COMPUTER IMPLEMENTED METHOD, COMPUTER-READABLE MEDIUM, COMPUTER PROGRAM PRODUCT AND CORRESPONDING SYSTEMS FOR THE PREDICTION OF DEFECTS IN WAFERS AND FOR OFFLINE WAFER-LESS PROCESS WINDOW QUALIFICATION

Abstract

A computer implemented method for inspecting a photolithography mask to predict defects in wafers, the method comprising: providing a model of the photolithography mask comprising one or more target features and one or more sub-resolution assist features, the photolithography mask being configured for printing of the one or more target features onto a wafer in a printing process using a photolithography system; identifying one or more critical locations in the model of the photolithography mask by verifying a predefined constraint concerning the target features and/or the sub-resolution assist features; generating an aerial image of the photolithography mask comprising the one or more identified critical locations by applying a model of the photolithography system to the photolithography mask; predicting defects in wafers by comparing the one or more identified critical locations of the aerial image to one or more corresponding locations of a reference image. Also disclosed are a computer-readable medium, a computer program product and corresponding systems for the prediction of defects in wafers and wafer-less process window qualification.

Claims

1. A computer implemented method for inspecting a photolithography mask to predict defects in wafers, the method comprising: providing a model of the photolithography mask comprising one or more target features and one or more sub-resolution assist features, the photolithography mask being configured for printing of the one or more target features onto a wafer in a printing process using a photolithography system; identifying one or more critical locations in the model of the photolithography mask by verifying a predefined constraint concerning the target features and/or the sub-resolution assist features; generating an aerial image of the photolithography mask comprising the one or more identified critical locations by applying a model of the photolithography system to the photolithography mask; and predicting defects in wafers by comparing the one or more identified critical locations of the aerial image to one or more corresponding locations of a reference image.

2. The method according to claim 1, wherein the predicted defects comprise sub-resolution assist feature related defects, that is defects resulting from sub-resolution assist features.

3. The method according to claim 1, wherein the predefined constraint comprises a space constraint between target features and/or sub-resolution assist features.

4. The method according to claim 1, wherein the space constraint comprises a minimum distance between target features and/or sub-resolution assist features.

5. The method according to claim 1, wherein the predefined constraint comprises an area constraint defining a minimum area of target features and/or sub-resolution assist features.

6. The method according to claim 1, wherein the predefined constraint comprises a size constraint defining a minimum size of one or more dimensions of target features and/or sub-resolution assist features.

7. The method according to claim 1, wherein the predefined constraint comprises a predefined shape of a target feature and/or a sub-resolution assist feature.

8. The method according to claim 1, wherein the predefined constraint comprises a predefined pattern of target features and/or sub-resolution assist features.

9. The method according to claim 1, wherein the predefined constraint comprises a predefined location of target features and/or sub-resolution assist features.

10. The method according to claim 9, wherein the predefined location comprises an alignment feature, which is used to align the photolithography mask to a wafer.

11. The method according to claim 9, wherein the photolithography mask is associated with a layer of a semiconductor design, and the predefined location comprises a target feature, which overlaps with another target feature, in particular in a layer above or below the layer associated with the photolithography mask.

12. The method according to claim 1, wherein each target feature and each sub-resolution assist feature is associated with a metric value and the predefined constraint is associated with a limit value, and wherein a predefined number of critical locations is selected from the identified one or more critical locations according to the distance of the metric value of the corresponding target feature or sub-resolution assist feature to the limit value.

13. The method according to claim 1, wherein the photolithography mask is associated with a layer of a semiconductor design, and the aerial image simulates back-scattering of light from printed layers below the layer associated with the photolithography mask.

14. The method according to claim 1, wherein the model of the photolithography mask comprises a CAD model comprising the one or more target features and the one or more sub-resolution assist features.

15. The method according to claim 1, wherein the reference image comprises a software simulation of an aerial image of the model of the photolithography mask comprising the one or more identified critical locations.

16. The method according to claim 1, wherein the reference image comprises an aerial image of a photolithography mask obtained from a database, the aerial image comprising target features substantially identical to the target features in the one or more identified critical locations of the photolithography mask.

17. The method according to claim 1, wherein the reference image comprises an aerial image of a second photolithography mask comprising target features substantially identical to the target features in the one or more identified critical locations of the photolithography mask.

18. The method according to claim 1, wherein the reference image comprises an acquired aerial image or a simulated aerial image of a different portion of the photolithography mask comprising target features substantially identical to the target features in the one or more identified critical locations of the photolithography mask.

19. The method according to claim 1, wherein the reference image is generated by a machine learning model, in particular by an autoencoder.

20. The method according to claim 1, wherein comparing the aerial image to the reference image comprises the computation of at least one distance measure, in particular at least one distance metric.

21. The method according to claim 1, wherein comparing the aerial image to the reference image comprises the application of a machine learning model.

22. The method according to claim 1, further comprising using the predicted defects to modify the model of the photolithography mask, in particular the sub-resolution assist features of the model of the photolithography mask.

