Methods and apparatus for analyzing images of microlithographically lithographically constructed components

Abstract

This invention relates to a method and apparatus for analyzing images of microlithographic microstructure components, wherein each of a plurality of pixels in the image is assigned an intensity value under various conditions. The method according to the invention includes the following steps: separating a plurality of edge segments in the image (step S130); classifying each of the separated edge segments as a relevant edge segment or an irrelevant edge segment (step S140); and determining continuous segments in the image based on the relevant edge segments (step S150).

Description

[0001] This application claims priority to German patent application DE 10 2021 113 764.0, filed on May 27, 2021. The contents of this application are incorporated herein by reference. Technical Field

[0002] The present invention relates to methods and apparatus for analyzing images of microlithographic microstructure components, particularly images of masks or wafers. Background Technology

[0003] For example, microlithography is used to manufacture microstructured components such as integrated circuits or LCDs. The microlithography process is performed in an apparatus called a projection illumination apparatus, which includes an illumination device and a projection lens. Here, an image of a mask (=mask master) illuminated by the illumination device is projected through the projection lens onto a substrate (e.g., a silicon wafer) coated with a photosensitive layer (photoresist) and positioned in the image plane of the projection lens, so as to transfer the mask structure onto the photosensitive coating of the substrate.

[0004] As the structural dimensions of both the masks used in photolithography and the microlithographic structured wafers become smaller and smaller, the challenges of analyzing, processing, or repairing these components in practice are becoming increasingly severe.

[0005] Furthermore, images obtained through microscopy using electron beams or ion beams are analyzed to determine any discrepancies between the corresponding measured images and the design images of the desired structure with a mask, and these are used as the basis for repairing the mask or wafer. The images to be analyzed (e.g., scanning electron microscopy (SEM) records of a mask or wafer) typically consist of multiple pixels, with each pixel assigned an intensity value as a "grayscale value".

[0006] In practice, various methods are known for distinguishing coated or structured areas of a mask or wafer from uncoated or unstructured areas through contour extraction or detection. Conventional methods include, for example, contour extraction or detection based on the second derivative of the grayscale contour, which is related to the application of various filtering mechanisms and algorithms for closing existing contours. Another known method is based on the intensity value itself (i.e., the zeroth derivative of the grayscale contour) exceeding a threshold.

[0007] It has been demonstrated in practice on numerous occasions that it is difficult to distinguish between segments to be separated (i.e., uncoated or structured regions versus coated or unstructured regions) during the analysis of images with strong pixelation and potentially high noise levels. This is especially true if the regions to be separated have approximately the same average brightness due to significant pixelation.

[0008] Against this backdrop, providing a robust method that delivers accurate results for a wide variety of scenarios or microscopic records presents significant challenges in practice. Summary of the Invention

[0009] One object of the present invention is to provide a method and apparatus for analyzing images of microlithographic microstructure components, which facilitates reliable characterization while at least partially avoiding the aforementioned problems.

[0010] This objective is achieved by methods and apparatus based on the features described below.

[0011] This invention particularly relates to a method for analyzing images of microlithographic microstructure components, wherein each of a plurality of pixels in the image is assigned an intensity value under various conditions, and the method includes the following steps:

[0012] - Separate multiple edge segments in an image;

[0013] - Classify each of the separated edge segments into relevant edge segments or irrelevant edge segments; and

[0014] - Determine continuous segments in the image based on relevant edge fragments.

[0015] In this context and below, "relevant" edge segment is understood to mean an edge segment that is considered to be the actual boundary between segments to be separated during image analysis according to the invention, i.e., the boundary between coated or structured regions and uncoated or unstructured regions.

[0016] The present invention is based in particular on the concept that all edge pixels are initially found in the microscopically acquired image to be analyzed, and multiple edge segments are separated from them so that irrelevant edge segments (which do not actually represent the boundary between coated or structured regions and uncoated or unstructured regions) are then removed from the separated edge segments.

[0017] In embodiments of the invention, as will be described in more detail below, separating edge segments specifically includes, after skeletonization of the pixelated image (during which any wider edge segment is replaced with an edge segment having only one pixel width), deleting or eliminating shorter branches (particularly branches having only one pixel length) and / or deleting or eliminating intersections (i.e., points where at least three lines intersect in the skeletonized image). The invention here specifically includes the principle that, although the elimination of branches and / or intersections has been previously described, the remaining edge segments in this case are initially retained, i.e., those not yet eliminated at this stage of the method, and any classification of them as "relevant" or "irrelevant" is only made during subsequent method steps.

