A computer-aided defect correction method of a charged particle beam device, defect correction device of a charged particle beam device and a computer program

Abstract

The present invention provides a computer-aided defect correction method of a charged particle beam device (1) comprising the consecutive steps of: irradiation of a sample (16) by a convergent charged particle beam, acquiring at least one diffraction pattern of the sample (16), analysis of at least one acquired diffraction pattern of the sample (16), and changing of the value of at least one operational parameter of the charged particle beam device (1) based on analysis of at least one acquired diffraction pattern of the sample (16). The analysis step is performed by a computing module (4) comprising an artificial intelligence correction module (5), wherein the output of the analysis of the at least one acquired diffraction pattern of the sample (16) by the artificial intelligence correction module (5) is a set of information based on which the computing module (4) calculates output values of operational parameters affecting at least one of the following defects: beam tilt relative to the optical axis (18) of the charged particle beam device (1), beam defocus and astigmatism. The invention also includes defect correction device of a charged particle beam device (1) and a computer program.

Description

[0001] A computer-aided defect correction method of a charged particle beam device, defect correction device of a charged particle beam device and a computer program

[0002] Field of the Invention

[0003] The invention relates to a method for correcting defects in a charged particle beam device, in particular a transmission electron microscope or a scanning transmission electron microscope. More particularly, the invention relates to the correction of beam tilt relative to the optical axis of a charged particle beam device, beam defocus or astigmatism.

[0004] Background of the Invention

[0005] In the current state of the art, transmission electron microscope solutions are known, which use the irradiation through the sample by a beam of charged particles - electrons. Electrons passing through the sample are detected by a detector and information about the internal structure of the sample is acquired. When an amorphous sample is irradiated by a convergent electron beam, a diffraction pattern is formed, which is commonly known in the field as a ronchigram. By analysing the ronchigram, i.e. its shape or position, it is possible to infer various optical defects of the microscope and to adjust the values of its operational parameters appropriately in order to improve the quality of the image. For example, it is known that the displacement of the ronchigram off the optical axis of the microscope in the diffraction plane of the microscope can be identified with the angle of the beam tilt relative to the optical axis. The elliptical shape of the ronchigram indicates the degree of astigmatism and the diameter of the ronchigram, specifically the diameter of its central part, indicates the defocus of the beam. The analysis of the Ronchigram is described e.g. in the article Introduction to the Ronchigram and its Calculation with Ronchigram.com (Schnitzer et al., 2019), while the authors have also created a web application ronchigram.com allowing the simulation of ronchigrams based on the input of individual parameters. This application can help in understanding the microscope correction, but the actual correction has to be done by the respective worker, as well as the input of the parameters into the application. Therefore, automatic correction of the relevant operational parameters to improve the quality of the image is not ensured.

[0006] Furthermore, Schnitzer et al. in Optimal STEM Convergence Angle Selection Using a Convolutional Neural Network and the Strehl Ratio (2020) and Maximal Resolution from the Ronchigram: Human vs. Deep Learning (2019) describe the use of convolutional neural networks for ronchigram analysis to find the optimal convergence angle of the beam relative to the optical axis of a corrected scanning electron microscope. Specifically, the convergence angle corresponding to the optimal resolution is sought. However, this document does not disclose corrections for the aforementioned defects (astigmatism and beam defocus), nor does it address the problem of automatic correction of beam tilt relative to the optical axis of the microscope. However, corrections of these phenomena are essential for fast and accurate measurements.

[0007] Other alternative ways to achieve correction of various parameters of a charged particle beam device are described in patent application EP4044210A1 , patent US9312094B2 and patent application EP4123681 A2.

[0008] Therefore, it would be desirable to come up with a solution that would allow automated adjustment of the operational parameters of the charged particle beam device, specifically to correct the beam tilt relative to the optical axis of the charged particle beam device, astigmatism and beam defocus to improve the quality of the image. of the invention

[0009] The above deficiencies are to some extent eliminated by a computer-aided defect correction method of charged particle beam devices comprising consecutive steps:

[0010] - irradiation of a sample by a convergent charged particle beam, wherein the irradiation being carried out at a set of input values of the operational parameters of the charged particle beam device,

[0011] - acquiring at least one diffraction pattern of the sample from the charged particle beam passed through the sample,

[0012] - analysis of at least one acquired diffraction pattern of the sample; and

[0013] - changing of the value of at least one operational parameter of the charged particle beam device based on analysis of at least one acquired diffraction pattern of the sample to correct at least one defect in the charged particle beam device.

