阅读说明：本技术 用于制造集成电路的热点避免方法 (Hot spot avoidance method for manufacturing integrated circuits ) 是由 刘翼硕 夏致群 周信廷 苏冠华 洪伟伦 陈志宏 陈科维 于 2021-01-19 设计创作，主要内容包括：本公开涉及用于制造集成电路的热点避免方法。一种方法,包括：从集成电路的布图中裁剪多个图像；生成第一多个散列值,其中每一者来自多个图像之一；加载存储在热点库中的第二多个散列值；以及将第一多个散列值中的每一者与第二多个散列值中的每一者进行比较。比较步骤包括计算第一多个散列值中的每一者与第二多个散列值中的每一者之间的相似度值。方法还包括将相似度值与预定阈值相似度值进行比较,并且响应于相似度值大于预定阈值相似度值的结果,记录具有该结果的相应图像的位置。该位置是相应图像在布图中的位置。(The present disclosure relates to a hotspot avoidance method for manufacturing integrated circuits. A method, comprising: cropping a plurality of images from a layout of an integrated circuit; generating a first plurality of hash values, each from one of a plurality of images; loading a second plurality of hash values stored in the hotspot library; and comparing each of the first plurality of hash values to each of the second plurality of hash values. The comparing step includes calculating a similarity value between each of the first plurality of hash values and each of the second plurality of hash values. The method further includes comparing the similarity value to a predetermined threshold similarity value and, in response to a result of the similarity value being greater than the predetermined threshold similarity value, recording a location of the corresponding image having the result. The position is the position of the corresponding image in the layout.)

1. A method of manufacturing a semiconductor device, comprising:

cropping a plurality of images from a layout of an integrated circuit;

generating a first plurality of hash values, each from one of the plurality of images;

loading a second plurality of hash values stored in the hotspot library;

comparing each of the first plurality of hash values to each of the second plurality of hash values, wherein the comparing comprises calculating a similarity value between each of the first plurality of hash values and each of the second plurality of hash values;

comparing the similarity value to a predetermined threshold similarity value; and

in response to a result of the similarity value being greater than the predetermined threshold similarity value, recording a location of the respective image with the result, wherein the location is a location of the respective image in the layout.

2. The method of claim 1, wherein the plurality of images form an array, and the location comprises a row number and a column number of the respective image in the array.

3. The method of claim 1, further comprising:

fabricating the integrated circuit on a wafer, wherein the fabricating includes performing a chemical mechanical polishing process on the wafer; and

finding a hotspot from the location, wherein the hotspot is a defect in the wafer resulting from the chemical mechanical polishing process.

4. The method of claim 1, wherein cropping the plurality of images comprises: the layout is divided into an array of images, and each image of the plurality of images in the array is cropped.

5. The method of claim 1, wherein the predetermined threshold similarity value is 0.9.

6. The method of claim 1, further comprising:

cropping an additional plurality of images from an additional layout of the additional integrated circuit;

generating a third plurality of hash values, each from one of the additional plurality of images;

comparing each of the third plurality of hash values to all hash values stored in the hotspot library to find a set of hash values that are similar to the third plurality of hash values; and

the similarity values for the set of hash values are ranked.

7. The method of claim 6, further comprising: a recipe associated with one hash value in the set of hash values is selected.

8. A method of manufacturing a semiconductor device, comprising:

cropping a plurality of images from a layout of an integrated circuit;

generating a plurality of hash values, each hash value from one of the plurality of images;

searching from a hotspot repository to find similar hash values similar to the plurality of hash values, wherein the hotspot repository stores hash values indexed to images with hotspots; and

marking locations of ones of the plurality of images associated with the similar hash values on a layout of the integrated circuit.

9. The method of claim 8, further comprising:

implementing a layout of the integrated circuit on a wafer, wherein the implementing includes performing a chemical mechanical polishing process on the wafer using a recipe; and

inspecting the location on the wafer to determine a hotspot at the location.

10. A system for fabricating a semiconductor device, comprising:

a library stored in a tangible medium, the library comprising a plurality of entries, each entry comprising:

a hash value;

an image associated with the hash value, wherein the image comprises a hotspot;

a recipe configured to reduce the hot spots; and

topology information of the hotspot.

Technical Field

The present disclosure relates to a hotspot avoidance method for manufacturing integrated circuits.

Background

In the manufacture of integrated circuits, process-related defects such as topological hot spots (hotspots) may be found after the corresponding processes are completed by physical measurements from the manufactured wafers. For example, to find defects associated with Chemical Mechanical Polishing (CMP), several stages must be performed to find topological defects, including a circuit design stage, a circuit layout stage, manufacturing and performing CMP on a physical wafer, and measuring the physical wafer. This process typically takes three months.

Disclosure of Invention

According to an embodiment of the present disclosure, there is provided a method of manufacturing a semiconductor device, including: cropping a plurality of images from a layout of an integrated circuit; generating a first plurality of hash values, each from one of the plurality of images; loading a second plurality of hash values stored in the hotspot library; comparing each of the first plurality of hash values to each of the second plurality of hash values, wherein the comparing comprises calculating a similarity value between each of the first plurality of hash values and each of the second plurality of hash values; comparing the similarity value to a predetermined threshold similarity value; and in response to a result of the similarity value being greater than the predetermined threshold similarity value, recording a location of the respective image with the result, wherein the location is a location of the respective image in the layout.

