Method for inspecting semiconductor samples layer by layer, and inspection apparatus for performing such method.

Abstract

For layer-by-layer (181 to 1818) inspection of a semiconductor sample (2) using focused ion beam etching and X-ray detection, an initial layer (18i) to be inspected is prepared via FIB etching. A surface area of ​​a target area volume (20) of the prepared layer (18i) of the sample (2) is aligned with the objective field of view of a scanning electron microscope (SEM). The electron energy of an electron beam (8) of the SEM is adjusted. The target area volume is probed by a scanning electron beam in the objective field of view. X-rays emanating from the aligned target area volume are detected. Detection signals acquired during the detection step are post-processed to deconvolve the detection signals into structured data derived from the sample structure in the target area volume. The next inspected layer is prepared by FIB etching and the steps "preparing" to "post-processing" are repeated until the layer-by-layer inspection of the overlapped target volumes of the sample (2) is completed. An inspection apparatus for carrying out such a method includes a focused ion beam source, a scanning electron microscope, an X-ray detection device, and a computer for performing post-processing, resulting in a method with improved volumetric spatial resolution within the volume of interest.

Description

[Technical Field] 【0001】 The contents of German Patent Application No. 10 2022 201 394.8 are incorporated by reference. 【0002】 The present invention relates to a method for inspecting semiconductor samples layer by layer using focused ion beam etching and X-ray detection. Furthermore, the present invention relates to an inspection apparatus for carrying out such a method. [Background technology] 【0003】 An inspection apparatus for inspecting semiconductor samples using a focused ion beam / SEM cross-beam method is disclosed in U.S. Patent No. 8,969,835. Inspection methods using X-ray spectrometry in SEM are known in the art. 【0004】 Details of focused ion beam tomography are disclosed in M. Cantoni and L. Holzer: Advances in 3D focused ion beam tomography, MRS Bulletin Vol. 39, pp. 354-360 (2014). Details of focused ion beam (FIB) tomography are disclosed in I. Utke (eds.), S. Moshkalev (eds.), and P. Russell (eds.), Oxford series on nanomanufacturing, Nanofabrication using focusing ion and electron beams, Principles and applications, 2012, pp. 410-435, and L. Holder and M. Cantoni: Review of FIB tomography. P. Burdet's doctoral dissertation (2012) from the Swiss Federal Institute of Technology Lausanne, "Three Dimensional Microanalysis by Energy Dispersive Spectrometry: Improved Data Processing," discloses an extended form of energy-dispersive spectroscopy (EDS) for 3D microanalysis using a combined microscopy technique, namely, an electron microscope equipped with a focused ion beam. In "Enhanced Quantification for 3D Energy Dispersive Spectrometry: Going Beyond the Limitation of Large Volume of X-Ray Emission," Microsc. Microanal. Vol. 20, pp. 1544-1555, 2014, a method developed to quantify three-dimensional energy-dispersive spectroscopy measurement data with voxel sizes smaller than the volume of the X-ray emission source is disclosed.P - Burdet, S. A. Croxall, P. A. Midgley: Enhanced quantification for 3D SEM - EDS: Using the full set of available X - ray lines, Ultramicroscopy 148, pp. 158 - 167, 2015 discloses a method for quantifying energy - dispersive spectra recorded in 3D using a scanning electron microscope, which uses all the available X - ray lines generated by the beam. L. Struder, A. Niculae, P. Holl, H. Soltau: Development of the Silicon Drift Detector for Electron Microscopy Applications, Microscopy Today, 28(5), September 2020 details the development of the silicon drift detector. U.S. Patent No. 6,924,484 discloses the evaluation of void characteristics of metal interconnect structures using X - ray emission analysis. International Application Publication No. 2019 / 071352 discloses a method for cross - sectional sample preparation. European Patent Application No. 2,557,584 discloses a charged - particle microscope imaging method. U.S. Patent No. 11,282,670 discloses the reconstruction of the slice depth of charged - particle images using model simulations to improve the generation of 3D sample images. U.S. Patent Application Publication No. 2022 / 0102121 discloses the reconstruction of the depth of 3D images of samples in a charged - particle system. 【Summary of the Invention】 【0005】 The object of the present invention is to improve the above - mentioned inspection method, in particular to improve the elemental information from the region of interest volume in order to improve the volume spatial resolution within the target volume. 【0006】 This object is achieved by a method according to the features of claim 1. 【0007】 The number of layers prepared in this method can be within the range of 2 to 100. In addition to the detection of X-rays by inelastic electron / material interaction, Auger electrons (AE) may also be detected. As a result of post-processing, the volume spatial resolution within the volume of interest is improved. Such volume spatial resolution may be better than 25 nm, better than 20 nm, better than 15 nm, and better than 10 nm. Practically, such volume spatial resolution is more than 1 nm. The definition of such volume spatial resolution is carried out by referring to the edge resolution criterion for distinguishing adjacent structures. Examples of edge resolution criteria are known in the art. 【0008】 The post-processing step may include an optimization algorithm. By this inspection method, it is possible to inspect chip structures, particularly semiconductor memory structures, more particularly flash memory structures, more particularly NAND structures, more particularly 3D-NAND structures, and more particularly vertical 3D-NAND structures. 【0009】 When using the wavelength-dependent X-ray detection according to claim 2, the versatility of this inspection method is further improved. In addition to spectral X-ray detection (EDX) for detecting inelastic scattered X-rays, Auger electrons (AE) can also be detected. As a result of such EDX and / or AE detection, elemental information is obtained from the target region volume. 【0010】 Typical detected X-ray energies can be in the range of 50 eV to 3 keV. The lines to be detected may include the K, L, M, and N lines of each element in the sample being inspected. In this regard, it is recalled that 277 eV is derived from carbon, 392 eV is derived from nitrogen, 452 eV is derived from titanium, 525 eV is derived from oxygen, 1486 eV is derived from aluminum, 1740 eV is derived from silicon, and 1775 eV is derived from tungsten. In one example, electron energies less than 1000 eV are used. Such electron energies improve the effect of post-processing deconvolution. Further improvement in volume spatial resolution is possible by utilizing wavelength-dependent X-ray detection and / or lower electron energies. 【0011】 In the post-processing step, it may be considered that the free interaction path length of the probe electrons depends on the material density and therefore on the elemental composition. 【0012】 Furthermore, this method may be used to examine the generation and / or quantity of dopants. 【0013】 The goal is to improve the 3D resolution in analytical tomography FIB-SEM imaging, i.e., the spatial three-dimensional acquisition of spectral X-ray data for material imaging (EDS / EDX) within the sample volume of bulk materials. The method is based on the effect of X-ray emission resulting from the excitation (and subsequent relaxation) of core-shell electrons due to the kinetic energy of electrons in the electron beam of the microscope. The excitation occurs within the sample volume, known as the interaction volume or the region of interest volume. 【0014】 Conventionally, one aspect that limits the resolution in this technique is the size of the interaction volume between the electron beam and the sample, that is, the size of the sample volume such that the primary electrons have sufficient energy to excite X-ray emission. 【0015】 The size of the interaction volume increases with increasing beam energy (acceleration voltage), which suggests a low value for high resolution. On the other hand, electrons need to have sufficient kinetic energy to excite core-shell electrons for a wide range of possible elements that may be present in the sample. For this reason, the practical minimum energy range is about 1 keV. In this case, the de Broglie wavelength is about 38 pm. However, depending on the element and how accurately the interaction volume range is measured, the size of the interaction volume can be on the order of about 20 nm. 【0016】 High-resolution EDX scanning may require low electron beam energies to minimize the size of the electron interaction volume in the sample. However, the efficiency of X-ray photon generation is very low at the low beam energies used in this example. Therefore, there has traditionally been a trade-off between resolution and acquisition time. 【0017】 By employing post-processing within the inspection method, an algorithmic super-resolution technique is obtained that improves the resolution of analytical EDX beyond the limitations of prior art. The spatial mixing behavior introduced by the interaction volume can be mathematically modeled by computational means in the post-processing step. Here, a reasonably realistic mathematical model of the measurement / interaction process can be used. 