Method for analyzing a fault location of a contact point and its possible cause, storage medium and evaluation device

Abstract

The invention relates to a method for detecting at least one defect (20) of a contact point, in particular at least one via (2), of a microelectronic component (1) and for determining its possible cause, wherein structural data (11) relating to the contact point are recorded, wherein the structural data (11) are supplied to an evaluation device (15) by means of which an evaluation of the structural data (11) is carried out to detect the at least one defect (20), wherein stress data (44) relating to a mechanical stress of a material present in and / or at the contact point are recorded, wherein the stress data (44) are supplied to the evaluation device (15) by means of which an evaluation of the stress data (44) is carried out to determine its possible cause.

Description

[0001] The present invention relates to a method for detecting at least one defect location of a contact point of a microelectronic component and for determining its possible cause.

[0002] Microelectronic components, comprising, for example, a plate-shaped silicon substrate on which microelectronic components are applied and / or formed, frequently have contact points for electrically connecting multiple semiconductor devices. Through-hole vias are often provided as contact points, by means of which an electrical connection is formed from a top to a bottom of this component. Through-hole vias typically have several superimposed layers that serve both insulating and contacting purposes. Such through-holes can either be filled with an electrically conductive material, such as copper, or have a hollow interior, the present invention relating to the latter case, in which the through-holes are not filled or only partially filled, at least during the manufacturing process.During the manufacturing process and during ongoing use, defects can occur in these layers, for example, due to thermal and / or mechanical stresses and / or chemical processes. Such defects can manifest as layer delamination and / or bulging or swelling and can negatively affect the electrical properties of the respective via. In the finished assembly state, defects may cause undesirable electrical properties, which can indicate their presence. However, it is desirable that any defects can be identified or detected before the final assembly state.

[0003] For detailed information regarding potential defects in vias, please refer to the following state of the art: - De Wolf, Ingrid, Kristof Croes, and Eric Beyne. „Expected failures in 3-D technology and related failure analysis challenges.“ IEEE Transactions on Components, Packaging and Manufacturing Technology 8.5 (2018): 711-718. - Vartanian, Victor, et al. „Metrology needs for through-silicon via fabrication.“ Journal of Micro / Nanolithography, MEMS, and MOEMS 13.1 (2014): 011206-011206. - Tsuto, Takashi, et al. „Advanced through-silicon via inspection for 3D integration.“ Transactions of The Japan Institute of Electronics Packaging 6.1 (2013): 13-17. - Li, Yan, Purushotham Kaushik Muthur Srinath, and Deepak Goyal. „A review of failure analysis methods for advanced 3D microelectronic packages.“ Journal of Electronic Materials 45 (2016): 116-124. - Bender, Hugo, et al. „Structural characterization of through silicon vias.“ Journal of Materials Science 47 (2012): 6497-6504.

[0004] Several conceivable methods for detecting potential defects are known from the prior art. Examples include electrical measurements, acoustic microscopy (SAM), particularly surface acoustic waves (SAW), infrared microscopy (IR microscopy), X-ray microscopy (XRM), microtomography (µ-CT), lock-in thermography (LIT), photoemission microscopy (PEM), time-of-flight secondary ion mass spectrometry (ToF-SIMS), and atomic force microscopy (AFM). Furthermore, non-destructive methods for detecting potential defects are also known, such as embedding a microelectronic component in resin and cutting and grinding this sample to obtain a cross-section, which can then be used for further investigations into the defect, for example, by scanning electron microscopy (SEM). For further details on the aforementioned possibilities, reference is made to the following prior art: - Kia, Alireza M., et al. “ToF-SIMS 3D analysis of thin films deposited in high aspect ratio structures via atomic layer deposition and chemical vapor deposition.” Nanomaterials 9.7 (2019): 1035. - de Veen, PJ, et al. “High-resolution X-ray computed tomography of through silicon vias for RF MEMS integrated passive device applications.” Microelectronics Reliability 55.9-10 (2015): 1644-1648. - Cassidy, C., et al. “Depth-resolved photoemission microscopy for localization of leakage currents in through silicon vias (TSVs).” 2009 16th IEEE International Symposium on the Physical and Failure Analysis of Integrated Circuits. IEEE, 2009.

[0005] Other preliminary publications, particularly concerning the resistance of vias to mechanical stress, include: - Jiang, Tengfei, et al. „Through-silicon via stress characteristics and reliability impact on 3D integrated circuits.“ Mrs Bulletin 40.3 (2015): 248-256. - Ryu, Suk-Kyu, et al. „Stress-Induced Delamination Of Through Silicon Via Structures.“ AIP Conference Proceedings. Vol. 1378. No. 1. American Institute of Physics, 2011. - De Wolf, Ingrid, et al. „In-depth Raman spectroscopy analysis of various parameters affecting the mechanical stress near the surface and bulk of Cu-TSVs.“ 2012 IEEE 62nd Electronic Components and Technology Conference. IEEE, 2012. - Zschenderlein, U., et al. „Residual stress investigations at TSVs in 3D micro structures by HR-XRD, Raman spectroscopy and fibDAC.“ 2014 IEEE 64th Electronic Components and Technology Conference (ECTC). IEEE, 2014. - Vogel, Dietmar, et al. „Stress analyses of high spatial resolution on TSV and BEoL structures.“ Microelectronics Reliability 54.9-10 (2014): 1963-1968

[0006] Several patent publications related to through-hole plating are also known, such as US 2014 / 0064445A1, US 2015 / 0192527A1, JP 2016-085045A, US 2015 / 0226676A1, CN 110911301A, US 10352879B2, WO 2021 / 084933A1, US 2017 / 0067732A1, and WO 2023 / 054910A. Further relevant patents include US 9810796B2, US 10088578B2, US 10156644B2, and US 10120082. B2, US 2022 / 0 066 054 A1, US 11 076 822 B2 and US 10 825 729 B2.

[0007] Other prior publications known from patent literature concerning the analysis of vias with regard to mechanical loads are CN 1 03 439 248 A, CN 1 02 944 336 A, CN 1 03 424 293 A, CN 1 09 297 627 A and KR 2023 168 910 A.

[0008] One conceivable approach for analyzing existing properties of components, in which the creation of a division plane involves the destruction of the respective component, is the method of electron backscatter diffraction, for which further details are referred to the following literature: - Britton, TB, and JLR Hickey. “Understanding deformation with high angular resolution electron backscatter diffraction (HR-EBSD).” IOP Conference Series: Materials Science and Engineering. Vol. 304. IOP Publishing, 2018. - Wilkinson, Angus J., Graham Meaden, and David J. Dingley. “High-resolution elastic strain measurement from electron backscatter diffraction patterns: New levels of sensitivity.” Ultramicroscopy 106.4-5 (2006): 307-313. - Wilkinson, Angus J., Graham Meaden, and David J. Dingley. “High resolution mapping of strains and rotations using electron backscatter diffraction.” Materials Science and Technology 22.11 (2006): 1271-1278.

[0009] The invention aims to provide an improved concept for the analysis of defects at contact points, particularly vias, of microelectronic components, especially with regard to obtaining the most reliable results possible.

[0010] According to the invention, the problem is solved in a method of the type mentioned at the outset by recording structural data relating to the contact point, wherein the structural data are supplied to an evaluation device by means of which an evaluation of the structural data is carried out to detect the at least one defect location, wherein load data relating to a mechanical load of a material present in and / or at the contact point are recorded, wherein the load data are supplied to the evaluation device by means of which an evaluation of the load data is carried out to determine its possible cause.