According to another embodiment of the present disclosure, there is provided a method of manufacturing a semiconductor device, including: cropping a plurality of images from a layout of an integrated circuit; generating a plurality of hash values, each hash value from one of the plurality of images; searching from a hotspot repository to find similar hash values similar to the plurality of hash values, wherein the hotspot repository stores hash values indexed to images with hotspots; and marking locations of ones of the plurality of images associated with the similar hash values on a layout of the integrated circuit.

According to still another embodiment of the present disclosure, there is provided a system for manufacturing a semiconductor device, including: a library stored in a tangible medium, the library comprising a plurality of entries, each entry comprising: a hash value; an image associated with the hash value, wherein the image comprises a hotspot; a recipe configured to reduce the hot spots; and topology information of the hotspot.

Drawings

Various aspects of the disclosure are best understood from the following detailed description when read with the accompanying drawing figures. Note that, in accordance with the standard practice in the industry, various features are not drawn to scale. In fact, the dimensions of the various features may be arbitrarily increased or reduced for clarity of discussion.

Fig. 1-4 illustrate cross-sectional views of a structure on which a chemical mechanical polishing process is performed and the results, according to some embodiments.

FIG. 5 illustrates a schematic flow in the design and manufacture of an integrated circuit according to some embodiments.

FIG. 6 illustrates a process flow for constructing a hotspot library, in accordance with some embodiments.

FIG. 7 illustrates an example image and a generated hash value according to some embodiments.

Fig. 8 illustrates a schematic diagram of an example wafer with hot spots, according to some embodiments.

FIG. 9 illustrates an example cropped image according to some embodiments.

FIG. 10 illustrates a grouping of hash values according to some embodiments.

Fig. 11 and 12 illustrate represented regions in a cropped image having different pattern densities and line widths according to some embodiments.

FIG. 13 illustrates a process flow for determining potential hotspots using a hotspot library, according to some embodiments.

FIG. 14 illustrates cropping the layout into a plurality of cropped images according to some embodiments.

FIG. 15 illustrates a process flow for finding a recipe (recipe) corresponding to a hotspot, in accordance with some embodiments.

FIG. 16 illustrates a graphical representation in finding a recipe according to some embodiments.

FIG. 17 illustrates an example recipe according to some embodiments.

FIG. 18 illustrates a process of improving a hotspot prevention model and improving a recipe, according to some embodiments.

FIG. 19 illustrates a system for performing tasks, in accordance with some embodiments.

Detailed Description

The following disclosure provides many different embodiments, or examples, for implementing different features of the invention. Specific examples of components and arrangements are described below to simplify the present disclosure. Of course, these are merely examples and are not intended to be limiting. For example, in the description that follows, forming a first feature over or on a second feature may include embodiments in which the first and second features are formed in direct contact, and may also include embodiments in which additional features may be formed between the first and second features, such that the first and second features may not be in direct contact. Further, the present disclosure may repeat reference numerals and/or letters in the various examples. This repetition is for the purpose of simplicity and clarity and does not in itself dictate a relationship between the various embodiments and/or configurations discussed.

Furthermore, spatially relative terms (e.g., "below," "beneath," "below," "above," "upper," etc.) may be used herein to readily describe one element or feature's relationship to another element(s) or feature(s) as illustrated in the figures. These spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. The device may be otherwise oriented (rotated 90 degrees or at other orientations) and the spatially relative descriptors used herein interpreted accordingly.

According to some embodiments, a hotspot avoidance method for manufacturing an integrated circuit is provided. According to some embodiments, systems and processes are shown for predicting hotspots and using the predicted hotspots to find the best recipe. Some variations of some embodiments are discussed. The embodiments discussed herein will provide examples to enable making or using the presently disclosed subject matter, and one of ordinary skill in the art will readily appreciate modifications that can be made while remaining within the intended scope of the various embodiments. Like reference numerals are used to indicate like elements throughout the various views and illustrative embodiments. Although method embodiments may be discussed as being performed in a particular order, other method embodiments may be performed in any logical order.

Throughout the specification, the term "hot spot" refers to a defect generated in an integrated circuit manufacturing process, rather than a design-related defect. In other words, the term "hot spot" refers to a process-related defect. Examples of hot spots are defects generated in a Chemical Mechanical Polishing (CMP) process, which will be discussed in detail with reference to fig. 1 to 4, while hot spots may also refer to other types of defects including, but not limited to, defects related to an etching process (e.g., portions that are desired to be removed but not removed in an etching process), defects related to a deposition process, and the like.

Fig. 1-4 illustrate deposition of some features, a CMP process, and several possible results from CMP. It should be understood that fig. 1-4 illustrate example structures of a CMP process for forming metal lines, and that the actual CMP process may be applied to different structures, different materials, etc. Referring to fig. 1, a wafer 10 is provided. The wafer 10 includes a base layer 20, which base layer 20 may include, for example, a silicon substrate and overlying structures and layers, and details are not shown. A plurality of trenches may be formed that extend into the dielectric layer of the base layer 20. A deposition process is then performed to deposit a glue layer 22 and a metal material 24 is deposited over the glue layer 22. According to some embodiments of the present disclosure, the glue layer 22 may be formed of or include titanium, titanium nitride, tantalum nitride, and the like. The filler material 24 may include copper, aluminum copper, or the like. Due to the topology of the trenches in the base layer 20, the deposited metal material 24 has a non-planar top surface that can reflect the topology of the base layer 20. A CMP process is performed to remove excess portions of metal material 24, resulting in a plurality of conductive features 26 (including 26A and 26B), which plurality of conductive features 26 may include metal lines, vias, contact plugs, and the like, as shown in fig. 2, 3, or 4.