【0018】 Approximations may be used during post-processing to derive a viable mathematical model for spatial inversion / deconvolution. This may include the reconstruction of high-resolution material maps from a 3D stack of spatial-spectral-resolved EDX measurements, i.e., from both spatial and spectral resolution. Such reconstructions may include formulations concerning elemental emission lines, which require processing of the acquired EDX spectra to remove contributions from different elemental lines to the observed spectra. Information from different modalities may be used to support the reconstruction task. 【0019】 A typical measurement setup may be a spatial-spectral-resolved 3D EDX scan, which can have any scanning geometry. The measurement geometry may be a cross-beam configuration (ion beam / electron beam) or a tilted FIB acquisition, i.e., a tilt angle between the ion (FIB) and electron (SEM) beams in the range of 0° to 90°. The results of the inspection method can be a corrected and sharpened map of material volume for various elements present in the sample. 【0020】 The post-processing refinements described in claims 3 to 7 enable further improvement of the volumetric spatial resolution and / or elemental information within the target region volume. 【0021】 The geometric input or other prior condition input described in claim 7 may be achieved by further independent measurements. SEM imaging may provide such further geometric input and / or prior condition input. The resolution of the post-processing step can be improved by using known geometries. Other material-dependent conditions obtained from a material library may contribute to the prior condition input of the post-processing step. 【0022】 The post-processing described in claim 8 is a modification for deconvolving the measurement data and for generating a sample image with advantageously high resolution. Further details relating to further steps of the method of claim 8 are referenced in European Patent No. 2 557 584. 【0023】 The value constraint described in claim 9 is particularly advantageous. 【0024】 The computational simulation described in claim 9 may be performed using Monte Carlo simulation and / or finite element analysis. 【0025】 The determination of the minimum divergence described in claim 8 may be the least squares distance, the Csiszar-Morimoto F divergence, the Bregman divergence, the Alpha-Beta divergence, the Bhattacharyya distance, the Cramer-Rao limit, and / or derivatives thereof. 【0026】 The beam parameters for a focused ion beam and / or electron beam can be selected from the following parameters: beam energy, beam focusing angle, and / or beam focusing depth. 【0027】 During the testing procedure, simulated radiation emitted from the sample during each measurement may be detected. Such induced radiation may be secondary electrons, backscattered electrons, and / or X-ray radiation. Intensity measurements and / or current measurements may be performed. 【0028】 The physical properties of the spatial variables may be the density of the action factor, the atomic density, and / or the secondary emission coefficient. 【0029】 The material removal method in each preparation step may be mechanical slicing with a cutting device, ion milling with an ion beam, ablation with an electromagnetic beam, beam-induced etching, chemical etching, and / or reactive etching. 【0030】 In the inspection method, the step of physically slicing may be combined with the step of computationally slicing. These physical / computational slicing steps may be repeated alternately. 【0031】 A further object of the present invention is to provide an inspection apparatus capable of performing an inspection method. 【0032】 This objective is achieved by the inspection apparatus described in claim 10. The advantages of such an inspection apparatus correspond to those described above with respect to the inspection method. This also applies to the inspection apparatus described in claim 11. 【0033】 The preparation and / or probing and / or detection geometries described in claims 12-14 have been shown to reduce or avoid unwanted contamination of the target area volume by ions and / or debris generated by the FIB etching preparation step. 【0034】 The angle between the etching plane and the initial bulk sample surface plane may be in the range of 20° to 40°, and may be approximately 35°. 【0035】 Exemplary embodiments of the present invention are described below with reference to the accompanying drawings. [Brief explanation of the drawing] 【0036】 [Figure 1]This is a schematic diagram of an inspection apparatus for performing an inspection method to inspect semiconductor samples layer by layer, including a focused ion beam (FIB) source, a scanning electron microscope (SEM), and an X-ray detector. [Figure 2] This figure shows the sample inspection geometry, which indicates the orientation of the FIB etching plane that defines the subsequent sample layer to be prepared by FIB etching, the initial bulk sample surface plane of the semiconductor sample to be inspected, and the SEM probe direction that coincides with the X-ray detection direction. [Figure 3] This is an axial cross-sectional view of a 3D vertical NAND flash memory design, where adjacent areas of different materials are depicted with different hatching. [Figure 4] This is a schematic cross-sectional view of the target area volume of the sample being examined, including a schematic sketch of the electron beam / sample interaction and the X-rays emitted from each interaction area of ​​such volume. [Figure 5] This figure, similar in depiction to Figure 4, shows that the size of the electron beam / sample interaction volume depends on the energy of the probing electrons. [Figure 6] This figure illustrates, similar to the depiction in Figure 5, that the interaction volume depends on the atomic number of the sample element present in the interaction volume. [Figure 7] This is a schematic diagram showing the FIB-SEM scanning sequence of a semiconductor sample, such as a memory cell capacitor, embodied as part of a vertical NAND design. In the left-to-right sequence, which is part of the inspection method, the electron beam / sample interaction volume moves from the upper layer where the memory cell capacitor is located to the bottom substrate layer. [Figure 8] Similar to Figure 4, this figure shows a Monte Carlo simulation of the electron beam / sample interaction volume, which may be part of the post-processing steps of the inspection method, with the electron beam energy increasing from top to bottom. The bottom panel of Figure 8 also depicts the X-ray path. [Figure 9] This is a schematic diagram of the volume of the target area at the vertical (Figures 9 and 10) and horizontal (Figures 11 and 12) boundary lines between lighter and heavier materials. [Figure 10] This is a schematic diagram of the volume of the target area at the vertical (Figures 9 and 10) and horizontal (Figures 11 and 12) boundary lines between lighter and heavier materials. [Figure 11] This is a schematic diagram of the volume of the target area at the vertical (Figures 9 and 10) and horizontal (Figures 11 and 12) boundary lines between lighter and heavier materials. [Figure 12] This is a schematic diagram of the volume of the target area at the vertical (Figures 9 and 10) and horizontal (Figures 11 and 12) boundary lines between lighter and heavier materials. [Figure 13] These figures, similar to Figures 9-12, show a comparison of the target region volume in lighter and heavier materials using the same probe electron energy. [Figure 14] This is a schematic diagram of a composite method that approximates the interaction volume region as a superposition of lighter and heavier material volume portions at a vertical boundary. [Figure 15] This is an SEM scan image through a semiconductor sample layer, depicting a central carbon region confined by two adjacent tungsten regions. [Figure 16] Figure 15 shows the results of the inspection method using the inspection device shown in Figure 1, relating to the tungsten / carbon boundary region on the left side. [Figure 17] This figure shows the results of the inspection method using the inspection apparatus shown in Figure 1, relating to the carbon / tungsten boundary region on the right side of Figure 15. [Modes for carrying out the invention] 【0037】 Figure 1 schematically shows an inspection apparatus 1 designed to inspect a semiconductor sample 2 layer by layer using focused ion beam (FIB) etching and X-ray spectroscopy. The inspection apparatus 1 includes a focused ion beam (FIB) source 3 and a scanning electron microscope (SEM) 4 including an electron optical system 5. Furthermore, the inspection apparatus 1 includes an X-ray detector 6 for detecting X-rays 7 (comparable to Figure 4) emitted from the sample 2, which are generated by probe electrons 8 (comparable to Figures 2 and 4) of the SEM 4. Furthermore, the inspection apparatus 1 includes a computer 9 for post-processing the detection signal acquired by the X-ray detector 6. 【0038】 The X-ray detection device 6 includes an X-ray spectrometer 10, schematically shown in Figure 1. Wavelength-dependent X-ray detection is possible using the X-ray spectrometer 10 of the X-ray detection device 6. 