Method for determining a probe tip shape

Abstract

A method of determining a tip shape indication of a tip shape of a probe tip disposed on a probe of a scanning probe microscope mapping a surface, the method comprising steps of receiving measurement data of a measurement of a surface topography of a surface; determining, based on the measurement data, the tip shape indication of the tip shape of the probe tip; outputting, based on the step of determining, the tip shape indication; wherein the step of determining the tip shape indication comprises providing the measurement data to a trained machine learning data processing model; wherein the trained machine learning data processing model performs data pattern recognition on the measurement data for deriving the tip shape indication from the measurement data.

Description

[0001] Title: METHOD FOR DETERMINING A PROBE TIP SHAPE

[0002] TECHNICAL FIELD AND BACKGROUND

[0003] The present disclosure relates to the field of scanning probe microscopy, a more specifically, to a method for determining a tip shape of a probe tip disposed on a probe of a scanning probe microscope, in particular an atomic force microscope.

[0004] Scanning probe microscopy is a useful technique for imaging topography of surfaces at atomic and molecular levels. A scanning probe microscope moves a tip of a probe across a sample surface. This scanning allows for a generation of detailed images of surface features at a nanometer scale. The obtained images are a convolution of the actual structures of the sample surface and a tip shape of the probe used to scan the sample surface. The resolution and accuracy of a scanning probe microscope measurements are therefore highly dependent on the tip condition, including its shape and sharpness. Over time, probe tips can undergo wear and damage during use, leading to a degradation in imaging quality and measurement accuracy. Monitoring and estimating the tip shape are essential for maintaining the reliability of scanning probe microscope measurements.

[0005] Traditional methods for assessing tip shape involve offline techniques such as imaging the tip shape using a scanning electron microscopy, which require the removal and examination of the probe tip outside of the scanning probe microscope. These methods are timeconsuming and prone to inaccuracies. In-situ monitoring methods during operation of the scanning probe microscope are limited to a blind tip estimation using an algorithm. This is relatively slow because of the computational complexity of applying the algorithm, is further relatively inaccurate, and sensitive to measurement noise. Therefore, existing techniques cannot provide real-time information on tip conditions of a probe tip disposed on a probe of a scanning probe microscope.

[0006] In view of the above, there remains need for a method of fast and accurate determination of a tip shape of a scanning microscope probe, in particular during its operation.

[0007] Aspects of the present disclosure relate to a method of determining a tip shape indication of a tip shape of a probe tip disposed on a probe of a scanning probe microscope, the scanning probe microscope being configured for moving the probe relative to a surface of a substrate in one or more directions, for mapping one or more surface structures on the surface. The method comprises steps of receiving, from a measurement data source, measurement data of a measurement of a surface topography of a surface, wherein the measurement is performed with the probe of the scanning probe microscope; determining, by a processor, based on the measurement data, the tip shape indication of the tip shape of the probe tip; and outputting, by the processor, based on the step of determining, the tip shape indication.

[0008] Advantageously, the step of determining the tip shape indication comprises providing, by the processor, the measurement data to a trained machine learning data processing model performing data pattern recognition on the measurement data for deriving the tip shape indication from the measurement data. By pre-training the machine learning data processing model, the step of determining the tip shape can be speeded up and carried out more accurately and even in real time. By utilizing data pattern recognition, the step of determining can be done based on available measurement data without the need of removing the probe from the microscope for inspection.

[0009] The term “tip shape indication” is intended to include any data or information that is suitable for providing information on the actual current shape of the probe tip. This may be a cross-sectional profile or only a part thereof (e.g. only the up-front wall profile of the probe tip in the scanning direction), or it may include any other two dimensional or three dimensional data on the shape of the probe tip. For example, the trained machine learning data processing model may have been trained on actual tip shape data, such as a 2D or 3D image of the probe tip obtained with another imaging device, to provide a tip shape indication based on or including accurate tip shape information. In another example, the trained machine learning data processing model may not have been trained on actual tip shape data, but e.g. on estimates of a tip shape obtained with an algorithm or based on different knowledge, to provide a tip shape indication that may be less accurate but good enough to keep an operator informed of the probable tip shape.

[0010] Furthermore, in various fields of application, and for example in certain industrial applications, the underlying surface topography of the surface of the substrate is always (or almost always) the same. For example, where the field of applications would be a manufacturing process of semiconductor structures, the same topography may have to be imaged over and over again by a scanning probe microscope in order to inspect each wafer produced. The topography may thus be more or less fixed, in which case the machine learning data processing model may be trained based on this one and the same topography. In such cases, it may not be of much added value to provide the surface topography as reference data to the trained machine learning data processing model. However, in other fields of applications, the surface topography may not always be the same. Therefore, in one aspect of the present disclosure, the data pattern recognition is based on reference data of the surface topography. By providing the machine learning data processing model with the reference data, the machine learning data processing model can use this reference data to recognize patterns in the measurement data that are indicative of structures that are suitable for deriving therefrom the actual tip shape, such as to provide the tip shape indication. This may be achieved independent of the orientation and / or composition of the surface topography in the measurement.