[0014] The essence of the method according to the present invention is that the step of analysing the at least one acquired diffraction pattern of the sample is performed by a computing module comprising an artificial intelligence correction module, wherein the artificial intelligence correction module comprises a classifier trained on the set of diffraction patterns, wherein the output of the analysis of the at least one acquired diffraction pattern of the sample by the artificial intelligence correction module is a set of information based on which the computing module calculates output values of operational parameters of the charged particle beam device affecting at least one of the following defects of the charged particle beam device: beam tilt relative to the optical axis of the charged particle beam device, beam defocus and astigmatism.

[0015] The advantage of the method according to the present invention is that it allows automatic correction of defects of the charged particle beam device, namely astigmatism and defocus of the charged particle beam, and also the tilt of the beam relative to the optical axis, whether the correction concerns only one of the above defects or a combination of several of these defects. This improves the display quality of the charged particle beam device and enables fast and accurate measurements. The output of the artificial intelligence correction module is the coordinates of the center of the at least one acquired diffraction pattern of the sample, wherein the set of output values of the operational parameters of the charged particle beam device affecting the beam tilt relative to the optical axis of the charged particle beam device is calculated based on the coordinates of the center of the at least one acquired diffraction pattern of the sample. The correction of the beam tilt relative to the optical axis of the charged particle beam device is performed by changing the operational parameter control signal of the centering element or by changing the operational parameter offset of the rasterizing element. By control signal is meant the current or voltage of the centering element or the rasterizing element, respectively, according to a particular embodiment of these elements. The correction of the beam tilt relative to the optical axis is thus implemented by changing the setting of the centering element or the rasterizing element.

[0016] The output of the artificial intelligence correction module is a vector comprising information about the direction and magnitude of the astigmatism, wherein the set of output values of the operational parameters of the charged particle beam device affecting the astigmatism is calculated based on the vector comprising information about the direction and magnitude of the astigmatism. The correction of the astigmatism is performed by changing the operational parameter control signal of the stigmator. The control signal of stigmator refers to the current or voltage of the stigmator according to the particular design of the element. The correction of astigmatism is thus implemented by changing the setting of the stigmator.

[0017] The output of the artificial intelligence correction module is a beam defocus value, wherein the set of output values of the operational parameters of the charged particle beam device affecting the beam defocus is calculated based on the beam defocus value that is the output of the artificial intelligence correction module. The correction of the beam defocus is performed by changing an operational parameter control signal of the objective lens or by changing an operational parameter control signal of the condensers. By control signal is meant the current or voltage of the objective lens and the condensers, respectively, i.e. the first condenser or the second condenser, or additional condensers if the device includes more than two condensers, according to the particular design of these elements. The correction of the beam defocus is thus implemented by changing the setting of the objective lens, the first condenser or the second condenser. Set of diffraction pattern images for training the classifier is advantageously generated based on a simulation model that reflects the operation of the charged particle beam device. In this way, various defects of the charged particle beam device can be simulated, but possible sample inhomogeneities, defects or irradiation at the interfaces of individual crystals, etc. can also be included in the simulations. Thus, a large set of diffraction pattern images for training the classifier can be easily and quickly acquired without the need to use actual images acquired by real and time-consuming experiments for training.

[0018] The diffraction pattern of the sample is acquired without an aperture in the illuminating part of the charged particle beam device. In this variant, the entire ronchigram is simply acquired.

[0019] The diffraction pattern of the sample is acquired in the presence of an aperture in the illuminating part of the charged particle beam device. The at least one acquired diffraction pattern of the sample is a series of diffractograms recorded at at least two values of the beam tilt angle relative to the optical axis of the charged particle beam device in the presence of an aperture in the illuminating part of the charged particle beam device, wherein the series of diffractograms is an input of the artificial intelligence correction module. However, the input to the artificial intelligence correction module may also be a ronchigram which composition or reconstruction, respectively, from these partial diffractograms occurs prior to insertion into the artificial intelligence correction module.

