size of the target image Sg is m*n, and the size of the template image g4 is m*n, that is, the target image Sg, the template image g4 and the local area H are of the same size; it is worth noting that the part of the grayscale value changed due to the edge rotation of the template image g3 with a changed grayscale value is eliminated here, which means that the part of the grayscale value change does not participate in the calculation, and at this time, the side length of the template image g4 is the same as the side length of the template image g1.

[0117] In the output image S1 of the camera, the local areas H are scanned sequentially.

[0118] In some embodiments, reference Fig.10 , Fig.11 and Fig.12 , the size of the output image S1 is M*N, and the local area H is scanned starting from the upper left corner of the output image S1, that is, the coordinates of the upper left corner of the local area H coincide with the coordinates of the upper left corner of the output image S1 as the initial point, and the scanning begins. During the scanning process, it is first scanned in the horizontal direction, that is, (1,1) to (M,1), and then it turns to (1,2) and scans to (M,2) after the scanning is completed, and finally it turns to (1,N) and scans to (M,N); specifically, in this process, each scanning interval is one pixel.

[0119] According to the normalized correlation coefficient formula, the correlation coefficient between each local area H and each template image g4 in the template library is calculated to obtain the correlation coefficient matrix ;

[0120] According to the correlation coefficient matrix It is determined whether the local area H in the output image S1 is the same as the template image g4. If they are the same, the position of the local area H is the position of the target image Sg.

[0121] The correlation coefficient calculation method and correlation coefficient matrix The judgment method is the same as the processing method when the angles between the target image Sg and the template image g1 are the same as above. The only difference here is that the definition of the selected template image is different, which will not be elaborated here.

[0122] According to the above calculation, it can be obtained that the coordinates of the upper left corner of the target image Sg in the output image S1 are (Xsg, Ysg); therefore, the geometric center coordinates of the target image Sg are (Xsg+m / 2, Ysg+n / 2).

[0123] Among them, the energy of the laser beam focused on the camera is larger than that of other beams received by the photosensitive surface of the camera. Therefore, the gray value of the laser beam received by the camera is larger (gray value 0 represents black, gray value 255 represents white). The laser spot image L can be located by the gray value, and the centroid coordinates of the laser spot image L can be accurately determined by combining the centroid algorithm. Specifically, the calculation of the centroid coordinates of the laser spot image L includes the following steps:

[0124] Scanning pixels of the output image S1;

[0125] Perform mean filtering on the output image S1 and select multiple pixels with the largest grayscale values;

[0126] The pixel grayscale values ​​in the neighborhood of the pixel point are counted, and the maximum mean value is calculated. The central area of ​​the laser spot image L is located near the pixel point corresponding to the maximum mean value;

[0127] Specifically, the coordinates of the upper left corner of the camera output image S1 are set to (1, 1);

[0128] Scan the image pixels (1, 1), (1, 2), (2, 1), (2, 2), (M, N) in a row cycle; where M is the long side of the output image S1, and N is the short side of the output image S1; calculate the grayscale histogram, and the histogram is Fig.15 As shown, since it is composed of multiple rectangular frames densely arranged, in this embodiment, the grayscale histogram is represented by the line connecting the top midpoints of each rectangular frame, wherein the abscissa of the grayscale histogram is the grayscale value, and the ordinate of the grayscale histogram is the number of pixels with changed grayscale values. Fig.15 The statistics of some pixel gray values ​​are omitted, that is, Fig.15 The figure in is a schematic diagram; based on this, the pixels corresponding to the maximum grayscale value corresponding to the number that is not zero can be found. In this embodiment, the ten pixels with the largest grayscale value are analyzed based on this principle;

[0129] The pixel coordinates selected above are taken as the base points, so the number of base points is 10. The pixel coordinates in the 24 neighborhoods of the 10 base point pixels are counted. The number of pixel coordinates does not exceed 15. Limiting to 24 pixels can effectively remove noise points and improve calculation efficiency. The mean of the required pixel points in the neighborhood of the 10 base point counted is calculated, and the maximum value of the mean is determined through a loop. The base point corresponding to the maximum value of the mean is the pixel point covered by the laser spot.

[0130] In some embodiments, the cyclic determination is to determine the neighborhood mean of the selected 10 base points.

[0131] Specifically, in order to further reduce the error, based on the above, calculating the centroid coordinates of the laser spot image L includes the following steps:

[0132] Calculate the gray value of the pixel in the neighborhood corresponding to the maximum mean value to decay to its gray value e -2 The pixel coordinates of

[0133] The pixel coordinates that decay to the gray value e-2 are marked as (mi, nj), where i and j are natural numbers, specifically, 1≤i≤M, 1≤j≤N;

[0134] The corresponding gray value is recorded as e i, j , i, j are natural numbers, specifically, 1≤i≤M, 1≤j≤N;

[0135] According to the centroid calculation method, the horizontal axis coordinate m of the centroid of the laser spot image L is obtained. α and the vertical axis coordinate n β .

