Method for measuring photomask

By recording and adapting the spatial image of the photomask, determining the structural edges, and correcting them in combination with the scanner and mask distortion equipment, the problem of registration error prediction and correction in the lithography process in the prior art is solved, and the light is improved Timing accuracy and process quality.

CN113340564BActive Publication Date: 2024-12-06CARL ZEISS SMT GMBH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110190657.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2020-02-18
Filing Date
2021-02-18
Publication Date
2024-12-06
Estimated Expiration
2041-02-18

AI Technical Summary

Technical Problem

The existing measurement methods are difficult to accurately predict the registration error of the photomask on the wafer during semiconductor lithography, and cannot effectively distinguish the distortion error and registration error in the imaging plane, resulting in the accumulation of errors when complex structures are formed, affecting light. Engraving accuracy.

Method used

By recording the spatial image of the photomask, defining the region of interest, determining the structural edges, and adapting them to the desired structure, moving the structural edges using separate registration measurement results, generating a spatial image with more precise information, and then Calculate corrected image errors and imaging aberrations, improve measurement results using simulated spatial images, and correct them in combination with scanners and mask distortion devices.

Benefits of technology

More accurate prediction and correction of on-chip registration errors is achieved, the accuracy of the lithography process is improved, error accumulation is reduced, and the quality of the alignment and adaptation process is enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113340564B_ABST
    Figure CN113340564B_ABST
Patent Text Reader

Abstract

The invention relates to a method for measuring a photomask for semiconductor lithography, comprising the following steps: - recording an aerial image of at least one area of ​​the photomask, - defining at least one region of interest (1), - determining a structure edge (3) in the at least one region of interest (1), - providing a desired structure (4) to be produced by the photomask, - adapting the determined structure edge (3) to the desired structure (4), - moving the adapted structure edge by means of the result of a separate registration measurement.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the priority of German patent application DE 10 2020 104167.5 filed on February 18, 2020, the contents of which are incorporated herein by reference in their entirety. Technical Field

[0003] The invention relates to a method for measuring a photomask for semiconductor lithography, in particular for determining the so-called registration error. Background Art

[0004] In this context, registration is to be understood as meaning the position of structures on the mask (and subsequently also on the wafer) relative to position marks on the photomask (so-called alignment marks). In a projection exposure apparatus, the alignment marks on the mask are particularly also used to set the mask position (and thus the imaged position on the wafer) relative to the imaging optics of the apparatus.

[0005] The registration error is therefore a deviation of the actual position of the structure on the mask relative to the intended position with respect to the respectively associated alignment mark and therefore with respect to the mask coordinate system. In particular in the case of so-called double exposures for the formation of complex structures on the wafer, i.e. in the case of a layer on the wafer being exposed successively via two different photomasks, such errors add up, so that in this case a correct registration or at least an accurate knowledge of the registration errors is particularly important.

[0006] Different systems, such as coordinate measuring machines, are currently available for determining the registration of photomasks. Such systems offer excellent spatial resolution down to the sub-nanometer range. For determining the registration, typically recorded aerial images of certain structures to be exposed (so-called features) or recorded aerial images of the edges of structures are used. However, the specified systems use illumination and optical units that deviate significantly from the subsequent conditions in the projection exposure apparatus. This results in the profiles created on the actual wafer deviating significantly from the profiles used to determine the registration in the coordinate measuring machine, both in terms of their relative position and in terms of their form. Therefore, the correlation between the registration measured on the mask plane and the actual registration generated on the wafer is not yet optimal.

[0007] By means of a system for producing an aerial image of the mask, a more realistic representation of the light intensity generated on the wafer can be achieved. The aforementioned systems differ in that they simulate the lighting and imaging conditions of a real projection exposure apparatus relatively well and thus help to predict the image generated on the wafer, in particular the image of the contours and edges, which is close to reality. However, these systems do not offer the option of high-precision measurements for registration. Furthermore, the correction of distortions in the recorded image is limited, since the errors caused by the distortions are similar to conventional registration errors and cannot be distinguished from registration errors in the imaging plane. Summary of the invention

[0008] The object of the invention is to specify a method which allows more accurate prediction of registration errors in the plane of a wafer during semiconductor lithography, without the need to expose the wafer for this purpose.