Different results may be achieved due to various factors, such as the topology of the base layer 20 (e.g., the density and width of the trenches), and the recipe of the CMP process, as shown in fig. 2, 3, or 4. Throughout the specification, the term "recipe" refers to a collection of process conditions, such as the number of (sub) steps in a CMP process, the type of slurry, the flow rate of the slurry, the downforce of the wafer relative to the polishing pad, the conditioning, the rotational speed, and the like. Thus, a recipe comprises a fixed set of process conditions. Another recipe may be considered to be generated when one or more of the process conditions of the recipe change.

Fig. 2 shows an ideal situation to be achieved. In fig. 2, the top surfaces of all of the resulting conductive features 26 (including 26A and 26B) are coplanar regardless of the width and pattern density of the conductive features 26.

Fig. 3 shows a practical situation which is not ideal but still acceptable. Due to the pattern loading effect, portions of metal material 24 having a higher density and/or larger width are polished more than portions of metal material 24 having a lower density and/or smaller width, thereby creating a dishing effect (dishing effect) that creates a recess. The depth D1 of the groove is less than the design specification and therefore no hot spot is generated. For example, design specifications may require a dish depth of less than about 10 nm. Since all grooves have a depth D1 less than specification, the result is acceptable and the grooves are not hot spots.

Fig. 4 shows a case where hot spots are generated in the case where the trenches are wide and/or the pattern density of the trenches is high. For example, the recess depth D2 of the wide trench is greater than the design specification (e.g., 10 nm). These irregular grooves can cause problems with subsequent processing, which may include circuit shorts, circuit opens, etc., depending on the particular circuit design. Throughout the specification, non-canonical grooves are used as example hotspots to explain the concepts of the present disclosure. In addition, it should be understood that fig. 3 and 4 illustrate overpolishing in a CMP process, while under-polishing may also occur with portions being less polished (and thus portions above the top surface of base layer 20), and hot spots may also be generated when the resulting bump (hump) is not normal. Hot spots may degrade production yield and need to be eliminated or at least reduced to within specification.

Fig. 5 shows a schematic flow in the design and manufacture of an integrated circuit according to some embodiments of the present disclosure. A circuit design is first provided (process 30) and the design may include a schematic of the circuit. Next, a layout of the circuit is prepared (process 32). From the layout, the hot spots and the locations of the hot spots in the layout are predicted using the models provided according to embodiments of the present disclosure (process 34). Process 300 in fig. 13 shows details of predicting hotspots. The generation, use, and improvement of the model are discussed in detail in subsequent paragraphs. Throughout the specification, the model is referred to as a hotspot prevention model.

After predicting the hotspots, a recipe (hereinafter referred to as the selected recipe) that produces the fewest number of hotspots is selected based on the predicted hotspots (process 36) such that by performing the CMP using the selected recipe, the number of hotspots is minimized and the severity (e.g., depth D1 and D2, as shown in fig. 4) is minimized. Process 400 in fig. 15 shows details of selecting a recipe. The recipe is then used to fabricate circuits on the wafer and to perform a CMP process on the physical wafer (process 38). It should be understood that no CMP process may be performed on any wafer implementing the particular layout until the point in time when the selected recipe is selected. After the CMP process, the resulting polished wafer may be tested (process 40) to verify the presence and location of hot spots. The test results may also be used to improve the hotspot prevention model, which is included in process 500 shown in fig. 18.

In subsequent paragraphs, a process flow 200 (FIG. 6) for constructing and improving a hotspot library, a process flow 300 (FIG. 13) for predicting hotspots on a circuit layout using a hotspot library, and a process flow 400 (FIG. 15) for suggesting a selected recipe are discussed in detail. These processes combine to provide a solution for predicting and eliminating (or at least minimizing) hot spots without actually performing a process (e.g., a CMP process) on a physical wafer.

Referring to FIG. 6, a process flow 200 for generating and improving a hotspot library is provided. Referring to process 202, a training layout of a chip implementing a circuit is provided. The chip layout may take the form of a Graphic Data System (GDS) format, or any other suitable format. Throughout the specification, the layout may alternatively be referred to as a GDS file. It should be understood that the training layout may be dedicated to generating hot spot libraries, but not for mass production of the product, or the training layout may be a production layout to be implemented on a production wafer.

The lab wafer is then fabricated to achieve the training GDS. Fig. 8 shows a schematic diagram of a respective wafer 42, the respective wafer 42 comprising a plurality of chips 44, wherein a layout (training GDS) is implemented in each chip 44. After the CMP process is performed, a test is performed to measure the surface topography of the wafer 42 (process 204 in fig. 6) and identify the hot spots 46 in the wafer 42. The location of the hotspot 46 in the wafer 42 is recorded, as shown by process 208 in fig. 6. Because multiple hot spots 46 may be found in process 204, multiple locations in wafer 42 are recorded.

