Ge-rich Ge2Sb2Te5(GGST) is considered as one of the best candidates for industrial phase change memory production. GGST memory cells are generally embedded with Si or Ti nitride layers to prevent oxidation, as it leads to an undesired decrease of the GGST crystallization temperature. Furthermore, GGST films are usually doped with elements such as N, C, O, or Bi, aiming to delay GGST crystallization during the fabrication process as well as during memory cell operation. In this work, ultrahigh vacuum thermal desorption spectroscopy (TDS) was performed during isochronal annealing of a N-doped GGST film covered by a 10 nm-thick TiNxlayer. Desorption is observed before GGST crystallization, but the comparison between TDS andin situx-ray diffraction measurements shows that the main desorption peak, observed between 653 K and 703 K, occurs after GGST full crystallization. The most prominent desorbing species are Ar, N2, H2, and H. These results show that the TiNxpolycrystalline layer cannot prevent N atoms from leaving the GGST layer during annealing, suggesting a progressive change of the N-doped GGST chemical composition during thermal annealing and crystallization.

Simulation of atomic redistribution in Ge-Sb-Te (GST)-based memory cells during SET/RESET cycling is needed in order to understand GST memory cell failure and to design improved non-volatile memories. However, this type of atomic scale simulations is extremely challenging. In this work, we propose to use a simplified GST system in order to catch the basics of atomic redistribution in Ge-rich GST (GrGST) films using atomistic kinetic Monte Carlo simulations. Comparison between experiments and simulations shows good agreements regarding the influence of Ge excess on GrGST crystallization, as well as concerning the GST growth kinetic in GrGST films, suggesting the crystallized GST ternary compound to be off-stoichiometric. According to the simulation of atomic redistribution in GrGST films during SET/RESET cycling, the film microstructure stabilized during cycling is significantly dependent of the GST ternary phase stoichiometry. The use of amorphous layers exhibiting the GST ternary phase stoichiometry placed at the bottom or at the top of the GrGST layer is shown to be a way of controlling the microstructure evolution of the film during cycling. The significant evolution of the local composition in the amorphous solution during cycling suggests a non-negligible variation of the crystallization temperature with operation time.

In microelectronics, recently developed 3D integration offers the possibility to stack the dice or wafers vertically instead of putting their different parts next to one another, in order to save space. As this method becomes of greater interest, the need for 3D imaging techniques becomes higher. We here report a study about different 3D characterization techniques applied to copper pillars, which are used to stack different dice together. Destructive techniques such as FIB/SEM, FIB/FIB, and PFIB/PFIB slice and view protocols have been assessed, as well as non-destructive ones, such as laboratory-based and synchrotron-based computed tomographies. A comparison of those techniques in the specific case of copper pillars is given, taking into account the constraints linked to the microelectronics industry, mainly concerning resolution and sample throughput. Laboratory-based imaging techniques are shown to be relevant in the case of punctual analyses, while synchrotron based tomographies offer highly resolved volumes for larger batches of samples.

The diffraction patterns acquired with a transmission electron microscope (TEM) contain Bragg reflections related to all the crystals superimposed in the thin foil and crossed by the electron beam. Regarding TEM-based orientation and phase characterisation techniques, the nondissociation of these signals is usually considered as the main limitation for the indexation of diffraction patterns. A new method to identify the information related to the distinct but overlapped grains is presented. It consists in subtracting the signature of the dominant crystal before reindexing the diffraction pattern. The method is coupled to the template matching algorithm used in a standard automated crystal orientation mapping tool (ACOM-TEM). The capabilities of the approach are illustrated with the characterisation of a NiSi thin film stacked on a monocrystalline Si layer. Then, a subtracting-indexing cycle applied to a 70 nm thick thin foil containing polycrystalline tungsten electrical contacts shows the capability of the technique to recognise small nondominant grains.

Abstract 3D integration takes more and more importance in the microelectronics industry. This paper focuses on two types or objects, which are copper pillars (25 micrometer of diameter) and hybrid bonding samples. It aims at a statistical morphology observation of hybrid bonding structures, which underwent an electromigration test at 350 deg C and 20 mA. The goal of the study is two-fold. It is both to limit the overall time needed to perform a whole process flow, from sample preparation to reconstructed volume, and to limit the time of human intervention. To achieve this goal, three strategies are presented: improving the sample preparation scheme, reducing the number of projections with iterative algorithms and the Structural SIMilarity function, and automating the post-processing. The post-processing of the data is fully automated and directly renders the reconstructed volume. The high signal to noise ratio allows for further segmentation and analysis.