[0020] The foregoing deficiencies are also eliminated to some extent by a defect correction device of a charged particle beam device adapted to perform the method according to the present invention, comprising a charged particle beam device and a control unit communicatively coupled to the charged particle beam device and adapted to control individual elements of the charged particle beam device, wherein the charged particle beam device comprises a detector for obtaining at least one diffraction pattern of the sample. The essence of the device according to the present invention is that the control unit comprises an artificial intelligence correction module comprising a classifier and adapted to analyse at least one acquired diffraction pattern of the sample.

[0021] The advantage of the device according to the present invention is that it enables automatic correction of defects of the charged particle beam device, namely astigmatism and defocus of the charged particle beam, as well as the beam tilt with relative to the optical axis, whether the correction concerns only one of the above defects or a combination of several of these defects. This improves the display quality of the charged particle beam device and enables fast and accurate measurements.

[0022] The control unit advantageously comprises a control module adapted to control individual elements of the charged particle beam device and a computing module, wherein the artificial intelligence correction module is part of the computing module. This arrangement ensures efficient calculation and reliable automatic adjustment of the output values of the operational parameters for the operation of the charged particle beam device.

[0023] The artificial intelligence correction module is advantageously implemented as a convolutional neural network, which enables fast and efficient analysis of the image input in order to correct the aforementioned defects of the charged particle beam device. It is a kind of neural network dedicated to work with image data.

[0024] The above deficiencies are also eliminated to some extent by a computer program comprising instructions that cause the device according to the present invention to perform the steps of the method according to the present invention.

[0025] Description of drawings

[0026] The summary of the invention is further explained by exemplary embodiments thereof, described with reference to the accompanying drawings, in which :

[0027] Fig. 1 is shown an illustrative block diagram of a defect correction device of a charged particle beam device according to the present invention,

[0028] Fig. 2 is shown a diagram of the charged particle beam device, namely a scanning transmission electron microscope, and

[0029] Fig. 3 is shown a flow diagram of a defect correction method of a charged particle beam device according to the present invention. Embodiments of the Invention

[0030] The invention will be further explained by reference to example embodiments with reference to the relevant drawings, which, however, have no limiting effect in terms of the scope of protection. In view of the logical continuity of the text, a defect correction device of a charged particle beam device 1 will first be described, followed by a computer-aided defect correction method of the charged particle beam device 1_.

[0031] The defect correction device of the charged particle beam device 1 comprises the charged particle beam device 1 and a control unit 2 communicatively coupled to the charged particle beam device 1, which is schematically shown in Fig. 1. The communication link is understood to be a link enabling the transmission of data, information or an electrical control signal between the communicatively coupled elements, wherein the communication link may also be implemented wirelessly. The control unit 2 in the first exemplary embodiment of the control unit 2 includes a control module 3 and a computing module 4. The control module 3 is adapted to control individual elements of the charged particle beam device 1, as will be described in more detail below. The computing module 4 includes an artificial intelligence correction module 5 adapted to analyse at least one acquired diffraction pattern of the sample 16, as will also be described in more detail below.

[0032] The artificial intelligence correction module 5 is implemented in the form of an artificial intelligence module, in particular a convolutional neural network, machine learning, deep learning, or other form of artificial intelligence that allows receiving image information as input, wherein the output is a set of new information acquired by analysing the image information by the artificial intelligence correction module 5. The artificial intelligence correction module 5 includes a classifier trained on a set of images of diffraction patterns. These diffraction patterns for training the classifier are acquired, in a first exemplary embodiment of the method for acquiring diffraction patterns, by simulation, that is, based on a simulation model reflecting the operation of the charged particle beam device T This simulation model may be defined more generally, i.e. it reflects the operation of the charged particle beam device 1 independently of its particular embodiment, or more simplistically, i.e. it does not reflect the operation of all parts of the charged particle beam device T Thus, the input of the simulation model is a set of values corresponding to the operational parameters of the charged particle beam device 1 or rate of its defects and the output is the simulated diffraction pattern. In an alternative exemplary embodiment of the method of acquiring diffraction patterns, the diffraction patterns for training are acquired by real experiments in which the operational parameters are varied, and are thus generated by a convergent beam of charged particles, such as electrons. It is also possible to combine both of these approaches to increase the set of diffraction pattern images and thus increase the accuracy of the trained classifier. Possible sample inhomogeneities 16, defects or irradiation at the interfaces of individual crystals, etc., are also included in the simulations. The training itself is performed by processing the images of the diffraction patterns, to which the set of values corresponding to the operational parameters is assigned.