23. The method according to claim 1, further comprising using the predicted defects to repair the photolithography mask.

24. A computer implemented method for wafer-less process window qualification for a photolithography mask and a photolithography system comprising: iterating the following steps: for at least one photolithography process parameter of the photolithography system selecting a photolithography process parameter value from a predefined range of photolithography process parameter values; inspecting the photolithography mask to predict defects in wafers using a computer-implemented method for inspecting a photolithography mask to predict defects in wafers according to claim 1, wherein the aerial image of the photolithography mask is generated based on the at least one selected photolithography process parameter value; and providing a process window for the at least one photolithography process parameter by analyzing the predicted defects for different photolithography process parameter values.

25. A computer-readable medium, having stored thereon a computer program executable by a computing device, the computer program comprising code for executing a method of claim 1.

26. A computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out a method of claim 1.

27. A photolithography mask inspection system for the prediction of defects, the system comprising: a subsystem for generating an aerial image of the photolithography mask; and a data analysis device comprising at least one memory and at least one processor configured to perform the steps of a computer implemented method of claim 1.

28. A wafer-less process window qualification system comprising: a subsystem for generating an aerial image of the photolithography mask; and a data analysis device comprising at least one memory and at least one processor configured to perform the steps of a computer implemented method of claim 24.

Description

BRIEF DESCRIPTION OF DRAWINGS

[0051] FIG. 1A shows a target pattern comprising target features within a model of a photolithography mask;

[0052] FIG. 1B shows the result of printing the target pattern in FIG. 1A to a wafer;

[0053] FIG. 1C illustrates an example of an OPC technique comprising the placement of sub-resolution assist features close to the target features yielding a corrected pattern;

[0054] FIG. 1D shows the result of printing the corrected pattern in FIG. 1C to a wafer, the printed corrected pattern comprising an SRAF related defect in the form of a bridge defect;

[0055] FIG. 2 shows a flowchart of a computer implemented method for inspecting a photolithography mask to predict defects, in particular sub-resolution assist feature related defects, in wafers according to a first embodiment of the invention;

[0056] FIG. 3 illustrates the prediction of defects based on a comparison of an aerial image of the photolithography mask to a reference image;

[0057] FIG. 4 illustrates the simulation of back-scattering during aerial image generation;

[0058] FIG. 5 shows a flowchart of a computer implemented method for wafer-less process window qualification for a photolithography mask and a photolithography system according to a second embodiment of the invention;

[0059] FIG. 6 illustrates a photolithography mask inspection system for the prediction of defects according to a fifth embodiment of the invention; and

[0060] FIG. 7 illustrates a wafer-less process window qualification system according to a sixth embodiment of the invention.

DETAILED DESCRIPTION

[0061] In the following, advantageous exemplary embodiments of the invention are described and schematically shown in the figures. Throughout the figures and the description, same reference numbers are used to describe same features or components. Dashed lines indicate optional elements.

[0062] FIG. 1A shows a target pattern 12 comprising target features 14 within a model 10 of a photolithography mask 76. Due to the sub-wavelength size of the target features 14, optical proximity effects occur during the photolithography process. Optical proximity effects are caused by the non-uniformity of energy intensity due to optical diffraction during the exposure process. As a result, images formed on the wafer do not fully reproduce the target pattern 12, since they deviate from their ideal dimensions and shape indicated by the model 10 of the photolithography mask 76. FIG. 1B shows the result of printing the target pattern 12 in FIG. 1A to a wafer. The printed pattern 16 comprises printed features 18, which deviate from the target features 14 due to optical proximity effects leading to, e.g., rounded edges, line thinning and line shortening. These deviations can significantly affect the performance of the semiconductor device. To compensate for these effects, optical proximity correction (OPC) techniques are applied to the model 10 of the photolithography mask 76. FIG. 1C illustrates an example of an OPC technique comprising the placement of sub-resolution assist features (SRAFs) 22 close to the target features 14 yielding a corrected pattern 20. SRAFs 22 may be provided in the form of scattering bars as shown in FIG. 1C. Since these additional features are sub-photolithographic, they will not be transferred to the wafer during printing. However, the placement of SRAFs 22 is difficult as there are infinitely many target feature combinations, and the SRAF models usually used for SRAF placement cannot account for all of these combinations. Therefore, defects 28 can occur on the wafer. For example, if SRAFs 22 are too close to a target feature 14 or to another SRAF 22, a defect 28 can occur on the wafer. Such defects 28 are SRAF related defects since they result from the placement of SRAFs. FIG. 1D shows the result of printing the corrected pattern 20 in FIG. 1C to a wafer, the printed corrected pattern 26 comprising an SRAF related defect 28 in the form of a bridge defect. In addition, the placement of SRAFs 22 influences the process window of the parameters of the manufacturing process. Hence, the placement of SRAFs 22 can be detrimental to the process window qualification and cause defects 28 in the wafer during the subsequent wafer manufacturing process. However, such SRAF related defects 28 are neither visible in the model 10 of the photolithography mask 76 nor can they be detected solely from an analysis of the wafer.

[0063] Process window qualification is commonly done by printing a number of wafers under different focus and exposure conditions and determining a window for each of these parameters within which the printing result is acceptable. However, this process is time consuming and occupies important resources of the fabs.

[0064] The novel methods and systems described herein solve these problems by acquiring an aerial image 44 of the photolithography mask 76. The generation of the aerial image 44 comprises separating target features 14 from SRAFs 22 based on optical intensity distributions in the aerial image 44 and excluding the SRAFs 22 from the aerial image 44. Based on the aerial image 44 defects 28 in the wafer can be predicted without actually printing the wafer. This information can be used for the prediction of defects 28, in particular SRAF related defects, as well as for wafer-less process window qualification. In order to obtain an efficient defect prediction method, critical locations 24 are identified in the model 10 of the photolithography mask 76, and only these critical locations 24 are further inspected for potential defects 28.