[0018] In the embodiments, the classification of the aforementioned “relevant” and “irrelevant” edge segments is based on the occurrence of the average intensity gradient of the separated edge segments, where values ​​exceeding or falling below appropriately defined thresholds can then be used as the basis for quantitative criteria.

[0019] In embodiments of the invention, the threshold is then preferably defined such that specified edge segments (i.e., associated average intensity gradients) are not considered when defining the threshold or are reduced in terms of gradient values.

[0020] In embodiments, edge segments that are not considered or to be eliminated can be particularly short edge segments, particularly high-contrast edge segments, particularly low-contrast edge segments, and / or edge segments located near relatively bright edge segments. To define the threshold, edge segments located near edge segments with relatively high contrast can also initially be reduced (and thus "weakened") in terms of average gradient value. Such preprocessing ensures that relatively short edge segments with relatively high intensity contrast, for example, are "eliminated" when defining the threshold for classification (the distinction between "relevant" and "irrelevant" edge segments), in order to avoid any falsification of a reasonable threshold due to "outliers" in the pixel distribution of average intensity gradient values.

[0021] In embodiments of the invention, continuous segments are determined based on relevant edge fragments without prior closure of gaps between edge fragments. In embodiments of the invention, as will be described in more detail below, even in the absence of fully closed edge paths, pixels located in interrupted regions between continuous edge fragments are allocated in a distance-based manner, specifically by determining for each pixel whether it is located in one region closer to an adjacent region or another region (i.e., closer to a coated or uncoated region). Not pre-closing the gaps between edge fragments before determining continuous segments has the following advantages: the method according to the invention is accelerated (i.e., the required computation time is reduced), and furthermore, the method is less prone to error because any errors associated with such steps can be avoided by selecting not to pre-close the edges.

[0022] According to one embodiment, irrelevant edge segments are also eliminated during the determination of consecutive segments.

[0023] According to one embodiment, after continuous segments have been determined, edge segments are combined into object edges and edge coordinates are calculated on a subpixel basis.

[0024] According to one embodiment, after edge fragments have been combined into object edges in this manner, irrelevant object edges are eliminated.

[0025] According to one embodiment, segmented images are calculated after irrelevant object edges have been eliminated.

[0026] According to one embodiment, image preprocessing is performed before separating multiple edge segments to reduce noise components.

[0027] According to one embodiment, separating multiple edge segments includes eliminating branches whose length is less than a specified value.

[0028] According to one embodiment, separating multiple edge segments includes eliminating at least three intersection points where the edges intersect.

[0029] According to one embodiment, the image is divided into two distinct regions by segmentation, and these two distinct regions differ in the materials located in the respective regions.

[0030] According to one embodiment, the microstructure component is a mask. In particular, the mask can be designed for operating wavelengths less than 250 nm, especially for operating wavelengths less than 200 nm, and even more particularly for operating wavelengths less than 15 nm.

[0031] According to another embodiment, the microstructure component is a wafer.

[0032] The present invention also relates to an apparatus for analyzing images of microlithographic microstructure components, wherein the apparatus is designed to perform a method having the above-described features.

[0033] Regarding the advantages and advantageous configurations of the device, refer to the above explanation in connection with the method according to the invention.

[0034] Other improvements to the invention can be obtained from the specification and dependent claims.

[0035] The invention will now be explained in more detail with reference to the exemplary embodiments shown in the accompanying drawings. Attached Figure Description

[0036] In the attached image:

[0037] Figure 1 A flowchart illustrating a possible sequence for one embodiment of the method according to the present invention is shown;

[0038] Figure 2 It was shown as Figure 1 Exemplary results of edge detection performed as part of the method according to the present invention;

[0039] Figures 3a-3b The results of separating edge segments as part of the method according to the invention are shown, including eliminating branches ( Figure 3a ) and intersection ( Figure 3b );

[0040] Figures 4a-4b This illustrates the method for clarifying branch elimination ( Figure 4a ) or intersection ( Figure 4b A schematic diagram;

[0041] Figure 5 A diagram is shown for illustrating threshold-based classification of edge segments that occur as part of the method according to the invention;

[0042] Figures 6a-6b A schematic diagram is shown illustrating the determination of continuous segments in the absence of a fully closed edge path, as part of the method according to the invention.