【0039】 Figure 1 further shows a stop mechanism 11 for defining the target region on sample 2. 【0040】 Sample 2 is located within the inspection chamber, which is bounded by chamber walls 12 and 13. 【0041】 The aperture mechanism 11 defines a process window 14 that can be bounded by movable aperture mechanism components, as schematically shown in Figure 1. 【0042】 Figure 2 schematically depicts the geometry of sample 2 on one hand and the inspection geometry of inspection device 1 on the other. The Cartesian xyz coordinate system in Figure 2 corresponds to that shown in Figure 1. 【0043】 Sample 2 has an initial bulk sample surface plane 15 located parallel to the xy plane. The electron beam from SEM 4, i.e., probe electrons 8, probes sample 2 at an angle of less than 90° from the initial bulk sample surface plane 15. Furthermore, the X-ray detector 6 detects X-rays 7 at an angle of less than 90° from the initial bulk sample surface plane 15. In other words, the direction of one SEM probe and the detection direction of the X-ray detector are either coincident or extend parallel to each other. 【0044】 The FIB 16 generated by the FIB source 3 etches the sample 2 in an etching plane 17 that includes an angle of 36° with respect to the xy initial bulk sample surface plane 15. Such an angle can be, for example, in the range of 20° to 40°. 【0045】 Figure 2 shows a temporary state of the inspection method using the inspection apparatus 1, in which initial layers 18 and 181 to be inspected on the semiconductor sample 2 via FIB 16 are prepared. The surface plane of these initial layers 18 coincides with the temporary etching plane 17. 【0046】 Because the etching plane 17 extends at an angle less than a certain angle with respect to the initial bulk sample surface plane 15, the alternating elemental layer structure of the first element A and the second element B in sample 2 lies along the initial layer 18. These A / B / A… elemental layer structures extend parallel to the initial bulk sample surface plane 15. Furthermore, a NAND structure 19 having a cylindrical boundary with a cylindrical axis extending parallel to the z-axis is cut at a certain angle within this initial layer 18. Due to the etching plane 17 being inclined with respect to the z-axis, the contour of the NAND structure 19 becomes elliptical as a result of such cylindrical cutting. 【0047】 After the layer is prepared, the surface area of ​​the target region volume 20, which has a droplet shape as shown by the dashed line in Figure 4, is aligned with the object field 21 of the SEM 4. The target region volume 20 is the volume in which electron / sample interactions occur, which are detected and analyzed during the inspection method. The object field 21 extends parallel to the xy plane. In Figure 4, the actual slope between the sample layer surface and the initial bulk sample surface plane is omitted. 【0048】 The electron energy of the electron beam 8 of the SEM4 is adjusted, and the target area volume 20 is probed with the electron beam 8 within the objective field of view 21. The X-rays 7 emitted from the aligned target area volume 20 are detected via the X-ray detector 6. 【0049】 Next, post-processing of the detection signal acquired during the preceding detection step is performed in order to spatially deconvolve this detection signal with a structural signal derived from the structure within the target region volume 20, which will be described in more detail later. 【0050】 Subsequently, the next layer 182 of sample 2 is prepared. This next layer 182 has an additional sample surface beneath the previously etched sample surface. This preparation of the next layer 182 is carried out by etching the additional sample surface of the next layer 182 through etching of sample 2 using a aligned focused ion beam 16. These alignment steps and post-processing steps are repeated until the inspection of each layer of the target volume of sample 2 is completed. Subsequent layers 183-18 are the result of repeating these steps. 18 However, this is further shown in Figure 2. The layer thickness is the same as the adjacent layer 18 i and 18 i+1 This is determined by the distance between them. With such layer thicknesses, an effective cutting depth d along the z-direction is obtained, which is also depicted for each layer in Figure 2. 【0051】 layer 18 i The number can range from 2 to 1000, particularly in the range of 2 to 500 or 2 to 100. 【0052】 The inspection apparatus 1 includes an alignment unit 21a for aligning the surface area of ​​the target region volume 20 of the sample 2 with the objective field of view 21 of the inspection apparatus 1, which may be defined by the process window 14. Figure 1 schematically depicts such an alignment unit 21a in an effective state for the sample 2. Using the alignment unit 21a, translation and / or tilting of the sample 2 with respect to the objective field of view 14 of the inspection apparatus 1 is possible via at least three degrees of freedom. Depending on the embodiment of the alignment unit 21a, such alignment is possible via four, five, or six degrees of freedom. 【0053】 Furthermore, the inspection apparatus 1 includes an adjustment unit 22 (see also Figure 1) for adjusting the electron energy of the electron beam 8 of the SEM 4. Such adjustment can be performed by controlling the electron acceleration voltage in the electron optical system 5. 【0054】 The detection device 6, including the FIB source 3, SEM 4, and spectrometer 10, the components of the aperture mechanism 11, the alignment unit 21a, and the adjustment unit 22 are signal-connected to a computer 9, which also functions as a control unit that controls the steps of the inspection method. 【0055】 Figure 3 schematically illustrates an example of a structure 23 within a semiconductor sample 2. Here again, the Cartesian xyz coordinate system in Figure 3 corresponds to that shown in Figures 1 and 2. Such a structure 23 can be realized as a 3D V-NAND structure. Such a structure includes a cylindrical SiO2 core 24 having a diameter d1. Such a core 24 is surrounded by a polysilicon sleeve 25, which may be doped to establish a transistor device. The top of the sleeve 25 is realized as a cover 26. The sleeve thickness of the sleeve 25 is shown as d2 in Figure 3. The height of the cover 26 is shown as h1 in Figure 3. 【0056】 Sleeve 25 is surrounded by a further SiO2 sleeve 27 with a thickness d3, and this sleeve 27 is similarly surrounded by a further Si2N3 sleeve 28 with a thickness d4. The resulting assemblies 24-28 constitute a plug hole with a volume of SiO2 bulk material 29. At the radial edge of this bulk volume 29 are further tungsten enclaves 30 and 31 with different radial extents. The height difference between the sleeve cover 26 and the upper of the two enclaves 30 and 31 is shown as h2 in Figure 3. The z-extension of the enclaves 30 and 31 is shown as t1 in Figure 3. The z-distance between adjacent enclaves 30 and 31 is shown as t2 in Figure 3. 【0057】 The typical thicknesses d2, d3, and / or d4 of the sleeve wall are within the range of 10 nm. 【0058】 The distance between the detached sections 30 and 31 and the outermost sleeve 28 is shown as d5 in Figure 3. 【0059】 The quantities to be measured by the inspection method performed by the inspection device 1 are the material structural dimensions d1-d5, h1, h2, t1, t2, and / or material (atomic type, elemental type, elemental composition type), and / or material composition, and / or material distribution of structural components 24-31, and / or the amount of dopant in one of these components 26-31. 【0060】 Figure 4 schematically shows the interaction process related to probing sample 2 using electron beam 8. Direct surface interactions at the layer surface within the objective field of view 21 where electron beam 8 strikes sample 2 generate Auger electrons (AEs) with discrete energies directly originating from each element of the material in the structure of sample 2. 【0061】 Secondary electrons SE, which carry topographic information of the surface structure within volume 32, are emitted from the first droplet-shaped volume portion 32 of the target region volume 20 located below the interacting surface. 【0062】 From a further droplet-shaped volume portion 33 located below the SE volume portion 32, which represents the next interaction path length between the probe electrons of the electron beam 8 and the sample 2, backscattered electrons BSE are emitted. From their energy, information about the atomic numbers of the elements contained within the volume portion 33, as well as further phase difference information, can be revealed. 【0063】 Characteristic EDX X-rays 7 with wavelengths derived from the atomic / elemental composition within a larger, droplet-shaped target region volume 20 are emitted, representing the next interaction path length between the probe electrons of the electron beam 8 and the sample 2. The depth of such a target region volume 20 depends, on the one hand, on the electron energy in the electron beam 8, and further on, on the other hand, on the other hand, on the other hand, on the atomic number of the elements present within the target region volume 20. 【0064】 Further radiation 34 (bremsstrahlung) and 35 (cathodoluminescence) are emitted from further droplet-shaped shells 36 and 37 located beneath the target region volume 20. These further radiations 34 and 35 can also be detected and analyzed within the inspection device 1, but are not particularly relevant to the following discussion. 