[0065] FIG. 2 shows a flowchart of a computer implemented method 30 for inspecting a photolithography mask 76 to predict defects 28, in particular SRAF related defects, in wafers, according to a first embodiment of the invention, the method comprising: in a model providing step 32, providing a model 10 of the photolithography mask 76 comprising one or more target features 14 and one or more SRAFs 22, the photolithography mask 76 being configured for printing of the one or more target features 14 onto a wafer in a printing process using a photolithography system; identifying one or more critical locations 24 in the model 10 of the photolithography mask 76 by verifying a predefined constraint concerning the target features 14 and/or the SRAFs 22 in a critical location identification step 34; generating an aerial image 44 of the photolithography mask 76 comprising the one or more identified critical locations 24 by applying a model of the photolithography system to the photolithography mask 76 in an aerial image generation step 36; and predicting defects 28 in wafers by comparing the one or more identified critical locations 24 of the aerial image 44 to one or more corresponding locations of a reference image 46 in a defect prediction step 38. This way, the defects 28 can be predicted without actually printing the target features 14 onto the wafers.

[0066] The predicted defects 28 can, optionally, be used in different ways, e.g., for modification of the model 10 of the photolithography mask 76 in a model modification step 40, for photolithography mask repair in a repair step 42 or for process-based mitigation in a process step 43.

[0067] In an example, the method, optionally comprises using the predicted defects 28 to modify the model 10 of the photolithography mask 76, in particular the one or more SRAFs 22 of the model 10 of the photolithography mask 76, in a model modification step 40. For example, the location and/or size of the target features 14 and/or the one or more SRAFs 22 can be modified, or target features 14 and/or SRAFs 22 can be removed from the model 10 of the photolithography mask 76, or target features 14 and/or SRAFs 22 can be added to the model 10 of the photolithography mask 76.

[0068] In an example, the method optionally comprises using the predicted defects 28 to repair the photolithography mask 76 in a repair step 42. To this end, the photolithography mask 76 is sent to a photolithography mask repair system.

[0069] In some implementations, the photolithography mask repair system can be configured to perform an electron beam-induced etching and/or deposition on the mask. The repair system can include, e.g., an electron source, which emits an electron beam that can be used to perform electron beam-induced etching or deposition on the mask. The repair system can include mechanisms for deflecting, focusing and/or adapting the electron beam. The repair system can be configured such that the electron beam is able to be incident on a defined point of incidence on the mask.

[0070] The repair system can include one or more containers for providing one or more deposition gases, which can be guided to the mask via one or more appropriate gas lines. The repair system can also include one or more containers for providing one or more etching gases, which can be provided on the mask via one or more appropriate gas lines. Further, the repair system can include one or more containers for providing one or more additive gases that can be supplied to be added to the one or more deposition gases and/or the one or more etching gases.

[0071] The repair system can include a user interface to allow an operator to, e.g., operate the repair system and/or read out data. The repair system can include a computer unit configured to cause the repair system to perform one or more of the methods described herein, based at least in part on an execution of an appropriate computer program. The repair system can also repair other types of objects (e.g., wafers) having integrated circuit patterns.

[0072] In an example, the method optionally comprises a process-based mitigation of the predicted defects 28 in a process mitigation step 43. During the process mitigation step 43 the predicted defects 28 are used to modify parameters of the manufacturing process, e.g., parameters of the lithography, the etching or the CMP process.

[0073] Based on the predefined constraint, critical locations 24 can be selected from the model 10 of the photolithography mask 76, and the prediction of defects 28 can be limited to these critical locations 24. In this way, the runtime of the defect prediction method can be reduced, since only the critical locations 24 are analyzed for defects. In addition, the defect analysis can be limited to specific defect types, e.g., bridge defects or overlay defects.

[0074] In an example of the first embodiment of the invention, the model 10 of the photolithography mask 76 comprises a CAD model comprising the one or more target features 14 and the one or more SRAFs 22. The model 10 of the photolithography mask 76 can be loaded from a database 86.

[0075] According to an example of the first embodiment of the invention, the predefined constraint comprises a space constraint between target features 14 and/or SRAFs 22. For example, the predefined constraint comprises a minimum distance between target features 14 and/or SRAFs 22, preferably a minimum distance below 60 nm, more preferably below 50 nm, most preferably below 40 nm. Target features 14 and/or SRAFs 22 having a distance below the minimum distance can, thus, be assigned to one or more critical locations 24. In this way, target features 14 and/or SRAFs 22 that are very close to each other and potentially interfere during printing form one or more critical locations 24, which are further analyzed with respect to the aerial image 44 of these critical locations 24. Alternatively, target features 14 or SRAFs 22 having a distance above a predetermined maximum distance to all other target features 14 or SRAFs 22 can be assigned to one or more critical locations 24. In this way, isolated target features 14 and/or SRAFs 22 are assigned to a critical location 24, which is further analyzed with respect to the aerial image 44.

[0076] According to an example of the first embodiment of the invention, the predefined constraint comprises an area constraint defining a minimum area of target features 14 and/or SRAFs 22. In this way, target features 14 and/or SRAFs 22 of small area, which could accidentally be removed during printing, are assigned to one or more critical locations 24.