[0136] That is, the base point corresponding to the maximum mean value above is taken as the initial point, and the pixel coordinates whose pixel grayscale values ​​in the neighborhood of the 10 base point pixels decay to the base point grayscale value e-2 are counted, so that the pixel coordinates and their grayscale values ​​within the required range are used as data, and the laser spot data coordinate system is marked as (m1, n1), (m2, n2),,, ( , ), the corresponding gray value is recorded as ,, , find the centroid corresponding to this data, the centroid calculation method is:

[0137]

[0138] in , They are respectively the horizontal and vertical coordinates of the center of mass of the coordinate system of the laser spot image L data.

[0139] This method has high calculation accuracy and strong resistance to stray light. The system and wafer are stationary and continuously collect 100 frames of image data to calculate their center of mass coordinates. The repeatability of the horizontal and vertical coordinates of the 100 frames of image data is ±0.1 pixel, thereby achieving accurate detection of the center of mass of the laser spot.

[0140] Based on the above, the horizontal axis coordinate of the center of mass of the laser spot image L can be calculated: , the vertical axis coordinate of the center of mass of the laser spot image L , the horizontal coordinate of the geometric center of the target image Sg is Xsg+m / 2, and the vertical coordinate of the geometric center of the target image Sg is Ysg+n / 2, so the pixel distance between them can be obtained;

[0141] Specifically, the horizontal axis coordinate of the center of mass of the laser spot image L is Subtract the horizontal axis coordinate Xsg+m / 2 of the geometric center of the target image Sg, and the pixel distance between the centroid of the laser spot image L and the geometric center of the target image Sg in the X-axis direction is ( -(Xsg+m / 2));

[0142] The vertical axis coordinate of the laser spot image L Subtract the vertical axis coordinate Ysg+n / 2 of the geometric center of the target image Sg, and the pixel distance between the centroid of the laser spot image L and the geometric center of the target image Sg in the Y-axis direction is ( -(Ysg+n / 2));

[0143] The moving distance of the laser beam is obtained by dividing the pixel distance by the system optical path magnification, and the moving direction of the laser beam is set according to the positive and negative values ​​of the calculation result to control the center of mass of the laser beam to move to the geometric center of the target pattern.

[0144] In some embodiments, the alignment accuracy between the laser spot image L and the target image Sg is within 1 pixel distance.

[0145] For ease of understanding, refer to Fig.13 and Fig.14 ,in Fig.13 Schematic diagram of the actual position of the laser spot image L and the target image Sg before alignment in the output image S1 according to the embodiment of the present invention; Fig.14 Schematic diagram of the actual position of the laser spot image L and the target image Sg after alignment in the output image S1 before the process according to an embodiment of the present invention.

[0146] In some embodiments, the magnification varies when the working distance of the entire system varies.

[0147] In some more specific embodiments, the magnification is 1.25 at a working distance of 360 mm, and the magnification is 0.54 at a working distance of 800 mm.

[0148] In some embodiments, reference Fig.16 The laser beam 2 emitted from the endpoint detection device 1 enters the etcher cavity 4 through the etcher window 3, wherein an electrostatic chuck 6 is provided in the etcher cavity 4, and the wafer 5 to be etched is placed on the electrostatic chuck 6. After passing through the etcher window 3, the laser beam 2 irradiates the surface of the wafer 5 to be etched, and after being reflected by the surface of the wafer 5 to be etched, it is imaged in the camera of the endpoint detection device 1. The pixel distance obtained by the centroid coordinates of the laser spot and the geometric center coordinates of the target image is used to control the laser beam 2 to move toward the target pattern.

[0149] In some embodiments, reference Fig.17, is a schematic diagram of etching grooves, wherein the etching depth of the etching groove 7 ranges from 100nm to 5000nm, and the etching accuracy is ±0.53nm.

[0150] The implementation principle of the automatic alignment method of laser interferometric endpoint detection in the embodiment of the present application is to determine whether there is a part in the output image S1 that is similar to the target image Sg or has a high similarity. This process uses the correlation coefficient matrix The results are used to judge when When the absolute value of is close to 1, the target image Sg in the output image S1 is very similar to the template image g1 or the template image g4. Then, a reasonable threshold is set to determine whether the local part analyzed in S is the same as the template image g1 or the template image g4.

[0151] In the above manner, the geometric center coordinates of the target image Sg in the camera output image S1 can be accurately detected only when the target image Sg and the template image g1 have the same angle.

[0152] In order to handle the situation when different template images g1 have different rotation angles, the present invention proposes template databaseization, that is, algorithmically rotating the target pattern Sg to establish multiple discrete template libraries to match the template image g1 in the camera output image S1 when it has different rotation angles and the correlation coefficient is close to 1 when the set threshold is met; in this process, since the rotation will generate new edges, it is necessary to remove these edges to obtain each new template, so as to accurately calculate the geometric center of the target image Sg in the camera output image S1.

[0153] Although the embodiments of the present invention are described in detail above, it is obvious to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations are within the scope and spirit of the present invention as described in the claims. Moreover, the present invention described herein may have other embodiments and may be implemented or realized in a variety of ways.

