Systems and methods for inspecting masks for EUV lithography

Abstract

To provide a system and method for inspecting a mask for EUV lithography.SOLUTION: The system comprises a first partial system for optical inspection of a mask, a second partial system for pre-classification of potential mask defects by assigning a confidence parameter to the potential mask defects, and a third partial system for checking the potential mask defects on the basis of the confidence parameter. The pre-classification of the potential mask defects based on machine learning is performed during the inspection of the mask for EUV lithography.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The contents of German Patent Application No. DE102020204508.9 are incorporated herein by reference.

[0002] The present invention relates to a system for inspecting a mask for EUV lithography. The present invention also relates to a method for inspecting a mask for EUV lithography. [Background technology]

[0003] Lithographic methods are used to manufacture microstructured or nanostructured components, such as memory chips. In this context, structures are imaged from a mask onto a wafer with the aid of a projection exposure apparatus. Before using the masks, they are tested with the aid of inspection systems, in particular APMI (Actinic Pattern Mask Inspection) systems, to ensure that they are suitable for the intended purpose. In particular, it is also possible to inspect the substrates provided for manufacturing the masks, also called mask blanks. For this purpose, ABI (Actinic Blank Inspection) systems can be used.

[0004] By way of example, an ABI system is known from US Patent Application Publication No. 2017 / 0235031. There is a constant need for improved systems and methods for inspecting masks for EUV lithography. Summary of the Invention

[0005] These objects are achieved by the features of the independent claims.

[0006] The core of the present invention consists in utilizing a multi-stage method or a system with multiple subsystems for inspecting lithography masks. In this context, potential mask defects are first identified and / or located with the aid of a first subsystem. Then, with the aid of a second subsystem, the potential mask defects are pre-classified and evaluated, in particular using a confidence parameter. A subset of the potential mask defects is then checked based on the pre-classification, in particular based on the confidence parameter of the pre-classification.

[0007] According to the invention, it has been recognized that it is advantageous to sort out from the potential mask defects identified in the first partial step those which make it possible to determine with a given degree of confidence that they are in fact mask defects without the need for further checks. Likewise, it is advantageous to sort out potential mask defects (so-called false positives) which make it possible to determine with a given degree of confidence that they are not defects or are defects not relevant for further use of the mask without the need for further checks.

[0008] In the check step performed by the third subsystem, it is only necessary to check a subset of the potential mask defects identified in the first method step, which leads to considerable time savings.

[0009] The system and method are equally suitable for inspecting masks and mask blanks, and in the following the term mask is understood to mean both the actual structured mask and the pre-structured mask, i.e. the mask blank.

[0010] EUV radiation is understood to mean electromagnetic radiation with a wavelength in the range of 5 nm to 30 nm, in particular it can relate to radiation with a wavelength of 13.5 nm or 7 nm.

[0011] According to one aspect of the present invention, the second subsystem for pre-classification of potential mask defects utilizes an automated image analysis method, in particular an automated pattern recognition method. In particular, this can be a fully automatic method. In particular, this can be a non-algorithmic method.

[0012] The second subsystem may also include algorithmic pre- or post-processing steps, in particular filtering and / or transformation steps, such as one or more Fourier transforms. In particular, the pre-classification may be performed based on data in the spatial or frequency domain.

[0013] According to a further aspect of the invention, the second subsystem for pre-classification of potential mask defects utilizes machine learning. In particular, the pre-classification can be performed in a software-based manner.

[0014] In particular, the second subsystem may comprise a data processing device, which in particular serves to process the data provided by the first subsystem in order to identify and / or locate potential mask defects.

[0015] The second subsystem may be data connected to the first subsystem, may be embodied as a component of the first subsystem, may be embodied as a component of a third subsystem, as a component of a common control device of the subsystems, or may be embodied as a separate subsystem.

[0016] The second system part can comprise a separate optical system, in particular an optical system for imaging the mask, in particular an actinic system.

[0017] In this context, an actinic radiation system is understood to be a system that uses illumination radiation of a wavelength that corresponds to the wavelength provided for the subsequent use of a mask in a projection exposure apparatus for structuring a wafer for imaging and / or testing.