[0077] According to an example of the first embodiment of the invention, the predefined constraint comprises a size constraint defining a minimum size of one or more dimensions of target features 14 and/or SRAFs 22. In this way, for example, design rules referring to feature sizes can be verified and compliance can be checked. For example, a design rule indicating a minimum feature size in one or more dimensions, e.g., 30 nm, can be checked.

[0078] According to an example of the first embodiment of the invention, the predefined constraint comprises a predefined shape of a target feature 14 and/or an SRAF 22. In this way, target features 14 and/or SRAFs 22 of a specific shape, which are known to the fab to cause defects 28 can be assigned to one or more critical locations 24. For example, L-shaped features or T-shaped features can be assigned to one or more critical locations 24.

[0079] According to an example of the first embodiment of the invention, the predefined constraint comprises a predefined pattern of target features 14 and/or SRAFs 22. In this way, specific patterns known to cause defects 28, e.g., logical patterns, form one or more critical locations 24.

[0080] According to an example of the first embodiment of the invention, the predefined constraint comprises a predefined location of target features 14 and/or SRAFs 22. In this way, for example, desired as well as undesired overlays of patterns can be verified. For example, the predefined location can comprise an alignment feature, which is used to align the photolithography mask 76 to a wafer. In this way, the correct printing of alignment features can be verified. In another example, the predefined location comprises a target feature 14, which overlaps with another target feature 14, in particular in a layer above or below the layer associated with the photolithography mask 76. In this way, target features 14 connecting different layers of the semiconductor device can be assigned to one or more critical locations 24. The predefined location can also comprise a target feature 14 and another overlapping target feature 14, which together form a structure and, thus, must be connected, e.g., a T-structure containing two straight lines, which must overlap to create a connection. This situation can, for example, occur in case of multi patterning. In another example, undesired connections between target features 14 can be checked, e.g., a target feature 14 of a metal layer and another target feature 14 of a VIA layer above or below, which must not be connected.

[0081] According to an example of the first embodiment of the invention, each target feature 14 and each SRAF 22 is associated with a metric value and the predefined constraint is associated with a limit value, and a predefined number of critical locations 24 is selected from the identified one or more critical locations 24 according to the distance of the metric value of the corresponding target features 14 or SRAFs 22 to the limit value. For example, the target features 14 and SRAFs 22 corresponding to a critical location 24 are associated with a metric value such as size, distance, area etc., and the predefined constraint comprises a minimum or maximum value, and the critical locations 24 are sorted according to the distance of the corresponding metric value to the minimum or maximum value, and the first elements from the list are selected. For example, if the predefined constraint comprises a minimum feature size of 30 nm, then all features below this feature size are identified as critical locations 24. Then the critical locations 24 are sorted according to increasing feature size, and the first N critical locations 24 are selected from this list. In this way, the most critical locations 24 among the critical locations 24 are selected for further analysis.

[0082] Optionally, common methods for etch and resist correction known to a person skilled in the art can be applied to the model 10 of the photolithography mask 76.

[0083] FIG. 3 illustrates the prediction of defects 28 based on a comparison of an aerial image 44 of the photolithography mask 76 to a reference image 46. To this end, the aerial image comprises predicted features 48, which are compared to the reference features 50 in the reference image 46. The comparison of the aerial image 44 to the reference image 46 can be carried out in various ways. For example, a difference image 51 of the aerial image 44 and the reference image 46 can be computed.

[0084] According to an example of the first embodiment of the invention, comparing the aerial image 44 to the reference image 46 comprises the computation of at least one distance measure, in particular at least one distance metric. For example, the pixel-wise sum of squared differences can be computed, which corresponds to the squared L.sub.2-metric.

[0085] According to an example of the first embodiment of the invention, comparing the aerial image 44 to the reference image 46 comprises the application of a machine learning model. The machine learning model, e.g., a convolutional neural network, can be trained to predict defects 28 from input images comprising the aerial image 44 and/or the reference image 46 and/or the difference image 51. The machine learning model can, for example, be trained on a number of pairs of aerial images 44 and reference images 46 with indicated defects 28. Alternatively, the machine learning model can be trained on difference images 51. The defect prediction can be used as a distance measure.

[0086] There are different ways of generating a reference image 46.

[0087] According to an example of the first embodiment of the invention, the reference image 46 comprises a software simulation of an aerial image of the model of the photolithography mask 76 comprising the one or more identified critical locations 24. The software simulates the photolithography process based on the model 10 of the photolithography mask 76. The simulated aerial image is compared to the acquired aerial image 44 and defects 28 are predicted based on the difference image 51.

[0088] According to an example of the first embodiment of the invention, the reference image 46 comprises an aerial image 44 of a photolithography mask 76 obtained from a database 86, the aerial image 44 comprising target features 14 substantially identical to the target features 14 in the one or more identified critical locations 24 of the photolithography mask 76. In this way, acquired aerial images 44 or software-based simulations of aerial images 44 comprising substantially identical target features 14 are used for the comparison. This way of comparing the aerial image 44 to a reference image 46 is a die-to-database method. The reference image 46 in the database 86 can, for example, comprise a simulation of the ideal target features 14 as modeled by the designer. Since the ideal target features 14 are known, each defect 28 can be predicted correctly during the comparison. In this way, the accuracy of the predicted defects 28 is improved. However, registration of the reference image 46 and the aerial image 44 is required. In this document, substantially identical means identical up to small variations that would not be seen as a defect or a different structure, e.g., including effects such as line edge roughness, corner rounding, noise, etc.

