CN115856000A - Re-inspection preparation method and re-inspection method of electron beam wafer defect re-inspection equipment - Google Patents

Abstract

The invention discloses a rechecking preparation method and a rechecking method of electron beam wafer defect rechecking equipment, wherein the rechecking preparation method comprises the following steps: after the wafer is loaded, initial inspection information is acquired, and the wafer alignment under an optical microscope system is completed, carrying out image gray level optimization of a scanning electron microscope; judging whether to carry out automatic focusing of the scanning electron microscope in the implementation of multi-level wafer alignment under the scanning electron microscope, finishing the level wafer alignment or the automatic focusing according to the judgment result, and finishing the level wafer alignment and the alignment of other higher level wafers; a reference point of the wafer coordinate system is determined. The method optimizes the preparation flow of the EBR equipment rechecking, organically inserts the content for judging whether to execute the rapid automatic focusing into the WA, removes the step disorder and redundancy in the prior art, reduces the risk of damaging the wafer caused by unnecessary SEM image acquisition, improves the efficiency and the accuracy of the rechecking preparation work, and also improves the throughput of the equipment.

Description

Re-inspection preparation method and re-inspection method of electron beam wafer defect re-inspection equipment

technical field

The invention relates to the field of defect detection of patterned wafers, in particular to a re-inspection preparation method and a re-inspection method of electron beam wafer defect re-inspection equipment.

Background technique

Wafer defect detection is included in the front-end process of large-scale integrated circuit (IC) manufacturing, mainly in-line (Inline) detection, which requires high speed. , referred to as BF) optical defect detection equipment and dark field (Dark Filed, referred to as DF) optical defect detection equipment, and electron beam wafer defect detection (e-beam inspection, referred to as EBI) equipment. Referring to FIG. 1 , many dies (Die) in the wafer 100 contain, for example, defects obtained from BF or DF, or EBI preliminary inspection equipment (different marks indicate different classification types of preliminary inspection defects). Subsequently, it is necessary to re-inspect the defects in the initial inspection results (Review), use higher-resolution images, such as up to 2nm, and detect the defect areas of the initial inspection, and perform feature extraction and classification to help IC Manufacturers discover problems in time, optimize the process, and improve yield. The re-inspection mainly uses electron beam wafer re-inspection (e-beam review, EBR for short) equipment.

Recheck equipment

Referring to Fig. 2A, the electron beam scanning wafer defect re-inspection equipment is for example EBR equipment 200, and its component comprises the scanning electron microscope (SEM) of core, and there is electron optical lens barrel (Column) 211 in the scanning electron microscope (SEM) that is a kind of scanning electron microscope (scanningelectron microscope) Microscope, referred to as SEM) system, and a common optical microscopic imaging system (Optical Microscope, namely OM) 212 parallel to it, OM system 212 is used for primary wafer alignment (Wafer Alignment, referred to as WA), and can have different magnifications the objective lens. The EBR equipment 200 also includes a mechanical motion platform (Stage) 213 for placing a wafer 214, a wafer handling system front-end module (EFEM) 215, which includes a robot (Robot) 2151 and a pre-aligner (Pre-aligner) 2152 , a wafer box (Cassette) 2153 can be placed outside it. There is also a vacuum transition chamber 216 between the EFEM 215 and the main chamber including the mechanical motion platform 213 , and there is a manipulator 217 in the main chamber for transferring wafers between the vacuum transition chamber 216 and the mechanical motion platform 213 . After the wafer 214 is loaded, it is placed on the electrostatic tray (E-chuck) on the mechanical movement platform 213 . In addition, the EBR device 200 also has an industrial computer 218 on which software 219 runs, including graphical user interface (GUI) software, system software, database, algorithm software, underlying hardware drivers, and communication software.

Electron optical column (Column) 211, that is, the SEM system, has a higher resolution than that of the initial inspection equipment EBI, for example, the beam spot size is in the range of 2nm–10nm, which is higher than the resolution of the initial inspection BF or DF equipment out one or two orders of magnitude. Referring to FIG. 2B , a schematic structural diagram of the SEM system in the EBR device, its main components include an electron source, a condenser lens, a beam-limiting diaphragm, a scanning coil, an electromagnetic objective lens, and a detector, which are responsible for the generation, focusing, and scanning of electron beams. And the detection of secondary electrons (Secondary Electron) and other electrons including reflected electrons (Backsatted Electron, BSE) generated from the sample, and then through the steps of signal amplification, analog-to-digital conversion (A/D), and signal processing, finally produce SEM images.

Retest preparation method

It is assumed that all parts of the EBR equipment have been calibrated, including the calibration of the required working parameters in the application, including the incident energy (Landing Energy), beam spot size (Beam Size), field of view (FOV) and so on. Like the vast majority of semiconductor equipment, the EBR equipment needs to create a work menu (Recipe) when performing the re-inspection, which includes sub-recipes to determine the steps of its work (often inline), including its own The work is the re-inspection work, and also includes the preparation work before the implementation of its own work. The Recipe created above also includes providing the parameters needed to execute the Recipe during the re-inspection of this type of wafer to be re-inspected, such as grain size/period (Die Pitch), if you want to re-inspect the memory (Memory ) type wafer also includes memory size/period (Cell Pitch), and preparations required for the re-inspection such as wafer alignment (Wafer Alignment, referred to as WA), image grayscale optimization (Image Grayscale Optimization, referred to as IGO) , Quick Auto Focus (QAF for short), etc., and the relevant parameters/contents required in the re-inspection preparation work, such as the templates and matching positions at all levels required for WA, and the determination of the working positions required for IGO and QAF .

The EBR equipment works on the production line to execute the recipe. Of course, it goes without saying at this time that the equipment must be in a normal state. The following are the main steps in the work of EBR equipment:

Steps of wafer loading

Referring to Figure 2A, EFEM takes the wafer out of the wafer cassette, pre-aligns it, and places the wafer on a mechanical motion platform through a vacuum transition chamber. Uncertainty, for example, within ±1° and 200μm respectively (the error range of high-end components may be slightly smaller), so WA, often multi-level WA, is required to correct its orientation and position to its own work, that is, defect re-inspection required precision.

Wafer Alignment Steps

This refers to the alignment of patterned wafers (Patterned Wafer). This is the first step that must be done after all wafers with patterns are loaded.

Referring to Figure 3, when creating a WA Recipe, the template image 300 is usually collected near the center of the wafer, from which a template such as 301 or 302 is selected, which is unique in the figure, and the brightness and contrast must meet the established requirements, and then move The distance between the wafer/mechanical motion platform 1 or a plurality of dies (not too far apart), collect the target image 310 , use a template such as 301 to perform template matching, and search for the best matching position 311 . This method is used regardless of OM images at various magnifications or WA of SEM images. For example, when looking at a high magnification SEM image, there is a template image 320 including a template 321 , and the corresponding matching position 331 is obtained from the template in the target image 330 . There are many commonly used template matching algorithms, such as image similarity algorithms, including cross-correlation (Normalized Cross Correlation, NCC) algorithm or feature-based template matching method, and in sub-pixel precision. In addition, the crystal grains on the wafer are arranged periodically in the X and Y directions, and there are streets between the grains, and there are standard sizes in the industry. Usually there are multiple levels of WA, the most common is 3 levels, the first level uses OM, the middle level uses low magnification SEM images, and the highest level uses high magnification SEM images. Referring to Fig. 4A, in all levels of WA, multiple Die positions along the same row (column) are used to match position results, and more template matching positions away from the center of the wafer are gradually obtained in turn, such as matching positions 411, 412, 413, 414, 415 , 416, joint participation, such as taking the points that match successfully to fit a straight line to obtain the wafer orientation angle θ, and correct it, such as rotating the wafer/mechanical motion platform in the opposite direction or in the wafer coordinate system and mechanical motion platform coordinates Establish a corresponding relationship between departments (depending on the specific system), and then switch to the next level of WA, and do it in the same way until all levels of WA are completed. There are many ways to select the target image acquisition position on the wafer during wafer alignment, not limited to the positions in the horizontal direction as shown in FIG. 4A , but they are all selected based on the periodicity of the crystal grains. For example, referring to Fig. 4B, positions 421, 422, 423, 424, 425 on the wafer (there are matching points in the X and Y directions) can also be used as target image acquisition positions for acquiring target images, and then template matching can also be realized Wafer alignment, the method is similar in principle, and Figure 4A will be used to help explain in the following text. After completing the current level of WA, transition to the next level of WA, and then use the same method to perform the next level of WA with higher resolution images, such as high magnification OM images or TDI/SEM images, until complete The final level of wafer alignment is usually imaged with a high magnification SEM to assure the precision of the EBR equipment's native work. The wafer alignment method in FIGS. 4A and 4B will be used as an example in FIG. 4A hereinafter.