[0011] For example, if during training the model would always be trained on measurement data of a same surface topography comprising ribbon structures with ground truth data of different probe shapes images, the trained machine learning data processing model will, after the training has been completed, yield a most probable tip shape as tip shape indication when being provided with arbitrary measurement data of such a surface topography comprising ribbon structures. However, during training the machine learning data processing model may also have been trained with a much larger data set comprising measurement data of many kinds of surface topographies, where the associated surface topographies are provided as reference data with each measurement data as part of the training set, and where the associated ground truth data associated with each measurement data would comprise the actual tip shape based on e.g. SEM image(s) of the probe tip. In that case, the trained machine learning data processing model will, after the training has been completed, yield a most probable tip shape as tip shape indication, when being provided with arbitrary measurement data and associated reference data of the surface topography that was measured. This could then be any surface topography. Although it is more difficult to train, the advantages are significant because it allows to obtain tip shape indications based on arbitrary measurements and associated surface topography design data. Hence, this provides for an even larger field of applications. A further possibility wherein reference data may eventually not be required during application of the trained machine learning data processing model, would be where the model is trained such that from the image it will autonomously recognize edges in the topography from which the tip shape may be derived.

[0012] In other aspect of the present disclosure, the measurement data is an image data of a topography image obtained using the scanning probe microscope. Advantageously, image recognition performed on the measurement data to directly obtain the tip shape indication therefrom can be done very fast on using a well-trained machine learning data processing model such as described above. Furthermore, by performing image recognition on a whole image, rather than a line scan or other data, one has the advantage of using all data available from the scanning probe microscope measurement, and therefore providing high accuracy in determining the tip shape.

[0013] Advantageously, edge structures in the topography image are utilized in the image recognition step to quickly determine the corresponding tip shape. When the probe encounters an edge structure, while moving in a specific direction across the surface, the image will provide a convolution of the shape of the edge and the shape of the probe tip in the direction of scanning. Thus, by scanning from left to right over a step- up, the tip shape of the right side of the tip will in convolution with the shape of the step-up be visible in the image. If, however, the probe tip will move from right to left, the shape of the left side of the probe will be visible in convolution with the shape of the edge. Therefore, in accordance with some implementations, the image recognition step comprises identifying one or more edge structures in the topography image, and relating the identified edge structures to a tip shape profile of the tip shape in a scanning direction of the probe tip at the identified edge structures relative to the surface of the substrate.

[0014] In another aspect of the present disclosure, the reference tip shape is analyzed such as to determine tip shape damage. Thus, the method may comprise a step of analyzing the tip shape indication based on a reference tip shape such as to determine tip shape damage data. The tip shape indication may for example be compared with data of an undamaged probe tip in order to yield actual probe tip abrasion damage therefrom. For example, in some embodiments, a notification of the tip shape damage may be generated, e.g. to inform an operator of the tip shape indication, to advice replacement of the probe. Furthermore, the tip shape damage may serve as an input for a measurement correction algorithm configured to correct the obtained scanning microscope measurement for the tip shape damage. Potentially, the determined tip shape damage data may also serve as input data for a probe tip repair process for enabling restoration or repair of the probe tip.

[0015] In some embodiments, a tip shape indication of a probe tip determined by the machine learning data processing model is provided to a measurement correction algorithm configured to correct the measurement data for the determined tip shape indication to generate corrected measurement data of the surface topography. By performing the correction algorithm, the measurement data may be corrected in real time for the actual shape of the probe to provide image data of the underlying surface structures.

[0016] Advantageously, the measurement data is received by the processor simultaneously with performing the measurement of the surface topography of the surface by the scanning probe microscope, thereby enabling in-situ real-time tip shape estimation during the measurement. Alternatively, or in addition, the measurement data is obtained from a data set stored in a data repository. For example, in some embodiments, the measurement data is received by the processor simultaneously with performing the measurement of the surface topography of the surface by the scanning probe microscope, and wherein the method is performed during said measurement. In other or further embodiments, the measurement data is obtained, by the processor, from a data set of measurement data of the scanning probe microscope, wherein the data set is stored in a data repository. In yet another aspect of the present disclosure, there is provided a method of training a machine learning data processing model. The machine learning data processing model is trained for enabling it to be usable in a method described above, in particular for determining a tip shape indication of a tip shape of a probe tip of a probe of a scanning probe microscope. The scanning probe microscope is configured for moving the probe relative to a surface of a substrate in the one or more directions, and for establishing — continuously or intermittently — contact between the probe tip and the surface for mapping one or more surface structures on the surface. The training method comprises: receiving by the machine learning data processing model, a training data set from a measurement data source. The training data set comprises measurement data of a plurality of measurements, wherein each measurement of the plurality of measurements is performed with a different probe tip shape or a different probe of the scanning probe microscope for determining a surface topography of a surface of a substrate. The training method further includes receiving by the machine learning data processing model, a ground truth data set. The ground truth data set may be obtained from a same data source as the measurement data, or from a different data source. The ground truth data set comprises ground truth data for each measurement of said plurality of measurements, in the sense that for each measurement the ground truth data is indicative of an actual tip shape of a probe tip of said respective probe of said scanning probe microscope during said measurement (i.e. with which that measurement has been performed). Furthermore, the training method includes training the machine learning data processing model based on the measurement data of the plurality of measurements and based on the ground truth data associated with each measurement, such as to enable the machine learning data processing model to perform data pattern recognition on the measurement data, for deriving the tip shape indication therefrom. Thus the machine learning data processing model is trained, using a plurality of measurements done with a plurality of different probe tip shapes or by a plurality of different probes. By referring to ‘different probe tip shapes’ it is meant that these include different stages of wear or abrasion of a same probe tip, or different probe tip shapes of different probes, or both. The training is performed for enabling the machine learning data processing model to perform data pattern recognition on the measurement data, for deriving the tip shape indication. Typically, the training is done based on a training data set and ground truth data set.

