Cross-sectional imaging method for examining volume in a wafer

Abstract

The invention relates to a dual-beam apparatus and three-dimensional circuit pattern inspection techniques by cross-sectional measurement of an inspection volume extending over large depths below the surface of a semiconductor wafer of more than 1 pm and more particularly to a method, computer program product and apparatus for generating 3D volume image data of deep inspection volumes inside a wafer without removing a sample from the wafer. The invention also relates to 3D volume image generation, cross-sectional image alignment methods for three-dimensional circuit pattern inspection with a dual-beam apparatus.

Description

Technical Field

[0001] This invention relates to a three-dimensional circuit pattern inspection and measurement technique by cross-section measurement of integrated circuits. More specifically, the invention relates to a three-dimensional circuit pattern inspection technique by cross-sectioning an inspection volume at a measurement location on a semiconductor wafer including integrated circuits, and even more specifically, to a method, computer program product, and corresponding semiconductor inspection equipment for obtaining a 3D volumetric image of the inspection volume at the measurement location on a semiconductor wafer. The method involves milling a cross-sectional surface into the inspection volume of the wafer at an inclined angle and imaging the inclined cross-sectional surface using a charged particle imaging microscope. The method, computer program product, and equipment can be used for quantitative metrology, defect detection, process monitoring, defect review, and inspection of integrated circuits within semiconductor wafers. Background Technology

[0002] Semiconductor structures are among the most delicate man-made structures, and contain only a very small number of defects. These rare defects are the features that defect detection, defect review, or quantitative metrology equipment is looking for. Semiconductor structures are fabricated based on existing knowledge. Semiconductor structures are made from sequences of layers parallel to the substrate. For example, in logic-type samples, metal lines extend parallel to the metal layers or in HAR (high aspect ratio) structures, and metal vias extend perpendicular to the metal layers. The angles between metal lines in different layers are either 0° or 90°. On the other hand, for VNAND-type structures, their cross-sections are known to be circular on average.

[0003] In integrated circuit manufacturing, feature sizes are becoming increasingly smaller. Current minimum feature sizes, or critical sizes, are below 10 nm, such as 7 nm or 5 nm, and are approaching below 3 nm in the near future. Therefore, measuring the edge shape of a pattern and determining the size of features or the roughness of line edges with high precision becomes challenging. The edge shape of a pattern or the roughness of a line is influenced by a variety of factors. Generally, the edge shape of a line or pattern can be affected by the properties of the material itself, photolithography exposure, or any other process steps involved (such as etching, deposition, or implantation). The measurement resolution of charged particle systems is typically limited by the sampling grating of a single image point or the residence time of each pixel on the sample, as well as the diameter of the charged particle beam. The sampling grating resolution can be set within the imaging system and can be adapted to the diameter of the charged particle beam on the sample. Typical grating resolutions are 2 nm or lower, but the grating resolution limit can be reduced without physical constraints. The charged particle beam diameter has a finite size, depending on the charged particle beam operating conditions and the lens. Beam resolution is limited to approximately half the beam diameter. Resolution can be below 2 nm, for example, even below 1 nm.

[0004] As the feature size of integrated semiconductor circuits shrinks and the resolution requirements of charged particle imaging systems increase, the inspection and 3D analysis of integrated semiconductor circuits within wafers become increasingly challenging. A semiconductor wafer, with a diameter of 300 mm, consists of several locations (so-called dies), each containing at least one integrated circuit pattern, such as for memory chips or processor chips. The semiconductor wafer undergoes approximately 1000 process steps, and within the wafer, approximately 100 or more parallel layers are formed, including transistor layers, line-interval layers, and interconnect layers, forming a 3D array of memory cells in a memory device.

[0005] A common method for generating 3D tomographic data from nanoscale semiconductor samples is the so-called slicing and imaging approach, detailed, for example, using a two-beam apparatus. In such an apparatus, two particle optics systems are arranged at an angle. The first particle optics system can be a scanning electron microscope (SEM). The second particle optics system can be a focused ion beam (FIB) system using, for example, gallium (Ga) ions. The gallium ion focused ion beam (FIB) is used to slice layers at the edges of the semiconductor sample piece by piece, and each section is imaged using a scanning electron microscope (SEM). The two particle optics systems can be oriented perpendicularly or at an angle between 45° and 90°. Figure 1 A schematic diagram of the slicing and imaging method is shown: using an FIB optical column 50 with a focused ion beam 51 in the z-direction and scanning in the yz plane, a thin layer is removed from a section passing through the semiconductor sample 10 to reveal a new front surface 52 as the cross-sectional surface 52. In the next step, for example, a SEM 40 is used for scanning imaging of the front surface of the cross-sectional surface 52. In this example, the SEM optical axis 42 is oriented parallel to the x-direction, and an image is generated by scanning the electron beam 44 along the scanning imaging line 46 in the yz plane. After performing a raster scan of the cross-sectional surface 52, a cross-sectional image slice 100.1 is formed. By repeating this method of FIB milling and SEM imaging through, for example, the front cross-sectional surfaces 53 and 54, cross-sectional image slices 100.2 and 100.3 with a distance d are obtained. Finally, a sequence of 2D cross-sectional images 1000 through the sample at different depths is obtained. The distance d between two subsequent image slices can be between 1 nm and tens of nm, for example, 30 nm. Based on the sequence of these 1000 2D cross-sectional images, a 3D image of the integrated semiconductor structure can be reconstructed.

[0006] Figure 1A slicing and imaging method is illustrated in an example of a bulk integrated semiconductor sample 10, which is removed from a semiconductor wafer using known techniques. A reference is formed on top of the bulk sample. Deriving the lateral position of each slice and the distance between layers is a common method using the so-called reference. US 9,633,819 B2 discloses an alignment method based on a guide structure (“reference”) exposed on top of the sample. US 7,348,556 describes alignment marks on a surface for determining three-dimensional surface roughness from a sequence of consecutive image slices.

[0007] Common slicing and imaging methods in the prior art are not suitable for inspecting the volume within a wafer. Using common slicing and imaging methods requires removing or extracting the sample from the semiconductor wafer before the slicing and imaging methods can be performed to obtain a 3D volumetric image of the sample. Therefore, the objective is to provide a slicing and imaging method suitable for generating 3D volumetric images of the volume within a wafer without removing the sample.

[0008] US 7,438,556 discloses a method for determining line edge or surface roughness using a dual-beam FIB / SEM tool. A series of cross-sectional surfaces are generated by FIB milling perpendicular to the sample surface. A sequence of cross-sections is generated at slice distances of approximately 10 nm or greater. A reference is applied to the top surface to determine the lateral position of the cross-sections. The SEM is tilted to the normal of the sample surface and used to form a sequence of cross-sectional images. Based on each cross-sectional image, the critical dimension (CD) of the feature in a direction parallel to the wafer surface is determined. The line edge roughness of the feature is determined from the CD, wherein the edge of the feature is determined with reference to a reference in each cross-sectional image. The position of the edge perpendicular to the sample surface is determined with reference to a known angle of the SEM and the reference. Therefore, it is not possible to accurately determine the depth of the feature. Thus, the proposed method is limited to 1D measurement of semiconductor features parallel to the sample surface. In particular, the proposed method does not provide a means for accurately determining the depth of semiconductor features below the wafer surface.

[0009] To analyze repetitive and deep semiconductor features, such as the HAR structure of memory devices, slicing and imaging under a single wedge-shaped cutting geometry with a small tilt angle using a FIB beam for milling has been proposed. For example, US 9,466,537 discloses a method for inspecting semiconductor devices with a molding layer. An inclined cross-sectional surface is formed through the molding layer by milling into it at an inclined angle. To obtain the desired tilt angle, the stage holding the sample under study is tilted between the milling and imaging steps. The obtained images of the cut or cross-sectional surface are analyzed, and the center position of the semiconductor feature is derived, for example, relative to a selected semiconductor feature used as a reference. It is claimed that process deviations in the manufacturing process can be derived from this analysis. However, it has been found that the analysis of a single cut or cross-section, and using the feature of interest as a reference, has limited accuracy and cannot provide the information needed to monitor the manufacturing process. Furthermore, the requirement to tilt the stage between milling and imaging is impractical for high-throughput inspection tasks.

[0010] US 10,026,590 discloses a similar method for examining features of interest by milling a single cross-sectional surface into a sample at a grazing angle, where the virtual feature is composed of cross-sections of different features of interest at varying depths. The depth is determined based on the lateral distance from the feature's cross-section to the edge of the groove. The depth resolution of the virtual feature can be improved through additional milling operations. An alignment reference perpendicular to the FIB beam is proposed to monitor the milling operation. However, it has proven impossible to perform milling at very small grazing angles, or this leads to inaccurate results in the milling operation, and the accuracy of deriving the virtual feature from a small number of cross-sections is limited, failing to provide the information needed to monitor the manufacturing process. Furthermore, determining the depth from the edge of the groove or notch is inaccurate, and applying the reference at a surface perpendicular to the FIB beam is a difficult and time-consuming process.

[0011] As the thickness of each layer in a multilayer stack of semiconductor devices decreases, the tilt angle required to mill the cross-sectional surface into the inspection volume of the wafer becomes increasingly smaller, for example, below 5° or even below 3°. One such method is described in US 9941096 BB. However, such small tilt angles are impractical for real-world mounting. Milling the cross-sectional surface into a deep semiconductor structure of approximately 5 μm or more in a multilayer structure at a very small angle, for example below 15°, requires milling very large surfaces over a considerable length in the direction of the milling beam using FIB, extending beyond 40 µm or even 100 µm to obtain deeper structures. Milling such large surfaces is time-consuming, and surface quality degrades sharply with increasing layer depth. Furthermore, large cross-sectional surfaces exceed the field of view of typical high-resolution charged particle imaging devices between 10 µm and 20 µm, and imaging large cross-sectional surfaces requires image stitching. US10184790 proposes an image stitching method to form a 2D image of a tilted surface, wherein a series of SEM images are obtained by lateral movement of the sample, and the images are stitched together to form a single 2D image of the tilted surface. The depth is measured at the edge of a reference trench, visible in an SEM image. Therefore, the depth determination is inaccurate. The obtained 2D information and limited accuracy are insufficient to meet current requirements. Therefore, the object of this invention is to provide a wafer inspection tool and method for generating 3D volumetric images, including high-quality imaging of deep structures.

[0012] US 2009 296073 describes a method for analyzing semiconductor features by milling a surface substantially parallel to the wafer surface. However, it is unclear how to obtain a surface with sufficient accuracy that is parallel to the wafer surface or even at very small angles of less than 5° or 10° by using FIB milling without removing the sample from the wafer.

[0013] Recent advancements have demanded higher accuracy at relatively high throughput, including the precise determination of the depth of semiconductor features beneath the wafer surface. Therefore, the objective is to provide a slicing and imaging method suitable for generating 3D volumetric images of a wafer inspection volume without removing the sample. The depth of features within the wafer inspection volume must be determined with high accuracy. Therefore, the objective is to provide a slicing and imaging method suitable for generating 3D volumetric images of a wafer volume that has high resolution in depth and does not require sample removal. Another objective of the present invention is to provide a wafer defect inspection apparatus and method capable of inspecting defects within the wafer inspection volume without removing a sample from the wafer.

[0014] Recently, the trend towards further integration of semiconductor circuits has led to taller stacks of alternating layers forming within or on silicon wafers. Current memory chips comprise up to one hundred or more different layers, such as 92 layers. Stacks of approximately 100 layers reach thicknesses exceeding 6 μm, and current and future stacks will reach heights of 10 μm. As stack height increases, imaging deep structures within the wafer becomes increasingly challenging for inspection volume. Therefore, the objective is to provide a wafer inspection tool and method for generating 3D volumetric images, including imaging deep structures without damaging the wafer.

[0015] In addition to increasing depth, the thickness of each layer becomes increasingly smaller. For inspection tasks, it is desirable to obtain cross-sectional images through multiple HAR structures (such as word lines or isolation layers) within a single layer. Therefore, the task is to provide a wafer inspection tool and method for generating 3D volumetric images, which generates cross-sections through HAR structures in a single layer without removing the sample from the wafer.

[0016] As depth increases, imaging the cross-sectional surface of an inspection volume with large depth extension using charged particle imaging beams becomes increasingly challenging. Therefore, the objective is to provide a dual-beam apparatus and method for generating 3D volumetric images for inspecting inspection volumes with large depth extensions within a wafer.

[0017] Typical wafer inspection tasks for in-line inspection require very high throughput. Therefore, another objective of this invention is to provide 3D volumetric inspection in semiconductor devices manufactured in-wafer with high throughput.

[0018] These tasks are solved by the present invention as described by the examples given in the embodiments of the invention. Summary of the Invention

[0019] The object of this invention is to provide a method for 3D inspection of an inspection volume in a wafer using a dual-beam apparatus, and a dual-beam apparatus configured for inspecting an inspection volume in a wafer without needing to extract or remove the inspection volume from the wafer. Another object of this invention is to provide a computer program product having program code for performing a method for 3D inspection of an inspection volume in a wafer using a dual-beam apparatus. According to an embodiment, the 3D inspection of the inspection volume in the wafer is performed using a wedge-cut geometry. The wedge-cut geometry allows for the study of inspection volumes with a lateral extension of approximately 10 μm-15 μm and a depth extension of 10 μm or more within the wafer without damaging the wafer or extracting the sample volume from the wafer. The slicing and imaging method under the wedge-cut geometry can generate a 3D representation or 3D volumetric image of the entire inspection volume of the wafer at a high lateral resolution of less than 5 nm, preferably less than 2 nm, and even more preferably less than 1 nm. For the entire inspection volume, a 3D volumetric image for 3D inspection is provided.

[0020] This invention provides a dual-beam apparatus and a 3D measurement method that generates 3D volumetric image data through cross-sectional measurement of integrated circuits without removing a sample from the wafer. Specifically, a dual-beam apparatus and a wafer inspection method are provided, featuring high resolution of the inspection volume extending at a large depth beneath the wafer surface. This invention provides a 3D measurement method for performing three-dimensional circuit pattern inspection of the inspection volume inside a wafer without removing a sample from the wafer. More specifically, this invention relates to a three-dimensional circuit pattern inspection technique by cross-sectional measurement of the inspection volume at a measurement location of a semiconductor wafer including integrated circuits, and more specifically, to a method, computer program product, and corresponding semiconductor inspection apparatus for obtaining a 3D volumetric image of the inspection volume at a measurement location of a semiconductor wafer without removing a sample from the wafer. The method uses second cross-sectional image features to determine the depth of a first cross-sectional image feature. Multiple second cross-sectional image features correspond to structures in layers within an integrated circuit, or generally correspond to structures of known or predetermined depths. In the example, the depth determination of the first cross-section is a relative depth determination with respect to multiple second cross-sectional image features. After acquiring and aligning at least one cross-sectional image slice and determining the depth of the first cross-sectional image features, the inspection volume at the measurement location of the wafer is evaluated, for example, the wafer's manufacturing errors. The manufacturing errors are analyzed, and, for example, a failure analysis of the manufactured wafer is performed. In an example, the manufacturing errors are analyzed, and, for example, specific manufacturing process steps used to manufacture the wafer are improved. This method, computer program product, and apparatus can be used for quantitative metrology, defect detection, process monitoring, defect inspection, and examination of integrated circuits within semiconductor wafers.

[0021] According to an embodiment of the present invention, a method for inspecting at least a first inspection volume of a wafer using a first dual-beam apparatus includes the step of loading a wafer onto a wafer support stage in the dual-beam apparatus, the dual-beam apparatus including at least a FIB pillar and a charged particle imaging device, wherein a first optical axis of the FIB pillar forms an angle GF with the surface of the wafer support stage, and a second optical axis of the charged particle imaging device forms an angle GE with the normal to the surface of the wafer support stage, the first and second optical axes intersecting. The wafer inspection method further includes the step of moving the wafer support stage to align a first measurement position on the wafer with the intersection of the dual-beam apparatus and milling a first cross-sectional surface in the first inspection volume using the FIB pillar at an angle GF. The wafer inspection method further includes the step of generating a first cross-sectional image slice of the first cross-sectional surface using the charged particle imaging device. The wafer inspection method further includes the step of obtaining performance indicators of a plurality of first semiconductor features in the first inspection volume, the step including analyzing at least one first cross-sectional image slice using prior information about the plurality of first semiconductor features. In the example, the first semiconductor feature is one of a via, a HAR structure, or a HAR channel, and the analysis step includes an image processing step to extract multiple first cross-sectional image features representing a cross section of multiple first semiconductor features at an angle GF, and the image processing includes at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation. In the example, the step of obtaining a performance metric further includes calculating at least one descriptive parameter of the first semiconductor feature from at least one of the multiple first cross-sectional image features, wherein the descriptive parameter is one of size, diameter, angle, area, shape, or volume. In the example, the step of obtaining a performance metric further includes calculating one of the average or statistical deviation of at least one descriptive parameter of the multiple first semiconductor features. In the example, the analysis step further includes generating a depth map Z(x, y) of the first cross-sectional image slice. In the example, the step of generating the depth map further includes identifying at least two second cross-sectional image features in the first cross-sectional image slice, and determining the depth map Z(x, y) from the lateral positions of the at least two second cross-sectional image features. The at least two second cross-sectional image features may represent integrated semiconductor structures or features at different depths within the examined volume. Multiple first semiconductor features may extend in a direction perpendicular to the wafer surface, and second cross-sectional image features may include cross-sections of semiconductor structures extending in a direction parallel to the wafer surface. In the example, the second cross-sectional image features include at least a cross-section of one of an isolation line or layer, a metal line or layer, or a semiconductor line or layer. The step of obtaining performance metrics may further include deriving a tilt angle deviation of the multiple first semiconductor features from their lateral positions, the tilt angle deviation being an angle of the first semiconductor feature relative to an axis perpendicular to the wafer surface.

[0022] In the example, the wafer inspection method further includes the step of deriving a 3D representation of a first plurality of semiconductor features from a plurality of first cross-sectional image features arranged with a tilt angle GF and a depth map Z(x,y).

[0023] In the example, the analysis step further includes comparing a first cross-sectional image slice with a 2D digital image slice of an inspection volume of a reference wafer or die. The 2D digital image slice may be a cross-sectional image slice obtained in previous measurements of the reference wafer or die and stored in memory. In the example, the 2D digital image slice is a virtual cross-sectional image slice, wherein the virtual cross-sectional image slice is generated from 3D volumetric image data stored in memory. In the example, the 3D volumetric image data is obtained in previous slice and image measurements of an inspection volume of the reference wafer or die and stored in memory. In the example, the previous slice and image measurements are performed using a second dual-beam apparatus including at least a second FIB column and a second charged particle imaging device. In the example, the previous slice and image measurements are performed by a first dual-beam apparatus.

[0024] In the example, the methods for wafer inspection also include:

[0025] --Move the wafer support stage so that the second measurement position of the wafer coincides with the intersection of the first dual-beam device.

[0026] --The second section surface is milled in the second inspection volume using a FIB column at an angle GF.

[0027] --Generate a second cross-sectional image slice of the second cross-sectional surface using a charged particle imaging device.

[0028] --Performance metrics of multiple first semiconductor features in first and second inspection volumes are obtained by analyzing first and second cross-sectional image slices with prior information about multiple first semiconductor features.

[0029] In the example, the step of rotating the wafer support stage relative to an axis perpendicular to the wafer support surface and deriving tilt angle deviations of multiple first semiconductor features between milling of a first cross-sectional surface in a first inspection volume and milling of a second cross-sectional surface in a second inspection volume may include analysis of first and second cross-sectional image slices.

[0030] The present invention also provides a wafer defect inspection apparatus, wherein the wafer defect inspection apparatus is configured to inspect an inspection volume in a wafer, wherein the inspection volume is not extracted from the wafer, the apparatus comprising:

[0031] --Focused ion beam (FIB) column, configured to mill and expose at least a first cross-sectional surface through a first inspection volume in the wafer at an angle GF;

[0032] --A charged particle imaging device is configured to image the at least first cross-sectional surface to form a first cross-sectional image slice;

[0033] --An image processing unit with software code installed is configured to determine multiple cross-sectional image features in at least a first cross-sectional image slice and to determine the depth of multiple cross-sectional features within an inspection volume, wherein the multiple cross-sectional image features are cross-sections of a semiconductor structure with a tilt angle GF within the inspection volume;

[0034] --A defect detection unit is configured to determine deviations from predetermined characteristics of a semiconductor structure within an inspection volume from multiple cross-sectional image features. In the example, an image processing unit with software code is further configured to calculate 3D representations of multiple first semiconductor structures from multiple first cross-sectional image features arranged at an angle GF. In the example, the wafer defect inspection apparatus also includes a memory for storing prior information. The memory device may be part of a control unit or part of the image processing unit of the control unit.

[0035] According to an embodiment of the present invention, a method for inspecting an inspection volume in a wafer using a dual-beam apparatus includes the following steps: loading the wafer onto a wafer support stage in a dual charged particle beam tool, and moving the wafer support stage so that a first measurement position of the wafer coincides with the intersection of the optical axes of the FIB and the charged particle imaging apparatus. The method further includes the step of obtaining a sequence or multiple N cross-sectional image slices, wherein the multiple N cross-sectional image slices include at least a first cross-sectional image slice and a second cross-sectional image slice in the inspection volume. The number N of cross-sectional image slices is at least N = 10, preferably N > 100, for example, N is about 300 or more, for example, N > 1000. The multiple cross-sectional image slices are obtained by subsequently exposing multiple N cross-sectional surfaces in the inspection volume by milling approximately into the inspection volume with an FIB pillar at angle GF and imaging each of the multiple N cross-sectional surfaces with the charged particle imaging apparatus. The multiple cross-sectional surfaces include at least first and second cross-sectional surfaces. The multiple cross-sectional image slices include at least first and second cross-sectional image slices. The method further includes determining at least one first cross-sectional image feature in the first and second cross-sectional image slices, and determining at least one second cross-sectional image feature in the first and second cross-sectional image slices; and determining the depth of at least one first cross-sectional image feature in the first cross-sectional image slice from the lateral position of at least one second cross-sectional image feature in the first cross-sectional image slice. In the example, the wafer is not moved during the step of obtaining a sequence of N cross-sectional image slices in the inspection volume.

[0036] Furthermore, the method for inspecting an inspection volume in a wafer may optionally include the step of performing a mutual lateral alignment of first and second cross-sectional image slices with at least one common cross-sectional image feature. In one example, the common cross-sectional image feature is a first cross-sectional image feature present in the first and second cross-sectional image slices. In another example, the common cross-sectional image feature is an image segment of an alignment feature provided or present near the inspection volume. The mutual lateral image alignment step may be further improved by subtracting or numerically compensating for image distortion deviations between at least the first and second cross-sectional image slices.

[0037] According to an embodiment of the method for inspecting the volume in a wafer, the depth of a first cross-sectional image feature in a first cross-sectional image slice is determined, and a 3D volumetric image of the inspected volume is derived using an algorithm employing several steps. The first step is cross-sectional image feature detection and classification, which detects cross-sectional image features and classifies them into first cross-sectional image features and second cross-sectional image features. The second step generates a depth map from the second cross-sectional image features in multiple cross-sectional image slices. The third step determines the depth of each first cross-sectional image feature based on the depth map. The fourth step generates a 3D volumetric image of the semiconductor structure of interest in the inspected volume. The fifth step derives the integrated circuit features or characteristics of the semiconductor structure of interest in the 3D volumetric image. The sixth step derives the defects of the integrated circuit features or characteristics of the semiconductor structure of interest in the 3D volumetric image. In this example of the method for inspecting the volume in a wafer, the tilt angle GF of the FIB beam used for milling multiple cross-sectional surfaces is adjustable. As the tilt angle GF changes, the coverage depth range of the inspected volume also changes. In this example of the method for inspecting the volume in a wafer, the distance between subsequent cross-sectional surfaces is adjustable. In this example, the distance between multiple cross-sectional surfaces is adjusted to be different for at least some distances between the cross-sectional image surfaces. By adjusting the distance, the throughput and resolution of the 3D volumetric image of the inspected volume can be fully or partially adjusted to meet the needs of the wafer inspection task.

[0038] According to an embodiment of a method for inspecting an inspection volume in a wafer, the step of determining a depth map or depth includes determining the lateral position of a second cross-sectional image feature. A depth map or depth of the first cross-sectional image feature in the first cross-sectional image slice is determined from the lateral difference between a first position of the second cross-sectional image feature in a first cross-sectional image slice and a second position of the second cross-sectional image feature in a second cross-sectional image slice. The depth determination of the second cross-sectional image feature does not utilize the first cross-sectional image feature, which represents, for example, a HAR channel of a memory device. Errors in the fabrication of the first cross-sectional image feature and the semiconductor structure represented by the first cross-sectional image feature, such as the entire three-dimensional HAR structure within the inspection volume, are determined with high precision. In an example of a method for inspecting an inspection volume in a wafer, tilt or wobble relative to the wafer surface, or alignment errors in HAR structure fabrication, or memory stack orientation, are determined with high precision and low ambiguity. In an example of a method for inspecting an inspection volume in a wafer, the HAR structure is studied and compared at multiple depths throughout the inspection volume of the entire wafer.

[0039] According to an embodiment of the method for inspecting an inspection volume in a wafer, the positions of two second cross-sectional image features are determined in a first cross-sectional image slice, wherein each second cross-sectional image feature represents an integrated semiconductor structure at a predetermined depth within the inspection volume. In the example, the step of determining the depth of the first cross-sectional image feature in the first cross-sectional image slice, or the depth map of the first cross-sectional image slice, is determined from the lateral positions of the two second cross-sectional image features.

[0040] In embodiments of a method for inspecting an inspection volume in a wafer, the method includes a further step of forming at least one alignment feature. The at least one alignment feature is formed or exposed near the inspection volume. The alignment feature is configured for mutual lateral alignment of a plurality of cross-sectional image slices, including first and second cross-sectional image slices. In embodiments, the alignment feature is fabricated above the inspection volume and configured to determine the location of an edge formed by the intersection of a cross-sectional surface and a wafer surface. In an example, the alignment feature is provided or exposed at one or more depths in an additional trench or trench to enable alignment at different imaging depths in the inspection volume outside the wafer surface. Furthermore, the method for inspecting an inspection volume in a wafer may optionally include a step of performing mutual lateral alignment of the first and second cross-sectional image slices with at least one common cross-sectional image feature. In an example, the common cross-sectional image feature is a first cross-sectional image feature present in the first and second cross-sectional image slices. In another example, the common cross-sectional image feature is an image segment of an alignment feature provided or present near the inspection volume. The mutual lateral image alignment step may be further improved by subtracting or numerically compensating for image distortion deviations between at least the first and second cross-sectional image slices.

[0041] According to an embodiment of the invention, the sequence of N cross-sectional image slices includes at least a first cross-sectional image slice and a second cross-sectional image slice for checking the volume, wherein the first cross-sectional image surface is milled with a greater extension in a direction perpendicular to the FIB beam compared to the second cross-sectional surface, such that after the second cross-sectional surface is formed, a parallel surface segment of the first cross-sectional surface is retained. At least one alignment feature may be formed on the parallel surface segment of the first cross-sectional surface and near the second cross-sectional surface for a first mutual lateral alignment of the first and second cross-sectional image slices.