【0065】 Figure 5 illustrates that the size of the target region volume 20 depends on the energy of the electrons in the electron beam 8. The three examples in Figure 5 are given for the same element that will be probed. 【0066】 Figure 5, on the left, shows the small target area volume 20 resulting from the low acceleration voltage set by the adjustment unit 22. 【0067】 In the center of Figure 5, a larger target area volume 20 is depicted, resulting from a moderate acceleration voltage of the adjustment unit 22. 【0068】 The right side of Figure 5 shows the larger target area volume 20 resulting from the high acceleration voltage of the adjustment unit 22. 【0069】 Figure 6 shows that the size of the target region volume 20 depends on the atomic number of the elements in the sample structure being probed. 【0070】 On the left side of Figure 6, a small target region volume 20 is shown, which results from probing an element with a high atomic number with a given, for example, a moderate accelerating voltage. 【0071】 In the center of Figure 6, a larger target region volume 20 is shown, resulting from probing elements with medium atomic numbers at the same accelerating voltage. 【0072】 The right side of Figure 6 shows the large target region volume 20 that occurs when elements with lower atomic numbers are probed with the same electron acceleration voltage. 【0073】 Therefore, the size of the target region volume 20 can also be used as an indicator of the elemental composition present in the sample to be examined. 【0074】 FIG. 7 shows an in-progress layer-by-layer FIB-SEM scan of a sample structure exemplified as sleeve 38 by one of sleeves 25, 27, 28 described above with reference to FIG. 3. 【0075】 On the left side of FIG. 7, for example, the situation after preparing the initial layer 18 to be examined i is shown (see also the above description regarding FIG. 2). What is shown is the target region volume 20, that is, the interaction volume within the sample 2 between an electron beam (not shown in FIG. 7) and the sample material. Such a target region volume has a quantitative volume X. 【0076】 The intersection of the electron beam 8 entering the first layer 18 i is represented by y. On the left side of FIG. 7, a location x as an exemplary location within the entire target region volume 20, which is the origin from which the radiation to be investigated is emitted, is also shown. From this location x, X-rays detected by the detection device 6 are emitted within the detection cone Ω. 【0077】 The target region volume 20 within the first layer 18 i includes a part from the material of the inner cylinder of the sleeve 38, a part of the sleeve 38 itself, a part of the material radially surrounding the sleeve 38, and further a part of the substrate material 39 below the sleeve 38. 【0078】 The central depiction in FIG. 7 shows the situation after preparing another layer 18 to be examined i+1 Therefore, the target region volume 20 in this case has not changed in size but has moved towards the substrate material 39. At this point, only a smaller portion of the target region volume 20 in the center of FIG. 7 includes the material of the inner cylinder or the sleeve 38. 【0079】 On the right side of FIG. 7, the next layer 18 i+2The situation after preparation is shown. At this point, the target region volume 20 has moved further toward the substrate 39. Of this target region volume 20, only a very small portion contains the material of the central cylinder. Layer 18 i+2 When probing, the material of sleeve 38 is hardly contained within the volume 20 of the area under investigation. 【0080】 As a result, layer 18 i , 18 i+1 By preparing each sample and carefully comparing the detected light rays, it becomes possible to estimate the structural and / or material composition of the sleeve structure within sample 2 in Figure 7. 【0081】 The interaction volume portions 32 and 33 within the target region volume 20 can be understood as volumes exhibiting kernel values ​​of the point spreading function. Such kernel values ​​are associated with the interaction parameters, and in particular with the energy of the incident electron beam 8, by defining a constrained point spreading function with kernel values ​​dependent on the type of interaction between the electron beam 8 and the sample, by defining spatial variables representing volume-dependent physical sample properties, and by defining the imaging properties of the measured electron beam and / or radiation emitted from the target region volume 20. Minimum divergence min D(M n ||K n *V) By determining this, it is possible to estimate the structural composition and / or material composition of the sleeve structure in particular within sample 2, and in the above formula, M n This represents the number of measurements for each of the electron beam 8 measurements with different measurement characteristics. K n This is the kernel value of the point spreading function that represents the behavior of the probing electron beam 8 in the bulk of sample 2, V is a spatial variable that expresses the physical properties of sample 2 as a function of its position within sample 2. In this regard, see European Patent No. 2 557 584. 【0082】 Figure 8 shows three different results from Monte Carlo simulations regarding the interaction between electrons from electron beam 8 and sample 2 within the target region volume 20. The trajectories of individual electrons are shown. 【0083】 Figure 8 above shows the simulation results using a low acceleration voltage for the probe electrons. 【0084】 The center of Figure 8 shows the situation using an intermediate acceleration voltage, while the lower part of Figure 8 shows the situation using a high acceleration voltage. The boundary of the target region volume 20 is also shown. Furthermore, the lower part of Figure 8 shows the path of the X-rays 7 toward the X-ray detector 40 of the X-ray detector device 6. 【0085】 During the post-processing of the detection signal acquired during the detection step, a geometry input or other priori condition inputs, particularly from further, preliminary measurements, may be used. 【0086】 Figures 9 to 11 illustrate examples of such a priori condition inputs. The solid line shows the initially assumed expansion of the target area volume 20. The dashed line shows a refined portion of the target area volume 20, taking into account priori knowledge about the distribution of various substances within the sample 2 being examined. 【0087】 Figure 9 shows the situation at the vertical boundary 41 between the lighter material A and the heavier material B. As explained above with respect to Figure 6, heavier materials with higher atomic numbers reduce the expansion of the target region volume (dashed line 20a in Figure 9 compared to the solid line for the target region volume 20). 【0088】 Figure 10 shows the opposite situation, where the main portion of the target region volume 20 is within the heavier material B, and the portion of the target region volume 20 crossing the vertical BA material boundary 41 is smaller. Since a larger portion of the target region volume 20 exists within the lighter material A, the actual resulting target region volume (line 20a in Figure 10) is larger than the initially assumed target region volume. 【0089】 Figures 11 and 12 show the condition of the horizontal layer boundary 42 between material A (lighter) and material B (heavier). 【0090】 In the situation shown in Figure 11, the uppermost layer of the heavier material B is separated from the lowermost layer of the lighter material A via a horizontal boundary 42. When the target region volume 20 enters the lighter material A, it expands into its actual shape 20a. 【0091】 In the situation shown in Figure 12, the uppermost layer is made of the lighter material A, and the lower layer, separated by the horizontal boundary 42, is made of the heavier material B. Therefore, the theoretical target region volume 20 is compressed to the actual target region volume 20a when it comes into contact with the heavier material B. 【0092】 Figure 13 shows a comparison of the target region volume 20a in the lighter material A and the target region volume 20B in the heavier material 20B, assuming that all other conditions of the incident electron beam 8 are the same, with a similar depiction to Figures 9-11. The target region volume 20A is more expanded compared to the more compressed target region volume 20B. As a result of such compression, the sample surface (corresponding layer 18) i Underneath the surface, the compressed target area volume 20B may expand more quickly. 【0093】 Figure 14 illustrates how, starting from knowledge of the position of the vertical boundary 41 equivalent to that in Figures 9 and 10, and further starting from knowledge of the extensions of the target area volumes 20A (lighter material) and 20B (heavier material), the approximation of the composite target area volume 20PP can be calculated in the post-processing step of the inspection method. 【0094】 Figure 14 is presented as an equation. 【0095】 Two "terms" are shown on the left. Under these a priori conditions, the contribution 20 of the volume of the target region of the lighter material A is on one side of the boundary 41. PP,A Enter the following. The second "term" on the left side of the equation in Figure 14 is the contribution of the volume of the area of ​​the heavier material B on the other side of the vertical boundary 41. PP,B This indicates. 