Method for inspecting a specimen using a charged particle beam device - Patents.com

Abstract

To provide a method for determining phases, i.e. chemical compositions, that are present in a sample, in a relatively short time.SOLUTION: A method of examining a sample using a charged particle beam apparatus includes the steps of: detecting, using a first detector, emissions of a first type from the sample in response to a charged particle beam illuminating the sample; and further acquiring spectral information on emissions of a second type from the sample. The step of acquiring spectral information includes the steps of: providing a spectral information prediction algorithm; and using the algorithm for predicting the spectral information based on detected emissions of the first type as an input parameter of the algorithm. Thus, EDS data can be gathered using only a BSE detector.SELECTED DRAWING: Figure 4c

Description

[Technical field]

[0001] overview The present invention relates to a method of inspecting a sample using a charged particle beam device, comprising the steps of providing a charged particle beam as well as a sample, illuminating the sample with the charged particle beam, and detecting, using a first detector, a first type of radiation from the sample in response to the charged particle beam illuminating the sample. [Background technology]

[0002] Charged particle microscopy is a well-known and increasingly important technique for imaging small objects, particularly in the form of electron microscopy. Historically, the basic type of electron microscope has evolved into several well-known instrument types, such as the Transmission Electron Microscope (TEM), the Scanning Electron Microscope (SEM), the Scanning Transmission Electron Microscope (STEM), and various subtypes, such as the so-called "dual beam" instruments (e.g., FIB-SEM) that employ a further "machining" Focused Ion Beam (FIB), enabling support, for example, of ion beam milling or Ion-Beam-Induced Deposition (IBID). Those skilled in the art will be familiar with the different types of charged particle microscopy.

[0003] Irradiation of a sample with a scanning electron beam stimulates the emission of "auxiliary" radiation from the sample in the form of secondary electrons, backscattered electrons, X-rays, and cathodoluminescence (infrared, visible, and / or ultraviolet photons). One or more components of this emitted radiation can be detected and used for sample analysis.

[0004] Typically in an SEM, backscattered electrons are detected by a solid-state detector, and each backscattered electron is amplified to create many electron-hole pairs in the semiconductor detector. The backscattered electron detector signal is used to form an image as the beam is scanned, and the brightness of each image point is determined by the number of backscattered electrons detected at the corresponding point on the sample as the primary beam moves across the sample. The image only provides information about the topology of the sample being inspected.

[0005] In a process called "energy dispersive X-ray spectroscopy" (also referred to as "EDS" or "EDX"), the energy of the X-rays coming from the sample in response to the electron beam is measured and plotted in a histogram to form a material-specific spectrum. The measured spectrum can be compared to known spectra of various elements to determine which elements and minerals are present in the sample.

[0006] One drawback of EDS is that it takes a significant amount of time to accumulate an X-ray spectrum of a sample. Typically, a grid with discrete analysis points is used. The electron beam dwells at each analysis point while the EDS detector records the X-rays. Once a sufficient number of X-ray counts have been recorded, the beam moves to the next analysis point. The signal from the EDS detector is fed to a signal processing unit which constructs an X-ray spectrum curve for each analysis point, which may then be queried against an extensive library of known mineral phases to select the one that best matches that analysis point. This known method is relatively time consuming to determine the phases present in a sample, i.e. the chemical composition.

[0007] In general, acquiring spectral information of a sample is relatively time consuming. Summary of the Invention

[0008] With the above in mind, it is an object of the present invention to provide an improved method of inspecting a sample using a charged particle beam device, where spectral information is used to inspect the sample. In particular, it is an object of the present invention to provide a method and apparatus for obtaining information about a sample more quickly and / or more accurately.

[0009] To this end, the present invention provides a method for inspecting a specimen using a charged particle beam device, as defined in claim 1. The method defined herein comprises the steps of providing a charged particle beam and using said beam to illuminate the specimen. The method further comprises the step of detecting, using a first detector, a first type of radiation from the specimen in response to the charged particle beam illuminating the specimen. Said first type of radiation relates to non-spectral information. In an embodiment, said first detector is configured to detect charged particles, in particular electrons such as backscattered electrons. Thus, the first detector may be a BSE detector.