[0017] In some cases, optionally, the training data set may further comprise reference data indicative of a surface topography of a surface associated with each measurement of the plurality of measurements. This may be directed at enabling the machine learning data processing model to automatically recognize, in an arbitrary surface topography, those structures from which it is possible to obtain tip shape information of the shape of the probe tip. In this connection, it is to be realized that not every structure on the surface of the substrate may be suitable for providing such tip shape data. Only those structures which are sufficiently pronounced such that imaging thereof leads to the convolution of the shape of the structure with the shape of the probe tip, are suitable candidates for obtaining therefrom the probe tip data. For example, these may include any of high aspect ratio structures, structures having walls that comprise a steep enough slope (i.e. steeper than the slope of the probe tip wall), or holes or trenches with such walls. By providing reference data indicative of the surface topography, e.g. a design topography of a semiconductor structure or a scanning electron microscope image of a surface, and using this reference data to train the machine learning data processing model, these suitable candidates may potentially be identified in arbitrary topographies after training. In some embodiments, the training is based on a training function that determines a difference between a tip shape indication determined by the machine learning data processing model and a tip shape of the respective probe tip comprised in the ground truth data received by the machine learning data processing model.

[0018] In some embodiments, for each measurement of said plurality of measurements, the tip shape comprised in the ground truth data is obtained based on an image of the probe tip obtained with an imaging device. Such an image provides accurate data on the tip shape from which the tip shape may accurately be determined. Using this data as ground truth data thereby enables the machine learning data processing model — once trained — to provide accurate tip shape indications. In some of these embodiments, the imaging device is at least one of a group comprising: a scanning electron microscope; a scanning ion microscope; or a scanning near-field optical microscope. Other nanometer scale imaging means my likewise be suitable for this task, and the above examples are merely meant to support the term ‘imaging device’ in connection herewith broadly.

[0019] In absence of more accurate ground truth data, or in those fields of application where less accuracy may be required or the requirements may be released a bit, the tip shape comprised in the ground truth data is a parametrized tip shape comprising at least two parameters. This is a less accurate manner of describing the shape of the tip. In these cases it may not be necessary to obtain a scanning electron microscope image to determine the actual tip shape. For example, in some of these embodiments, the parametrized tip shape may be determined by an erosion algorithm configured to estimate the tip shape abrasion damage. Such an erosion algorithm could have been determined or tuned, for example, based on deconvolution of the measurement data with a reference surface topography comprised in the reference data mentioned before. The tip shape comprised in the ground truth data may also be a simulated tip shape, e.g. obtained by simulating typical or expected wear of the probe tip based on various input parameters. The outcome provided by the trained machine learning data processing model in the method according to the first aspect in these latter examples and embodiments may be less accurate than by using data from real scanning electron microscope images, but may be suitable for the purpose intended and may be easier to obtain. Hence this does provide workable and convenient implementations of the claimed invention.

[0020] By providing a scanning probe microscope with the trained machine learning data processing model, real-time information on tip conditions, e.g. tip shape, of a probe tip disposed on a probe of the scanning probe microscope can be determined during operation of the scanning probe microscope, e.g. simultaneously with a measurement of a surface topography. Advantageously, an operator of the scanning probe microscope can be informed about the tip conditions and / or advised to replace the probe. Alternatively, or in addition, a processor of the scanning probe microscope can immediately correct measurement data comprising the measurement for the determined tip shape.

[0021] Brief description of the drawings

[0022] These and other features, aspects, and advantages of the apparatus, systems and methods of the present disclosure will become better understood from the following description, appended claims, and accompanying drawing wherein:

[0023] FIG 1A illustrates a method of determining a tip shape of a probe tip;

[0024] FIG IB illustrates a method of training a machine learning data processing model;

[0025] FIGs 2A-B illustrate a tip shape-based correction of surface topography measurements; and FIGs 3A-C illustrate a parametrized, simulated, and deconvolved tip shape.