Steps for setting a reference point

After the wafer is aligned, the position of the reference point on the wafer can be determined, which is used to define the origin of the wafer coordinate system. Usually, its position and template image have been determined when the recipe is created, for example, with a special logo (Alignment Mark, AM for short) on the Street near the center of the wafer or an image of a predetermined pattern, usually a high-magnification SEM image, The position of the reference point can be determined by performing template matching, and then the wafer coordinate system is established.

Steps for Auto Focus

The EBR device also requires Quick Auto Focus (QAF) before it can perform its job. This is usually because each wafer thickness varies and the SEM system operating parameters also gradually drift over time (at the same operating variable setting, the focus becomes worse, but also partly because of the wafer thickness/height, that is, the distance from the wafer surface to the SEM system change in working distance). The depth of focus range 501 of the SEM is shown in FIG. 5A . The so-called fast means that usually only the current Im control variable in the SEM electromagnetic objective lens is changed, which is suitable for focal length adjustment in a small range (when the defocus is not serious and there is no obvious astigmatism). made on circle. It is different from the comprehensive and wide-ranging focus calibration before the equipment works. Calibration AutoFocus (CAF for short) is also automatic, which also includes calibration of astigmatism (Astigmatism, that is, the degree of focus in different directions is different. At this time, the electron beam The beam spot is non-circular), so a pair of control voltages are also involved, and the calibration results make the three optimal at the same time. For any one of the variables, the so-called optimum is to make its focus reach its peak value. Compared with the above-mentioned QAF, the search range of the control variable change is wider. The specific amount of focus is called Focus Metric, or FM for short. It usually measures certain features in the image such as edges (including edges extracted with different edge extraction operators/kernels and scales) to examine its amount. There are many algorithms. Because of it is an extensive quantity, the FM of different measurement objects is usually not comparable, and the results of different FM algorithms are not comparable. CAF is typically done not on a wafer but on a high-strength, durable metal target mounted on a mechanical motion platform, as shown as target 502 in Figure 5A. CAF is only system calibration, and when the equipment works on the wafer, it still needs to focus on the wafer surface. In addition, there is a rougher wafer height/Z-direction calibration (Calibration), and it can be assumed that there is no problem here.

Going back to the details of QAF, if the focus control variable is searched for the best FM in a large range, the result is shown in curve 503 . Searching in a large area is time-consuming, seriously affects the throughput of the equipment (Throughout), and is easy to damage the wafer. And if it is a small-scale search for the best FM, as shown in the curve 504, only a few points are needed, for example, at least 3 points are needed (corresponding to the same position on the wafer, and at least 3 frames of images under different control variables are collected). Determine the peak value of the curve, assuming that the system focus degree changes/drifts little, that is, obtain the control variable corresponding to the best focus degree, such as the coil current Im of the aforementioned electromagnetic objective lens. However, if the offset of the electron optical system (referring to the SEM system on the EBR device including software, hardware and firmware) is too large relative to the Im search step size, usually 3 points (corresponding to collecting 3 frames of images at the same position on the wafer) are not enough , which may not include the peak, and if the step size of the control variable changes is too large, the accuracy of the QAF result will be poor, and the system will fail to achieve the best focus.

The industry usually chooses to do QAF on the area containing X and Y direction features (edges) on the wafer. Referring to FIG. 5B, there are multiple dies 510, WA matching points (positions) 511 and 512, just like the matching position 411 in FIG. 4 , 412, QAF image acquisition position 515 or 516 (image acquisition position 517 in the figure is the position used for image grayscale optimization, ignore it for now), can use matching position 511 or 512 in WA to determine QAF (also includes The relative position (Offset) of image acquisition required by the IGO to be mentioned in this article.

The calculation of the focus degree FM mostly examines the image definition/sharpness (Sharpness), and also corresponds to the target shape, such as calculating the one-dimensional (1D) gradient/edge extraction of the image:

为卷积运算的符号，I为SEM图像。灰度图像梯度/边缘提取算法很多，不再赘述。Among them, ▽ represents the first-order gradient operation, or it can be the second-order Laplacian operation, or other gradient operations, and G is a Gaussian function.

is the symbol of convolution operation, and I is the SEM image. There are many grayscale image gradient/edge extraction algorithms, so I won't repeat them here.

In addition, because only QAF is involved in this article, for the sake of convenience, autofocus and fast autofocus will not be distinguished in the text in the future, and both refer to the aforementioned QAF.

Steps of image grayscale optimization

Before the EBR equipment re-inspects the defects on the wafer, image grayscale optimization (ImageGrayscale Optimization, referred to as IGO) should be performed, otherwise, referring to Figure 6A, taking the 8-bit (bit) image data type as an example, there will be The grayscale histogram of the SEM image is too narrow (such as curve 601), that is, the grayscale resolution is low, the high-end grayscale is saturated (such as curve 602), and the low-end grayscale is saturated (such as curve 603), which is not conducive to the detection of defects, namely It is difficult to detect defects and/or misjudgment is easy to occur, and the gray histogram of an ideal SEM image should be like the curve 604 . Usually, even for the same type of wafer, the gray distribution of the SEM images of different wafers at the same position may be different. The image grayscale optimization is not a simple software-only image histogram manipulation (Histogram Manipulation), because it will seriously affect the setting of the defect grayscale threshold and increase image noise. The industry usually optimizes the signal amplifier (refer to Figure 2B ) in the working parameters, such as gain and offset, because the gray value I of the final image pixel has such a relationship with the two:

I∝(gainV+offset)

Usually, the image grayscale optimization position is selected on the wafer, such as position 517 in FIG. 5, so that the SEM image collected thereon represents the material with the highest grayscale, followed by the material with the lowest grayscale, as shown in FIG. 6B, schematically The SEM image collected at one image acquisition position, which represents the region 621 with the highest gray level and the area 622 with the lowest gray level, is at the same image acquisition position on the wafer, otherwise two different image acquisition positions are required, this are all doable. Then set the gray level optimization target including the lowest gray level Lo, the highest gray level Ho, contrast Co=Ho–Lo. Then establish an evaluation function with the actual L, H, or/and C=HL measured in the image, for example, it can be w1(C–Co) ² +w2(H–Ho) ² +w3(L–Lo) ² , where the weight Factors w1, w2, w3 take values in [0,1], and their sum is 1. Among them, iterative regression methods such as gradient descent method are used to obtain the gain and offset values closest to the target. This is also a prior art, and will not be repeated here.

current issue

The above-mentioned methods are all prior art. There are the following obvious problems, especially in the preparation of the EBR equipment for its own work, including:

1. The preparations of the above-mentioned parts are independent of each other, which not only wastes time, but also leads to poor results due to disordered order. For example, QAF is sometimes done before SEM WA, sometimes after SEM WA, sometimes both before and after SEM WA, and sometimes not on the wafer, but on a special sample (Chip) on the mechanical motion platform Do QAF on the top, and then go back to the wafer to do it. In addition, the timing of IGOs is also uncertain and often unreasonable. In short, in the prior art, the above-mentioned preparation steps of each part are redundant and repeated, and the efficiency is low, which not only reduces the throughput of the equipment, but also often affects the working accuracy due to poor effect.

2. Especially when doing Quick Auto Focus (QAF), the existing technologies are all carried out on the area to be re-inspected on the wafer. During the process, multiple frames of images need to be collected. For example, QAF needs to collect at least 3 frames of images, and if the control variable changes If the step size is not selected properly (not a small probability event, because it is often only based on experience), more images need to be collected to judge the relative best focus condition. Therefore, it not only reduces the throughput of the equipment (Throughput), but also increases the risk of damage to the wafer (repeatedly scanning the polarizable dielectric material at a fixed position leads to local charging or charging);

3. In addition, there are other minor problems in the prior art, including: whether it is doing QAF or IGO, no attention is paid to avoiding defects, especially large defects that can interfere with the results, resulting in QAF or IGO in some cases IGO results are inaccurate, thereby affecting the accuracy of the EBR device's own work.

Contents of the invention

The object of the present invention is to provide a kind of re-inspection preparation method and re-inspection method of electron beam wafer defect re-inspection equipment, in order to solve the problems 1 and Question 2.

To achieve this purpose, the embodiments of the present invention adopt the following technical solutions:

On the one hand, a re-inspection preparation method for electron beam wafer defect re-inspection equipment is disclosed, including: after wafer loading and initial inspection information are obtained and wafer alignment under the optical microscope system is completed, scanning electron microscopy is performed Image grayscale optimization; in the implementation of the multi-level wafer alignment under the scanning electron microscope, it is judged whether to perform automatic focusing of the scanning electron microscope, and the wafer alignment or the automatic focusing of this level is completed according to the judgment result, and the level is completed. Wafer alignment and the rest of the more advanced wafer alignment; determine the reference point for the wafer coordinate system.

Further, the method includes the following steps executed in sequence:.