[0010] The methods defined herein further include obtaining spectral information relating to a second type of radiation from the sample in response to the charged particle beam illuminating the sample, the second type of radiation being different from the first type of radiation. The spectral information may, in an embodiment, relate to X-rays emanating from the sample.

[0011] According to the invention, obtaining the spectral information comprises providing a spectral information prediction algorithm and using the algorithm to predict the spectral information based on the detected radiation of the first type as an input parameter of the algorithm.

[0012] Thus, as defined herein, the obtained radiation of a first type, e.g., backscattered electrons, is used as an input parameter of a spectral information prediction algorithm to obtain spectral information related to a second type of radiation. Thus, in effect, the spectral information can be obtained without the need for a second detector to detect the second type of radiation, and in principle, only one detector is required to obtain the information regarding the first type of radiation and the spectral information.

[0013] The object of the invention is thus achieved. Advantageous embodiments are described below.

[0014] It should be noted that multiple detectors of a first type may be used to obtain the first type of radiation of interest, and embodiments may include bright field, dark field, or high angle annular dark field STEM detectors in a TEM.

[0015] In an embodiment of the method, a charged particle beam is scanned over the sample. The radiation detected by the first detector and therefore the spectral information determined in the spectral information prediction algorithm can be associated with a specific scanning beam position, i.e. with a specific position on the sample. This means that the spectral information can be determined for the corresponding positions on the sample as well. The spectral information obtained for the different positions can be compared with each other and one or more specific phases can be assigned to these different positions. The already assigned phases can be used to establish estimated phases for other positions, for example based on proximity to these already assigned phases, based on partially obtained spectral profiles obtained for other positions, or based on other parameters.

[0016] The first detector may be a backscattered electron detector. The backscattering of electrons depends on the atomic number of the elements at the surface and the geometrical relationship between the surface, the primary beam and the detector. Hence, the backscattered electron image shows contour information, i.e., the boundaries between regions of different composition, and cross-sectional information. Obtaining a backscattered electron image is much faster than obtaining a sufficient number of X-rays to collect a complete spectrum at each point, since it requires only collecting a sufficient number of electrons at each point to produce a reasonable contrast between points with different properties. Also, the probability of an electron being backscattered is higher than the probability of an electron producing a characteristic X-ray emission of a particular frequency. Obtaining sufficient backscattered electron image data at a single dwell point typically requires less than one microsecond, whereas obtaining enough X-rays to obtain an analyzable spectrum at a single dwell point typically requires more than one millisecond. Thus, by using a relatively fast backscattered electron image as an input for obtaining the spectral information of the sample, it is possible to obtain accurate and fast spectral information of the sample, depending on the accuracy and speed of the spectral information prediction algorithm.

[0017] In an embodiment, the algorithm is configured to identify inter-spatial data relationships for the radiation of the first type, which can be used to obtain spectral information more quickly and accurately using a spectral information prediction algorithm. In an embodiment, the algorithm is then configured to use the identified inter-spatial data relationships to predict the spectral information at a given sample location.

[0018] In an embodiment, the algorithm is configured to use the first type of radiation to identify regions of the sample having similar properties. As described above, backscattered electron images can show contour information, i.e., boundaries between regions of different composition, and cross-sectional information. This information can be used by the algorithm to identify regions having similar properties.

[0019] In an embodiment, the algorithm includes at least one neural network. The neural network can be trained using sample data that includes information about a first type of radiation and spectral information related to a second type of radiation, where these first and second types of radiation are physically obtained during data acquisition. This allows the neural network to be trained to predict the spectral information based on the first type of radiation as an input parameter.

[0020] The spectral information may include elemental information of the sample. The sample may in principle be any sample, although good results have been obtained with material samples such as geological and semiconductor samples. However, other samples are also contemplated.