[0033] The computing module 4 is communicatively coupled to the control module 3, which allows the output of the artificial intelligence correction module 5, which is first processed by the computing module 4, to be sent to the control module 3. The control module 3_is further communicatively coupled to the charged particle beam device 1, which allows the operational parameters of the charged particle beam device 1 to be adjusted based on the output of the computing module 4, i.e. based on a set of output values of the operational parameters, the acquisition of which will be described in more detail below. Thus, the computing module 4 is also connected to the charged particle beam device 1 via the control module 3, but may alternatively also be connected directly to the charged particle beam device 1, as indicated by the dashed line in schematic Fig. 1 . The control unit 2 is the term used to cover the two aforementioned modules, namely the control module 3 and the computing module 4, and may take the form of a physical control unit, but may also be understood as a virtual unit comprising the two aforementioned modules.

[0034] The charged particle beam device 1 in its first exemplary embodiment according to Fig. 2 is implemented as a scanning transmission electron microscope (also referred to as a rasterizing transmission electron microscope, STEM) and includes a column 6 in which the individual elements of the charged particle beam device 1 are arranged. The charged particle beam device 1 comprises a charged particle source 7, in a first exemplary embodiment of the charged particle beam device 1 an electron source, which emits charged particles in the form of a charged particle beam. The charged particles are first accelerated in the space between the charged particle source 7 and an anode 8, since the charged particle source 7 acts as a cathode. Subsequently, the charged particle beam is deflected, focused or otherwise generally influenced and modified by a set of other elements, for example successively by a centering element 9, which in its first exemplary embodiment includes two centering coils, a first condenser 10, an aperture 1 1 , a second condenser 12, a stigmator 13, a rasterizing element 14, which in its first exemplary embodiment includes two rasterizing coils, and the objective lens 15. Alternatively, a different number of condensers is used, for example four condensers. The dashed line shows the optical axis 18 of the charged particle beam device 1.

[0035] The charged particle beam device 1 further comprises a table or holder for mounting the sample 16, the table or holder being adapted for translation along at least two mutually perpendicular axes and for tilting about at least one axis, and a detector 17 for acquiring a diffraction pattern of the sample 6. The aforementioned part of the charged particle beam device which comprising elements for deflecting, directing or focusing the charged particle beam in front of the sample 16, is called the illuminating part. However, the charged particle beam device 1 may also include similar components in the region downstream of the sample 16, i.e., in the region between the sample 16 and the detector 17, which is referred to as the projection part. The elements in this area, called projection lenses, ensure that the beam is properly magnified and focused and that the majority of the charged particles reach the detector 17. Thus, the projection part ensures that the back focal plane is projected onto the detector 17.

[0036] The detector 17 is positioned at a distance behind the sample 16 such that it is able to detect the diffraction pattern. In the case where the objective lens of the charged particle beam device has a so-called immersion field, the detector 17 is located in the back focal plane where this diffraction pattern is formed. If the objective lens of the charged particle beam device does not have such an immersion field, the detector 17 is placed at a suitable distance behind the sample 16 so that the diffraction pattern is visible on the detector 17. By sample 16 is meant here an observed object, whether with unknown or known properties. A special case of the sample 16 is a calibration element corresponding to the sample 16 with known properties, wherein more particularly the calibration element is meant to be a structure with known dimensions and diffraction parameters. In the first exemplary embodiment of the charged particle beam device 1 shown in Fig. 2, the detector 17 is a camera positioned downstream of the sample 16 in the direction of passage of the charged particle beam, wherein the detector 17 detects the charged particles passed through the sample 6, thereby acquiring a diffraction pattern of the sample 1.6. This diffraction pattern of the sample 16 is a so-called ronchigram, or alternatively a series of diffractograms recorded at at least two values of the beam tilt angle relative to the optical axis 18 of the charged particle beam device 1 in the presence of the aperture 1 1 in the illuminating part of the charged particle beam device 1, as will be described below. Alternatively, the charged particle beam device 1 may be implemented as a non-STEM type of microscope, such as a TEM or SEM, if it is adapted to record diffraction patterns.