[0026] Detailed description

[0027] Terminology used for describing particular embodiments is not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. The term "and / or" includes any and all combinations of one or more of the associated listed items. It will be understood that the terms "comprises" and / or "comprising" specify the presence of stated features but do not preclude the presence or addition of one or more other features. It will be further understood that when a particular step of a method is referred to as subsequent to another step, it can directly follow said other step or one or more intermediate steps may be carried out before carrying out the particular step, unless specified otherwise. Likewise it will be understood that when a connection between structures or components is described, this connection may be established directly or through intermediate structures or components unless specified otherwise.

[0028] The invention is described more fully hereinafter with reference to the accompanying drawings, in which embodiments of the invention are shown. In the drawings, the absolute and relative sizes of systems, components, layers, and regions may be exaggerated for clarity. Embodiments may be described with reference to schematic and / or crosssection illustrations of possibly idealized embodiments and intermediate structures of the invention. In the description and drawings, like numbers refer to like elements throughout. Relative terms as well as derivatives thereof should be construed to refer to the orientation as then described or as shown in the drawing under discussion. These relative terms are for convenience of description and do not require that the system be constructed or operated in a particular orientation unless stated otherwise.

[0029] FIG 1A schematically illustrates a method of determining D a tip shape indication 12i of a tip shape 12 of a probe tip 14 disposed on a probe 11 of a scanning probe microscope (SPM) 10. The scanning probe microscope

[0030] 10 includes an actuator system 25, e.g. of piezo type, for moving the probe

[0031] 11 relative to a surface 16 of a substrate 15 in various directions, for mapping surface structures 17 on the surface 16. The substrate 15 is supported on a substrate carrier 26, which may be part of the SPM 10 or may be a separate entity. Instead of or in addition to moving the probe 11 with the actuator system 25, also the substrate 15 may be moved by the carrier 26. During measurement of the substrate surface 16, the probe 11 is moved in a direction parallel to the surface 16 while the probe 11 is operated to continuously or intermittently establish contact between the probe tip 14 and the surface 16. The probe tip 14 thereby encounters the presence of structures 17. To very accurately determine the Z-position of the probe tip 14 upon contact with the surface 16, the SPM 10 includes an optical beam deflector arrangement comprising an optical beam source 18 for generating a beam 19, which impinges on the backside of the probe 11 near the tip 14. This side of the probe 11 includes a specular reflective surface from which the beam 19 is reflected towards a four quadrant optical sensor 23. The beam 19 forms a spot on the sensor 23 that partially overlaps each of the four quadrants of the sensor 23. If the angle of incidence changes as a result of Z-position differences of the probe tip 14 due to encountering a structure 17, the ratio of overlap of each quadrant changes with which the angle of incidence of the beam 19 may be accurately determined. The Z-position may be determined accurately by correcting the Z of the probe with the Z- actuator of the actuator arrangement 25, until the position of the spot on the sensor 23 is such that the ratio of overlap with each quadrant is in its starting position again. The SPM is operated using a controller or processor 24.

[0032] As may be understood, the probe tip 14 of probe 11, due to its intermittent contact with the surface 16 will be subject to abrasion causing the probe tip to wear. This results in the tip shape 12 of the probe tip 14 to change while it is in use. For example, a common probe tip 14 originally has a cone shape, but in use the sides of the cone may wear off resulting in the side walls to become curved or slightly concave. This is illustrated in the enlargement of the probe tip 14 in the dotted circle of figure 1A.

[0033] The method of determining D the tip shape 12 of a probe tip 14 in order to provide tip shape data 12i (also referred to as tip shape indication 12i) comprises steps of receiving Rm, determining D, and outputting 0. The method may be performed by the controller or processor 24 of the SPM, but may alternatively or additionally also be performed on a different computing system including a controller and in communicative connection with a measurement data source 21 or directly with the SPM 10. In one embodiment, the step of receiving Rm comprises receiving, from the measurement data source 20, measurement data 20 of a measurement 13 of the surface topography S of a surface 16. The measurement 13 is performed with the probe 11 of the scanning probe microscope 10. The step of determining D the tip shape indication 12i of the tip shape 12 is performed by the processor 24 based on the measurement data 20. The measurement data 20 may be obtained directly from the SPM 10, and the whole method may be performed locally on the SPM 10 in some implementations. For example, the measurement data 20 may be analyzed in real-time during measurement in order to also in real-time be able to take proper action. Such proper action for example may be the generation of a notification — also in real time — to an operator that the probe tip 14 is too badly eroded and must be replaced. This is greatly advantageous and is something that cannot be done in a conventional SPM system, because conventionally this always requires the probe tip 14 to be subject to inspection for which the probe tip 14 must be taken off the SPM 10. Due to the application of a trained machine learning data processing model, the measurement data (e.g. image data) from the SPM may be analyzed in real-time to detect probe tip wear, e.g. abrasion, during use in real-time (i.e. to detect immediately when the shape 12 of the tip 14 is damaged too much). Another direct advantage is that this also enables — at all times during the measurement and again in real-time — to correct the measurement for the current actual tip shape 12.