[0018] In particular, the second system subsystem may comprise an EUV system.

[0019] According to a further aspect of the invention, the second and third system parts can use the same optical system for checking the mask. In particular, the second and third system parts can have a common optical system. This reduces the structural costs of the entire system.

[0020] In particular, the second system part and the third system part can be formed by a common optical system, in particular by the same optical system, which is used in different measurement modes.

[0021] The second and third system subsystems can also have independent optics, which allows the second and third inspection steps to be performed in parallel, thereby increasing the overall throughput of the system.

[0022] According to a further aspect of the invention, the second system subsystem has a memory for storing a database containing the pre-classified mask defects. In particular, the second system subsystem can have a memory containing the database containing the pre-classified mask defects. The database can also be exchangeable. In particular, the database can be updateable and / or upgradeable.

[0023] In particular, it has been found that potential mask defects identified and / or located by the first system subsystem can be compared with pre-classified mask defects from a database, facilitating a fast yet reliable classification of mask defects.

[0024] In particular, the database can be upgradeable, which improves the predictability of mask defect classification.

[0025] The databases can also be interchangeable, allowing for specific databases to be maintained that can be used for different mask types, e.g., masks containing different structural elements, thereby further improving the reliability of the pre-classification.

[0026] According to a further aspect of the invention, the second subsystem is embodied to divide the potential mask defects into at least two classes, in particular into two, three or more classes.

[0027] In this context, one class contains mask defects that do not require further checking, and a separate class contains mask defects that should be checked by a third subsystem.

[0028] In particular, defects that are definitely incorrectly identified (so-called false positives) and defects that are definitely correctly identified are classified as a first class, which includes mask defects that do not require further checking.

[0029] The remaining defects, for which no statement can be made with sufficient confidence, are classified as a further class.

[0030] The classification of potential defects relating to different classes may be determined based on one or more parameters, which may be continuous or discrete parameters.

[0031] It has been found that the pre-classification, which is usually carried out with the aid of the second subsystem, allows a significant reduction in the number of mask defects checked in the third inspection step.

[0032] The ratio of the number of mask defects checked with the aid of the third system subsystem or in the third inspection step to the total number of mask defects identified and / or located using the first system subsystem or in the first inspection step is, in particular, 20% or less, in particular 10% or less, in particular 5% or less, in particular 3% or less, in particular 2% or less, in particular 1% or less.

[0033] According to a further aspect of the invention, the classification speed v2 of the second subsystem is greater than the checking speed v3 of the third subsystem, v2>v3.

[0034] The classification speed v2 of the second subsystem is in particular at least 500 classified defects per hour, in particular at least 750 classified defects per hour, in particular at least 1000 classified defects per hour, in particular at least 1250 classified defects per hour, in particular at least 1500 classified defects per hour, in particular at least 2000 classified defects per hour.

[0035] In particular, the following applies: v2 / v3 > 1.1, in particular v2 / v3 > 1.2, in particular v2 / v3 > 1.5, in particular v2 / v3 > 2, in particular v2 / v3 > 2, in particular v2 / v3 > 3, in particular v2 / v3 > 5, in particular v2 / v3 > 10. In particular, v2 / v3 > 1 / (1 - a) applies, where a specifies the expected ratio of the number of potential mask defects checked by the third subsystem to the number of potential mask defects classified by the second subsystem. In this context, the number of potential mask defects classified by the second subsystem is exactly equal to the number of potential mask defects identified and / or located by the first subsystem. The ratio a can be determined experimentally based on the mask being checked.

[0036] According to a further aspect of the invention, the first system part for inspecting masks uses illumination light having a wavelength longer than 30 nm, in particular longer than 100 nm. In particular, the first system part can comprise a non-actinic optical system. In particular, the first system part can comprise a DUV system. In this respect, reference is made by way of example to US Pat. No. 8,103,086 B2. In particular, the first system part uses illumination light having a wavelength of 193 nm or longer for inspecting masks.

[0037] According to a further aspect of the invention, the second partial system for pre-classifying potential mask defects and / or the third partial system for checking a subset of the potential mask defects comprises an optical system having an illumination light beam with a wavelength in the EUV range.