[0042] According to embodiments of the present invention, precise alignment of multiple cross-sectional image slices of an inspection volume below the wafer surface is obtained. A first coarse alignment of the first and second cross-sectional image slices is performed, for example, by additional alignment features formed near the inspection volume. Using the first alignment, a mapping of cross-sectional image features in the first and second cross-sectional image slices is obtained. A second precise alignment of the first and second cross-sectional image slices utilizes cross-sectional image features of the semiconductor structure within the inspection volume below the wafer surface, and achieves mutual positional accuracy of less than 5 nm, 3 nm, or even less than 2 nm. In an example, the second precise alignment includes calculating a first displacement of a first cross-sectional image feature between the first and second cross-sectional image slices. Calculate the second displacement of the second cross-sectional image features between the first cross-sectional image slice and the second cross-sectional image slice. Determine the distance d between the first and second cross-sectional image slices, and determine the mutual lateral displacement vector between the first and second cross-sectional image slices. Therefore, this two-step alignment process avoids registration errors and enables precise alignment.

[0043] In an example of a method for examining an examination volume in a wafer, the step of obtaining a sequence of N cross-sectional image slices includes acquiring multiple laterally displaced image segments and stitching the multiple laterally displaced image segments to form cross-sectional image slices. In the example, acquiring multiple image segments in a charged particle imaging microscope involves changing the focal position of the charged particle imaging device for at least a subset of the multiple image segments. Therefore, high-resolution imaging is maintained even if the examination volume has a large extension in a direction perpendicular to the wafer surface.

[0044] In an embodiment of the method for inspecting an inspection volume within a wafer, a fan-beam tomography (FMT) method is applied. In this embodiment, the multiple N cross-sectional image slices or a sequence of N cross-sectional image slices within the inspection volume include: a focused ion beam scanning a FIB column in a first direction by a scanning unit to expose a first cross-sectional surface within the inspection volume by FIB milling; the focused ion beam tilting in a second direction perpendicular to the first direction by the scanning unit; and the focused ion beam scanning in the first direction by the scanning unit to expose a second cross-sectional surface within the inspection volume by FIB milling, such that the first and second cross-sectional surfaces form different angles with the wafer surface, approximately a tilt angle GF. Using the fan-beam tomography method, the multiple cross-sectional surfaces form different angles with the wafer surface, having an angular extension GZ centered on the tilt angle GF. In the fan-beam tomography method, the wafer does not move during the step of obtaining multiple cross-sectional image slices within the inspection volume.

[0045] In embodiments, the cross-sectional imaging method in a wedge-cut geometry and the method for inspecting the inspection volume in a wafer include distortion compensation. Image distortion in the cross-sectional image slice is generated, for example, by a charged particle imaging beam at an angle GE deviating from 0°, such as for a beam angle of 10° or greater. Other sources of image distortion are errors in the image scanning unit of the charged particle imaging device, dynamic changes in the focal position of the charged particle imaging device, or deviations from the planar shape of the cross-sectional surface. Image distortion in the cross-sectional image slice is determined, for example, from a predetermined angle GE or from features of a second cross-sectional image, compared to prior knowledge of the semiconductor design, and the image distortion is digitally compensated.

[0046] Embodiments of the present invention include algorithms and methods for generating 3D volumetric images of an inspection volume. In a first step, cross-sectional image features are detected in multiple cross-sectional image slices, for example, by object detection methods known in the art. The cross-sectional image features are further classified in feature classification, and the cross-sectional image features are classified into first cross-sectional image features and second cross-sectional image features. In a second step, a depth map is generated for each cross-sectional image slice from the second cross-sectional image features representing a known or reference depth as described above. The depth map may be an absolute depth map in nm, or a relative depth map scaled relative to a reference given by a relative depth of a second integrated semiconductor feature (e.g., a feature in a specific layer or multiple planar layers constituting an integrated circuit). In a third step, the depth of each first cross-sectional image feature is determined based on the depth map. A fourth step is to generate a 3D volumetric image of the inspection volume, including depth information of multiple first cross-sectional image features from multiple depth maps of each of the multiple cross-sectional image slices. The 3D volumetric image of the inspection volume is generated based on the depth maps along with the multiple first cross-sectional image features of the multiple cross-sectional image slices, for example, by projection and interpolation of virtual cross-sectional image slices. The fifth step is to derive the characteristics of the integrated circuit structure of interest from the 3D volumetric image, such as features or characteristics of the HAR structure, like tilt angles or wobble. The sixth step is to derive defects in the integrated circuit features or characteristics from the 3D volumetric image. In the example, integrated circuit features deviating from their lateral design position or from a depth position outside the wafer surface correspond to manufacturing errors, for example, in the HAR structure. The detection of manufacturing errors is a particularly interesting aspect of wafer defect inspection. Using a method that determines the depth of the first cross-sectional image features through the features of the second cross-sectional image, it is possible to determine the manufacturing errors of the HAR structure with high accuracy, including global offset errors, such as the global lateral offset of all HAR structures.

[0047] In an embodiment, a method is provided to obtain a virtual cross-sectional image or a sequence of virtual cross-sectional images from a set of cross-sectional image slices, each virtual image slice comprising a plurality of virtual cross-sectional image pixels. The method includes the steps of: obtaining a sequence of N cross-sectional image slices by alternately imaging and milling a sequence of N cross-sectional surfaces into an inspection volume inside a wafer at an angle GF; and determining a first orientation direction for a first semiconductor feature, the first semiconductor feature forming a first plurality of first cross-sectional image features in the sequence of N cross-sectional image slices. The method of obtaining a virtual cross-sectional image or a sequence of virtual cross-sectional images further includes the step of calculating a virtual cross-sectional image perpendicular to the first orientation direction, wherein for each virtual cross-sectional image pixel, a pixel value is calculated by projection of at least a subset of the N cross-sectional image slices in the sequence of N cross-sectional image slices onto the first orientation direction and by interpolation of pixel values ​​from the projections of the at least a subset of the cross-sectional image slices. In the method of obtaining at least one virtual cross-sectional image, the number N of cross-sectional image slices is at least N = 10, preferably N > 100, for example, N is approximately 1000 or greater.

[0048] In an example of a method for obtaining at least one virtual cross-sectional image, for each virtual cross-sectional image pixel, a subset of at least one cross-sectional image slices is selected by evaluating the distance of each cross-sectional image slice in a sequence of N cross-sectional image slices to the virtual cross-sectional image pixel in a first orientation direction and selecting at least one first cross-sectional image slice with the minimum distance. In this example, a second cross-sectional image slice of the subset of at least one cross-sectional image slices is selected accordingly as the cross-sectional image slice with the second minimum distance. In a further example, additional cross-sectional image slices of the subset of at least one cross-sectional image slices are selected in a sequence of increasing distances to the virtual cross-sectional image pixel in the first orientation direction.

[0049] In an embodiment of the method for obtaining at least one virtual cross-sectional image, the steps of projecting at least one subset of cross-sectional image slices and interpolating pixel values ​​from the projection of the subset of at least one cross-sectional image slices include projecting and interpolating at least one subset of a first plurality of first cross-sectional image features to form a third plurality of first cross-sectional image features in the virtual image slices. In the example, the projection and interpolation steps are combined with at least one of feature extraction, thresholding operation, contour interpolation, or model-based interpolation. First cross-sectional image features are detected in a subset of at least one cross-sectional image slice through feature extraction, thresholding operation, contour interpolation, or model-based interpolation, and the third plurality of first cross-sectional image features in the virtual image slices are interpolated with high precision.

[0050] In an embodiment of the method for obtaining at least one virtual cross-sectional image, the method further includes generating a depth map for each of a sequence of N cross-sectional image slices. The steps, where for each of the N cross-sectional image slices, the index... In the example, there is an index. Depth map of each of the cross-sectional image slices It is generated from multiple second-section image features, which represent cross-sections passing through second semiconductor features oriented in a second orientation direction perpendicular to the first orientation direction. In the example, it has an index. Depth map of each cross-sectional image slice It is generated by determining the depth of the first cross-sectional image feature in the cross-sectional image slice from the lateral position of at least two second cross-sectional image features in the cross-sectional image slice.

[0051] In the example, a depth map of the generated virtual cross-section image is used. According to the depth map and with index Multiple depth maps of each of N cross-sectional image slices By evaluating each depth map of each cross-sectional image slice Depth map of virtual cross-section image pixels The distance in the first orientation direction is used to select at least one subset of cross-sectional image slices by selecting at least the first cross-sectional image slice with the minimum distance.

[0052] In the example, a second cross-sectional image slice of a subset of at least one cross-sectional image slice is selected as the cross-sectional image slice with a second minimum distance. In a further example, another cross-sectional image slice of a subset of at least one cross-sectional image slice is selected in a sequence of increasing distances to virtual cross-sectional image pixels in the first orientation direction.

[0053] In an embodiment of the method for obtaining at least one virtual cross-sectional image, the first semiconductor feature includes at least one of a via, a HAR structure, or a HAR channel of an integrated semiconductor circuit within the inspection volume of the wafer. The second semiconductor feature includes at least one of an isolation line or layer, a metal line or layer, a word line, or a semiconductor line or layer of an integrated semiconductor circuit within the inspection volume of the wafer. The method further includes the steps of determining at least one first cross-sectional image feature in each of a sequence of N cross-sectional image slices and determining at least one second cross-sectional image feature in each of a sequence of N cross-sectional image slices.

[0054] In an embodiment of the method for obtaining at least one virtual cross-sectional image, the first orientation direction is the z-direction perpendicular to the wafer surface, and the virtual cross-sectional image slices are calculated at a depth ZV below the wafer surface in a plane parallel to the wafer surface. In another embodiment of the method for obtaining at least one virtual cross-sectional image, for each virtual cross-sectional image pixel coordinate (x, y), a subset of at least one cross-sectional image slice is determined by selecting at least the m-th cross-sectional image slice with the smallest distance to depth ZV, such that the distance... It has an index All depth maps The minimum value. In the example, the second and additional cross-sectional image slices, which are subsets of at least one cross-sectional image slice, are arranged according to their distance from the virtual cross-sectional image pixel in the Z direction. The additional sequence is selected. In the example, the depth ZV is adjusted according to the depth of layers parallel to the wafer surface, wherein these layers are formed by second semiconductor features oriented in a second orientation direction parallel to the wafer surface. In the example, a first virtual cross-sectional image slice is calculated at a depth ZV1 in an isolation layer between two adjacent metal layers or word lines. In the example, a second virtual cross-sectional image slice is calculated at a depth ZV2 within the metal layer of the word line.

[0055] In an embodiment of the method for obtaining at least one virtual cross-sectional image, the optical axis of the charged particle beam imaging system used to acquire a sequence of N cross-sectional image slices is oriented perpendicular to the wafer surface, such that for the angle GE between the optical axis and the z-axis perpendicular to the wafer surface, angle... In this example, the charged particle imaging device is a helium ion microscope (HIM).

[0056] In an embodiment of the method for obtaining at least one virtual cross-sectional image, the method further includes the steps of forming at least one alignment feature near the inspection volume, configured to form at least one common cross-sectional image feature for mutual lateral alignment, and performing mutual lateral alignment on each of a sequence of N cross-sectional image slices having at least one common cross-sectional image feature. In an example, the mutual lateral image alignment step includes subtracting image distortion bias.

[0057] The method for obtaining at least one virtual cross-sectional image further includes the steps of: loading a wafer onto a wafer support stage in a dual-beam apparatus, the dual-beam apparatus comprising at least a FIB pillar and a charged particle imaging device, wherein a first optical axis of the FIB pillar forms an angle GF with the surface of the wafer support stage, and a second optical axis of the charged particle imaging device forms an angle GE with the normal to the surface of the wafer support stage, the first and second optical axes intersecting; and the step of moving the wafer support stage to align a first measurement position on the wafer with the intersection of the dual-beam apparatus. In the example, the wafer is not moved during the step of obtaining a sequence of N cross-sectional image slices in the examination volume.

[0058] According to an embodiment, a method for obtaining at least one 2D virtual cross-sectional image or a set of 2D virtual cross-sectional images from a set of cross-sectional image slices includes the following steps:

[0059] --A sequence of N cross-sectional image slices is obtained by alternately imaging and milling N cross-sectional surfaces into the inspection volume inside the wafer at an angle GF.

[0060] --Determine the first orientation of the feature of interest or the first semiconductor structure, wherein the first semiconductor feature forms a first plurality of first cross-sectional image features in a sequence of N cross-sectional image slices.

[0061] --Calculate virtual cross-sectional images perpendicular to the first orientation direction, where each virtual image slice includes multiple virtual cross-sectional image pixels.

[0062] Specifically, for each virtual cross-sectional image pixel, the pixel value is calculated by interpolating the pixel value from the projection of at least one subset of cross-sectional image slices of a sequence of N cross-sectional image slices in a first orientation direction. For each virtual cross-sectional image pixel, a subset of at least one cross-sectional image slice is determined by calculating the distance from each of the N cross-sectional image slices in the sequence to the virtual cross-sectional image pixel in the first orientation direction and selecting at least one first cross-sectional image slice with the minimum distance. In the example, a second cross-sectional image slice of the subset of at least one cross-sectional image slice is selected accordingly as the cross-sectional image slice with the second minimum distance. Further cross-sectional image slices of the subset of at least one cross-sectional image slice can be selected accordingly in a sequence of increasing distances from the virtual cross-sectional image pixel in the first orientation direction.

[0063] In the example, methods for obtaining at least one 2D virtual cross-sectional image or a collection of 2D virtual cross-sectional images include:

[0064] --Determine at least one first cross-sectional image feature in each of a sequence of N cross-sectional image slices; and

[0065] --Determine at least one second cross-sectional image feature in each of a sequence of N cross-sectional image slices, and derive the depth of at least one first cross-sectional image feature relative to the depth of the at least one second cross-sectional image feature. The depth of the at least one second cross-sectional image feature is known prior or used as a reference. Further details of the depth map generation method are described in examples and embodiments.

[0066] In embodiments of the wafer defect inspection apparatus and the method for inspecting an inspection volume within a wafer, the charged particle imaging apparatus is a scanning helium ion microscope. The cross-sectional surface is exposed at approximately an angle GF and extends approximately at an angle GF throughout the wafer surface of the inspection volume. The scanning helium ion microscope provides the necessary depth of focus and achieves high-resolution imaging of the cross-sectional surface in a single image scan without changing the focus of the charged particle imaging apparatus, even if the inspection volume has a large extension exceeding 2 µm in a direction perpendicular to the wafer surface.

[0067] According to embodiments of the present invention, a wafer inspection method for 3D inspection of an inspection volume in a wafer is disclosed, wherein the inspection volume is not extracted or removed from the wafer, and the inspection volume has a depth range greater than 1 µm, preferably greater than 2 µm, greater than 6 µm, or even 10 µm. The method includes the steps of: generating an image with a lateral resolution of less than 2 nm, preferably less than 1 nm, or even less than 0.5 nm through 3D inspection, wherein 3D inspection of the inspection volume is performed at the wafer, and wherein the image is a 3D image. The method further includes the step of milling at least one cross-sectional surface through the depth range of the inspection volume using a FIB pillar. The image is obtained by acquiring at least one image through at least one cross-sectional surface of the inspection volume using a helium ion microscope in a single image scan. The method further includes image processing, performing at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation. The method further includes the step of calculating one of a depth map, a 3D volume image, or a virtual cross-sectional image. The method further includes the step of calculating one of the dimensions, area, diameter, angle, or shape of a semiconductor feature. The method further includes the step of calculating one of the average value and statistical deviation of a plurality of semiconductor features.

[0068] In an embodiment of the present invention, a method for inspecting a wafer using a dual-beam apparatus is provided. The dual-beam apparatus includes a focused ion beam column (FIB) and a helium ion microscope (HIM). The optical axis of the FIB is arranged at an angle GF between 30° and 45° with the support surface of the wafer stage. The optical axis of the helium ion microscope is arranged perpendicular to the support surface. The optical axes of the FIB and HIM intersect. The method for inspecting the wafer further includes the following steps:

[0069] --The wafer is positioned at a first measurement location using a wafer stage with a wafer support, the wafer support being configured to hold the wafer at the intersection.

[0070] --The first cross-sectional surface is milled through the first inspection volume using an FIB pillar at an angle GF, wherein the depth extension LZ below the wafer surface is greater than 1 µm, and

[0071] --The surface of the first cross-section is imaged using a helium ion microscope in a single image scan to form a high-resolution cross-sectional image slice.

[0072] In the example, the first inspection volume has a depth range greater than 2 µm, greater than 6 µm, or even 10 µm, and the HIM is configured to generate an image with a lateral resolution of less than 2 nm, preferably less than 1 nm, or even less than 0.5 nm in a single image scan of the cross-sectional surface of the inspection volume at an angle GF.

[0073] In an embodiment of the present invention, a method for inspecting the inspection volume below the surface of a wafer using a dual-beam apparatus is disclosed. The dual-beam apparatus includes a focused ion beam column (FIB) and a helium ion microscope (HIM). The optical axis of the FIB is arranged at an angle GF greater than 30° to the support surface of the wafer stage, and the optical axis of the helium ion microscope is arranged perpendicular to the support surface. The optical axes of the FIB and HIM intersect. The method includes the following steps:

[0074] --The wafer is mounted on the wafer support stage in the dual-beam device.

[0075] --The wafer is positioned at a first measurement location using a wafer stage with a wafer support, the wafer support being configured to hold the wafer at the intersection.

[0076] --Using FIB pillars, multiple N cross-sectional surfaces are milled alternately at an angle GF within the inspection volume, where the depth extension LZ below the wafer surface is greater than 1 µm, and

[0077] --The surface of each cross section is imaged using a helium ion microscope in a single image scan to form multiple N high-resolution cross-sectional image slices.

[0078] In the example, the number N of cross-sectional image slices is at least 1. Preferred Even more preferably, N is about 1000 or greater. In an alternative example, the number of milled surfaces N is less than 50, preferably less than 20, and the inspection volume is divided into multiple B blocks arranged diagonally across the inspection volume.

[0079] In the example, the milling of multiple N cross-sectional surfaces has a depth extension LZ greater than 2µm, preferably greater than 6µm, such as 10µm, below the wafer surface. In the example, the method also includes the step of calculating at least a virtual cross-sectional image parallel to the wafer surface from multiple N cross-sectional image slices. For example, a first virtual cross-sectional image is calculated in a conductive layer or word line, and a second virtual cross-sectional image is calculated in an isolation layer.

[0080] In the example, the method further includes the step of performing mutual lateral alignment of each of the N cross-sectional image slices in a sequence with at least one common cross-sectional image feature. In the example, the method includes the step of forming at least one alignment feature near an inspection volume configured to form at least one common cross-sectional image feature, and performing mutual lateral alignment of the N cross-sectional image slices with the at least one common cross-sectional image feature. The method also includes the following steps:

[0081] --Generate an indexed array for each of the N cross-sectional image slices in the sequence. Depth map of each of the N cross-sectional image slices ;

[0082] --Determine at least one second cross-sectional image feature in each of a sequence of N cross-sectional image slices, the at least one second cross-sectional image feature representing a cross section through a second semiconductor feature oriented parallel to the wafer surface, wherein the index Depth map of each cross-sectional image slice It is generated from the lateral position of at least one second cross-sectional image feature.

[0083] In one embodiment, the cross-sectional imaging method in a wedge-cut geometry for a dual-beam apparatus is provided with additional grooves adjacent to the inspection volume to reduce debris. The additional grooves are fabricated by FIB milling in a first direction and include a first proximal groove and a second distal groove. After fabricating the additional grooves, the wafer is rotated 90°, and the inspection volume is inspected using cross-sectional imaging techniques in a wedge-cut geometry. Debris generated during milling of the cross-sectional surface is collected in the distal additional groove.

[0084] A wafer inspection apparatus according to an embodiment for inspecting an inspection volume at an inspection location within a wedge-cut geometry includes a six-axis wafer stage with a wafer support configured to hold the wafer, wherein the wafer surface rests on a support surface of the wafer support. The wafer inspection apparatus is configured to inspect an inspection volume within the wafer, which is not extracted from the wafer, and includes a dual-beam device. The dual-beam device includes a focused ion beam (FIB) column configured to approximate milling and exposing at least a first cross-sectional surface within a first inspection volume inside the wafer at an angle GF relative to the wafer surface, and a charged particle imaging device configured for imaging. The charged particle imaging device is, for example, a scanning electron microscope (SEM) or a helium ion microscope (HIM), wherein a first optical axis of the FIB column forms an angle GF with the surface of the wafer support, and a second optical axis of the charged particle imaging device forms an angle GE with the normal to the surface of the wafer support. The first and second optical axes of the column of the dual-beam device intersect. The wafer inspection apparatus also includes a control unit configured to perform and control any method of inspecting the inspection volume within the wafer. The wafer inspection apparatus also includes an image processing unit. The control unit and image processing unit include a processor and a memory with software code installed, configured to execute the method of the present invention, including the image processing step and the depth map generation step described above.

[0085] In embodiments of the invention, a wafer defect inspection apparatus is configured to mill and expose a plurality of cross-sectional surfaces, including at least first and second cross-sectional surfaces in a wafer. The wafer defect inspection apparatus is further configured to image the at least first and second cross-sectional surfaces to form first and second cross-sectional image slices. The wafer defect inspection apparatus also includes an image processing unit with software code mounted thereon, configured to determine cross-sectional image features in the at least first and second cross-sectional image slices and to determine the depth of cross-sectional features within an inspection volume, the cross-sectional image features being cross-sections of a semiconductor structure within the inspection volume. The wafer defect inspection apparatus also includes a defect detection unit configured to determine deviations from predetermined characteristics of the semiconductor structure within the inspection volume based on the cross-sectional image features.

[0086] In the example, the wafer defect inspection apparatus includes a wafer stage for holding the wafer, a focused ion beam (FIB) column configured to mill and expose a sequence of N cross-sectional surfaces within an inspection volume of the wafer at an angle GF relative to the wafer surface, wherein the depth extends beyond 2 µm, and a charged particle imaging device including a charged particle imaging device configured to image the multiple N cross-sectional surfaces using a single image scan to form each of a multiple N cross-sectional image slices. The wafer defect inspection apparatus also includes an image processing unit with software code installed, configured to determine a virtual cross-sectional image from the sequence of N cross-sectional image slices, wherein the virtual cross-sectional image is oriented parallel to the surface of the wafer.

[0087] In the example, the wafer defect inspection apparatus further includes: a focused ion beam (FIB) column configured to mill and expose a sequence of N cross-sectional surfaces within an inspection volume inside the wafer at an angle GF relative to the wafer surface during use; a charged particle imaging device configured to image the sequence of N cross-sectional surfaces during use to form a sequence of N cross-sectional image slices; and an image processing unit with software code installed, configured to determine a virtual cross-sectional image from the sequence of N cross-sectional image slices during use, wherein the virtual cross-sectional image is oriented parallel to the wafer surface. The wafer defect inspection apparatus with a dual-beam device includes an operation unit and an image processor configured to perform a method for obtaining a virtual cross-sectional image from the set of cross-sectional image slices.

[0088] In some embodiments or examples of the invention, the charged particle imaging beam is a scanning electron beam. In another example, the charged particle imaging beam is an ion beam from an ion microscope, such as a helium ion beam from a helium ion microscope (HIM). Resolutions achieved using HIM via slicing and imaging methods are below 1 nm, preferably below 0.5 nm, or even more preferably 0.25 nm. HIM also provides higher material contrast. HIM also provides a large depth of focus up to 10 micrometers, which is very large compared to a scanning electron beam, and particularly greater than the depth range of cross-sectional surfaces milled into the wafer surface at a milling angle GF preferably between 30° and 80°, exceeding 1 µm, 2 µm, or even reaching 10 m. Typically, the depth extension LZ of the examined volume below the wafer surface is greater than 1 µm, preferably greater than 2 µm, and the HIM is configured to have a depth of focus (DOF) exceeding the depth extension LZ, such that cross-sectional image slices are obtained by a single image scan using the HIM beam, without the need for image stitching. Other examples of imaging ion beam microscopy include other inert gases as ions, such as neon or a mixture of helium and neon.

[0089] In an embodiment, the dual-beam apparatus for wafer inspection includes (a) a wafer stage for supporting the wafer, (b) gallium FIB pillars arranged at an angle GF relative to the support surface of the sample support stage, configured to mill at least a first cross-sectional surface into the wafer at the angle GF, (c) a helium ion beam pillar arranged perpendicular to the support surface of the wafer stage, configured to generate a first cross-sectional image slice of the cross-sectional surface in a single image scan, (d) a secondary electron detector configured to collect multiple secondary electrons during a single image scan, and (e) an operation unit configured to operate and control the dual-beam apparatus during use. The FIBs with an angle GF between 30° and 40° are configured to mill the cross-sectional surface to a depth range extending below the wafer surface exceeding 1µm, 2µm, or even 6µm. The operation unit further includes (f) an image processing unit for performing at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation. The image processing unit is also configured to calculate one of a depth map, a 3D volumetric image, or a virtual cross-sectional image. The operating unit also includes (g) a defect detection unit for calculating one of the dimensions, area, diameter, angle, or shape of a semiconductor feature. The defect detection unit is also configured to calculate one of the average values ​​or statistical deviations of multiple semiconductor features.

[0090] A wafer inspection apparatus according to an embodiment for inspecting an inspection volume at an inspection location within a wedge-cut geometry includes a control unit for controlling a FIB column and a HIM beam. The control unit is configured to mill multiple N cross-sectional surfaces within the inspection volume approximately alternately at an angle GF using the FIB beam, and to image each cross-sectional surface using the HIM beam in a single image scan to form multiple N high-resolution cross-sectional image slices with a resolution better than 2 nm, preferably better than 1 nm. The inspection volume has a depth range LZ greater than 1 µm, greater than 3 µm, for example 6 µm, or even 10 µm along the optical axis of the HIM. Imaging of each cross-sectional surface to form multiple N high-resolution cross-sectional image slices is achieved using an HIM beam with a depth of focus (DOF) exceeding the depth extension LZ. The control unit is configured to scan the HIM beam in a single image scan over an area exceeding approximately 10 μm in lateral dimension LX or LY (LX or LY) of the inspection volume, and to collect multiple secondary electrons in a time-sequential manner using a secondary electron detector.

[0091] In an example of the present invention, a dual-beam apparatus for inspecting an inspection volume below the surface of a wafer is disclosed, the dual-beam apparatus comprising:

[0092] --A wafer stage with a wafer support, the wafer support being configured to hold a wafer on a support surface of the wafer support during use.

[0093] --Focused ion beam column (FIB) with its optical axis arranged at an angle GF of more than 30° relative to the support surface of the wafer stage.

[0094] --A helium ion microscope (HIM) with its optical axis perpendicular to the support surface, and the optical axis of the FIB column intersecting with the HIM.

[0095] --The stage control unit is configured to position the first measurement position of the wafer at the intersection point during use, and

[0096] --A control unit for controlling the FIB beam and the HIM beam is configured to approximate milling of multiple N cross-sectional surfaces in an inspection volume at an angle GF using the FIB beam, and to image each cross-sectional surface by scanning with the HIM beam to form multiple N high-resolution cross-sectional image slices, wherein the inspection volume has a depth extension LZ greater than 1µm, preferably greater than 2µm, below the wafer surface. In the example, the depth extension LZ in the direction exceeds 5µm, for example, up to 10µm. By approximating milling of multiple N cross-sectional surfaces in the inspection volume at an angle GF with the FIB beam, it means that due to the beam divergence of the FIB, the actual angle of the cross-sectional surface can deviate from the milling angle GF by several degrees, for example, 1° to 4°. The control unit is also configured to scan the HIM beam in a single scan during use over an area of ​​more than approximately 5µm to 10µm in lateral dimension LX or LY (LX or LY) of the inspection volume, and to collect multiple secondary electrons in a time-sequential manner using a secondary electron detector. The HIM is configured to have a depth of focus (DOF) exceeding the depth extension LZ, such that each of the plurality of N high-resolution cross-sectional image slices is obtained by a single image scan using a high-resolution HIM beam having a resolution of approximately less than 2 nm, such as 1 nm or even 0.5 nm. In the example, the dual-beam device according to the embodiment further includes an image processing unit with software code installed, configured to determine at least a first virtual cross-sectional image from the plurality of N cross-sectional image slices, wherein the at least first virtual cross-sectional image is oriented parallel to the wafer surface.