[0034] Preferably, the step of determining D the tip shape indication 12i comprises providing, by the processor, the measurement data 20 to the trained machine learning data processing model 30 performing data pattern recognition on the measurement data 20 for deriving the tip shape indication 12i therefrom. This may be provided by a convolutional neural network (CNN) 30, or by another suitable type of machine learning data processing model. Preferably, the measurement data 20 is received by the processor simultaneously with performing the measurement of the surface topography S of the surface by the scanning probe microscope 10, and wherein the method is performed during said measurement. Alternatively, or in addition, the measurement data 20 is obtained, by the processor, from a data set of measurement data of the scanning probe microscope 10. This data set may be stored in the measurement data source 21 shown as the data repository in figure 1A. In one embodiment, the data set is stored in a local and / or remote data repository. In other or further embodiment, the scanning probe microscope 10 comprises an atomic force microscope. In one embodiment, the measurement data source 21 is a memory of the atomic force microscope 10.

[0035] In other or further embodiments, the surface topography S comprises a topography of a wafer, including semiconductor structures such as chips, memories and / or other kind of semiconductor structures. The present concept is not limited to this field of application, and could well be applied to different topographies. Alternatively, or in addition, the surface topography S comprises a topography of a nanostructure and / or a nanomaterial.

[0036] In some embodiments, the machine learning data processing model comprises a convolutional neural network and / or other kind of deep learning model. In one embodiment, the machine learning data processing model 30 is trained by the means of supervised or semi- supervised machine learning technique. Alternatively, or in addition, is the machine learning data processing model 30 trained using unsupervised and / or reinforcement learning technique. In some, other or further implementations, the data pattern recognition comprises identification of patterns or trends within the measurement data 20. In one embodiment, the data pattern recognition is additionally based on reference data S’ of the surface topography S. In other or further embodiment, the reference data S’ of the surface topography S comprises a blueprint of the surface topography S, e.g. a design file and / or simulation of the actual surface topography comprised in the measurement by the scanning probe microscope. This may be used by the machine learning data processing model 30 to learn identifying edges or structures that enable to derive the tip shape 12 therefrom to provide the indication 12i. Alternatively, or in addition, the reference data S’ comprises a simulation of the structure surface topography S.

[0037] In yet other or further embodiments, the measurement data 20 is an image data of a topography image obtained using the scanning probe microscope 10, and wherein the data pattern recognition performed by the trained machine learning data processing model 30 comprises an image recognition step. In one embodiment, is the image recognition step based on feature extraction and / or classification. Alternatively, or in addition is the image recognition step based on segmentation and / or object detection. In some embodiments, the image recognition step comprises identifying one or more edge structures in the topography image, e.g. with the help of the surface topography reference data S, and relating the identified edge structures to a tip shape profile of the tip shape 12 in a scanning direction of the probe tip relative to the surface of the substrate. The machine learning data processing model may be enabled, by properly training it with surface topography reference data S, to automatically recognize and focus, in an arbitrary surface topography, on those structures 17 from which it is possible to obtain tip shape information of the shape 12 of the probe tip 14. In this connection, it is to be realized that not every structure 17 on the surface of the substrate may be suitable for providing such tip shape data 12i. Only those structures 17 which are sufficiently pronounced such that imaging thereof leads to the convolution of the shape of the structure 17 with the shape 12 of the probe tip 14, are suitable candidates for obtaining therefrom the probe tip data 12i. For example, these may include any of high aspect ratio structures, structures having walls that comprise a steep enough slope (i.e. steeper than the slope of the probe tip wall), or holes or trenches with such walls. By providing reference data indicative of the surface topography S, e.g. a design topography of a semiconductor structure or a scanning electron microscope image of a surface, and using this reference data to train the machine learning data processing model 30, these suitable candidates may potentially be identified in arbitrary topographies after training. Alternatively, or in addition, the recognition steps comprises identifying one or more textures and / or colors.

[0038] In other or further embodiments, the method of determining a tip shape indication 12i further comprises a step of analyzing the tip shape indication 12i based on a reference tip shape such as to determine tip shape damage data. In one embodiment, the method further comprises, based on the tip shape damage data, generating a notification comprising at least one element of a group comprising: data informing an operator of the tip shape indication; an indication advising replacement of the probe, an indication of the tip shape damage or tip shape damage data; input data for a probe tip repair process for enabling restoration or repair of the probe tip based on the tip shape damage data; or input data for a measurement correction algorithm.