Next, as shown in process 210 in FIG. 6, for each hot spot 46 found, an image is cropped from the layout, which may be in the form of a line GDS file. For example, FIG. 9 illustrates an example cropped image. Assuming that a hot spot 46 is found at location 48, the image surrounding the hot spot 46 is cropped from the row layout data. The image may be rectangular and may be square. The length L1 and width W1 of the cropped image are selected so that the surrounding environment around hotspot 46 is large enough to include surrounding features whose pattern and density may produce hotspot 46, but not so large as to include features that are detrimental to the generation of hotspot 46. For example, the length L1 and the width W1 may be in a range between about 64 μm and about 512 μm. Because single or multiple hot spots 46 may be found in process 204, single or multiple images may be cropped from the layout of wafer 42.

Referring to process 212 in FIG. 6, a hash value is generated from the cropped image. FIG. 7 shows an example for generating a hash value from an image. Because the image cannot be indexed and searched, the image is represented by a hash value, which is the only digital representation of the image. The hash values and images are in a one-to-one correspondence such that the same image will generate the same hash value, while different images will generate different hash values. Further, images similar to each other will generate similar hash values, and the similarity of the hash values can be calculated. The similarity of the hash values also indicates the similarity of the images. For example, the hash values may have a similarity between 0 and 1, where a value of 0 indicates that the images are completely different from each other and a value of 1 indicates that the images are the same. Generating hash values from images and calculating the similarity of the hash values may be performed using existing algorithms and tools. For example, the Discrete Cosine Transform (DCT) algorithm used by perceptual hashing (pHash) is a known available algorithm.

The hash value may be obtained by an intermediate value represented by a two-dimensional matrix, which is then converted into a hash value represented by a series of numbers and letters. For example, FIG. 7 shows three example images, image A, image B, and image C. The details of the image are not shown. Image a shows a person sitting on snow with a thick garment, with a tree in the snow. Image B is similar to image a, except that image B has been equalized from image a, with color and contrast adjusted. Image C shows the face with goggles on the forehead and flames surrounding the face. On the right side of each of image a, image B and image C, an 8 × 8 two-dimensional matrix is provided, which is generated from the respective image and/or two-dimensional matrix. Hash values (which include numbers and letters) are shown on the right side of the respective matrices. Referring back to FIG. 6, when a single or multiple hot spots 46 are found in process 204, the single or multiple images are cropped and a single or multiple hash values are generated in process 212.

Referring to process 214 in FIG. 6, the plurality of hash values are grouped into one or more groups by a grouping algorithm, where the grouping is based on similarity of hash values, similar hash values being grouped in the same hash group. An example grouping algorithm is explained using fig. 10. Fig. 10 shows a plurality of circles shown in a two-dimensional space for visually showing the hash value, each circle representing the hash value generated from the trimmed image. In the grouping algorithm, a plurality of hash values (which include hash values H1 through H13 shown as an example) are processed one by one. Assume that the order of processing is the serial number of the hash value (e.g., 1 to 13). In processing the hash value H1, because there are no other hash values, and there are no previously generated hash groups, a first hash group G1 is generated and the hash value H1 is placed in the first group G1. The hash value of the first placement H1 is considered to be the center of the first group G1.

Next, the second hash value H2 is processed. A similarity value between hash value H2 and the center of group G1 (centered at H1) is calculated. Assuming that the similarity value is greater than the predetermined threshold similarity value, hash values H1 and H2 are considered similar to each other, and hash value H2 belongs to the hash group G1. Throughout the specification, two hash values having a similarity value greater than a predetermined threshold similarity value are referred to as similar hash values. Their corresponding images are also referred to as similar images. Hash value H2 is added to hash group G1. According to some embodiments, the threshold similarity value is 0.9, and other values may be used.

Assuming that the hash value of the next process is H3, a similarity value between the hash value H3 and the center H1 of the hash group G1 is calculated. Further assuming that the similarity value is equal to or less than the predetermined threshold similarity value, hash values H1 and H3 are considered dissimilar, and hash value H3 does not belong to hash group G1. Thus, a second hash group G2 will be generated and hash value H3 placed in hash group G2. Hash value H3 is the center of hash group G2.

In subsequent processes, each of the remaining hash values H4 through H13 is processed one by one to calculate their similarity to the center of an existing hash group (e.g., G1 and G2), so that it can be determined to which hash group the newly processed hash value belongs, or whether a new hash group should be generated. Fig. 10 shows an example in which hash value H12 is dissimilar from any hub (e.g., H1 and H3), such that an additional hash group G3 is generated and hash value H12 is placed in hash group G3. The other hash values H4-H11 and H13 are in hash group G1 or G2.

Referring to process 216 in FIG. 6, the center of each hash group is obtained, which may be the hash value first placed in each hash group. After the centers of hash groups are obtained, the non-center hash values are discarded, as each center is similar to, and can represent, the other hash values in its group. In other words, the cropped image represented by the discarded hash value in the same hash group is similar to the cropped image represented by the hash value at the center of the hash group. The hash values of the central hash values in different hash groups are dissimilar to each other. Otherwise, if the two central hash values are similar to each other, then the two central hash values will have been placed in the same hash group, and as a result, only one of the hash values will be central, while the other hash value will be discarded.