【0096】 The right-hand side of the equation shown in Figure 14 is the resulting composite target region volume 20 pp This shows that this is the contribution of the volume of the target region 20 pp,A and 20 pp,B It is the sum of. 【0097】 In one example of the present invention, the composition is described by a superposition of homogeneous material scattering cross-sections. According to the present invention, such a simplified model configuration enables traceable optimization. In another example, the estimated densities of materials A and B do not affect the basic shape of the target region volume 20. Such a method can be used as an alternative to Monte Carlo simulation methods (see Figure 8 above). 【0098】 Figure 15 shows the prepared layer 18 of Exemplary Example 2 as a result of conventional SEM scanning. i A conventional SEM image is shown. The total horizontal width of the SEM scan shown is approximately 6 μm. Figure 15 shows this layer 18. i The drawing is shown from above, meaning that the xy-plane of the xyz coordinate system introduced earlier coincides with the drawing plane. The surrounding matrix material 21, which may be SiO2, is shown. In the center of Figure 15, from left to right, the material arrangement of tungsten ("W"), carbon ("C"), and tungsten ("W") again is shown. Layer 18 shown in Figure 15 i This is an example of the elemental structure being examined within sample 2. 【0099】 Figure 16 shows the detection and post-processing results of the inspection method according to the present invention, performed using inspection apparatus 1 on the W / C SEM / EDX scan line 43 shown at the W / C boundary in Figure 15. The total length of the scan line 43, shown as nm horizontal coordinates in Figure 16, is 250 nm. 【0100】 The measurement was performed using a probe electron energy of approximately 3 keV from electron beam 8. The EDX intensity I measured in the X-ray bandwidth corresponding to the boundary [270 eV, 290 eV] is shown as line 44 in Figure 16. 【0101】 In the post-processing step of the inspection method according to the present invention, a convolutional EDX intensity I 45 is generated from the measured EDX intensity I 44 by smoothing the riffle of the measured EDX intensity I 44. 【0102】 Figure 16 further shows the spatial inverse convolution EDX intensity I 46, which is the final result of the post-processing step of the inspection method. Unlike the measured EDX intensity I 44 and convolution EDX intensity I 45, the inverse convolution EDX intensity I 46 shows a sharp increase at the W / C boundary location along the W / C scan line 43. Such a sharp increase corresponds to a resolution of approximately 20 nm in terms of W / C boundary reproducibility. 【0103】 Figure 17 shows the results of X-ray detection via the C / W boundary scan line 47 of the SEM / EDX shown in Figure 15. The total length of scan line 47 is also approximately 250 nm. Figure 17 shows the situation when tungsten is detected again using a 2 keV probe electron energy of electron beam 8, but this time X-rays corresponding to EDX energies with bandwidths of [1760 eV, 1780 eV] are detected. Here, the measured spatial spectrum 44, convolutional spatial spectrum 45, and inverse convolutional spatial spectrum 46 have edge sharpness that is almost the same to varying degrees, and again yield a resolution of approximately 20 nm. 【0104】 The vertical coordinates in Figures 16 and 17 represent the normalized X-ray intensity. 【0105】 Comparing the results in Figures 16 and 17, it can be seen that the spatial deconvolution in the post-processing step of the inspection method according to the present invention is particularly advantageous for lower X-ray energies detected, especially those originating from the elements carbon (approximately 280 eV), nitrogen (approximately 390 eV), titanium (approximately 450 eV), and oxygen (approximately 530 eV). 【0106】 As discussed above, Figures 4-6 show sketches of the interaction volume between the electron beam 8 and the material sample 2, where the material sample 2 is assumed to be homogeneous. The sketches are considered in three dimensions, with a certain point in the volume being x 【number】 or y 【number】 It is represented as follows. In the following analytical methods, the scattering cross-section (electron / electron or electron / X-ray) within the interaction volume is σ Si This is represented as (x;y), where x is the evaluation point in the volume and y is the target point of the electron beam. The subscript Si may indicate the material, for example, silicon in this case. 【0107】 As discussed above, in the example of the present invention, sample layer 18 i A FIB-SEM configuration is used, employing destructive 3D scanning of the sample volume by repeatedly removing the material. Therefore, the target point y 【number】 This can also be considered three-dimensional; please compare it with Figure 7. 【0108】 Generally, the scattering cross-section changes spatially, as evidenced by its dependence on y; that is, unless scanning a homogeneous material, the shape of the interaction volume usually changes with different scanning positions. Furthermore, throughout the entire scan, the sample surface 18 relative to beam 8... i The inclination is fixed, which is due to the parameters of the scanning device and the sample / device geometry. These device / geometry parameters are collectively referred to in the following analysis. 【0109】 【number】 It is called that. 【0110】 At a fixed scanning position y', the cross-sectional function is, in this case, a three-dimensional scalar function. 【number】 The actual form of the function is the material density distribution ρ1(x),…,ρ within the interaction volume X, i.e., the volume of the target region 20 (compared with Figure 7 left). N It depends on (x), where ρ i (x) represents the three-dimensional matter density, where N material species exist within the volume. 【0111】 The scattering cross-section is also parameterized by λ, i.e., the assumed X-ray energy in the EDX measurement performed using the inspection device 1. 【0112】 Given these premises, the measurement is 【number】 (1) It is formulated as follows, and in the above equation, the scattering cross-section σ is given by the material density function ρ1(x),…,ρ N (x), that is, the (unknown) distribution of elements / atoms in the sample and the parameters of the device / geometry. 【number】 It functions as a spatially angularly varying kernel that depends on the device settings and geometry stored in . The sample and device settings are separated (by ";") from the other parameters in the scattering cross-section argument list because they are considered to be fixed in a given experiment. The solid angle Ω (compare with Figure 7) represents the detector geometry, and ω is the differential solid angle toward the detector (multiple scattering is omitted). 【0113】 Equation (1) generally describes the multimode imaging technique. According to the multimode imaging technique, multiple intensities of electromagnetic radiation, including X-ray radiation, with different spectral ranges, here denoted by λ, are detected. The multimode imaging technique can also consider further secondary radiation, such as scattered electrons or secondary electrons. The present invention provides a method for generating an image of a sample with higher resolution by utilizing the multimode imaging technique and computer-aided inversion of the multimode imaging technique. In some cases, computer-aided inversion of the multimode imaging technique is improved by utilizing prior information about the sample being examined, such as CAD information or the sample's known material composition. In some cases, computer-aided inversion of the multimode imaging technique is improved by utilizing prior information about the material-specific spectral range of X-rays and typical scattering cross-sections. In some cases, computer-aided inversion of the multimode imaging technique is improved by using the FIB-SEM slice imaging method described above. The following are some examples of inspection methods that utilize computer-aided inversion of multimode imaging techniques. 【0114】 In the following, the spatially varying scattering cross-section within the interaction volume is referred to as the kernel. 【number】 This refers to instrument / geometry imaging parameters that affect the scattering cross-section, particularly Electron energy of probe electrons in electron beam 8 (acceleration voltage), unit [kV], Electron beam 8 beam current, unit [nA], The integral time of the detector 40 of the detection device 6, in units [s], For example, the sample layer 18 relative to the electron beam 8, measured with respect to the surface layer normal. i The angle of inclination, in units [deg] Collect them. 【0115】 It should be noted that the kernel is a vector value because, upon excitation by electron beam 8, X-rays 7 with different energies E=hν, where ν=c / λ, are generated. Here, ν is the photon frequency, c is the speed of light, h is Planck's constant, and λ is the photon wavelength. For reference: 【0116】 【number】 (2) In the above equation, the wavelength is given in units of [nm]. 【0117】 The kernel is generally unknown and depends on several parameters. Below, we will examine the main effects and their impact on the superresolution problem. 