[0011] The invention is based in particular on the idea of ​​obtaining as much information as possible regarding the presence, type, properties, and any causes of a potential defect during the analysis of both structural and stress data. The evaluation of the structural and stress data is preferably carried out by combining, and thus in particular simultaneously evaluating, the data available for the respective defect. Up to now, structural data has typically been used exclusively for detecting the presence of a defect, whereas stress data has been used exclusively in connection with stress-specific conditions in the area of ​​the contact point or via.In contrast, combining this data allows, in particular, the consideration of the insights gained from the load data regarding the load-specific state, specifically whether and under what conditions a defect occurs. Conversely, the structural data can also reveal, in principle, the load-specific circumstance that must have caused the defect to occur. Ideally, the results obtained during the execution of the inventive method for structurally identical contact points can be used in subsequent analyses to draw conclusions about the mechanisms underlying the defect formation with minimal effort, based solely on the structural data, without the need to determine the load data.The results obtained using the inventive method can also be used to improve the manufacturing of the microelectronic component or the contact point with regard to preventing the occurrence of defects. For example, if it turns out that a large proportion of the defects are due to a specific mechanism, such as stress-induced layer delamination, then the manufacturing process can be specifically modified so that this mechanism occurs to a lesser extent, i.e., so that, for example, lower stresses occur in the area of ​​the contact point.

[0012] Another key aspect of the inventive method concerns the automated evaluation of structural and load data using the evaluation device and, consequently, a computer. This enables a large number of contact points to be checked for defects in a short time. In particular, it allows for the precise identification of those contact points where a defect is present.

[0013] In principle, the method according to the invention is generally conceivable for contact points of microelectronic components. The contact points provided in the implementation of the method according to the invention are particularly preferably the vias already described above. In principle, the contact points within the scope of the present invention can also be direct copper-to-copper bonds (direct Cu-Cu bond) or micro-copper bonds (µ-Cu bond). Since the analysis of vias is particularly preferred according to the invention, this application is referred to more extensively below; however, it is expressly pointed out that the aspects described accordingly can also be provided for in the analysis of contact points generally, where appropriate.

[0014] Preferably, the microelectronic component comprises a substrate, wherein the at least one via is formed by a recess in the substrate. The substrate is made of plastic or, preferably, silicon. To produce the via, the recess, which is, for example, a blind hole, can be created in the substrate, for example, by etching. Subsequently, an insulating layer can be applied there, for example, by chemical, in particular electrochemical, or physical deposition of an insulating material, to create electrical insulation between the via and the substrate material. In the next step, a barrier layer can be applied to the insulating layer, for example, to prevent material diffusion and to mechanically reinforce the via.Finally, a conductive layer made of an electrically conductive material, such as a metal like copper, is deposited onto the barrier layer, for example, by chemical, especially electrochemical, or physical deposition. Thus, the hollow interior of the via can be bounded by the conductive layer. For chemical deposition, chemical vapor deposition (CVD) can be used, for example. For physical deposition, physical vapor deposition (PVD) can be used, for example. The conductive layer provides the desired electrical conductivity of the via. The interior can be filled with an electrically conductive material, such as a metal like copper, or it can remain empty.The aforementioned layers are located either only within the depression or also in a lateral area around an opening of the via.

[0015] The conductive layer can cover or form a bottom and a side wall or walls of the recess. If the via is hollow, the recess can be open on one side. The carrier is preferably plate-shaped, with the opening located in the region of one of the two planar surfaces of the carrier. The carrier material can be ground down on the surface opposite the surface containing the opening, exposing the bottom of the conductive layer as a contact pin. The interior can have a geometry that is at least substantially cylindrical.

[0016] The structural data can be image data, and the sensory acquisition of this data using an imaging device typically does not require excessive effort. Nevertheless, this data basis is fundamentally suitable for reliably detecting and defining any defects. The image data can depict the physical structure of the contact point or an area of ​​the microelectronic component encompassing the contact point. The image data can be in the form of at least one two-dimensional image, i.e., at least one two-dimensional array, comprising several pixels. Alternatively, or in addition, the image data can be in the form of at least one three-dimensional image, i.e., at least one three-dimensional array, comprising several voxels.Each pixel or voxel can be assigned a value for a property of the structure of the contact point at that location, such as a value relating to a color and / or an intensity or a brightness and / or the like.

[0017] The load data pertains to an existing mechanical load or stress within and / or at the respective contact point. The load data can specify the magnitude and / or direction of an existing internal stress or residual stress in the material of the contact point or the carrier, preferably as a function of the respective position. The load data can also be provided as image data, with each pixel or voxel assigned at least one value that characterizes the load present there.

[0018] Preferably, the structural data are acquired using a non-destructive testing method. This means that the structural integrity of the microelectronic component is maintained even after the structural data has been acquired. Specifically, the contact point can be exposed to radiation, particularly electromagnetic radiation, and the resulting back radiation, which arises, for example, from reflection and / or diffraction, is detected by sensors to determine the structural data. Specific examples are given below, and all conceivable combinations of the measurement methods mentioned below can be implemented within the scope of the present invention.

[0019] It is conceivable that the structural data could be acquired using optical microscopy data via an optical microscopy setup, specifically concerning at least a portion of the visually visible range of the electromagnetic spectrum. The microscopy setup could comprise a microscope and a camera, with the structural data acquired as image data representing a magnified view of the contact point obtained through the microscope. Particularly since the contact point, or rather its opening, has a diameter of less than 100 µm, a magnification of 500 to 1000 times achieved with the microscope is advantageous. To ensure optimal illumination during image acquisition, a Köhler illumination system could be used.The microscope can comprise several components, such as an eyepiece or eyepiece lens, an objective turret or lens changer, at least one objective lens, a tube lens, and a stage. The stage can include a mounting device for attaching the microelectronic component. The stage can be movable along the optical axis of the microscope, particularly to position the contact point at a focal point of the microscope. The viewing direction of the microscopy device, or the optical axis of the microscope, can be used to acquire image data through the opening into the interior of the via. The viewing direction of the image acquisition device during the acquisition of structural data and a longitudinal axis of the cylindrical cavity can be identical.During the acquisition of structural data, the viewing direction of the microscopy device is preferably at least substantially perpendicular to the base of the via.

[0020] With respect to the plane perpendicular to the optical axis of the microscope, the stage can be displaceable in at least one spatial direction. By moving the stage along this at least one spatial direction, several contact points of the microelectronic component can be successively brought into the field of view of the microscope or camera, so that multiple data sets relating to image or structural data of one of the contact points can be acquired sequentially. To achieve this displaceability, the stage can be coupled to at least one actuator, in particular an electromechanical one.Thus, a control device, which is in particular the evaluation device, can be set up to generate control signals for controlling at least one actuator, so that several data sets of structural data from several contact points of the microelectronic component held by means of the object table can be automatically and successively recorded.

[0021] It is conceivable that the microscopy apparatus comprises at least one light source for generating light, wherein the structural data is acquired by illuminating the contact point with the light generated by the at least one light source. The light source is or comprises, for example, at least one light-emitting diode and / or at least one halogen lamp. The light can be monochromatic, with the wavelength being located, in particular, in the visually visible region of the electromagnetic spectrum. The light can be polychromatic, with its spectral distribution being located, in particular, at least partially, in the visually visible region of the electromagnetic spectrum. The microscopy apparatus can have several light sources located at different positions in order to illuminate the contact point from different directions.If only one light source is provided, it is preferably arranged at least substantially centrally above the opening, in particular on an extension of the longitudinal axis of the cylinder describing the shape of the interior.