[0038] In particular, the second system part and / or the third system part can comprise actinic optics.

[0039] In particular, the second system part and / or the third system part can comprise an actinic radiation aerial imaging system, i.e., an optical system with an illumination light beam having a wavelength in the EUV range for generating an aerial image of the mask. In particular, this can be a scanning system. Regarding the actinic radiation aerial imaging system, reference is made, by way of example, to DE102010029049A1. According to a further aspect of the invention, the system comprises a memory unit data-connected to the first system part. In particular, the memory unit serves to store data records containing data on potential mask defects.

[0040] In particular, the memory unit can be or is capable of being data connected to the second system part.

[0041] The data relating to the potential mask defects identified by the first system subsystem can be stored in a physical storage medium or in a virtual memory, in particular, where the virtual memory can simplify the data transmission from the first system subsystem to the second system subsystem.

[0042] The data of the pre-classification of the potential mask defects identified by the second system subsystem can be stored in a physical storage medium or in a virtual memory, where storage in a virtual memory simplifies data transmission from the second system subsystem to a third system subsystem.

[0043] According to a further aspect of the invention, the second inspection step includes an automated image analysis method for pre-classifying potential mask defects.

[0044] Image analysis methods may include pre-processing steps, such as filtering or transformation steps, in particular one or more Fourier transforms.

[0045] Image analysis can be performed in the spatial or frequency domain.

[0046] According to one aspect of the present invention, the second inspection step for pre-classifying potential mask defects comprises a machine learning-based method. In particular, this can be a method based on supervised learning. It can also be related to reinforcement learning. In this context, results from the third inspection step can be used for further training of the second inspection step. In this context, different types of masks can be treated differently, in particular, which can lead to an improved reliability of the pre-classification.

[0047] According to a further aspect of the invention, the distribution of one-dimensional and / or two-dimensional structures in the image of the mask recorded in the first inspection step is analyzed with a view to pre-classifying potential mask defects.

[0048] These structures can be analyzed in either the spatial or frequency domain.

[0049] In particular, the one-dimensional structures can be critical dimensions (CDs). In particular, the two-dimensional structures can be contours of the image of the mask recorded in the first inspection step, or its Fourier transform. Also, the properties of these structures can be analyzed, for example their diameters.

[0050] According to a further aspect of the invention, the intensity distribution of the mask image recorded in the first inspection step is analyzed with a view to pre-classifying potential mask defects. In particular, the mask image can be analyzed pixel by pixel. Upsampling or downsampling is possible.

[0051] One or more of the subsystems 2, 3, 4 may be imaging systems, each recording the entire image of the mask or mask blank in a single exposure, or may be scanning systems, in which the image of the mask or mask blank is collected from multiple exposures.

[0052] According to a further aspect of the invention, the second inspection step includes one or more pre-processing steps, in particular one or more correction steps for correcting the image or images recorded in the first inspection step with respect to photon noise.

[0053] According to a further aspect of the invention, the second inspection step comprises a comparison step for comparing one or more images of the mask recorded in the first inspection step with data from a database.

[0054] According to a further aspect of the invention, the third inspection step comprises a method for verifying and / or analyzing the aerial image using actinic methods, in particular illumination light in the EUV range.

[0055] According to a further aspect of the invention, the first inspection step comprises a non-actinic method, in particular a method for verifying an image of the mask using illumination light in the DUV wavelength range, in particular illumination light with a wavelength of 193 nm.

[0056] In particular, it has been found that the method according to the invention can include at least two optical inspection steps in which the image of the mask is viewed and / or analyzed using illumination light of different wavelengths, which firstly reduces the time required to inspect the mask and secondly improves the reliability of the characterization of mask defects.

[0057] Further advantages, details and features of the invention become apparent from the description of exemplary embodiments with reference to the drawings, in which: [Brief explanation of the drawings]

[0058] [Figure 1] 1A-1C are schematic diagrams illustrating steps in a method for inspecting a mask for EUV lithography. [Figure 2] FIG. 1 is a schematic diagram illustrating components of a system for inspecting masks for EUV lithography. DETAILED DESCRIPTION OF THE INVENTION

[0059] FIG. 2 shows a schematic diagram of a system 1 for inspecting masks for EUV lithography.