【0118】 The fundamental physical principle of the kernel is the (multiple) scattering process of primary electrons used in probing the sample (Figure 8). Generally, there are three scattering processes with interaction products (see also Figure 4): 1. Electrons: When electrons interact with a sample, they are scattered through various underlying mechanisms, and primary electrons lose energy in this process. 2.X-ray: (a) Since electrons are charged, when their velocity or direction changes, they emit electromagnetic radiation. This radiation has a continuous spectrum and is called bremsstrahlung radiation. (b) The second X-ray source is the ionization of atoms in the material being examined. X-rays are generated from transitions in the inner electron shells of atoms in the material (photons are produced when they relax to the ground state). The ionization spectrum is highly irregular and characteristic of the material. This is also called characteristic radiation. The probability of the radiation transition is given by w. (c) Ionized atoms may also be relaxed through a non-radioactive process called the Coster-Kronig transition, which results in the emission of Auger electrons (AE). The probability of the non-radioactive transition is given by 1-w. 3. Visible Light: When the kinetic energy of electrons becomes sufficiently low, only the outer electron shells of the atom become excitable, producing visible light emission known as cathodoluminescence. This is low-resolution (emitted from a large interaction volume), but when resolved by spectroscopy, it contains information about the atomic species. 【0119】 Due to the size limitations of the detector, it is not possible to capture all backscattered electrons or photons. The modified kernel is limited to the collected secondary radiation and is restricted to describing only a portion of the X-ray photons that hit, for example, detector 40. 【0120】 Furthermore, an electric current may be generated within the material, which could interact with the incident electrons. Also, if the electrons are not "discharged" quickly enough, a charging phenomenon can occur. 【0121】 Because X-rays can be multiple-scattered, they can also produce fluorescence, meaning they can introduce a Stokes shift to the measurement wavelength. This usually results in new spectral peaks at lower energies. 【0122】 Device / Geometry / Imaging / Sample Parameters 【number】 The main effects are as follows: Excitation energy: The energy of electrons in the primary electron beam affects the size of the interaction volume; see Figure 5. This is actually affected by the acceleration voltage. Higher primary electron energy results in a larger interaction volume, i.e., a larger kernel, which means a lower resolution for the EDX image / 3D stack. 【0123】 Since the accelerating voltage is an instrument parameter of the SEM, the primary electron energy or excitation energy is, 1. Adjustable throughout at least one SEM scan, and 2. These fluctuations can be modeled, for example, through simulation. 【0124】 Beam current and exposure time: These parameters determine the overall electron flux to the material. These primarily contribute to the signal-to-noise ratio (SNR) in the measurement. Below, a good SNR is assumed, i.e., a measurement dominated by photon shot noise, i.e., Poisson noise with a large average value. 【0125】 Sample tilt: The sample tilt affects the interaction volume because it introduces an asymmetric situation with respect to the surface normal of sample layer 181. Some electrons have a shorter effective path away from the sample than others. The tilt angle is assumed to be fixed during scanning. Therefore, its effect can be included in the simulation. 【0126】 Detector efficiency and / or detector geometry can be handled through simulation. 【0127】 Other examples may include other secondary radiation, such as backscattered or secondary electrons, and electromagnetic radiation below the X-ray regimen. 【0128】 The second class of effects originates from the sample composition. The main influencing factors are the atomic and molecular species present within the target region volume 20. These have a significant impact on the electron-X-ray cross-section. 【0129】 Material (atomic number): The shape of the target region volume 20 depends on the atomic number of the material. Heavier nuclei result in smaller interaction volumes compared to lighter elements; see Figure 6. 【0130】 Examples of inspection methods according to this disclosure are provided below. By performing approximations, the mathematical properties of the optimization problem are improved and the computational requirements are reduced. In some examples, the inspection method typically utilizes prior information or model-based assumptions about the material or material composition of the sample to be inspected. 【0131】 Monte Carlo simulation is a forward model that can be used to simulate the physics of electron microscopes. Software available for Monte Carlo simulation is well-established and known in this field. 【0132】 Simplification is necessary to adapt the results of Monte Carlo simulations to the post-processing steps of the inspection method. 【0133】 In general, the dependence of the collection angle on the solid angle Ω (see Figure 7) is not explained below. 【0134】 According to the first example, the scattering cross-section 【number】 This is simplified according to the example shown in Figure 14. In such a case, the cross-sectional area is 【number】 (3) Simplified as above, the spatially varying cross-sectional area σ is equal to the local material density ρ. i It is described as the sum of homogeneous cross-sectional areas σ of a single material scaled by ρ. This approximate model assumes that X-ray emission can only be considered for a single type of material. i (x) should be noted as potentially scaling characteristic X-ray emission to zero in regions where certain substances are absent. 【0135】 In this case, equation 1 represents a spatially varying operation such as convolution: 【number】 (4) It can be interpreted as follows. According to the present invention, equation 4 is the function ρ i The problem can be solved for the following: In the discretized setting of the first example, the optimization problem is 【number】 (5) This can be written as follows, where matrix A i ρ is the cross-sectional area of ​​the material. i This is the discretized linear operator of Equation 4 relating to the material density ρ, where ||·|| is the material density ρ i The norm for measuring the predicted spectral deviation resulting from the measured value I(E2=I) is "priors," where ρ is the perturbed material density. i This is a general term for the a priori conditions imposed upon something. 【0136】 Equation (5) can be interpreted as referring to a critical condition where excitation allows the photon to reach the detector. 【0137】 To calculate the target region volume 20 and the resulting intensity I(y,λ) with an appropriate SNR, a long residence time of, for example, several minutes may be required for a single point y, and this may be necessary in 3D (layer 18 i ,18 1+1 ,…) This process would have to be repeated for every FIB-SEM sample location in the data stack (e.g., on the order of 1 million samples). This effort can be mitigated by utilizing the symmetry of the sample geometry and / or scanning configuration, parallel processing, etc. 【0138】 Further examples of the inspection method according to the present invention are as follows: The starting point is a known theoretical design, provided, for example, by CAD files and material data, whose implementation deviates slightly from the complete model formulation. In this case, the inversion of Equation 1 may be directly calculated using important prior knowledge introduced at small actor deviations from the prior model in the actual sample. 【0139】 Furthermore, the following presents a more approximate scheme that can offer broader applicability by reducing the amount of prior knowledge required. 【0140】 An example of such further approximations is σ iEssentially, this calculation ignores material heterogeneity (Equation 4). In particular, in this example, the effect of material boundaries is ignored when determining the shape of the cross-sectional function. 【0141】 This model still follows Equation 4. However, the cross-sectional area function σ i The calculations required to determine this become independent of the sample at this point, and the cross-sectional shape becomes spatially invariant. 【0142】 The model in Equation 1 is therefore 【number】 (6) This represents a 3D spatial convolution for each spectral channel λ. This has the added advantage of enabling a fast FFT-based implementation of the 3D convolution. Furthermore, there is no registration requirement between the simulation / reconstruction and the experimental measurement. This model can be used in conjunction with the optimization scheme of Equation 5. 【0143】 Not only is the simulation preparation time reduced, but it is no longer necessary to know the sample composition in advance. The approximations are as follows: a) the interaction volume does not deform as it approaches the material boundary, and b) X-ray absorption between the generation site and the detector ignores the spatial structure of the sample, and instead assumes that the absorption cross-section of the photon-emitting material is maintained over the entire volume (excluding the free space between the sample and the detector, which is accurately modeled). 【0144】 Adverse effects may be partially mitigated by the prior information discussed elsewhere in this specification. 