[0022] Preferably, the light generated by the light source is polarized. Linearly, circularly, or elliptically polarized light can be used. Tests have shown that defects at contact points, especially vias, become particularly apparent in the structural data or optical microscopy data when using polarized light. In particular, the edge regions of the base where the side wall(s) meet appear with particularly strong contrast in the structural data when using circularly polarized light, making defects in these areas especially easy to identify. This advantageous effect is particularly pronounced when using circularly polarized light.

[0023] Preferably, optical microscopy data is acquired using bright-field microscopy. This means that all light reflected at the contact point is captured by the microscopy apparatus or camera. However, dark-field microscopy is also conceivable, in which light of a specific wavelength range or with a specific polarization is used to acquire the image data. For this purpose, appropriately sized filters and / or apertures can be incorporated into the beam path of the microscope and / or camera. Other operating modes generally feasible with microscopes can also be used for acquiring the structural data, such as modes for generating differential interference contrast, phase contrast, or similar techniques.

[0024] In addition or alternatively, X-ray microscopy data can be acquired using an X-ray microscopy device. X-ray microscopy uses X-rays instead of visible light, which allows for higher resolution and greater information depth. Fresnel zone plates can be used to focus the X-rays. X-ray microscopes can achieve a resolution of 20 to 30 nm. The X-ray microscopy device preferably comprises a scanning X-ray microscope with which the sample is scanned point by point. For this purpose, the X-ray microscopy device includes an X-ray source that generates the X-rays used for this scanning.

[0025] Alternatively, or in addition, it is conceivable that the structural data could be acquired using computed tomography (CT) data from a CT scanner, particularly an X-ray CT scanner. During CT data acquisition, the sample is positioned between a radiation source, particularly an X-ray source, and a radiation detector, with the sample and the system comprising the X-ray source and the radiation detector being moved, particularly rotated, relative to each other. This allows the density of the sample to be determined in space, resulting in a stack of images. Further processing of this data can yield three-dimensional image data as the structural data.

[0026] Structural data, particularly acoustic surface waves, can be acquired using an acoustic microscopy device. For this purpose, high-frequency sound, especially ultrasound, is generated by a transmitter within the acoustic microscopy device and introduced into or reflected from the sample, allowing for the detection of existing defects and / or material properties. Specifically, measurement data regarding the transit time, amplitude, and / or polarity of the received sound wave are determined.

[0027] Alternatively, or in addition, it is conceivable that the structural data could be acquired using infrared microscopy. In this method, the sample is illuminated with infrared radiation, and due to the interaction of this radiation with molecules in the sample, absorption occurs with respect to certain frequency ranges and / or different reflectances are observed. Specifically, a position-dependent transmittance and / or reflectance of the infrared radiation in the sample can be determined.

[0028] Furthermore, it is conceivable that the structural data could be acquired using lock-in thermography. This involves thermally or energetically exciting the sample, thereby inducing a heat flow. Layers and / or defects within the sample lead to inhomogeneities with respect to the heat flow, which can then be detected using a thermal imaging camera of the lock-in thermography system.

[0029] Furthermore, structural data can be acquired using photoemission microscopy (PEM) using a PEM setup. PEM data represent a two-dimensional intensity distribution of photoelectrons on the sample's surface. For this purpose, the sample is irradiated with light, particularly UV light, from a UV source within the PEM setup, causing the photoelectrons to be emitted from the sample via the photoelectric effect. These emitted photoelectrons are then collected using a strong electrostatic field, and the resulting electron image can optionally be magnified using electron lenses. The photoelectrons then strike a fluorescent screen, creating an image that can be captured using a camera, particularly a CCD camera.

[0030] According to the invention, it can be provided that the structural data is used to check whether at least one defect is present at the respective contact point, with the load data only being recorded in this case. In this embodiment, the load data is determined only if a defect is actually present at the respective contact point, or if the structural data at least imply its presence. This approach is particularly advantageous when the aim is to specifically record information relating exclusively to existing defects.

[0031] Alternatively, it is conceivable that the stress data is acquired independently of whether the structural data imply the presence of a defect. Specifically, the stress data can be acquired after the structural data in any case. In this embodiment, the stress data is also acquired at contact points where no defect is present. This approach is particularly useful when, in addition to the specific information regarding the defect, information is also required on how the stress state at the contact point can manifest itself without a defect having occurred.

[0032] Within the framework of both of the aforementioned alternatives, it is conceivable that the inventive method is carried out for a large number of contact points, in particular vias. In particular, these are contact points of the same design, so that statistically reliable conclusions regarding the results for contact points of this design can be obtained from the findings. For example, the structural data can be recorded for several thousand contact points, with the recording of the stress data also being carried out for all of these contact points or for at least a subset of these contact points, for example, at least for those where the structural data indicates the presence of at least one defect.

[0033] Regarding the stress data, it may be necessary to acquire it using a destructive testing method. A destructive testing method is understood to be a procedure in which the acquisition of the respective data requires, or results in, the destruction of the sample. In particular, this necessitates cutting or dissecting the sample, for example, to acquire data concerning its interior.

[0034] It is conceivable that, as part of the stress data acquisition, sample preparation of the microelectronic component is carried out, including embedding the microelectronic component in an investment material, particularly acrylic or synthetic resin. The investment material preserves the stress states present in the sample, which are to be determined during the stress data acquisition, and in particular prevents these states from being altered by the creation of a parting line, for example, using a mechanical tool.

[0035] In this way, a separation plane can be created in the microelectronic component during the acquisition of stress data. In this embodiment, the sample or microelectronic component is specifically cut open to create the exposed separation or cutting plane, which would otherwise be embedded within the sample. The microelectronic component, or more precisely, the substrate, is in particular a solid, for example a crystalline one, which is cut into several pieces during the formation of the separation plane and / or whose surface is removed by a machining process.

[0036] Since the stress data to be recorded pertains to the contact point or any fault present there, the separation plane preferably runs through the contact point, particularly through the via. It is also conceivable that the separation plane bypasses the contact point, particularly the via. In order to record the stress data relating to the contact point, a sufficiently small distance between the contact point, particularly the conductor layer or the insulating layer, is required. This distance can be between 1 and 20 µm.This distance, or rather the thickness of the layer of material remaining between the parting line and the contact point, particularly the hollow interior of the via, is preferably thin enough to allow the measurement principle used to acquire the load data to capture the corresponding load data for this layer. On the other hand, the distance is large enough to prevent any thinning of this material in this area caused by the formation of the parting line from resulting in a densification of the stress lines and thus an increase in the existing stress. Regarding the specific method of creating the parting line, it may be achieved by laser cutting and / or grinding and / or milling and / or lapping and / or, in particular, chemical-mechanical and / or ion beam-based polishing and / or ion beam etching.Stress data can be acquired using Raman spectroscopy equipment. Raman spectroscopy, based on the Raman effect, is a spectroscopic method used to investigate vibrational, rotational, and other low-energy transitions in molecules. Monochromatic light, particularly laser light, generated by a light source in the Raman spectroscopy equipment, is directed onto the sample. The light is largely scattered elastically due to Rayleigh scattering, without a change in energy. However, a small fraction of the light is scattered in all directions, resulting in energy transfer between the light and the molecules of the sample. This inelastic scattering leads to a change in the frequency of the scattered light.This frequency shift corresponds to the energy of the vibrations or rotations in the molecules and provides specific information about the molecular structure and composition of the sample. From this, information regarding the locally present stress state can be obtained.