[0037] A computer-aided defect correction method according to the present invention, using a trained classifier of the artificial intelligence correction module 5, includes firstly the steps of irradiating the sample 16 with a convergent charged particle beam and then the step of acquiring at least one diffraction pattern of the sample 16. This at least one diffraction pattern of the sample 16 is acquired based on the irradiation of the sample by the convergent charged particle beam, wherein the irradiation is performed at a set of input values of the operational parameters of the charged particle beam device 1. Irradiation of the sample 16 with the convergent charged particle beam results in diffraction. The particles passed through sample 16 affected by diffraction are detected by detector 17 and the diffraction pattern of sample 16 is recorded.

[0038] The diffraction pattern of the sample 16 is acquired by at least one of the following procedures when the charged particle beam device 1 is operated. The first possibility is acquiring the diffraction pattern without the presence of the aperture 1 1 in the illuminating part of the charged particle beam device . This diffraction pattern acquired by the procedure according to the first possibility is the so-called ronchigram, known from the state of the art. The second possibility is to produce a series of diffraction patterns, so- called diffractograms, acquired for at least two values of the beam tilt angle relative to the optical axis 18 of the charged particle beam device 1 in the presence of the aperture 1 1 in the illumination part of the charged particle beam device T A ronchigram is subsequently composed of these partial diffraction patterns. The third possibility is to form a series of diffraction patterns, called diffractograms, acquired at at least two values of the beam tilt angle relative to the optical axis 18 of the charged particle beam device 1 in the presence of the aperture 1 1 in the illumination part of the charged particle beam device 1, whereby this series of diffractograms serves as input to the artificial intelligence correction module 5. Thus, in the case of the first and second possibility, the directly acquired or composed ronchigram serves as input to the artificial intelligence correction module 5, while in the third possibility, a series of diffractograms serves as input to the artificial intelligence correction module 5 without the composition of the ronchigram itself. The aperture 1 1 used in the second and third possibility must be small enough to crop the beam, wherein a specific suitable dimensions would be obvious to one skilled in the art and known from the state of the art or could also be determined from performed experiments.

[0039] The acquired diffraction pattern of the sample 16 is sent from the detector 17 to the computing module 4, more particularly to the artificial intelligence correction module 5, and an analysis of the acquired diffraction pattern of the sample 16 is performed. As also illustrated in the flow diagram of Fig. 3, this analysis includes first a step of acquiring a set of information including at least one of the following information: a coordinate of the center of the diffraction pattern of the sample 6, a vector including information about the direction and magnitude of the astigmatism, and a beam defocus value, this step being performed by the artificial intelligence correction module 5, and then a step of calculating a set of output values of the operational parameters, which is performed by the computing module 4 but no longer directly by the artificial intelligence correction module 5.

[0040] The analysis of the diffraction pattern of the sample 16 takes advantage of the fact that the shape and position of the diffraction pattern of the sample 16 are dependent on the values of the operational parameters of the charged particle beam device 1, and the defects of the charged particle beam device 1 can be identified from the shape and position of the diffraction pattern of the sample 6. Specifically, the beam tilt relative to the optical axis 18 of the charged particle beam device 1 causes the centre of the diffraction pattern to be deflected away from the optical axis 18 of the charged particle beam device 1. The symmetry of the standard circular diffraction pattern may be further disturbed by the astigmatism of the beam, wherein the degree of astigmatism is related to the magnitude of asymmetry of the circular shape of the diffraction pattern, which becomes elliptical. The size, respectively diameter of the diffraction pattern, namely the diameter of its central part, then corresponds to the degree of defocus of the beam. Thus, each of the above mentioned defects (beam tilt relative to the optical axis 18 of the charged particle beam device 1, beam defocus and astigmatism) can be corrected by a suitable change in the relevant operational parameters, whereby a set of new, output values of these operational parameters is calculated by the calculation module 4 on the basis of information (coordinates of the centre of the diffraction pattern, a vector including information of the magnitude and direction of astigmatism and beam defocus) from the correction module 5 of the artificial intelligence. The actual calculation of the set of output values of the operational parameters is performed by a well-known procedure, which will also be approximately described below.