[0021] The method may include the further step of accepting or rejecting the sample based on the acquired spectral information. This is particularly advantageous in a process control setting where samples may be obtained from a manufacturing process, and the methods defined herein may be used to collect (predict) spectral information about the sample, and the results may be used to accept or reject the sample, and thus the manufacturing process or part of it. For example, if the sample contains too much of a particular material, it may be concluded that a particular manufacturing process did not produce satisfactory results, and then the part of the product may be further investigated.

[0022] In an embodiment, the method includes a further step of outputting data related to the predicted spectral information, the output including at least one of an image of the sample, elemental information of the sample, and / or a rating for the sample.

[0023] According to an aspect, the present disclosure provides a charged particle beam device for inspecting a specimen using the above method, comprising: a charged particle beam source for emitting a charged particle beam; a sample holder for holding a sample; an illuminator for directing a charged particle beam emitted from a charged particle beam source onto a sample; a first detector for detecting a first type of radiation arising from the sample in response to illumination of the sample by charged particles emitted from the charged particle source; a control unit connected to the first detector and configured to control at least a part of the operation of the charged particle beam device.

[0024] As defined herein, a charged particle beam device, such as a charged particle microscope, is configured to execute the method according to one or more of the preceding claims. This means that the control unit may be configured to use the first type of radiation in a spectral information prediction algorithm for determining the spectral information of the sample. Alternatively (or additionally), the information of the first type of radiation may be transferred to an external or cloud-based processing unit, which is able to calculate the spectral information. The results may then be transferred back to the relevant user, for example on a screen located near the charged particle beam device.

[0025] When using a cloud-based processing unit, it becomes possible to connect multiple charged particle beam devices to a single spectral information prediction algorithm. This allows the determined spectral information to remain accurate and up-to-date with the latest insights and updates of the spectral information prediction algorithm. Updates of the algorithm can be made, after which charged particle beam devices connected to an external processing unit, such as a cloud-based processing unit, can benefit from this updated algorithm. Data provided by the charged particle beam devices, which may include spectral information, may be used for training purposes, for example, in the case where the spectral information prediction algorithm is a neural network. For example, it is conceivable that whenever a new type of sample is being studied or examined, the spectral information for these charged particle beam devices is recorded for calibration purposes.

[0026] In an embodiment, the charged particle beam device includes at least one detector including the first detector, and all of the at least one detector are non-spectral detectors. Thus, the charged particle beam device can be free from any spectral detector, and the charged particle beam device can rely only on a spectral information prediction algorithm to obtain spectral information about the sample.

[0027] According to an aspect, there is provided a method for developing an algorithm, in particular a method for training a neural network, the algorithm being adapted for use in the above method, comprising the steps of: - providing sample data relating to a first type of radiation from the sample; - providing corresponding spectral information regarding a second type of radiation from the sample; - using the provided data to develop the algorithm, in particular to train the neural network, where a first type of radiation is used as input and spectral information is used as desired output.

[0028] The developed methods may also be used to update spectral information prediction algorithms in a setup where multiple charged particle beam devices are connected to one or more processing units, such as a cloud-based processing unit. [Brief description of the drawings]

[0029] The invention will now be elucidated in detail on the basis of exemplary embodiments and the accompanying schematic drawings. [Figure 1] 1 shows a longitudinal cross-sectional view of a charged particle beam device according to a first embodiment of the invention; [Diagram 2] 2 shows a longitudinal cross-sectional view of a charged particle beam device according to a second embodiment of the invention; [Diagram 3] 1 shows a schematic overview of a neural network architecture that can be used as a spectral information prediction algorithm. [Figure 4a] Respectively, a) an electron microscope image, b) an EDS image obtained with an EDS detector, and c) an EDS image obtained using the spectral information prediction algorithm defined herein. [Figure 4b] Respectively, a) an electron microscope image, b) an EDS image obtained with an EDS detector, and c) an EDS image obtained using the spectral information prediction algorithm defined herein. [Figure 4c] Respectively, a) an electron microscope image, b) an EDS image obtained with an EDS detector, and c) an EDS image obtained using the spectral information prediction algorithm defined herein. [Diagram 5] 1 shows an example of a neural network architecture that can be used in the spectral information prediction algorithm defined herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0030] FIG. 1 (not to scale) is a highly schematic illustration of an embodiment of a charged particle microscope M according to an embodiment of the present invention. More specifically, the illustration shows an embodiment of a transmission microscope M, in this case a TEM / STEM (although in the context of the present invention it could just as well be, for example, a SEM (see FIG. 2) or an ion-based microscope). In FIG. 1, within a vacuum enclosure 2, an electron source 4 produces an electron beam B, which propagates along an electron optical axis B', passes an electron optical illuminator 6 and serves to direct / focus the electrons onto a selected portion of a sample S (which may for example be (locally) thinned / flattened). A deflector 8 is also illustrated, which can be used to (among other things) effect a scanning movement of the beam B.