The spatial resolution and the indexing quality obtained with an automated orientation and phase mapping tool are analyzed for different Transmission Electron Microscope (TEM) illumination settings. The electron probe size and convergence angle are studied for two TEM configuration modes referred as microprobe and nanoprobe modes. Using a 10μmC2 aperture in a FEI Tecnai F20 (S)TEM, the nanoprobe mode is used to get a small convergent electron beam while the microprobe mode provides a nearly parallel illumination at the cost of a larger probe size. The nanoprobe configuration enables to increase the spatial resolution (∼1nm vs 3nm) but also affects the fraction of mis-indexed points (15% vs 1%). Indexing errors are attributed to the increase by a factor of three of the convergence angle with respect to the microprobe mode. While intermediate optimum settings may be found and are potentially achievable on electron microscopes providing a 'free lens' control or a larger choice of C2 apertures, it is emphasized that the spatial resolution cannot be considered without reference to the indexing quality and, consequently to the convergence angle.

Abstract To get both the resolution and the field of view needed, 3Di devices are characterized in this paper using phase-contrast X-ray tomography performed in a synchrotron source. The paper shows how the synchrotron-based tomography can be routinely used as a tool for failure analysis, and how some strategies can be applied to make those analyses more time-efficient and automatic without any loss of resolution. It presents and assesses the possibilities offered by a synchrotron radiation facility such as European Synchrotron Radiation Facility for the field of failure analysis in microelectronics. The paper illustrates those possibilities through two main examples, based on two different types of connection of bottom and top tiers in 3D integration, either thermocompression with copper pillars or hybrid bonding using copper pads. Several strategies have been successfully tested for the data acquisition to be faster and to limit the needed human intervention as much as possible.

3D stacked Image sensor is the stacking of a BackSide Illuminated (BSI) CMOS Image Sensor on a logic die. It enables compact size, higher performances and additional functionalities compared to standard BSI sensors. The highest footprint reduction is obtained with 3D hybrid bonding with metal interconnects between top and bottom tiers. Hybrid bonding process with oxide / copper direct bonding allows the highest scalability of interconnect pitch. In this study we present the morphological and electrical characterizations of a test vehicle. The hybrid bonding of wafers from two different technology nodes is performed using a dual damascene integration for the hybrid bonding level. The main parameters to assess the bonding interface quality are analyzed such as the influence of the pad design, the impact of reworkability and wafer-to-wafer overlays. The process robustness is studied through reliability tests and electromigration measurements.

Materials microstructure is a source of variability in devices performance considering today’s transistor feature sizes. As the well-known Electron BackScatter Diffraction (EBSD) technique [1] shows limitations to characterize grains orientation smaller than a hundred nanometers, new tools have been developed over the past few years to characterize the texture of crystalline materials at nanometer scale. Among them is the ASTAR tool based on the acquisition and indexation of Precession Electron Diffraction (PED) patterns acquired in scanning mode [2]. Although its use has proved to be efficient to analyze nano-crystalline materials [3], indexation issues appear when crystal grains size is significantly smaller than the lamella thickness. For such cases, the acquired images are composed of a superimposition of several diffraction patterns. Strong expertise and time is required to prepare ultra-thin TEM lamellae adapted for the analysis of current nanometer scale transistor devices so that only the feature of interest remains in the thickness while left crystalline. As an alternative, a dedicated procedure is proposed to overcome crystal overlapping issues and exploit volume information: each diffraction pattern is iteratively re-indexed after subtraction of the reflections related to the main—or previous— solution. In this work, this technique is used to study the microstructure of a material when several crystalline phases remain in the TEM lamella thickness. The sample observed was a planar section of nickel silicide (NiSi) thin films (~30nm) on Silicon-On-Insulator (SOI) where some monocrystalline silicon (Si) remained in the thickness of a typical 50-70nm lamella. A scheme of the studied sample is shown in figure 1a. The

X-ray tomography is widely used in materials science. However, X-ray scanners are often based on polychromatic radiation that creates artifacts such as dark streaks. We show this artifact is not always due to beam hardening. It may appear when scanning samples with high-Z elements inside a low-Z matrix because of the high-Z element absorption edge: X-rays whose energy is above this edge are strongly absorbed, violating the exponential decay assumption for reconstruction algorithms and generating dark streaks. A method is proposed to limit the absorption edge effect and is applied on a microelectronic case to suppress dark streaks between interconnections. (C) 2015 AIP Publishing LLC.