[0089] According to an example of the first embodiment of the invention, the reference image 46 comprises an aerial image 44 of a second photolithography mask 76 comprising target features 14 substantially identical to the target features 14 in the one or more identified critical locations 24 of the photolithography mask 76. This way of comparing the aerial image 44 to a reference image 46 is a die-to-die method. In this way, the accuracy of the predicted defects 28 is improved. As there is no need for a model of the ideal target features 14, the effort for the applicant is reduced. However, defects 28 present in both photolithography masks 76 may not be predicted, and registration of the reference image 46 and the aerial image 44 is still required.

[0090] According to an example of the first embodiment of the invention, the reference image 46 comprises an acquired aerial image 44 or a simulated aerial image 44 of a different portion of the photolithography mask 76 comprising target features 14 substantially identical to the target features 14 in the one or more identified critical locations 24 of the photolithography mask 76. The simulated aerial image 44 can be generated by use of a software-based simulation. This way of comparing the aerial image 44 to a reference image 46 is an intra-die method. In this way, the accuracy of the predicted defects 28 is improved. As there is no need for a model of the ideal target features 14 nor for a second simulated or acquired aerial image, the effort for the applicant is even further reduced. However, repeater defects occurring in various locations of the photolithography mask 76 may not be predicted, and registration of the different portions of the photolithography mask 76 is still required.

[0091] According to an example of the first embodiment of the invention, the reference image 46 is generated by a machine learning model, in particular by an autoencoder. The machine learning model can be trained to generate the reference image 46 from the aerial image 44 based on training data comprising substantially defect-free aerial images 44. Substantially defect-free here means that less than 5% of the pixels of each aerial image 44 are part of a defect 28. By comparing the aerial image 44 to the machine learning model generated reference image 46, defects 28 can be predicted. This way of comparing the aerial image 44 to a reference image 46 is a single-die method. In this way, the accuracy of the predicted defects 28 is improved. At the same time, the effort for the applicant for defect prediction is reduced most, as no reference image 46 needs to be provided, nor is registration necessary.

[0092] The machine learning model can comprise an autoencoder. An autoencoder is a type of artificial neural network used in unsupervised learning to learn efficient codings of unlabeled data. An autoencoder comprises two main parts: an encoder that maps the input into a code, and a decoder that maps the code to a reconstruction of the input. The input of the autoencoder can, for example, comprise an aerial image 44 or a subset thereof. The encoder neural network and the decoder neural network can be trained to minimize a difference between the reconstructed representation of the input and the input itself. The code typically is a representation of the input with lower dimensionality and can, thus, be viewed as a compressed version of the input. For this reason, autoencoders are forced to reconstruct the input approximately, preserving only the most relevant aspects of the input in the reconstruction.

[0093] Therefore, autoencoders can be used for the prediction of defects 28. Defects 28 generally concern rare deviations from the norm within an aerial image 44. Due to the rarity of their occurrence the autoencoder will not reconstruct this kind of information, thus suppressing defects 28 in the reconstruction. Defects 28 can then be predicted by comparing the imperfect reconstruction of the aerial image 44 to the aerial image 44. The larger the difference between them, the more likely a defect 28 is present.

[0094] Instead of applying the autoencoder to the full aerial image 44, the aerial image 44 can be subdivided into smaller subsets. The autoencoder can then be applied to each of the subsets separately. In this way, the required complexity of the autoencoder is reduced due to the reduced size of the input, thus improving the accuracy of the reconstruction and the defect detection.

[0095] According to an example of the first embodiment of the invention, the photolithography mask 76 is associated with a layer of a semiconductor design, and the aerial image 44 simulates back-scattering 54 of light from printed layers below the layer associated with the photolithography mask 76. FIG. 4 illustrates the simulation of back-scattering 54 during aerial image generation. Incoming light 52 due to the application of photolithography mask n+1 is reflected by the previously printed layers, that is from the first layer 56, the second layer 58 and all layers up to the n-th layer 60.

[0096] Commonly, defect detection methods in wafers analyze short loop wafers, i.e., only single layers of a wafer without taking into account light reflected from the layers below. Therefore, by simulating back-scattering 54 from layers below the layer associated with the photolithography mask 76 the aerial image 44 allows for a more accurate simulation of the printed semiconductor structures on the wafer and, therefore, allows for a more accurate prediction of defects 28.

[0097] Traditionally, process window definition forms one of the most important steps of high-volume manufacturing. It takes about two weeks to qualify a process window for a particular part. Due to the ever-shrinking integrated circuit structures, this process has become more chip specific. Even within a technology node, multiple qualifications for multiple chips need to be done. Currently there is no way to predict the process window at a theoretical or simulation level. Thus, fabs have to spare tools, time and resources for this exploratory work. Therefore, a wafer-less process window qualification method is desirable. Such a method can be obtained based on the computer implemented method for inspecting a photolithography mask 76 to predict defects 28 in wafers according to the first embodiment of the invention.