[0041] The above-mentioned operational parameters of the charged particle beam device 1 are parameters selected from the group including: an operational parameter control signal of the stigmator 13, which affects the astigmatism defect, an operational parameter control signal of the objective lens 15, an operational parameter control signal of the condensers 10, 12, i.e. control signal of the first condenser 10 or the control signal of the second condenser 12, or also the control signal of the other condensers, if the charged particle beam device IJncludes them, wherein these operational parameters affect the defocus defect of the beam, and the operational parameter control signal of the centering element 9_and the operational parameter offset of the rasterizing element 1.4, which affect the defect of the beam tilt relative to the optical axis 18 of the charged particle beam device 1.

[0042] The correction of the beam tilt relative to the optical axis 18 of the charged particle beam device 1 will now be described in more detail. First, the coordinates of the center of the acquired diffraction pattern are detected by the artificial intelligence correction module 5, and then the deviation of these coordinates from the optical axis 18 of the charged particle beam device 1 is calculated by the calculation module 4. Subsequently, based on the deviation calculated in the previous step, the computing module 4 calculates an output value of at least one operational parameter affecting the beam tilt relative to the optical axis 18 of the charged particle beam device 1_so that the output value of the operational parameter affecting the beam tilt relative to the optical axis 18 of the charged particle beam device 1 affects the shape of the diffraction pattern so that the position of the centre of the diffraction pattern is at least closer to, but ideally on, the optical axis 18 of the charged particle beam device T The output value of the operational parameter affecting the beam tilt angle relative to the optical axis 18 of the charged particle beam device 1 is sent to the control module 3 and is subsequently set by this module for further operation of the charged particle beam device 1_. The minimization of the said deviation is realized by means of a suitable optimization algorithm and is performed cyclically. That is to say, firstly the operational parameters of the charged particle beam device 1 are changed, then the deviation is again calculated and the output values of the operational parameters are calculated, which are again sent to the control module 3 and set on the charged particle beam device 1 until the deviation is at the desired minimum value.

[0043] The operational parameter that affects the value of the beam tilt angle relative to the optical axis 18 of the charged particle beam device 1 is the control signal of the centering element 9. Specifically, when the centering element 9 is made so that it includes centering coils, this control signal is the current through the centering coils. Alternatively, the centering element 9 may also be implemented as an electrostatic dipole or a combination thereof, in which case said control signal is a voltage on the element, not the current. Another operational parameter which affects the value of the beam tilt angle relative to the optical axis 18 of the charged particle beam device 1 is the offset of the rasterizing element 14, for example the offset of the rasterizing coils.

[0044] The correction of the beam astigmatism, which is manifested by the asymmetry of the diffraction pattern, will now be described in more detail. Firstly, a vector including information about the direction and magnitude of the astigmatism is acquired by the correction module 5 of the artificial intelligence, i.e. specifically, two values which define this vector are acquired. These values correspond, for example, to the x- and y- components of the vector. Subsequently, the magnitude of the vector as the square root of the sum of the squares of the acquired values and also the direction of this vector are calculated by computational module 4, whereby the magnitude of this vector corresponds to the degree of astigmatism and the direction of this vector corresponds to the direction of astigmatism. Using a suitable optimization algorithm, the output value of the operational parameter affecting the astigmatism of the beam is then calculated, wherein the output value of the operational parameter affecting the astigmatism of the beam affects the shape of the diffraction pattern such that its formerly elliptical shape is at least more circular, respectively the longer of the axes of the elliptical shape is either shortened or the shorter of the axes of the elliptical shape is lengthened, at the best it is circular. The output value of the operational parameter affecting the astigmatism is sent to the control module 3 and is subsequently set by this module for further operation of the charged particle beam device T The optimization algorithm operates cyclically until the degree of the astigmatism is at the desired minimum value, similarly to the beam tilt correction described above. The operational parameter that affects the value of the astigmatism defect is the control signal of the stigmator 13. In a first exemplary embodiment of the stigmator 13, the stigmator 13 is implemented to include eight coils, wherein four coils provide astigmatism correction in one direction and four additional coils (rotated 45° relative to the others) provide astigmatism correction in the second direction. Thus, in this embodiment, the control signal of the stigmator 13 is the current through the coils of the stigmator 13. Alternatively, the stigmator 13 may also be implemented with a different number of coils, namely at least four coils, the various possible embodiments of the stigmator 13 being known in the prior art. Alternatively, the stigmator 13 may also be implemented electrostatically, in which case the control signal of the stigmator 13 would be a voltage on the element.