[0043] Figure 7 A schematic diagram illustrating further image processing as part of the active contouring method is shown; and

[0044] Figures 8a-8b A schematic diagram is shown to illustrate another embodiment of the method according to the invention when applied to the analysis of SEM images of a wafer. Detailed Implementation

[0045] The following will refer to Figure 1 The flowchart shown and Figure 2 The embodiments of the method according to the present invention are explained in more detail with reference to the schematic diagram in Figure 8.

[0046] according to Figure 1 Initially, in step S100, an image acquired by a microscope is provided, such as an image of a mask or wafer recorded by a scanning electron microscope (SEM).

[0047] The image is then analyzed using the method according to the invention, which has method steps described below, such that regions coated or carrying structures are distinguished from unstructured or uncoated regions in various cases. Regarding coated regions, the term "segmentation" will be used hereinafter, defined by "edge." Furthermore, such segments of edges will be referred to below as "edge fragments." Additionally, the image to be analyzed consists of multiple pixels, each assigned an intensity value (as a "grayscale value").

[0048] The image to be analyzed typically has significant noise components (where, for example, pixels with low intensity values ​​and pixels with relatively high intensity values ​​can be directly adjacent to each other). To reduce the noise, image preprocessing is initially performed in step S110, where any suitable image smoothing methods can be combined in principle. Suitable methods include, for example, binning, Gaussian filtering, low-pass filtering, etc. As an example only, in this case, four (or possibly more or fewer) mutually adjacent pixels can be replaced by a single pixel, which is then assigned the average intensity value of the four pixels.

[0049] In the subsequent step S120, edge pixels are initially identified or extracted from the corresponding preprocessed or smoothed image. To capture as many edge pixels as possible in this case, it is preferable to use multiple edge extraction methods known in each case, or to apply the same edge extraction method multiple times with different parameters. Suitable known methods include, for example, "Canny," "Laplacian of Gaussian," "Sobel," etc.

[0050] The edges found or extracted in step S120 typically also have—such as Figure 2 As shown in the example, the width is at least partially greater than the width of one pixel (and may vary along the corresponding edge).

[0051] In the subsequent step S130, the edge pixels found in the previous step S120 are initially skeletonized, where wider edge segments are replaced in each case by edge segments with a width of only one pixel (see [link to previous step]). Figure 3a In this case, in particular, two or more narrow edge segments (each with a width of one pixel) can replace the original wider edge segment.

[0052] Furthermore, branches or intersections are also processed in step S130. In this case, relatively short branches (especially those with a length of only one pixel) are deleted, such as... Figure 4a The diagram is schematic and shown in a highly simplified manner. Conversely, if the branch has a larger length (e.g., more than one pixel), the relevant intersection is removed, resulting in the line originally originating from that intersection now appearing as separate edge fragments, such as... Figure 4b The diagram is schematic and shown in a highly simplified form. This takes into account the fact that in a “binary” image with only two different types of regions (e.g., coated and uncoated mask regions), the actual existence of such intersections is theoretically excluded; in other words, at least one of the lines intersecting at said intersection is “irrelevant” in this respect, meaning that the line does not represent any actual transition between the coated and uncoated regions.

[0053] Then, refer to again Figure 1In a further step S140, the previously separated edge segments are classified as either "relevant edge segments" (which essentially represent the transition between coated and uncoated areas) or "irrelevant edge segments" (i.e., edge segments without such a transition). For this classification (based on "eliminating" irrelevant or "erroneous" edge segments), in each case, the average intensity gradient is applied to the individual edge segment and compared with a threshold that has been determined in an appropriate manner. Here, only edge segments with an average intensity gradient exceeding the determined threshold are considered "relevant" in the aforementioned sense.

[0054] Figure 5 The diagram shown is merely an example illustrating a potential definition of a threshold based on an exemplary distribution of the average intensity gradient values ​​of edge segments appearing in an image, where each index on the horizontal axis corresponds to an edge segment, and where the associated values ​​of the average intensity gradient are plotted on the vertical axis.

