JP2012178159A - Flaw inspection method and flaw inspection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for automating setting of a cell area to be set by a recipe of a flaw inspection device.SOLUTION: For discrimination of a Cell Mat Area, a method of scanning an image using a difference between distribution features of Gray Levels of the Cell Mat Area and a non-Cell Mat Area to divide the Cell Mat Area and the non-Cell Mat Area from the result is adopted. Concretely, a threshold to be a standard for discriminating start and finish points of a Cell Mat is calculated in an Area where only a Memory Cell is present and the start and finish points are subsequently searched by adapting the threshold and connected respectively to form a Cell Area.

Description

  The present invention relates to a defect inspection apparatus and a defect inspection method for detecting foreign matter on the surface of semiconductor wafers, photomasks, magnetic disks, liquid crystal substrates, etc., and a pattern inspection method, and in particular, a defect inspection method provided with means for automatically determining an inspection area. And a defect inspection apparatus.

  In order to inspect defects generated in wafers (wafers) in a semiconductor defect inspection apparatus, it is necessary to create a recipe (recipe) in which an inspection area and inspection conditions are set. In this case, the defect inspection can be automatically performed, but it is the current technical level that most of the recipe creation for executing the automatic inspection needs to be judged by human manual work.

  Setting up Cell Mat is a time-consuming task that makes it difficult for people who are not familiar with inspection equipment. In order to perform high-sensitivity inspection with a defect inspection system, it is necessary to divide the die into periodic Memory Cell Mat AREA and non-Memory Cell Mat Area and set different inspection methods and different inspection sensitivities for each Area There is. For example, in the case of Memory Device, Die is classified into Cell Area, Sens-Amp Area, SWD Area, Peri Area, etc. When inspecting, Cell Mat Area is inspected by Cell comparison, and other Areas are inspected by Die comparison. The method is commonly used.

  In the conventional technique, humans manually measure the distance between the Cell Mat, which is the unit of the Cell Area, and the next Cell Mat, and repeat the characteristics of the Cell Mat. A method was used in which the start and end of the Cell Mat confirmed with the eye were clicked with the mouse and copied N times by a few minutes.

  For example, a method as described in Patent Document 1 is known as a method of obtaining a density histogram of an image of an object to be inspected and determining the outer peripheral edge of the chip based on the degree of dispersion of the histogram. However, the identification of the cell area provided inside one chip (which may be referred to as Die) has a small density difference and it is difficult to set a threshold value for determining the boundary. It was.

Japanese Patent No. 3219094

  As mentioned above, if a human manually sets Cell Mat, three major problems will occur.

  The first is that the accuracy of the inspection area drops when a human clicks the mouse to set it, and fine area settings such as Cell Edge cannot be made. In other words, in recent semiconductor products, defect generation suppression at the edge of the Cell Mat is an important factor for improving the yield, but there is a problem that the inspection of the edge of the Cell becomes impossible because the accuracy cannot be secured. .

  Secondly, it takes a lot of time to set the inspection area, and the recipe creation time greatly increases.

  Third, in order to manually set the inspection area, an engineer who has received related training is required.

  The objective of this invention is providing the defect inspection apparatus and defect inspection method which solved the said subject.

  In order to solve the above-described problems, the above-described various problems appear in the conventional method in which human manual calculation and confirmation with the eyes and designation by clicking with the mouse are performed. As a solution to this, in the present invention, (Auto Threshold Search Area) scan is performed in a specific cell mat area, and a threshold value serving as a reference for detecting a pixel (start point) from which the cell mat starts and a pixel (end point) from which the cell mat ends is detected. Ask. After that, a specific 1 Die scan in the wafer including the same Auto Threshold Search Area is performed. The scanning results are compared in units of cell pitches, and a plurality of start points and end points of the cell mat are acquired based on the threshold value obtained in the auto threshold search area.

  Cell Mat start and end points are detected in the X and Y directions to detect a plurality of cell mats. Thereafter, the cell mat area is determined by connecting the detected start point and end point.