[0098] FIG. 5 shows a flowchart of a computer implemented method 62 for wafer-less process window qualification for a photolithography mask 76 and a photolithography system according to a second embodiment of the invention, the method comprising an iteration 66 of the following steps: for at least one photolithography process parameter of the photolithography system selecting a photolithography process parameter value from a predefined range of photolithography process parameter values in a parameter selection step 64, and, in a defect prediction step 68, inspecting the photolithography mask 76 to predict defects 28 in wafers using any of the computer-implemented methods 30 for inspecting a photolithography mask 76 to predict defects 28 in wafers described above, wherein the aerial image 44 of the photolithography mask 76 is generated based on the at least one selected photolithography process parameter value. After the final iteration, a process window for the at least one photolithography process parameter is determined by analyzing the predicted defects 28 for different photolithography process parameter values in a process window qualification step 70. Based on the qualified process window, parameter values for the parameters of the photolithography process can be selected and applied during high volume manufacturing of semiconductor devices.

[0099] Due to the simulation of the photolithography process for the photolithography mask 76 and the photolithography system by generating the aerial image 44, an actual printing of wafers is not required for process window qualification. Thus, the process window qualification can be carried out offline. Therefore, the fabs save a lot of time, tools and resources.

[0100] The above-described wafer-less process window qualification method according to the second embodiment of the invention has several advantages: firstly, the process window qualification method is based on an aerial image 44 generated from the photolithography mask 76 itself instead of a mere model 10 of the photolithography mask 76. Thus, error sources such as mask imperfections and registration errors are taken into account leading to defect predictions of improved accuracy. Secondly, random process defects during wafer printing mostly resulting from nuisances, which are not detrimental to the functioning of the semiconductor device, do not influence the process window qualification. Thirdly, the aerial image 44 takes into account intensity models of the layers below the layer associated with the photolithography mask 76 and, thus, predicts defects 28 with an improved accuracy. Fourthly, stochastic effects of the scanner, which pose a common problem during wafer-based process window qualification, do not influence the wafer-less process window qualification.

[0101] The wafer-less process window qualification can be a substitute for wafer-based process window qualification, or it can be used as a reference process for subsequent processes.

[0102] According to a third embodiment of the invention, a computer-readable medium is disclosed, having stored thereon a computer program executable by a computing device, the computer program comprising code for executing any of the methods described above.

[0103] According to a fourth embodiment of the invention, a computer program product is disclosed comprising instructions which, when the program is executed by a computer, cause the computer to carry out any of the methods described above.

[0104] FIG. 6 illustrates a photolithography mask inspection system 72 for the prediction of defects 28 according to a fifth embodiment of the invention, the system comprising: a subsystem 74 for generating an aerial image 44 of the photolithography mask 76, and a data analysis device 78 comprising at least one memory 80 and at least one processor 82 configured to perform the steps of a computer implemented method 30 according to any of the methods for inspecting a photolithography mask 76 to predict defects 28 in wafers described above. The subsystem 74 can include, e.g., a charged coupled device (CCD) image sensor or a complementary metal-oxide-semiconductor (CMOS) image sensor having an array of pixels or light sensing elements. The aerial image 44 is transferred to the data analysis device 78 via an interface 84. The photolithography mask inspection system 72 can comprise a database 86 and/or a user interface 88.

[0105] The subsystem 74 can provide an aerial image 44 to the data analysis device 78. The data analysis device 78 includes a processor 82, e.g., implemented as a central processing unit (CPU) or graphics processing unit (GPU). The processor 82 can receive the aerial image 44 via an interface 84. The processor 82 can load program code from a memory 80. The processor 82 can execute the program code. Upon executing the program code, the processor 82 performs techniques such as described herein, e.g., inspecting a photolithography mask 76 to predict defects 28 in wafers, computing an aerial image 44, a reference image 46 or a difference image 51, evaluating a predefined constraint, etc. For example, the processor 82 can perform any of the above-described examples of the computer implemented method 30 shown in FIG. 2 upon loading program code from the memory 80.

[0106] FIG. 7 illustrates a wafer-less process window qualification system 90 comprising a subsystem 74 for generating an aerial image 44 of the photolithography mask 76 and a data analysis device 78 comprising at least one memory 80 and at least one processor 82 configured to perform the steps of the computer implemented method 62 for wafer-less process window qualification for a photolithography mask 76 and a photolithography system as described above. The aerial image 44 is transferred to the data analysis device 78 via an interface 84. The photolithography mask inspection system 72 can comprise a database 86 and/or a user interface 88.

[0107] The subsystem 74 can provide an aerial image 44 to the data analysis device 78. The data analysis device 78 includes a processor 82, e.g., implemented as a CPU or GPU. The processor 82 can receive the aerial image 44 via an interface 84. The processor 82 can load program code from a memory 80. The processor 82 can execute the program code. Upon executing the program code, the processor 82 performs techniques such as described herein, e.g., wafer-less process window qualification for a photolithography mask 76 and a photolithography system. For example, the processor 82 can perform any of the above-described examples of the computer implemented method 62 shown in FIG. 5 upon loading program code from the memory 80.

[0108] In some implementations, a computing device can include one or more data processors. Each data processor can include one or more processor cores, and each processor core can include logic circuitry for processing data. For example, a data processor can include an arithmetic and logic unit (ALU), a control unit, and various registers. Each data processor can include cache memory. Each data processor can include a system-on-chip (SoC) that includes multiple processor cores, random access memory, graphics processing units, one or more controllers, and one or more communication modules. Each data processor can include millions or billions of transistors.