[0055] The threshold is then preferably defined according to the invention such that specified edge segments (i.e., the associated average intensity gradient) are not considered when defining the threshold or are reduced in terms of gradient value (i.e., "weakened"). In particular, particularly short edge segments, especially high-contrast edge segments, particularly low-contrast edge segments, and / or edge segments located near relatively bright edge segments may not be considered during the definition of the threshold. Furthermore, even edge segments near relatively bright edge segments may be initially "weakened," i.e., reduced in terms of the value of the average intensity gradient, before the threshold is defined.

[0056] The preprocessing described above has the advantage that any “outliers” or corresponding values ​​of the average intensity gradient in the edge segments can be initially eliminated, resulting in a reduction in the dispersion of the corresponding average intensity gradient values ​​of the ultimately distinguished regions (i.e., “bright” and “dark”), or a “homogenization” of the range of related values.

[0057] As described below, continuous segments are determined based on previously classified related edge segments. This determination of continuous segments is then preferably performed without any prior closure that may still exist between the related edge segments, i.e., before any fully closed edge paths are formed (see below). Figure 6a ).

[0058] Refer again Figure 1 In step S150, edge fragments are clustered to image objects, where irrelevant clusters or portions thereof will be further removed.

[0059] Specifically, in step S150, pixels located in the interrupted region between consecutive edge segments are assigned to one or the other of adjacent regions (i.e., coated or uncoated regions, or bright or dark segments) in an interval-based manner. This distance-based assignment, particularly by using additional virtual edges, each with a pixel width, to supplement existing edge segments on either side in each case, allows the nearest edge segment located in the interrupted region to be determined, regardless of whether it is closer to one region (e.g., a "bright" or uncoated region) or closer to another region (e.g., a "dark" or coated region). Depending on the result of this distance comparison, each pixel is then assigned to one or the other region, such that, as a result of step S150, an effectively closed segment is obtained (see...). Figure 6b ).

[0060] As described above, the method according to the invention omits the prior closure of edge paths or edge segments in the creation of closed segments, thus accelerating or simplifying the process and avoiding errors that may be associated with closed edge paths.

[0061] Since all edge fragments are now available “against the object background” (i.e., can now be considered as an integral part of the coated or uncoated area in each case), any remaining objects or edge fragments that do not actually represent a transition between the coated and uncoated areas can now also be eliminated in the image in step S150. These can be, in particular, objects with relatively few edge pixels, objects with only one non-closed edge fragment, or edge fragments within objects branching from a closed polygon chain. The fact that the elimination of individual objects or edge fragments occurs only at this stage of the method (i.e., not in one of the previous steps) has a favorable effect on the reliability of the method according to the invention, because unauthorized deletion of a specified object or edge fragment—which is unauthorized because it occurs when the background of the object is unknown—is avoided.

[0062] Next, in step S160, edge coordinates are calculated on a subpixel-by-subpixel basis. For this purpose, edge fragments are combined into object edges. Subpixel-accurate position calculation can preferably be achieved using an active contouring method. Here, for the purpose of obtaining the most constant external energy along the edge, gradient image smoothing and / or homogenization can be implemented in various cases. Furthermore, the support points of the edge path between some or all iterations of the active contouring method can be adapted such that, in various cases, the distance between two support points along the corresponding edge is substantially constant. Moreover, even between some or all iterations of the active contouring method, the support points of the edge path can be adapted such that the distance between two support points along the edge is reduced in regions with relatively more significant edge curvature. Alternatively, the subpixel-by-subpixel calculation of edge position can occur by shifting the edge in its normal direction to the location of the maximum gradient.

[0063] Next, in step S170, object edges are selected. For this, irrelevant object edges can be eliminated in a manner similar to step S140. Next, in step S180, a segmented image is calculated based on the object edges, where the hue can be determined based on the intensity gradient along the corresponding edge. Furthermore, sub-pixel information of the edge locations can be represented based on the grayscale values ​​in the resulting image.

[0064] Figures 8a-8b A schematic diagram illustrating another embodiment of the method according to the invention is shown, wherein the image to be analyzed is a SEM image of a wafer, compared to the previously described exemplary embodiments.

[0065] in this regard, Figure 8a With Figure 2 A similar approach is shown to illustrate the results obtained as a result of step S120 (i.e., as a result of edge extraction), where the Canny algorithm, used only as an example in this case, is used to perform edge detection or extraction. (See from...) Figure 8a Specifically, the edges detected here have interruptions in the regions indicated by “A” and “B”, respectively.