  According to the present invention, the setting of the cell area set by the recipe of the defect inspection apparatus can be automated and can be set accurately.

  In other words, even when a person who is not skilled in setting the inspection area creates the recipe by automating the cell mat area setting, the inspection area setting can be created easily and quickly. In addition, since the inspection area is set by the Gray Level characteristic analysis, it is possible to accurately set up to the edge of the cell mat, and it is possible to inspect the cell mat edge. In addition, it can be expected that the time and labor required to create the recipe will be greatly reduced.

The whole block diagram which detects each start point and end point of Cell Mat by selecting Memory Cell Area from Chip in Wafer, performing X direction scan and Y direction scan. An example of a TDI image in the Memory Cell Area for obtaining the thresholds necessary to detect the start and end points of each cell mat, and the TDI image's Gray Level as the Y axis and the scan X position as the X axis Graph. The 1-line portion of the TDI image in FIG. 2 and the lower graph were matrixed in units of cell pitch. A figure in which a total of m Cell Pitch is composed of n Pixels and a set is created with values in the same order in the matrix. The figure which calculated | required each of the maximum value in the set made in FIG. The figure which calculated | required each of the minimum value in the set made in FIG. The figure which represented the value which pulled the maximum value calculated | required in FIG. 4, and the minimum value calculated | required in FIG. 5 in Cell_Threshold. The figure represented as an actual TDI image of the Die Start Area and CellMat start and end points existing in one Die.

  The overall configuration of the present invention will be described with reference to FIG.

  1 shown in FIG. 1 is Wafer which is a semiconductor material. When various semiconductor processes are run through this wafer and finally cut in die units, it becomes a single semiconductor chip. That is, the basic unit generally referred to as one chip is Die shown in 2.

  In the case of DRAM, a large number of elements (transistors) are formed in one Die. Generally, a group of minimum units in which these elements are collected and arranged is called a cell mat. One of the Cell Mats is shown in FIG. In the Die shown in FIG. 1, there are four groups of 25 cell mats, and there are a total of 100 cell mats.

  In order to accurately distinguish the Cell Mat from the non-Cell Mat, the start point (4, 9 to 18 in FIG. 1 and 45 in FIG. 7) and the end point (5, 19 to 28 in FIG. 1 and 46 in FIG. 7) ) Need to be distinguished.

  In FIG. 1, 9 to 18 indicate the start points of the Cell Mat, and 19 to 28 indicate the end points of the Cell Mat. A method for detecting the start point and end point of each Cell Mat from the scan results of 7 (X scan) and 8 (Y scan or X scan) will be described below.

  First, since the start point and end point are detected from the data (Gray Level) obtained from the scans 7 and 8 instead of being determined by humans, a gray level threshold is required to detect the start point and end point. become. An area for obtaining this threshold is called an Auto Threshold Search Area, and only this designation is performed by a person searching for a Memory Cell Mat. In FIG. 1, this Auto Threshold Search Area is shown as 6, and 6 is further enlarged and shown as 29 in FIG.

  The 29 images obtained from the scan are digital images, and each pixel has a value from 0 to 255 called Gray Level. In FIG. 2, first, the gray level is taken from the bottom line of 32 in the 29 Auto Threshold Search Area, and the gray levels for the X pixel positions are arranged in a matrix in units of cell pitches and arranged as shown in FIG. This data is collected and shown in a graph and an equation as shown in the graph 33 in FIG. 2 and the equation in FIG. However, the graph shown in FIG. 33 and the formula shown in FIG. 3 are written on the assumption that one cell pitch has n pixels. In the equation of FIG. 3, the matrix is arranged because there are m cell pitches composed of n pixels.

  In 34, only the gray level of the pixel located at the beginning of the cell pitch is collected into each GL_Matrix_1 among the pixels constituting one cell pitch. In 35, each of the following Pixels was collected into GL_Matrix_2. When this definition is performed up to n-th as in 36, the n-th pixel becomes GL_Matrix_n.