[0045] Now the beam defocus correction will be described in more detail. Firstly, the beam defocus value is acquired by the artificial intelligence correction module 5 and an interval of acceptable beam defocus values is determined, wherein the beam defocus is based on the diameter of the diffraction pattern. Subsequently, an output value of the operational parameter affecting the defect of the beam defocus is calculated by the computation module 4 by means of an optimization algorithm and is sent to the control module 3 and subsequently set by this module for further operation of the charged particle beam device 1 so that the value of the beam defocus is zero or is another desired value, e.g. to lie within a defined interval of acceptable values of the beam defocus. The optimization algorithm operates cyclically until the output value of the operational parameter influencing the beam defocus defect is such that the beam defocus value falls within a defined interval of acceptable values of diffraction pattern diameters.

[0046] The operational parameter that affects the value of the beam defocus defect is the control signal of the objective lens 15. Specifically, if the objective lens 15 is implemented as an electromagnetic lens, the control signal is the current through the coil of the objective lens 15. Alternatively, the objective lens 15 may also be implemented as an electrostatic lens, in which case the control signal is the voltage on the element, not the current. Other operational parameters which affect the value of the beam defocus defect are the control signal of the condensers 10, 12, i.e., exemplary the control signal of the first condenser 10 or the control signal of the second condenser 12, or the control signal of other condensers if the charged particle beam device 1 includes them. This control signal is the current through the coil or coils of the condensers 10, 12 if they are implemented as electromagnetic lenses. Alternatively, the condensers 10, 12 may be made as electrostatic lenses, in which case said control signal is a voltage on these elements, not the current.

[0047] Further, any combination of at least two of the aforementioned defects in the form of the diffraction pattern may be detected by the correction module 5 of the artificial intelligence, whereby the output of the calculation module 4 is a corresponding number of output values of the operational parameters affecting the given types of defects, whereby the output values of the operational parameters are sent to the control module 3 and are subsequently set by this module for further operation of the charged particle beam device 1 .

[0048] Industrial applicability

[0049] The present invention can be used for automated adjustment of operational parameters and correction of defects in a scanning transmission electron microscope, or other charged particle beam device, to improve the quality of imaging. The present invention can also be used for correcting defects in charged particle beam devices with a different than aforementioned arrangement of the illuminating part.

[0050] List of reference signs

[0051] 1 - charged particle beam device

[0052] 2 - control unit

[0053] 3 - control module

[0054] 4 - computing module

[0055] 5 - artificial intelligence correction module

[0056] 6 - column

[0057] 7 - charged particle source

[0058] 8 - anode

[0059] 9 - centering element

[0060] 10 - the first condenser

[0061] 11 - aperture

[0062] 12 - second condenser

[0063] 13 - stigmator

[0064] 14 - rasterizing element

[0065] 15 - objective lens

[0066] 16 - sample

[0067] 17 - detector

[0068] 18 - optical axis of the charged particle beam device

Claims

PATENT CLAIMS1 . A computer-aided defect correction method of a charged particle beam device (1 ) comprising the consecutive steps of:- irradiation of a sample (16) by a convergent charged particle beam, wherein the irradiation being carried out at a set of input values of the operational parameters of the charged particle beam device (1 ),- acquiring at least one diffraction pattern of the sample (16) from the charged particle beam passed through the sample (16),- analysis of at least one acquired diffraction pattern of the sample (16); and- changing of the value of at least one operational parameter of the charged particle beam device (1 ) based on analysis of at least one acquired diffraction pattern of the sample (16) to correct at least one defect in the charged particle beam device (1 ), characterized in that the step of analysing the at least one acquired diffraction pattern of the sample (16) is performed by a computing module (4) comprising an artificial intelligence correction module (5), wherein the artificial intelligence correction module (5) comprises a classifier trained on the set of diffraction patterns, wherein the output of the analysis of the at least one acquired diffraction pattern of the sample (16) by the artificial intelligence correction module (5) is a set of information based on which the computing module (4) calculates output values of operational parameters of the charged particle beam device (1 ) affecting at least one of the following defects of the charged particle beam device (1 ): beam tilt relative to the optical axis (18) of the charged particle beam device (1 ), beam defocus and astigmatism.