[0066] In principle, in the implementation examples, it can be based on Figure 8a The image presented in the previous reference is implemented. Figure 1-7 The exemplary embodiments described are similar to further processing. However, in another exemplary embodiment, the "repair" or closure of edge paths can be implemented in regions "A" and "B". This can be accomplished again by applying a relatively simple threshold-based method to the direct environment of the interruption located in regions "A" and "B". For this purpose, after identifying the interrupted regions "A" and "B", trimming is performed according to... Figure 8aThe image. In each detail thus produced, edges or contours are generated using a threshold-based method known per se, where the grayscale value of the pixel located in the end segment is used as a threshold for edge detection or contour generation. Therefore, the edges or contours extracted in this method necessarily extend through the end segment of the original edge. If according to Figure 8b If two details with additional extracted edges or contours overlap, this in the exemplary embodiment results in a connection between previously existing end segments, and thus a "repair" of the edge path in the interrupted area. Subsequently, edge skeletonization can be performed on the edge path that has been "repaired" in this way, similar to the previous reference. Figure 1 The exemplary embodiments described in the following figures are used to define, for example, their width as, in each case, one pixel.

[0067] Although the invention has been described based on specified embodiments, many variations and alternative embodiments will be apparent to those skilled in the art, for example, by combining and / or exchanging features of individual embodiments. Therefore, it will be apparent to those skilled in the art that the invention also covers such variations and alternative embodiments, and that the scope of the invention is limited only to the appended claims and their equivalents.

Claims

1. A method for analyzing an image of a microlithographic microstructure component, wherein, In the image, each of a plurality of pixels is assigned an intensity value in each case, wherein the method comprises the following steps: - separating a plurality of edge segments in the image; - classifying each of the separated edge segments as a relevant edge segment or as an irrelevant edge segment; and - determining a continuous segmentation in the image on the basis of the relevant edge segments; - wherein, for determining a continuous segmentation, pixels located in a surrounding area of an edge segment in each case are assigned to a respective one of two areas separated by the edge segment in an interval-based manner; and - wherein the continuous segmentation is determined without previously closing gaps existing between the edge segments.

2. The method of claim 1, wherein, The classification of each of the separated edge segments is effected on the basis of an average intensity gradient of the separated edge segment.

3. The method of claim 2, wherein, Each of the separated edge segments is classified on the basis of whether the respective average intensity gradient of the separated edge segment exceeds a threshold value.

4. The method of claim 3, wherein, For the definition of the threshold value, some of the separated edge segments are not taken into account.

5. The method according to any one of claims 1 to 4, characterized in that, Irrelevant edge segments are also eliminated during the determination of the continuous segmentation.

6. The method according to any one of claims 1 to 4, characterized in that, After the determination of the continuous segmentation, the edge segments are combined into object edges and edge coordinates are calculated in a sub-pixel manner.

7. The method of claim 6, wherein, After the combination of the edge segments into object edges, irrelevant object edges are eliminated.

8. The method of claim 7, wherein, After the elimination of the irrelevant object edges, a segmentation image is calculated.

9. The method according to any one of claims 1 to 4, characterized in that, Before the separation of the plurality of edge segments, an image pre-processing for reducing noise components is carried out.

10. The method according to any one of claims 1 to 4, characterized in that, The separation of the plurality of edge segments comprises eliminating branches having a length below a specified value.

11. The method according to any one of claims 1 to 4, characterized in that, The separation of the plurality of edge segments comprises eliminating intersections of at least three edges.

12. The method according to any one of claims 1 to 4, characterized in that, The image is divided into a total of two different areas by the segmentation, which differ in terms of the material located in the respective area.

13. The method according to any one of claims 1 to 4, characterized in that, The microstructured component is a mask.

14. The method of claim 13, wherein, The mask is designed for a working wavelength of less than 250 nm.

15. The method of claim 14, wherein, The working wavelength is less than 200 nm.

16. The method of claim 14, wherein, The working wavelength is less than 15 nm.

17. The method of any one of claims 1 to 4, wherein, The microstructured component is a wafer.

18. An apparatus for analyzing an image of a microlithographic microstructure component, characterized by The apparatus is designed to carry out the method as claimed in any one of claims 1 to 17.

Images