  Next, as shown in FIGS. 4 and 5, the maximum Gray Level and the minimum Gray Level in GL_Matrix_1 are obtained and set as GL_Matrix_Max (1) and GL_Matrix_Min (1). Similarly, GL_Matrix_Max (2), GL_Matrix_Max (3) ,,,,, GL_MATRIX_MAX (N) and GL_Matrix_Min (2), GL_Matrix_Min (3), ,,,, GL_Matrix_Min (n) are obtained.

  In FIG. 6, GL_Matrix_Min (1) is subtracted from GL_Matrix_Max (1) obtained by the equations of FIGS. 4 and 5 to make Cell_Threshold1, and similarly Cell_Threshold2 to Cell_Thresholdn are obtained. Finally, the maximum value up to n Cell_Thresholds is calculated and defined as Cell_Treshold_Line1.

  Next, the gray level of the line on 1 Pixel of the bottom 1 Line shown in 32 of FIG. 2 is taken, and Cell_Treshold_Line 2 is finally calculated through FIGS. 3, 4, 5 and 6.

  If the same calculation is performed for P lines in the Y direction from No. 32 to No. 30, the Cell_Treshold from Cell_Treshold_Line1 to Cell_Treshold_LineP can be calculated. Each Cell_Threshold_Line is used as a threshold for each Line that determines the start and end points of the Cell Mat.

  Next, the same width (P Pixel) in the Y direction as the Auto Threshold Search Area shown in 6 of FIG. 1 and 29 of FIG. 2 is scanned by 1 Die as shown in 7 of FIG.

  The purpose of scanning is to detect the start point (No. 9 to No. 18) and the end point (No. 19 to No. 28). A method for detecting the respective start and end points will be described below. First, Gray level data for the X position is obtained as a matrix in units of cell pitches and 1 Die is acquired. Subtraction is performed for each cell pitch from the Die Start position 44 in FIG. 7 from the acquired matrix data, and it is determined whether the absolute value of the calculation result is larger or smaller than the Cell_Treshold_Line of the corresponding line. The start point and end point candidates are determined based on the result of this determination. In the following, a method for determining the true start point and end point from the start point and end point candidates will be described.

  First, a pixel whose absolute value does not exceed a threshold value (Cell_Treshold_Line) is searched by subtracting from the Die Start Pixel in matrix units of Cell Pitch intervals. If the consecutive 3 pixels including the pixel do not exceed Cell_Threshold, the pixel is defined as S_Pixel_Candidate that represents the start point candidate. Then, when 90% of the Pixel of the Cell Pitch comparison result next to the Cell Pitch including S_Pixel_Candidate is a comparison result that does not exceed the threshold, this Pixel is defined as S_Pixel. Then, if the comparison between the cell pitches is compared to the end of 1Die, a plurality of S_Pixels can be found. This can be expressed as (1) to (4) below.

(1) First, from the Die Start Pixel, the gray level data in the Cell Mat Image that are matrixed in Cell Pitch units are subtracted at Cell Pitch intervals. Then, a pixel whose absolute value of the subtraction difference does not exceed the threshold value (Cell_Treshold_Line) is searched.
(2) If the consecutive 3 pixels including the pixel do not exceed Cell_Threshold, the pixel is defined as S_Pixel_Candidate that represents the start point candidate.
(3) Then, when 90% of the Pixel of the image subjected to the subtraction after the subtraction for obtaining S_Pixel_Candidate does not exceed the threshold (Cell_Treshold_Line), this S_Pixel_Candidate is defined as S_Pixel.
(4) Subsequently, when subtracting the gray level data in the cell pitch unit and subtracting at the cell pitch interval and comparing the absolute value of the subtraction difference with the threshold to the end of 1 Die, multiple S_Pixels Can find out.