S1. Load the wafer and obtain initial inspection information;

S2. Complete the wafer alignment under the optical microscope system;

S3. Complete image grayscale optimization for the scanning electron microscope;

S4. Perform wafer alignment and focus inspection under the scanning electron microscope, including:

S4.1. Start the first-level wafer alignment under the scanning electron microscope;

S4.2, and check the focusing degree of the scanning electron microscope, judge whether to implement automatic focusing, implement automatic focusing or enter the next step according to the judgment result;

S4.3. Complete the first-level wafer alignment of the scanning electron microscope;

S4.4. Complete the remaining higher-level wafer alignment of the scanning electron microscope;

S5. Determine the reference point of the wafer coordinate system.

Further, when performing the image grayscale optimization and/or automatic focusing, the image is collected using a position with no defect or a defect size smaller than or equal to a preset size.

Further, in creating the wafer alignment work menu, not only the template used for SEM alignment and the matching position on the wafer, that is, the target image acquisition position, but also the complete template image containing the template are reserved.

Further, in step 4.1, the first target image is collected at the first matching position; in step 4.2, the focus degree inspection is started after the first target image is collected, and when the focus degree is checked, use The target image that is first successfully matched in wafer alignment and that overlaps with the template image meets the predetermined conditions is the focused working target image, and the focus degree of the overlapping area between the target image and the template image is compared, and according to the comparison The results and the established thresholds are used to judge whether to implement auto-focus.

Further, in step S4.4, use the same focus degree inspection method as step S4.2 to judge whether to implement the autofocus, implement autofocus or complete the current level of wafer alignment according to the judgment result.

Further, when a work menu is created, the main template image, the template image positioned at a set distance around the main template image, and the respective templates in the formed multi-frame template images are saved in the work menu; when the work menu is executed , in step 4.1, capture the first target image at the first matching position, in step 4.2, start the inspection of the focus degree after capturing the first target image, if the template in the main template image If the matching with the first target image fails or the overlapping area of the two does not meet the predetermined threshold condition, the template image around the main template image is used for matching with the first target image, and the successfully matched template image with the largest overlapping area is selected for comparison. It is judged whether to implement auto-focus according to the result of the comparison and the predetermined threshold for the focus degree of the overlapping area of the first target image.

Further, select a local area in the overlapping area, compare the focus degrees of the template image obtained in the local area and the target image, and judge whether to implement automatic focusing according to the comparison result and a predetermined threshold.

Further, the local area is a reduced overlapping area, so that the proportion of image features in the local area is higher.

Further, the local area is to select two independent areas in the overlapping area, respectively, for calculating the focus degrees Fx and Fy in the orthogonal x and y directions, respectively.

Further, the first focus ratio S1 is obtained based on the focus degree Fx1 of the template image in the x direction and the focus degree Fy1 in the y direction, and the first focus ratio S1 is obtained based on the focus degree Fx2 of the target image in the x direction and the focus degree Fy2 in the y direction. The second focus ratio S2 compares the first focus ratio S1 with the second focus ratio S2, and generates a warning message according to the comparison result.

Further, the automatic focusing includes: based on the template image, the target image that is successfully matched in the current-level wafer alignment of the scanning electron microscope and that overlaps with the template image satisfies a predetermined condition, that is, the focused working target image, the focused working target image The acquisition position on the wafer is the focus work position. Starting from the initial focus control variable i0 in the current wafer alignment work menu, change the value of the focus control variable according to the predetermined search step and collect 1 or more frames of search Image, compare the focus degree in the overlapping area between the template image and the search image, search for the corresponding focus degree control variable value when the comparison result satisfies the threshold condition, and use it to set the scanning electron microscope, so as to complete the Autofocus described above.

Further, use the focus difference Δf in the overlapping area of the template image and the successfully matched target image whose overlapping area satisfies the predetermined condition, and combine the empirical formula to determine the search step size, and use the search step size to search The best focus.

Further, start searching on the side of the initial focus control variable i0 according to a predetermined direction, change the control variable by one step, obtain the control variable i1, collect the corresponding first search image, and directly compare the first A search image and the focus degree of the template image in the overlapping area, if the threshold condition is met, the SEM is set with the control variable i1 to complete the automatic focus, otherwise it is determined based on the focus difference between the current target image and the first search image Change the direction of the control variable, and then change the focus control variable with a step length to obtain the control variable i2, collect the corresponding second search image, and directly compare it with the focus of the template image in the overlapping area, if the required The threshold condition is then set with the control variable i2 to complete the automatic focusing, otherwise use the initial focus control variable i0, focus control variable i1 and focus control variable i2 and the corresponding focus value to fit the secondary The focus degree control variable corresponding to the best focus degree is obtained from the curve, and the scanning electron microscope is set according to it to complete automatic focusing.

On the other hand, a re-inspection method of electron beam wafer defect re-inspection equipment is disclosed, which includes: using the above-mentioned method to prepare for defect re-inspection on a patterned wafer, and then performing the work of the re-inspection equipment.

Beneficial effect:

The embodiment of the present invention optimizes the EBR equipment re-inspection preparation process, intersperses the content of judging whether to execute fast auto-focus QAF into WA, removes the step disorder and redundancy in the prior art, and reduces the inappropriateness in the prior art. The risk of damage to the wafer caused by excessive collection of SEM images improves the efficiency and accuracy of the re-inspection preparation work, and also improves the throughput of the EBR equipment, which solves the above-mentioned problems 1 and 2.

In addition, when performing the image grayscale optimization and/or automatic focusing, the image is collected using a position with no defect or a defect size smaller than or equal to a preset size, so the embodiment of the present invention also solves the above-mentioned problem 3.

Description of drawings

FIG. 1 is a schematic diagram of a wafer to be re-inspected with initial inspection defects in the prior art;

2A is a schematic diagram of an example of electron beam scanning wafer defect re-inspection equipment in the prior art;

2B is a schematic structural diagram of an SEM system in an electron beam scanning wafer defect re-inspection device in the prior art;

3 is a schematic diagram of a template selection method in wafer alignment in the prior art;

FIG. 4A is a schematic diagram of matching positions in a certain level of wafer alignment in the prior art;

FIG. 4B is a schematic diagram of another matching position in a certain level of wafer alignment in the prior art;

FIG. 5A is a schematic diagram of the focusing principle and method of electron beam scanning wafer defect re-inspection equipment in the prior art;

FIG. 5B is a schematic diagram of selecting an image acquisition position for automatic focusing or image optimization on the wafer in the prior art;

6A is a schematic diagram of a problem gray histogram and a correct gray histogram of an SEM image in the prior art;

6B is a schematic diagram of selecting an image optimization position in a representative area on a wafer in the prior art;

Fig. 7 is the schematic diagram of the main re-inspection method of EBR equipment in the prior art;

FIG. 8 is a schematic diagram of the process of electron beam scanning wafer defect re-inspection equipment re-inspection preparation process in an embodiment of the present invention;

FIG. 9A is a schematic diagram of an overlapping area after matching a template image and a target image in an embodiment of the present invention;

9B is a schematic diagram of the overlapping area of the wafer alignment template image and the target image at different matching positions in an embodiment of the present invention;

FIG. 9C is a schematic diagram of further selecting and reducing sub-regions in the overlapping region in an embodiment of the present invention;

FIG. 9D is a schematic diagram of further selecting two independent sub-regions in the overlapping region in an embodiment of the present invention;

9E is a schematic diagram of collecting multiple frames of template images at multiple positions on the wafer in an embodiment of the present invention;

10A-10B are schematic diagrams of the relationship between focus control variables and focus in different situations in the prior art;

10C-FIG. 10F are schematic diagrams of the relationship between focus control variables and focus under different conditions in the embodiments of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the accompanying drawings.

The present invention all relates to the patterned wafer (Patterned Wafer) which is the most common in application. Wafer Alignment (WaferAlignment, referred to as WA) is the first step after the wafer is loaded, and it is a prerequisite for the work of the EBR equipment, and it is also the only way. Therefore, the main idea in the present invention is to use as much as possible WA, making the EBR equipment ready for its own work as much as possible during the implementation of WA. This saves time, increases device throughput, and guarantees accuracy.

At this time, the EBR equipment should be in a normal state, that is, each part has been calibrated, and its work menu (Recipe) state can be executed, and the specific work menu has been determined at this time, covering the preparation work of the equipment and the work of the job. step.

An embodiment of the present invention provides a re-inspection preparation method for electron beam wafer defect re-inspection equipment, including:

After the wafer is loaded and the initial inspection information is obtained and the wafer alignment under the optical microscope system is completed, the image grayscale of the scanning electron microscope is optimized; it is judged during the implementation of the multi-level wafer alignment under the scanning electron microscope Whether to perform automatic focusing of the scanning electron microscope, complete the wafer alignment of this level or the automatic focusing according to the judgment result, complete the wafer alignment of this level and other higher-level wafer alignments; determine the reference point of the wafer coordinate system.

In one embodiment, referring to FIG. 8 , the re-examination preparation method includes the following steps executed in sequence:

S1 loads the wafer and obtains initial inspection information.