[0031] The sample S is held in a sample holder H, which can be positioned in multiple degrees of freedom by a positioning device / stage A that moves a cradle in which the holder H is (removably) mounted; for example, the sample holder H can be equipped with fingers that can be moved in the XY plane (among other things) (see Cartesian coordinate system in the depiction, usually movements parallel to Z and tilts about X / Y are also possible). Such movements allow different parts of the sample S to be illuminated / imaged / inspected by an electron beam B moving (in Z direction) along an axis B' (and / or a scanning movement can be performed as an alternative to electron beam scanning). If desired, an optional cooling device (not depicted) can be brought in intimate thermal contact with the sample holder H, thereby maintaining the sample holder H (and the sample S thereon) at, for example, cryogenic temperatures.

[0032] The electron beam B interacts with the sample S in such a way as to cause it to emit various types of "stimulated" radiation, including (for example) secondary electrons, backscattered electrons, X-rays, and light radiation (cathodoluminescence). If desired, one or more of these types of radiation can be detected with the aid of an analysis device 22, which can be, for example, a combination of a scintillator / photomultiplier tube or an EDX or EDS (energy dispersive X-ray spectroscopy) module, and in such a case an image can be constructed using essentially the same principles as in an SEM. However, as defined herein, such an EDX or EDS module is in principle not actually required to obtain this spectral information, as will be explained later. It is also possible to consider electrons that traverse (pass) the sample S, exit / emit from the sample, and continue to propagate (effectively, generally with some degree of deflection / scattering) along the axis B'. Such a transmitted electron flux enters an imaging system (projection lens) 24, which typically comprises various electrostatic / magnetic lenses, deflectors, collectors (stigmators, etc.), etc. In standard (non-scattering) TEM mode, this imaging system 24 is capable of focusing an electron beam onto a fluorescent screen 26 which can be retracted / extracted (as indicated diagrammatically by arrow 26') to bring it out of the path of axis B', if desired. On the screen 26, an image (or diffractogram) of (a part of) the sample S is formed by the imaging system 24, which can be seen through a viewing port 28 located in a suitable part of the wall of the housing 2. The retraction mechanism of the screen 26 is, for example, mechanical and / or electrical in nature and is not shown here.

[0033] As an alternative to viewing an image on screen 26, one can instead take advantage of the fact that the focusing depth of the electron flux leaving imaging system 24 is typically quite deep (e.g., on the order of one meter). As a result, various other types of analytical devices (such as those listed below) can be used downstream of screen 26. -TEM camera 30. At the position of camera 30 the electron flux can form a still image (or diffractogram) which can be processed by controller / processor 20 and displayed on a display device 14, such as a flat panel display. When not required, camera 30 can be retracted / extended (as shown diagrammatically by arrow 30') to move the camera off axis B'.

[0034] - STEM camera 32. The output from camera 32 can be recorded as a function of the (X,Y) scanning position of beam B on sample S, and an image can be constructed that is a "map" of the output from camera 32 as a function of X,Y. Camera 32 can also be an Electron Microscope Pixel Array Detector (EMPAD), but can comprise a single pixel, e.g., 20 mm in diameter, as opposed to the pixel matrix typically present in camera 30. Furthermore, camera 32 generally can be a 10 mm pixel array, e.g., ... 2 Much higher acquisition rates (e.g., 10 images / sec) 6 points / sec). Again, if not required, camera 32 may be retracted / pulled out (as indicated diagrammatically by arrow 32') out of the way of axis B' (such retraction may not be necessary, for example, in the case of a donut-shaped annular dark field camera 32, where a central hole may allow for bundle passage when the camera is not in use).