  At the same time as the calculation to find S_Pixel, the calculation to find E_Pixel that shows the end of Cell Mat is also performed. That is, if the comparison between Cell Pitch from Die's 44th Start Pixel is continued, it is possible to find a Pixel exceeding the threshold. If the consecutive 3 pixels including the pixel exceed the Cell_Threshold, the pixel is defined as E_Pixel_Candidate that represents the end point candidate. After that, when 90% of the comparison result pixel of the next cell pitch including the E_Pixel_Candidate exceeds the threshold, it is defined as E_Pixel, and a plurality of E_Pixels are searched by the cell pitch comparison up to the end of 1Die. This can be expressed as the following (5) to (8).

(5) Pixels exceeding the threshold can be found by continuously subtracting the gray level data from Cell No. 44 from Die's No. 44 start pixel at Cell Pitch intervals. .
(6) Then, if the consecutive 3 pixels including the pixel exceed the Cell_Threshold, the pixel is defined as E_Pixel_Candidate that represents the end point candidate.
(7) After that, when 90% of the Pixel of the image subjected to the subtraction after the subtraction for obtaining E_Pixel_Candidate exceeds the threshold, it is defined as E_Pixel.
(8) Subsequent subtraction at the cell pitch interval between the gray level data matrixed in units of cell pitches, and comparing the absolute value of the subtraction difference with the threshold to the end of 1 Die, multiple E_Pixels Can find out.

  However, if E_Pixel is detected first before S_Pixel is detected, it returns to the S_Pixel detection loop and the next cell pitch comparison of Cell Pitch that contains S_Pixel that meets the condition of continuous 3 Pixels that does not exceed the threshold (Cell_Treshold_Line) S_Pixel is detected by reducing the condition that 90% of the resulting Pixel does not exceed the threshold by 10%.

  In other words, since E_Pixel may be detected first before S_Pixel is detected, in this case, the S_Pixel is detected by returning to the S_Pixel detection loop described above and setting the condition of 90% as 80%. Still, if S_Pixel is not detected, S_Pixel is detected by setting the condition of 80% as 70%. The condition is reduced every 10% until S_Pixel can be detected.

  Similarly, when S_Pixel appears after E_Pixel is not detected after S_Pixel, it returns to the E_Pixel detection loop and the next cell of Cell Pitch that contains E_Pixel that meets the condition of continuous 3 Pixels exceeding the threshold (Cell_Treshold_Line1) The E_Pixel is detected by reducing the condition that 90% of the Pitch comparison result Pixel should exceed the threshold by 10%.

  That is, there is a case where S_Pixel appears after S_Pixel is not detected, and in this case, the E_Pixel is detected by setting the condition of 90% as 80% by returning to the E_Pixel detection loop described above. If E_Pixel is still not detected, E_Pixel is detected by setting the condition of 80% as 70%. The condition is reduced every 10% until E_Pixel can be detected.

  By using the above method, each Cell_Treshold_Line can be applied to each Line to find each start point and each end point in the X direction. That is, if P lines are calculated from the scan 7 in FIG. 1, 10 start points (S_Pixel_1 to S_Pixel_10) and 10 end points (E_Pixel_1 to E_Pixel_10) can be detected for each P lines.

  Next, S_Pixel_1 obtained from the obtained P lines finally determines S_Pixel_1 as the X position where the position of the Pixel is the best match. Similarly, the X position that most closely matches the other start points and end points is finally defined as S_Pixel and E_Pixel.

  Next, in order to calculate the start and end points of the Cell Mat in the Y direction, a partial X scan or Y scan is performed as shown in FIG. Here, the start point and end point in the Y direction are calculated by the same method as the method in which the image obtained from the scan is rotated 90 degrees to determine the start point and end point in the X direction.

  The cell mat is defined between S_Pixel and E_Pixel by connecting the start and end points in the XY direction finally calculated.

  By such a method, a defect inspection apparatus capable of performing more accurate defect inspection can be realized by determining the start point and end point of the Cell Mat and applying different inspection methods and inspection conditions depending on the type of Cell.