The EBR equipment is placed on the mechanical motion platform 213 according to the procedure in the Recipe for the re-inspected wafer to be loaded. At the same time, the EBR equipment system software opens the initial inspection result file corresponding to the wafer, and obtains the initial inspection information, including related wafer information and initial inspection results, including the location of the initial inspection defect, which is required for subsequent re-inspection / own work, In one embodiment, the current preparatory work also needs to know the position of the defect in the initial inspection, because the image acquisition in the IGO and QAF needs to avoid the defect, mainly to avoid the area exceeding the predetermined threshold and the intensity exceeding the predetermined threshold. Interference of defects, specifically, when performing the image grayscale optimization and/or automatic focusing, images are collected at positions with no defects or defect sizes smaller than or equal to a preset size, so the embodiment of the present invention solves the above-mentioned problem 3.

S2 completes the wafer alignment under the optical microscope system and completes the OM WA.

The wafer alignment (Wafer Alignment, WA) of the EBR equipment all starts from OM WA. Different devices may include 1 or more stages of OM WA at different magnifications. In one embodiment, at least 1 stage and at most 3 stages, with 1 or 2 stages being most common. The OMWA process itself is the same as that in the prior art, as described in the background art, including in each level of OMWA, collecting the target image at the matching position on the wafer (ie, the target image acquisition position) to create the recipe Determine and save the template, perform template matching/searching, in one embodiment, fit a straight line with multiple successful matching positions, as shown in Figure 4A, finally determine the wafer orientation and correct it, such as rotating the mechanical movement platform, And pass the verification; In another embodiment, obtain the coordinate transformation matrix of the wafer coordinate system and the mechanical motion platform coordinate system according to the successful matching position, save the coordinate transformation matrix, so that the EBR equipment knows the current level of wafer alignment. Coordinate transformation matrix, EBR equipment can perform defect re-inspection based on the coordinate transformation matrix when performing its own work in the future.

S3 completes the image grayscale optimization of the scanning electron microscope to complete the IGO.

Since IGO does not have high requirements on image location, it can be done after OM WA. After the OM WA is completed, move the wafer to the predetermined IGO position in the Recipe, such as position 517 in Figure 5, which has a fixed relative relationship to the SEM WA position (for example, HMSEM WA position 511 in Figure 5, namely offset, where HM That is, high magnification) to collect images, first check whether the standard is met according to the established contrast and/or grayscale requirements (C, H, L requirements), if the standard is met, the IGO will be completed, and it will not take too much time, otherwise continue, method and background The method in the prior art described in the technology is the same, and also optimizes the gain and offset of the system amplifier circuit, so it is not repeated here. The difference is that, in an embodiment of the present invention, if the predetermined IGO position contains large defects obtained in the initial inspection, it should be avoided, and the same position in the adjacent die can be selected to collect images for IGO.

S4 performs wafer alignment and focus inspection under the scanning electron microscope, that is, a composite step of performing SEM WA and focus inspection.

It includes the following sub-steps:

S4.1 Start the first-level wafer alignment under the scanning electron microscope, that is, start LM SEM WA.

After all OM WA is completed, the system switches to the LM SEM image acquisition mode. Usually, SEM WA has multiple levels, the most common is 2 levels. For example, the primary level is a low magnification (Low Magnification, LM) SEM image, and the second level is High magnification (High Magnification, referred to as HM) SEM images, always first low and then high. The following take Level 2 SEM WA as an example, which is applicable to situations with more than Level 2. Referring to Fig. 4, in step S4.1, move the wafer to the first matching position 411 of LM SEM WA, start to collect the first target image of this level of WA at the first matching position 411, use the template reserved in the Recipe to The search is to perform template matching (note that unlike the prior art, in the embodiment of the present invention, the corresponding template image that generates the template is also retained in the Recipe, that is, not only the SEM Align the template used and the matching position on the wafer, that is, the target image acquisition position, and also retain the complete template image containing the template, which is useful later), as in the first matching position 411 in FIG. 4 Collect the first target image, but do not rush to collect the rest of the target images (because in the embodiment of the present invention, the focus inspection of the EBR device needs to be interspersed in WA, so as to simplify and optimize the process, improve efficiency and even the throughput of the entire device) to complete the current level of WA, but start from the current target image, such as the first target image collected from the first matching position 411, to perform QAF, that is, system focus-related work (including focus inspection and fast Autofocus (QAF), although ultimately still need to be the same as in the prior art, take a series of target images such as each matching position 412, 413, ... 416 in Fig. 4 to collect target images respectively, and perform template matching respectively, using Match the successful matching position and fit the straight line to complete the current WA.

S4.2 Check the focus of the scanning electron microscope, judge whether to implement QAF, implement QAF or enter the next step according to the judgment result.

In step S4.2 of an embodiment, the focus degree FM is checked with the current target image, for example, the first target image, that is, the focus degree check starts after the first target image is captured. It needs to be explained that if the first target image is not suitable at this time (for example, the matching of the template to it fails, or the relative displacement between the template image and the target image is too large and the overlapping area is too small to reach the predetermined threshold), it is used to check the focus or For automatic focusing, the direct countermeasure is to use the next target image, for example, to collect the target image at the second matching position 412 (also closest to the first matching position 411 and the center of the wafer), that is, the second target image, which can theoretically be sequentially It is the 2nd, 3rd... target image, until the area of the overlapping area reaches the predetermined requirement, that is to say, use (and only use) the target that is successfully matched first encountered in wafer alignment and whose overlapping area with the template image meets the predetermined conditions image, the target image is determined as the focus work target image, and the corresponding image acquisition position on the wafer is determined as the focus work position, and the focus degree of the overlapping area between the target image and the template image is compared, and according to the comparison The results and the established thresholds are used to determine whether to implement autofocus. In one embodiment, because the target image acquisition position will be properly adjusted according to the difference of the previous matching position, the matching result will not be bad (that is, the overlapping area between the template image and the target image will not be small), usually Only the second target image and the image acquisition position, ie, the second matching position, need to be used. It should be noted that here it seems that one more frame of image is collected, but because the image collection involved is still a necessary part of WA (just not the first frame of image) and because WA itself has a relatively high focus requirement as mentioned above Low, so there is no additional image acquisition in general, that is, no additional time-consuming. Of course, there are other countermeasures in the article, including preparing a multi-frame template image when creating a recipe, and using the one that matches the current target image with the largest overlapping area.

Supplementary explanation of the meaning of the content of the above overlapping areas. The overlapping area refers to the overlapping area between the template image and the target image when the matching positions of the template in the template image and the matching positions in the target image are superimposed when the template matching is successful. Assuming the position (x0, y0) of a template in the template image (whether using the upper left corner of the template or the center of the template), use the template to search for a target image to perform template matching, and find the best similarity threshold condition. Match the position (x1, y1), then there is a relative displacement between the template image and the target image (the relative rotation between the two images does not need to be considered in a single image of WA), dx=x1–x0, dy=y1-y0, assuming two The image size is the same, width and height are (W, H), and the area of the overlapping area is (W-dx)×(H-dy), as shown in the shaded part in Figure 9A, that is, between the template image 901 and the first target image 902 There is an overlapping area 903 . In the embodiment of the present invention, there are predetermined threshold requirements for the area of the overlapping region, including an absolute threshold and/or a relative threshold (a percentage of the original image area W×H, for example, 65%). Figure 9B shows the above-mentioned situation where the overlapping area between the template image and the first target image is insufficient, wherein 910 is the template image, 911 is the first target image, and the overlapping area 9101 of the two does not meet the predetermined threshold requirements, while the second The overlapping area 9102 of the target image 912 and the template image 910 meets the predetermined threshold requirements, and can be used for focus inspection or automatic focusing. At this time, the second target image 912 is the focus working target image, and its acquisition position is the second match in WA. The position is the focused working position. As mentioned above, in theory, subsequent target images in WA, such as the sixth target image 916, can also be used if the overlapping area 9106 of the template image 910 meets the predetermined threshold requirements, but it is not the preferred solution in practical applications, usually Not required.

It needs to be emphasized that when choosing to do QAF in a certain level of WA, there is only one matching position in the WA as the focusing work position, and only one WA target image is used as the focusing work target image, and no more are needed.

As mentioned in the background art, in performing QAF, a certain Focus Metrics (FM for short) algorithm is usually calculated and selected for a given region in a given image. The normal (the EBR equipment calibrated in time) system focus will not deviate seriously from the normal state, that is, there will be no serious drift, and there is no need for a large-scale search (like the curve 503 in Figure 5A), but fine-tuning is still required . However, in the embodiment of the present invention, unlike the prior art, at least 3 frames of images need to be collected to judge whether the focus is good or not, and the prior art is relative, and often requires more than 3 frames of images. However, in the embodiment of the present invention, relatively few images need to be collected, and each frame of image at the focus working position can be used to directly determine whether the focus is up to standard. Compared with the prior art, since there is no absolute reference, it is not known whether the FM corresponding to any frame of images is the best, and the best of them can only be determined when there are at least 3 frames of images. And I don't know how good it is. If the search step is too large, the best one can always be found at 3 points, but the accuracy may be poor, and it is difficult to achieve both.