[0035] As an alternative to imaging using the cameras 30 or 32, a spectroscopic device 34 can also be invoked, which can be, for example, an EELS module.

[0036] It should be noted that the order / position of items 30, 32, and 34 is not strict and many possible variations are contemplated. For example, spectroscopic device 34 could be integrated with imaging system 24.

[0037] In the illustrated embodiment, the microscope M further comprises a retractable X-ray Computed Tomography (CT) module, generally designated by the reference numeral 40. Computed tomography (also called tomographic imaging) uses electron sources and (diametrically opposed) detectors to interrogate a sample along different lines of sight to obtain an insightful view of the sample from different perspectives.

[0038] It should be noted that a controller (computer processor) 20 is connected to the various components shown via control lines (bus) 20'. This controller 20 may provide various functions such as synchronizing actions, providing set points, processing signals, performing calculations, and displaying messages / information on a display device (not shown). Needless to say, the controller 20 (shown diagrammatically) may be (partially) inside or outside the housing 2 and may have a unitary or composite structure, as desired. The controller, as shown in this embodiment, comprises a data processing device P configured for carrying out the method defined herein.

[0039] Those skilled in the art will understand that the interior of the housing 2 does not have to be kept in strict vacuum conditions. For example, in so-called "environmental TEM / STEM", a background atmosphere of a given gas is deliberately introduced / maintained inside the housing 2. Those skilled in the art will also understand that in practice, it may be advantageous, if possible, to restrict the volume of the housing 2 so as to essentially embrace the axis B', taking the form of a small tube (e.g. of dimensions of the order of 1 cm), transparent to the employed electron beam, but extending outwards to accommodate structures such as the electron source 4, the sample holder H, the screen 26, the cameras 30, 32, the spectrometer 34, etc.

[0040] Referring now to Figure 2, another embodiment of the device according to the invention is shown. Figure 2 (not to scale) is a very schematic representation of a charged particle microscope M according to the invention, more specifically showing an embodiment of a non-transmission microscope M, in this case an SEM (although in the context of the invention, for example, an ion-based microscope may be equally useful). In the figure, parts corresponding to items in Figure 1 are indicated using the same reference numbers and are not considered separately here. The following parts (among others) are added to Figure 1: -2a: vacuum port, which can be opened to introduce / remove items (components, samples) to / from the interior of the vacuum chamber 2 or in addition thereto, for example, auxiliary devices / modules can be attached. The microscope M can be equipped with several such ports 2a if desired. - 10a, 10b: Schematically depicted lenses / optical elements in illuminator 6 -12: A voltage source that allows the sample holder H or at least the sample S to be biased (floated) to a certain potential relative to ground, if desired. -14: Display such as FPD or CRT. - 22a, 22b: (Segmented electron detector 22a, with a number of independent detection segments (e.g. quadrants) arranged around a central aperture 22b (allowing the passage of the beam B). Such a detector can be used, for example, to study the (angular dependence of) the flux of output (secondary or backscattered) electrons emerging from the sample S.

[0041] Here too, a controller 20 is present. The controller is connected to a display 14, which may be connectable to a data processing device P configured to execute the methods defined herein. In the embodiment shown, the data processing device P is a separate structure that does not form part of the controller or even part of the microscope P. The data processing device P may be local or cloud-based and, in principle, is not limited in location. It is noted that in all embodiments described herein, the data processing unit P may be part of the charged particle beam device, such as a charged particle microscope, or may be located externally.

[0042] Turning now to Figure 3, an embodiment of the method defined herein is shown in schematic form. The method 101 is used for inspecting a specimen using a charged particle beam device, - detecting 110, using a first detector, a first type of radiation from the sample in response to a charged particle beam illuminating the sample; - acquiring 120 spectral information on a second type of radiation from the sample in response to the charged particle beam illuminating the sample.