Referring to FIG. 6, in process 218, a hotspot library (which includes a database) entry is composed for each hash value that is not discarded, which is the center of the hash group. According to some embodiments, a plurality of recipes are generated, as shown by process 206 in FIG. 6. The generation and modification of multiple recipes is discussed with reference to FIG. 18. The plurality of recipes may also include empirical recipes known to eliminate hot spots for certain types of images. Each recipe in the plurality of recipes corresponds to a test GDS having its hash value as shown in fig. 18. The hash value of the undiscarded center of the hash group is compared (by calculating a similarity value) to the hash value of the GDS file corresponding to the recipe, and the recipe (whose corresponding test GDS is closest to the center hash value) is associated with each center hash value. Each central hash value will be associated with a recipe.

In addition to the recipe, a cropped image (from which the corresponding center hash value was generated) is associated with the center hash value. Also, as will be discussed in subsequent paragraphs, expected topological information (e.g., whether the hot spot is under-polished or over-polished, and the recess depth or bump height) is also associated with the recipe (as will be discussed with reference to fig. 18). Expected topology information can also be obtained in the process shown in fig. 18. Thus, each hotspot library entry includes a hash value, a corresponding cropped image, a corresponding recipe, and corresponding topology information. A plurality of hotspot library entries are generated. The index of the hotspot library entry may be a hash value. These hotspot library entries are stored in a database in hotspot library 222, as shown by process 220 in figure 6.

As also shown in FIG. 6, a hotspot prevention model 223 can be constructed and updated by the process in process flow 200. The hotspot prevention model 223 incorporates the previously discussed relationships between GDS files and hotspots and uses GDS files or cropped images (or their corresponding hash values) as input parameters and outputs hotspots as output parameters.

Fig. 11 and 12 schematically show example cropped images saved with hash values in a hotspot repository. It will be appreciated that fig. 11 and 12 are schematic, in that some of the larger areas 52 are shown in outline, and some of the smaller areas are not shown in outline. Further, inside each region, there are a plurality of patterns such as parallel stripes, and the pattern in the illustrated region 52 is not illustrated. The size, shape and density of the patterns in the illustrated region 52 may be different from one another. The different patterns, pattern densities, etc. that form the environment surrounding the hot spot 46 are the cause of the hot spot. For example, in fig. 11, region 50 has a smaller line width that is much smaller than the line width of its surrounding region 52. Region 50 may also have a higher pattern density that is much higher than the pattern density of its surrounding region 52. This creates a hot spot 46. Hot spots are expected to occur when similar images with similar environments are found in other GDS files.

FIG. 12 illustrates a cropped image stored in the hotspot library. Similarly, region 52 is shown showing the outline of some of the larger regions, and the outline of some of the smaller regions is not shown. Further, inside each region 52, there are a plurality of patterns such as parallel stripes, not shown. The size, shape, and density of the patterns in the illustrated area 52 may be different from one another, thereby creating respective hot spots 46.

In the previous discussion, it was assumed that at the beginning of the process flow 200, the hotspot library 222 has not been generated and that no hash group and central hash value has been previously generated. Thus, a new hash group will be generated and the hot spot library will be generated from scratch. Once the hotspot library 222 is generated, the hotspot library 222 may be continually refined using a new training GDS file, which may be a mass produced GDS file, or a GDS file dedicated for training purposes rather than production. The processes 202, 204, 208, 210, 212, and 214 in FIG. 6 will be repeated for the new GDS file. Accordingly, a plurality of new hotspots are found from the newly manufactured wafer implementing the new GDS file, and thus, a plurality of new images are cropped. Then, a plurality of new hash values are generated from the newly cropped image. The newly generated hash values are then processed one by one and the similarity of the newly generated hash values to the existing hub hash values (stored in the hotspot repository 222) is calculated. It will be appreciated that at this point, the stored central hash values remain theoretically the center of the hash group, except that each hash group is a single member group (non-central members have been discarded) with only one hash value remaining (which is the center of the previously processed hash group). A similarity is calculated to each of the hub hash values stored in the hotspot repository 222 to determine whether the newly processed hash value belongs to an existing hash group. If the newly processed hash value belongs to one of the existing groups, then the newly processed hash value and its corresponding cropped image, recipe, topology information, etc. will be discarded because similar hotspots already exist in the hotspot repository 222. If the newly processed hash value is not similar to any stored central hash value, the newly processed hash value and its corresponding cropped image, recipe, topology information will be saved as a new entry in the hotspot repository 222. Through this process, the hotspot library 222 can be improved.

FIG. 13 shows a process flow 300, where the process flow 300 is a process flow for using the hotspot library 222 to determine possible hotspots in a new GDS file (new layout). Referring to process 302, a new GDS file (layout) is provided. Next, in process 304, the new GDS file is cropped into a plurality of cropped images, each of which has a size of L2 XW 2, as shown in FIG. 14. For example, FIG. 14 shows a design of layout 55, which layout 55 is divided into an array of images 56 having a length L2 and a width W2. According to some embodiments, length L2 is equal to length L1 (fig. 8), and width W2 is equal to width W1. According to other embodiments, length L2 may be greater than or less than length L1 (fig. 8), and width W2 may be greater than or less than width W1.

Referring to process 306, each cropped image 56 is processed to generate a hash value. The method of generating the hash value is similar to that discussed with reference to process 212 in fig. 6 and, therefore, is not repeated.

