Method of analysis of a sample surface
The method uses a charged particle beam device to divide sample images into matrix windows and compare sub-histograms for accurate alignment mark detection and surface analysis, addressing the challenges of diverse sample surfaces and enabling precise navigation and automation.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- TESCAN GRP AS
- Filing Date
- 2025-12-09
- Publication Date
- 2026-06-18
AI Technical Summary
Existing methods struggle to accurately determine the size of alignment marks on sample surfaces in charged particle beam devices, particularly in topographically or materially diverse samples, and to identify the nature of the sample surface, which is crucial for precise navigation and automated detection.
A method using a charged particle beam device with an evaluation unit that divides the sample image into matrix windows, creates sub-histograms, and compares them to a stored histogram to determine the similarity, allowing for the detection of alignment marks and surface properties regardless of the image's histogram distribution.
Enables reliable detection of the smallest possible alignment marks and identification of surface properties, ensuring precise navigation and automation in diverse samples without relying on normal distribution assumptions.
Smart Images
Figure CZ2025050100_18062026_PF_FP_ABST
Abstract
Description
[0001] Method of analysis of a sample surface
[0002] Technical Field
[0003] The invention relates to a method of analysis of a sample surface for the purpose of determining the appropriate size of alignment marks and the nature of the sample surface, particularly in charged particle beam devices.
[0004] Background of the Invention
[0005] Alignment marks play an important role in the operation of focused ion beam scanning electron microscopes (FIB-SEM). These marks are strategically placed on the sample and serve many purposes that increase the precision and efficiency of imaging and machining. First, they are essential for precise navigation across the sample. This ensures correct targeting of the region of interest, which is necessary for precise navigation of the manipulator and for sample machining, for example when producing lamellae from the sample and subsequently extracting them. Second, the alignment marks facilitate reproducibility by enabling precise re-targeting of a specific region of the sample. This is particularly important in experiments which require multiple imaging steps. Third, the alignment marks enable automation of FIB-SEM processes. They allow the system to automatically recognize and adjust the position of the sample. They also make it possible to observe or machine the sample in different devices. Thanks to the alignment marks, the same region is always targeted.
[0006] Currently, however, the problem lies in correctly determining the size of the alignment mark. In general, it is desirable to create the smallest possible alignment marks, since larger marks take up more space, are more time-consuming to produce, and consume more material. On the other hand, the smaller the alignment marks, the more difficult they can be to detect, particularly in samples with a topographically or materially diverse surface. This is a problem especially in automated detection of alignment marks, where such detection is generally more difficult for the algorithms, as the alignment marks may merge with the background.
[0007] It may also be preferable to know the nature of the sample surface for various other applications. The nature of the surface means whether it is topographically diverse, whether the sample is heterogeneous or homogeneous, or whether it contains any contaminants or other defects. Currently, a method for detecting foreign objects on a sample is known from the document JP2799921 B2, which uses histogram analysis and applying different weights to identify foreign objects on the sample surface with a histogram that otherwise has a normal distribution. In this case, the captured image is divided into sub-histograms, and an overall mean value and variance value are obtained by summing the individual sub-histograms. If, at a certain location, the deviation from the mean value and variance value is greater than a defined one, it is determined that a foreign object is present in the given region. The disadvantage of this method consists in its limitation to a normal distribution of the histogram, relative to which the deviation is determined, since any deviation from a normal distribution is considered to be a foreign object. However, a real sample may also have a different histogram distribution. Therefore, this is not a universally applicable method.
[0008] It would be desirable to come up with a solution that would allow detection of the sample surface nature and determination of the minimum required size of the alignment mark regardless of the histogram distribution of the image of the sample.