[0039] FIGs 2A-B illustrate a tip shape-based correction of surface topography measurements. In some embodiments, the tip shape indication 12i is provided to a measurement correction algorithm 50 configured to correct the measurement data 20 for the determined tip shape indication 12i to generate corrected measurement data 13c of the surface topography S. For example, the measurement correction algorithm 50 comprises deconvolution of the measurement 13 with the tip shape indication 12d determined by the machine learning data processing model 30.

[0040] In some embodiments, the machine learning data processing model 30 is trained according to a method illustrated in FIG IB. The purpose of the method of training the machine learning data processing model is to enable the machine learning data processing model 30 to be usable for determining D a tip shape indication 12i of a tip shape 12 of a probe tip disposed on a probe 11 of a scanning probe microscope 10 moving the probe relative to a surface of a substrate in one or more directions, thereby mapping surface topography S of one or more surface structures on the surface.

[0041] In some embodiments, the method of training the machine learning data processing model comprises steps of receiving Rm, Rg, and training T. In one embodiment, the step of receiving Rm comprises receiving, from a measurement data source 21, by the machine learning data processing model 30, a training data set, the training data set comprising measurement data 20 of a plurality of measurements 13a, 13b, 13c. In another or further embodiment, each measurement of the plurality of measurements 13a, 13b, 13c is performed with a respective probe 11a, lib, 11c of a scanning probe microscope 10 for determining surface topography S of a surface of a substrate. In one embodiment, the step of receiving Rg comprises receiving, from a ground truth data source 40, by the machine learning data processing model 30, a ground truth data set, the ground truth data set comprising ground truth data 41 for each measurement of said plurality of measurements 13a, 13b, 13c. In another or further embodiment, for each measurement is the ground truth data 41 indicative of an actual tip shape 12a’, 12b’, 12c’ of a probe tip of said respective probe 11a’, lib’, 11c’ of said scanning probe microscope 10 during said measurement. Preferably, the ground truth data 41 of the tip shape 12 comprises scanning electron microscope images of the tip shape 12. In other or further embodiments, the step of training T comprises training the machine learning data processing model 30 based on the measurement data 20 of a plurality of measurements 13a, 13b, 13c and based on the ground truth data 41 associated with each measurement of said plurality of measurements 13a, 13b, 13c, such as to enable the machine learning data processing model 30 to perform data pattern recognition on the measurement data 20, for deriving the tip shape indication 12i therefrom.

[0042] In further embodiments, the training data set further comprises reference data S’ indicative of a surface topography S of a surface associated with each measurement of the plurality of measurements 13a, 13b, 13c.

[0043] In yet further embodiments, the training T is based on a training function that determines a difference between a tip shape indication 12i determined by the machine learning data processing model 30 and a tip shape 12 of the respective probe tip comprised in the ground truth data 41 received by the machine learning data processing model 30. For example, the difference is determined by a cost / award function. In some embodiments, the plurality of probe tips 11a, lib, 11c has different tip shapes, thereby providing variation of data to the machine learning data processing model 30, e.g. a tip shape 12 of an unused new probe and a tip shape 12 of a heavily used deteriorated probe, or anything there in-between. In other or further embodiments, the plurality of probe tips 11a, lib, 11c comprises probe tips of the same kind e.g. shape, intended use, design, material, dimensions and a different degree of deterioration. Alternatively, or in addition, the plurality of probe tips 11a, lib, 11c comprises probe tips of different kinds and the same degree of deterioration.

[0044] In other or further embodiments, the step of determining D of the tip shape 12 is based on determining a relation, pattern, and / or transformation function between the measurement 13 of the surface topography S comprised in the training data set and a tip shape 12 of a probe tip of the plurality of probe tips 11a, lib, 11c.

[0045] In yet further embodiments, the ground truth data set comprises parameter values of a known tip shape, scanning electron microscope images of a tip shape, history of a tip usage, and / or a tip shape based on a measurement 13 of a reference surface topography deconvolved with a reference data S’ indicative of the reference surface topography S.

[0046] In some embodiments, as shown in FIG 3A, the tip shape indication 12i determined by the machine learning data processing model 30 and / or the tip shape 12 comprised in the ground truth data 41 is a parametrized tip shape comprising at least two parameters A,R. In one embodiment, A specifies a distance to a point along a longitudinal axis of the tip shape 12, e.g. from the tip of the tip shape 12, and R specifies distance from the point along the longitudinal axis of the tip shape 12 to an edge of the tip shape 12, e.g. radius R1 of the tip shape 12 at half of its height Al.

[0047] In other or further embodiments, as shown in FIG 3B, the tip shape 12 comprised in the ground truth data 41 is a modelled tip shape 12s. In one embodiment, a simulation is configured to simulate scanning of the reference surface topography S by a plurality of modelled different tip shapes to provide a plurality of simulated measurements of the modelled reference surface topography S. In another or further embodiment, the plurality of modelled different tip shapes may be used as ground truth data. This ground truth data is associated with the plurality of simulated measurements of the modelled reference surface topography S and may be used to train the machine learning data processing model. This training is directed at learning the machine learning data processing model to predict, based on the simulated measurements, the correct modelled tip shape (i.e. associated with said simulated measurement) as tip shape indication..