[0097] The optical axis of the FIB pillar is arranged at an angle GF to the wafer support, and the angle GF can be between 30° and 80°, for example, GF is approximately 30° to 45°. In another example, the angle GF between the first optical axis of the FIB pillar and the surface of the wafer support is in the range of 45° to 80°. In an alternative example, the angle GF between the first optical axis of the FIB pillar and the surface of the wafer support is in the range of 8° to 30°, for example, in the range of 8° to 15°.

[0098] The optical axis of a charged particle imaging device can be arranged at an angle perpendicular to the wafer support stage, wherein... In the example, the optical axis of the charged particle imaging device is arranged approximately perpendicular to the wafer, or the angle GE between the optical axis of the charged particle imaging device and the normal to the surface of the wafer support stage is less than 10°.

[0099] Another method of the present invention is a method for generating a measurement recipe for inspecting representative inspection volumes in a wafer. A method for generating a measurement recipe includes the following steps:

[0100] --Generate the first set of critical design performance metrics D1, which includes dimensional and CAD image data of key design features in the volume.

[0101] --Generate a second set of key manufacturing performance metrics, D2, including 3D volumetric image data obtained during manufacturing process development.

[0102] --Derive measurement formula R, which includes the step of obtaining at least one of a third set of performance indicators D3.

[0103] --The steps of generating a second set of key manufacturing performance indicators D2 and measuring recipe R include obtaining at least a first cross-sectional image slice of at least a first cross-sectional surface through a representative inspection volume, wherein the depth extension below the wafer surface is... At least one of the steps of obtaining at least one performance metric D1 to D3 includes an image processing step of extracting multiple first cross-sectional image features representing a cross section of multiple first semiconductor features at an angle GF, and the image processing includes at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation. At least one of the steps of obtaining at least one performance metric D1 to D3 includes a step of calculating at least one descriptive parameter of at least one first semiconductor feature from at least one of the multiple first cross-sectional image features, wherein the descriptive parameter is one of size, diameter, angle, area, shape, or volume. At least one step of obtaining at least one performance metric D1 to D3 further includes a step of calculating one of the average or statistical deviation of at least one descriptive parameter of the multiple first semiconductor features.

[0104] The present invention also provides a computer program product having program code for performing any of the methods of the present invention. For example, a computer program product is provided having program code for performing a method for inspecting volume in a wafer. For example, a computer program product is provided having program code for performing any method steps of a method for obtaining at least one virtual cross-sectional image. For example, the computer program product includes elements controlling the generation of a set of N cross-sectional image slices, elements selecting a first orientation direction, and elements calculating at least one virtual cross-sectional image from the set of N cross-sectional image slices by projection and interpolation in the first orientation direction.

[0105] According to this embodiment, a dual-beam device is provided, including a control unit and an image processing unit, configured to perform any of the method steps described above.

[0106] Throughout the embodiments, the first cross-sectional image feature includes at least a cross-section of a semiconductor structure extending in a direction parallel to an axis perpendicular to the wafer surface, and the second cross-sectional image feature includes at least a cross-section of a semiconductor structure extending in a direction parallel to the wafer surface. For example, the first common cross-sectional image feature includes at least a cross-section of one of the following: a via, a HAR structure, or a HAR channel of an integrated semiconductor circuit. For example, the second cross-sectional image feature includes at least a cross-section of one of the following: an isolation line or layer, a metal line or layer, a word line, or a semiconductor line or layer of an integrated semiconductor circuit.

[0107] The embodiments or examples described above can be combined with each other, either wholly or in part. This also applies to the examples or embodiments describing different aspects of the invention. Attached Figure Description

[0108] The invention will be more fully understood by referring to the following figures:

[0109] Figure 1 This is an illustration of a cross-sectional imaging technique used to interpret a sample extracted from a wafer.

[0110] Figure 2 This is an illustration of a dual-beam device applied to a wafer, configured for cross-sectional imaging technology of wedge-shaped cutting geometry.

[0111] Figure 3 Details of the cross-sectional imaging technique in wedge-cut geometry are shown.

[0112] Figure 4 Other aspects of cross-sectional imaging techniques in wedge-cut geometry are illustrated.

[0113] Figure 5 This is a diagram illustrating the method for determining the depth of cross-sectional image features in an x-direction view.

[0114] Figure 6 This is an illustration of a method for determining the depth of cross-sectional image features in an example of two cross-sectional image slices.

[0115] Figure 7 This is an illustration of a cross-sectional imaging technique that utilizes wedge-cut geometry with additional alignment features.

[0116] Figure 8 This is an illustration of a cross-sectional imaging technique that utilizes image segment stitching to create a wedge-shaped cut geometry.

[0117] Figure 9 This is an illustration of a cross-sectional imaging technique in a wedge-cut geometry that utilizes additional alignment features at different depths.

[0118] Figure 10 This is an illustration of a cross-sectional imaging technique in a wedge-cut geometry using the fan-beam computed tomography method.

[0119] Figure 11 This is an illustration of a cross-sectional imaging technique in a wedge-shaped cutting geometry with additional grooves for reducing debris.

[0120] Figure 12 This is an illustration of a cross-sectional imaging technique in a wedge-cut geometry with distortion compensation.

[0121] Figure 13 This is an illustration of a cross-sectional imaging technique in a wedge-cut geometry with additional alignment features manufactured above the inspection volume.

[0122] Figure 14 This is a diagram illustrating the steps of a cross-sectional imaging technique in wedge-shaped cutting geometry.

[0123] Figure 15 Operation and control unit.

[0124] Figure 16 This is a diagram illustrating the interpolation of virtual image slices from a sequence of cross-sectional image slices.

[0125] Figure 17 This is another illustration of the interpolation of virtual image slices from a sequence of cross-sectional image slices from a curved cross-sectional surface.

[0126] Figure 18 is a diagram of a dual-beam setup using a helium ion microscope.

[0127] Figure 19 illustrates the slicing and imaging method using the dual-beam device according to Figure 18, and the results of single-scan image acquisition and virtual cross-sectional image calculation in alternating layers of the memory device.

[0128] Figure 20A statistical evaluation of the HAR characteristics of memory devices is shown.

[0129] Figure 21 This demonstrates a time-efficient method for checking the volume of a wafer.

[0130] Figure 22 The inspection of a depth inspection volume with a large milling angle GF is shown.

[0131] Figure 23 A flowchart illustrating the process of generating a monitoring recipe for online wafer inspection is shown.

[0132] Figure 24 An analysis of the inspection volume based on prior information is shown.

[0133] Figure 25 An example of a monitoring formulation for analyzing the tilt of a HAR structure is shown.

[0134] Figure 26 This is an illustration of a cross-sectional imaging technique in a wedge-cut geometry that utilizes additional alignment features on a preceding or first cross-sectional surface with a large extension.

[0135] Figure 27 This is an illustration of a method for determining the depth of cross-sectional image features. Detailed Implementation

[0136] Figure 1A schematic diagram of a common cross-sectional imaging method for obtaining a 3D volumetric image of an integrated semiconductor sample is shown. The cross-sectional method, also known as the slicing and imaging method, achieves three-dimensional (3D) volumetric image acquisition through a “step-by-step repetition” approach. First, the integrated semiconductor sample is prepared for the common cross-sectional imaging method using methods known in the art. Throughout this disclosure, “cross-sectional image” and “slice” will be used as synonyms. In one step, a thin surface layer or “slice” of material is removed. This material slice can be removed in several ways known in the art, including milling using a focused ion beam or polishing at a grazing angle by a focused ion beam (FIB) column 50. For example, a focused ion beam 51 propagates almost parallel to the z-axis and scans in the y-direction to mill through the top surface of the sample 10 and expose a new cross-sectional surface 52 in the yz plane. As a result, the newly exposed cross-sectional surface 52 is available for imaging. In a subsequent step, the cross-sectional surface layer 52 is raster-scanned using a charged particle beam (CPB) imaging system 40, such as a scanning electron microscope (SEM) or a second FIB, to obtain a cross-sectional image slice 100.1. The optical axis 42 of the charged particle imaging system 40 can be arranged parallel to the x-direction or tilted at an angle relative to the x-direction. A detector (not shown) collects secondary and backscattered electrons to reveal material contrasts within the integrated semiconductor sample, visible as different gray levels in cross-sectional image slice 100.1. Metallic structures generate brighter measurement results. The surface layer removal and cross-sectional imaging process is repeated through cross-sectional surfaces 53 and 54, and other equally spaced cross-sectional surfaces, to obtain a sequence of 2D cross-sectional image slices 1000 (including, for example, N cross-sectional image slices 100.2, 100.3, ... 100.N at different depths) to construct a three-dimensional 3D dataset. A representative cross-sectional image slice 100.1 was obtained by measuring a commercial Intel processor integrated semiconductor chip with 14 nm technology.

[0137] Using this method, at least first and second cross-sectional image slices are generated by subsequently milling the cross-sectional surfaces into the integrated semiconductor sample using a focused ion beam to expose or make a sequence of cross-sectional surfaces available for imaging, and each cross-sectional surface of the integrated semiconductor sample is imaged using a charged particle beam imaging system 40. A 3D image of the integrated semiconductor structure is reconstructed from a sequence of N 2D cross-sectional image slices 1000. The distance d between the cross-sectional image slices 100.1, 100.2, and 100.3 can be controlled by a FIB milling or polishing process and can be between 1 nm and 30 nm.

[0138] In the example above, the cross-sectional image plane is oriented perpendicular to the top surface 55 of the integrated semiconductor wafer, where the normal to the top surface 55 is oriented parallel to the z-direction, as shown below. Figure 1As shown. This results in 2D cross-sectional image slices oriented parallel to the yz plane, or in other words, the cross-sectional image plane includes the z-axis or wafer normal axis, and the imaging direction x is parallel to the wafer surface. Therefore, conventional slicing and imaging methods in this conventional geometry are only suitable for samples extracted from wafers.

[0139] The objective of this invention is to provide a slicing and imaging method suitable for inspecting volumes within a wafer. In a first embodiment of the invention, the method for generating 3D volumetric images utilizes a slicing and imaging approach that applies a so-called "wedge cutting" method or wedge cutting geometry to an inspection volume within the wafer without requiring sample removal from the wafer. The slicing and imaging method is applied to inspection volumes several µm in size, such as 5 µm to 10 µm lateral extensions in a 200 mm or 300 mm wafer, without requiring sample removal from the wafer. Grooves or edges are milled into the top surface of the integrated semiconductor wafer so that the cross-sectional surface can be angled to the top surface. The 3D volumetric image of the inspection volume is acquired at a limited number of measurement locations, such as representative locations on the die, for example at a process control monitor (PCM), or at locations identified by other inspection tools. The slicing and imaging method will only partially damage the wafer, while other dies remain usable, or the wafer can still be used for further processing.

[0140] Figure 2 A first embodiment of the invention and an example of a wedge-shaped cutting geometry are illustrated. A wafer 8 is provided with several measurement positions 6.1 and 6.2, wherein the measurement positions are defined, for example, in a position diagram generated by inspection tools or design information. The wafer 8 is placed on a wafer support stage 15. Measurement position 6.1 of the wafer 8 is aligned at the intersection 43 of a dual-beam apparatus with a five-axis wafer stage (not shown), which includes an FIB column 50 with an FIB optical axis 48 and a charged particle beam (CPB) imaging system 40 with an optical axis 42. At the intersection 43 of the two optical axes of the FIB and CPB imaging systems, the wafer surface is arranged at an angle GF relative to the FIB axis 48. The FIB axis 48 and the CPB imaging system axis 42 include an angle GFE, and the CPB imaging system axis forms an angle GE with the z-axis perpendicular to the wafer plane. Using the FIB 51, the surface of the wafer 8 is impacted at an angle GF, and the inclined cross-section surface is milled into the wafer at inspection position 6.1 at an approximately inclined angle GF by ion beam milling. Figure 2 In the example, the tilt angle GF is approximately 30°. Due to the beam divergence of focused ion beams (e.g., gallium ion beams), the actual tilt angle of the tilted cross-section surface can deviate from the tilt angle GF by as much as 1 to 4°. An image of the milled surface is acquired using a charged particle beam imaging system 40 with a tilt angle GE relative to the wafer normal. Figure 2In the example, the angle GE is less than 15°. During imaging, a charged particle beam is scanned by the scanning unit of the charged particle beam imaging system 40 along a scanning path on the cross-sectional surface of the wafer at measurement position 6.1, generating secondary particles and scattered particles. Particle detector 17 collects at least some of the secondary and scattered particles and communicates the particle count to control unit 19. Control unit 19 controls the charged particle beam imaging column 40 of FIB 50 and is connected to control unit 16 to control the position of the wafer mounted on the wafer support via a wafer stage (not shown). Control unit 19 communicates with operation unit 2, which triggers, for example, the placement and alignment of the wafer 8 at measurement position 6.1 at intersection 43 by moving the wafer stage, and repeatedly triggers FIB milling, image acquisition, and stage movement operations.

[0141] Each intersecting surface is imaged using a substantially vertical charged particle imaging beam 44, such as a scanning electron beam or any other charged particle beam microscope (e.g., helium ion microscopy (HIM)).

[0142] In the second embodiment, a slicing and imaging method for a wedge-shaped cutting geometry is provided. By repeating the slicing and imaging method for the wedge-shaped cutting geometry, cross-sectional surfaces 52, 53, and 54 (see...) are generated. Figure 1 and 3 The image slices are stacked into N cross-sectional image slices, and a 3D volume image of the inspection volume of wafer 8 at measurement position 6.1 is generated. Figure 3 A wedge-cut geometry is shown in an example of a 3D memory stack. The wafer surface is milled with a FIB beam 51 at an angle GF less than 80° relative to the wafer surface (e.g., 30° angle GF), but other angles less than 45° are also possible, such as 40° or 36°. In the example, a small milling angle GF between 8° and 45° is preferred because it allows for better imaging of the first cross-sectional image features (e.g., HAR structures) at high resolution. In another example, a steeper milling angle GF (e.g., greater than 30° or greater than 45°) is preferred. A steeper milling angle GF between 45° and 80° allows for cutting through the depth of the inspection volume with a small lateral dimension. With a steeper milling angle, greater depths can be achieved, and 3D volumetric images of deep semiconductor stacks with depths greater than 6µm (e.g., 10µm or 15µm depths in thick or deep semiconductor stacks) can be generated. At steeper angles, the lateral extension of the cross-sectional surface is kept below, for example, 30µm, preferably below 20µm, so that cross-sectional image slices can be obtained with high throughput without image stitching.

[0143] Depending on the chosen coordinate system, wafer surface 55 coincides with the XY plane. The memory stack extends in the Z direction, perpendicular to wafer surface 55. Using FIB beam 51, a new cross-sectional surface 52 is generated in the wafer, wherein cross-sectional surface 52 is inclined at approximately angle GF relative to the wafer surface. Cross-sectional surface 52 is scanned, for example, by SEM beam 44. Figure 3 In the example, the SEM beam 44 is arranged to be perpendicularly incident on the wafer surface 55, generating a high-resolution image. The cross-sectional image slice includes a first cross-sectional image feature formed by intersections with high aspect ratio (HAR) structures or vias (e.g., first cross-sectional image features of HAR structures 4.1, 4.2, and 4.3) and a layer comprising, for example, SiO2, SiN-, or tungsten wires. The second cross-sectional image features are formed by the intersection. Some lines are also referred to as "word lines". The maximum number of layers M is typically greater than 50, for example greater than 100 or even greater than 200. HAR structures and layers extend through most of the volume in the wafer, but may include gaps. HAR structures typically have a diameter of less than 100 nm, for example about 80 nm, or for example 40 nm. Therefore, the cross-sectional image slice contains the first cross-sectional image features as cross-sections or intersections of the HAR channel footprints at different depths (Z) at various XY locations. In the case of cylindrical vertical memory channels, the obtained first cross-sectional image features are circular or elliptical structures at different depths determined by the position of the structures on the inclined cross-sectional surface 52.

[0144] The choice of slice distance d is a balance between the required sampling and throughput. The thickness d or minimum distance d between two adjacent cross-sectional image slices is adjusted to values ​​typically on the order of several nm, such as 30 nm, 20 nm, 10 nm, 5 nm, 4 nm, or even smaller. Once the material layer of predetermined thickness d has been removed with FIB, the next cross-sectional surface 53 is exposed and can be imaged with an almost perpendicular imaging beam 44. Multiple N cross-sectional image slices acquired in this way cover the inspection volume of wafer 8 at measurement location 6.1 and are used to form a high 3D resolution 3D volumetric image, for example, below 10 nm, preferably below 5 nm, and to reconstruct the characteristics of the semiconductor structure of interest, such as memory stacks within the inspection volume. Multiple N cross-sectional image slices can comprise a small number of images, for example... or Up to several hundred cross-sectional image slices, for example Or more. The volume being examined typically has in the xy plane. The lateral extension of the cross-sectional surface is 2µm to 15µm below the wafer surface, but the lateral extension of the inspection volume can also be significantly larger, reaching a lateral dimension of approximately 1 mm. Preferably, the lateral extension of a single cross-sectional surface is less than 30 μm, for example, less than 20 μm. In some examples, for memory devices including memory HAR structures, it is sufficient to provide at least three cross-sectional image slices for each HAR structure. Therefore, the preferred slice distance d is less than 30 nm.

[0145] In the example, features and 3D locations of the semiconductor structure of interest, such as the location of HAR channels, are detected using image processing methods, such as detection from the centroid of the HAR. Further descriptions of 3D volumetric image generation, including image processing methods and feature-based alignment, are found in U.S. Provisional Application No. 62 / 858.470 and German Patent Application No. 10 2019 006645.6, both of which are incorporated herein by reference in their entirety. It should be noted that the layers and HAR structures do not need to extend through the entire measured volume.

[0146] Now, in a third embodiment of the invention, a method for recovering information regarding the relative position, depth, and orientation of cross-sectional image features of a HAR structure within an inspection volume relative to each other is described. The angle GF between the FIB direction and the wafer surface 55 (XY plane) controls the depth LZ, and the scan controls the field of view (FoV) and total lateral area LX and LY of the imaging beam 44 in x and y, which can be used for scanning imaging using the imaging beam 44. Proper positioning of image slices relative to each other is generally referred to as alignment, which includes registering features of the 2D image in a first step and recovering the correct distance d between subsequent cross-sectional image slices in a second step. The first alignment method is as follows: Figure 4 As shown. A columnar HAR structure, such as a channel or channel aperture, denoted by the number 75 in one example, extends through the volume of a wafer in the z-direction, such as a memory chip. The HAR structure is oriented perpendicular to the wafer surface and becomes visible in a cross-sectional image slice obtained by imaging beam 44 (see...). Figure 3 As shown in the figure above, the cross-sectional surface is oriented at a predefined angle relative to the approximate GF of the HAR structure. Figure 4 In the example, the tilt angle GF is approximately 25°. Figure 4b illustrates two examples of two consecutive 2D cross-sectional image slices with indices n and n+1, where the first cross-sectional image feature of the HAR channel is indicated by 77.1 to 77.5. In this example, cross-sectional image slices n and n+1 are obtained perpendicular to the cross-sectional image surfaces n and n+1, where the imaging beam 44 is positioned at an angle GFE of approximately 90° relative to the ion beam 51, and where the image coordinate systems X', Y', Z' are rotated by an angle GE on the X-axis relative to the wafer coordinate system (X, Y, Z). Each of the cross-sectional surfaces n and n+1 forms an edge with the top surface of the wafer 55, and two examples of the edges are indicated by reference numerals 76.1 and 76.2. In this example, the first HAR structure intersects with the first cross-sectional image feature 77.1 in the first cross-sectional image slice with index n, and again with the first cross-sectional image feature 77.2 in the second image slice with index n+1. The first cross-sectional image features 77.1 and 77.2 are common cross-sectional image features of the two image slices, shifted by the lateral displacement vector dy'. Due to lateral displacement, some cross-sectional image features of the HAR structure disappear. For example, cross-sectional image feature 77.4 of the nth image almost disappears in the (n+1)th cross-sectional image slice, where only a portion of cross-sectional image feature 77.5 is visible. Typically, the apparent lateral displacement dy' in the image coordinate system is determined by the tilt angle GF of the inclined cross-sectional surface, the distance d between image slices, and the angle GE of the imaging beam 44. Given. The difference in the lateral coordinates corresponds to the difference in depth dz beyond the wafer surface, where In the example, the depth dz variation of the corresponding first cross-sectional image features 77.1, 77.2 perpendicular to the wafer surface 55 is derived from the lateral displacement dy' of the first cross-sectional image features 77.1, 77.2 in at least two images n and n+1 of a plurality of images, which are generated by FIB beam milling at a predetermined angle GF and imaging at a predetermined angle GE.

[0147] According to Figure 4 In the depth determination method explained in the example, the imaged charged particle beam 44 is projected at a predetermined angle GE (see example). Figure 3 The normal to the wafer surface 55 is oriented, which corresponds to the z-axis in the selected coordinate system. This method utilizes the apparent lateral displacement of a first cross-sectional image feature among multiple cross-sectional images, where the first cross-sectional image feature is a cross-section of a vertically oriented structure in the semiconductor wafer, such as a via or HAR structure. The vertically oriented via or HAR structure is oriented perpendicular to the wafer surface 55. However, utilizing according to... Figure 4 In the example method shown, the lateral displacement error of the first cross-sectional image features affects the depth determination of the first cross-sectional image features. Therefore, according to Figure 4The method for depth determination using features from a first cross-sectional image, as shown in the third embodiment, has limited accuracy, especially for deep structures. For example, the surface of a real cross-sectional image deviates from a planar shape, and this deviation increases with depth. Furthermore, HAR structures deviate from being completely perpendicular to the wafer surface.

[0148] Therefore, another objective of the present invention is to provide a method for determining the location of features in a first cross-sectional image with even higher precision. A second example of the depth determination method of the present invention provides a solution. The method is described as using an imaging beam of charged particles 44 oriented with a normal parallel to the wafer surface 55, thus a predetermined angle GE is selected to be approximately . Figure 5 This setup is illustrated. However, the method is not limited to imaging charged particle beams 44 oriented perpendicular to the wafer surface 55, and angles GE deviating from 0° are also possible.

[0149] Two cross-sectional surfaces 52 and 53 are shown, which are inclined relative to the wafer surface 55 at a predetermined angle GF. Figure 2 In this example, the tilt angle GF is approximately 26°, but other tilt angles, such as 30° or greater, are possible. Two HAR structures 75.1 and 75.2, along with a set of layers L1 to L4, are shown. Because the imaging charged particle beam 44 is incident perpendicular to the wafer surface 55, theoretically, the first cross-sectional image features 77.1 and 77.2 of the perfect HAR structure 75.1 appear at the same y-coordinate in the nth and (n+1th)th images, and it is impossible to determine the depth from the apparent lateral displacement of the cross-sectional image features 77.1 and 77.2 in the nth and (n+1th)th images. If the first cross-sectional image features 77.1 and 77.2 of the real HAR structure 75.1 appear at different y-coordinates in the nth and (n+1th)th images, any depth determination based on different y-coordinates would be misleading. The depth determination according to the second example utilizes the second cross-sectional image features of multiple layers, including layers L1 to L4, from multiple N cross-sectional image slices, including cross-sectional image slices n and n+1. For example, the second cross-sectional image feature of layer L4, which has a depth Z41 below the wafer surface and is located on the upper surface 78 (see references 73.1 and 73.2), has a y' coordinate Y41' in the nth cross-sectional image. n And it has y' coordinate Y41' in the (n+1)th cross-sectional image. n+1 For example, the second cross-sectional image feature 73.3 of the lower surface 72 of layer L1 at a depth of Z12 below the wafer surface has a y' coordinate Y12'n in the nth cross-sectional image. Knowing the depths of at least two edges of the two layers, such as depth Z41 and depth Z12, the y coordinates in the cross-sectional image slices can be converted to z coordinates, and a depth map can be derived. For example, the depth z of the centroid 79.1 of the first cross-sectional image feature 77.1 in the nth image is derived from the y-coordinate of the second cross-sectional image feature 73.3 and the y-coordinate y' of the centroid 79.1, where In an alternative example, the depth z of the centroid 79.1 of the first cross-sectional image feature 77.1 in the nth image is derived from the y-coordinates of at least two second cross-sectional image features 73.1 and 73.3, and the depth map pass Obtain the local displacement of the location of the second cross-sectional image feature, for example, in two consecutive cross-sectional image slices n and (n+1) with y-coordinate Y41'. n and Y41' n+1 The local displacement D2 of the upper boundary surface of layer L4 at the location can be used, for example, as a reference for the lateral alignment of depth maps of continuous cross-sectional image slices n and (n+1), or as a verification of the local milling angle GF or the local slice distance dz. Using the latter method, precise knowledge of the milling angle GF is not required, and a depth map can be derived for each cross-sectional image slice with index n. In this example, the cross-sectional image slices are assumed to be parallel to the x-direction, thus having the same depth for each x-coordinate for a given y-coordinate, hence the depth map. It depends only on y, but is independent of x. In another example, depth This is also different for each x-coordinate, and a depth map is generated for each x-coordinate in a similar manner. Using the method of the third embodiment, a depth map is generated for each of the multiple N cross-sectional image slices. Furthermore, the precise depth and location of cross-sectional image features can be determined from multiple cross-sectional image slices obtained under a wedge-cut geometry. An example using the method of the third embodiment is that a depth map is generated for the cross-sectional image slice from the lateral position of a second cross-sectional image feature within the slice. Furthermore, for cross-sectional image slices obtained under wedge-shaped cutting geometry, the depth and position of the first cross-sectional image features can be accurately determined.

[0150] Figure 6 The nth and (n+1)th cross-sectional images are shown, which are milled by FIB beam 51 at a predetermined angle GF and by angle. Imaging charged particle beam 44 oriented along the z-direction or perpendicular to the wafer surface 55 (see Figure 5The image is obtained through imaging. The vertical HAR structure appears as a first cross-sectional image feature in the cross-sectional image slices, such as first cross-sectional image features 77.1, 77.2, and 77.3. Since the imaging charged particle beam 44 is parallel to the orientation of the HAR structure, the first cross-sectional image features representing, for example, an ideal HAR structure will appear at the same y-coordinate. Deviations from a constant position correspond to manufacturing errors, such as tilting or "wobbling" of the HAR structure. For example, the first cross-sectional image features of the ideal HAR structures 77.1 and 77.2 are centered on the line 80 at the same y-coordinate of the nth and (n+1)th image slices. The first cross-sectional image features of the HAR structure, such as 77.1 to 77.3, therefore do not allow determination of the depth of the cross-sectional image features within the wafer, nor do they allow determination of the slice distance d or slice angle GF. The cross-sectional image slices also include multiple second cross-sectional image features of multiple layers (including, for example, layers L1 to L5), such as second cross-sectional image features 73.1 and 73.2 of layer L4. In the cross-sectional image slices, the layer structure is shown as a high-contrast segment along the X-direction. However, the positions of these stripes or second cross-sectional image features representing multiple layers (layers L1 to L5 are shown here) relative to the first cross-sectional image features change with each cross-sectional image slice. As each layer intersects the image plane with increasing depth, the positions of the second cross-sectional image features change from image slice n to image slice n+1 in a predetermined manner. The upper surface of layer L4, indicated by reference numerals 78.1 and 78.2, is shifted by a distance D2 in the y-direction, thereby, for example, by equation Calculate the slice distance d or slice angle GF. Based on the determined position of the second cross-sectional image features (e.g., 78.1 and 78.2), the two depth maps of the two consecutive cross-sectional image slices can be aligned relative to the Z coordinate. It can also export multiple depth maps for multiple cross-sectional image slices. .