[0109] In some implementations, the program code can be stored in one or more machine-readable storage devices, such as hard drives, magnetic disks, solid state drives, magneto-optical disks, or optical disks. Alternatively or in addition, the program code can be encoded on a propagated signal that is an artificially generated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a programmable processor.

[0110] The program code can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0111] In some implementations, the processes described above can be implemented using software for execution on one or more mobile computing devices, one or more local computing devices, and/or one or more remote computing devices (which can be, e.g., cloud computing devices). For instance, the software forms procedures in one or more computer programs that execute on one or more programmed or programmable computer systems, either in the mobile computing devices, local computing devices, or remote computing systems (which may be of various architectures such as distributed, client/server, grid, or cloud), each including at least one data processor, at least one data storage system (including volatile and non-volatile memory and/or storage elements), at least one wired or wireless input device or port, and at least one wired or wireless output device or port.

[0112] Reference throughout this specification to an embodiment or an example or an aspect means that a particular feature, structure or characteristic described in connection with the embodiment, example or aspect is included in at least one embodiment, example or aspect. Thus, appearances of the phrases according to an embodiment, according to an example or according to an aspect in various places throughout this specification are not necessarily all referring to the same embodiment, example or aspect, but may refer to different embodiments. Furthermore, the particular features or characteristics may be combined in any suitable manner, as would be apparent to one of ordinary skill in the art from this disclosure, in one or more embodiments.

[0113] Furthermore, while some embodiments, examples or aspects described herein include some but not other features included in other embodiments, examples or aspects combinations of features of different embodiments, examples or aspects are meant to be within the scope of the claims, and form different embodiments, as would be understood by those skilled in the art.

[0114] Embodiments and examples of the invention can be described by the following clauses: [0115] 1. A computer implemented method 30 for inspecting a photolithography mask 76 to predict defects 28 in wafers, the method comprising: [0116] Providing a model 10 of the photolithography mask 76 comprising one or more target features 14 and one or more sub-resolution assist features 22, the photolithography mask 76 being configured for printing of the one or more target features 14 onto a wafer in a printing process using a photolithography system; [0117] Identifying one or more critical locations 24 in the model 10 of the photolithography mask 76 by verifying a predefined constraint concerning the target features 14 and/or the sub-resolution assist features 22; [0118] Generating an aerial image 44 of the photolithography mask 76 comprising the one or more identified critical locations 24 by applying a model of the photolithography system to the photolithography mask 76; [0119] Predicting defects 28 in wafers by comparing the one or more identified critical locations 24 of the aerial image 44 to one or more corresponding locations of a reference image 46. [0120] 2. The method according to clause 1, wherein the predicted defects 28 comprise sub-resolution assist feature related defects, that is defects 28 resulting from sub-resolution assist features 22. [0121] 3. The method according to clause 1 or 2, wherein the predefined constraint comprises a space constraint between target features 14 and/or sub-resolution assist features 22. [0122] 4. The method according to any one of the preceding clauses, wherein the space constraint comprises a minimum distance between target features 14 and/or sub-resolution assist features 22, preferably a minimum distance below 60 nm, more preferably below 50 nm, most preferably below 40 nm. [0123] 5. The method according to any one of the preceding clauses, wherein the predefined constraint comprises an area constraint defining a minimum area of target features 14 and/or sub-resolution assist features 22. [0124] 6. The method according to any one of the preceding clauses, wherein the predefined constraint comprises a size constraint defining a minimum size of one or more dimensions of target features 14 and/or sub-resolution assist features 22. [0125] 7. The method according to any one of the preceding clauses, wherein the predefined constraint comprises a predefined shape of a target feature 14 and/or a sub-resolution assist feature 22. [0126] 8. The method according to any one of the preceding clauses, wherein the predefined constraint comprises a predefined pattern of target features 14 and/or sub-resolution assist features 22. [0127] 9. The method according to any one of the preceding clauses, wherein the predefined constraint comprises a predefined location of target features 14 and/or sub-resolution assist features 22. [0128] 10. The method according to clause 9, wherein the predefined location comprises an alignment feature, which is used to align the photolithography mask 76 to a wafer. [0129] 11. The method according to clause 9, wherein the photolithography mask 76 is associated with a layer of a semiconductor design, and the predefined location comprises a target feature 14, which overlaps with another target feature 14, in particular in a layer above or below the layer associated with the photolithography mask 76. [0130] 12. The method according to any one of the preceding clauses, wherein each target feature 14 and each sub-resolution assist feature 22 is associated with a metric value and the predefined constraint is associated with a limit value, and wherein a predefined number of critical locations 24 is selected from the identified one or more critical locations 24 according to the distance of the metric value of the corresponding target feature 14 or sub-resolution assist feature 22 to the limit value. [0131] 13. The method according to any one of the preceding clauses, wherein the photolithography mask 76 is associated with a layer of a semiconductor design, and the aerial image 44 simulates back-scattering 54 of light 52 from printed layers below the layer associated with the photolithography mask 76. [0132] 14. The method according to any one of the preceding clauses, wherein the model 10 of the photolithography mask 76 comprises a CAD model comprising the one or more target features 14 and the one or more sub-resolution assist features 22. [0133] 15. The method according to any one of the preceding clauses, wherein the reference image 46 comprises a software simulation of an aerial image 44 of the model 10 of the photolithography mask 76 comprising the one or more identified critical locations 24. [0134] 16. The method according to any one of the preceding clauses, wherein the reference image 46 comprises an aerial image 44 of a photolithography mask 76 obtained from a database 86, the aerial image 44 comprising target features 14 substantially identical to the target features 14 in the one or more identified critical locations 24 of the photolithography mask 76. [0135] 17. The method according to any one of the preceding clauses, wherein the reference image 46 comprises an aerial image 44 of a second photolithography mask 76 comprising target features 14 substantially identical to the target features 14 in the one or more identified critical locations 24 of the photolithography mask 76. [0136] 18. The method according to any one of the preceding clauses, wherein the reference image 46 comprises an acquired aerial image 44 or a simulated aerial image 44 of a different portion of the photolithography mask 76 comprising target features 14 substantially identical to the target features 14 in the one or more identified critical locations 24 of the photolithography mask 76. [0137] 19. The method according to any one of the preceding clauses, wherein the reference image 46 is generated by a machine learning model, in particular by an autoencoder. [0138] 20. The method according to any one of the preceding clauses, wherein comparing the aerial image 44 to the reference image 46 comprises the computation of at least one distance measure, in particular at least one distance metric. [0139] 21. The method according to any one of the preceding clauses, wherein comparing the aerial image 44 to the reference image 46 comprises the application of a machine learning model. [0140] 22. The method according to any one of the preceding clauses, further comprising using the predicted defects 28 to modify the model 10 of the photolithography mask 76, in particular the sub-resolution assist features 22 of the model 10 of the photolithography mask 76. [0141] 23. The method according to any one of the preceding clauses, further comprising using the predicted defects 28 to repair the photolithography mask 76. [0142] 24. A computer implemented method 62 for wafer-less process window qualification for a photolithography mask 76 and a photolithography system comprising: [0143] Iterating the following steps: [0144] For at least one photolithography process parameter of the photolithography system selecting a photolithography process parameter value from a predefined range of photolithography process parameter values; [0145] Inspecting the photolithography mask 76 to predict defects 28 in wafers using a computer-implemented method 30 for inspecting a photolithography mask 76 to predict defects 28 in wafers according to any one of clauses 1 to 21, wherein the aerial image 44 of the photolithography mask 76 is generated based on the at least one selected photolithography process parameter value; [0146] Providing a process window for the at least one photolithography process parameter by analyzing the predicted defects 28 for different photolithography process parameter values. [0147] 25. A computer-readable medium, having stored thereon a computer program executable by a computing device, the computer program comprising code for executing a method of any one of clauses 1 to 24. [0148] 26. A computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out a method of any one of clauses 1 to 24. [0149] 27. A photolithography mask inspection system 72 for the prediction of defects 28, the system comprising: [0150] a subsystem 74 for generating an aerial image 44 of the photolithography mask 76; [0151] a data analysis device 78 comprising at least one memory 80 and at least one processor 82 configured to perform the steps of a computer implemented method 30 according to any one of clauses 1 to 23. [0152] 28. A wafer-less process window qualification system 90 comprising: [0153] a subsystem 74 for generating an aerial image 44 of the photolithography mask 76; [0154] a data analysis device 78 comprising at least one memory 80 and at least one processor 82 configured to perform the steps of a computer implemented method 30 according to clause 24.