[0009] This object is achieved by a method having the characteristics described below. The description below relates to advantageous developments and variants of the invention.

[0010] The method for measuring a photomask for semiconductor lithography according to the present invention comprises the following steps:

[0011] - recording an aerial image of at least one region of the photomask,

[0012] - defining at least one region of interest (ROI), i.e. a region of the aerial image containing a structure of interest,

[0013] - determining structure edges, i.e. boundaries of the structure of interest, in at least one region of interest,

[0014] - providing the desired structure produced by the photomask,

[0015] - adapting the determined structure edges to the desired structures; these structures may also be referred to as "lithography targets",

[0016] - The adapted structure edges are moved by the results of separate registration measurements.

[0017] Due to optical effects and mask design rules, the above-mentioned lithography targets deviate slightly in their contours from the contours of the corresponding structures on the aerial image. Due to this deviation, small positioning errors of the measured structures occur. Therefore, the structures determined from the aerial image are in their correct positions - except for the above-mentioned small errors.

[0018] Thus, in this way a data record or information item is obtained which contains a more precise information item about expected errors, in particular registration errors of the mask. The result can therefore in particular contain a real aerial image of the mask with largely correct registration information.

[0019] In order to improve the quality of the alignment and adaptation process, a computational correction of image errors, aberrations and imaging aberrations of the system used to generate the aerial image can be carried out after the generation of the aerial image.

[0020] In a variant of the invention, an improvement in the measurement result can be achieved by using an aerial image simulated on the basis of a mask design in the step of adapting the determined structure edges.

[0021] For the simulated aerial image, it can be assumed that the structures simulated with this are in their correct positions, although they deviate slightly from the actual aerial image. However, since optical effects and design rules have been taken into account in the simulation, the deviations are small. In addition, alignment based on the overall profile of the intensity levels is more accurate than alignment based on the profile. For these two reasons, the alignment of structures measured from the aerial image is improved.

[0022] Here, a simulated aerial image can be generated for the registration measurement system and / or the mask inspection system. In other words, a simulated aerial image is generated with the aid of the imaging properties of the registration measurement system and / or the mask inspection system to form the basis for a corresponding simulation.

[0023] In a further step, a correction strategy can be developed based on the results in order to correct the measured errors. Here, the effects of the errors measured on the image in the wafer or the effects of errors of the mask itself can be corrected.

[0024] In a first variant, the determined errors can be corrected by means of correction options inherent in the scanner. For this purpose, the determined structural offsets serve, for example, as input parameters for the scanner software which subsequently creates a strategy by which the distortions are corrected in situ when the wafer is exposed in the scanner by dynamically adapting suitable scanner parameters.

[0025] In other cases, equipment for physical mask distortion can be used, such as the RegC or ForTune systems from Carl Zeiss SMT GmbH. Such equipment is able to physically stretch the mask at points. When the above results are input, the equipment-specific software can develop the appropriate mask processing.

[0026] Ideally, the options of scanner and mask distortion device should be considered in combination in order to obtain the best correction result.

[0027] Furthermore, not every offset on the mask is equally critical for the function of the semiconductor structure to be manufactured. Small offsets at critical points should be evaluated as more important than larger offsets at non-critical points. Therefore, it may be advantageous to weight the offsets determined at a certain point by a measure of the criticality of the point in various ways. Thus, the creation of the correction strategy may be further optimized by taking more account of critical points than less critical points.

[0028] The above-mentioned embodiments relate to the case where the mask and wafer exposure generate the chip plane. This is called a single exposure method or single patterning.

[0029] However, it is not uncommon for a chip plane to be produced by two or more exposures (i.e., for the so-called double or multiple exposure method or double or multiple patterning to be used). Typically, two or more masks are also used in the process. Each of these masks contributes to the overall image of the chip plane, which has the distortion of the mask itself.