[0043] As indicated previously, prior art methods have used two detectors, such as a backscattered electron detector for detecting a first type of radiation in the form of backscattered electrons, and an EDS detector for detecting a second type of radiation in the form of X-rays. However, the method defined herein does not in principle require the use of a spectral detector, even though such a detector may of course be present. To this end, the method defined herein uses the following steps to obtain 104 spectral information: a first step is providing 102 a spectral information prediction algorithm, and a second step is using 103 said algorithm to predict said spectral information. The spectral information prediction algorithm uses the first type of detected radiation as an input parameter of said algorithm. For this reason, the spectral information prediction algorithm is configured to predict spectral information based on the first type of detected radiation, for example in the form of backscattered electrons. In an embodiment, the spectral information is therefore related to EDS information predicted based on the detected backscattered electrons.

[0044] Turning now to Figures 4a-c, examples of results achievable with the spectral information prediction algorithm are shown. Figure 4a shows a backscattered electron image of a semiconductor sample, where the chip structure and layout are recognizable to those skilled in the art. Figure 4b shows an EDS image of the sample shown in Figure 4a, where the use of an EDS detector has been performed. It can be seen that in many places the data is relatively noisy, which makes it difficult to accurately and confidently identify the material in some areas. Figure 4c shows an EDS image of the sample shown in Figure 4a, where the use of the spectral information prediction algorithm as defined herein has been performed. It can be seen that the signal-to-noise ratio is improved for the measured EDS data.

[0045] The algorithm used to predict the spectral data can be configured to identify inter-spatial data relationships in the radiation of the first type. In the example shown in Fig. 4c, the algorithm can identify certain regions in the EM image (Fig. 4a) that have common characteristics based on which the algorithm can more quickly and / or more accurately identify these regions as having the same or similar spectral data. Thus, the algorithm can be configured to use the identified inter-spatial data relationships to predict the spectral information at a given sample location.

[0046] In one embodiment, the algorithm includes at least one neural network. As known to those skilled in the art, a neural network (NN), also referred to as an artificial neural network (ANN) or simulated neural network (SNN), is an interconnected group of artificial neurons that uses mathematical or computational models for information processing based on a connectionist approach to computation. An artificial neural network, as defined herein, is an adaptive system that changes its structure based on external or internal information flowing through the network. In more practical terms, a neural network is a nonlinear statistical data modeling or decision-making tool that can be used to model complex relationships between inputs and outputs or to find patterns in data. Learning with neural networks is particularly useful in applications where the complexity of the data or the task makes the design of such functions by hand impractical. Thus, the application of neural networks is advantageous for optimizing alignment procedures in any charged particle beam device. In the present disclosure, a neural network can be trained on sample data that includes data related to a first type of radiation (e.g., backscattered electron images) as input parameters and data related to a second type of radiation (spectral data, e.g., EDS data) as desired output parameters. This data can be readily obtained and therefore used for training purposes in a fast and reliable manner. Thus, the spectral information prediction algorithm can be - providing sample data relating to a first type of radiation from the sample, particularly where said first type of radiation is related to non-spectral data; - providing corresponding spectral information on a second type of radiation from the sample, in particular when said spectral information is related to the measured spectral information; - using the provided data to develop the algorithm, in particular to train the neural network, where the first type of radiation is used as input and the spectral information is used as the desired output.

[0047] Figure 5 shows an example of a neural network architecture that can be used in the spectral information prediction algorithm defined herein. The neural network utilizes the concept of convolutional layers, attention blocks, and an autoencoder / decoder architecture. The MobileNet V2 architecture is used as the neural network backbone, and a depthwise separable convolutional layer is added, resulting in a faster and more powerful encoder / decoder network. Attention-augmented convolutional blocks are added to the bottleneck of the encoder / decoder architecture, which better captures pixel-wise relationships across the image. The key idea behind self-attention is to generate a weighted average of values ​​calculated from hidden units. The weights used in the weighted average operation are dynamically generated via a similarity function between hidden units. Notably, this allows self-attention to capture long-range interactions.