Claims

1. An automatic alignment method for laser interferometric endpoint detection, characterized in that: The following steps are involved: The wafer surface is illuminated and the laser beam is directed onto the wafer surface; Receiving an output image S1 formed by the illuminated area on the wafer surface, wherein the output image S1 includes a target image Sg formed by a target pattern on the wafer and a laser spot image L reflected by the wafer; Calculate the geometric center coordinates of the target image Sg in the output image S1 and the centroid coordinates of the laser spot image L to obtain the pixel distance corresponding to the two coordinates; According to the pixel distance, the laser beam is controlled to move toward the target pattern on the wafer surface so that the centroid of the laser spot image L coincides with the geometric center of the target image Sg, thereby achieving automatic alignment; Calculating the geometric center coordinates of the target image Sg includes the following steps: Intercepting a target image Sg from the output image S1 to obtain a template image g1; Cutting out a local area H from the output image S1, and calculating the position of the target image Sg in the output image S1 according to the correlation coefficient between the local area H and the template image g1; Scanning the target image Sg comprises the following steps: When the target image Sg and the template image g1 have different angles; The initial template image g1 is rotated by using the forward mapping method to obtain the rotated template image g2, and the template image g3 is obtained by using the reverse mapping method and the bilinear interpolation algorithm; Eliminate the portion of the edge of the template image g3 where the grayscale value changes due to the rotation and obtain the template image g4, so that the grayscale value in the rotated template image g4 is consistent with the grayscale value in the template image g3 before the rotation; The template images g4 at multiple different angles form a template library; A local area H is intercepted from the output image S1, and the position of the target image Sg in the output image S1 is calculated according to the correlation coefficient between the local area H and a plurality of template images g4 in the template library.

2. The automatic alignment method according to claim 1, characterized in that: Scanning the target image Sg comprises the following steps: When the target image Sg and the template image g1 have the same angle; Setting the size of the local area H to m*n, and sequentially scanning the local area H in the output image S1; The correlation coefficient between each local area H and the template image g1 is calculated according to the normalized correlation coefficient formula to obtain a correlation coefficient matrix ; According to the correlation coefficient matrix It is determined whether the local area H in the output image S1 is the same as the template image g1. If they are the same, the position of the local area H is the position of the target image Sg.

3. The automatic alignment method according to claim 1, characterized in that: The step of scanning the target image Sg according to the template image g4 comprises: Setting the size of the local area H to m*n, and sequentially scanning the local area H in the output image S1; The correlation coefficient between each local area H and each template image g4 in the template library is calculated according to the normalized correlation coefficient formula to obtain a correlation coefficient matrix ; According to the correlation coefficient matrix It is determined whether the local area H in the output image S1 is the same as the template image g4. If they are the same, the position of the local area H is the position of the target image Sg.

4. The automatic alignment method according to claim 2 or 3, characterized in that: The coordinates of the upper left corner of the target image Sg in the output image S1 are (Xsg, Ysg); The geometric center coordinates of the target image Sg are (Xsg+m / 2, Ysg+n / 2).

5. The automatic alignment method according to claim 4, characterized in that: Calculating the centroid coordinates of the laser spot image L includes the following steps: Scanning pixels of the output image S1; Perform mean filtering on the output image S1 and select multiple pixels with the largest grayscale values; The pixel grayscale values ​​in the neighborhood of the pixel point are counted to calculate the maximum mean value. The central area of ​​the laser spot image L is located near the pixel point corresponding to the maximum mean value.

6. The automatic alignment method according to claim 5, characterized in that: Calculating the centroid coordinates of the laser spot image L includes the following steps: Calculate the gray value of the pixel in the neighborhood corresponding to the maximum mean value to decay to its gray value e -2 The pixel coordinates of will decay to its gray value e -2 The pixel coordinates are marked as (m i , n j ), i, j are natural numbers; The corresponding gray value is recorded as e i, j , i, j are natural numbers; According to the centroid calculation method, the horizontal axis coordinate m of the centroid of the laser spot image L is obtained. α and the vertical axis coordinate n β .

7. The automatic alignment method according to claim 6, characterized in that: Subtract the horizontal axis coordinate of the geometric center of the target image Sg from the horizontal axis coordinate of the center of mass of the laser spot image L, and obtain the pixel distance between the center of mass of the laser spot image L and the geometric center of the target image Sg in the X-axis direction: ( -(Xsg+m / 2)); Subtract the vertical axis coordinate of the geometric center of the target image Sg from the vertical axis coordinate of the laser spot image L to obtain the pixel distance between the centroid of the laser spot image L and the geometric center of the target image Sg in the Y-axis direction: ( -(Ysg+n / 2)); The pixel distance is divided by the system optical path magnification to obtain the moving distance of the laser beam, and the moving direction of the laser beam is set according to the positive and negative values ​​of the calculation result to control the center of mass of the laser beam to move to the geometric center of the target pattern.

8. The automatic alignment method according to claim 2 or 3, characterized in that: When the correlation coefficient matrix When it is close to 1, the local area H in the output image S1 is the position of the target image Sg.

Images