[0060] The system 1 includes a first subsystem 2, a second subsystem 3, and a third subsystem 4.

[0061] Preferably, so-called multi-layer defects of a mask, in particular a mask blank, are checked separately with the aid of a separate system, in particular an actinic blank inspection tool (ABI tool).

[0062] The first system part 2 is responsible for the optical inspection of the mask, in particular for identifying and / or locating potential mask defects.

[0063] The first system part 2 includes in particular an inspection system for inspecting masks using illumination light in the DUV range, in particular with an illumination light having a wavelength of 193 nm. For details of such a system, reference should be made, exemplarily and representatively, to U.S. Pat. No. 8,103,086 (B2), which is incorporated herein by reference.

[0064] The second system subsystem 3 serves in particular to pre-classify potential mask defects identified and / or located by the first system subsystem 2 .

[0065] The second system part 3 can include an optical system with a wavelength in the EUV range.

[0066] In particular, the third subsystem 4 serves to check for potential mask defects.

[0067] The third system part 4 comprises, in particular, an optical system for checking the mask using wavelengths in the EUV range. For details of such a system, reference should be made, exemplarily and representatively, to DE 10 2010 029 049 A1, which is incorporated herein by reference.

[0068] As illustrated diagrammatically in Figure 2, the first system subsystem 2 is data-connected to a first memory unit 5. In turn, the first memory unit 5 is data-connected to a second system subsystem 3.

[0069] The first memory unit 5 can be embodied as a separate memory unit, in particular as a separate storage medium or as a virtual memory, or as part of the first system subsystem 2 or of the second system subsystem 3.

[0070] The first memory unit 5 serves to store data ascertained by the first system subsystem 2 in order to identify and / or locate potential mask defects, the corresponding data serving as input for further inspection of the mask with the aid of the second system subsystem 3.

[0071] The second system part 3 is data connected to a second memory unit 6 .

[0072] The second memory unit 6 can be embodied as a separate memory unit, in particular as a separate storage medium or as a virtual memory, and can also be embodied as a component of the second system subsystem 3 or as a component of the third system subsystem 4.

[0073] The second system part 3 may include an optical system for testing the mask, in particular for generating and analyzing an aerial image of the mask, and in particular may be an EUV system.

[0074] The third system part 4 comprises an optical system for checking and analyzing potential mask defects, in particular the optical system of the third system part 4 is an actinic radiation system.

[0075] The second system subsystem 3 comprises a memory 7 for storing a database containing pre-classified mask defects. The memory 7 may be embodied separately from the second system subsystem 3.

[0076] In the following, the basic steps of the method for inspecting a mask will be described with reference to FIG.

[0077] The method is a multi-stage method, in particular it comprises a first inspection step 8, a second inspection step 9 and a third inspection step 10.

[0078] The test steps 8, 9, 10 may include one or more sub-steps.

[0079] A first inspection step 8 serves in particular to record an image of the supplied mask in order to identify and / or locate potential mask defects in the supplied mask.

[0080] The second inspection step 9 serves, inter alia, to pre-classify potential mask defects.

[0081] The third inspection step 10 serves specifically to check a subset of potential mask defects.

[0082] In a first inspection step 8, the structured lithographic mask is analyzed, in particular with the aid of an optical system. In a first inspection step 8, a list of potential mask defects is created, in particular with the aid of a DUV system. The list of potential mask defects can be stored in a first memory unit 5.

[0083] In a second inspection step 9, the mask is tested using actinic optics, in particular an EUV system, in this context, in particular the potential defects according to the list created in the first inspection step 8 are pre-classified.

[0084] In particular, it is possible to divide the potential defects identified in the first inspection step 8 into at least two classes, one class containing only potential defects that do not require further checking, and the other class containing potential defects for which it is not possible to make a final assessment with sufficient confidence as to whether they are in fact defects, in particular defects related to the intended use of the mask.