In an embodiment, the specific content related to the focus inspection is introduced below.

First of all, in the embodiment of the present invention, it is useful to not only keep the template itself but also keep the entire template image in the WA Recipe when creating the WA Recipe. For example, in FIG. 3 , not only the template 301 but also the entire template image 300 are kept. Of course, the matching position is the target image acquisition position, for example, the matching positions 411-416 in FIG. 4 should have been kept in the recipe. The matching position 411 or 412 is usually the original template image acquisition position when the recipe is created (it corresponds to the target image acquisition position when the recipe is executed). Usually it should be as close to the center of the wafer as possible. In addition, when creating a recipe, the focus of the template image is the best, whether it is through automatic focus or manual focus, it can be guaranteed and is necessary.

When it’s time to check the focus (note, before implementing WA in LM SEM WA, the image contrast has been checked, regardless of whether IGO has been done before), move the wafer to the first matching position (for example, the first matching position 411) Collecting the first target image and performing template matching is the first step in WA. It needs to be added that the template matching algorithm in the present invention allows a certain degree of out-of-focus, that is, Defocus/Blurring, for the image used by LM SEM WA at this time, that is, its own (positioning work) does not have particularly strict requirements on the image focus degree, usually The requirements for focusing degree are much lower than that of the EBR equipment for its own work. In the existing image processing/machine vision technology, there are many techniques for template matching between images with a limited degree of Blurring, such as some algorithms based on out-of-focus models or symmetry of basic features in images, and some commercial software can also support focusing A function for template matching between images with varying degrees of variation. Before going back to the supplementary explanation about template matching in the case of out-of-focus, in the step of checking the focus, use the template saved in the recipe, such as template 301 in FIG. 3, to the same position on the current wafer when the recipe is currently executed, for example The first matching position 411 in FIG. 4 (or similar positions, that is, other positions that can be used as matching points required by the current WA) collects the first target image of the current WA. As mentioned above, this position has been reserved in the recipe, which is usually the first matching position (for example, the first matching position 411 in FIG. 4 ). Then use the template to perform template matching/searching in the first target image, usually with a high probability of success, and in case of special accidents, such as damage to the part of the wafer that makes the matching unsuccessful, you can still use the template in the remaining matching positions such as Fig. Matching positions 412, 413, . . . , 416 in 4 (in an embodiment, positions that differ by an integer multiple of the grain period are acceptable, but those close to the center of the wafer should be preferred) to collect target images for the matching. Referring to FIG. 9A, the matching result gives the translation between the template image and the target image (because the wafer is pre-aligned through the aforementioned, the relative rotation between the two single-frame images is negligible), the template image 901 and the target image 902 Overlap area 903 . Then the focus degree (FocusMetric, FM for short) of the image is calculated only from the overlapping area 903 . At this time, it also includes calculating the 1D focus degree Fx, Fy in the X and Y directions respectively or calculating the 2D focus degree F based on the two. As a supplementary note, although the method of obtaining FM is the same as that in the prior art, the meaning here is different. Because the FM value is relative, there is no absolute standard, and the same target object can only give the highest FM among them. Therefore, in the prior art, at least 3 frames of images are required to give the highest FM among them, and it cannot be determined that the result is the best. If the control variable search step is too large, the accuracy of the result will be poor. However, in the embodiment of the present invention, since the template image is focused (whether automatically or manually) when creating the recipe, corresponding to the best focus degree, an absolute reference is provided. In the embodiment of the present invention, since the template image and the target image are not completely overlapped, the objects involved in the calculation of FM are not completely the same, and are not completely comparable. Therefore, it is necessary to use the range of the overlapping area to calculate the The focus degrees of the template image and the target image (for example, the first target image), so that both (Fx1, Fy1) and (Fx2, Fy2) of 1D and F1, F2 of 2D are comparable. For images using WA, since the template matching itself has requirements for the characteristics of the X and Y directions, it can also guarantee the calculation of FM in the X and Y directions, that is, Fx and Fy. Here (Fx1, Fy1) and F1 can represent the 1D and 2D focus degrees of the template image within the above overlapping range, and (Fx2, Fy2) and F2 can represent the 1D and 2D focusing degrees of the target image matching the template image within the above overlapping range. 2D focus. The focus degrees are all represented by a predetermined FM. Because the template image when creating the recipe is clear, its focus (Fx1, Fy1) or F1 is a good reference (standard), and at this time, the same area (limited to the overlapping area) in the target image when the recipe is executed (Fx2 , Fy2) or F2 can be directly compared with (Fx1, Fy1) or F1, no matter what FM algorithm is used, then according to the comparison result, it can be determined whether the current focus is qualified, that is, whether its change (variation) is within Within the predetermined threshold range (the depth of focus of the SEM system has a certain range, within which it is considered to be up to standard). Considering that different FM algorithms have different results for the same image, the threshold can use a relatively variable threshold TFr to examine changes in the focus F

And the change of the same 1D focus Fx, Fy

In this way, Fx, Fy, and F can share a threshold TFr, which is dimensionless; in addition, the above-mentioned overlapping area can be further optimized, including appropriately reducing it to obtain a reduced overlapping area, that is, a sub-area in the original overlapping area, as shown in Figure 9C The overlapping area 903 is reduced to the overlapping area 904, so that the relative ratio/concentration of features (edges, corners) 909 is higher. It should be added that even if the same position on the wafer (such as 515 in Figure 5) is used to repeatedly capture multiple frames (usually more than 3 frames) to obtain FM for comparison in the prior art, the area for calculating FM should be optimized instead of using the entire image . As mentioned above, in case the template matching between the template in WA and the first target image fails or the overlapping area between the template image and the first target image does not meet the established threshold requirements, the countermeasures include using subsequent target images to finally determine the focus work The target image and the corresponding focusing working position are used for subsequent QAF.

Optionally, the embodiment of the present invention further includes optimization of the overlapped areas satisfying the conditions, and selecting a local area therefrom for use in the focus degree calculation during the focus degree inspection/autofocus. In one embodiment, a local area is selected in the overlapping area, the focus degrees of the template image obtained in the local area and the target image are compared, and whether to implement automatic focusing is determined according to the comparison result and a predetermined threshold. One of the methods, the local area is a reduced overlapping area, so that the proportion of image features in this local area is higher, specifically, referring to image 9C, and further selecting a local area 904 in the overlapping area 903, wherein 909 is an image feature Another one of the optimal selection method, the local area is to select two independent areas (i.e. two independent sub-areas) in the overlapped area respectively, which are used to calculate the focus of the orthogonal x and y directions respectively Degree Fx, Fy, with reference to Fig. 9D, in above-mentioned overlap area 903, further determine the independent measurement area 905 and 906 of X and Y direction focus degree Fx and Fy (that is two independent sub-areas 905 and 906, relatively described overlap area , such as relative to its center or upper left corner position, has a fixed displacement relationship), which are used to calculate Fx and Fy respectively. Its purpose is to make the proportion of the features contained in the independent measurement areas 905 and 906 used to calculate Fx/Fy higher than that of the entire overlapping area, so that the data accuracy of Fx and Fy is higher; the two can also overlap, The width and height of the area 905 used to calculate the focus degree Fx are Dxw and Dxh respectively, and the width and height of the area 906 used to calculate the focus degree Fy are Dyw and Dyh respectively, both of which include vertical/Y and horizontal/X in the image The basic principle of selecting the independent area 905/906 for calculating Fx/Fy is to make the features in the X/Y direction (mainly the edge in the image, that is, the area with a large grayscale change) account for a larger proportion or discard it. Some areas that contribute little to Fx/Fy.

Optionally, in an embodiment, it can be further improved. In the calculation of the above-mentioned focus degree, no matter whether the optimized overlapping area is used or not and what kind of optimization method (reduced or independent in the X and Y directions), when the obtained The ratio of Fx and Fy S=Fx/Fy or S=Fy/Fx (as long as one of them is selected, only S=Fy/Fx will be used as an example in the future), the ratio of the degree of focus obtained in the template image S1 (Fy1/Fx1) is compared with S2 (Fy2/Fx2) obtained in the target image, if the difference between the two exceeds a predetermined threshold, even if the respective changes of Fx and Fy have not exceeded the predetermined threshold, report to the system , that is, a warning is sent to the system to remind the system to perform relevant calibration (Calibration) in time, such as calibrating system astigmatism (Astigmatism), because the astigmatism itself is a system problem that has nothing to do with the wafer and cannot be solved by QAF. Specifically, the first focus ratio S1 is obtained based on the focus degree Fx1 of the template image in the x direction and the focus degree Fy1 in the y direction, and based on the focus degree Fx2 of the target image in the x direction and the focus degree Fy2 in the y direction A second focus ratio S2 is obtained, the first focus ratio S1 is compared with the second focus ratio S2, and a warning message is generated according to the comparison result. As a supplementary note, the S=Fy/Fx obtained in different areas are not comparable, but in the same area, for example, the S obtained in the above overlapping area is comparable, and there can be a predetermined relative threshold TSr as above

In an embodiment, if the difference between Fx and Fy of the overlapping area is too large, so that the ratio of the two, that is, the ratio S of the focus degree, deviates too much from 1, the algorithm can choose to automatically adjust the shape of the area for calculating Fx, Fy appropriately, For example, appropriately change (increase or decrease) the overlapping area 903 in one direction, or change the widths of the independent focus measurement areas 905, 906, mainly Dxh and Dyw, which can be easily realized in software when the overlapping area is known. As above, even if the same position on the wafer (such as the image acquisition position 515 in Figure 5B) is used to repeatedly capture multiple frames (usually more than 3 frames) to obtain FM for comparison in the prior art, the area for calculating FM should be optimized instead of Use the entire graph.