[0030] In this case, both the absolute distortion and the distortion difference have an effect on the subsequent semiconductor structure. Since the distortion or error is generated individually for each mask according to the described method, the distortion difference can also be calculated by simple mathematical subtraction of the individual distortions, and in this respect the correction strategy can be optimized later on the basis of the scanner and / or the mask. In this case, each mask is not corrected individually to the greatest possible extent, but rather each mask receives a correction that generates the smallest possible combined total distortion on the subsequent semiconductor structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Exemplary embodiments of the present invention are explained in more detail below with reference to the accompanying drawings. In the drawings:

[0032] Figure 1 An exemplary aerial image of a mask is shown,

[0033] Figure 2 Schematically shows Figure 1 Image errors in the illustrated aerial image, in particular distortion,

[0034] Figure 3 shows the subdivision of the aerial image into different regions of interest,

[0035] Figure 4 shows the extracted structural outline,

[0036] Figure 5 shows the results of an optional matching of the mask design to the aerial image,

[0037] Figure 6 shows the desired structure on the wafer,

[0038] Figure 7 A schematic diagram is shown in which the results of the separate registration measurements are presented together with the results obtained up to this point. DETAILED DESCRIPTION

[0039] Figure 1The result of recording a real mask by a mask inspection system is shown. An aerial image of the mask as it would essentially also appear in a scanner is illustrated, but recorded by the mask inspection system under conditions equivalent to the scanner. The mask inspection system creates conditions equivalent to the scanner by reproducing the scanner to the best possible extent on the illumination side. This includes, as an example, a flexible setting of the illumination pattern (the so-called setup) and of the polarization pattern using a light source with the same or very similar wavelength to that used in the scanner. The intensity distribution - as it would also be projected into the photoresist of the wafer to be exposed - is recognizable.

[0040] Figure 2 The distortion that can be traced back to the mask inspection system is schematically shown, and from Figure 1 The distortions are subtracted from the illustrated image for image correction purposes. Since these distortions are only relevant to the mask inspection system, they are irrelevant to the subsequent use of the mask in the scanner and therefore irrelevant to chip manufacturing. Therefore, it is advantageous to separate the distortions into those caused by the mask itself and those caused by the mask inspection system.

[0041] Figure 3 The diagram shows the subdivision of the distortion-corrected image into different sub-regions, so-called "regions of interest" 1, which are drawn using dashed lines in the diagram. These regions of interest 1 (hereinafter denoted "ROI") are essentially defined by containing self-contained features or feature groups of interest, i.e. certain structures 2 or groups of structures generated on the wafer and thus created on the mask.

[0042] Figure 4 The result of a first contour extraction for an individual ROI 1 is shown. Here, the contour is extracted with the aid of a certain intensity value (the so-called threshold value initially defined), which represents a measure of the edge created on a real wafer. It goes without saying that process-specific parameters of the lithography process (such as the type of photoresist used) also have an influence on the threshold value. A large number of edge detection methods are known from image processing (e.g., the canny method, the Sobel method, gradient-based methods, etc.). Another method from this group can also be used instead of the threshold value. Furthermore, the contour can also be derived by artificial intelligence from a data record trained with a suitable aerial image. As a result, a structure edge 3 is obtained, which is illustrated as a solid line in the individual ROI, i.e., an edge 3 that delimits the extracted contour.

[0043] Figure 5 The results of an optional matching of the mask design to the aerial image are shown.

[0044] Figure 6The desired structures in the wafer plane, i.e. the lithography targets 4, are shown, as structures essentially delimited by straight paths. The edges 3 measured in the mask inspection system are now fitted to these structures 4. In particular, the threshold values ​​and the x and y positions as fitting parameters become problematic. According to the fitting process, the best possible superposition of the contour actually measured in the wafer plane relative to the lithography target 4 is obtained. The described exact adaptation is made possible by the use of a mask inspection system, since, as already mentioned, the mask inspection system reproduces the properties of the scanner to a large extent. Therefore, in the fitting process the relative position in the wafer plane (x and y position) and the threshold values ​​with their influence on the edge position are varied until the specified optimization criterion has been met. Since the x and y positions are taken into account as parameters of the fitting process, the displacement of the structure edges can no longer be inferred from the result with sufficient accuracy. In particular, a method based on the optimization of the edges can be carried out for any ROI. Figure 6 This facilitates for the first time the combination of the actual contour generated in the scanner (including all optical imaging effects) and its position on the mask.