  1 ... Wafer, 2 ... Die, 3 ... Cell Mat, 4 ... Start point of Cell Mat, 5 ... End point of Cell Mat, 6 ... Auto Threshold Search Area, 7 ... X scan, 8 ... Y scan, 9-18 ... Cell Mat Start point, 19 to 28 ... Cell Mat end point, 29 ... Auto Threshold Search Area TDI Image, 30 ... PLine scan, 31 ... One Cell Pitch, 32 ... 1Line scan, 33 ... 1Line scan result Gray level graph, 34 ... m Cell Pitch (constituted in n Pixels), a collection of each set of first Pixels, expressed as GL_Matrix_1, 35 ... m Cell Pitch ( Figure showing GL_Matrix_2's set of collections of each of the second pixels that make up n pixels), 36th nth piece that makes up m cell pitches (composed of n pixels) Fig. 37: Auto Threshold Search Area, 38 ... X scan of the Pth line during 1Die scan, 39 ... X scan of the second line from the bottom during 1Die scan, 40 ... X scan of the bottom line during 1Die scan, 41 to 43: Scan in the Y direction when scanning for 1 Die, 44: One Die Start Pixel represented by TDI image, 45 ... Cell Mat start point represented by TDI image, 46 ... Cell Mat represented by TDI image end point.

Claims (10)

In defect inspection equipment,
Obtaining an image of a group formed by collecting and arranging a plurality of elements, dividing the gray level of the image into predetermined intervals, performing subtraction for each predetermined interval between the divided gray levels, A defect inspection apparatus comprising a processing unit that obtains at least one position of a start point and an end point of the group by comparing an absolute value of a subtraction result with a threshold value. The defect inspection apparatus according to claim 1,
The processor is
The absolute value is less than or equal to the threshold;
Furthermore, when a predetermined pixel after the pixel for which the absolute value is obtained is equal to or less than the threshold value, the pixel having the absolute value is set as the candidate for the starting point. The defect inspection apparatus according to claim 2,
The processor is
An inspection apparatus that recognizes the candidate for the start point as the start point if the gray level of a pixel that has been subtracted after subtraction for obtaining the absolute value is equal to or less than a predetermined ratio of the threshold value. The defect inspection apparatus according to claim 1,
The processor is
The absolute value is greater than the threshold;
Further, when a predetermined pixel after the pixel from which the absolute value is obtained is larger than the threshold value, the pixel having the absolute value is set as a candidate for the end point. The defect inspection apparatus according to claim 4,
The processor is
An inspection apparatus that recognizes the end point candidate as the end point if the gray level of a pixel that has undergone subtraction after subtraction for obtaining the absolute value is greater than a predetermined ratio of the threshold value. In the inspection condition setting method for defect inspection,
Obtain an image of the group formed by collecting and arranging multiple elements,
The gray level of the image is divided into predetermined intervals,
Subtraction is performed at the predetermined intervals between the separated gray levels,
An inspection condition setting method characterized by obtaining at least one position of a start point and an end point of the group by comparing an absolute value of a result of the subtraction and a threshold value. In the inspection condition setting method according to claim 6,
The absolute value is less than or equal to the threshold;
Furthermore, when a predetermined pixel after the pixel from which the absolute value is obtained is equal to or less than the threshold value, a pixel having the absolute value is set as a candidate for the start point. The inspection condition setting method according to claim 7,
An inspection condition setting method, wherein a start point candidate is recognized as the start point if a gray level of a pixel after subtraction after obtaining the absolute value is not more than a predetermined ratio of the threshold value. In the inspection condition setting method according to claim 6,
The absolute value is greater than the threshold;
Furthermore, when a predetermined pixel after the pixel from which the absolute value is obtained is larger than the threshold value, a pixel having the absolute value is set as the end point candidate. The defect inspection apparatus according to claim 9,
An inspection condition setting method, wherein the end point candidate is recognized as the end point if the gray level of the pixel after subtraction after obtaining the absolute value is larger than a predetermined ratio of the threshold value.

Images