[0151] By extracting features from the second cross-sectional image features, such as edge detection or centroid calculation and image analysis, and by understanding the depth of the second cross-sectional image features from prior information (e.g., from design information), the lateral position and depth of the first cross-sectional image features within a cross-sectional image slice can be determined with high precision. The depth of any boundary or surface of each layer (e.g., layers L1 to L5) extending parallel to the wafer surface is typically known with very high precision, and this depth is constant over a large area of ​​the wafer due to the planar manufacturing techniques involved in wafer fabrication. In the example, when the depth of the second cross-sectional image features, such as the layer or the upper or lower surface of the layer, is unknown, the method for depth determination is applied. Even without knowing the precise depth value in nm, the depth relative to the boundary or surface of the layer can be specified with high precision. In this example, the depth of the first cross-sectional image feature within the cross-sectional image slice is given relative to the depth of the second cross-sectional image feature boundary (e.g., the upper and lower surfaces of multiple layers). In other words, the depth of the first cross-sectional image feature is given relative to the depth of the second cross-sectional image feature, such as above, beyond, at the same depth, or on the depth scale, according to multiple second cross-sectional image features L1 to LM with M layers.

[0152] In a fourth embodiment of the present invention, a method and structure for aligning cross-sectional image slices are provided. Figure 7Several further aspects of the method for 3D volume inspection in a wafer are illustrated. The inspection volume 13 at measurement location 6.1 is marked, for example, by alignment marks 24. After placing the wafer with measurement location 6.1 below the intersection of FIB beam 51 and charged particle imaging beam 44, the wedge above the first cross-sectional surface 52 is first removed, and a first cross-sectional image slice is obtained by scanning the charged particle imaging beam 44. After imaging the first cross-sectional surface 52, FIB milling is repeated at an angle GF relative to the wafer surface 55, and the charged particle imaging beam 44 images multiple cross-sectional surfaces including cross-sectional surfaces 52, 53, and 54, forming multiple cross-sectional image slices in the slicing and imaging method described above. In the example, the slicing and imaging method in the wedge-cut geometry includes additional alignment features for mutual alignment of the multiple cross-sectional images. Before milling the cross-sectional surfaces, additional alignment features, such as alignment marks 20.1, 20.2, and 20.3, are created on the wafer surface 55 by depositing a layer and structuring the layer with an alignment pattern. Examples of alignment patterns can be crosses, as shown in alignment feature 20.1, but any alignment pattern known in the art is also applicable. Alignment features can also be, for example, alignment grooves or edges 22 milled into the boundary of a wedge cut, in the portion where the FIB does not change when the slice is milled through the inspection volume 13 at measurement position 6.1. Alignment feature 22 is not limited to edges, but can also be, for example, a serrated feature. Alignment features can also be integrated circuit features exposed on the wafer surface near the inspection volume, such as integrated circuit structure 25. Alignment features as described above form common cross-sectional image features in several cross-sectional image slices and allow for additional methods for lateral alignment of cross-sectional image slices. Coordinates of a first cross-sectional image feature representing the semiconductor structure of interest are obtained with higher precision, and a depth map is derived from the Y coordinates of a representative second cross-sectional image feature with higher precision. .

[0153] Alignment features exist as common cross-sectional image features across all cross-sectional image slices and can be analyzed in each individual cross-sectional image slice using conventional edge detection techniques (e.g., gradient-based contour extraction). Edge locations are known to be most accurately determined in directions orthogonal to the edge. In the example, the accuracy of edge localization is improved in at least two directions by adding features, such as at least two alignment features. In the example, the additional alignment features 20.1, 20.2, and 20.3 are used for lateral image alignment of the cross-sectional image slices. Alignment features 20 and 22 enable the reconstruction of a 3D volume from multiple tilted cross-sectional image slices, independent of the geometry of the semiconductor within the examined volume. In the example, the aforementioned additional alignment features are configured for high-contrast imaging utilizing charged particle imaging beams. High contrast and good visibility of alignment patterns or edges, for example, are achieved by locally coating the wafer surface at the measurement location prior to the patterning process or by milling a wedge into the wafer surface with a material that provides high material contrast relative to the wafer sample material silicon. Suitable coating materials are, for example, platinum (Pt) or carbon (C) or a combination of both.

[0154] In addition to or in addition to the manufactured alignment marks, integrated circuit features present and visible on the wafer surface can be used as additional alignment marks. For example, visible structure 25 is formed by a HAR structure present on the wafer surface near the inspection volume and is visible in a cross-sectional image slice in a manner similar to alignment marks 20.1, 20.3, and 24. In this example, the visible structure 25 of the integrated circuit is used as an additional alignment feature and is used for lateral image alignment of the cross-sectional image slice. For example, the surface of the wafer near the inspection volume is processed to expose the integrated circuit structure 25 such that the exposed integrated circuit structure 25 is visible to charged particle imaging pillars and can be used for alignment via common cross-sectional image features.

[0155] The fifth embodiment describes another aspect of the invention. A cross-sectional surface milled by the FIB beam 51 is imaged by a charged particle imaging beam 44 using scanning imaging. In this example, the cross-sectional surface is not perpendicular to the charged particle imaging beam 44, and the field of view (FOV) of a single scan operation may be too small, or the cross-sectional surface may be outside the depth of focus of the charged particle imaging beam 44, in order to obtain an image in a single scan operation. In this case, at least two scanned image segments are obtained by the charged particle imaging beam 44, and the at least two image segments are stitched together. Figure 8An example is illustrated. After surface 53 is milled and exposed by an FIB beam (not shown), a first image segment 26.1 is obtained by scanning with a charged particle imaging beam 44.1 at a first location. The first image segment 26.1 includes, for example, alignment features 20.1 on the surface of wafer 55. The wafer, mounted on a wafer stage (not shown), is then laterally displaced, and additional image segments, such as image segments 26.2 and 26.3, are obtained by scanning the charged particle beam 44.1 at a second or third location. Multiple image segments 26.2, 26.3 are also configured to overlap each other to achieve mutual alignment of at least each pair of image segments using cross-sectional image features of structures present in the overlapping region of adjacent image segments. “Top-down” imaging of the inclined cross-sectional surface requires a certain depth of focus (DoF) perpendicular to the charged particle imaging beam. When the depth of focus of the imaging beam 44 is insufficient, for example, if the DOF of the imaging beam 44 is less than the LZ, the FoV of the scanning imaging beam 44 is split into smaller subfields, and sequential imaging is achieved using progressive focus adjustment. Typically, the depth of focus of a charged particle imaging beam is limited by resolution requirements. For high-resolution imaging below 2 nm (e.g., 1 nm), the depth of focus decreases. For example, in scanning electron microscopy (SEM), the depth of focus at 1 nm resolution is less than 50 nm. As will be explained in more detail below, for examination volumes with a large depth extension LZ, it is therefore preferable to use helium ion microscopy, which provides a greater depth of focus of approximately 10 μm at a high resolution of approximately 1 nm.

[0156] In an example where the depth of focus (DOF) of a charged particle imaging beam with a focal length (LZ) exceeding the desired resolution of less than 2 nm (e.g., 1 nm) for a charged particle imaging beam 44.2 is shifted to a deeper location within the wafer, resulting in image segment 28. The extension of image segment 28 is chosen such that the depth extension 30 of image segment 28 relative to the cross-sectional image surface 53 is less than the depth of focus (DoF) of the charged particle imaging beam 44 at the desired resolution. It should be noted that, for example, SEM also provides dynamic focus adjustment during image scanning; however, it has a limited z-position range of approximately 1 µm to 2 µm and requires changing the focus position, for example, by moving the wafer stage, such that the depth extension 30 of image segment 28 is less than the dynamic focus adjustment range of the charged particle imaging beam 44. By adjusting the z- or focus position, for example by moving the wafer stage in the z-direction, the desired resolution can be maintained throughout the imaging process of the inclined cross-sectional surface 53. A smaller second trench can also be generated near the inspection volume, and at least one additional alignment feature can be generated at the bottom of the second trench. Therefore, several alignment marks are configured at a predetermined depth to overcome the DoF limitation of the imaging beam 44. In the example, such as... Figure 9As shown, additional alignment features 32.1 and 32.2 are placed in three depth levels. For example, alignment feature 32.2 is fabricated at depth level DLZ (reference numeral 34) for aligning image segments at depth level DLZ. To achieve this, small trenches at the desired depth DLZ are milled in the wafer surface, and alignment patterns are constructed at depth level DLZ. In the example, alignment features 32.1 or 32.2 at different depth levels are integrated circuit features of the integrated semiconductor close to the inspection volume. Alignment features 32.1 or 32.2 at different depth levels form common image features in a sequence of cross-sectional image slices and enable precise alignment of the sequence of cross-sectional image slices to generate a 3D volumetric image with high precision. In the example, the image stitching method and the image alignment method with additional alignment features provided at different depths are applied together with a dual-beam device of SEM having FIB pillars arranged at angles between 8° and 45° as a charged particle beam imaging system. In the example, an image alignment method with additional alignment features provided at different depths is applied to the inspection of depth inspection volumes that extend below the wafer surface by more than 5µm, such as 6µm or 10µm or more.

[0157] Figure 26 Another example of additional alignment features according to the fourth embodiment is shown. In this example, the angle GFE between the charged particle imaging beam 44 and the FIB 51 used for slicing multiple surface sections is... The angle GF is chosen to be approximately 30°, but is typically between 25° and 60°. The plane formed by the FIB 51 and the imaging charged particle column 44 is orthogonal to the wafer surface 55. As a result, the imaging beam is orthogonal to the imaging cross-sectional surface (slice). For simplicity, only a single cross-sectional surface 52 is shown. With this setup, the entire cross-sectional surface can be kept at the “focal point” of the imaging beam 44. This is particularly important for an electron beam acting as an imaging charged particle column with a limited depth of focus, approximately 200 nm. The large angle GFE between FIB 51 and imaging column 44 reduces the possible volume conflict between the two columns.

[0158] In the case of particularly deep samples (>10 μm), a steeper cutting angle (e.g., >60∘) will allow for a cut covering the entire depth range, thus keeping the extension of the cross-sectional surface 52 within the field of view (FOV) of approximately 10 μm to 15 μm typical of single-beam charged particle microscopy. Conversely, shallower angles, such as a GF angle below 20°, will result in a very extended field of view and will require image stitching. Therefore, in one example, a tilt angle GF between 25° and 60° is preferred.

[0159] Initial alignment of cross-sectional image slices in a direction orthogonal to the imaging charged particle beam (i.e., parallel to the slice) can be achieved using alignment marks 10.1 or 20.2. Figure 26 In the example, alignment marks 20.1 and 20.2 are formed on the first section 96, which extends further in the x-direction by FIB44 compared to a plurality of second section image slices (e.g., section 52) below the first section 96. With this arrangement, at least the parallel surface segment 23.1 is implemented as part of the first section surface 96, remaining at least on one side of the second section surface (e.g., section surface 52). In the parallel surface segment 23.1, alignment features 20.1 can be formed, for example, by deposition and etching. A second parallel surface segment 23.2, comprising additional alignment features 20.2, can be implemented.

[0160] Due to various factors, such as image distortion or variations in the focal plane from the first cross-sectional surface 96 to the second cross-sectional surface 52, the alignment of multiple cross-sectional images using the aforementioned alignment method with additional alignment features such as features 20.1 and 20.2 is generally found to be too coarse for 3D volumetric image reconstruction with lateral resolutions below a few nm, such as 5 nm, 3 nm, or even below 2 nm. Therefore, according to a fourth embodiment of the invention, an additional method based on precise alignment of the wafer's structure or semiconductor features within the examined volume is applied. The precise alignment method also includes the calculation of the actual slice thickness d. Through the first coarse alignment step, registration errors due to repeated patterns are avoided, and mapping or registration of cross-sectional image features within the cross-sectional image slices is obtained. Through the second step or precise alignment, the relative positional accuracy of each pair of cross-sectional image slices reaches an accuracy below 5 nm, 3 nm, or even below 2 nm.

[0161] The integrated circuit structure inside a chip typically includes a vertical structure (e.g., Figure 26 The 3D memory channels or vias (4.1 and 4.2) and horizontal structures (e.g., “word lines” or metal lines L1 to Ln) can be used for precise alignment and slice thickness determination, as described above. Figure 27 An example is shown. Figure 27 A cross-section of a wafer according to a wedge-cutting method is shown. The horizontal and vertical structures are formed by the horizontal edges 80 of the layers and, for example, HAR structures (e.g., Figure 26 The vertical edge 78 of element 4.1 or 4.2 in the image is indicated. The Y' components of the positions of the horizontal edge 80 and the vertical edge 78 in the cross-sectional image slice at index (n+1) are compared with the Y' components of the positions of the horizontal edge 80 and the vertical edge 78 in the cross-sectional image slice at index n, and then... ("WL" = "wordline") and ("Ch" = "channel") is used to calculate the difference. The Y' axis lies in the plane of the cross-sectional image slice. Imperfect alignment results in additional displacement. This is equal for all structures in the cross-sectional image slice. Therefore, the actual displacements of the horizontal and vertical structures considered above are...

[0162]

[0163]

[0164] Where d is the actual distance between the two cross-sectional surfaces. Note that displacement...

[0165] It is negative because the Y-axis points downwards along the slice.

[0166] Therefore, the displacement error can be calculated. And the slice thickness d, they are from measurements and And the lateral position of the tilt angle GF reconstructed cross-sectional image slice (n+1) relative to the cross-sectional image slice (n) in the volume is necessary:

[0167]

[0168]

[0169] The slices are orthogonal to... Figure 27 The displacement in the X direction of the plane in the middle image is simply equal to the structure and The corresponding measured displacement along the X-axis. The small change in the tilt angle GF affects the displacement error. The calculation of thickness d has only a limited impact. Therefore, even without knowing the tilt angle with higher precision, such as 1°, for example, having... It can also calculate, for example, displacement error with an accuracy of less than 1 nm. .

[0170] Multiple cross-sectional image slices typically do not form perfectly parallel planes, but rather possess certain morphologies, such as... Figure 17 As shown. To account for the morphology, the above equations can be solved locally at multiple locations within the cross-sectional image slices. Therefore, the slice thickness d can be calculated separately for multiple locations of each pair of cross-sectional image slices. Furthermore, the lateral displacement can be calculated locally. and Furthermore, potential image distortion can be considered.

[0171] By using the coarse and fine alignment methods described above, precise alignment of multiple cross-sectional image slices was obtained, and high-precision 3D volumetric image reconstruction with a resolution of less than 5nm, less than 3nm, and even less than 2nm was achieved.

[0172] In the method for generating a 3D volumetric image of the inspection volume at measurement location 6.1 on the wafer, a plurality of subsequent cross-sectional surfaces having a distance d are milled by a FIB beam 51, which is configured to be at a predetermined angle GF relative to the wafer surface. In the example coordinate system, the FIB beam 51 scans in the x-direction, and the cross-sectional surfaces are tilted at an angle approximately equal to GF. “Approximately equal” means that due to the beam divergence of the FIB, the actual angle of the cross-sectional surfaces can deviate from the milling angle GF by a few degrees between 1° and 4°. In the example, the generation of a plurality of subsequent cross-sectional surfaces with a distance d less than 15 nm, less than 5 nm, or even lower can be achieved by lateral displacement of the wafer stage in the y-direction or z-direction. In the sixth embodiment of the invention, a plurality of subsequent cross-sectional surfaces with a distance d of approximately 5 nm are obtained by milling with the FIB beam 51 without lateral or vertical movement of the wafer stage. Figure 10 As shown, the FIB beam scans along the x and z directions around the FIB beam coincidence point 58, without the wafer being displaced by the wafer stage. This fan-shaped tomography method applied to wedge-cut geometry allows for very precise and rapid adjustment of the slice distance d of subsequent cross-sectional image slices. The milling angle is not constant at a predetermined angle GF, but varies around the predetermined angle GF with the difference 64 between the minimum and maximum milling angles, where the angle spread GZ is approximately 1 mrad. Therefore, the distance between adjacent image slices varies. As an example, milling can be configured such that the slice distance 60.1 between two adjacent cross-sectional image surfaces (e.g., cross-sectional image surface 53 and cross-sectional image surface 54) is constant at the surface of wafer 55. Depending on the different milling angles, the slice distance 62.1 at the bottom of the examination volume varies from slice to slice. For example, the first cross-sectional image surface 53 is milled with FIB beam 51.1 at a first milling angle 66, and the second cross-sectional image surface 54 is milled with FIB beam 51.2 at a second milling angle 68, such that the distance between two adjacent cross-sectional image surfaces in the wafer plane is dy1 (reference numeral 60.1), and the distance between two adjacent cross-sectional image surfaces at a depth LZ below the wafer plane is dy2 (reference numeral 62.1) greater than dy1. Therefore, the precise control of the mutual distance between the cross-sectional surfaces can be maintained within nm precision. In this case, the resulting multiple cross-sectional slices are not parallel to each other. The alignment and depth determination method according to the above embodiments of the present invention still applies in the same way.

[0173] In the example, the method of tilting the FIB beam through the scanning unit of the FIB pillar is used to adjust the milling angle GF. In the example, the distance between continuous cross-sectional surfaces is adjustable. In the example, for at least some distances of the cross-sectional image surfaces, the distances between multiple cross-sectional surfaces are adjusted to be different. By adjusting the distance and angle GF, the throughput and resolution of the 3D volumetric image of the inspected volume are fully or partially adjusted to suit the needs of the wafer inspection task.

[0174] One problem is achieving uniform milling over a large area of ​​multiple cross-sectional image surfaces. In the seventh embodiment, an arrangement and method are disclosed to avoid the deposition of large amounts of debris on inclined cross-sectional surfaces. Figure 11 As shown, at measurement position 6.1, a first proximal trench 92 is first milled into the wafer surface adjacent to inspection volume 13, followed by a second distal trench 94, such that inspection volume 13 is located between the first and second trenches 92 and 94. The wafer is then rotated 90°, and multiple cross-sectional image surfaces, including cross-sectional image surfaces 53 and 54, are milled into inspection volume 13 by FIB beam 51. For each cross-sectional surface, the FIB beam is scanned in the x-direction with angular spread GX (reference numeral 90) and tilted in the z-direction with angular spread DZ (reference numeral 64) by scanning mechanism of the FIB pillar. Multiple cross-sectional images are obtained by a charged particle imaging beam (not shown) from a vertical direction, optionally together with imaging of an alignment feature (e.g., alignment feature 20.1). Debris generated during FIB milling accumulates in the distal trench 94, located relative to the FIB beam in inspection volume 13. The spacing between the two trenches 92 and 94 is configured to accommodate the examination volume 13. The width and depth of the proximal trench 92 are adjusted to achieve the slicing angle of the aforementioned approximate angle GF. The depth of the distal trench 94 is adjusted to achieve the slicing angle of the aforementioned approximate angle GF while maintaining debris deposition in the distal trench 94. Therefore, the slicing and imaging method in the wedge-cut geometry maintains low debris and high accuracy. In the example, two or more trenches with different orientations and spacings are configured to generate multiple sub-volumes of desired size and orientation for 3D tomography. Alignment features, such as alignment feature 20.1, are placed near the examination volume 13. Similar to... Figure 9 As shown in the example, at least alignment feature 20.1 can be manufactured at depth DLZ.

[0175] In the eighth embodiment, the imaging using the charged particle imaging device is degraded due to image distortion. Similarly, the method steps of the above embodiments and examples are similarly applicable to configurations of charged particle imaging devices having an angle GE > 0°. In the example, the optical axis of the charged particle imaging device is arranged at an angle GE to the normal to the wafer surface. In this case, the imaging coordinate system (X', Y') is rotated by an angle GE relative to the wafer surface, and the cross-sectional image is distorted in the Y direction. The cross-sectional image slice is digitally transformed by a deformable image transformation by applying a first image magnification in the x direction, different from the second image magnification in the y direction. In the example, the scanning unit of the charged particle beam imaging device is degraded due to distortion, such as trapezoidal distortion. The cross-sectional image slice is digitally transformed by distortion compensation to compensate for the image distortion. The results of the deformable image transformation and distortion compensation are as follows: Figure 12 As shown. Figure 12 The left side shows a cross-sectional image slice with image distortion, which is produced by image acquisition at angle GE, and is shown as trapezoidal distortion as an example. Figure 12 The right side is used Figure 6 The reference figures show cross-sectional image slices after distortion compensation.

[0176] In the ninth embodiment, additional alignment features are provided on top of the inspection volume. Figure 13 This is an illustration of the distance determination between subsequent cross-sectional image slices, which are tilted relative to the top surface of the integrated semiconductor sample by approximately a predetermined master milling angle GF. The cross-sectional image slices are tilted towards the top surface by approximately angle GF. Additional alignment features may also be the edges of the cross-sectional surfaces having wafer surface 55, such as the edge 76.1 between surface 52 and wafer surface 55 (see...). Figure 7Edge 76.1 is removed during subsequent FIB milling of the subsequent cross-sectional image surface (e.g., surface 53). However, additional edges 76.3 and 76.4 provide a measurement of the slice distance d. In the example, alignment features 38.1 and 38.2 are provided on the top surface of the wafer at the location of inspection volume 13 prior to milling. Alignment features 38.1 and 38.2 are not parallel but are inclined to each other at an angle GV and have different distances in the x-direction at each edge of each respective cross-sectional image surface, for example, distance 36.4 at edge 76.4 of cross-sectional image surface 54. The y-position of edge 76.4 can be derived with high precision based on the predetermined geometry and position of alignment features 38.1 and 38.2 and distance 36.4. In another example, the distance between two cross-sections having alignment features 38.1 and 38.2 from the first cross-sectional image surface 53 to the adjacent second cross-sectional image surface 54 changes from distance 36.3 to distance 36.4 by an amount of dx. At each cross-sectional surface, the distances between alignment features 38.1 and 38.2 are measured at the corresponding edges of the cross-sectional surface and wafer surface 55. The variation of distance dx is determined, and the angle GV is known; therefore, the distance dy in the y-direction between subsequent edges of the cross-sectional image can be calculated using the following formula:

[0177]

[0178] To determine the distance d between slices, consider the milling angle GF and

[0179] .

[0180] In wedge-cut geometry, dx itself or any error contained in the measurement of dx is now reduced by a factor. Therefore, using the additional references 38.1 and 38.2, the slice thickness d is derived, and the positional variation of the y-position between second cross-sectional image features in adjacent cross-sectional image slices can be derived with high precision. Thus, the depth of the second cross-sectional image features and the depth map of each cross-sectional image slice are derived with higher precision. Based on the depth or depth map of the second cross-sectional image features, the depth of the first cross-sectional image features is derived with high precision.

[0181] A tenth embodiment of a method for 3D inspection of a wafer's 3D inspection volume at least at the measurement location includes... Figure 14The steps are shown below. In step S1, the wafer is loaded onto a wafer support stage on a wafer carrier and placed into the vacuum chamber of a dual-beam apparatus. The dual-beam apparatus includes a focused ion beam (FIB) column, whose first optical axis is arranged at an angle GF with the surface of the wafer support stage or the wafer surface when loaded onto the wafer support stage. The measuring instrument also includes an imaging charged particle beam, such as a SEM or a helium ion microscope, whose second optical axis forms an angle GE with the normal to the surface of the wafer support stage and is arranged at an angle GFE with the FIB column, and the first and second optical axes intersect.

[0182] The location of the inspection volume at the first measurement position on the wafer is determined, and the wafer is placed by moving the wafer stage so that the first measurement position is located at the intersection of the FIB column and the charged particle imaging column. Optionally, the wafer orientation can be further adjusted by rotating the wafer support stage of the wafer stage.

[0183] In optional step S3, at least one alignment feature, such as an alignment mark, is fabricated near the measurement location. In optional step S3, a coating may be locally deposited on the wafer surface at the measurement location to cover the inspection volume, as well as the alignment mark for generating high imaging contrast, or to achieve high contrast between the sample surface and sections not covered during milling. Such as Figure 7 Additional alignment features of alignment mark 20.1 or alignment mark 22 are configured such that at least one alignment feature is visible in at least two cross-sectional image slices, forming a common cross-sectional image feature for mutual alignment of at least two cross-sectional image slices. In the example, the additional alignment features are, for example... Figure 7 Alignment mark 20.1 or alignment mark 22 is configured such that at least one alignment feature is visible in all cross-sectional image slices for mutual alignment of all cross-sectional image slices. In the example, the alignment feature is fabricated above the inspection volume and is configured to determine the positions of a plurality of edges formed by the intersection of a plurality of cross-sectional surfaces, including first and second cross-sectional surfaces, with the wafer surface, including the first and second edges.

[0184] In step S5, the first wedge is removed from the top surface of the wafer by FIB milling to expose the first cross-sectional surface for cross-sectional imaging. The FIB milling is monitored, for example, by a charged particle imaging beam.

[0185] In step S7, a slicing and imaging method is performed in the wedge-cut geometry, and multiple cross-sectional images are generated. FIB milling of the new cross-sectional surfaces is monitored, for example, by a charged particle imaging beam, including monitoring the alignment features created in step S3 to control the distance d between subsequent image slices. This results in a sequence or multiple N cross-sectional image slices comprising at least a first cross-sectional image slice and a second cross-sectional image slice within the inspection volume. Obtaining the first and second cross-sectional image slices involves subsequently exposing at least the first and second cross-sectional surfaces in the inspection volume by milling approximately into the inspection volume at angle GF using an FIB pillar, and imaging the at least first and second cross-sectional surfaces with a charged particle imaging device to obtain at least the first and second cross-sectional image slices. Typically, the number N of cross-sectional image slices is at least N = 10, preferably N > 100, and even more preferably N is about 1000 or greater.

[0186] In the example, a slicing and imaging method is performed in the fan-shaped tomography method, which includes scanning the focused ion beam of the FIB column in a first direction by a scanning unit to expose a first cross-sectional surface within the examination volume, tilting the focused ion beam in a second direction perpendicular to the first direction by a scanning unit, and scanning the focused ion beam in the first direction by a scanning unit to expose a second cross-sectional surface within the examination volume, such that a plurality of cross-sectional surfaces including the first and second cross-sectional surfaces form different angles approximately tilted by an angle GF with respect to the wafer surface.

[0187] In step S8, cross-sectional image features are detected and classified, such that the cross-sectional image features are classified into first cross-sectional image features and second cross-sectional image features, as well as optional alignment features that form common cross-sectional image features. For cross-sectional image feature detection and classification, methods known in the art can be applied, such as edge detection, image comparison, feature extraction, object detection, including combinations thereof, also known as machine learning algorithms. Therefore, optionally, common cross-sectional image features are detected and classified in the first and second cross-sectional image slices.