[0085] In particular, classified into the first class are potential defects that have a sufficient probability that they are not real defects (so-called false positives), in particular with a confidence level of at least 95%, in particular at least 97%, in particular at least 99%.Furthermore, classified into the first class are potential defects that are real defects, in particular defects related to the intended use of the mask, with a confidence level of at least 95%, in particular at least 97%, in particular at least 99%.

[0086] The remaining potential defects for which no final statement with at least 95% confidence is possible are classified as a second class.

[0087] The second inspection step 9 particularly includes fast disposition of potential mask defects, with a classification rate of at least 500 classified defects per hour, particularly at least 750 classified defects per hour, particularly at least 1000 classified defects per hour, particularly at least 1250 classified defects per hour, particularly at least 1500 classified defects per hour, particularly at least 2000 classified defects per hour.

[0088] In particular, machine learning based methods are used to pre-classify potential mask defects in the second inspection step 9.

[0089] Different details and aspects of the method steps for pre-classifying potential mask defects are described below.

[0090] The first system subsystem can have a rapid-action mode. In this mode of operation, the image in the first inspection step can be recorded particularly quickly. In this context, the accuracy, particularly the resolution, required for the mode used for accurate optical analysis of potential defects can be dispensed with. It is sufficient to be able to demarcate the image of the potential defect from photon noise with sufficient reliability.

[0091] The second inspection step 9 may include pattern recognition, which may include thresholding and image contour analysis.

[0092] During the image contour analysis, one-dimensional and / or two-dimensional structures and / or logical structures can be analyzed, in this context logical structures being understood as asymmetric or non-periodic structures.

[0093] In particular, it allows for automated thresholding to measure the critical dimension (CD) for each structural element in the mask's overall image. Analysis of the distribution of critical dimension values ​​was found to have clearly defined peaks in the frequency domain. Deviations from these peaks may indicate potential defects. The power spectral density (PSD) of the critical dimensions in the aerial image was found to have a continuum and peaks, and the characteristics of the peaks may be affected by defects.

[0094] Furthermore, the properties of these peaks can be analyzed, for example their full width at half maximum (FWHM), and in this context, photon noise can be taken into account.

[0095] Training of the system for machine learning can be done in the spatial or frequency domain.

[0096] Furthermore, a correction method for correcting photon noise can be performed before the image analysis in the second inspection step 9 .

[0097] According to one variant, the variations in the image contour or frequency distribution of the measured structural parameters of possibly preprocessed images, in particular noise-corrected images, can be analyzed and compared with data in a database of defect-free structures and defects.

[0098] In particular, the second inspection step 9 may include a comparison step for comparing the potential defects, in particular certain parameters of the potential defects, with corresponding data in a database.

[0099] Pre-classification can be performed based on the distribution of identified critical dimensions or specific parameters of the contour in the spatial or frequency domain. For example, low-intensity peaks in the power spectral density (PSD) can indicate defects. Pre-classification based on this is possible after appropriate prior training.

[0100] As an alternative to pre-classification based on critical dimensions or contours, it may also be possible to evaluate the intensity distribution or the frequency of changes in this distribution. In particular, this evaluation may be performed pixel by pixel. Upsampling or downsampling is also possible. Again, in this context, corrections may be made to take photon noise into account. The distribution analyzed in this way may be compared with data in a database used to train the system for machine learning methods. In this regard, it should be noted that, in particular due to contours, defects have different frequencies relative to regular mask structures, in particular relative to the frequencies of regular, especially periodic, mask structures.

[0101] Further processing steps, such as smoothing methods (image smoothing) or low-pass filter steps, can be provided in all methods, in particular to filter out noise.

[0102] According to a further alternative, it may be possible to directly compare the images of the mask, and in particular the images of the mask defects, identified in the second inspection step 9 with the images in the database.

[0103] The database can be continuously replenished throughout the method, as a result of which it is possible to continuously improve, and in particular further increase the reliability of, the pre-classification in the second inspection step 9. This results in an average decrease in the ratio of potential mask defects checked in the third inspection step 10 to the number of potential mask defects identified overall in the first inspection step 8.