[0009] Summary of the Invention
[0010] The above problems are solved by a method of analysis of a sample surface using a charged particle beam device comprising at least one charged particle source, at least one signal particle detector, a stage adapted for placing a sample, the sample placed on the stage, and an evaluation unit for performing the method, wherein the evaluation unit contains a memory; the essence of which lies in the fact that the memory stores information about at least one matrix window size, a limit number of sub-histograms different from the histogram of the acquired image of the sample, and a decision criterion where two histograms are still similar, wherein the method comprises a first step of acquiring the image of the sample by irradiating the sample with a charged particle beam at such irradiation parameters where the acquired image of the sample has a histogram without saturated areas, and it further comprises a second step comprising sequentially performed sub-steps: a sub-step of dividing the acquired image of the sample by the evaluation unit into a matrix with a matrix window size corresponding to the matrix window size stored in the memory, a sub-step of creating sub-histograms of the individual matrix windows, a sub-step of calculating the similarity of the individual sub-histograms to the histogram of the acquired image of the sample based on the decision criterion, a sub-step of issuing information that the sub-histograms are not similar to the histogram of the acquired image of the sample at the given matrix window size if the number of sub-histograms different from the histogram of the acquired image of the sample exceeds the limit number of subhistograms different from the histogram of the acquired image of the sample stored in the memory, or that the sub-histograms are similar to the histogram of the acquired image of the sample at the given matrix window size if the number of sub-histograms different from the histogram of the acquired image of the sample does not exceed the limit number of sub-histograms different from the histogram of the acquired image of the sample stored in the memory.
[0011] The method accomplishes the above objective by utilizing the first and second steps and sub-steps described, which makes it possible to apply this method to a sample with a surface whose image has an arbitrary histogram distribution after imaging. Simultaneously, based on the information whether the sub-histograms are similar to the histogram of the acquired image of the sample at the given matrix window size, it may be determined whether, for example, an alignment mark of a size corresponding to the matrix window size will be usable, i.e. detectable, for a sample with such a surface. In one of the preferred embodiments of the method, the second step is performed repeatedly, with the matrix window size corresponding to the smallest one of the matrix window sizes stored in the memory being selected first, and then each time a successively larger one of the matrix window sizes stored in the memory than in the previous embodiment of the second step, wherein the repetition is performed until either information that the sub-histograms are similar to the histogram of the acquired image of the sample is issued, or until the matrix window with the largest one of the matrix window sizes stored in the memory is selected. The advantage of such repetition of the second step is the possibility of determining the smallest possible size of the alignment mark without affecting the reliability of the detection of this mark.
[0012] In another possible preferred embodiment of the method, the second step is performed repeatedly, with the matrix window size corresponding to the largest one of the matrix window sizes stored in the memory being selected first, and then each time a successively smaller one of the matrix window sizes stored in the memory than in the previous embodiment of the second step, wherein the repetition is performed until either information that the sub-histograms are not similar to the histogram of the acquired image of the sample is issued, or until the matrix window with the smallest one of the matrix window sizes stored in the memory is selected. The advantage of such repetition is the possibility of finding, for example, a foreign object on the sample surface of a size approximately corresponding to the matrix window size.
[0013] In a preferred embodiment of the matrix window size, the matrix window size stored in the memory corresponds to the size of the alignment mark. The advantage of this embodiment is that it establishes applicability of conventionally used sizes of alignment marks for the given sample surface.
[0014] Description of Drawings
[0015] A summary of the invention is further clarified using exemplary embodiments thereof, which are described with reference to the accompanying drawings. For the sake of clarity, only those parts of the device that are important in terms of the principle of the present invention are shown in the drawings. Here: fig. 1 shows schematically the charged particle beam device; fig. 2 shows the acquired image of the sample; fig. 3 shows the histogram of the acquired image of the sample; fig. 4 indicates a division of the acquired image of the sample into a 3x3 matrix; figs. 5, 7, 9, 11 , 13, 15, 17, 19, and 21 successively show the individual matrix windows according to the division of the acquired image of the sample into a 3x3 matrix line by line; figs. 6, 8, 10, 12, 14, 16, 18, 20, and 22 show histograms corresponding to the individual matrix windows according to the division of the acquired image of the sample into a 3x3 matrix; fig. 23 indicates the division of the acquired image of the sample into a 2x2 matrix;
[0016] -figs. 24, 26, 28, and 30 successively show the individual matrix windows according to the division of the acquired image of the sample into a 2x2 matrix line by line; figs. 25, 27, 29, and 31 show histograms corresponding to the individual matrix windows according to the division of the acquired image of the sample into a 2x2 matrix;
[0017] Exemplary Embodiments of the Invention
[0018] Said embodiments show exemplary variants of the embodiments of the invention, which, however, have no limiting effect from the point of view of the scope of protection.