[0188] In step S9, multiple cross-sectional images are aligned with each other such that multiple first cross-sectional image features of a vertical structure, such as a HAR structure or a via, appear at nearly the same lateral coordinates, e.g., xy coordinates. In the example, multiple cross-sectional images are aligned using at least one additional alignment feature or a common cross-sectional image feature. Multiple cross-sectional images are aligned with each other using at least one previously generated alignment feature such that multiple first cross-sectional image features of a vertical structure, such as a HAR structure or a via, appear at similar lateral coordinates. Local lateral position errors of the first cross-sectional image features (e.g., cross-sections of individual HAR structures) are detected by utilizing at least one additional alignment feature. Global wobble errors of the HAR structure at depth z perpendicular to the wafer surface are detected by determining the cross-sectional position of the HAR structure relative to the alignment mark.

[0189] In the example, mutual lateral image alignment includes subtraction or digital compensation of image distortion deviations between at least the first and second cross-sectional image slices.

[0190] In step S10, the location of the second cross-sectional feature in each cross-sectional image slice is determined. Create a depth map In the example, any slice of the cross-sectional image The depth z of a point is derived from the lateral displacement of the second cross-sectional image feature. This second cross-sectional image feature corresponds, for example, to the layer structure of a semiconductor device oriented parallel to the wafer surface. In the example, any point within the cross-sectional image slice... The depth Z of the point is extrapolated from the lateral position of a plurality of second cross-sectional image features parallel to the wafer surface on the layer surface, which are formed by multiple layers of an integrated semiconductor device.

[0191] Therefore, in step S10, multiple depth maps of multiple N cross-sectional image slices are generated. , where index .

[0192] In the example, depth determination and depth map creation via second cross-sectional image features are combined with alignment features, such as additional alignment markers or reference-based alignment. Integrated semiconductor samples may include highly repetitive features, such as gates in a gate layer, which can lead to blurred image registration of the second cross-sectional image features. Typically, according to any embodiment of the invention, coarse registration using alignment features or references formed on top of the integrated semiconductor sample can reduce blurring and improve the speed of fine image registration of the second cross-sectional image features used for depth determination.

[0193] In step S11, the depth of a first cross-sectional image feature in at least one cross-sectional image slice is derived from the depth map. In this example, steps S10 and S11 form a single step of depth determination to determine the depth of at least one first cross-sectional image feature from at least one second cross-sectional image feature.

[0194] In the example, the step of determining the depth of the first cross-sectional image feature includes determining the lateral difference between the first position of the second cross-sectional image feature in the cross-sectional image slice and the second position of the second cross-sectional image feature in the second cross-sectional image slice.

[0195] In the example, at least two second cross-sectional image features are determined within a first cross-sectional image slice, where each second cross-sectional image feature represents an integrated semiconductor structure at a different depth within the examined volume. The step of determining the depth of the first cross-sectional image feature includes determining the lateral positions of the at least two second cross-sectional image features.

[0196] Depth determination via the first cross-sectional image features, obtained through the second cross-sectional image features, involves statistical averaging and is therefore more robust to, for example, image noise. Due to the typically larger number of layers, such as 5, 10, or up to 100 layers, the cross-sectional image includes multiple second cross-sectional image features, and the statistics of depth determination are improved by using depth determination via multiple second cross-sectional image features. Since defects are few, they do not affect the overall quality of the depth determination method. Furthermore, defect candidates can be detected as outliers of the expected depth based on statistical evaluation.

[0197] In the above embodiments and examples, the cross-sectional surface is considered a planar surface, exposed by FIB milling and polishing. In another example, due to errors or degradation during the milling and polishing process, at least the cross-sectional surface is curved. Therefore, the corresponding cross-sectional image slice is an image of the curved surface, and thus includes distortion and artifacts. In the example, cross-sectional image features are used to determine the curvature of the cross-sectional surface, and the determined surface curvature is applied to correct the distortion of the cross-sectional image slice.

[0198] In step S12, a 3D volumetric image of the inspection volume is generated, including 3D positional information of the first cross-sectional image features. In step S12, depth information of multiple first cross-sectional image features from multiple depth maps of each of multiple cross-sectional image slices is combined. The depth maps and the multiple first cross-sectional image features from the multiple cross-sectional image slices are combined to form the 3D volumetric image of the inspection volume. Corresponding integrated circuit features to the first cross-sectional image features are derived to form a 3D volumetric image of the corresponding integrated circuit features. In a further step, defects or deviations of the corresponding integrated circuit features, or characteristics of the corresponding integrated circuit features, are derived from the 3D volumetric image.

[0199] In the example, the second cross-sectional image features are analyzed, and the local cross-sectional surface slope can be determined based on the local variation in the width of the second cross-sectional image features of the structure in the representation layer. In the example, the information on the local cross-sectional surface slope is used for iterative milling in feedback loop correction. In the example, the digital image transformation is achieved by performing a transfer from a distorted image slice to an undistorted image slice obtained from the planar cross-sectional surface. In the example, the cross-sectional surface is curved in the x and y directions. As in the previous embodiments, the position or edge of the second cross-sectional image features is used for distortion correction, and the corresponding z-coordinate of each pixel of the corresponding cross-sectional image slice is extracted, including the z-position of the first cross-sectional image features.

[0200] In one aspect of the invention, detection results at different measurement locations on the wafer are compared, for example, by deriving integrated circuit features and characteristics in 3D based on slicing and image methods within wedge-cut geometry at different measurement locations. The features and characteristics of the integrated circuit can be, for example, tilt or wobbling errors of the HAR structure. Determining the depth of the integrated circuit features represented by the first cross-sectional image features at different measurement locations using the method described above allows for precise comparisons, including comparisons of the depth of integrated circuit features and characteristics. The second cross-sectional image features, representing lines or layers parallel to the wafer surface, form a common reference for determining depth at different measurement locations. Therefore, tilt and wobbling of the perpendicular HAR structure can be determined in 3D at different locations or measurement points on the wafer without requiring an additional common alignment reference.

[0201] In the eleventh embodiment, a wafer defect inspection apparatus is provided, configured to inspect inspection volumes within a wafer without removing a sample from the wafer. Figure 2Some aspects of the wafer defect inspection apparatus are explained below. The wafer defect inspection apparatus includes: a focused ion beam column 50 configured to mill and expose at least first and second cross-sectional image surfaces in an inspection volume of wafer 8; and a charged particle imaging device 40 configured to image multiple cross-sectional surfaces, including at least a first and second cross-sectional surface, to form multiple cross-sectional image slices, including the first and second cross-sectional image slices. The wafer defect inspection apparatus also includes an operation unit 2 that triggers, for example, the placement and alignment of a measurement position 6.1 of wafer 8 at an intersection by moving a wafer stage via a control unit 19 and a stage control unit 16. The operation unit 2 is equipped with software operation code and triggers, during operation, inspection of the inspection volume using slicing and imaging methods in a wedge-cut geometry, including repetitive operations of FIB milling, image acquisition, and stage movement. The operation unit communicates with the control unit 19, which controls the charged particle beam imaging column 40 of the FIB 50 and is connected to the control unit 16 to control the position of the wafer mounted on a wafer support via a wafer stage (not shown). Figure 15 Other aspects of the wafer defect inspection apparatus are illustrated. In this example, the control unit 19 does not communicate directly with the charged particle beam column, but rather with a charged particle beam control unit 86, which controls the operation of the charged particle beam column, such as a charged particle imaging device 40 or a FIB 50. The operation unit 2 includes an image processing unit 82 with computer programs and software program code installed, for receiving image information from the charged particle detector 17 via the control unit 19, synchronized with the scanning operation of the imaging charged particle beam scanned by the scanning unit (not shown) of the charged particle imaging device 40. During use, the image processing unit 82 and the software program code installed in the image processing unit 82 are configured to determine, during use, first and second cross-sectional image features in at least first and second cross-sectional image slices by feature detection and classification, as described above in step S8. The image processing unit 82 and the software program code installed in the image processing unit 82 are also configured during use to determine the depth of the first and second cross-sectional features within the inspection volume, the first and second cross-sectional image features being cross-sections of an integrated semiconductor structure within the inspection volume. Image processing unit 82 and software code implement image processing methods as described above, such as corner or edge detection, thresholding, morphological operations, or similar operations, which are well known in the art. Image processing has recently been improved by increasing computational speed, for example by using computer clusters comprising hundreds of processors in image processing unit 82. Image processing methods for extracting features or structures from integrated semiconductor samples may also involve machine learning algorithms or be replaced by machine learning algorithms.

[0202] In the example, the integrated semiconductor structure is a previously known structure. Design information or 3D CAD information can be used to improve edge extraction of metal lines and HAR channels, layer surface location extraction, and high-precision depth determination. For example, CAD information can be used to identify the location of the metal line ends, which should therefore not be visible in the cross-sectional image. This reduces outliers in the image processing method. Furthermore, the depth of the second cross-sectional image features is typically determined with high precision from the 3D CAD information. The image processing unit 82 and the software program code installed in the image processing unit 82 enable the determination of the depth of the second cross-sectional features within the inspection volume during use by comparing with the 3D CAD information, and, for example, the derivation of the depth of the first cross-sectional image features according to the method described above.

[0203] The operation unit 2 also includes a defect detection unit 84, which is configured to determine deviations from predetermined 3D characteristics of the semiconductor structure within the inspection volume during use, for example, by comparing first cross-sectional image features with 3D CAD data or library data of predetermined 3D characteristics. Furthermore, defect candidates can be detected as outliers in statistical evaluation.

[0204] The operation unit 2 also includes an interface and communication unit 88, which communicates with external input and output devices, such as user control terminals, tablet computers, databases, chip processors, or manufacturing operating systems.

[0205] To obtain information about the inspection volume within the wafer, such as defects or deviations from the desired shape of the semiconductor structure, a sequence of inclined cross-sectional surfaces is sequentially milled and imaged "piece by piece." Once a material layer of adjustable thickness is removed by FIB milling, a vertical charged particle imaging beam is used to acquire new cross-sectional image slices. Multiple N cross-sectional image slices acquired in this manner cover the inspection volume within the wafer, generating 3D volumetric image data with lateral pixel dimensions and a distance dz between consecutive image slices. The 3D volumetric image data is analyzed and used to reconstruct, for example, the characteristics of a semiconductor memory stack within the inspection volume. Utilizing the embodiments of the invention described above, a reconstruction is provided that includes the recovery of information about mutual positions and orientations, including the depth of the individual cross-sectional image features of interest within the inspection volume.

[0206] In the twelfth embodiment, 3D volumetric image data is generated from a sequence of N cross-sectional image slices by resampling or interpolating cross-sectional image slices from an inclined cross-sectional surface to a regular grating. The interpolation artifacts in the 3D volumetric image generated from the sequence of N cross-sectional image slices are reduced using the interpolation method described below. Embodiments of 3D volumetric image data generation include methods for obtaining at least a 2D virtual cross-sectional image from a set of cross-sectional image slices. Integrated semiconductor elements in semiconductor devices typically have predetermined shapes and orientations. They are typically arranged in layers parallel to or extending perpendicular to the wafer surface. Examples of such elements are as described above and... Figure 3 The diagram shows a memory channel or HAR structure and layer with metal wires. In the proposed method for rapid 3D inspection, virtual cross-sectional image slices are generated in a plane orthogonal to the direction of the semiconductor feature of interest. Figure 16 An example is shown. Figure 16 Three representative cross-sectional surfaces 52, 53, and 54 are shown, which were sequentially formed by FIB milling at approximately a grazing angle GF within an examination volume (not shown) below the wafer surface 55. From each cross-sectional surface 52, 53, and 54, cross-sectional image slices with indices n-1, n, and n+1 were generated by imaging charged particle microscopy, such as scanning electron microscope beam 44 or HIM. The cross-sectional surfaces were milled by an ion beam (not shown) at an angle GF at a relative distance dz in the z-direction, and the z-distance was achieved, for example, by lateral movement of the wafer stage (not shown) or by combining the above with a fan-shaped tomography method (see above). Figure 10 The FIB is controlled by the scan offset in the z-direction, as described above. Figure 16 In the example shown, the first distance in the z-direction between the first cross-sectional surface 52 and the second cross-sectional surface 53 is... The second distance in the z-direction is different from that between the second section 53 and the third section 54. The distance dz is in the range of a few nm, such as 5 nm, 6 nm, or even larger, such as 10 nm. For example, Approximately 6nm, and Approximately 7nm. Each cross-sectional image is scanned using a lateral pixel grating in the xy plane, where the pixel grating is approximately 2nm, 1nm, or even smaller, such as 0.5nm. The projection lines of the pixel grating in the y-direction on cross-sectional surfaces 52, 53, and 54 are shown by dashed lines 123 extending in the x-direction (only the index 123 of surface 52 is shown). An example of a virtual cross-sectional image slice 121 perpendicular to the z-axis is shown. The virtual image slice 121 is parallel to the wafer surface 55 at depth ZV. The pixel value at point C in the virtual cross-sectional image slice 121 is obtained by projecting the pixel values ​​of adjacent cross-sectional image slices in the direction of the structure of interest. In this example, the structure of interest is a HAR structure oriented perpendicular to the wafer surface. Therefore, the projection or interpolation direction is parallel to the z-axis. Arrows 125.1, 125.2, 125.3, and 125.4 show several examples of interpolation directions. In the example, the shortest distance from point C in the virtual image slice to the adjacent cross-sectional image surface is determined in the projection or interpolation direction. In this example, the first cross-sectional surface 52 is closest to point C, and the pixel value at pixel position C of the virtual cross-sectional image slice 121 is selected to be the same as the pixel value at pixel A of the first cross-sectional image slice of the first cross-sectional surface 52. Pixel A of the cross-sectional image n-1 of the first cross-sectional image plane 52 and pixel C of the virtual cross-sectional image slice 121 have the same x and y coordinates, and the distance AC in the z direction is the minimum distance among all distances in the z direction between the virtual image slice 121 and each of the plurality of cross-sectional surfaces including cross-sectional surfaces 52, 53, and 54.

[0207] In another example, the pixel value at pixel C of the virtual cross-sectional image slice 121 is interpolated from two next cross-sectional surfaces (here, the first cross-sectional surface 52 and the second cross-sectional surface 53). The first pixel value A of the first cross-sectional image slice with index n-1 and the second pixel value B of the second cross-sectional image slice with index n are located at the same absolute lateral coordinates of the first and second cross-sectional surfaces 52 and 53, and are used to calculate the pixel value at pixel C in the virtual cross-sectional image slice. The interpolation can be, for example, linear interpolation or weighted interpolation, and the probability of the pixel value at pixel C is calculated. Since the expected pixel value at pixel C is typically a binary result, which is the pixel value of the first pixel A or the second pixel B, representing the measurement result for a specific material, the interpolation can be combined with a thresholding operation.

[0208] In the example, more than two pixel values ​​from more than two cross-sectional image slices of corresponding adjacent cross-sectional surfaces can be used for interpolation, for example, by high-order polynomial interpolation. In this example, the pixel size on the virtual cross-sectional image slice 121 is selected to be equal to the pixel size of cross-sectional image slices n-1, n, and n+1 obtained from the plurality of cross-sectional surfaces 52, 53, and 54. For example, the scanning of the third cross-sectional surface 54 by the imaging charged particle microscope 44 (e.g., electron beam 44) is accordingly adjusted in the y-direction such that the third cross-sectional image slice with label n+1 is generated relative to a common transverse coordinate system having the same absolute pixel coordinates as the first and second cross-sectional image slices with labels n-1 and n. In an alternative example, the transverse pixel raster of at least one of the plurality of cross-sectional image slices is digitally adjusted and resampled such that the virtual cross-section 121 is obtained from at least two cross-sectional image slices by projection or interpolation along the z-direction. In an alternative example, the projection or interpolation direction between the first pixel position A on the first cross-sectional surface and the second pixel position B on the second cross-sectional surface is tilted relative to the projection direction (here, the z-direction), and the pixel value at pixel position C of the virtual image slice is interpolated accordingly.

[0209] By employing a method of calculating at least a virtual cross-sectional image slice through interpolation in a predetermined projection direction (e.g., the z-direction perpendicular to the wafer surface 55 and parallel to the orientation of a first set of semiconductor features), interpolation artifacts are reduced, and the virtual cross-sectional image is an accurate representation of the cross-section of the first set of semiconductor features. The interpolation method in a projection direction parallel to the orientation of the first semiconductor feature is particularly useful for analyzing HAR structures in a wafer, where it is impossible to obtain a cross-sectional image perpendicular to multiple HAR structures via FIB milling.

[0210] For each virtual cross-sectional image pixel, a subset of at least one cross-sectional image slice is determined by calculating the distance from each of the N cross-sectional image slices in the sequence to the virtual cross-sectional image pixel in the first orientation direction and selecting at least the first cross-sectional image slice with the minimum distance. In the example, a second cross-sectional image slice of the subset of at least one cross-sectional image slice is selected accordingly as the cross-sectional image slice with the second minimum distance. Further cross-sectional image slices of the subset of at least one cross-sectional image slice can be selected accordingly in order of increasing distance from the virtual cross-sectional image pixel in the first orientation direction.

[0211] In the example, a depth map is generated for each virtual cross-section image. A depth map is generated through geometric construction, ensuring the virtual cross-sectional image is oriented perpendicular to a first orientation direction. For each pixel value of the virtual cross-sectional image, distance is evaluated. And by determining the distance The minimum slice index m is used to select at least one subset of cross-sectional image slices. Second and additional cross-sectional image slices of this subset can be selected accordingly based on their distance from the virtual cross-sectional image pixels in the Z direction. Choose according to the order in which they are added.

[0212] In an example of a method for obtaining at least a 2D virtual cross-sectional image, the first orientation direction is the z-direction perpendicular to the wafer surface, and virtual cross-sectional image slices are calculated on a plane parallel to the wafer surface at a constant depth ZV below the wafer surface. For each virtual cross-sectional image pixel having lateral coordinates (x, y), at least a subset of the m-th cross-sectional image slice is determined by selecting at least one cross-sectional image slice with the minimum distance to the plane at depth ZV, such that the distance... All depth maps The minimum value, where the index is... The second and additional cross-sectional image slices, which are a subset of at least one cross-sectional image slice, can be correspondingly arranged according to their distance from the virtual cross-sectional image pixel in the Z direction. Choose according to the order in which they are added.

[0213] In the example, a depth map is generated for each virtual cross-section image. A depth map is generated through geometric construction, ensuring the virtual cross-sectional image is oriented perpendicular to a first orientation direction. For each pixel value of the virtual cross-sectional image, distance is evaluated. And by determining the distance The minimum slice index m is used to select at least one subset of cross-sectional image slices. Second and additional cross-sectional image slices from this subset can be selected accordingly based on their distance from the virtual cross-sectional image pixels in the Z direction. Choose according to the order in which they are added.

[0214] In the example, the pixel values ​​of the virtual cross-section image are interpolated from a subset of at least one cross-section image slice in the first orientation direction via geometric projection. The angle between the virtual cross-section and the product corresponding to the m-th cross-section image slice is determined. And divide the pixel coordinates of the virtual image slice by Calculate the projected lateral coordinates of the m-th cross-sectional image slice in the plane. Next, calculate the pixel values ​​of the virtual image slice by interpolating the pixel values ​​of the adjacent pixels of the m-th cross-sectional image slice to the projected lateral coordinates.

[0215] In the example, the optical axis of the charged particle imaging system used to acquire a sequence of N cross-sectional image slices is oriented perpendicular to the wafer surface, such that for the angle GE between the optical axis and the z-axis perpendicular to the wafer surface, In this example, the lateral coordinates (x, y) of at least one virtual cross-sectional image and the lateral coordinates of the sequence of N cross-sectional image slices can be the same, and the angles... It is approximately equal to the angle GF. This example is as follows: Figure 16 As shown. A method for obtaining at least a 2D virtual cross-sectional image may include the step of forming at least one alignment feature near the inspection volume, the alignment feature being configured to form at least one common cross-sectional image feature for mutual lateral alignment of each cross-sectional image slice. The mutual lateral image alignment step may include subtracting the image distortion bias of each cross-sectional image slice.

[0216] In an example of a method for obtaining at least a 2D virtual cross-sectional image, interpolation includes at least one of feature extraction, thresholding, contour interpolation, or model-based interpolation. Therefore, interpolation with high accuracy for a first semiconductor structure or feature of interest is provided. A third cross-sectional image feature representing a first semiconductor structure oriented in a first orientation direction in the virtual cross-sectional image is interpolated from first cross-sectional image features from a subset of at least one cross-sectional image slice with high accuracy and reduced interpolation artifacts.

[0217] A virtual cross-sectional image is calculated from a set of N cross-sectional image slices at an arbitrary depth ZV within an examined volume using a method for obtaining at least one 2D virtual cross-sectional image or a set of 2D virtual cross-sectional images. For example, the depth ZV is selected based on the depth of layers parallel to the wafer surface, wherein these layers are formed by second semiconductor features oriented in a second orientation direction parallel to the wafer surface. For example, in two adjacent equivalent layers such as word lines (e.g.,...) Figure 5 The first virtual cross-sectional image slice is calculated at a depth ZV1 between layers L2 and L3, and the layer or word line (e.g.) Figure 5 The second virtual cross-section image slice is calculated at a depth ZV2 inside layer L4.

[0218] index Depth map of each cross-sectional image slice The second cross-sectional image features, calculated from cross-sectional image slices, represent a cross-sectional image of a second semiconductor feature oriented in a second orientation direction parallel to the wafer surface. The projection or interpolation method relies on the actual depth information of each pixel in multiple cross-sectional image slices. The depth information or depth map of each pixel in a cross-sectional image slice with index n... It is derived through the methods described in the above embodiments and examples. In the example of HAR structure analysis, the depth corresponds to each pixel position in the first cross-sectional image of the HAR structure. It is derived from the second cross-sectional image features, which represent structures that extend parallel to the wafer surface, such as multiple layers or word lines of a 3D memory device as described above.

[0219] A collection of virtual image slices (e.g., slice 121) provides a reliable representation of cross-sectional images of the investigated semiconductor structure or feature (e.g., a HAR structure) and can be used for statistical analysis or defect analysis. In an example of a method for obtaining a collection of 2D virtual cross-sectional images, faster data acquisition is achieved, and the distance dz in the z-direction, or the distance d perpendicular to each cross-sectional image slice, is chosen to be substantially larger than the pixel size of the corresponding cross-sectional image. Pixel sizes are typically less than 2 nm, such as 1 nm, 0.5 nm, or even 0.25 nm. For rapid data acquisition, the slice distance dz perpendicular to the wafer surface is selected and adjusted to be greater than 6 nm, such as 8 nm, 10 nm, or even larger. By selecting slice thicknesses or distances dz significantly larger than the image pixel size, the number of milling operations using FIB is significantly reduced, and rapid 3D image acquisition of the wafer interior inspection volume is achieved. The voxel dimensions of the 3D volumetric data have significantly different extensions along different axes, and the pixel size or resolution in the lateral direction differs from the pixel size or resolution in the depth direction. This method enables the creation of multiple virtual cross-sectional image slices with predetermined distances in the z-direction adapted to the lateral pixel spacing in the x and y directions, and generates isotropic 3D voxels of a 3D volumetric image for examining the volume. Since interpolation is performed in a first direction of the semiconductor feature of interest (e.g., a HAR structure) and can be optionally combined with thresholding operations or other methods of contour interpolation, interpolation artifacts are reduced. Using the described method, a first cross-sectional image of the semiconductor element in the virtual image slice is calculated from a first cross-sectional image of a subset of multiple cross-sectional image slices, where the depth distance, or z, is larger than the lateral pixel spacing.

[0220] In the example, a predetermined set of virtual cross-sectional images is derived from a reduced subset of planes orthogonal to the direction of the HAR structure, such as virtual cross-sectional image 121 including a first cross-sectional image of the HAR structure. The virtual cross-sectional image slices are calculated for a small number of z-planes parallel to the wafer surface, for example, lateral pixel gratings with a z-interval dz greater than, for example, 1 nm. In the example, the z-interval of multiple virtual cross-sectional images is selected based on the number of layers parallel to the wafer surface, and the computation time is significantly reduced. Therefore, the distance in the z-direction between the reduced sets of virtual cross-sectional surfaces is selected, for example, between 5 nm and 50 nm, such as 10 nm or 25 nm. In the example, the distance in the z-direction is variable and depends on the depth of the layers extending parallel to the wafer surface.

[0221] In an example of a method for obtaining a set of 2D virtual cross-sectional images, multiple N cross-sectional image slices are generated by scanning multiple N cross-sectional surfaces using an imaging charged particle microscope, and a set of F virtual cross-sectional image slices is calculated, where F < N. In the example, the maximum number of layers in the semiconductor memory stack is M (see...). Figure 3 In one example, F is chosen to be equal to M, such that virtual cross-sectional image slices are generated within each layer. In another example, F is chosen to be equal to the number of word lines, such that virtual cross-sectional image slices are generated within each word line layer and each isolation layer therebetween. In each example, the number of horizontal pixels far exceeds the number of cross-sectional image slices N and the number of virtual cross-sectional image slices F.

[0222] Multiple precise planar cross-sectional surfaces are generated through high-precision milling, including subsequent polishing of multiple cross-sectional surfaces via FIB. However, precise milling, including polishing, is a time-consuming process. In high-throughput applications such as wafer inspection, it is desirable to acquire 3D volumetric images of the inspection volume within the wafer at higher speeds. In the thirteenth embodiment of the invention, if a rapid milling operation is applied, the surface shape of each individual cross-sectional surface may deviate from a perfectly planar surface after rapid milling due to milling artifacts (e.g., introduced by FIB divergence, imperfect FIB control) or due to the effects of the interaction between the FIB and the milling material within the inspection volume. An example of this effect is the well-known curtain effect. Figure 17 A modification to the interpolation method described for generating virtual slices is shown. The pixel value at point C in virtual slice 121 is obtained by interpolating the pixel values ​​of pixels at points A and / or B of cross-sectional image slices obtained from curved cross-sectional surfaces 52, 53, and 54, where the pixel values ​​of the nth cross-sectional image slice are interpolated. The z-coordinate is derived using the method of the present invention described above. The depth or z-coordinate of an image pixel representing, for example, a HAR structure, is derived from a known depth or relative to an image pixel representing, for example, a word line perpendicular to the orientation of the HAR structure. In an example, a method for obtaining a virtual cross-sectional image or a sequence of virtual cross-sectional images from a set of cross-sectional image slices, each virtual image slice comprising a plurality of virtual cross-sectional image pixels, the method comprising the steps of: obtaining a sequence of N cross-sectional image slices by alternately imaging and milling a sequence of N cross-sectional surfaces into an inspection volume inside a wafer at an angle GF, wherein for each virtual cross-sectional image pixel, a pixel value is calculated by projection of at least a subset of the N cross-sectional image slices of the sequence of cross-sectional image slices in a first orientation direction and by interpolation of pixel values ​​from the projection of the at least a subset of the cross-sectional image slices. The number of cross-sectional image slices N is at least N = 10, preferably N > 100, and even more preferably N is about 1000 or greater.