【0145】 In actual detectors, an ideal X-ray transition line that is infinitely narrow and easily distinguishable in the spectrum cannot be seen, and there is a limit to the linewidth that the detector can resolve. This is called the spectral response of the EDX sensor. The spectral response can also be described by a Gaussian distribution with varying dispersion for different detection energies. As a result of the limitations of the spectral response, nearby X-ray transition peaks may blur and blend together; this is a process called spectral convolution. 【0146】 The actually recorded spectral intensity is therefore given by using either Model Equation 4 or Equation 6 for the emitted X-ray emission I(y,λ), and furthermore, the bremsstrahlung component E, which has been previously ignored. bs (y,λ) is introduced here: I c (y,λ)=∫I(y,λ')r(λ,λ')dλ'+∫E bs (y,λ')r(λ,λ')dλ' (7) In the above equation, I c (y,λ) is the photon count recorded for energy λ at sample position y, and r(λ,λ') is the energy-dependent spectral response function of the sensor. The integral in Equation 7 is divided into two parts due to different physical processes, as can be seen later. 【0147】 The process of resolving peak mixing is known in this case as spectral deconvolution. 【0148】 One method allows for the direct inversion of equation 7. However, the data is usually very noisy, and the kernel attenuates high frequencies, making the inversion unstable and amplifying the noise. In practice, improper conditions prevent the achievement of high spectral resolution. Further examples of solving equation (7) according to the present invention are shown below. 【0149】 Integration into 3D optimization: Analytically combining Equation 7 with either Equation 4 or Equation 6 to derive an optimization method like the one in Equation 5 allows for a direct computation method for the post-processing deconvolution step, but this requires relatively high computational power. 【0150】 Inverse convolution using known materials I: Known emission line energies: Here, we assume that the elements present in the sample are known. In this case, L discrete numbers of X-ray transition lines λ' of emitted radiation. i In addition to =1…L, there is a continuous bremsstrahlung background that is ignored in the first step of the derivation. In that case, the first integral of Equation 7 becomes the sum of: 【0151】 【number】 (8) Measured count I c (y,λ), the spectral response function r of the sensor, and the X-ray transition line λ' i Since the coefficients I(y,λ') are known, i )=:e i It is possible to calculate (y), and the previous notation means that this quantity is the wavelength λ' i A specific Gaussian r(λ,λ') of the sensor response centered around i This emphasizes that ) = : r(λ) is simply a scalar coefficient. Equation 8 therefore describes a linear system at any sample location y. The linear system covers only the spectrum, i.e., the energy dimension. Equation 8 can be considered a fitting of the measured spectrum having a known set of emission peaks. This becomes more apparent by writing it in reduced notation and ignoring the spatial y dependence: 【0152】 【number】 (9) Since a typical spectrum has more λ samples than LX-line transition lines, the linear system is 【number】 (10) The solution must be written as the least squares method, and the bolded symbols in the above equation are the vector versions of the quantities introduced above. The matrix R contains the sampled sensor response function r in its columns. The value e i The fact that ≥0 is known should be further utilized. This can be done by adding an optimization constraint e≧0 and solving it using a quadratic programming solver instead of performing an unconstrained least-squares fit. 【0153】 Up to this point, we have ignored the continuous bremsstrahlung background (Equation 7, second term). Therefore, directly applying Equation 10 would attempt to fit the bremsstrahlung component to a discrete set of broadened X-ray transitions, resulting in a biased estimate. 【0154】 However, as further examples show, the optimized formulation of Equation 10 also makes it easy to include additional information. The bremsstrahlung component is modeled as a smooth function superimposed on the broadened emission line. The bremsstrahlung is expressed as a linear combination of K basis functions, which are convolved with a spectrally varying sensor response function r(λ,λ'): 【0155】 【number】 Here, b k φ is a coefficient, k (λ) is the basis function for the bremsstrahlung background. Thus, equation 10 can be written as follows: 【number】 (11) The matrix Φ contains discretized convolutional bremsstrahlung basis functions in its columns, and the vector b is the coefficient b k Collect the following. There are various ways to select a bremsstrahlung basis, and a standard basis is a polynomial basis. 【0156】 Another possibility is the truncated power law spectrum, which has been proposed as suitable for the low [keV] range. Such truncated power law spectra are well known to experts from their applications in astronomy and astrophysics. Another flexible option is to statistically reduce them by simulating a large number of Monte Carlo bremsstrahlung spectra and using PCA (principal component analysis). Such reduction can help dramatically reduce the amount of data processed. As a result, computation time can be significantly reduced. 【0157】 In one variation, an inverse convolution is performed using an unknown material, but given a superset of candidate elements. This setup is an extension of the methods already discussed. It is assumed that a superset of target elements for a particular application is known, but not all of such elements need to be present in the specific sample. However, there should be no missing elements from this superset of elements. In this case, there are examples of variable selection and fitting, for example, a denoising algorithm based on nonlinear total variation is known to experts. That is, the algorithm must select the correct elements and apply a fit as in Equation 11. A standard selection method is the regression reduction and selection algorithm known as LASSO. This can be implemented by L1 regularization. This optimization method is written as follows: 【0158】 【number】 (12) Here, γ is a tuning parameter, and choosing a larger value forces a sparser solution (i.e., a solution with more zero coefficients). An alternative solution is L0 regularization, however, this results in a computationally expensive exhaustive search procedure. 【0159】 Using known emission line energies and their relative ratios, deconvolution by the underlying material can be performed. According to the technique proposed in the previous paragraph, the algorithm selects X-ray emission lines in arbitrary ratios to fit the data. In reality, these ratios are not arbitrary and follow specific distributions that are difficult to quantify because the elements exist across different spatial structures as arbitrary mixtures. For this reason, some flexibility may be introduced into the algorithm to select lower-probability peak ratios if a better data fit can be achieved. In this case, a simple extension of the above scheme is possible, i.e., individual radiative peak responses belonging to common elements r can be extracted, for example, from scanning or simulating a homogeneous bulk material. i The use of linear combinations is excluded. Elemental response j is 【0160】 【number】 Assuming that it can be expressed as above, element j is indicated by a superscript in the above equation. 【0161】 Rather than imposing strict constraints, α i We can assume that this is a vector pointing in a direction in which the coefficients of the individual emission line responses (part of the coefficient vector e in Equation 10 and its variant) are likely to covariate. This can be done by encouraging solutions close to the expected subspace of the coefficients of variation. 【0162】 【number】 (13) Let the column include the direction of variation for each elemental coefficient. Here, M elements are represented by emission lines N1, N2, ..., NM. 【number】 is the corresponding coefficient. Π=P(P T P) -1 P T (14) is the orthogonal projection operator over the encoded subspace in matrix P. The optimization problem, equation 10, can be rewritten as follows, using an additional regularizer that penalizes solutions far from the expected subspace of variation: 【0163】 【number】 (15) Equation 15, in contrast to equations 10-12, forces specific elemental responses with different a priori, nearly known relative peak heights for each individual element. This is therefore more stable against accidental substitution of emission lines. Consider the case of a tungsten / silicon mixture. The X-ray lines of W M5-N6+7 (1773.60 [eV]) and Si K-L2+3 (1739.70 [eV]) are spectrally close to each other than the spectral response width of a typical SDD detector that can be used in X-ray detector 6 (e.g., 122 [eV] FWHM for Mn Kα). Therefore, a single peak is observed that may consist of Si, W, or both. Identifying the elements by observing a single peak is difficult. However, W has a further isolated group of peaks W M4-N2+M5-N3 (1380.00 [eV] and 1383.90 [eV]), whose presence and height indicate a) the presence of W and b) its approximate relative amount. Equation 15 utilizes this reasoning, but in Equation 10 and its variants, the peaks are fitted individually, which may lead to misidentification of the elements from which they originate. 