[0037] Alternatively, or in addition to the stress data, electron backscatter diffraction data can be acquired using an electron backscatter diffraction device, in particular a scanning electron microscope. Due to the comparatively low information depth, this approach requires a particularly thin layer of material remaining between the interface and the contact point, especially the via. During this process, electrons are backscattered from the sample, and the resulting interference effects allow conclusions to be drawn about the crystalline structure of the sample material. This also provides information regarding the local stress state.

[0038] Preferably, the evaluation unit performs segmentation of the structural data and / or the stress data, particularly if the respective data consists of images. The resulting image segments can be assigned to defect locations and / or predefined areas of the contact point. The image segments generated during segmentation are contiguous areas within the respective image data or image that are assigned to a defect location or a predefined area of ​​the contact point. If the contact point is the via, possible predefined areas include the base, the side wall(s), or a surface of the microelectronic component or substrate adjacent to the via. Specifically, the pixels or voxels, or each pixel or voxel, are assigned to one of the image segments.The allocation is preferably carried out automatically using the evaluation unit during the evaluation of the data.

[0039] In the method according to the invention, the evaluation unit can be used to classify the detected defect location, or at least one of the detected defect locations, with regard to the type and / or cause of the respective defect location. Thus, the appearance and / or position of the defect location, for example in the structural data and optionally taking into account the additional information obtained from the stress data, can be evaluated with regard to the presence of typical features that are typically present in a certain class of defect locations, whereby the defect location or the associated image segment can be assigned to one of several predefined classifications. The classification is also preferably carried out automatically by the evaluation unit during the data evaluation.

[0040] It is particularly preferred, based on the classification relating to the type of defect, that the respective defect is a layer delamination and / or a bulge, especially of the via. With regard to layer delamination, it is conceivable that two adjacent layers of the aforementioned type detach from each other, particularly due to shear forces occurring and caused by existing stress states, which can lead to a change in the geometric structure and ultimately the electrical properties of the via. With regard to bulge, it is conceivable that the delamination causes the outermost layer to bulge into the interior. Inconsistencies in the chemical or physical deposition can also lead to thickened areas, particularly of the conductor layer, which also results in bulging, especially frequently in the base region.

[0041] Furthermore, or alternatively, the classification relating to the cause of the defect can indicate whether it involves mechanical stress and / or a chemical process. These classifications relating to the cause of the defect are particularly relevant for defects manifesting as layer delamination. As already mentioned, mechanical stresses often exert shear forces on the layers, which can cause them to detach from one another. Similarly, chemical processes in areas where adjacent layers meet can cause the adhesion between these layers to break down, leading to their detachment. If the layers in question were produced by chemical and / or physical deposition, the classification relating to the cause of the defect can indicate whether it involves a defective deposition process.whether it involves a mechanical stress and / or a chemical process. This classification, which relates to the cause of the defect, applies particularly to bulges.

[0042] The evaluation of structural and / or load data to identify at least one defect and / or determine its possible cause can be performed using artificial intelligence. Compared to evaluation of the available data by a human operator, the use of artificial intelligence increases efficiency with regard to the resources required, especially time. Furthermore, artificial intelligence advantageously enables the consideration of diverse circumstances and interrelationships that would be overwhelming for a human operator. Consequently, the result of the image data evaluation—that is, the identification of any defects and, if applicable, segmentation and / or classification—is as promising as possible in terms of the correct identification of defects.Particularly preferred is the implementation of artificial intelligence as a trained model generated by machine learning. In this embodiment, the evaluation of structural and load data is performed simultaneously, i.e., by combining the structural and load data, so that all results obtained can be determined based on all available data.

[0043] Furthermore, the present invention relates to a computer-implemented method for generating a trained model which, within the framework of the evaluation of the structural data and / or the load data, can be used to detect the at least one defect location and / or to determine its possible cause according to the method described above, wherein the method comprises the following steps: - Specifying at least one training input data set, - Specifying a training result that is assigned to at least one training input data set, - Training a model based on at least one training input data set and the training result, thereby obtaining the trained model.

[0044] The model undergoes training, or machine learning. This allows the trained model to perform cognitive functions that correspond to, or at least closely resemble, human thinking. Through training, the model is fundamentally capable of uncovering and utilizing previously unrecognized relationships and patterns. Thus, the model's ability to determine quantities, circumstances, and / or relationships can be further developed and improved through training. The training process can be carried out in separate steps or cycles, with the model continuously improving. Specifically, supervised training can be performed, but unsupervised training is also conceivable.

[0045] Real, historical datasets, specifically structural and stress data from contact points, can be used as training input data. If supervised training is performed, user-defined results can be specified as ideal solutions, which then serve as the training results. The results generated by the model during training can be compared to these ideal solutions, with the goal of minimizing the deviations between the training results and the generated results. When using a neural network-based model, the image data, or rather the training input datasets, are fed into the model via an input state, and the evaluation results are provided via an output state of the model.A large number of further layers can be present between these, each comprising a certain number of nodes, between which connections are formed during training to create a neural network. Details regarding this are well known to those skilled in the art and are therefore not explained further. It should also be noted that all advantages, features, and aspects discussed in connection with the inventive method described above are equally transferable to this inventive method and vice versa.

[0046] Furthermore, the present invention relates to a method for detecting at least one defect at a contact point, in particular at least one via of a microelectronic component, and for determining its possible cause, wherein, within the framework of a non-destructive testing method, structural data relating to the contact point are acquired, wherein the structural data are supplied to an evaluation device, by means of which an evaluation of the structural data is carried out to detect the at least one defect and to determine its possible cause, wherein the evaluation of the structural data is carried out using artificial intelligence, wherein the artificial intelligence is based on a trained evaluation model generated by means of machine learning.The training of the evaluation model was carried out using at least one evaluation training input data set and at least one evaluation training result, which were available during the execution of the inventive method described above. All advantages, features, and aspects explained in connection with the inventive methods described above are equally transferable to this inventive method and vice versa.

[0047] This procedure outlines a possible way to further utilize the results obtained during the above-described process. Specifically, the structural data obtained through this process, along with the results derived from it using the load data, serve as a training dataset for the evaluation model and the artificial intelligence. The goal is to enable the evaluation model to generate the desired results based solely on the structural data, without requiring the load data.Since stress data is typically acquired through destructive testing, this approach offers a way to obtain results of the same scope as the method described above, particularly regarding the potential cause of the detected defect, without requiring the destruction of the microelectronic component to obtain the structural data. The present method thus provides a means of non-destructive quality control for microelectronic components. The microelectronic components and their contact points used here are preferably identical in design and structure to those obtained during the evaluation training process.

[0048] Furthermore, the present invention relates to a computer-implemented method for generating a trained evaluation model which can be used in the evaluation of the structural data to detect the at least one defect location and / or to determine its possible cause in the method just described, wherein the method comprises the following steps: - Specifying at least one evaluation training input data set comprising structural data that were used as input data in the previously carried out, introductory method according to the invention, - Specifying an evaluation training result that is assigned to the at least one evaluation training input data set and was obtained as a result data set relating to the detection of the at least one fault location and / or the determination of its possible cause based on the input data within the framework of the previously carried out, introductory, inventive method, - Training a model based on at least one evaluation training input data set and at least one evaluation training result, thereby obtaining the trained evaluation model.

[0049] All advantages, features and aspects explained in connection with the above-described methods according to the invention are equally transferable to this method according to the invention and vice versa.