[0223] In the above embodiments, the requirement for inspection tasks with high resolution and large depth extension (LZ) in the wafer is achieved through slicing and imaging methods under a wedge cutting geometry with an angle (GF) greater than 30°. As in the fifth embodiment, combined with... Figure 8 As described, for high-resolution requirements, SEM must acquire a large number of images with different focal positions. For example... Figure 12 In conjunction with the eighth embodiment, examples of embodiments using SEM for imaging also require image processing, such as stitching and distortion compensation. In another example, the optical axis of the SEM is adjusted by angle GE such that the angle GFE between the SEM and the FIB is approximately 90° (see...). Figure 2 and 4 (b) However, in embodiments and examples with SEM, image acquisition and lateral alignment of multiple N cross-sectional image slices with a large depth extension exceeding, for example, 1 µm are time-consuming and require additional time for focus adjustment and image processing. Furthermore, accuracy is also reduced. In the fourteenth embodiment, the charged particle imaging device is a helium ion microscope (HIM) with a depth of field up to 10 m. The depth of field of a high-resolution SEM is approximately DOF ~ 10-20 nm. Typically, the DoF of a charged particle imaging microscope is limited by the desired resolution. Using HIM, at a large depth of field greater than 6 μm, the resolution remains below 1 nm and reaches 10 μm or greater. HIM offers a further advantage: imaging is dominated by secondary electrons and less by backscattered charged particles. Therefore, the imaging contrast is less susceptible to morphological effects. Furthermore, using helium ions, any charge on the wafer is always positive, and contrast variations due to charging effects on the wafer surface are minimized.

[0224] Therefore, the dual-beam device 1 with HIM according to FIG. 18 provides a preferred embodiment for inspection volumes with large depth extensions. In this fourteenth embodiment, a wafer inspection apparatus and method are provided for three-dimensional circuit pattern inspection of an inspection volume within a wafer without removing a sample from the wafer. This apparatus is configured and suitable for thick or deep stacks of layers, such as having a depth extension or height of approximately 6 µm (e.g., 10 µm or 15 µm) below the surface of a processed semiconductor wafer. As described above, slicing and imaging methods in a wedge-cut geometry avoid damage to the wafer or removal of the inspection volume. In the fourteenth embodiment, with an inclination angle GF of 20°, preferably greater than 30°, and even more preferably greater than 35°, for milling the FIB pillars into the wafer surface 55, a lateral extension of the cross-sectional surface of less than 25 µm is achieved, for example, a lateral dimension LY of approximately 20 µm or less, and image acquisition is performed very quickly and efficiently without image stitching. The challenge of wafer inspection tasks is the requirement for high resolution, such as 2 nm, 1 nm, or 0.5 nm or even below 0.5 nm, and the required depth range within the wafer reaching 15 µm below the top surface of the wafer. According to the fourteenth embodiment, high resolution and the required depth of focus for the charged particle imaging beam are achieved by using a helium ion microscope (HIM). The helium ion microscope provides the required resolution of approximately 1 nm or less at the required large depth of focus and allows for image acquisition of each cross-sectional surface under a wedge-cut geometry at the required resolution in a single image. A “single-image scan” refers to an image acquisition scan without changing the focal plane or performing challenging focus control or image stitching over a large depth range. During a single image scan, the wafer does not move, and the focal plane of the charged particle imaging beam does not change.

[0225] By utilizing a dual-beam apparatus for wafer inspection employing HIM (Heat-In-Image) technology, high-resolution and high-throughput 3D volumetric inspection of the inspection volume is provided, particularly for deep inspection volumes with depths greater than 100 nm and resolutions less than 1 nm, such as depths LZ greater than 1 µm. In the example, inspection of a memory device with a HAR structure extending approximately 6 µm below the wafer's top surface is provided without damaging the wafer or removing the inspection volume from it. HIM in a specific arrangement of the dual-beam apparatus provides high-resolution imaging of multiple N cross-sectional surfaces with depths exceeding 5 µm (e.g., 6 µm) and generates multiple N cross-sectional image slices with resolutions less than 1 nm.

[0226] By alternately slicing multiple cross-sectional surfaces with Ga-FIB 50 at approximately 36° and imaging with HIM, the slicing and imaging were implemented to the bottom of the HAR channel of the memory device. Figure 18a and Figure 18bDetails of the fourteenth embodiment are explained. Both figures illustrate different aspects of the dual-beam apparatus 1 used for examining deep volumes within a wafer. A helium ion microscope (HIM) 140 with an optical axis 142 is perpendicular to the wafer support stage (see figure). Figure 18b The wafer support stage holds the wafer, with its top surface 55 perpendicular to the optical axis 142 of the HIM 140. Gallium-FIB pillars 50 are oriented with an angle of inclination GF of approximately 36° relative to the top surface 55. Within an inspection volume of approximately 10µm × 10µm × 6µm lateral extension LX × LY × LZ, first and second cross-sectional surfaces among a plurality of N cross-sectional surfaces are indicated by indices n and (n+1). These cross-sectional surfaces show cross-sectional features, in this example, where the first cross-sectional feature of the HAR structure is oriented perpendicular to the top surface 55 of the wafer. The first HAR cross-section 177.1 is close to the wafer surface 55 with a shallow depth, and the second HAR cross-section 177.2 is at a greater depth, such as approximately 6µm or more at a depth LZ. Both surface details are within the depth of field or depth of focus (DoF) of the HIM 140 of approximately 10µm or more. Due to the 36° angle of inclination, for a depth LZ of 6µm, the lateral extension of the cross-sectional surface LY is only about 10m. In another example, for to the depth, to obtain Lateral extension. For approximately This LZ depth structure, with a lateral image extension of, for example, 15µm, is possible at larger tilt angles, such as 45°. The slice distance dz is, for example, approximately 10nm or selected according to sampling requirements, and a smaller slice distance can be selected for large tilt angles. In each case, each cross-sectional image surface can be acquired by a single image scan of HIM without changing the focus, and without the need for distortion compensation or image stitching.

[0227] To align multiple N cross-sectional image slices, alignment marks, such as alignment mark 148, are provided on the top surface 55 of the wafer. Both the alignment marks on the top surface 55 of the wafer and the angled cross-sectional surface are well within the DOF of the HIM, and the multiple N cross-sectional image slices can be aligned with high precision in the lateral direction. For example, the second HAR cross-section 177.2 at a large depth and the alignment feature 148 on the wafer surface 55 are both traversed by the fast scan line 152 of the HIM 140 and are both within the depth of focus (DoF) of the HIM 140. The unique advantage of the large depth of field of the HIM 140 is that it can image structures or features at different depths within a depth range exceeding 1µm (e.g., 2µm, 5µm, or even 10µm), and the alignment feature or reference 148 fabricated on the top of the wafer surface 55.

[0228] A dual-beam apparatus for inspecting the volume at the inspection location of a wafer with a wedge-shaped cutting geometry 1 (see Figure 18a The system includes: a six-axis wafer stage 155 having a wafer support stage 15 configured to hold a wafer having a wafer surface 55 on a support surface 152 of the wafer support stage 15; a FIB column 50 arranged at an angle GF relative to the support surface 152 of the wafer support stage 15; a helium ion microscope (HIM) 140 having an optical axis 142 arranged perpendicular to the support surface 152, the optical axis of the FIB column 50 and the HIM 140 forming an intersection point 43; and a platform control unit 16 configured to position a first measurement position 6.1 of the wafer 8 at the intersection point 43 during use; and a control unit 19 for controlling the FIB column 50 and the HIM 140, configured to alternately mill multiple N cross-sectional surfaces in the inspection volume with an FIB beam 51 at an angle GF and image each cross-sectional surface with the HIM beam 144 to form a high-resolution multiple N cross-sectional image slice with a resolution better than 2 nm, preferably better than 1 nm, thereby allowing the inspection volume to be milled in the HIM beam 144. The optical axis 142 of the HIM 140 has a depth extension LZ greater than 1µm, for example, 3µm. In the example, the depth extension LZ in the direction of the optical axis 142 of the HIM 140 exceeds 5µm, for example, 6µm or even 10µm. Imaging of each cross-sectional surface is achieved by the HIM beam 144 having a depth of focus (DOF) exceeding the depth extension LZ to form multiple high-resolution N cross-sectional image slices. The control unit 19 is configured to scan the HIM beam 144 over a region with a lateral dimension LX or LY of an examination volume exceeding approximately 5µm to 10µm (LX or LY) during use, and to collect multiple secondary electrons in a time-sequential manner using the secondary electron detector 17.

[0229] In the example, the angle GF formed between the FIB beam 51 and the support surface 152 of the wafer support stage 15 is between 30° and 60°, for example, 36° or 45°. In the example, the dual-beam device 1 also includes an operation unit 2 configured to align a plurality of N cross-sectional image slices with at least one common cross-sectional image feature. In the example, the control unit 19 is also configured to generate at least one alignment mark 148 on the wafer surface 55 during use, forming at least one common cross-sectional image feature for the mutual alignment of the plurality of N cross-sectional image slices. In the example, the operation unit 2 also includes an image processing unit 82 (see...). Figure 15 (), used to calculate at least a virtual cross-sectional image slice in a plane parallel to the wafer surface 55 of the inspection volume during use.

[0230] Utilizing the large depth of focus and telecentric configuration of the HIM 140, cross-sectional image slices can be imaged with high precision and without distortion without altering the focal plane of the charged particle imaging microscope. Leveraging a large depth extension exceeding 1µm, such as 6µm or even up to 10µm, a highly efficient and rapid workflow allows for the acquisition of 3D information about the inspection volume within the wafer. For example, for memory devices with multiple HAR structures, virtual image slices can be generated in each alternating layer, such as the word line layer and the isolation layer between word lines. Since the cross-sectional images are obtained from a tilted cross-sectional surface, where the HIM is perpendicular to the wafer surface or parallel to the HAR structure, image processing time is reduced and interpolation artifacts are minimized. Optional alignment structures 148 formed on the wafer surface 55 are precisely within the depth of focus of the HIM and can be used for mutual alignment of cross-sectional image slices.

[0231] Figure 19 illustrates some aspects of a fourteenth embodiment of an inspection volume for a large z-extension LZ. The inspection volume 160 has a large z-extension LZ between 2µm and 10µm. In the inspection volume of the memory device, a plurality of alternating layers 162 are arranged parallel to the wafer surface 55. A plurality of HAR structures are arranged orthogonally to the wafer surface 55. A plurality of N cross-sectional surfaces are formed in the inspection volume 160. The cross-sectional surfaces at indices (n-1), n, and (n+1) are illustrated. Each cross-sectional surface is milled into the inspection volume at an angle GF (e.g., 36°) approximately between 30° and 60°. The distance between consecutive cross-sectional surfaces is chosen to be less than 12nm. By controlling the slice distance, for example by a smaller distance and a larger number of slices, the accuracy of 3D volumetric image acquisition can be further improved, and image processing errors can be reduced. Alternating with milling, each cross-sectional surface is imaged by HIM 144, and a plurality of N cross-sectional image slices are formed, such as the cross-sectional image slice at index n in Figure 19b. The cross-sectional image slices include multiple cross-sectional features of the HAR structure, as shown in 177.1, and cross-sectional features of multiple layers, such as the conductive layer cross-section 175. Each of the multiple N cross-sectional image slices includes image detail 150 representing additional alignment feature 148, and each of the multiple N cross-sectional image slices is laterally aligned with image detail 150 representing additional alignment feature 148, which is not corrected by subsequent milling of the cross-sectional image surface. Virtual cross-sections are calculated in the first conductive layer or word line to form virtual cross-sectional image 166.1, and virtual cross-sections are calculated in the second isolation layer to form virtual cross-sectional image 166.2. Thus, multiple virtual cross-sections are calculated (Figures 19c and 19d).

[0232] like Figure 20 As shown, image processing is used to analyze each of multiple cross-sectional images or virtual cross-sections, and statistical characteristics of the HAR structure are derived. Figure 20Two examples of the analysis results are shown. On the left, the distribution of the radius and the average radius of multiple HAR channels 164 within a depth range from 0µm to 4.5µm is shown. On the right, the distribution of the ellipticity and the average ellipticity of multiple HAR channels 164 within a depth range from 0µm to 4.5µm is shown. Other parameters of the multiple HAR channels 164 can be obtained and shown in a similar manner.

[0233] Figure 21 A fifteenth embodiment of inspection of an inspection volume with a large depth extension is shown. In this embodiment, the time required to inspect an inspection volume with a large depth extension is reduced by dividing the inspection volume into multiple blocks B 181.1, ... 181.b, ... 181.B. For simplicity, three blocks are illustrated. The multiple blocks 181 are arranged diagonally along the milling direction of the FIB beam. Therefore, the number of multiple cross-sectional surfaces to be milled and imaged is reduced to a lower number, and the time required for inspection is reduced. In each block 181, multiple, for example, 70 HAR features 164 are analyzed in a specific depth range lz1 to lzB. The block size is adjusted according to the milling angle GF of the cross-sectional surface, the HAR feature size, and the number of HAR cross-sections 177 required for statistical analysis in each depth range. Therefore, the preferred slice distance d is less than 30 nm. Multiple virtual cross-sections are calculated in different depth ranges of multiple layers 162. Therefore, the number of milling operations is reduced by more than three times, and the operation time required to inspect the volume is reduced by more than three times. In the example, for a memory device that includes a memory HAR structure, it is sufficient to provide at least three cross-sectional image slices for each HAR structure.

[0234] Figure 22 A sixteenth embodiment of a dual-beam apparatus for inspecting a deep and small inspection volume 191 over a large depth extension is shown. In this example, the angle GF is selected as a large angle between 45° and 80°, such as a milling angle of approximately 76°, to slice the HAR features of the memory device at a large angle DF across the entire depth range. The axis of the charged particle microscope 195 is arranged at an angle GE to the normal of the wafer surface 55, such as 40° or greater. In this example, wafer damage is minimized.

[0235] Typically, wafer inspection for monitoring and controlling semiconductor features on or within semiconductor wafers in high-volume manufacturing (HVM) requires very high speeds and throughput. In the seventeenth embodiment, the development of a monitoring formulation with a dual-beam device according to any other embodiment of the invention is described. Figure 23The seventeenth embodiment is shown. The development of a semiconductor device generally follows three phases from M1 to M3. In the Research and Development (RnD) phase M1, a functional prototype of the new semiconductor device is developed. In step M1.1, the design of the semiconductor device is developed. In step M1.2, a functional demonstrator of the new semiconductor device or a new design feature of the new semiconductor device is manufactured, and its functionality is tested in step M1.3. During phase M1, new design features and new process steps are tested. For example, according to any of the embodiments described above, or using the sample detachment method described in the prior art, a 3D volumetric analysis of the demonstrator is performed. This determines and provides a list of key design parameters and key design performance indicators D1.

[0236] In stage M2, improvements to the manufacturing process are provided. During step M2.1, new manufacturing processes are developed, or existing processes are improved. In step M2.2, comprehensive 3D analysis or 3D volumetric imaging is required, for example, through any of the methods or apparatus described in the above embodiments. For example, a new set of representative defect patterns is obtained, and a list of key manufacturing performance indicators D2 is generated from the representative defect patterns. If the manufacturing during stage M2 proves critical, the process can return to stage M1, and design changes can be implemented.

[0237] Comprehensive 3D measurements were performed in both stages M1 and M2 to understand the process and fully identify potential defects. In step M2.2, multiple manufacturing steps were monitored and multiple 3D measurements were performed; it is advantageous if the measurement time during step M2.2 is shorter than, for example, the measurement time in step M1.3. A preferred method for 3D inspection during stage M1 or M2 is the generation of 3D volumetric images under a wedge-cut geometry, as described in one of the embodiments above. In the "wedge-cut" method, as... Figure 3 As shown, a 3D memory stack is milled using FIB at an angle less than 80° to the wafer surface. In an example of a memory device, the inclined cross-sectional surface 52 of the trench contains memory channels and word line footprints. The cross-sectional surface is imaged from the "top" using a vertical charged particle imaging beam (e.g., a helium ion or electron beam). The corresponding cross-sectional image slice contains projections of the channel footprints at corresponding depths (Z) in the XY plane. In the case of cylindrical vertical memory channels, the cross-sectional image slice contains circular footprints at different depths determined by the positions of the footprints on the inclined surface 52 of the trench, such as... Figure 6As shown. The angle GF between the FIB beam 51 and the wafer surface (XY plane) controls the depth range covered by the imaging beam's field of view and the total lateral area available for imaging. "Top-down" imaging of an inclined cross-section surface requires a certain depth of focus (DoF) for the vertical beam. In cases where the DoF is insufficient, the FoV can be divided into smaller subfields, and sequential imaging can be achieved using progressive focus adjustments, as shown in... Figure 8 The fifth embodiment in the examples is explained. Alternatively, the HIM according to the fourteenth embodiment described in FIG18 can be used. Based on the measurements, multiple performance parameters and statistical analyses can be performed, such as generating virtual cross-sectional image slices, and the evaluation can include, for example... Figure 20 As described in [the text].

[0238] Phase M3 is High-Volume Manufacturing (HVM) of semiconductor devices. In the HVM phase, measurement time is reduced, which helps maximize throughput. In step M3.1, the manufacturing process developed in phase M2 is applied, and in step M3.2, a limited number of metric or monitoring measurements are routinely performed. Based on the key design performance metrics D1 and key manufacturing performance metrics D2 derived during phases M1 and M2, and as described above, a measurement or monitoring formula R is defined and then routinely executed in phase M3 to monitor the manufacturing process.

[0239] Based on key design performance metrics D1 and key manufacturing performance metrics D2, a measurement or monitoring recipe R is generated for rapid and high-throughput monitoring of the HVM stage M3. Key design performance metrics D1 and key manufacturing performance metrics D2 can include a large amount of data, including 2D and 3D image data, 2D virtual image slices, and dimensional and statistical analyses of data from semiconductor devices manufactured with ideal performance and those manufactured with performance degradation due to typical manufacturing errors. Key design performance metrics D1 and key manufacturing performance metrics D2 can also include simulations provided from databases, such as CAD simulations or analytical data from other semiconductor devices. Representative defect patterns and specific measurement tasks in the inspection volume at a specific inspection location are derived and adhered to in the measurement recipe R. For example, volumetric effects in the inspection volume can typically be represented by one or two representative cross-sectional image slices of the inspection volume. Based on the 3D volumetric image obtained through the above-described slicing and imaging methods, the preferred angle and orientation of at least one representative cross-sectional image slice are determined, and the preferred angle and orientation of the cross-sectional surface at a specific inspection location on the wafer are also determined. The preferred angle and orientation of the cross-sectional surface at a specific inspection location are examples of prior information that utilizes prior knowledge of the semiconductor structure and features within the inspection volume.

[0240] The measurement recipe R used for HVM is derived, for example, by means of machine learning or deep learning applied to key design performance metrics D1 and key manufacturing performance metrics D2, and defines a simplified set of representative measurements for HVM in stage M3, including a set of monitoring or HVM performance metrics D3 for monitoring recipe R. During HVM in step M3, the measurement or monitoring recipe R is applied, and multiple actual values ​​of monitoring metrics D3 are generated. The actual values ​​of HVM performance metrics D3 can include a large amount of data, including 2D and 3D image data, semiconductor devices manufactured with ideal HVM performance, and semiconductor devices manufactured with performance degradation due to HVM manufacturing errors, along with the measurement dimensions and statistical analysis of these data. Multiple actual values ​​of HVM performance metrics D3 can be fed back to the measurement or monitoring recipe generation in step R, and the measurement or monitoring recipe generation in step R can be verified or implemented at regular time intervals.

[0241] Therefore, the method for generating a measurement or monitoring recipe according to the seventeenth embodiment includes the following steps: generating a first set of critical design performance indicators D1, the set of critical design performance indicators D1 including CAD image data and dimensions of critical design features in an inspection volume; generating a second set of critical manufacturing performance indicators D2, including 3D volumetric image data obtained during manufacturing process development; and deriving a measurement recipe R, the measurement recipe R including a step for obtaining at least one of a third set of performance indicators D3, wherein the step of generating the second set of critical manufacturing performance indicators D2 or the monitoring recipe R includes obtaining at least a first cross-sectional image slice of at least a first cross-sectional surface through a representative predetermined inspection volume, wherein the depth extension below the wafer surface... At least one of the steps of obtaining at least one performance metric D1 to D3 includes an image processing step to extract multiple first cross-sectional image features representing a cross section of multiple first semiconductor features at an angle GF, the image processing including at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation. At least one of the steps of obtaining at least one performance metric D1 to D3 includes a step of calculating at least one descriptive parameter of at least one first semiconductor feature from at least one of the multiple first cross-sectional image features, wherein the descriptive parameter is one of size, diameter, angle, area, shape, or volume. At least one of the steps of obtaining at least one performance metric D1 to D3 further includes a step of calculating one of the average or statistical deviations of at least one descriptive parameter of the multiple first semiconductor features. Measurement scheme R may include measurements according to any of the above embodiments, but preferably the method described in the eighteenth embodiment. In HVM stage M3, a full 3D tomography scan may consume an unacceptable amount of time, while a single wedge section is possible. Utilizing the monitoring recipe generation described above, a single wedge cut can provide the information required for process monitoring. In the eighteenth embodiment, a method for extracting 3D information from the analysis of a single wedge cut through a 3D memory stack is provided.

[0242] The single wedge cut monitoring method according to the eighteenth embodiment provides information about a single tilted plane 51 within a 3D memory stack. Compared to 3D tomography of a stack using parallel slices, such as those employing sequential milling and imaging, the information obtained from a single cut is insufficient for a complete 3D reconstruction of the stack geometry. However, the single wedge cut method offers the advantage of shorter milling and image acquisition times compared to milling and imaging multiple slices. Therefore, applying the single wedge cut monitoring method in HVM stage M3 is advantageous. The orientation of the FIB and imaging beam allows for the milling of localized trenches without further damage to the wafer (the wafer sample does not "detach"), which is crucial for in-line wafer inspection. Figure 24A single wedge-cut monitoring method is illustrated in an example of multiple HAR structures in a memory device. A single interface 51 is milled by a FIB 55 at an angle GF relative to a wafer surface (not shown) parallel to the y-axis. A charged particle imaging device (not shown) is arranged in the z-direction and generates a digital image of surface 51, including multiple cross-sectional image features 203.1…203.4 of HAR structures 205.1…205.4 oriented perpendicular to the wafer surface. By combining the footprints 203.1 to 203.4 of the cross-sectional image features of the individual memory or HAR channels 205.1…205.4, a reconstruction of the representative vertical HAR profile 201 of the average HAR channel is obtained. To reconstruct the vertical profile of the representative vertical HAR profile 201 (diameter or more generally, the shape of the representative vertical HAR profile as a function of Z), the imaging footprints 203.1 to 203.4 of the individual channels at tilt angles corresponding to angle GF can be combined to form a shape such as… Figure 24 The representative vertical HAR profile 201 is shown. The Z coordinates of the image pixels in the digital image can be reconstructed from their XY coordinates on the top-down image, for example, using the methods described above, such as... Figure 5 , 6 The method utilizes cross-sectional image features of word lines in the inclined surface 51 shown in Figures 12 and 19b. Prior information generated in stage M2 is applied, such as information that all channels within the field of view ideally have the same vertical profile or predefined variations in the vertical profile. In a single cross-sectional surface, the cross-sections of the HAR structure lie at different depths, and the depth-dependent characteristics of the HAR channels can be trained or learned from the 3D volumetric inspection in stage M2. In a simple example, the diameter of the HAR channels remains constant with respect to Z, and the axes of all HAR channels are parallel to each other. If a more complex vertical profile of the average channel is predicted from the 3D volumetric inspection during stage M2, the assumed profile can be applied to correct for critical dimensions and proximity measurements at different depths.

[0243] If we assume the channels form a regular grid in the XY plane, the wedge-cut data also allows the lateral displacement of the average channel 201 to be reconstructed as a function of Z (the so-called "tilt" or "sway"). In the example, the memory channels are designed in step M1 to form a hexagonal grid in the lateral plane. For each channel footprint, including footprints or cross-sectional image segments 203.1 to 203.4 in the image, the centroid is measured. (For multiple HAR channels) Each cross-sectional image segment has an index i). As mentioned above, word lines can be used for any pair. Determine the Z coordinate of the centroid of the footprint If all channels in FoV have the same shape and orientation, then the centroid of the channel footprint at a given z is... The lateral position can be defined as

[0244]

[0245] and

[0246]

[0247] in and It describes the lateral displacement of the average channel with depth; and Describes a certain reference depth The ideal lateral position of the channel. Through this reconstruction, the function... and The tilt and "oscillation" of the average channel in FoV were determined. This can be accomplished by solving the following overdetermined system of equations (e.g., by...). minimize):

[0248]

[0249] for and Assume they form a predefined regular grid, such as a hexagonal grid. Therefore, for a sufficient number of measured channel centroids, the system of equations can be solved to find... and .

[0250] Therefore, digital images can be used to reconstruct the 3D geometry of a 3D memory stack under certain statistical assumptions or prior information obtained during 3D volumetric inspection in stages M1 and M2. For example, defect types or defect features can be identified from digital images of a single surface 51 using prior information or machine learning methods. A set of representative digital images of a single surface is obtained from 3D volumetric images of typical defects in stages M1 and M2 and used to train a machine learning algorithm. The 3D volumetric defect type is determined from the digital images of a single intersecting surface 51 based on a single surface and analysis according to prior information or machine learning algorithms. Figure 25 A simplified example is shown. Figure 25 Figure a shows multiple HAR structures (three indicated by reference numeral 205) within an examination volume 160 extending laterally LX and LY, and their corresponding depth range LZ. The multiple HAR structures are tilted relative to the z-axis. Figure 25 b illustrates a first wedge cut of the first cross-sectional surface 51.1 milled by FIB 50 oriented in a local wafer coordinate system in the yz plane. A first digital image 207.1 obtained by charged particle imaging device 40 or 140 is shown in... Figure 25As shown on the right side of b, multiple cross-sectional image features 203.1 represent cross-sections of multiple HAR structures 205. The tilt of the multiple HAR channels 205 is derived from a digital image of the cross-sectional surface 51.1 by varying the spacing between multiple lines 209.1 of the parallel rows of HAR structures parallel to the x-direction, along with prior information about the design distance of the HAR channels.

[0251] To further analyze the channel tilt, a second cross-sectional surface 51.2 is generated and imaged, wherein the orientation of the wedge cut changes between the first and second cross-sectional surfaces. For example, between the generation of the first intersecting surface and the generation of the first digital image, and between the generation of the second intersecting surface and the generation of the second digital image, the two intersecting surfaces 51.1 and 51.2 are etched and imaged as the wafer rotates about the wafer normal or z-axis. Thus, two intersecting surfaces are formed, wherein the first intersecting surface is rotated, for example, by 90° relative to the second intersecting surface. The results of the generation of the second intersecting surface and the generation of the second digital image are... Figure 25 c is shown in a local wafer coordinate system. For illustration, the wafer is not rotated, but the orientation of the imaging device is rotated so that the FIB pillar 50 is now oriented in the xy plane of the local wafer coordinate system. The second cross-sectional surface 51.2 is again generated at angle GF, but rotated by a predetermined angle about the z-axis relative to the first cross-sectional surface 51.1. In this example, the predetermined angle is 90°, but other angles are possible. The corresponding second digital image 207.2 of the second cross-sectional surface 51.2 is again obtained by the charged particle imaging device 40 or 140 and includes multiple cross-sectional image features 203.2, which represent cross-sections passing through multiple HAR structures at different depths. The line 209.2 through the parallel rows of HAR structures is now tilted relative to the x-axis, and the tilt of the multiple HAR channels 205 is derived from it, together with prior information about the design distance of the HAR channels. Through the first and second inspections under the wedge-cut geometry, the tilt angles and orientations of the multiple HAR structures in the inspection volume 160 can be fully determined from only two cross-sectional measurements.