[0048] The task at hand is posed in the embodiment as a semantic segmentation problem, which aims to assign a semantic label to every pixel in an image. The algorithm differs from classical semantic segmentation in that multiple labels can be assigned to each pixel (multiple chemical elements can be present at the same pixel location), the pixel intensities of the predicted analysis map (neural network output) have different values, and it provides a quantitative result (atomic percentage of a defined chemical element).

[0049] The method defined herein is particularly useful when samples that do not show much variability need to be examined. For example, in quality control processes where examination may lead to the acceptance or rejection of a sample or process, or when a single type of sample is being investigated. Examples may include semiconductor samples and mineralogy samples. However, it is believed that the spectral information prediction algorithm can also be trained for a wide variety of samples.

[0050] Charged particle beam devices such as those shown in Figures 1 and 2 may be configured to perform the methods defined herein. The controller may be configured to predict, for example, using a spectral information prediction algorithm, spectral information based on a first type of detected radiation, the first type of detected radiation being associated with non-spectral information such as backscattered electrons. A spectral detector may be included and may be used, for example, as an input parameter to the spectral information prediction algorithm, although this is not required by the present disclosure. Thus, any detector available in a charged particle beam device may be associated with a non-spectral detector.

[0051] The desired protection is provided by the appended claims.

Claims

1. 1. A method of inspecting a specimen using a charged particle beam device, comprising: - detecting, using a first detector, a first type of radiation from the sample in response to a charged particle beam illuminating the sample; acquiring spectral information on a second type of radiation from the sample in response to the charged particle beam illuminating the sample, The step of obtaining spectral information further comprises: - providing an algorithm for predicting spectral information comprising at least one neural network, said neural network being trained using sample data comprising data related to said first type of radiation as input parameters and data related to said second type of radiation as desired output parameters; using said algorithm to predict said spectral information based on the detected radiation of said first type as an input parameter of said algorithm, The method, wherein the algorithm is configured to identify inter-spatial data relationships in the radiation of the first type.

2. The method of claim 1 , wherein the algorithm is configured to identify inter-spatial data relationships in the radiation of the first type.

3. The method of claim 2 , wherein the algorithm is configured to use the identified inter-spatial data relationships to predict the spectral information at a given sample location.

4. A method according to any preceding claim, wherein the algorithm is configured to use the first type of radiation to identify regions of the sample having similar properties.

5. The method of any one of claims 1 to 4, wherein the input parameters include a backscattered electron image and the output parameters include EDS data.

6. The method according to any one of claims 1 to 5, wherein the sample is a semiconductor sample.

7. The method of any one of claims 1 to 6, wherein the spectral information comprises elemental information of the sample.

8. The method of any one of claims 1 to 7, comprising the further step of accepting or rejecting the sample based on the acquired spectral information.

9. 9. The method of any one of claims 1 to 8, comprising the further step of outputting data related to the predicted spectral information, said output comprising at least one of an image of the sample, elemental information of the sample, and / or a rating for the sample.

10. Charged particle beam device for inspecting a specimen using the method according to any one of claims 1 to 9, comprising: a charged particle beam source for emitting a charged particle beam, a sample holder for holding a sample, an illuminator for directing the charged particle beam emitted from the charged particle beam source onto the sample; a first detector for detecting a first type of radiation arising from the sample in response to illumination of the sample by charged particles emitted from the charged particle beam source; a control unit connected to the first detector and configured to control at least a part of the operation of the charged particle beam device, Charged particle beam device, characterized in that the charged particle beam device is configured to carry out the method according to one or more of claims 1 to 9.

11. The charged particle beam device of claim 10, wherein the charged particle beam device comprises at least one detector including the first detector, and wherein all of the at least one detectors consist of non-spectral detectors.

12. A method for developing an algorithm, in particular a method for training a neural network, for use in the method according to any one of claims 1 to 9, said method comprising the steps of: - providing sample data relating to a first type of radiation from the sample; - providing corresponding spectral information regarding a second type of radiation from said sample; - using the provided data to develop the algorithm, in particular for training the neural network, wherein a first type of radiation is used as input and spectral information is used as desired output.

Images