[0019] The method of analysis of the sample surface is performed using the device with the charged particle beam visible in fig. 1 . The particles are generated by at least one charged particle source 1_. The charged particle source is a source of electrons or a source of ions.
[0020] The device further contains at least one column 2 and a chamber 3. The column 2 is connected to the chamber 3. The charged particle source 1_ is placed in the column 2. A set 7 of elements for shaping and directing the particles is further placed in the column 2. The elements for shaping and directing the particles are, for example, electromagnetic lenses, stigmators, electrodes at a potential, apertures, and other commonly used components for particle optics. In case the device contains two columns connected to the chamber 3, these are connected such that their optical axes form an angle greater than 0° and smaller than 180°. An example of such a device is a scanning electron microscope with a second column producing a focused ion beam.
[0021] The charged particle source 1_ irradiates a certain region on the sample referred to as the field of view using the set 7 of elements for shaping and directing the particles.
[0022] The charged particle beam device further contains at least one signal particle detector 8. In particular, the signal particles are secondary or backscattered particles, or other particles emitted by the sample 5 due to interaction with the incident charged particle beam or particles backscattered by the sample 5. The signal particle detector 8 may be placed in the chamber 3 or in the column 2. The signal particle detector 8 is used to acquire the image of the sample 5 following the irradiation of the sample 5 by the charged particle beam.
[0023] A stage 4 for placing at least one sample 5 is further located in the chamber 3. The stage 4 is adapted to move along at least two mutually perpendicular axes. In an exemplary embodiment of the stage 4, the stage 4 is adapted to move along three mutually perpendicular axes. In another exemplary embodiment of the stage 4, the stage 4 is further adapted to rotate around at least one axis and tilt around at least one axis. The exemplary embodiments of the stage 4 may be arbitrarily combined. The stage 4 is adapted for placement of at least one sample 5.
[0024] In an exemplary embodiment, the device further contains a system for supplying a gaseous precursor containing a reservoir 9 containing at least partially gaseous precursor and a system 10 for conducting the gaseous precursor connected to the reservoir 9 and opening, at its other end, into the chamber 3.
[0025] The device further comprises an evaluation unit for performing the method. The evaluation unit contains memory.
[0026] Information about at least one matrix window size is stored in the memory. If multiple matrix window sizes are stored in the memory, these sizes are always different. In the first preferred embodiment of the matrix window size, the matrix window size corresponds to the size of the alignment mark. The alignment mark is an orientation element created artificially on the sample intended for orientation on the sample 5, typically with a specific shape. The alignment mark may be created by adding material to the sample 5, e.g., using the system for supplying the gaseous precursor, or conversely by removing material from the sample 5. The size of the alignment mark is selected according to the sample 5 surface. If the sample 5 surface is heterogeneous, topographically diverse, or there is a foreign object on the surface, or any combination of said factors, a larger size of the alignment marks is usually selected to ensure, for example, their correct identification by machine vision. In the second exemplary embodiment of the matrix window size, the matrix window size corresponds to the size of the entire acquired image of the sample 5 divided by an integer greater than 1. This approach can be used, for example, to determine the size of foreign objects on the sample 5 surface.