[0252] In the example, the tilt of HAR channel 205 could be a result of an misaligned etching process, and the tilt angle of the HAR structure depends on its location on the wafer; for example, the radial tilt angle increases with increasing distance from the wafer center. Other examples are systematic errors in alignment, according to which all HAR structures remain parallel in one direction with equal tilt angles. By generating several representative cross-sectional surfaces through milling and imaging at predetermined locations on the wafer, the characteristics or origin of defects can be derived.

[0253] In the example above, the multiple HAR structures within the examined volume are tilted and parallel. Therefore, the cross-sections of the HAR structures exhibit different spacing in the x and y directions. In other examples, for instance, if the HAR structures are tilted relative to a common center of all HAR axes, or randomly tilted, the characteristics of all tilts can be evaluated by generating at least two cross-sectional surfaces at different orientations, for example, by rotating the wafer between the generation of the first and second cross-sectional images.

[0254] As described in German patent application 102019006645.6 filed on September 20, 2019 and PCT application PCT / EP2020 / 000101 filed on May 15, 2020, channel footprints detected in top-down images of wedge-shaped cut sections can be used to study the lateral variation of channel diameter (CD uniformity) and the proximity of adjacent channels, both of which are incorporated herein by reference.

[0255] Therefore, a method for inspecting at least a first inspection volume of a wafer according to a monitored formulation includes the following steps: loading the wafer onto a wafer support stage in a dual-beam apparatus; moving the wafer support stage so that a first measurement position on the wafer coincides with the intersection of the dual-beam apparatus; milling a first cross-sectional surface in the first inspection volume at an inclined angle using a FIB column; generating a first cross-sectional image slice of the first cross-sectional surface using a charged particle imaging device; and obtaining performance indicators of a plurality of first semiconductor features in the first inspection volume, including the step of analyzing the first cross-sectional image slice with prior information about the plurality of first semiconductor features. The analysis step may include an image processing step to extract a plurality of first cross-sectional image features representing a cross-section of the plurality of first semiconductor features at an inclined angle GF, the image processing including at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation. The step of obtaining performance indicators may further include the step of calculating at least one descriptive parameter of the first semiconductor feature from at least one of the plurality of first cross-sectional image features, the descriptive parameter being one of size, diameter, angle, area, shape, or volume. It may also include the step of calculating one of the average or statistical deviation of at least one descriptive parameter of the plurality of first semiconductor features. According to a third embodiment, the analysis step may further include generating a depth map of the first cross-sectional image slice. To generate a depth map, the method may further include (1) determining at least two second cross-sectional image features in a first cross-sectional image slice, and (2) determining a depth map from the lateral positions of the at least two second cross-sectional image features. At least two second cross-sectional image features represent integrated semiconductor structures at different depths within the inspected volume. In the example, the step of obtaining performance metrics includes deriving tilt angle deviations of multiple first semiconductor features from their lateral positions, where the tilt angle deviation is the angle of the first semiconductor feature relative to an axis perpendicular to the wafer surface. This can be achieved from the tilt angle GF and depth map. Multiple first cross-sectional image features are arranged to generate a 3D representation of a first plurality of semiconductor features. The first cross-sectional image slices can be compared with 2D digital image slices of an inspection volume of a reference wafer or die, wherein the 2D digital image slices are cross-sectional image slices or virtual cross-sectional image slices obtained in previous measurements of the reference wafer or die and stored in memory, and wherein the virtual cross-sectional image slices are generated from 3D volumetric image data stored in memory. The 3D volumetric image data is obtained in previous slice and image measurements of the inspection volume of the reference wafer or die and stored in memory, for example, during stage M2. The previous slice and image measurements can be performed using a second dual-beam apparatus including at least a second FIB column and a second charged particle imaging device, or using the same or a first dual-beam apparatus.

[0256] In an example of monitoring formulation R, the formulation includes moving a wafer support stage to align a second measurement position of the wafer with the intersection of a first dual-beam device, milling a second cross-sectional surface in a second inspection volume at an angle GF, and generating a second cross-sectional image slice of the second cross-sectional surface using a charged particle imaging device. In this example, moving the wafer includes rotation of the wafer support stage relative to an axis perpendicular to the wafer support surface. Performance metrics of the multiple first semiconductor features in the first and second inspection volumes are obtained by analyzing the first and second cross-sectional image slices with prior information about multiple first semiconductor features. In this example, the angle deviation of the multiple first semiconductor features includes analysis of the first and second cross-sectional image slices.

[0257] A wafer defect inspection apparatus includes a focused ion beam (FIB) column and a charged particle imaging device. The FIB column is configured to mill and expose at least a first cross-sectional surface through a first inspection volume in the wafer at an angle GF. The charged particle imaging device is configured to image the at least first cross-sectional surface to form a first cross-sectional image slice. It also includes an image processing unit with software code mounted on it, configured to determine a plurality of cross-sectional image features in the at least first cross-sectional image slice and to determine the depth of the plurality of cross-sectional image features within the inspection volume, the plurality of cross-sectional image features being cross-sections of a semiconductor structure at an angle GF within the inspection volume; and a defect detection unit configured to determine deviations from predetermined characteristics of the semiconductor structure within the inspection volume from the plurality of cross-sectional image features. The wafer defect inspection apparatus is configured to inspect an inspection volume in a wafer, whereby the inspection volume is not extracted from the wafer. The wafer defect inspection apparatus includes a memory for storing prior information.

[0258] The fiber-in-the-wall (FIB) used for milling inclined sections into a wafer is described in some examples as a gallium FIB, which is a common FIB used for milling. Other examples of FIBs for milling may utilize other materials, such as gases generated in a gas field ion source (GFIS), such as hydrogen, helium, or neon. In the examples, the dual-beam device of the present invention includes an FIB for milling and a charged particle imaging beam for imaging, both utilizing inert gases such as He and Ne for milling and imaging.

[0259] The above embodiments and examples are intended only as illustrative examples of the present invention. Although embodiments are described in the example of a semiconductor structure as a probe, the methods and apparatus of at least some embodiments can also be applied to materials or probes of comparable structures that allow the depth to be determined from second cross-sectional image features representing layers or known depths in a sample. Those skilled in the art can make changes, modifications, variations, and combinations to the specific embodiments and examples without departing from the scope defined by the appended terms.

[0260] This invention is further described by the following sets of clauses:

[0261] Clause 1: A method for inspecting at least a first inspection volume of a wafer using a first dual-beam apparatus, comprising:

[0262] --The wafer is mounted on a wafer support stage in a dual-beam apparatus, which includes at least a FIB pillar and a charged particle imaging device. The first optical axis of the FIB pillar forms an angle GF with the surface of the wafer support stage, and the second optical axis of the charged particle imaging device forms an angle GE with the normal to the surface of the wafer support stage. The first and second optical axes intersect.

[0263] --Move the wafer support stage to align the first measurement position on the wafer with the intersection of the dual-beam device.

[0264] --The first section surface is milled at an angle GF in the first inspection volume using a FIB column.

[0265] --Generate a first cross-sectional image slice of the first cross-sectional surface using a charged particle imaging device.

[0266] --The step of obtaining performance metrics for multiple first semiconductor features in a first inspection volume includes analyzing a first cross-sectional image slice using prior information about the multiple first semiconductor features.

[0267] Clause 2: The method according to Clause 1, wherein the first semiconductor feature is one of a via, a HAR structure, or a HAR channel.

[0268] Clause 3: The method according to any one of Clauses 1 or 2, wherein the analysis step includes an image processing step to extract a plurality of first cross-sectional image features representing a plurality of first semiconductor features at a tilt angle GF, the image processing including at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation.

[0269] Clause 4: The method according to Clause 3, wherein the step of obtaining the performance index further comprises the step of calculating at least a descriptive parameter of the first semiconductor feature from at least one of a plurality of first cross-sectional image features, the descriptive parameter being one of size, diameter, angle, area, shape or volume.

[0270] Clause 5: The method according to Clause 4, wherein the step of obtaining the performance index further includes the step of calculating the average or statistical deviation of at least one of the descriptive parameters of the plurality of first semiconductor features.

[0271] Clause 6: The method according to any one of Clauses 3 to 5, wherein the analysis step further includes generating a depth map of the first cross-sectional image slice. .

[0272] Clause 7: The method described in Clause 6 further comprises the step of generating a depth map:

[0273] --Determine at least two second cross-sectional image features in a slice of the first cross-sectional image, and

[0274] --Determine the depth map from the lateral positions of at least two second-section image features. .

[0275] Clause 8: The method according to Clause 7, wherein each of at least two second cross-sectional image features represents an integrated semiconductor structure at different depths within the inspection volume.

[0276] Clause 9: The method according to Clause 7 or 8, wherein each of the plurality of first semiconductor features extends in a direction perpendicular to the wafer surface, and the second cross-sectional image feature includes a cross-section of the semiconductor structure extending in a direction parallel to the wafer surface.

[0277] Clause 10: The method according to any one of Clauses 7 to 9, wherein the second cross-sectional image features include at least a cross-section of one of an isolation line or layer, a metal line or layer, or a semiconductor line or layer.

[0278] Clause 11: The method according to any one of Clauses 6 to 10, wherein the step of obtaining the performance index includes deriving the tilt angle deviation of a plurality of first semiconductor features from the lateral positions of a plurality of first cross-sectional image features, the tilt angle deviation being the angle of the first semiconductor feature relative to an axis perpendicular to the wafer surface.

[0279] Clause 12: The method according to any one of Clauses 4 to 11 further includes, from the tilt angle GF and with depth map The steps of deriving a 3D representation of a plurality of semiconductor features from multiple first cross-sectional image features.

[0280] Clause 13: The method according to any one of Clauses 1 to 12, wherein the analysis step further comprises comparing the first cross-sectional image slice with a 2D digital image slice of the inspection volume through a reference wafer or die.

[0281] Clause 14: The method according to Clause 13, wherein the 2D digital image slice is a cross-sectional image slice obtained from a previous measurement of a reference wafer or die and stored in memory.

[0282] Clause 15: The method according to Clause 13, wherein the 2D digital image slice is a virtual cross-sectional image slice, and wherein the virtual cross-sectional image slice is generated from 3D volumetric image data stored in memory.

[0283] Clause 16: The method according to Clause 15, wherein the 3D volumetric image data is obtained from previous slices and image measurements of the inspection volume of a reference wafer or die and stored in memory.

[0284] Clause 17: The method according to Clause 16, wherein the prior slicing and image measurements are performed using a second dual-beam device, the second dual-beam device comprising at least a second FIB column and a second charged particle imaging device.

[0285] Clause 18: The method described in accordance with Clause 16 or 17, wherein the prior slicing and image measurements are performed by a first dual-beam device.

[0286] Clause 19: The method according to any one of Clauses 1 to 18 further includes:

[0287] --Move the wafer support stage so that the second measurement position of the wafer coincides with the intersection of the first dual-beam device.

[0288] --The second section surface is milled in the second inspection volume using a FIB column at an angle GF.

[0289] --Generate a second cross-sectional image slice of the second cross-sectional surface using a charged particle imaging device.

[0290] --Performance metrics of multiple first semiconductor features in first and second inspection volumes are obtained by analyzing first and second cross-sectional image slices with prior information about multiple first semiconductor features.

[0291] Clause 20: The method according to Clause 19, wherein between milling of the first cross-sectional surface in the first inspection volume and milling of the second cross-sectional surface in the second inspection volume, the wafer support stage rotates relative to an axis perpendicular to the wafer support surface.

[0292] Clause 21: The method according to any one of Clauses 19 to 20 further includes the step of deriving tilt angle deviations of a plurality of first semiconductor features, including analysis of first and second cross-sectional image slices.

[0293] Clause 22: The method according to any one of Clauses 1 to 21, wherein the inclination angle GF of the FIB column is between 30° and 80°, preferably between 30° and 45°.

[0294] Clause 23: The method according to any one of Clauses 1 to 22, wherein the charged particle imaging device is oriented perpendicular to the wafer surface. .

[0295] Clause 24: The method according to any one of Clauses 1 to 23, wherein the charged particle imaging device is a helium ion microscope (HIM).

[0296] Clause 25: The method according to Clause 24, wherein the first cross-sectional image slice is obtained by obtaining an image of the first cross-sectional surface by examining the volume in a single image scan.

[0297] Clause 26: A method for inspecting a wafer using a dual-beam apparatus, the dual-beam apparatus comprising a focused ion beam column (FIB) and a helium ion microscope (HIM), wherein the optical axis of the FIB is arranged at an angle GF between 30° and 45° with respect to the support surface of the wafer stage, and the optical axis of the helium ion microscope is arranged perpendicular to the support surface, the optical axes of the FIB column and the HIM forming an intersection.

[0298] Includes the following steps:

[0299] --The wafer is positioned at a first measurement location using a wafer stage with a wafer support, the wafer support being configured to hold the wafer at the intersection.

[0300] --The first cross-sectional surface is milled through the first inspection volume using an FIB pillar at an angle GF, wherein the depth extension LZ below the wafer surface is greater than 1µm, and

[0301] --The surface of the first cross-section is imaged using a helium ion microscope in a single image scan to form a high-resolution cross-sectional image slice.

[0302] Clause 27: The method according to Clause 26, wherein the first inspection volume has a depth range greater than 2µm, greater than 6µm, or even 10µm.

[0303] Clause 28: The method according to any one of Clauses 26-27, wherein in the imaging step, the HIM is configured to generate an image with a lateral resolution of less than 2 nm, preferably less than 1 nm or even less than 0.5 nm.

[0304] Clause 29: A computer program product having program code for performing the method according to any one of Clauses 1 to 28.

[0305] Clause 30: A dual-beam device configured to perform the method according to any one of Clauses 1 to 28.

[0306] Clause 31: A wafer defect inspection apparatus, comprising:

[0307] --Focused ion beam (FIB) column, configured to mill and expose at least a first cross-sectional surface through a first inspection volume in the wafer at an angle GF;

[0308] --A charged particle imaging device, configured to image at least a first cross-sectional surface to form a first cross-sectional image slice;

[0309] --An image processing unit with software code installed is configured to determine multiple cross-sectional image features in at least a first cross-sectional image slice and to determine the depth of multiple cross-sectional image features within an inspection volume, wherein the multiple cross-sectional image features are cross-sections of a semiconductor structure at an inclination angle GF within the inspection volume;

[0310] --The defect detection unit is configured to determine deviations from predetermined characteristics of the semiconductor structure within the inspection volume from multiple cross-sectional image features;

[0311] The wafer defect inspection equipment is configured to inspect the inspection volume within the wafer, which is not extracted from the wafer.

[0312] Clause 32: The device according to Clause 31, wherein the image processing unit with software code installed is further configured to calculate a 3D representation of a plurality of first semiconductor structures from a plurality of first cross-sectional image features arranged at an angle GF.

[0313] Clause 33: The device described in Clause 31 or 32 further includes a memory for storing prior information.

[0314] Clause 34: The apparatus according to any one of Clauses 31-33, wherein the charged particle imaging apparatus further includes a wafer support stage.

[0315] Clause 35: The apparatus according to Clause 34, wherein the optical axis of the FIB pillar is arranged at an angle GF relative to the wafer support stage, the angle being between 30° and 80°, preferably between 30° and 45°.

[0316] Clause 36: The apparatus according to Clause 34 or 35, wherein the optical axis of the charged particle imaging apparatus is arranged at an angle perpendicular to the wafer support stage.

[0317] Clause 37: The device pursuant to any one of Clauses 31 to 36, wherein the charged particle imaging device is a helium ion microscope (HIM).

[0318] Clause 38: The apparatus according to Clause 37, wherein the inspection volume has a depth extension LZ greater than 1µm, preferably greater than 2µm, below the wafer surface, and the HIM is configured to have a depth of focus (DOF) exceeding the depth extension LZ, such that a cross-sectional image slice is obtained by a single image scan utilizing the HIM beam.

[0319] Clause 39: The device according to any one of Clauses 31 to 38 has a control unit with software code installed to perform the method according to any one of Clauses 1 to 28.

[0320] Clause 40: A method for generating a measurement recipe for inspecting a representative inspection volume in a wafer, comprising the following steps:

[0321] --Generate the first set of critical design performance metrics D1, which includes CAD image data of the examined volume and dimensions of critical design features.

[0322] --Generate a second set of key manufacturing performance metrics, D2, including 3D volumetric image data obtained during manufacturing process development.

[0323] --Derive measurement formula R, which includes the step of obtaining at least one of a third set of performance indicators D3.

[0324] --The step of generating a second set of key manufacturing performance indicators D2, the process recipe R includes obtaining at least a first cross-sectional image slice of at least a first cross-sectional surface through a representative inspection volume, wherein the depth extension below the wafer surface is... , .

[0325] Clause 41: The method according to Clause 41, wherein at least one of the steps of obtaining at least one of performance metrics D1 to D3 includes an image processing step to extract a plurality of first cross-sectional image features representing a plurality of first semiconductor features at a tilt angle GF, the image processing including at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation.

[0326] Clause 42: The method according to Clause 41, wherein at least one of the steps of obtaining at least one of performance indicators D1 to D3 includes the step of calculating at least a descriptive parameter of at least one first semiconductor feature from at least one of a plurality of first cross-sectional image features, wherein the descriptive parameter is one of size, diameter, angle, area, shape or volume.

[0327] Clause 43: The method according to Clause 42, wherein at least one of the steps of obtaining at least one of the performance metrics D1 to D3 further comprises the step of calculating the average or statistical deviation of at least one descriptive parameter of the plurality of first semiconductor features.

[0328] Clause 44: The method according to any one of Clauses 41 to 43, wherein at least one of the steps of obtaining at least one of performance indicators D1 to D3 includes the method steps of any one of Clauses 1 to 28.

[0329] Clause 45: A method for inspecting an inspection volume in a wafer using a dual-beam apparatus, the dual-beam apparatus comprising at least a FIB pillar and a charged particle imaging device, wherein a first optical axis of the FIB pillar forms an angle GF with the surface of a wafer support stage, and a second optical axis of the charged particle imaging device forms an angle GE with the normal to the surface of the wafer support stage, the first and second optical axes intersecting at a point, the method comprising the steps of:

[0330] --Load the wafer onto the wafer support stage in the dual-beam device;

[0331] --Move the wafer support stage to align the first measurement position on the wafer with the intersection point;

[0332] --A sequence of N cross-sectional image slices including at least a first cross-sectional image slice and a second cross-sectional image slice is obtained in the examination volume, wherein obtaining the first and second cross-sectional image slices includes subsequently exposing at least the first and second cross-sectional surfaces in the examination volume by milling into the examination volume approximately at angle GF using FIB pillars, and imaging the at least the first and second cross-sectional surfaces with a charged particle imaging device to obtain at least the first and second cross-sectional image slices.

[0333] --Determine at least one first cross-sectional image feature in the first and second cross-sectional image slices;

[0334] --Determine at least one second cross-sectional image feature in the first and second cross-sectional image slices; and

[0335] --Determine the depth of at least one first cross-sectional image feature in the first cross-sectional image slice from the lateral position of at least one second cross-sectional image feature in the first cross-sectional image slice.

[0336] Clause 46: The method according to Clause 45 further includes the step of performing mutual lateral alignment of the first and second cross-sectional image slices with at least one common cross-sectional image feature.

[0337] Clause 47: The method according to Clause 46, wherein the step of aligning the images laterally includes subtracting the image distortion deviation between at least the first and second cross-sectional image slices.

[0338] Clause 48: The method according to any one of Clauses 45 to 47, wherein the step of determining the depth of at least one first cross-sectional image feature comprises determining at least a first position of at least one second position of a second cross-sectional image feature in a first cross-sectional image slice and at least a lateral difference between at least one second position of a second cross-sectional image feature in a second cross-sectional image slice.

[0339] Clause 49: The method according to any one of Clauses 45 to 48, wherein in the step of determining at least one second cross-sectional image feature, at least two second cross-sectional image features are determined in a slice of a first cross-sectional image, wherein each of the second cross-sectional image features represents an integrated semiconductor structure at a different depth within the inspection volume.

[0340] Clause 50: The method according to Clause 49, wherein the step of determining the depth of at least one first cross-sectional image feature includes determining the depth of at least one first cross-sectional image feature in a first cross-sectional image slice from the lateral positions of at least two second cross-sectional image features.

[0341] Clause 51: The method according to any one of Clauses 45 to 50, wherein the optical axis of the charged particle imaging device is perpendicular to the wafer arrangement.

[0342] Clause 52: The method according to any one of Clauses 45 to 51, wherein the charged particle imaging device is a scanning helium ion microscope.

[0343] Clause 53: The method according to any one of Clauses 45 to 52 further includes the step of forming at least one alignment feature near the inspection volume, the alignment feature being configured for mutual lateral alignment of the first and second cross-sectional image slices.

[0344] Clause 54: The method according to Clause 53, wherein alignment features are fabricated above the inspection volume and configured to determine the positions of the first and second edges formed by the intersection of the first and second cross-sectional surfaces with the wafer surface.

[0345] Clause 55: The method according to Clause 53, wherein a first cross-sectional surface is formed within the wafer, having a greater extension in the x-direction perpendicular to the FIB beam compared to a second cross-sectional surface, such that after the second cross-sectional surface is formed, a parallel surface segment of the first surface segment is retained, and at least one alignment feature is formed on the remaining parallel surface segment of the first surface segment.

[0346] Clause 56: The method according to any one of Clauses 45 to 55, wherein the tilt angle GF between the first optical axis of the FIB pillar and the surface of the wafer support stage is in the range of 8° to 45°, more preferably in the range of 8° to 30°, and even more preferably in the range of 8° to 15°.

[0347] Clause 57: The method according to any one of Clauses 45 to 55, wherein the tilt angle GF between the first optical axis of the FIB pillar and the surface of the wafer support stage is between 25° and 60°, more preferably between 25° and 45°, and even more preferably between 30° and 40°.

[0348] Clause 58: The method according to any one of Clauses 45 to 57, wherein the step of obtaining at least first and second cross-sectional image slices in the inspection volume comprises scanning the focused ion beam of the FIB column in a first direction by a scanning unit to expose the first cross-sectional surface within the inspection volume, tilting the focused ion beam in a second direction perpendicular to the first direction by a scanning unit, and scanning the focused ion beam in the first direction by a scanning unit to expose the second cross-sectional surface within the inspection volume, such that the first and second cross-sectional surfaces form different angles with respect to the wafer surface at approximately a tilt angle GF.

[0349] Clause 59: The method according to Clause 59, wherein the wafer is not moved during the step of obtaining at least a first cross-sectional image slice and a second cross-sectional image slice in the inspection volume.

[0350] Clause 60: The method according to any one of Clauses 45 to 59, wherein the first cross-sectional image feature includes at least a cross-section of a semiconductor structure extending in a direction parallel to an axis perpendicular to the wafer surface, and the second cross-sectional image feature includes at least a cross-section of a semiconductor structure extending in a direction parallel to the wafer surface.

[0351] Clause 61: The method according to any one of Clauses 45 to 60, wherein the first common cross-sectional image feature includes at least a cross-section of one of the vias, HAR structures, or HAR channels of an integrated semiconductor circuit.

[0352] Clause 62: The method according to any one of Clauses 45 to 61, wherein the second cross-sectional image features include at least a cross-section of an isolation line or layer, a metal line or layer, or a semiconductor line or layer of an integrated semiconductor circuit.

[0353] Clause 63: The method according to any one of Clauses 45 to 62, wherein the number N of cross-sectional image slices is at least N = 10, preferably N > 100, and even more preferably N is about 1000 or greater.

[0354] Clause 64: A computer program product having program code for performing the method according to any one of Clauses 45 to 63.

[0355] Clause 65: A dual-beam device configured to perform the method according to any one of Clauses 45 to 63.

[0356] Clause 66: A wafer defect inspection apparatus configured to perform the method according to any one of Clauses 45 to 63, comprising:

[0357] --Focused ion beam column, configured to mill and expose at least first and second cross-sectional image surfaces in an inspection volume within a wafer;

[0358] --A charged particle imaging device, configured to image at least first and second cross-sectional surfaces to form first and second cross-sectional image slices;

[0359] --An image processing unit with software code installed is configured to determine cross-sectional image features in at least first and second cross-sectional image slices and to determine the depth of the cross-sectional image features within the inspection volume, the cross-sectional image features being cross-sections of semiconductor structures within the inspection volume;

[0360] --Defect detection unit, configured to determine from cross-sectional image features a deviation from predetermined characteristics of the semiconductor structure within the inspection volume; wherein the wafer defect inspection device is configured to inspect the inspection volume within the wafer, which is not extracted from the wafer.

[0361] Clause 67: A dual-beam apparatus (1) for inspecting an inspection volume below the wafer surface (55) of a wafer (8), comprising:

[0362] --A wafer stage (155) having a wafer support (15), the wafer support (15) being configured to hold a wafer (8) on a support surface (152) of the wafer support 15 during use.

[0363] --Focused ion beam column (FIB) (50), whose optical axis is arranged at an angle GF of more than 30° relative to the support surface (152) of the wafer support stage 15.

[0364] --A helium ion microscope (HIM) (140) having an optical axis (142) arranged perpendicular to the support surface (152), the optical axis of the FIB column (50) and the HIM (140) intersecting at point (43).

[0365] --The stage control unit (16) is configured to position the first measurement position (6.1) of the wafer (8) at the intersection (43) during use, and

[0366] --Control unit (19) for controlling FIB pillar (50) and HIM (140) is configured to mill multiple N cross-sectional surfaces in the inspection volume approximately alternately at angle GF using FIB beam (51) and to image each cross-sectional surface by scanning using HIM beam (144) to form multiple N high-resolution cross-sectional image slices, wherein the inspection volume has a depth extension LZ greater than 1 μm, preferably greater than 2 μm, below the wafer surface (55).

[0367] Clause 68: The dual-beam device (1) according to Clause 67, wherein the depth extension LZ in one direction exceeds 5 μm, for example 6 μm or 10 μm.

[0368] Clause 69: The dual-beam device (1) according to any one of Clauses 67 to 68, wherein the control unit (19) is further configured to scan the HIM beam (144) over a region of lateral dimension LX or LY of an examination volume exceeding about 5 μm to 10 μm (LX or LY) during use, and to collect a plurality of secondary electrons in a time-sequential manner with a secondary electron detector (17).

[0369] Clause 70: A dual-beam device (1) according to any one of Clauses 67 to 69, wherein the HIM (140) is configured to have a depth of focus (DOF) exceeding the depth extension LZ, such that each of a plurality of N high-resolution cross-sectional image slices is obtained by a single image scan utilizing the HIM beam (144).

[0370] Clause 71: The dual-beam device (1) according to any one of Clauses 67 to 70 further includes an image processing unit with software code installed, which is configured to determine at least a first virtual cross-sectional image from a plurality of N cross-sectional image slices, wherein the first virtual cross-sectional image is oriented parallel to the surface (55) of the wafer (8).

[0371] Clause 72: A method for inspecting an inspection volume below the wafer surface (55) of a wafer (8) using a dual-beam apparatus comprising a focused ion beam column (FIB) (50) and a helium ion microscope (HIM) (140), wherein the optical axis of the focused ion beam column (FIB) (50) is arranged at an angle GF greater than 30° with the support surface (152) of the wafer support stage (15), and the optical axis (142) of the helium ion microscope (HIM) (140) is arranged perpendicular to the support surface (152), and the optical axes of the FIB column (50) and the HIM (140) intersect at a point (43), the method comprising the following steps:

[0372] --The wafer (8) is positioned at a first measurement position (6.1) using a wafer stage (155), the wafer stage (155) having a wafer support stage (15) configured to hold the wafer (8) at the intersection (43).