[0050] Furthermore, the present invention relates to a computer-readable storage medium. According to the invention, the problem is solved in such a storage medium by the fact that it comprises instructions which, when executed by means of a processing unit designed as a computer, cause the processing unit to carry out at least one of the previously described methods. All advantages, features, and aspects described in connection with the methods of the invention described above are equally transferable to the storage medium of the invention and vice versa.

[0051] Finally, the present invention relates to an evaluation device. According to the invention, the problem is solved in such an evaluation device by the fact that it comprises a computer-readable storage medium as described in the preceding passage and a processing device, wherein the instructions, when executed by means of the processing device designed as a computer, cause the processing device to carry out at least one of the previously described methods. All advantages, features, and aspects set forth above in connection with the methods and storage medium according to the invention are equally transferable to the evaluation device according to the invention, and vice versa.

[0052] Further advantages, features, and details of the invention will become apparent from the exemplary embodiments presented below and from the figures. These show schematically: Fig. 1 a cutaway view of an area of ​​a microelectronic component in the area of ​​a contact point designed as a through-hole, Fig. 2 a flowchart of a method according to the invention in an exemplary embodiment, Fig. 3 an order to carry out a first step of the procedure of Fig. 2, comprising an evaluation device according to an exemplary embodiment with a computer-readable storage medium according to an exemplary embodiment according to an exemplary embodiment, Fig. 4 - 6 Representations of exemplary structural data, i.e., optical microscopy data, which were obtained during the execution of the procedure based on the Fig. The procedures described in section 2 can be used to obtain the winnings. Fig. 7. Perspective view of a prepared and cut-open microelectronic component, as used in the execution of the procedure based on the Fig. The process described in section 2 is generated, and Fig. 8 - 10 each show a flowchart of further methods according to the invention, each according to an exemplary embodiment.

[0053] Fig. Figure 1 shows a sectional view of a microelectronic component 1 in the area of ​​a contact point, which in this case is a via 2, wherein the microelectronic component 1 is exemplified as a component of a detector of a computed tomography system. The microelectronic component 1 comprises, although in Fig. 1 only one is shown and as further explained below. Fig. 3 will become apparent, several vias arranged along a row 2. The following based on the in Fig. The aspects shown in section 1 and 2 apply equally to all other vias 2 of the microelectronic component 1.

[0054] It should be noted that the embodiment described below is specifically illustrated by the contact points provided as the vias 2, although the method according to the invention can also be carried out generally for contact points of microelectronic components. The contact points can also be direct copper-to-copper bonds (direct Cu-Cu bond) or micro-copper bonds (µ-Cu bond).

[0055] The microelectronic component 1 comprises a plate-shaped silicon substrate 3, wherein the via 2 is designed as a blind hole in the substrate 3, which was etched into the silicon. In the area of ​​the via 2, several superimposed layers 4, 5, 6 are provided, namely an insulating layer 4, a barrier layer 5, and a conductive layer 6. The insulating layer 4 was applied to the silicon of the substrate 3 by chemical, in particular electrochemical, and / or physical deposition of an insulating material and serves as electrical insulation between the via 2 and the substrate 3. The barrier layer 5 located above the insulating layer 4 serves to prevent material diffusion and to mechanically reinforce the via 2.The uppermost layer is a conductive layer 6 and consists of an electrically conductive material, namely a metal such as copper, and is formed by chemical, in particular electrochemical, and / or physical deposition. The conductive layer 6 provides the desired electrical conductivity of the via 2. Layers 4, 5, and 6 are located within the recess and also extend laterally around an opening of the via 2. The opening is located in the region of one of the two planar surfaces 18, 19 of the substrate 3. Fig. In the state shown in Figure 1, the surface 19 of the carrier 3 opposite the opening of the via 2 has carrier material that still needs to be ground off. This results in the conductor layer 6 forming an exposed contact point on the surface 19 that forms the underside of the carrier 3. Regarding chemical and physical deposition, it is provided that chemical and physical vapor deposition, respectively, has been carried out.

[0056] The via 2 has a hollow interior 7, open at the top and with an at least substantially cylindrical geometry. The hollow interior 7 is bounded by the conductor layer 6, which consists of a base 8 and a hollow cylindrical side wall 9. The via 2 enables electrical contact between the two opposing, planar surfaces 18, 19 of the substrate 3.

[0057] The aim of the inventive method described below, according to one embodiment, is to automatically detect any defects 20 and their causes in the vias 2. A flowchart of this method, comprising steps 28 to 30, is shown in the Fig. 2 shown. Typical defect locations 20 are, for example, layer delaminations and / or bulges or bulges affecting layers 4, 5, 6, which negatively affect the conductivity and thus the functionality of the via 2. Fig. Figure 3 shows a setup for carrying out the method, in which several vias 2 of the microelectronic component 1 are visible according to this illustration. Furthermore, the lower part of the carrier 3 is ground down in the shown state, so that the bottoms 8 of the vias 2 are exposed downwards as contact points.

[0058] In the first step 28 of the procedure, structural data 11 relating to the structure of the respective via 2 are determined using non-destructive testing methods. The structural data 11, obtained as image data, comprise optical microscopy data 31, X-ray microscopy data 32, computed tomography data 33, acoustic microscopy data 34, infrared microscopy data 35, lock-in thermography data 36, ​​and photoemission microscopy data 37, with details regarding the optical microscopy data 31 being explained in particular below. However, it is also conceivable that not all, but only a portion of the aforementioned data 31–37 are acquired.

[0059] Thus, to record the structural data, 11 is used in the Fig. The recognition device 23 shown in Figure 3 includes, among other things, an optical microscopy device 10 by means of which the optical microscopy data 31 are acquired. The optical microscopy device 10 comprises a microscope 12 and a camera 13, wherein the optical microscopy data 31 relating to the visually visible part of the electromagnetic spectrum are a magnified representation or a corresponding image of the respective via 2 obtained by means of the microscope 12. The via 2 has a diameter of, for example, approximately 50 µm, whereby a magnification of 500 to 1000 times is achieved by means of the microscope 12. In the illustrated embodiment, the viewing direction of the optical microscopy device 10 corresponds to an optical axis of the microscope 12, which in this case extends along the vertical direction of the Fig. 2 extends. Thus, the viewing direction of the optical microscopy device 10 is at least essentially perpendicular to the base 8 of the via 2, so that the optical microscopy data 31 each show the hollow interior 7 of the respective via 2.

[0060] The microscope 12 comprises several components, which for the sake of clarity are shown in the Fig. The following components are not shown: an eyepiece or eyepiece lens, an objective turret or lens changer, objective lenses, and a tube lens. A stage 14 with a mounting device for attaching the microelectronic component 1 is also provided. To position the via 2 at a focal point of the microscope 12, the stage 14 is movable along the optical axis of the microscope 12. With respect to the plane perpendicular to the optical axis of the microscope 12, the stage 14 is also movable in the two other spatial directions. This allows the vias 2 to be successively brought into the field of view of the microscope 12. To achieve this movement along the aforementioned directions, the stage 14 is coupled to several electromechanical actuators, which can be controlled by a control unit, which in this case also serves as an evaluation unit 15.For this purpose, the control device generates control signals output to the actuators, which cause the vias 2 to be brought successively into the field of view of the microscope 12 in order to successively acquire several sets of optical microscopy data 31.