[0048] In yet further embodiments, as shown in FIG 30, the tip shape 12 comprised in the ground truth data 41 is an eroded tip shape 12e determined by an erosion algorithm 60 configured to estimate the tip shape 12 based on deconvolution of a measurement 13 with a respective reference surface topography S comprised in the reference data S’.

[0049] In some embodiments, the method of determining a tip shape of a probe tip is provided to a scanning probe microscope 10. In one embodiment, the scanning probe microscope comprises a processor communicatively connected to a measurement data source 21 and configured to obtain measurement data 20 comprising a measurement 13 of one or more surface structures on the surface by moving the probe tip relative to a surface of a substrate in one or more directions, wherein the probe tip is disposed on a probe 11 of the scanning probe microscope 10. In another or further embodiment, the processor is configured to provide a tip shape indication 12i of the tip shape 12 of the probe tip by receiving Rm, from the measurement data source 21, measurement data 20 of the measurement 13 of the surface topography S of the surface; and determining D, by the processor, based on the measurement data 20, the tip shape indication 12i of the tip shape 12 of the probe tip. In other or further embodiments, the measurement data source 21 further comprises a trained machine learning data processing model 30. In one embodiment, the step of determining D the tip shape indication 12i comprises providing, by the processor, the measurement data 20 to the trained machine learning data processing model 30. In another or further embodiment, the trained machine learning data processing model 30 performs data pattern recognition on the measurement data 20 for deriving the tip shape indication 12i from the measurement data 20.

[0050] In yet further embodiments, the processor of the scanning probe microscope 10 is configured to output 0 a notification. In one embodiment, the notification comprises at least one element of a group comprising: data informing an operator of the tip shape indication; an indication advising replacement of the probe, an indication of the tip shape damage or tip shape damage data.

[0051] In yet further embodiments, the processor of the scanning probe microscope 10 is configured to input data for a probe tip repair process for enabling restoration or repair of the probe tip based on the tip shape damage data; or input data for a measurement correction algorithm 50. In one embodiment, the measurement correction algorithm 50 is configured to correct the measurement data 20 for the determined tip shape indication 12i to generate corrected measurement data 13c of the surface topography S. For example, the measurement data 20 is corrected using an inverse modeling algorithm, deconvolution, and / or other kind of image correction algorithms.

[0052] For the purpose of clarity and a concise description, features are described herein as part of the same or separate embodiments, however, it will be appreciated that the scope of the invention may include embodiments having combinations of all or some of the features described. In interpreting the appended claims, it should be understood that the word "comprising" does not exclude the presence of other elements or acts than those listed in a given claim; the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements; any reference signs in the claims do not limit their scope; several "means" may be represented by the same or different item(s) or implemented structure or function; any of the disclosed devices or portions thereof may be combined together or separated into further portions unless specifically stated otherwise. Where one claim refers to another claim, this may indicate synergetic advantage achieved by the combination of their respective features. But the mere fact that certain measures are recited in mutually different claims does not indicate that a combination of these measures cannot also be used to advantage. The present embodiments may thus include all working combinations of the claims wherein each claim can in principle refer to any preceding claim unless clearly excluded by context.

Claims

CLAIMS1. A method of determining a tip shape indication of a tip shape of a probe tip disposed on a probe of a scanning probe microscope, the scanning probe microscope being configured for moving the probe relative to a surface of a substrate in one or more directions, and for establishing contact between the probe tip and the surface for mapping one or more surface structures on the surface, the method comprising steps of receiving, from a measurement data source, measurement data of a measurement of a surface topography of a surface, wherein the measurement is performed with the probe of the scanning probe microscope; determining, by a processor, based on the measurement data, the tip shape indication of the tip shape of the probe tip; outputting, by the processor, based on the step of determining, the tip shape indication; wherein the step of determining the tip shape indication comprises providing, by the processor, the measurement data to a trained machine learning data processing model, wherein the trained machine learning data processing model performs data pattern recognition on the measurement data for deriving the tip shape indication from the measurement data.

2. Method according to claim 1, wherein the measurement data is an image data of a topography image obtained using the scanning probe microscope, and wherein the data pattern recognition performed by the trained machine learning data processing model comprises an image recognition step.

3. Method according to claim 2, wherein the image recognition step comprises identifying one or more edge structures in the topography image, and relating the identified edge structures to a tip shape profile of the tip shape in a scanning direction of the probe tip at the identified edge structures relative to the surface of the substrate.

4. Method according to any one or more of the preceding claims, further comprising obtaining from the measurement data source, reference data of the surface topography, wherein the step of data pattern recognition is further based on the reference data.

5. Method according to any one or more of the preceding claims, wherein the method further comprises a step of analyzing the tip shape indication based on a reference tip shape such as to determine tip shape damage data.