[0373] --Multiple N cross-sectional surfaces are alternately milled in the inspection volume using FIB pillars (50) at approximately angle GF, with a depth extension LZ greater than 1 μm below the wafer surface (55), and

[0374] --Image each cross-sectional surface using a helium ion microscope (140) in a single image scan to form multiple N high-resolution cross-sectional image slices.

[0375] Clause 73: The method according to Clause 72, wherein the milling of the plurality of N cross-sectional surfaces has a depth extension LZ greater than 2 μm, preferably greater than 6 μm, more preferably greater than 10 μm below the wafer surface (55).

[0376] Clause 74: The method described pursuant to Clause 72 or 73 further includes:

[0377] --Calculate at least a virtual cross-sectional image parallel to the wafer surface (55) from multiple N cross-sectional image slices.

[0378] Clause 75: The method according to Clause 74, wherein a first virtual cross-sectional image is calculated in a conductive layer or word line.

[0379] Clause 76: The method according to any one of Clauses 74 or 75, wherein a second virtual cross-sectional image is calculated in an isolation layer.

[0380] Clause 77: The method according to any one of Clauses 72 or 76 further includes the step of performing a step of mutually lateral alignment of each of the N cross-sectional image slices with at least one common cross-sectional image feature.

[0381] Clause 78: The method according to any one of Clauses 72 or 77 further includes:

[0382] -- At least one alignment feature is formed near the inspection volume, the alignment feature being configured to form at least one common cross-sectional image feature, and

[0383] --Perform a sequence of N cross-sectional image slices with mutual lateral alignment of at least one common cross-sectional image feature.

[0384] Clause 79: The method according to any one of Clauses 72 or 78 further includes:

[0385] --The wafer is mounted on the wafer support stage in the dual-beam device.

[0386] Clause 80: The method pursuant to any one of Clauses 72 or 79 further comprises:

[0387] -- Generate a depth map Z(x,y;n) for each of the N cross-sectional image slices, where for each of the N cross-sectional image slices, the index is... .

[0388] Clause 81: The method described pursuant to Clause 80 further includes:

[0389] --Determine at least one second cross-sectional image feature in each of a sequence of N cross-sectional image slices, the at least one second cross-sectional image feature representing a cross section through a second semiconductor feature oriented parallel to the wafer surface 55.

[0390] Clause 82: The method described in accordance with Clause 81, wherein the index Depth map of each cross-sectional image slice It is generated from the lateral position of at least one second cross-sectional image feature.

[0391] Clause 83: The method according to any one of Clauses 72 or 82, wherein the number N of cross-sectional image slices is at least N = 10, preferably N > 100, and even more preferably N is about 1000 or greater.

[0392] Clause 84: The method according to any one of Clauses 72 or 82, wherein the number N of milled surfaces is less than 50, preferably less than 20, and the inspection volume is divided into a plurality of B blocks 181.1…181.B arranged diagonally by the inspection space.

[0393] Clause 85: A method for obtaining at least a virtual cross-sectional image from a set of cross-sectional image slices, comprising the steps of:

[0394] --A sequence of N cross-sectional image slices is obtained by sequentially imaging N cross-sectional surfaces at an angle GF and milling them into the inspection volume inside the wafer.

[0395] --Determine the first orientation direction of the first semiconductor feature, which forms a first plurality of first cross-sectional image features in a sequence of N cross-sectional image slices.

[0396] -- Calculate a virtual cross-sectional image perpendicular to a first orientation direction, the virtual cross-sectional image comprising a plurality of virtual cross-sectional image pixels, wherein for each virtual cross-sectional image pixel, a pixel value is calculated by projection of at least a subset of cross-sectional image slices of a sequence of N cross-sectional image slices onto the first orientation direction and by interpolation of pixel values ​​from the projection of the subset of at least one cross-sectional image slices.

[0397] Clause 86: The method according to Clause 85, wherein for each virtual cross-sectional image pixel, at least one subset of cross-sectional image slices is selected by evaluating the distance of each of the N cross-sectional image slices in the sequence to the virtual cross-sectional image pixel in the first orientation direction and selecting at least the first cross-sectional image slice with the minimum distance.

[0398] Clause 87: The method according to Clause 86, wherein a second cross-sectional image slice of a subset of at least one cross-sectional image slice is selected as the cross-sectional image slice having a second minimum distance.

[0399] Clause 88: The method according to any one of Clauses 86 to 87, wherein further cross-sectional image slices of a subset of at least one cross-sectional image slice are selected in order of increasing distance to the virtual cross-sectional image pixel in a first orientation direction.

[0400] Clause 89: The method according to any one of Clauses 85 to 88, wherein the step of projecting a subset of at least one cross-sectional image slice and interpolating pixel values ​​from the projection of the subset of at least one cross-sectional image slice includes projecting and interpolating at least one subset of a first plurality of first cross-sectional image features to form a third plurality of first cross-sectional image features in a virtual image slice.

[0401] Clause 90: The method of any one of Clauses 85 to 89, wherein the projection and interpolation steps are combined with at least one of feature extraction, thresholding operation, contour interpolation or model-based interpolation.

[0402] Clause 91: The method according to any one of Clauses 85 to 90 further includes

[0403] --Generate a depth map for each of the N cross-sectional image slices in the sequence. For each of the N cross-sectional image slices, the index is... .

[0404] Clause 92: The method described in accordance with Clause 91, wherein the index Depth map of each cross-sectional image slice It is generated by multiple second cross-sectional image features, which represent cross-sections of second semiconductor features oriented in a second orientation direction perpendicular to the first orientation direction.

[0405] Clause 93: The method described pursuant to Clauses 91 to 92, wherein the index Depth map of each cross-sectional image slice It is generated by determining the depth of the first cross-sectional image feature in a cross-sectional image slice from the lateral positions of at least two second cross-sectional image features.

[0406] Clause 94: The method according to any one of Clauses 85 to 93, wherein the first semiconductor feature includes at least one of a via, a HAR structure, or a HAR channel of an integrated semiconductor circuit within the inspection volume of the wafer.

[0407] Clause 95: The method according to Clause 94, wherein the second semiconductor feature includes at least one of an isolation line or layer, a metal line or layer, or a semiconductor line or layer of an integrated semiconductor circuit within the inspection volume of the wafer.

[0408] Clause 96: The method according to any one of Clauses 93 to 95 further includes:

[0409] --Determine at least one first cross-sectional image feature in each of the sequence of N cross-sectional image slices;

[0410] --Determine at least one second cross-sectional image feature in each of the sequence of N cross-sectional image slices.

[0411] Clause 97: The method according to any one of Clauses 85 to 96, wherein the first orientation direction is the z-direction perpendicular to the wafer surface, and a virtual cross-sectional image slice is calculated at a depth ZV below the wafer surface in a plane parallel to the wafer surface.

[0412] Clause 98: The method according to Clause 97, wherein for each virtual cross-sectional image pixel coordinate (x, y), at least a subset of cross-sectional image slices is determined by selecting at least the m-th cross-sectional image slice having the minimum distance to depth ZV, such that the distance... All depth maps The minimum value, where the index is... .

[0413] Clause 99: The method according to Clause 98, wherein a second and additional cross-sectional image slices, a subset of at least one cross-sectional image slice, are located at a distance in the Z direction from the virtual cross-sectional image pixel. The order in which they are added is selected.

[0414] Clause 100: The method according to any one of Clauses 97 to 99, wherein the depth ZV is adjusted according to the depth of layers parallel to the wafer surface, wherein these layers are formed by second semiconductor features oriented in a second orientation direction parallel to the wafer surface.

[0415] Clause 101: The method according to any one of Clauses 97 to 100, wherein a first virtual cross-sectional image slice is calculated at a depth ZV1 in an isolation layer between two adjacent metal layers or word lines.

[0416] Clause 102: The method according to any one of Clauses 97 to 101, wherein a second virtual cross-sectional image slice is calculated at a depth ZV2 inside the metal layer of the word line.

[0417] Clause 103: The method according to any one of Clauses 85 to 102, wherein the optical axis of the charged particle beam imaging system for acquiring a sequence of N cross-sectional image slices is oriented perpendicular to the wafer surface such that for the angle GE between the optical axis and the z-axis perpendicular to the wafer surface, the angle GE = 0°.

[0418] Clause 104: The method according to any one of Clauses 85 to 103 further includes the step of performing a step of mutually lateral alignment of each of the N cross-sectional image slices with at least one common cross-sectional image feature.

[0419] Clause 105: The method according to any one of Clauses 85 to 104 further includes the step of forming at least one alignment feature near the inspection volume, the at least one alignment feature being configured to form at least one common cross-sectional image feature for lateral alignment with each other.

[0420] Clause 106: The method according to any one of Clauses 104 to 105, wherein the step of aligning the images laterally includes subtracting the image distortion bias.

[0421] Clause 107: The method according to any one of Clauses 85 to 106, wherein the number N of cross-sectional image slices is at least N = 10, preferably N > 100, and even more preferably N is about 1000 or greater.

[0422] Clause 108: The method according to any one of Clauses 85 to 107 further includes:

[0423] --The wafer is mounted on a wafer support stage in a dual-beam apparatus, which includes at least a FIB pillar and a charged particle imaging device. The first optical axis of the FIB pillar forms an angle GF with the surface of the wafer support stage, and the second optical axis of the charged particle imaging device forms an angle GE with the normal to the surface of the wafer support stage. The first and second optical axes intersect.

[0424] --Move the wafer support stage to align the first measurement position on the wafer with the intersection of the dual-beam device.

[0425] Clause 109: The method of Clause 108, wherein the wafer is not moved during the step of obtaining a sequence of N cross-sectional image slices in the inspection volume.

[0426] Clause 110: A computer program product having program code for performing the method according to any one of Clauses 85 to 109.

[0427] Clause 111: A dual-beam device configured to perform the method according to any one of Clauses 85 to 109.

[0428] Clause 112: A wafer defect inspection apparatus configured to perform the method according to any one of Clauses 85 to 109, comprising:

[0429] --A wafer stage used to hold the wafers.

[0430] --Focused ion beam (FIB) column, which is configured to mill and expose a sequence of N cross-sectional surfaces in an inspection volume inside the wafer at an angle GF relative to the surface of the wafer;

[0431] --A charged particle imaging device configured to image a sequence of N cross-sectional surfaces to form a sequence of N cross-sectional image slices;

[0432] --An image processing unit with software code installed is configured to determine a virtual cross-sectional image from a sequence of N cross-sectional image slices, wherein the virtual cross-sectional image is oriented parallel to the surface of the wafer.

[0433] Clause 113: The wafer defect inspection apparatus according to Clause 112, wherein the tilt angle GF of the FIB pillar is between 30° and 80°, preferably between 30° and 45°.

[0434] Clause 114: A wafer defect inspection apparatus as described in Clause 112 or 113, wherein the charged particle imaging apparatus is oriented perpendicular to the wafer surface.

[0435] Clause 115: Wafer defect inspection apparatus according to any one of Clauses 112 to 114, wherein the charged particle imaging apparatus is a helium ion microscope (HIM).

[0436] Clause 116: A dual-beam apparatus for wafer inspection, comprising:

[0437] --A wafer stage for supporting wafers.

[0438] Gallium FIB pillars are arranged at an angle GF to the support surface of the sample support stage, and are configured to mill at least the cross-sectional surface into the wafer at an angle GF.

[0439] --Helium ion beam columns, arranged perpendicular to the support surface of the wafer stage, are configured to generate cross-sectional image slices of the cross-sectional surface through a single image scan.

[0440] --A secondary electron detector is configured to collect multiple secondary electrons during a single image scan.

[0441] --The operation unit is configured to operate and control the dual-beam device during use.

[0442] Clause 117: The dual-beam apparatus according to Clause 116, wherein the FIB is configured to mill the cross-sectional surface to a depth extending below the wafer surface to a depth exceeding 1 μm, 2 μm, or even 6 μm.

[0443] Clause 118: The dual-beam device according to any one of Clauses 116 to 17, wherein the operating unit further includes an image processing unit for performing at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation.

[0444] Clause 119: The dual-beam device according to Clause 118, wherein the image processing unit is further configured to calculate one of a depth map, a 3D volumetric image, or a virtual cross-sectional image.

[0445] Clause 120: A dual-beam device according to any one of Clauses 118 to 19, wherein the operating unit further comprises a defect detection unit for calculating one of the dimensions, area, diameter, angle, or shape of a semiconductor feature.

[0446] Clause 121: The dual-beam device according to Clause 120, wherein the defect detection unit is further configured to calculate one of the average values ​​and statistical deviations of a plurality of semiconductor features.

[0447] Clause 122: A wafer inspection method for 3D inspection of an inspection volume in a wafer, the inspection volume having a depth range greater than 1 μm, preferably greater than 2 μm, greater than 6 μm, or even 10 μm.

[0448] -- 3D inspection generates images with a lateral resolution of less than 2nm, preferably less than 1nm, or even less than 0.5nm, wherein 3D inspection of the inspection volume is performed at the wafer.

[0449] Clause 123: The method described in accordance with Clause 122, wherein the image is a 3D image.

[0450] Clause 124: The method according to any one of Clauses 122 or 123 further includes the step of milling at least the cross-sectional surface through the depth range of the inspection volume with a FIB column.

[0451] Clause 125: The method according to Clause 124, wherein an image is obtained by using a helium ion microscope in a single image scan to obtain at least an image of at least one cross-sectional surface of the examination volume.

[0452] Clause 126: The method according to any one of Clauses 122 or 125 further includes image processing, performing at least one of feature extraction, edge detection, pattern recognition, or pixel interpolation.

[0453] Clause 127: The method according to any one of Clauses 122 or 126 further includes the step of calculating one of a depth map, a 3D volumetric image, or a virtual cross-sectional image.

[0454] Clause 128: The method according to any one of Clauses 122 or 127 further includes the step of calculating one of the dimensions, area, diameter, angle, or shape of the semiconductor feature.

[0455] Clause 129: The method according to any one of Clauses 122 or 128 further includes the step of calculating an average value of a plurality of semiconductor characteristics, or one of the statistical deviations.

[0456] Clause 130: A method for forming a 3D volumetric image from a sequence of cross-sectional image slices, comprising:

[0457] --A sequence of N cross-sectional image slices, including at least a first cross-sectional image slice and a second cross-sectional image slice of the examination volume, wherein obtaining the first and second cross-sectional image slices includes subsequently exposing at least the first and second cross-sectional surfaces in the examination volume by milling into the examination volume approximately at angle GF using a FIB column, and imaging the at least the first and second cross-sectional surfaces with a charged particle imaging device to obtain at least the first and second cross-sectional image slices; wherein the first cross-sectional image surface is milled with a greater extension in a direction perpendicular to the FIB beam compared to the second cross-sectional surface, such that after the second cross-sectional surface is formed, a parallel surface segment of the first cross-sectional surface is retained.

[0458] Clause 131: The method according to Clause 130 further includes the step of forming at least one alignment feature on a parallel surface segment of the first cross-sectional surface for a first mutual lateral alignment of the first and second cross-sectional image slices.

[0459] Clause 132: The method described in accordance with Clause 132 further includes the following steps:

[0460] --Determine at least one first cross-sectional image feature in the first and second cross-sectional image slices;

[0461] --Based on the first mutual lateral alignment, perform a second mutual lateral alignment of the first and second cross-sectional image slices to achieve a predetermined mutual positional accuracy of the first and second cross-sectional image slices.

[0462] Clause 133: The method according to Clause 133, wherein the first cross-sectional image feature is a cross-section of a via, a HAR structure, or a HAR channel.

[0463] Clause 134: The method described pursuant to Clause 132 or 133 further includes the following steps:

[0464] --Determine at least one second cross-sectional image feature in the first and second cross-sectional image slices; and

[0465] --Determine the depth of at least one first cross-sectional image feature in the first cross-sectional image slice from the lateral position of at least one second cross-sectional image feature in the first cross-sectional image slice.

[0466] --Determine the depth of at least one first cross-sectional image feature in the second cross-sectional image slice from the lateral position of at least one second cross-sectional image feature in the second cross-sectional image slice.

[0467] --Consider the depth in the second mutual lateral alignment to achieve a mutual positional accuracy of at least 5nm, 3nm or even 2nm for the first and second cross-sectional image slices.

[0468] Clause 135: The method according to Clause 134, wherein at least one second cross-sectional image feature is formed by a cross-section of one of an isolation line or layer, a metal line or layer, or a semiconductor line or layer.

[0469] Clause 136: The method described pursuant to Clause 134 or 135 further includes:

[0470] --Calculate the first displacement of the first cross-sectional image features between the first cross-sectional image slice and the second cross-sectional image slice. ,

[0471] --Calculate the second displacement of the second cross-sectional image features between the first cross-sectional image slice and the second cross-sectional image slice. ,

[0472] --Determine the distance d between the first and second cross-sectional image slices, and

[0473] --Determine the mutual lateral displacement vector between the first and second cross-sectional image slices .

[0474] Clause 137: The method according to any one of Clauses 130 to 136, wherein the tilt angle GF is adjusted between 25° and 45°, or between 30° and 36°.

[0475] Clause 138: A method for forming a precise alignment of multiple cross-sectional image slices, the multiple cross-sectional image slices being obtained from an inspection volume below the wafer surface, the method comprising:

[0476] --A sequence of cross-sectional image slices is formed by continuously milling the cross-sectional surfaces using FIB pillars arranged at an angle GF to the wafer surface, and imaging each cross-sectional surface using charged particle imaging pillars arranged at an angle GE to the normal to the wafer surface. This sequence of cross-sectional image slices includes a first cross-sectional image slice and a second cross-sectional image slice.

[0477] --Perform a first coarse alignment of the first cross-sectional image slice and the second cross-sectional image slice to obtain a mapping of cross-sectional image features in the first cross-sectional image slice and the second cross-sectional image slice.

[0478] --Perform a second precise alignment of the first cross-sectional image slice and the second cross-sectional image slice to achieve mutual positional accuracy of less than 5nm, 3nm or even less than 2nm.

[0479] Clause 139: The method described pursuant to Clause 138 further includes the following steps:

[0480] --Including forming a first cross-sectional surface by milling, which has a greater extension in the direction perpendicular to the FIB beam compared to the second cross-sectional surface, such that after forming the second cross-sectional surface, the parallel surface segment of the first cross-sectional surface is retained.

[0481] -- At least one alignment feature is formed on a parallel surface segment of the first cross-sectional surface for a first mutual lateral alignment of the first cross-sectional image slice and the second cross-sectional image slice.

[0482] Clause 140: The method described pursuant to Clauses 138 or 139 further includes the following steps:

[0483] --Determine at least one first cross-sectional image feature from the first cross-sectional image slice and the second cross-sectional image slice.

[0484] --Determine at least one second cross-sectional image feature from the first cross-sectional image slice and the second cross-sectional image slice; and

[0485] --Determine the depth of at least one first cross-sectional image feature in the first cross-sectional image slice from the lateral position of at least one second cross-sectional image feature in the first cross-sectional image slice.

[0486] --Determine the depth of at least one first cross-sectional image feature in the second cross-sectional image slice from the lateral position of at least one second cross-sectional image feature in the second cross-sectional image slice.

[0487] --Consider the depth in the second precise alignment.

[0488] Clause 141: The method described pursuant to Clause 140 further includes:

[0489] --Calculate the first displacement of the first cross-sectional image features between the first cross-sectional image slice and the second cross-sectional image slice. ,

[0490] --Calculate the second displacement of the second cross-sectional image features between the first cross-sectional image slice and the second cross-sectional image slice. ,

[0491] --Determine the distance d between the first cross-sectional image slice and the second cross-sectional image slice, and

[0492] --Determine the mutual lateral displacement vector between the first cross-sectional image slice and the second cross-sectional image slice. .

[0493] Clause 142: The method according to any one of Clauses 140 to 141, wherein at least one first cross-sectional image feature is a cross-section of a via, a HAR structure, or a HAR channel.

[0494] Clause 143: The method according to any one of Clauses 140 to 142, wherein at least one second cross-sectional image feature is formed by a cross-section of one of an isolation line or layer, a metal line or layer, or a semiconductor line or layer.

[0495] Clause 144: The method according to any one of Clauses 138 to 143, wherein the tilt angle GF is adjusted between 25° and 45°, or between 30° and 36°.

[0496] Clause 145: The method according to any one of Clauses 140 to 144, wherein the angle GE is adjusted to 0°.

[0497] Clause 146: A computer program product having program code for performing the method according to any one of Clauses 130 to 146.

[0498] Clause 147: A dual-beam device configured to perform the method according to any one of Clauses 130 to 146.

[0499] List of reference numbers:

[0500] 1. Dual-beam device

[0501] 2 operating units

[0502] 4.1, 4.2, 4.3 Features of the First Cross-Section Image

[0503] 6.1, 6.2 Measurement locations

[0504] 8 chips

[0505] 10 semiconductor samples

[0506] 11-section surface

[0507] 13. Check volume

[0508] 15-chip support platform

[0509] 16-stage control unit

[0510] 17 Secondary Electronic Detector

[0511] 19 control units

[0512] 20 Alignment Marks

[0513] 22 Align with the groove or edge

[0514] 23 Parallel surface segment of the first cross-section surface

[0515] 24 alignment marks

[0516] 25 Integrated Circuit Characteristics

[0517] Image segments 26.1, 26.2, and 26.3

[0518] 28 image segments

[0519] 30 Depth Extension

[0520] 32.1, 32.2, 32.3 Alignment Features

[0521] 34-level DLZ

[0522] 36 Distance to the aligned feature

[0523] 38 alignment features

[0524] 40 Charged Particle Beam (CPB) Imaging System

[0525] 42 Imaging System Optical Axis

[0526] 43 intersections

[0527] 44 electron beam

[0528] 46 scan imaging lines

[0529] 48 FIB optical axis

[0530] 50 FIB column

[0531] 51 Focused Ion Beam

[0532] 52 cross-section surface

[0533] 53 cross-section surface

[0534] 54 cross-section surface

[0535] 55 wafer top surface

[0536] 58 FIB beam coincidence point

[0537] Slice distance at the surface of 60 wafers

[0538] 62. Check the slice distance at the bottom of the volume.

[0539] 64-angle extended GZ

[0540] 66 First milling angle

[0541] 68 Second milling angle

[0542] 72 layers, lower surface of L1

[0543] 73.1, 73.2, 73.3 Features of the Second Cross-Section Image

[0544] The upper boundary of floor 74, L4

[0545] 75 columnar HAR structure

[0546] Top edges of the surfaces of sections 76.1 and 76.2

[0547] 77.1, 77.2, and 77.3 Cross-sectional image segments of the HAR channels

[0548] 78 HAR structure vertical edge

[0549] 79.1 Centroid of cross-sectional image features

[0550] 80th floor horizontal edge

[0551] 82 Image Processing Units

[0552] 84 Defect Detection Unit

[0553] 86 Charged Particle Column Control Unit

[0554] 88 interface unit

[0555] 90 FIB angular extension in x

[0556] 92 Proximal trench

[0557] 94 distal trench

[0558] 96 First cross-section surface

[0559] 2D cross-sectional images of 100.1, 100.2, and 100.3

[0560] A sequence of 1000 2D cross-sectional images

[0561] 121 Virtual cross-sectional image slices

[0562] Pixel raster in the 123 y direction

[0563] Interpolation direction of 125 HAR structure

[0564] 140 Helium Ion Microscopy (HIM)

[0565] 142 HIM optical axis

[0566] 146 Check location

[0567] 148 alignment features

[0568] 150 aligned with feature 148 image

[0569] A single scan line in the x-direction of 152

[0570] 155 chip stage

[0571] 160 Check Volume

[0572] More than 162 alternating layers

[0573] More than 164 HAR structures

[0574] Virtual sections 166.1 and 166.2

[0575] 175 Cross-section of conductive layer or word line

[0576] 177.1, 177.2 Cross sections of the HAR structure

[0577] Virtual cross-sections of the HAR structure, 179.1 and 179.2

[0578] 181.1, 181.b Check the volume of the block

[0579] 191 deep inspection volume

[0580] 201 representative vertical HAR profiles

[0581] 203 cross-sectional image features

[0582] 205 HAR structure

[0583] 207 Digital Images

Claims

1. A method of forming a 3D volume image from a sequence of cross-sectional image slices, comprising: obtaining a sequence of N cross-sectional image slices comprising at least a first cross-sectional image slice and a second cross-sectional image slice of an examination volume, wherein obtaining the first cross-sectional image slice and the second cross-sectional image slice comprises subsequently exposing at least a first cross-sectional surface and a second cross-sectional surface in the examination volume by milling into the examination volume with a FIB column approximated at an angle GF and imaging the at least first and second cross-sectional surface with a charged particle imaging device to obtain the at least first and second cross-sectional image slices; wherein the first cross-sectional image surface is milled with a greater extension in a direction perpendicular to the FIB beam compared to the second cross-sectional surface such that after forming the second cross-sectional surface a parallel surface section of the first cross-sectional surface remains, and forming at least one alignment feature on the parallel surface section of the first cross-sectional surface for a first mutual lateral alignment of the first cross-sectional image slice and the second cross-sectional image slice.

2. The method according to claim 1, further comprising the steps of determining at least one first cross-sectional image feature in the first cross-sectional image slice and the second cross-sectional image slice; performing a second mutual lateral alignment of the first cross-sectional image slice and the second cross-sectional image slice based on the first mutual lateral alignment to achieve a predetermined mutual positional accuracy of the first cross-sectional image slice and the second cross-sectional image slice.

3. The method of claim 2, wherein, the first cross-sectional image feature is a cross-section of a via, a HAR structure or a HAR channel.

4. The method according to claim 2 or 3, further comprising the steps of: determining at least one second cross-sectional image feature in the first cross-sectional image slice and the second cross-sectional image slice; and determining a depth of the at least one first cross-sectional image feature in the first cross-sectional image slice from a lateral position of the at least one second cross-sectional image feature in the first cross-sectional image slice, determining a depth of the at least one first cross-sectional image feature in the second cross-sectional image slice from a lateral position of the at least one second cross-sectional image feature in the second cross-sectional image slice, considering the depths in the second mutual lateral alignment to achieve a mutual positional accuracy of the at least first cross-sectional image slice and the second cross-sectional image slice of below 5 nm, below 3 nm or even below 2 nm.

5. The method of claim 4, wherein, the at least one second cross-sectional image feature is formed by a cross-section of one of an isolation line or layer, a metal line or layer, a semiconductor line or layer.

6. The method according to claim 4, further comprising: computing a first displacement of the first cross-sectional image feature between the first cross-sectional image slice and the second cross-sectional image slice , computing a second displacement of the second cross-sectional image feature between the first cross-sectional image slice and the second cross-sectional image slice , determining a distance d between the first cross-sectional image slice and the second cross-sectional image slice, and determining a mutual lateral displacement vector between the first and second cross-sectional image slices .

7. The method of any one of claims 1 to 3, wherein, the tilt angle GF is adjusted between 25° and 45° or between 30° and 36°.

8. A computer program product having a program code for performing the method according to any one of claims 1 to 7.

9. A dual beam device configured to perform the method according to any one of claims 1 to 7.

Images