2. The method according to claim 1 , characterized in that the output of the artificial intelligence correction module (5) is the coordinates of the center of the at least one acquired diffraction pattern of the sample (16), wherein the set of output values of the operational parameters of the charged particle beam device (1 ) affecting the beam tilt relative to the optical axis (18) of the charged particle beam device (1 ) iscalculated based on the coordinates of the center of the at least one acquired diffraction pattern of the sample (16).

3. The method according to claim 2, characterized in that the correction of the beam tilt relative to the optical axis (18) of the charged particle beam device (1 ) is performed by changing the operational parameter control signal of the centering element (9) or by changing the operational parameter offset of the rasterizing element (14).

4. The method according to any one of the preceding claims 1 to 3, characterized in that the output of the artificial intelligence correction module (5) is a vector comprising information about the direction and magnitude of the astigmatism, wherein the set of output values of the operational parameters of the charged particle beam device (1 ) affecting the astigmatism is calculated based on the vector comprising information about the direction and magnitude of the astigmatism.

5. The method according to claim 4, characterized in that the correction of the astigmatism is performed by changing the operational parameter control signal of the stigmator (13).

6. The method according to any one of the preceding claims 1 to 5, characterized in that the output of the artificial intelligence correction module (5) is a beam defocus value, wherein the set of output values of the operational parameters of the charged particle beam device (1 ) affecting the beam defocus is calculated based on the beam defocus value that is the output of the artificial intelligence correction module (5).

7. The method according to claim 6, characterized in that the correction of the beam defocus is performed by changing an operational parameter control signal of the objective lens (15) or by changing an operational parameter control signal of the condensers (10, 12).

8. The method according to any one of the preceding claims 1 to 7, characterized in that the set of diffraction pattern images for training the classifier is generated based on a simulation model that reflects the operation of the charged particle beam device (1 ).

9. The method according to any one of the preceding claims 1 to 8, characterized in that the diffraction pattern of the sample (16) is acquired without an aperture (11 ) in the illuminating part of the charged particle beam device (1 ).

10. The method according to any one of the preceding claims 1 to 8, characterized in that the diffraction pattern of the sample (16) is acquired in the presence of an aperture (1 1 ) in the illuminating part of the charged particle beam device (1 ).1 1 .The method according to claim 10, characterized in that the at least one acquired diffraction pattern of the sample (16) is a series of diffractograms recorded at at least two values of the beam tilt angle relative to the optical axis (18) of the charged particle beam device (1 ) in the presence of an aperture (1 1 ) in the illuminating part of the charged particle beam device (1 ), wherein the series of diffractograms is an input of the artificial intelligence correction module (5).

12. A defect correction device of a charged particle beam device (1 ) adapted to perform the method according to any one of the preceding claims 1 to 11 , comprising a charged particle beam device (1 ) and a control unit (2) communicatively coupled to the charged particle beam device (1 ) and adapted to control individual elements of the charged particle beam device (1 ), wherein the charged particle beam device (1 ) comprises a detector (17) for obtaining at least one diffraction pattern of the sample (16), characterized in that the control unit (2) comprises an artificial intelligence correction module (5) comprising a classifier and adapted to analyse at least one acquired diffraction pattern of the sample (16).

13. The device according to claim 12, characterized in that the control unit (2) comprises a control module (3) adapted to control individual elements of the charged particle beam device (1 ) and a computing module (4), wherein the artificial intelligence correction module (5) is part of the computing module (4).

14. The device according to any one of the preceding claims 12 or 13, characterized in that the artificial intelligence correction module (5) is implemented as a convolutional neural network.

15. A computer program characterized in that it includes instructions that cause the device according to any one of claims 12 to 14 to perform the method steps according to any one of claims 1 to 1 1 .

Images