[0061] The acquisition of optical microscopy data 31 is exemplified by bright-field microscopy, in which all light of the visually visible part of the electromagnetic spectrum is reflected at the via 2 and subsequently captured by the optical microscopy device 10. A Köhler illumination system is provided for the microscope 12 to ensure optimal illumination. The optical microscopy device 10, or the microscope 12, comprises several light sources 16 designed as LEDs and / or halogen lamps for generating circularly polarized light 17. The image data 11 is acquired under illumination of the via 2 by means of this light 17. Instead of circularly polarized light 17, linearly or elliptically polarized light or unpolarized light can also be used.

[0062] The structural data 11, further determined as the X-ray microscopy data 32, are acquired by means of an X-ray microscopy device 38 of the recognition device 23. This comprises an X-ray source and a scanning X-ray microscope, with which the via 2 is scanned point by point. Two-dimensional image data are obtained as the X-ray microscopy data 32.

[0063] The structural data 11, further determined as computed tomography data 33, are acquired using a computed tomography unit 39 of the recognition unit 23, which in this case is an X-ray computed tomography unit. For this purpose, the microelectronic component 1 is positioned between an X-ray source and a radiation detector of the computed tomography unit 39, whereby the microelectronic component 1 and the X-ray source together with the radiation detector are rotated relative to each other. Three-dimensional image data are obtained in this way as the computed tomography data 33.

[0064] The structural data 11, which are further determined as acoustic microscopy data 34, are acquired by means of an acoustic microscopy device 40 of the detection device 23. This includes an ultrasound transmitter, whereby the generated sound waves are introduced into or reflected from the microelectronic component 1, so that two- and / or three-dimensional image data, on the basis of which defects and / or material properties can be detected, in particular, are generated as acoustic microscopy data 34.

[0065] The structural data 11, further determined as infrared microscopy data 35, are acquired using an infrared microscopy device 41 of the detection device 23. This device comprises an infrared source, whereby the via 2 is illuminated with the generated infrared rays. Image data, as infrared microscopy data 35, are thus acquired, relating to a position-dependent transmittance or reflectance of the infrared radiation.

[0066] The structural data 11, which are further determined as the lock-in thermography data 36, ​​are acquired by means of a lock-in thermography device 42 of the detection device 23. For this purpose, a heat flow is introduced into the microelectronic component 1, whereby layers or defects in the interior of the microelectronic component 1 lead to inhomogeneities with respect to the heat flow, which in turn are acquired by means of a thermal imaging camera of the lock-in thermography device 42 in the form of image data.

[0067] The structural data 11, further determined as the photoemission microscopy data 37, are acquired using a photoemission microscopy device 43 of the detection device 23. For this purpose, the via 2 is irradiated with UV radiation generated by a UV source of the photoemission microscopy device 43. The resulting photoelectrons are visualized using a fluorescent screen, whereby the photoemission microscopy data 37 are available as image data, which are acquired using a CCD camera of the photoemission microscopy device 43.

[0068] Images of the recorded structural data 11 are used for various cases in the Fig. Figures 4 to 6 are shown. These figures illustrate the optical microscopy data 31. The optical microscopy data 31 assigned to each via 2 are available as a two-dimensional image, i.e., a two-dimensional array, comprising several pixels. Each pixel is assigned a value for color and a value for brightness. The optical microscopy data 31 are available for each of the vias 2 as an image of the respective hollow interior 7. In the case of the via 2 of the based on the Fig. The optical microscopy data 31 shown in section 4 reveals no defect 20. In the via 2 of the [unclear text] based on the Fig. The optical microscopy data 31 shown in section 5 reveals a central bulge 21 in the area of ​​the base 9, representing a defect 20. During the through-hole plating 2, the defect 20 was identified based on the Fig. 6 shown optical microscopy data 31 show several layer detachments 22 in the area of ​​the side wall 9 as defect locations 20.

[0069] To detect these defects 20, the structural data 11 are supplied to the evaluation device 15 according to the invention, which is implemented according to one embodiment. The evaluation device 15 performs a computer-implemented and automated evaluation of the structural data 11. The evaluation device 15 comprises a computer-readable storage medium 24 according to one embodiment and a processing unit 25. Instructions 26 are stored on the storage medium 24, which, when executed by the processing unit 25 (designed as a computer), cause it to perform the evaluation of the structural data 11 as explained below. The evaluation of the structural data 11 is carried out using instructions 26 implementing an artificial intelligence 27. The artificial intelligence 27 is implemented as a trained model generated by machine learning.

[0070] During the evaluation of the structural data 11, the image data, or structural data 11, is first segmented using artificial intelligence 27. For the resulting image segments, it is then determined whether the area of ​​the image represented in the respective segment is the floor 8, the side wall 9, or the surface 18. Furthermore, it is determined whether the respective segment indicates a defect 20. The image segments resulting from the segmentation are contiguous areas within the respective structural data 11, or the respective image, that are assigned to one of these identified areas.

[0071] In the second step 29, stress data 44 are recorded using a destructive testing method, at least for those vias 2 for which the presence of at least one defect 20 has been detected. For this purpose, sample preparation of the microelectronic component 1 is first carried out, in which the affected microelectronic component 1 is embedded in an investment material, namely acrylic or synthetic resin. Subsequently, a parting line 45 is created in the microelectronic component 1. For this purpose, the embedded microelectronic component 1 or its substrate 3 is cut open to form the exposed parting line 45.Das Aufschneiden des mikroelektronischen Bauteils 1 respektive das Freilegen der Trennebene 45 und eine gegebenenfalls erfolgende Glättung der Trennebene 45 erfolgt mittels Laserschneidens und / oder Schleifens und / oder Fräsens und / oder Läppens und / oder, insbesondere chemisch-mechanischen und / oder ionenstrahlbasierten, Polierens und / oder lonenstrahlätzens.

[0072] The cutaway microelectronic component 1 is shown in a perspective view in Fig. Figure 7 shows that although the interface 45 can pass through the via 2, it is designed to bypass the via 2, leaving a gap 46 between the via 2 and the interface 45. Since the stress data 44 are to be recorded for the area of ​​the substrate material 3 remaining between the via 2 and the interface 45, the gap 46 is sufficiently large to prevent any thinning of this material at this point caused by the formation of the interface 45, thus avoiding any increase in stress concentrations and consequently any increase in the existing stress. On the other hand, the gap 46 is small enough to allow for the recording of stress data according to a suitable measurement principle. Specifically, the gap 46 is, for example, 20 µm.

[0073] The stress data 44 are Raman spectroscopy data 47 acquired using a Raman spectroscopy device not shown in detail in the figures. For this purpose, laser light is directed onto the microelectronic component 1 in the area of ​​the via 2 or the material remaining between the interface 45 and the via 2, whereby an inelastically scattered portion of the light is evaluated with respect to a frequency shift in order to obtain information about the molecular structure and composition and thus the locally present stress or strain state. The Raman spectroscopy data 47 are also ultimately available as image data.

[0074] Furthermore, electron backscatter diffraction data 48 are acquired as stress data 44 using an electron backscatter diffraction device comprising a scanning electron microscope (not shown in detail in the figures). For this purpose, interference effects of backscattered electrons are evaluated to obtain specific information about the crystalline structure of the material of the microelectronic component 1 and thus the locally present stress state. The electron backscatter diffraction data 48 are also ultimately available as image data.

[0075] In the next step 30, artificial intelligence 27 is used to classify any detected defects 20 with regard to the type of defect 20 and the cause that led to its formation. For the latter purpose, the load data 44 are primarily used. If a classification has already been performed within the framework of the segmentation described above, and thus only based on the structural data 11, the corresponding results are now verified and corrected if necessary. The segmentation of the structural data 11 can also be performed using the load data 44, allowing the results obtained so far to be verified and corrected if necessary. Furthermore, segmentation can also be performed using the load data 44, which is available as image data.In principle, the evaluation of the structural data 11 and the load data 44 is carried out simultaneously, i.e. by combining all available data 11, 44, so that all results obtained are determined in principle depending on all available data sets.