[0045] Figure 7 The diagram shows further steps in which the results of the separate registration measurements are applied to the results obtained so far. Shown in the figure (particularly in the detailed enlargement illustrated at the bottom of the figure) is a lithography target 4 which can be identified by its straight-line boundaries, the fitted contour 3 from the aerial image measurement (illustrated with dashed lines) and the contour 5 shifted by the separately measured registration errors. Here, the registration errors can be determined in particular by a coordinate measuring machine, naturally, other processes for determining the registration errors are also conceivable. The image obtained is distinguished by its rich information content. Firstly, it contains a true representation of the contour expected on the wafer (obtained from the aerial image of the mask inspection system), and secondly, it also contains the registration errors expected for each region of interest. In particular, if structures on the same layer of the wafer are intended to be generated by double exposure from two different masks, the following situation may occur: the different but existing registration errors of the masks will add up and, in combination with the situation of the real contour of the features deviating from the lithography target, lead to the generation of structures with defects. Since foreseeable problems can be identified in a timely manner based on mask measurement according to the invention, effective and appropriate mask correction can be undertaken by the known system.

[0046] However, the above method assumes that the aerial image recorded by the mask inspection system and the aerial image recorded by the registration measurement system do not differ from each other or differ only slightly. However, since the aforementioned systems differ significantly, especially in terms of illumination and numerical aperture, this assumption is not quite correct. Therefore, there is a possibility that the registration errors determined by the registration measurement system are not completely transferable to the aerial image generated by the mask inspection system.

[0047] In order to solve this problem, in addition to the known method steps, a registration image and a mask inspection system image are generated from the mask design by simulation, in particular by Kirchhoff simulation or rigorous simulation. The simulation of the registration image supplies values ​​for the registration; in particular, known imaging aberrations such as distortions are also taken into account, for example. The simulation is usually carried out in any case for the purpose of determining the registration. From the comparison of the two simulated images with one another, differential errors of the registration measurement between the registration measurement system and the mask inspection system can be derived, which can be taken into account during further measurement and qualification of the mask.

[0048] In particular, the x / y adaptation is firstly carried out based on the simulated aerial image (and no longer on the lithography target).

[0049] Then, in a second step and without further adaptation of the x and y values, the threshold values ​​are adapted such that the CD values ​​of the respectively considered structure are obtained from the lithography target to the greatest extent possible.

[0050] Thus, since the simulation of the mask design makes errors of the OPC (ie software based on the fact that the mask design is defined starting from the lithography target) identifiable, a representation of the offset of the determined structure relative to the lithography target is obtained, among other things.

[0051] Reference numerals list

[0052] 1 Region of Interest (ROI)

[0053] 2 Structure

[0054] 3 Structural edge

[0055] 4 Lithography Target

[0056] 5. Offset profile from registration measurement

Claims

1. A method for measuring a photomask for semiconductor lithography, comprising the following steps: - recording an aerial image of at least one area of ​​the photomask to generate a recorded aerial image of the at least one area of ​​the photomask, - defining at least one region of interest (1) of said photomask, - determining a structure edge (3) in at least one region of interest (1) of the photomask based on the recorded aerial image to generate a determined structure edge of the photomask, - providing the desired structure to be produced by said photomask (4), - adapting a determined structure edge (3) in at least one region of interest of the photomask to the desired structure (4) to be produced by the photomask to generate an adapted structure edge of the photomask, wherein the determined structure edge of the photomask is fitted to the desired structure to be produced by the photomask, - Shifting the adapted structure edges of the photomask by the results of the separate registration measurements.

2. The method according to claim 1, in, After the recording of the aerial image, there is a computational correction of the image aberrations of the system for recording the aerial image.

3. The method according to claim 1 or 2, in, In the step of adapting the determined structure edges (3), an aerial image simulated on the basis of the mask design is used.

4. The method according to claim 3, in, A simulated aerial image is generated for a registration measurement system and / or a mask inspection system.

5. The method according to claim 1 or 2, in, The determined structure edges are used to create a correction strategy for the mask or for the exposure process of a scanner.

6. The method according to claim 5, in, One correction strategy for the mask is to physically stretch the mask in a targeted manner.

7. The method according to claim 5, It is characterized in that A correction strategy is created for the matching of at least two masks.