[0104] It has been found that with the aid of pre-classification in the second inspection step 9, it is possible to reduce the number of potential mask defects checked in the third inspection step 10 from the thousands of potential mask defects identified in the first inspection step 8 to less than 200, in particular less than 100.

[0105] Actinic methods, particularly actinic aerial imaging, are provided in a third inspection step 10 to check for potential mask defects.

[0106] The overall result of the inspection method, and in particular the three inspection steps 8, 9 and 10, is a statement regarding the relevance of the mask defects for the intended use of the mask. In particular, it can be determined based on this overall result whether the mask meets certain specified quality criteria.

Claims

1. 1. A system for inspecting a mask for EUV lithography, comprising: 1.

1. A first system part (2) for optical inspection of a mask for identifying and / or locating potential mask defects; 1.

2. A second system subsystem (3) for pre-classification of the potential mask defects, and 1.

3. A third subsystem (4) for checking the potential mask defects Equipped with 1.

4. The second subsystem (3) is embodied to assign to the potential mask defects identified and / or located by the first subsystem (2) a confidence parameter for characterizing the reliability of the identification and / or the relevance of the defect for a subsequent application; 1.

5. The third subsystem (4) is controllable such that a subset of the mask defects identified and / or located by the first subsystem (2) is checked based on confidence parameters assigned to the potential mask defects by the second subsystem (3); 1.

6. System, characterized in that the second system part (3) has a memory for storing a database with pre-classified mask defects.

2. 2. The system (1) according to claim 1, characterized in that the second subsystem (3) for the pre-classification of the potential mask defects utilizes automated image analysis methods.

3. The system (1) according to claim 1 or 2, characterized in that the second subsystem (3) for the pre-classification of the potential mask defects makes use of machine learning.

4. The system (1) according to any one of claims 1 to 3, characterized in that the second subsystem (3) is implemented to divide the potential mask defects into two, three, four or more classes.

5. The system (1) according to any one of claims 1 to 4, characterized in that the second subsystem (3) has a classification speed v2 and the third subsystem (4) has a check speed v3, where the following applies: v2>v3.

6. The system (1) according to any one of claims 1 to 5, characterized in that the first system part (2) for inspecting the mask uses an illumination light beam with a wavelength longer than 30 nm.

7. 7. The system (1) according to claim 1, characterized in that the second system part (3) for pre-classification of the potential mask defects and / or the third system part (4) for checking a subset of the potential mask defects comprise an optical system with an illumination light beam having a wavelength in the EUV range.

8. 1. A method for inspecting an EUV lithography mask, comprising: 8.

1. Preparing a mask for EUV lithography; 8.

2. A first inspection step carried out with the aid of a first system subsystem for recording an image of the prepared mask and for identifying and / or locating potential mask defects in the prepared mask; 8.

3. A second inspection step performed with the aid of a second subsystem for preclassifying the potential mask defects into at least two non-empty subsets, and 8.

4. A third inspection step for checking the potential mask defects in one of the subsets identified in the second inspection step. Including, 8.

5. The second inspection step for pre-classifying the potential mask defects includes a machine learning based method; 8.6 Method, wherein the number of mask defects checked in the third inspection step is reduced by pre-classification performed with the aid of the second system subsystem.

9. 9. The method of claim 8, wherein the distribution of one-dimensional and / or two-dimensional structures in the image of the mask recorded in the first inspection step is analyzed for the purpose of pre-classification of the potential mask defects.

10. 10. A method according to claim 8 or 9, characterized in that the intensity distribution in the image of the mask recorded in the first inspection step is analysed with a view to pre-classifying the potential mask defects.

11. A method according to any one of claims 8 to 10, characterized in that the second inspection step comprises a comparison step in which the recorded image of the mask is compared with data in a database.

12. The method according to any one of claims 8 to 11, characterized in that the third inspection step (4) comprises an actinic method.

13. The method according to any one of claims 8 to 12, characterized in that the first inspection step (2) comprises a non-actinic method.

14. A method described in any one of claims 8 to 13, characterized in that the ratio of the number of mask defects checked in the third inspection step to the total number of mask defects identified and / or located in the first inspection step is 20% or less.

Images