【0164】 Standard Prior Information: The use of prior information (a priori conditions) is particularly advantageous. This refers to the term abbreviated as "priors" in Equation 5. Such priors may include smoothing priors such as the L2 norm for the gradient of the reconstruction function (material density), edge-preserving priors such as total variation known in denoising algorithms based on nonlinear total variation, and small coefficient priors such as Tychonov regularization. The reconstruction scheme according to the present invention is advantageous in using these or similar prior information. 【0165】 Different modes in a multimode electron microscope have different spatial resolutions and intensity characteristics. As discussed above, multimode imaging techniques are not limited to the analysis of spectrally resolved X-ray intensities. In some cases, intensity images based on backscattered electrons (BSE) are distinguished by their kinetic energy from secondary electrons (SE). The BSE intensity contrast is governed by the number of protons of the elements in the sample (Z contrast). The contrast is further dependent on the beam energy. 【0166】 Therefore, edges in BSE intensity images (and also in SE images, although less suitable) can be used as an indicator of chemical contrast. 【0167】 Furthermore, BSE has a small interaction volume in all three spatial dimensions, providing higher spatial resolution. 【0168】 Simultaneous BSE images may provide appropriate additional information that can enhance the robustness of the proposed super-resolution multimode imaging scheme. SE / BSE images can be used as guide images for image cleaning. SE / BSE images may be proposed as guide images for cleaning EDX material maps (using joint bidirectional filtering) and improving their resolution. Similarly, in STEM (scanning transmission electromagnetism), HAADF (High-angle annular dark-field imaging) has good material contrast and high resolution, and therefore it has been proposed to use it for cleaning spectral images, in this case by non-local mean filtering. 【0169】 Because BSE images offer higher resolution than typical EDX spectral channels, they can be used as prior information for edge locations. This can be modeled as an additional prior within the reconstruction framework. 【0170】 Such priors can be used in combination with the optimization scheme of Equation 5, for example, in the spatial convolution model of Equation 6. In that case, the term "priors" will be replaced below. 【number】 (16) This formulation modifies the edge term of the prior art model introduced in the context of segmentation. The function g is the guide image I g For example, it has a low value at edge positions, such as those determined by the BSE image. For example, g(x)=exp(-a|∇I|) where a=10 and b=0,55. x | b ) is used. This has high values ​​in other image regions. As a result, the function ρ i The jump in is preferably facilitated in the region where the value of g(x) is low, while in other regions, the constant function ρ i This is exacerbated. It would be useful to have a user-adjustable regularization parameter that multiplies the prior term (Equation 16). 【0171】 In conclusion, techniques are described that improve volumetric spatial resolution by using knowledge of the interaction volume between the electron microscope and the material sample—via spatial deconvolution—and further improve spectral deconvolution, i.e., the identification of material emission lines from measured and recorded EDX spectra. One technique discussed above is interleaved EDX imaging and optimization fitting (particularly compare equations (5) and (10) above) to obtain structural 3D information. Furthermore, knowledge of a material library of substances that may be present in the sample is utilized. This knowledge, in particular, allows for the acquisition of stable solutions.

Claims

[Claim 1] Using focused ion beam (FIB) etching and X-ray detection, a semiconductor sample (2) having a sample surface is subjected to layer (18 i ) from layer (18 i+1 A method for inspecting each layer of a sample (2), wherein the sample (2) has structures (23, 38) with different elemental compositions, and the method comprises the following steps, namely, The layer to be inspected (18) of the semiconductor sample (2) 1 The steps include preparing the sample by etching the initial sample surface using a focused ion beam (FIB), The prepared layer (18) of the sample (2) i The steps include aligning the surface area of ​​the target region volume (20) with the objective field of view (21) of a scanning electron microscope (SEM), The steps include adjusting the electron energy of the electron beam (8) of the SEM, The steps include: probing the target area volume (20) within the objective field of view (21) with the electron beam (8); The steps include detecting X-rays (7) emitted from the aligned target region volume (20), The detection signal acquired during the detection step is post-processed in order to spatially deconvolve the detection signal with structural data derived from the sample structure within the target region volume (20). The steps from the "prepare" step to the "post-processing" step are repeated until the inspection of each layer of the superimposed target volume of the sample (2) is completed, It has, The FIB source (3) that generates the focused ion beam etches the sample (2) in an etching plane (17) that includes an angle of 20° to 60° with respect to the initial bulk sample surface plane (15). method. [Claim 2] The method according to claim 1, wherein X-ray detection is performed in a wavelength-dependent manner. [Claim 3] The method according to claim 1 or 2, wherein the post-processing takes into account the volume interaction between the electron beam (8) and the sample (2) in the target region volume (20). [Claim 4] The method according to claim 1, wherein the post-processing takes into account elemental mapping of elements in the probed sample (2) in the target region volume (20). [Claim 5] The method according to claim 2, wherein the post-processing includes spectral deconvolution of the detected X-rays (7). [Claim 6] The method according to claim 1, wherein the post-processing includes a Monte Carlo simulation of the interaction between probe electrons and the material of the sample. [Claim 7] The method according to claim 1, wherein the post-processing includes inputting geometry or other priori conditions from further measurements. [Claim 8] During the aforementioned inspection, the following steps are taken, namely: For each value of n, a kernel value K represents the behavior of the probing beam in the bulk sample for a given beam parameter value. n The steps include defining a point spread function having, The steps include: defining a spatial variable V that expresses the physical properties of the sample as a function of its position in the bulk; For each value of n, Q n = K n * K that forms a V n The steps include defining an imaging quantity having a value Qn which is a multidimensional convolution of and V, For each value of n, M n and Q n the minimum divergence between min D (M) n ∥K n *V) A step in which the above formula determines the value K by calculation, wherein the above formula is equal to the value K n The steps involved in solving and determining V while imposing constraints on it, The method according to claim 1, wherein the following is performed. [Claim 9] The aforementioned value K n The aforementioned constraints on are: At least value K n A set of computational simulations, At least value K n Empirical decision of the set, At least value K n Modeling of a point spread function as a parameterized function with a limited number of modeling parameters that can be estimated based on a set of parameters, A theoretically possible value K that is determined to be physically meaningless. n A restriction on the logical solution space such that it is discarded. Value K n By applying extrapolation and / or interpolation to the first set of values, the value K n Interference of the second set, The method according to claim 8, derived using at least one method selected from the group including the following. [Claim 10] Focused ion beam (FIB) source (3), Scanning electron microscope (SEM) (4) and An X-ray detection device (6) for detecting X-rays (7) emitted from the sample (2) and generated by the probe electrons of the SEM (4), A computer (9) for performing the post-processing of the detection signal acquired by the X-ray detection device (6), Equipped with, The inspection apparatus is configured such that the FIB source (3) etches the sample (2) on an etching plane (17) that includes an angle of 20° to 60° with respect to the initial bulk sample surface plane (15). An inspection apparatus (1) for performing the method according to claim 1. [Claim 11] The inspection apparatus according to claim 10, wherein the X-ray detection device (6) includes an X-ray spectrometer (10). [Claim 12] The inspection apparatus according to claim 10 or 11, wherein the SEM (4) is positioned to probe the sample (2) at an angle of less than 90° from the initial bulk sample surface plane (15). [Claim 13] The inspection apparatus according to claim 10, wherein the X-ray detection device (6) is arranged to detect the X-rays (7) at an angle of less than 90° when measured from the surface plane (15) of the initial bulk sample.

Images