[0155] In some examples, the invention relates to a computer implemented method 30 for inspecting a photolithography mask 76 to predict defects 28 in wafers, the method comprising: providing a model 10 of the photolithography mask 76 comprising one or more target features 14 and one or more sub-resolution assist features 22, the photolithography mask 76 being configured for printing of the one or more target features 14 onto a wafer in a printing process using a photolithography system; identifying one or more critical locations 24 in the model 10 of the photolithography mask 76 by verifying a predefined constraint concerning the target features 14 and/or the sub-resolution assist features 22; generating an aerial image 44 of the photolithography mask 76 comprising the one or more identified critical locations 24 by applying a model of the photolithography system to the photolithography mask 76; predicting defects 28 in wafers by comparing the one or more identified critical locations 24 of the aerial image 44 to one or more corresponding locations of a reference image 46. The invention also relates to a computer-readable medium, a computer program product and corresponding systems for the prediction of defects 28 in wafers and wafer-less process window qualification.

TABLE-US-00001 Reference number list 10 Model 12 Target pattern 14 Target feature 16 Printed pattern 18 Printed feature 20 Corrected pattern 22 Sub-resolution assist feature 24 Critical location 26 Printed corrected pattern 28 Defect 30 Computer implemented method 32 Model providing step 34 Critical location identification step 36 Aerial image generation step 38 Defect prediction step 40 Model modification step 42 Repair step 43 Process step 44 Aerial image 46 Reference image 48 Predicted feature 50 Reference feature 51 Difference image 52 Light 54 Back-scattering 56 First layer 58 Second layer 60 n-th layer 62 Computer implemented method 64 Parameter selection step 66 Iteration 68 Defect prediction step 70 Process window determination step 72 Photolithography mask inspection system 74 Subsystem 76 Photolithography mask 78 Data analysis device 80 Memory 82 Processor 84 Interface 86 Database 88 User interface 90 Wafer-less process window qualification system