Optionally, there can be more than one template in the template image (do it when creating a recipe), which is helpful for successful matching when the image quality is relatively poor. Of course, this is in the LM SEM WA image relatively rare. In addition, there may be more than one template image collection position to collect multiple template images. In such a plurality of template images, there are m templates in total, m>1. When creating a work menu, save the main template image, the template image positioned at the peripheral setting distance of the main template image and the respective templates in the multi-frame template images formed in the work menu; when executing the work menu, in step In 4.1, the first target image is collected at the first matching position. In step 4.2, the focus degree inspection is started after the first target image is collected. If the matching of the target image fails or the overlapping area of the two does not meet the predetermined threshold condition, use the template image around the main template image and the first target image for matching, and select the successfully matched template image with the largest overlapping area (of course, in the subsequent WA This template is also used for all matching), compare the focus degree of the overlapping area with the first target image, and judge whether to implement auto-focus according to the result of the comparison and a predetermined threshold. At this time, the focus work target image is the first target image in WA, and the focus work position is the first matching position in WA. Referring to Fig. 9E, for example, there are multiple template image acquisition positions t1-t9 (not shown in the figure) in the shape of overlapping nine-square grid 920, and the corresponding main template image T1 is collected at the position t1 in the center (the solid line part in the center of the figure), which is For the original template image, collect the remaining corresponding template images T2-T9 (indicated by dotted lines in the figure) at the peripheral positions t2-t9 of t1. The collected images T2-T9 partially overlap with the main template image T1 respectively, and are not spliced. Select m in total. After the template (931 in the figure), all (referring to the template image and template) are saved in the recipe. This is helpful for successful matching when the image quality is relatively poor, and only with successful matching can there be the overlapping region. Of course this is also rare when LM SEM WA images. In addition, it is not particularly important to spend more time when creating a recipe (collecting multiple frames of template images), but what is important is the time when the recipe is executed, which is related to the throughput index of the device.

Optionally, in one embodiment, when comparing the overlapping area of the selected template image and the target image for focus degree comparison, if a certain target image has a large defect in the overlapping area (during the re-inspection, each Defect position and size, as a reference), avoid using the target image, for example, take the target image of the same position on the adjacent and unused die instead. In addition, for any area where IGO or QAF needs to be done outside WA, the position of the large defect given in the initial inspection result should be avoided (the simple method is to change the same position on the adjacent die).

Then, based on the inspection results, the following steps are decided, including the implementation of QAF:

If the focus degree inspection is qualified, in the embodiment of the present invention, there is no need to perform any fast automatic focus QAF, and the subsequent wafer alignment is directly carried out, that is, the target image is collected at the subsequent matching position, and the current stage is completed, that is, LM SEM WA, and the existing The same as in the technology, go to the matching positions 411,...416 to collect target images for template matching, and use the matching positions to fit the straight line. I won’t repeat it here, and complete the more advanced HM SEM WA, that is, all levels of WA. Thereafter, under normal circumstances, QAF does not need to be performed on the wafer to be tested.

Otherwise, QAF is performed in the current LM SEM state. The automatic focusing includes: based on the template image, the target image that is successfully matched in the current wafer alignment of the scanning electron microscope and meets the predetermined conditions in the overlapping area with the template image is the focusing work target image, and the focus work target image is on the wafer The acquisition position above is the focus working position. Starting from the initial focus control variable i0 in the current wafer alignment work menu, change the value of the focus control variable according to the predetermined search step and collect one or more frames of search images. Comparing the focus degree of the overlapping area of the template image and the search image, searching for the corresponding focus degree control variable value when the comparison result satisfies the threshold condition, and using it to set the scanning electron microscope, thereby completing the automatic focusing . In the embodiment of the present invention, it is not necessary to collect images in a predetermined area on the wafer for QAF as in the prior art, but to directly use it at the position where the WA stays, that is, the focusing work position. This can not only save steps including mechanical movement, but also the WA image acquisition area ensures that there are enough features in the image in the X, Y direction to be used for the calculation of Fx, Fy, which can be located at the first target image position (usually that is to create the Recipe When the template image is collected, as mentioned above, in case the first target image is not suitable, the second target image can still be used, and its large overlapping area can be guaranteed, and it is also a necessary step in WA. The above-mentioned preparatory work does not increase additional time-consuming as a whole, and there is also the option of the above-mentioned multiple templates (fixed at the first matching position as the focus working position).

And in the embodiment of the present invention, when doing QAF at a predetermined position, unlike the prior art, the fixed search step size (which is inefficient and imprecise) is not used for the control variable, but according to the focus degree FM obtained in the overlapping area It is determined by comparison, which is more accurate and can support subsequent optimization/efficient search for focus control variables. For example, use the focus difference Δf in the overlapping area of the template image and the successfully matched target image whose overlapping area satisfies the predetermined condition, and combine the empirical formula to determine the search step size, and use the search step size to search for all the best focus. With reference to Fig. 10A-Fig. 10F, curve 1010 in Fig. 10A is the situation that the step size of control variable Im is too large in the prior art, and wherein i0 is the current focus degree control variable value that is to collect described template image when creating Recipe (guarantee The value of the control variable when the focus is up to standard/best), but now it is not up to standard after the previous focus test (either due to the drift of the system itself or the change of the working distance from the wafer surface to the SEM system, there is still a step in the actual application. grows more or worse), ih is the optimal position of the actual focus to be found, Δi is the fixed step size of the control variable, i1, i2, i3 are the values of the control variable in the interval Δi, and need to be in these The image obtained under the value is searched for the current best position ih; the curve 1020 is the situation where the step size of the control variable Im is too small in the prior art, and the value of the control variable between i0 and i1 and between i0 and i2 are all step sizes Δi, so that more frames of images need to be collected. Although the optimal position ih can always be approached in the end, too long time consumption will not only cause the throughput of the EBR equipment to exceed the standard, but also may damage the wafer part. In the embodiment of the present invention, based on conditions such as also benefiting from the preservation of the template image (not just the template itself), the focus of the known template image is up to standard and optimal for the target/pattern on a given wafer, the WA template is used The overlapping area generated by the matching of the image and the target image makes the focus of the template image and the target image comparable within the overlapping area. The following methods can be used to 1) obtain a suitable/more accurate search step size, and further 2) support optimized The search for the best control variable.

Referring to the curve 1030 in FIG. 10C , in the embodiment of the present invention, the appropriate search step size Δi must first be determined, and the focus degree fm of the template image in the overlapping area represents the best in history, but its corresponding control variable is the current acquisition target The control variable i0 used in the image is no longer optimal. It gives the FM result f0 in the overlapping area, and the difference between the two is Δf=fm–f0. At this time, according to a mapping relationship based on experiments and experience, For example, a look up table (LUT for short) in the form of discrete data can obtain a suitable control variable step size Δi=LUT(Δf). In addition, because the result of FM obtained by changing the control variable in a small range is close to a quadratic curve (note: the curves in Figure 10A-Figure 10F are only schematic, not strictly quadratic), it can also be determined based on experiments and experience The parameters of the quadratic function (Func), then Δi=Func(Δf).

In an embodiment of the present invention, the obtained more accurate step size Δi of the control variable is continued to support subsequent optimization searches. Considering that if the FM of the target image (in the overlapping area) is close enough to the template image, that is, the difference is within the above threshold, it will pass directly (will not go here) in the previous inspection of the focus degree, that is, i0 is still suitable. Acquire any additional images; otherwise stop here and need to do QAF. In one embodiment, search is started on the side of the initial focus control variable i0 according to a predetermined direction, and the control variable is changed by one step to obtain the control variable i1, and the corresponding first search image is collected for direct comparison The focus degree of the first search image and the template image in the overlapping area, if the threshold condition is satisfied, the scanning electron microscope is set with the control variable i1 to complete the automatic focusing, otherwise based on the focus degree of the current target image and the first search image difference to determine the direction in which the control variable changes, and then change the focus control variable by one step to obtain the control variable i2, collect the corresponding second search image, and directly compare it with the focus of the template image in the overlapping area, If the threshold condition is met, the SEM is set with the control variable i2 to complete the automatic focusing; otherwise, the initial focus control variable i0, the focus control variable i1, the focus control variable i2 and the corresponding focus value are used to simulate Combining the quadratic curve to obtain the focus degree control variable corresponding to the best focus degree, and using it to set the scanning electron microscope to complete automatic focusing.