[0027] The memory further stores the limit number of sub-histograms different from the histogram of the acquired image of the sample 5. The number of sub-histograms different from the histogram of the acquired image of the sample 5 is pre-determined by the user, for example based on a requirement for the reliability of the detection of the alignment mark by machine vision in case the matrix window size corresponds to the possible size of the alignment mark, where if a more reliable detection of the alignment mark is required, a lower limit number of sub-histograms different from the histogram of the acquired image of the sample 5 is selected. The memory further stores a decision criterion where two histograms are still considered similar. The decision criterion is pre-determined by the user, for example based on a requirement for the reliability of the detection of the alignment mark by machine vision, or based on a requirement for the precision of determining one of the properties of the sample 5 surface from a group of properties including heterogeneous surface, topographically diverse surface, or surface with a foreign object, or a combination of the above properties.
[0028] The method comprises a first step and a second step.
[0029] In the first step of the method, the image of the sample 5 is acquired by irradiating the sample 5 by the charged particle beam and detecting the generated signal particles by the signal particle detector 8. The sample 5 is irradiated at such irradiation parameters that the acquired image of the sample 5 has a histogram without saturated areas. The histogram without saturated areas is a histogram with distribution of data which is free of extreme values. The irradiation parameters include the accelerating voltage of the charged particles, the current of the charged particle beam, the exposure time of each pixel, and others.
[0030] The second step of the method further comprises the following sequentially performed sub-steps.
[0031] A sub-step of dividing the acquired image of the sample 5 by the evaluation unit into a matrix with a matrix window size corresponding to the matrix window size stored in the memory.
[0032] The second step further comprises a sub-step of creating sub-histograms of the individual matrix windows.
[0033] The second step further comprises a sub-step of calculating the similarity of the individual sub-histograms to the histogram of the acquired image of the sample 5 based on the decision criterion stored in the memory. The similarity of the individual subhistograms to the histogram of the acquired image of the sample 5 may be calculated using any suitable statistical method. Such statistical methods include, for example, chi- squared test, Bhattacharyya method, Hellinger method, Anderson-Darling test. Kolmogorov-Smirnov test or Kullback-Leibler test. The decision criterion depends on the statistical method used, e.g., in the case of the chi-squared test, the decision criterion is the significance level. The similarity of the sub-histograms is compared to the histogram of the entire, i.e. , the acquired image of the sample 5, from which the sub-histograms originate. The sub-histograms are compared with the histogram of the acquired image of the sample 5, since if the sub-histograms were compared to each other, the result, their similarity, could be affected, for example, by differences in brightness in different areas of the sample 5, thus, even though it is not a case where the sample 5 surface, at the given matrix window size, has at least one property from the group of properties including heterogeneous surface, topographically diverse surface, or surface with a foreign object, an incorrect evaluation could occur. The histogram of the entire acquired image of the sample 5 is influenced by the entire sample 5 surface, and therefore, there is a greater probability of a correct evaluation of its similarity to the individual parts, the individual subhistograms, than if the sub-histograms were compared to each other.
[0034] The second step further comprises a sub-step of issuing information that the subhistograms are not similar to the histogram of the acquired image of the sample 5 at the given matrix window size if the number of sub-histograms different from the histogram of the acquired image of the sample 5 exceeds the limit number of sub-histograms different from the histogram of the acquired image of the sample 5 stored in the memory, or that the sub-histograms are similar to the histogram of the acquired image of the sample 5 at the given matrix window size if the number of sub-histograms different from the histogram of the acquired image of the sample 5 does not exceed the limit number of sub-histograms different from the histogram of the acquired image of the sample 5 stored in the memory.
[0035] The second step of the method may be performed repeatedly. In the first exemplary embodiment of the repetition of the second step, first the matrix window size corresponding to the smallest one of the matrix window sizes stored in the memory is selected, and then each time a successively larger one of the matrix window sizes stored in the memory than in the previous embodiment of the second step. The repetition is performed until either information that the sub-histograms are similar to the histogram of the acquired image of the sample 5 is issued, or until the matrix window with the largest one of the matrix window sizes stored in the memory is selected. In the second exemplary embodiment of the repetition of the second step, first the matrix window size corresponding to the largest one of the matrix window sizes stored in the memory is selected, and then each time a successively smaller one of the matrix window sizes stored in the memory than in the previous embodiment of the second step, wherein the repetition is performed until either information that the sub-histograms are not similar to the histogram of the acquired image of the sample 5 is issued, or until the matrix window with the smallest one of the matrix window sizes stored in the memory is selected. In this exemplary embodiment, information on specifically which sub-histogram(s) is / are not similar to the histogram of the acquired image of the sample 5 may be further issued. This information is then used to precisely determine the area on the sample 5 surface that has one of the properties from the group of properties including heterogeneous surface, topographically diverse surface, or surface with a foreign object, or an arbitrary combination thereof.