[0076] Regarding the classification of the type of defect 22, it is indicated whether the respective defect 20 is a layer delamination 22 and / or a bulge 21. Regarding the classification of the cause of the defect 20, it is indicated whether it relates to mechanical stress and / or a chemical process, or whether it relates to a defective deposition process. In particular, if the recording of the stress data 44 has also been carried out for vias 2 where no defect 20 is present, then further information can be obtained as to the stress level in the area of ​​the via 2 at which the formation of a stress-induced defect 20 is not to be expected.

[0077] The results generated in this way are finally output via an output device not shown in detail in the figures and saved as a result data set 49, thus making it available to a user for further evaluation.

[0078] A further method according to the invention, based on an exemplary embodiment, is described below. This method is aimed at generating a trained model that is used in the evaluation of data 11 and 44 described above. A flowchart of this method, comprising steps 50 and 51, is shown in Fig. Figure 8 shows that, in the first step, 50 training input data sets 52 are provided, which are in the form of structural data 11 and load data 44 as described above. These data sets 52 are historical data, with an associated training result 53 specified for each input data set 52. The training result 53 indicates whether a defect 20 is present at the through-hole 2 assigned to the respective training input data set 52 and, if so, what the corresponding results should ideally look like with regard to classification. The results 54 obtained in the next step 51 using the artificial intelligence 27 or the model to be trained are compared with the training results 53, with one objective being to minimize the deviations between the training results 53 and the results 54 generated by the artificial intelligence 27.This will preserve the trained model described above.

[0079] Fig. Figure 9 shows a flowchart of a further method according to the invention in an exemplary embodiment, comprising steps 55 and 56, which is aimed at detecting at least one defect 20 of a via 2 of a microelectronic component 1 and at determining its possible cause. In the first step 55, structural data 11 relating to the via 2 are acquired using a non-destructive testing method, wherein the structural data are supplied to an evaluation device 15, by means of which an evaluation of the structural data 11 is carried out to detect the at least one defect 20 and to determine its possible cause. This method thus corresponds in principle to the one already described in the Fig. The procedures described in section 2 differ, however, in that the recording of the load data 44 and thus the destruction of the respective microelectronic component 1 are not provided for. This is based on the Fig. The method described in section 9 is preferably intended to be used in the course of quality control, in which the microelectronic components 1 to be examined are subsequently put to a specific use.

[0080] To avoid the problem of results that are actually obtained from the load data 44 not being obtained, the evaluation of the structural data 11 is carried out using an artificial evaluation intelligence 57, which is based on a trained evaluation model generated by machine learning, whereby the training of the evaluation model is carried out on the data 11, 44 as well as the result data set 49, which are carried out as part of the execution of the based on the Fig. The two procedures described above were obtained. Within the framework of the process based on the Fig. Section 9 of the procedure outlined provides a specific possibility regarding the further use of the results, which arises during the implementation of the procedure based on the Fig. The results were obtained using the procedures described in section 2. This enables the evaluation model or artificial intelligence 57 to obtain the corresponding results based solely on the structural data 11, without requiring the load data 44. In this case, the results obtained are also output via the output device and / or saved as a result data set 58.

[0081] To obtain robust and statistically reliable results, the following methods were used in the study based on the Fig. The procedures described in section 2 utilize a large number of vias 2, namely several thousand, for the corresponding analysis of the defect locations 20. In the case of the microelectronic components 1, which are used within the framework of the Fig. The methods described in section 9 are those of the same design as the microelectronic components 1, which are used in the context of the Fig. The two procedures described above were used.

[0082] Finally, based on the information in the Fig. Figure 10 illustrates a further method according to the invention in an exemplary embodiment, comprising steps 59 and 60, which is based on the generation of the artificial evaluation intelligence 57 and the trained evaluation model, respectively, which is used in the execution of the process based on the Fig. The procedure described in section 9 is used. In principle, this corresponds to the following: Fig. 10 explained procedures based on the Fig. 8 procedures explained.

[0083] In the first step, 59 evaluation training input data sets 61 are specified, which are the structural data 11 that are used in the execution of the based on the Fig.The procedures described in section 9 were used. Furthermore, evaluation training results 62 are specified, each assigned to one of the evaluation training input datasets 61, where the evaluation training input datasets 61 are the result datasets 49 obtained from one of the evaluation training input datasets 61 or structural data 11, respectively. The results 63 obtained in the next step 60 by means of the artificial evaluation intelligence 57 or the evaluation model to be trained are compared with the evaluation training results 62, with one objective being to minimize the deviations in this regard. This results in the trained evaluation model.

[0084] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] US 2014 / 0 064 445 A1

[0006] US 2015 / 0 192 527 A1

[0006] JP 2016 - 085 045 A

[0006] US 2015 / 0 226 676 A1

[0006] CN 1 10 911 301 A

[0006] US 10 352 879 B2

[0006] WO 2021 / 084 933 A1

[0006] US 2017 / 0 067 732 A1

[0006] WO 2023 / 054 910

[0006] US 9 810 796 B2

[0006] US 10 088 578 B2

[0006] US 10 156 644 B2

[0006] US 10 120 082 B2

[0006] US 2022 / 0 066 054 A1

[0006] US 11 076 822 B2

[0006] US 10 825 729 B2

[0006] CN 1 03 439 248 A

[0007] CN 1 02 944 336 A

[0007] CN 1 03 424 293 A

[0007] CN 1 09 297 627 A

[0007] KR 2023 168 910

[0007] Zitierte Nicht-Patentliteratur

[0000] - De Wolf, Ingrid, Kristof Croes, and Eric Beyne. „Expected failures in 3-D technology and related failure analysis challenges.“ IEEE Transactions on Components, Packaging and Manufacturing Technology 8.5 (2018): 711-718

[0003] Vartanian, Victor, et al. „Metrology needs for through-silicon via fabrication.“ Journal of Micro / Nanolithography, MEMS, and MOEMS 13.1 (2014): 011206-011206

[0003] Tsuto, Takashi, et al. „Advanced through-silicon via inspection for 3D integration.“ Transactions of The Japan Institute of Electronics Packaging 6.1 (2013): 13-17

[0003] Li, Yan, Purushotham Kaushik Muthur Srinath, and Deepak Goyal. „A review of failure analysis methods for advanced 3D microelectronic packages.“ Journal of Electronic Materials 45 (2016): 116-124

[0003] - Bender, Hugo, et al. „Structural characterization of through silicon vias.“ Journal of Materials Science 47 (2012): 6497-6504

[0003] Kia, Alireza M., et al. „ToF-SIMS 3D analysis of thin films deposited in high aspect ratio structures via atomic layer deposition and chemical vapor deposition.“ Nanomaterials 9.7 (2019): 1035

[0004] de Veen, P. J., et al. „High-resolution X-ray computed tomography of through silicon vias for RF MEMS integrated passive device applications.“ Microelectronics Reliability 55.9-10 (2015): 1644-1648

[0004] Cassidy, C., et al. „Depth-resolved photoemission microscopy for localization of leakage currents in through Silicon Vias (TSVs).“ 2009 16th IEEE International Symposium on the Physical and Failure Analysis of Integrated Circuits. IEEE, 2009