Specifically, the steps of the QAF process include: 1) After obtaining the appropriate control variable search step Δi by the above method, search for i0-Δi and i0+Δi on both sides of the current control variable i0, and start with a fixed direction during specific implementation. If it is fixed in the i0-Δi direction, in the best case, a control variable change/image acquisition will find around ih (as mentioned earlier, the focal depth of the SEM system has a certain range, and it is considered to be up to standard in it), at the focus working position The image is collected, and its FM directly meets the standard (that is, the difference between the FM of the template image and the template image is within the above-mentioned threshold range. Note that it is limited to the overlapping area, and the above-mentioned threshold is, for example, a relative threshold), so QAF is completed; At the beginning of Δi, the result is much worse, that is, the above-mentioned FM obtained is smaller than f0, then go to the opposite direction, that is, i0+Δi direction (it is stipulated in the present invention, always change the control variable Δi distance to the direction with a higher degree of focus between the two ), at this time near ih, collect the image at the focus working position to obtain the FM in the overlapping area with the template image, if the FM directly meets the standard, then it is the same as in 1), and QAF is also completed, and there are 2 image acquisitions at this time , this situation is shown in the curve 1040 in Fig. 10D, its step size Δ1 in the left/smaller direction of i0 is Δi, and the step size Δ2 on the right side can also be Δi (but it can also be based on Δi based on the previous FM comparison results are fine-tuned appropriately); 3) If the FM results given by the control variables on the left and right sides of i0 are not directly up to the standard (the difference with the template image in the overlapping area exceeds the above relative threshold), these 3 points can be used, As shown in Fig. 10E curve 1050 in i0, i1, the result of i2 point fits a quadratic curve to obtain optimal control variable ih, according to the above-mentioned control variable step size determination method and search method in an embodiment of the present invention, above-mentioned 3 points Can guarantee the best focus must be among them. In addition, it should be noted that at this time, the three frames of images are all from the same position on the wafer, that is, the focusing work position, so it is not necessary to use overlapping regions when comparing the degree of focus. In one embodiment, of course, sub-regions are used, such as appropriately reducing the calculation of FM. It is also feasible to increase the proportion of features in the area, and then use it to set the SEM system to complete QAF. In addition, it is assumed that i0 is on the left side of ih (the search starts from a fixed direction, always from the left side). In order to illustrate that the method is applicable to various situations, the curve 1060 in FIG. The situation on the right side of ih (the search method is fixed, it still starts from a fixed direction, always starts from the left side, and the focus control variable changes direction to the direction of i0-Δi and i0 corresponding to the relatively large focus degree, in In Figure 10F, Δi is represented by Δ1), the results of the position il of i0-Δi on the left side and the position ir of i0+Δi on the right side can fit the results of 3 points as above and fit a quadratic curve to obtain the best To control the variable ih, and then complete QAF, it can also be based on the current situation, that is, when the FM given at il is better than the FM value corresponding to the current i0, then the better side can be separated by a step, that is, i0-2Δi The image is collected at position il2, and the results of three points i0-2Δi, i0-Δi, and i0 can also fit a quadratic curve to obtain the optimal control variable ih. It can be seen that in the above situations, two additional frames of images usually need to be collected, and it is best to complete it at one time, which is obviously superior to the prior art. In terms of probability and statistics, the efficiency of the method in an embodiment of the present invention is more than 70% higher than that in the prior art, and the accuracy of the result is higher. It needs to be added that, regarding the search method, various examples are given in one embodiment of the present invention but no further limitation is given. Based on the step size obtained by the method in one embodiment of the present invention, there can be many search methods, all of which belong to the present invention. within the scope of the invention.

In this way, QAF is completed in the whole necessary LM SEM WA process, without any additional image acquisition, saving the time of QAF.

S4.3 Completion of the first-level wafer alignment of the scanning electron microscope is the completion of LM SEM WA.

The steps required for LM SEM WA are the same as those in the prior art, except that the preparation work of the EBR equipment is interspersed, including checking the required focus and performing QAF if necessary. After the LM SEM WA is completed, switch from the LM SEM WA to the HM SEM WA, mainly the switching of the working parameters in the electron optical system 211, which are the same as those in the prior art, and then go to the next step.

S4.4 Complete the rest of the more advanced wafer alignment for the SEM.

In one embodiment, in step S4.4, HM SEM WA is completed.

Specifically, HM SEMWA is performed after the system switches to the HM SEM image acquisition mode after LM SEM WA is completed. It is the same as in the prior art. It also moves the wafer to the first matching position of the HM SEM WA on the wafer (the WA at each level in the Recipe has its own template and matching position), and starts to collect the first target of the WA at this level. Image, such as 411 in Figure 4A, but when necessary, you can choose the same as LM SEM WA, do not rush to collect the rest of the target image, but start QAF, including checking the focus status and implementing QAF if necessary, also need The preparation part of the EBR equipment is interspersed, although ultimately it is necessary to take a series of target images as in the prior art, for example, to collect target images at matching positions 412, 413, . . . 416 in FIG. 4 . That is, in one embodiment, in step S4.4, use the same focus inspection method as step S4.2 to judge whether to implement the autofocus, implement autofocus or complete the current level of wafer alignment according to the judgment result . But for many similar equipment, due to the LM SEM QAF, when it comes to the HM SEM, there is no need for QAF, and the HM SEM WA can be completed directly. If there is a choice to do QAF in LM or HM WA, you should choose to do it in LM SEMWA, because the features in LM SEM images are more abundant, it is more suitable for QAF, and the FOV is larger, the energy of the electron beam is relatively less concentrated, and it is easier to Not easy to damage the wafer surface.

S5 determines the reference point of the wafer coordinate system.

So far, the EBR equipment has completed all the WA, and at the same time completed all the preparations required by the EBR equipment for its own work, including the required SEM image contrast and focus reaching the established requirements. Then, the wafer coordinate reference point can be further determined as required, and the method is the same as that in the prior art, and will not be repeated here.

In this way, the preparatory work for the re-inspection of the EBR equipment is completed.

Next, the device will perform its own work in accordance with the established content in the recipe. The job of the EBR equipment is the core re-inspection work, and the content about the re-inspection method will be introduced below.

At present, wafer re-inspection objects are mainly graphics wafers, which can be roughly divided into two categories, logic circuits such as CPU, DSP, GPU, etc., and memory such as DRAM, NAND, etc., usually have different re-inspection methods. Carried out after the EBR tool is loaded and aligned on the wafer. At present, the EBR re-inspection methods in the industry mainly include Cell-to-cell (CTC) detection method for memory wafers, and Die-to-Die (DTD), Die-to-database (DTDB) and other detection methods for logic circuits. class wafer.

DTD method

Referring to FIG. 7 , the die-to-die (DTD) re-inspection method is the most commonly used, and mainly includes collecting the image 701 to be re-inspected at the position of the defect to be re-inspected on the wafer 700, and the image 701 to be re-inspected is collected at a distance of dr (dr≥1 ) Die) collect reference image 702 at positions without initial detection defects, and use corresponding image processing algorithms to compare the two to detect defects.

CTC method

Continuing to refer to FIG. 7 , the cell-to-cell (CTC) re-inspection method mainly includes collecting a single frame of an image to be re-inspected 703 at the defect position to be re-inspected, i.e. the initial inspection, under an appropriate field of view/FOV. For the periodicity of the memory unit (cell), compare the image difference of adjacent cells (there can be multiple specific algorithms) and use the corresponding image processing algorithm to detect defects. Since the comparison objects (different cells) come from the same frame of images, the positions are very close, and the algorithm is relatively less disturbed (such as the thickness of the wafer and its surface film layer, the difference in the sampling position of the wafer, etc.), usually when encountering an available CTC, try to May be used.

DTDB method

Continuing to refer to FIG. 7 , the die-to-database (DTDB) re-inspection method includes collecting an image 705 to be re-inspected at the position of the defect to be re-inspected, that is, the initial inspection, but unlike the reference image collected in DTD, it is designed by IC The synthesized image is used as a synthesized reference image, such as the synthesized reference image 706 in the figure, and then the same method as DTD is used to detect defects. The IC design information usually comes from Graphic Design System, referred to as GDS file, currently in GDSII version format. Usually, the IC circuit analyzed by GDS needs to be converted into an image or image profile close to the SEM image under the imaging conditions of the system through image processing. Then it is compared with the reference image to be re-inspected and synthesized, and the defect is detected with an algorithm similar to DTD.