[0036] A specific exemplary embodiment of the method will be described below. In this exemplary embodiment, two matrix window sizes corresponding to two sizes of the alignment marks are stored in the memory. In this specific exemplary embodiment, the 1stmatrix window size and thus the potential alignment marks correspond to the total width of the acquired image of the sample 5 in this exemplary embodiment divided by 3; the 2ndmatrix window size and thus the potential alignment marks correspond to the total width of the acquired image of the sample 5 in this exemplary embodiment divided by 2. In this specific exemplary embodiment, the statistical method used to calculate the similarity of the individual sub-histograms to the histogram of the acquired image 5 will be the chi-squared test. The decision criterion of a significance level of 0.001 and the limit number of a maximum of two sub-histograms different from the histogram of the acquired image of the sample 5 are further stored in the memory.
[0037] In the sub-histograms visible in figs. 6, 8, 10, 12, 14, 16, 18, 20, 22, 25, 27, 29, and 31 , the lighter gray shows the histogram of the corresponding matrix window always visible in the preceding figs. 5, 7, 9, 11 , 13, 15, 17, 19, 21 , 24, 26, 28, and 30, and the darker gray shows the histogram of the acquired image of the sample 5 of fig. 3.
[0038] In the first step, the sample placed on the stage 4 in the chamber 3 of the charged particle beam device is irradiated with the charged particle beam, in this specific exemplary embodiment an electron beam. The signal particle detector 8 is then used to acquire the image of the sample 5 visible in fig. 2 with the histogram visible in fig. 3.
[0039] In the second step, a sequence of sub-steps is performed. In the first sub-step, the acquired image of the sample 5 is divided by the evaluation unit into a 3x3 matrix, which corresponds to the smallest one of the stored matrix window sizes, as indicated in fig. 4 and as seen in figs. 5, 7, 9, 11 , 13, 15, 17, 19, and 21. In the second sub-step, subhistograms of the individual matrix windows are created, as seen in figs. 6, 8, 10, 12, 14, 16, 18, 20, and 22. In the third sub-step, the similarity of the individual sub-histograms to the histogram of the acquired image of the sample 5 of fig. 3 is calculated based on the significance level stored in the memory. This calculation shows that the 4 sub-histograms are not similar to the histogram of the acquired image of the sample 5 of fig. 3. In the fourth sub-step, information that the sub-histograms are not similar to the histogram of the acquired image of the sample 5 of fig. 3 at the matrix window size is issued, since the number of different sub-histograms is 4 and the limit number of possible different subhistograms is 2. Therefore, the sample surface at this matrix window size has one of the properties from the group of properties including heterogeneous surface, topographically diverse surface, or surface with a foreign object, and thus it is not possible to use the alignment mark of this size corresponding to the matrix window size used in this embodiment of the second step, as it would not be recognized on this sample surface.