Referring to process 308 in FIG. 13, the (central) hash value stored in hotspot repository 222 is loaded into a computer and corresponding software. Each newly generated hash value is compared to each hash value loaded from the hotspot library 222 to compare their similarity, as shown in process 310. For example, when the newly generated hash value is similar to one loaded from the hotspot library, it is determined that the newly generated hash value and the corresponding cropped image have been represented by the similar hash value stored in the hotspot library 222 (process 312). It is also determined that hot spot(s) may be generated in the corresponding cropped image. Accordingly, the location of the respective image in the respective GDS file is marked (process 314). For example, when the trimming image is located in the 2 nd row and 3 rd column of the array divided from the layout 55 in fig. 14, the position (2, 3) will be marked. By comparing all newly generated hash values (of the cropped image) with all hash values in the library, a list of all hotspots (if any) in the GDS 55 (fig. 14) will be generated and the corresponding location of each hotspot will be marked in the corresponding GDS. The marked location may be used in the future. For example, after the CMP process, the portion of the chip at the marked location is examined to determine whether the hot spot has been successfully eliminated or at least reduced.

On the other hand, if none of the newly generated hash values are similar to any hash values loaded from the hotspot library, then it is determined that the newly generated hash values and the corresponding cropped image are not represented by any hotspot library entry in the hotspot library 222 (process 312). In other words, the hash value is not found from the GDS file, and the process may end (process 316). Thus, the layout can be manufactured without paying attention to the hot spot.

Fig. 15 shows a process flow 400 for determining and suggesting recipes for performing a CMP process in order to eliminate or at least reduce hot spots found in the process flow 300 shown in fig. 13. Referring to process 402, a list of hotspots generated in process 314 in process flow 300 (FIG. 13) is obtained. If no hotspot is found, the process flow ends. If one or more hotspots are found, a comparison is made to find hotspots in hotspot library 222 that are similar to the found hotspots. To perform the comparison, the hash values stored in the hotspot repository 222 are first loaded into the respective tools and computers, and similarity values between the found hash values of the hotspots and the respective hash values stored in the hotspot repository are calculated (by calculating the similarity values), as shown in process 406. Depending on the total number of hotspots found in the GDS, there may be one or more similarity values, each corresponding to one found hotspot. The hotspots (hash values thereof) are ranked according to the respective similarity values (process 406), with hotspots with higher similarity values having a higher ranking and higher priority than hotspots with lower similarity values.

Referring to process 408 in process flow 400, the GDS file, topology information, and corresponding recipe corresponding to the ranked hotspot are found from hotspot repository 222, for example, by indexing to the corresponding stored central hash value in hotspot repository 222. The topology information is analyzed and one of the found recipes is selected (process 410). According to some embodiments, the selected recipe is the recipe corresponding to the highest ranked hash value. According to some embodiments, the selected recipe is the recipe corresponding to the one hash value of the non-highest permutation, taking into account other factors. The selected recipe may thus be used to perform a CMP process on a corresponding physical wafer.

FIG. 16 shows a graphical representation of processes 406 and 410 in process flow 400 in FIG. 15. As shown in FIG. 16, a plurality of hash values (represented by a two-dimensional matrix thereof) of the found hotspots are generated, which corresponds to process 404 in FIG. 15. Next, according further to process 406 in FIG. 15, similarity values between the discovered hotspots and their respective representative hotspots in the hotspot library 222 are calculated and the discovered hotspots (and their hash values) are ranked. As shown in fig. 16, the order of the hash values shown is rearranged to display the arrangement. FIG. 16 also shows a plurality of recipes and GDS files corresponding to the arranged hash values. Next, one of the recipes is selected (process 410), and the selected recipe is recipe B in this example. In other embodiments, the formula with the highest ranking (formula a) may be selected.

FIG. 17 shows an example recipe for a CMP process. Each recipe may comprise (sub-) steps performed in the CMP process, as well as parameters to be used in each step. In the example shown, there are four steps, step 1, step 2, step 3 and step 4, each performed with a plurality of parameters, and the parameters are changed between the steps. For example, there may be head rotation (head rotation) a, head rotation B, slurry flow a (flow rate of the first slurry), slurry flow B (flow rate of the second slurry), down force of the wafer head (on the polishing pad), trim on/off (whether the pad conditioner is on or off), and so forth. The X-axis represents the time of the CMP process and the Y-axis represents the parameters and their corresponding values. For example, for each value, at any time, when a respective strip is present, the respective parameter is turned on, and if the strip is wide (in the Y direction), the respective parameter has a higher value. For example, slurry B has a higher flow rate in the initial stage of step 1, and is then turned off for the remainder of step 1. Slurry B has a relatively small flow rate throughout step 2 and is shut off throughout step 3 and throughout step 4. The pad conditioner (represented by trim on/off) is opened with a relatively low down force during steps 1 and 2 and is opened with a relatively high down force during steps 3 and 4. Throughout this specification, when a recipe is referred to as being adjusted, this means a combination of adjustment steps, parameters, and parameter values, which means that when any parameter is adjusted, the recipe is considered to be adjusted.

FIG. 18 illustrates a process flow 500 for improving the process of formulating and training the hotspot prevention model 223 (FIG. 6). Throughout the specification, the hotspot prevention model 223 may alternatively be referred to as a Machine Learning (ML) model, as the model may be improved by learning in the improvement process of the process flow 500. The modified recipes (recipes A, B, C and D) may be used as the stored recipe in process 206 in FIG. 6.