Post-processing

The results of the various inspections include a defect image (Difference Image), such as the difference between the image to be inspected in the DTD and the aligned reference image, in which there are pixels of suspected defects whose intensity reaches or exceeds a predetermined threshold, and then post-processed (Post Processing), so that the defect pixels are connected to form a defect area (Blob), and its basic characteristics include position, boundary, circumscribed rectangle, number of pixels, etc. The detection usually uses a high magnification (High Magnification, HM for short) SEM image, and of course a corresponding WA precision is required to ensure the precision of the position to be re-examined. Regardless of which of the above methods is used to complete the detection of a single defect to be re-inspected/initially inspected, a SEM image with a higher magnification (Ultra High Magnification, referred to as UHM) is usually collected, and then the defect features are extracted from it and the defect is classified. For defect re-inspection, the EBR equipment continues to re-inspect the next defect to be re-inspected. The intensity may be the above-mentioned difference image (Difference Image), for example, the absolute value of pixels of the difference between images from two Dies or images from two Cells after alignment.

In this embodiment, the preparation method for re-inspection described above is used to prepare for defect re-inspection on the patterned wafer, and then perform the work of the re-inspection equipment. The specific steps of the re-examination method include:

a. Switch the SEM system to the appropriate magnification. It is best to choose the same magnification/FOV as the HM WA for the detection equipment, which is much more convenient;

b. Execute at least one of the above-mentioned CTC, DTD, and DTDB defect detection methods, visit each defect location to be re-inspected one by one, collect images, and perform re-inspection with the above-mentioned method;

c. The above-mentioned re-inspection includes post-processing of defective pixels, including filtering, connecting, screening, and integrating them;

d. For each re-inspection, collect higher-resolution SEM images of major defects in the image for defect classification (usually with pre-trained models, including traditional Machine Learning (ML) based on selected feature extraction method and/or deep learning (Deep Learning, referred to as DL) method, there are many in the prior art, and will not be described in detail;

e. End the re-inspection of the wafer, save the test result, unload the wafer, and prepare for the re-inspection of the next wafer.

So far, it can be seen that the method in the above-mentioned embodiment of the present invention optimizes the EBR equipment re-inspection preparation process as a whole, especially intersperses the judgment whether to perform fast auto-focus QAF in WA, and eliminates the step disorder in the prior art and redundancy, reducing the risk of damage to the wafer caused by improperly collecting too many SEM images in the prior art, improving the efficiency and accuracy of the re-inspection preparation work, and also improving the throughput of the equipment, that is, solving Question 1 and Question 2 above.

In addition, when performing the image grayscale optimization and/or automatic focusing, the image is collected using a position with no defect or a defect size smaller than or equal to a preset size, so the embodiment of the present invention also solves the above-mentioned problem 3.

The above embodiments only set forth the basic principles and characteristics of the present invention. The present invention is not limited by the above embodiments. Without departing from the spirit and scope of the present invention, based on the embodiments of the present invention, those of ordinary skill in the art will All other implementation modes obtained under the premise of making creative work belong to the scope of protection of the present invention.

Claims (15)

1. A rechecking preparation method of an electron beam wafer defect rechecking device is characterized by comprising the following steps: after the wafer is loaded, initial inspection information is acquired, and the wafer alignment under an optical microscope system is completed, carrying out image gray level optimization of a scanning electron microscope; judging whether to carry out automatic focusing of the scanning electron microscope in the implementation of multi-level wafer alignment under the scanning electron microscope, finishing the level wafer alignment or the automatic focusing according to the judgment result, and finishing the level wafer alignment and the alignment of other higher level wafers; a reference point of the wafer coordinate system is determined.

2. The method according to claim 1, characterized in that it comprises the following steps performed in sequence: .

S1, loading a wafer and acquiring initial inspection information;

s2, completing wafer alignment under an optical microscope system;

s3, completing image gray level optimization on the scanning electron microscope;

s4, carrying out wafer alignment and focusing degree inspection under the scanning electron microscope, comprising the following steps:

s4.1, starting to align the first-stage wafer under the scanning electron microscope;

s4.2, checking the focusing degree of the scanning electron microscope in the scanning electron microscope, judging whether to implement automatic focusing,

implementing automatic focusing or entering the next step according to the judgment result;

s4.3, finishing the first-stage wafer alignment of the scanning electron microscope;

s4.4, finishing the alignment of the rest higher-level wafers of the scanning electron microscope;

and S5, determining a reference point of a wafer coordinate system.

3. The method according to claim 1, wherein when performing the image gray scale optimization and/or auto-focusing, an image is acquired using a position where no defect or a defect size is equal to or less than a preset size.

4. The method according to claim 2, characterized in that not only the template used for scanning electron microscope alignment and the matching position on the wafer, i.e. the target image capture position, but also the complete template image containing the template are retained in the creation of the wafer alignment work menu.

5. The method according to claim 2, characterized in that in step 4.1, a first target image is acquired at a first matching position; in step 4.2, the test of the focusing degree is started after the first target image is collected, when the focusing degree is tested, the focusing degree of the overlapping area of the target image and the template image is compared by using a focusing working target image which is the target image which is successfully matched with the template image in the wafer alignment and satisfies the set conditions, and whether to implement automatic focusing is judged according to the comparison result and the set threshold value.

6. The method of claim 5, wherein in step S4.4, the same test method of the degree of focus as that in step S4.2 is used to determine whether to perform the auto-focusing, and based on the determination result, the auto-focusing is performed or the current stage of wafer alignment is completed.

7. The method according to claim 2, wherein when creating a work menu, respective templates of the master template image, the template images located at a set distance from the periphery of the master template image, and the formed multi-frame template images are saved in the work menu; when the work menu is executed, a first target image is collected at a first matching position in step 4.1, the test of the focusing degree is started after the first target image is collected in step 4.2, if the matching of the template in the main template image and the first target image fails or the overlapping area of the template and the first target image does not meet the condition of a set threshold value, the template image at the periphery of the main template image and the first target image are used for matching, the successfully matched template image with the largest overlapping area is selected, the focusing degree of the overlapping area of the successfully matched template image and the first target image is compared, and whether the automatic focusing is implemented or not is judged according to the comparison result and the set threshold value.

8. The method according to claim 5 or 6, wherein a local region is selected in the overlap region, the degree of focusing of the template image and the target image obtained in the local region is compared, and whether or not to perform autofocus is determined based on the result of the comparison and a predetermined threshold.

9. The method of claim 8, wherein the local region is a reduced overlap region such that the image feature ratio of the local region is higher.

10. The method of claim 8, wherein the local area is two separate areas selected from the overlapping area for calculating the focus Fx, fy in orthogonal x, y directions.

11. The method according to claim 5, wherein a first focusing power ratio S1 is obtained based on the focusing power Fx1 of the template image in the x direction and the focusing power Fy1 in the y direction, a second focusing power ratio S2 is obtained based on the focusing power Fx2 of the target image in the x direction and the focusing power Fy2 in the y direction, the first focusing power ratio S1 and the second focusing power ratio S2 are compared, and warning information is generated according to the comparison result.

12. The method of claim 5, wherein the autofocusing comprises: based on template image, target image which is successfully matched in current-level wafer alignment of scanning electron microscope and meets set conditions with the overlapping area of the template image, namely the focusing work target image, and the collection position of the focusing work target image on the wafer, namely the focusing work position, taking an initial focusing degree control variable i0 in a current execution wafer alignment work menu as a starting point, changing a focusing degree control variable value according to a set search step length and collecting 1 or more frames of search images, comparing the focusing degree in the overlapping area of the template image and the search images, searching the corresponding focusing degree control variable value when the comparison result meets threshold conditions, and setting the scanning electron microscope by using the focusing degree control variable value, thereby completing the automatic focusing.

13. The method of claim 12, wherein the focus difference Δ f in the overlapping area of the template image and the target image of the matching success that satisfies the predetermined condition in the overlapping area is used, and an empirical formula is combined to determine the search step, and the best focus is searched for with the search step.

14. The method according to claim 12 or 13, wherein the search is started in a predetermined direction on the side of the initial focusing power control variable i0, the control variable is changed by one step to obtain a control variable i1, a corresponding first search image is acquired, the degree of focus of the first search image and the template image in the overlapping region is directly compared, if the threshold condition is satisfied, a scanning electron microscope is set by the control variable i1 to complete the autofocusing, otherwise, the direction of change of the control variable is determined based on the difference in the degree of focus between the current target image and the first search image, the focusing power control variable is changed by one step to obtain a control variable i2, a corresponding second search image is acquired, the degree of focus of the second search image and the template image in the overlapping region is directly compared, if the threshold condition is satisfied, the scanning electron microscope is set by the control variable i2 to complete the autofocusing, otherwise, the focusing power control variable corresponding to the optimal degree of focus is obtained by using the initial focusing power control variable i0, the focusing power control variable i1, the focusing power control variable i2, and a corresponding focusing power control curve, and the autofocusing electron microscope is set by the same.

15. A rechecking method of an electron beam wafer defect rechecking device is characterized by comprising the following steps: preparing a patterned wafer for defect review using the method of any of claims 1-14, and then performing job-time work of the review equipment.

Images