[0040] Since information that the sub-histograms are not similar to the histogram of the acquired image of the sample 5 of fig. 3 was issued, and since another matrix window size is stored in the memory that is larger than the one used in the first embodiment of the second step, the second step is performed repeatedly with this larger matrix window size. In the first sub-step, the acquired image of the sample 5 is divided by the evaluation unit into a 2x2 matrix, which corresponds to the second smallest one, i.e. , larger one, of the stored matrix window sizes, as indicated in fig. 23 and as seen in figs. 24, 26, 28, and 30. In the second sub-step, sub-histograms of the individual matrix windows are created, as seen in figs. 25, 27, 29, and 31. In the third sub-step, the similarity of the individual sub-histograms to the histogram of the acquired image of the sample 5 of fig. 3 is calculated based on the significance level stored in the memory. In this case, the calculation shows that 0 sub-histograms are not similar to the histogram of the acquired image of the sample 5 of fig. 3. In the fourth sub-step, information that the sub-histograms are similar to the histogram of the acquired image of the sample 5 of fig. 3 at the matrix window size is issued, since the number of different sub-histograms is 0 and the limit number of possible different sub-histograms is 2. Therefore, the sample surface at this matrix window size does not have any of the properties from the group of properties including heterogeneous surface, topographically diverse surface, or surface with a foreign object, and thus it is possible to use the alignment mark of this size, as it would be recognized on this sample surface.
[0041] Industrial Applicability
[0042] The method described above may be used with a device other than the charged particle beam device, for example a device using a beam of photons.
[0043] List of Reference Signs
[0044] 1 - Charged particle source
[0045] 2 - Column
[0046] 3 - Chamber 4 - Stage
[0047] 5 - Sample
[0048] 6 - Manipulator
[0049] 7 - Set of elements for shaping and directing the particles
[0050] 8 - Signal particle detector 9 - Reservoir
[0051] 10 - System for conducting the gaseous precursor
Claims
CLAIMS1 . A method of analysis of a sample surface using a charged particle beam device comprising at least one charged particle source (1 ), at least one signal particle detector (8), a stage (4) adapted for placing a sample (5), the sample (5) placed on the stage (4), and an evaluation unit for performing the method, wherein the evaluation unit contains a memory, characterized in that the memory stores information about at least one matrix window size, a limit number of subhistograms different from the histogram of the acquired image of the sample (5), and a decision criterion where two histograms are still similar, wherein the method comprises a first step of acquiring the image of the sample (5) by irradiating the sample (5) with a charged particle beam at such irradiation parameters where the acquired image of the sample (5) has a histogram without saturated areas, and it further comprises a second step comprising sequentially performed sub-steps:- a sub-step of dividing the acquired image (5) by the evaluation unit into a matrix with a matrix window size corresponding to the matrix window size stored in the memory,- a sub-step of creating sub-histograms of the individual matrix windows,- a sub-step of calculating the similarity of the individual sub-histograms to the histogram of the acquired image of the sample (5) based on the decision criterion,- a sub-step of issuing information that the sub-histograms are not similar to the histogram of the acquired image of the sample (5) at the given matrix window size, if the number of sub-histograms different from the histogram of the acquired image of the sample (5) exceeds the limit number of sub-histograms different from the histogram of the acquired image of the sample (5) stored in the memory, or that the sub-histograms are similar to the histogram of the acquired image of the sample (5) at the given matrix window size, if the number of sub-histograms different from the histogram of the acquired image of the sample (5) does not exceed the limit number of sub-histograms different from the histogram of the acquired image of the sample (5) stored in the memory.and it further comprises a third step of determining the size of the alignment mark corresponding to the matrix window size stored in the memory.
2. The method of analysis of a sample surface using a charged particle beam device according to claim 1 , characterized in that the second step is performed repeatedly, with the matrix window size corresponding to the smallest one of the matrix window sizes stored in the memory being selected first, and then each time a successively larger one of the matrix window sizes stored in the memory than in the previous embodiment of the second step, wherein the repetition is performed until either information that the sub-histograms are similar to the histogram of the acquired image of the sample (5) is issued, or until the matrix window with the largest one of the matrix window sizes stored in the memory is selected.
3. The method of analysis of a sample surface using a charged particle beam device according to claim 1 , characterized in that the second step is performed repeatedly, with the matrix window size corresponding to the largest one of the matrix window sizes stored in the memory being selected first, and then each time a successively smaller one of the matrix window sizes stored in the memory than in the previous embodiment of the second step, wherein the repetition is performed until either information that the sub-histograms are not similar to the histogram of the acquired image of the sample (5) is issued, or until the matrix window with the smallest one of the matrix window sizes stored in the memory is selected.