Referring to fig. 18, a plurality of GDSs (layouts) A, B, C and D are provided. GDS a, B, C, and D may be the same as each other, may be slightly different from each other, or may be completely different from each other. Each recipe improvement process will improve the recipe through iteration. For example, GDS a is provided (process 502) and fed to hotspot prevention model 223 (fig. 6), such that hotspots that may occur are generated and output by hotspot prevention model 223. Then, recipe A is suggested, and the generation of the suggested (selected) recipe A is shown in process 400 (FIG. 15). The selected recipe a is then used to perform a CMP process on the physical wafer implementing GDS a. Measurements are then performed on the wafer to determine hot spots and topology information on the wafer, and wafer results are generated (process 506). During the measurement, locations on the wafer, which are marked in process 314 (FIG. 13), will be examined to determine whether the various hot spots found in process 312 (FIG. 13) have been eliminated or at least reduced. If the hot spots are eliminated or at least reduced, then it is determined that the recipe is beneficial and may or may not be further improved. If the hot spots are not reduced or even worsened, then other recipes are needed.

Depending on the wafer results, the training data 508 (which may include the measurements) is fed back to the hotspot prevention model 223 and the hotspot prevention model 223 is updated (process 510). For example, when the measurements indicate that some new hotspots are found, or some expected hotspots do not exist, the hotspot prevention model will be updated so that the hotspot prevention model will output the newly found hotspots, and no longer output a model that does not exist.

Further, based on the measurement results, recipe a may be modified, for example, for eliminating the remaining found hot spots. Then, another wafer may be fabricated and CMP performed using the modified recipe a, and measurements may be performed to determine hotspot and topology information. This constitutes an iteration, and the iteration may continue until the result is satisfactory.

The GDS a, B, C, D may be the same as each other and the initial recipe A, B, C, D may be selected to be different so that the recipes may be improved in different directions and, ultimately, the best recipe may be selected among the modified recipes A, B, C and D (each recipe being modified in its own iteration). GDS A's, B, C, D that are different from each other may also be used so that the resulting model may cover different layouts and more recipes may be generated for different layouts.

FIG. 19 schematically illustrates a tool 600 for performing the tasks described above, including but not limited to computing, determining, and storing the hotspot library 222. For example, the processes illustrated in process flows 200, 300, 400, and 500 may all be performed using a computer (processor) 602 that includes hardware and software (computer program code). The program code of the tool 600 may be embodied on a non-transitory storage medium such as a hard disk drive, diskette, and the like. The hotspot library 222, which may be implemented in a storage device such as a hard disk, is electrically and signally connected to the computer 602 for saving and retrieval.

Embodiments of the present disclosure have some advantageous features. By predicting hot spots and selecting recipes to reduce/eliminate the predicted hot spots, hot spots can be found without manufacturing physical wafers and measuring the physical wafers. The first physical wafer may be manufactured using a recipe that is expected to eliminate potential hot spots. The manufacturing cycle time can be significantly reduced, for example, by one third.

According to some embodiments of the disclosure, a method comprises: cropping a plurality of images from a layout of an integrated circuit; generating a first plurality of hash values, each from one of a plurality of images; loading a second plurality of hash values stored in the hotspot library; comparing each of the first plurality of hash values to each of the second plurality of hash values, wherein comparing comprises calculating a similarity value between each of the first plurality of hash values and each of the second plurality of hash values; comparing the similarity value with a predetermined threshold similarity value; in response to a result of the similarity value being greater than a predetermined threshold similarity value, a location of the respective image having the result is recorded, wherein the location is a location of the respective image in the layout. In an embodiment, the plurality of images form an array, and the location includes a row number and a column number of the respective image in the array. In an embodiment, the method further comprises: fabricating integrated circuits on a wafer, wherein the fabricating includes performing a chemical mechanical polishing process on the wafer; and finding a hot spot from the location, wherein the hot spot is a defect in the wafer due to the chemical mechanical polishing process. In an embodiment, cropping the plurality of images comprises: the layout is divided into an array of images, and each of the plurality of images in the array is cropped. In an embodiment, the predetermined threshold similarity value is 0.9. In an embodiment, the method further comprises: cropping an additional plurality of images from an additional layout of the additional integrated circuit; generating a third plurality of hash values, each from one of the additional plurality of images; comparing each of the third plurality of hash values to all hash values stored in the hotspot library to find a set of hash values that are similar to the third plurality of hash values; and ranking the similarity values of the set of hash values. In an embodiment, the method further comprises selecting a recipe associated with one hash value of the set of hash values.

According to some embodiments of the disclosure, a method comprises: cropping a plurality of images from a layout of an integrated circuit; generating a plurality of hash values, each hash value from one of the plurality of images; searching from a hotspot repository to find similar hash values similar to the plurality of hash values, wherein the hotspot repository stores hash values indexed to images with hotspots; and marking locations of ones of the plurality of images associated with similar hash values on a layout of the integrated circuit. In an embodiment, the method further comprises: implementing a layout of an integrated circuit on a wafer, wherein implementing includes performing a chemical mechanical polishing process on the wafer using a recipe; the location on the wafer is inspected to determine the hot spot at that location. In an embodiment, the method further comprises: the recipe is determined based on the similar hash values that have been found. In an embodiment, the similar hash values are associated with a plurality of recipes in the hotspot library, and wherein the recipe is selected from the plurality of recipes. In an embodiment, the formulation comprises: a first duration and flow rate of the slurry used in the chemical mechanical polishing process, and respective second durations and magnitudes of the conditioning and downforce used in the chemical mechanical polishing process. In an embodiment, each image of the plurality of images has a square shape with a length and width in a range between about 64 μm and about 256 μm. In an embodiment, the hotspot library comprises a plurality of entries, each entry comprising a hash value, an image, a recipe, and topology information. In an embodiment, the hotspot library is indexed by a hash value.