[0004] Jiang, Tengfei, et al. „Through-silicon via stress characteristics and reliability impact on 3D integrated circuits.“ Mrs Bulletin 40.3 (2015): 248-256

[0005] Ryu, Suk-Kyu, et al. „Stress-Induced Delamination Of Through Silicon Via Structures.“ AIP Conference Proceedings. Vol. 1378. No. 1. American Institute of Physics, 2011

[0005] - De Wolf, Ingrid, et al. „In-depth Raman spectroscopy analysis of various parameters affecting the mechanical stress near the surface and bulk of Cu-TSVs.“ 2012 IEEE 62nd Electronic Components and Technology Conference. IEEE, 2012

[0005] - Zschenderlein, U., et al. „Residual stress investigations at TSVs in 3D micro structures by HR-XRD, Raman spectroscopy and fibDAC.“ 2014 IEEE 64th Electronic Components and Technology Conference (ECTC). IEEE, 2014

[0005] Vogel, Dietmar, et al. „Stress analyses of high spatial resolution on TSV and BEoL structures.“ Microelectronics Reliability 54.9-10 (2014): 1963-1968

[0005] - Britton, T. B., and J. L. R. Hickey. „Understanding deformation with high angular resolution electron backscatter diffraction (HR-EBSD).“ IOP Conference Series: Materials Science and Engineering. Vol. 304. IOP Publishing, 2018

[0008] Wilkinson, Angus J., Graham Meaden, and David J. Dingley. „High-resolution elastic strain measurement from electron backscatter diffraction patterns: New levels of sensitivity.“ Ultramicroscopy 106.4-5 (2006): 307-313

[0008] Wilkinson, Angus J., Graham Meaden, and David J. Dingley. „High resolution mapping of strains and rotations using electron backscatter diffraction.“ Materials Science and Technology 22.11 (2006): 1271-1278

[0008]

Claims

[1] Method for detecting at least one defect (20) of a contact point, in particular at least one via (2), of a microelectronic component (1) and for determining its possible cause, wherein structural data (11) relating to the contact point are recorded, wherein the structural data (11) are supplied to an evaluation device (15) by means of which an evaluation of the structural data (11) is carried out to detect the at least one defect (20), wherein stress data (44) relating to a mechanical stress of a material present in and / or at the contact point are recorded, wherein the stress data (44) are supplied to the evaluation device (15) by means of which an evaluation of the stress data (44) is carried out to determine its possible cause. [2] Method according to claim 1, characterized by, that the structural data (11) are recorded using a non-destructive testing method. [3] Method according to claim 2, characterized by , that as the structural data (11) - optical microscopy data (31) using an optical microscopy device (10), relating in particular to at least a part of the visually visible range of the electromagnetic spectrum, and / or - X-ray microscopy data (32) using an X-ray microscopy device (38) and / or - Computed tomography data (33) using a computed tomography device (39) and / or - acoustic microscopy data (34), in particular relating to acoustic surface waves, using an acoustic microscopy device (40) and / or - Infrared microscopy data (35) using an infrared microscopy device (41) and / or - Lock-in thermography data (36) using a lock-in thermography device (42) and / or - Photoemission microscopy data (37) are acquired using a photoemission microscopy device (43). [4] Method according to any of the preceding claims, characterized by , that - either by checking the structural data (11) to see if at least one fault location (20) is present at the respective contact point, whereby the load data (44) are only recorded in this case, - or the load data (44) are in any case recorded after the structural data (11) have been recorded. [5] Method according to any of the preceding claims, characterized by , that the stress data (44) are collected using a destructive testing method. [6] Method according to claim 5, characterized by, that as part of the acquisition of the stress data (44) a sample preparation is carried out on the microelectronic component (1), comprising embedding the microelectronic component (1) in an investment material, in particular in acrylic or synthetic resin, wherein a parting line (45) is created on the microelectronic component (1) as part of the acquisition of the stress data (44), wherein the parting line (45) extends through or past the contact point, in particular at a distance (46) between one and 20 micrometers, wherein the parting line (45) is created by means of laser cutting and / or grinding and / or milling and / or lapping and / or, in particular chemical-mechanical and / or ion beam-based, polishing and / or ion beam etching. [7] Method according to any of the preceding claims, characterized by , that as the load data (44) - Raman spectroscopy data (47) using a Raman spectroscopy device and / or - Electron backscatter diffraction data (48) are acquired using an electron backscatter diffraction device. [8] Method according to any of the preceding claims, characterized by , that by means of the evaluation device (15) a classification with regard to the type and / or the cause of the respective fault location (20) is carried out for the detected fault location (20) or for at least one of the detected fault locations (20). [9] Method according to any of the preceding claims, characterized by, that the classification relating to the type of defect (20) indicates whether the defect (20) is a layer delamination (22) and / or a bulge (21), and the classification relating to the cause of the defect (20) indicates whether it is a mechanical stress and / or a chemical process. [10] Method according to any of the preceding claims, characterized by , that the evaluation of the structural data (11) and / or the load data (44) to detect the at least one fault location (20) and / or to determine its possible cause is carried out using artificial intelligence (27), in particular a trained model generated by machine learning. [11] Computer-implemented method for generating a trained model that can be used in the evaluation of the structural data (11) and / or the load data (44) to detect the at least one defect location (20) and / or to determine its possible cause in the method according to any one of claims 1 to 10, wherein the method comprises the following steps: - Specifying at least one training input data set (52), - Specifying a training result (53) that is assigned to at least one training input data set (53), - Training a model based on the at least one training input data set (52) and the at least one training output (53), thereby obtaining the trained model. [12] Method for detecting at least one defect (20) of a contact point, in particular at least one via (2), of a microelectronic component (1) and for determining its possible cause, wherein structural data (11) relating to the contact point are recorded as part of a non-destructive testing procedure, wherein the structural data (11) are supplied to an evaluation device (15) by means of which an evaluation of the structural data (11) is carried out to detect the at least one defect (20) and to determine its possible cause, wherein the evaluation of the structural data (11) is carried out using an artificial evaluation intelligence (57), wherein the artificial evaluation intelligence (57) is based on a trained evaluation model generated by means of machine learning,wherein the training of the evaluation model was carried out using at least one evaluation-training input data set (61) and at least one evaluation-training result (62) which were available during the execution of the method according to one of claims 1 to 10. [13] Computer-implemented method for generating a trained evaluation model that can be used in the evaluation of the structural data (11) to detect the at least one defect location (20) and / or to determine its possible cause in the method according to claim 12, wherein the method comprises the following steps: - Specifying at least one evaluation training input data set (61) comprising structural data (11) that were used as input data in the previously carried out method according to one of claims 1 to 10, - Specifying an evaluation training result (62) that is assigned to the at least one evaluation training input data set (61) and was obtained as a result data set (49) relating to the detection of the at least one fault location (20) and / or the determination of its possible cause based on the input data within the framework of the previously carried out method according to one of claims 1 to 10, - Training a model based on the at least one evaluation training input data set (61) and the at least one evaluation training result (62), thereby obtaining the trained evaluation model. [14] Computer-readable storage medium (24) comprising instructions (26) which, when executed by means of a processing device (25) designed as a computer, cause the processing device (25) to carry out the method according to any of the preceding claims. [15] Evaluation device comprising a computer-readable storage medium (24) according to claim 14 and a processing device (25), wherein the instructions (26), when executed by means of the processing device (25) designed as a computer, cause the processing device (25) to carry out the method according to any one of claims 1 to 13.

Images