6. Method according to claim 5, further comprising, based on the tip shape damage data, generating a notification, wherein the notification comprises at least one element of a group comprising: data informing an operator of the tip shape indication; an indication advising replacement of the probe, an indication of the tip shape damage or tip shape damage data; input data for a probe tip repair process for enabling restoration or repair of the probe tip based on the tip shape damage data; or input data for a measurement correction algorithm.

7. Method according to any one or more of the preceding claims, wherein a tip shape indication of a probe tip determined by the machine learning data processing model is provided to a measurement correction algorithm configured to correct the measurement data for the determined tip shape indication to generate corrected measurement data of the surface topography.

8. Method according to any one or more of the preceding claims, wherein at least one of: the measurement data is received by the processor simultaneously with performing the measurement of the surface topography of the surface by the scanning probe microscope, and wherein the method is performed during said measurement; or the measurement data is obtained, by the processor, from a data set of measurement data of the scanning probe microscope, wherein the measurement data source is a data repository.

9. Method of training a machine learning data processing model, for enabling the machine learning data processing model to be usable in a method according to any one or more of the preceding claims, for determining a tip shape indication of a tip shape of a probe tip disposed on a probe of a scanning probe microscope, the scanning probe microscope being configured for moving the probe relative to a surface of a substrate in the one or more directions, and for establishing contact between the probe tip and the surface for mapping one or more surface structures on the surface, wherein the method comprises: receiving, from a measurement data source, by the machine learning data processing model, a training data set, the training data set comprising measurement data of a plurality of measurements, wherein each measurement of the plurality of measurements is performed with a respective probe of a scanning probe microscope for determining a surface topography of a surface of a substrate, receiving, from a ground truth data source, by the machine learning data processing model, a ground truth data set, the ground truth data set comprising ground truth data for each measurement of said plurality of measurements, wherein for each measurement the ground truth data is indicative of an actual tip shape of a probe tipof said respective probe of said scanning probe microscope during said measurement; and training the machine learning data processing model based on the measurement data of the plurality of measurements and based on the ground truth data associated with each measurement of said plurality of measurements, such as to enable the machine learning data processing model to perform data pattern recognition on the measurement data, for deriving the tip shape indication therefrom.

10. Method according to claim 9, wherein the training data set further comprises reference data indicative of the surface topography of the surface associated with each measurement of the plurality of measurements.

11. Method according to any of the two preceding claims, wherein the training is based on a training function that determines a difference between a tip shape indication determined by the machine learning data processing model and a tip shape of the respective probe tip comprised in the ground truth data received by the machine learning data processing model.

12. Method according to any one or more of claims 9-11, wherein for each measurement of said plurality of measurements, the tip shape comprised in the ground truth data is obtained based on an image of the probe tip obtained with an imaging device.

13. Method according to claim 12, wherein the imaging device is at least one of a group comprising: a scanning electron microscope; a scanning ion microscope; or a scanning near-field optical microscope.

14. Method according to any one or more of claims 9-11, wherein the tip shape comprised in the ground truth data is a parametrized tip shape comprising at least two parameters.

15. Method according to claim 14 or any other claim dependent on claim 12, wherein the parametrized tip shape is determined by an erosion algorithm configured to estimate a tip shape abrasion, for example based on deconvolution of the measurement data with a reference surface topography comprised in the reference data.

16. Scanning probe microscope comprising a probe including a probe tip, wherein the scanning probe microscope includes or cooperates with a substrate carrier for supporting a substrate, the scanning probe microscope being configured for moving the probe relative to a surface of the substrate in one or more directions, and for establishing contact between the probe tip and the surface for mapping one or more surface structures on the surface, wherein the scanning probe microscope further comprises a processor communicatively connected to a measurement data source, wherein the processor is configured to provide a tip shape indication of a tip shape of the probe tip, and wherein for providing the tip shape indication the processor is configured for: receiving, from the measurement data source, measurement data of a measurement of a surface topography of the surface; and determining, by the processor, based on the measurement data, the tip shape indication of the tip shape of the probe tip; wherein the measurement data source further comprises a trained machine learning data processing model, and wherein for determining the tip shape indication the processor is configured for providing the measurement data to the trained machine learning data processing model, for performing by the trained machine learning data processing model data pattern recognition on the measurement data for deriving the tip shape indication from the measurement data.

17. Scanning probe microscope according to claim 16, wherein the processor is further configured for analyzing the tip shape indication based on a reference tip shape such as to determine tip shape damage data.

18. Scanning probe microscope according to claim 17, wherein the processor is configured for, based on the tip shape damage data, output a notification, wherein the notification comprises at least one element of a group comprising: data informing an operator of the tip shape indication; an indication advising replacement of the probe, an indication of the tip shape damage or tip shape damage data.

19. Scanning probe microscope according to any of the claims 16-18, wherein the processor is configured to perform a measurement correction algorithm based on the tip shape indication, wherein the measurement correction algorithm is configured to correct the measurement data for the determined tip shape indication to generate corrected measurement data of the surface topography.

Images