According to some embodiments of the disclosure, a system comprises: a library stored in a tangible medium, the library comprising a plurality of entries, each entry comprising: a hash value; an image associated with the hash value, wherein the image includes a hotspot; a recipe configured to reduce hotspots; and topology information of the hotspot. In an embodiment, the system further comprises: a tool comprising software, wherein the software is configured to generate a hash value from an image. In an embodiment, the similarity value for any pair of hash values in the plurality of entries stored in the repository is less than about 0.9. In an embodiment, the hot spot comprises a groove or a protrusion that appears at the center of the image. In an embodiment, the recipe includes process conditions configured to reduce hot spots.

The foregoing has outlined features of several embodiments so that those skilled in the art may better understand the aspects of the present disclosure. Those skilled in the art should appreciate that they may readily use the present disclosure as a basis for designing or modifying other processes and structures for carrying out the same purposes and/or achieving the same advantages of the embodiments introduced herein. Those skilled in the art should also realize that such equivalent constructions do not depart from the spirit and scope of the present disclosure, and that they may make various changes, substitutions, and alterations herein without departing from the spirit and scope of the present disclosure.

Example 1. a method of manufacturing a semiconductor device, comprising: cropping a plurality of images from a layout of an integrated circuit; generating a first plurality of hash values, each from one of the plurality of images; loading a second plurality of hash values stored in the hotspot library; comparing each of the first plurality of hash values to each of the second plurality of hash values, wherein the comparing comprises calculating a similarity value between each of the first plurality of hash values and each of the second plurality of hash values; comparing the similarity value to a predetermined threshold similarity value; and in response to a result of the similarity value being greater than the predetermined threshold similarity value, recording a location of the respective image with the result, wherein the location is a location of the respective image in the layout.

Example 2. the method of example 1, wherein the plurality of images form an array, and the location includes a row number and a column number of the respective image in the array.

Example 3. the method of example 1, further comprising: fabricating the integrated circuit on a wafer, wherein the fabricating includes performing a chemical mechanical polishing process on the wafer; and finding a hot spot from the location, wherein the hot spot is a defect in the wafer resulting from the chemical mechanical polishing process.

Example 4. the method of example 1, wherein cropping the plurality of images comprises: the layout is divided into an array of images, and each image of the plurality of images in the array is cropped.

Example 5. the method of example 1, wherein the predetermined threshold similarity value is 0.9.

Example 6. the method of example 1, further comprising: cropping an additional plurality of images from an additional layout of the additional integrated circuit; generating a third plurality of hash values, each from one of the additional plurality of images; comparing each of the third plurality of hash values to all hash values stored in the hotspot library to find a set of hash values that are similar to the third plurality of hash values; and ranking the similarity values of the set of hash values.

Example 7. the method of example 6, further comprising: a recipe associated with one hash value in the set of hash values is selected.

Example 8 a method of fabricating a semiconductor device, comprising: cropping a plurality of images from a layout of an integrated circuit; generating a plurality of hash values, each hash value from one of the plurality of images; searching from a hotspot repository to find similar hash values similar to the plurality of hash values, wherein the hotspot repository stores hash values indexed to images with hotspots; and marking locations of ones of the plurality of images associated with the similar hash values on a layout of the integrated circuit.

Example 9. the method of example 8, further comprising: implementing a layout of the integrated circuit on a wafer, wherein the implementing includes performing a chemical mechanical polishing process on the wafer using a recipe; and inspecting the location on the wafer to determine a hotspot at the location.

Example 10. the method of example 9, further comprising: determining the recipe based on the similar hash values that have been found.

Example 11. the method of example 10, wherein the similar hash values are associated with a plurality of recipes in the hotspot library, and wherein the recipe is selected from the plurality of recipes.

Example 12. the method of example 9, wherein the recipe comprises: a first duration and flow rate of slurry used in the chemical mechanical polishing process, and respective second durations and magnitudes of conditioning and downforce used in the chemical mechanical polishing process.

Example 13. the method of example 8, wherein each image of the plurality of images has a square with a length and width in a range between 64 μ ι η and 256 μ ι η.

Example 14. the method of example 8, wherein the hotspot library comprises a plurality of entries, each entry comprising a hash value, an image, a recipe, and topological information.

Example 15. the method of example 14, wherein the hotspot library is indexed by a hash value.

Example 16. a system for fabricating a semiconductor device, comprising: a library stored in a tangible medium, the library comprising a plurality of entries, each entry comprising: a hash value; an image associated with the hash value, wherein the image comprises a hotspot; a recipe configured to reduce the hot spots; and topology information of the hotspot.

Example 17. the system of example 16, further comprising: a tool comprising software, wherein the software is configured to generate the hash value from the image.

Example 18. the system of example 16, wherein a similarity value of any pair of hash values in the plurality of entries stored in the repository is less than 0.9.

Example 19. the system of example 16, wherein the hotspot comprises a groove or protrusion that occurs at a center of the image.

Example 20. the system of example 16, wherein the recipe comprises process conditions configured to reduce the hot spot.