Total variation minimization is applied to the particular case of X-ray tomography in a scanning electron microscope. To prove the efficiency of this reconstruction method, noise-free and noisy data based on the Shepp & Logan phantom have been simulated. These simulations confirm that Total variation minimization-reconstruction algorithm better manages data containing low number of projections with respect to simultaneous iterative reconstruction technique or filtered backprojection, even in the presence of noise. The algorithm has been applied to real data sets, with a low angular sampling and a high level of noise. Two samples containing micro-interconnections have been analyzed and 3D reconstructions show that Total variation minimization-based algorithm performs well even with 60 projections in order to properly recover a 500 nm diameter void inside a copper interconnection.

In microelectronics, more and more attention is paid to the physical characterization of interconnections, to get a better understanding of reliability issues like voiding, cracking and performance degradation. Those interconnections have a 3D architecture with features in the deep sub-mu m range, requiring a probe with high spatial resolution and high penetration depth. Third generation synchrotron sources are the ideal candidate for that, and we show hereafter the potential of synchrotron radiation-based hard x-ray nanotomography to investigate the 2D and 3D morphology of through silicon vias (TSVs) and copper pillars, using projection (holotomography) and scanning (fluorescence) 2D and 3D imaging, based on a series of experiments performed at the ESRF. In particular, we highlight the benefits of the method to characterize voids, but also the distribution of intermetallics in copper pillars, which play a critical role for the device reliability. Beyond morphological imaging, an original acquisition scheme based on scanning Laue tomography is introduced. It consists in performing a raster scan (z,theta) of a sample illuminated by a synchrotron polychromatic radiation while recording diffraction data. After processing and image reconstruction, it allows for 3D reconstruction of grain orientation, strain and stress in copper TSV and also in the surrounding Si matrix.

The potential of X-ray nanotomography hosted in a SEM in presented in this paper. In order to improve the detail detectability of this system, which is directly related to the X-ray source size, thin metal layers have been studied and installed in the equipment. A 3D resolution pattern has been created in order to determine the smallest detectable features by this setup. This sample is a 25 μm diameter copper pillar in which size-controlled holes have been milled using a plasma-focused ion beam. This pattern has then been scanned and the resulting 3D reconstruction demonstrates that the instrument is able to detect 500 nm diameter voids in a copper interconnection, as used in 3D integration.

3D integration is a promising technology to overcome miniaturization challenges and chips densification. It consists in increasing the number of components by realizing vertical pileups of interconnected chips. Solder balls, through silicon vias (TSVs) and copper pillars are widely used for that purpose. It is essential to characterize these metallic structures in order to validate the fabrication process and to ensure an optimal connection between the different chips. Here, X-ray tomography is proposed as a non-destructive technique to investigate metallic interconnections. In particular, the potential of an innovative computerized tomography (CT) system, the X-ray tomography hosted in a scanning electron microscope (SEM), is presented. Since sample preparation is an important step of the X-ray characterization process, the use of a Plasma Focused Ion Beam (P-FIB) to extract large chunks out of the bulk silicon is reported. We expose an original work towards systematic studies: in order to validate a new copper filling chemistry, five samples containing TSVs at various filling rates have been scanned and the 3D results are shown. Chunks containing completely filled TSVs have then been scanned and their resulting 3D reconstructions demonstrate that the instrument is able to detect 500 nm diameter voids. The advantages and limitations of this characterization method are finally pointed out.

As of 45 nm node, ultralow k (ULK) materials (εr < 2.55) are widely used in microelectronic interconnects to reduce signal propagation delay. In order to get such low dielectric constant, most of these ULK are obtained using subtractive method: after co-deposition of matrix and porogen a UV cure process is used to remove labile species (porogen). During this process, porosity is created and films are densified as shrinkage increases. This paper presents an in-depth study of the UV cure process. Two kinetic models are presented to describe shrinkage rate: the nth order and the autocatalytic model. These models also used to describe photo-polymerization shrinkage-strain of dental composite, gives predictions that are in good agreement with experimental data. We find that nth-order model is best suited to describe ULK transformations during UV cure process. CHx bonds loss and carbon content can also be modeled using same kinetic model. In order to be able to build the model, initial SiOCH-CxHy film has to be over-cured to determine the maximum conversion that film can reach when submitted to extra-long cure times. Results show that such SiOCH materials have to be cured more than 250 times their nominal cure time to reach maximum shrinkage (∼50%) and a complete loss of carbon, making them looking like SiO2. When structural changes are observed at different temperature, activation energy of reactions can be determined. In this way, shrinkage reaction has an activation energy close to 1000–1100 kJ/mol and CHx bonds or carbon loss reactions have activation energy in the range of 250–400 kJ/mol, which is pretty close to C-H binding energy.

The growing demand for wireless multimedia applications (smartphones, tablets, digital cameras) requires the development of devices combining both high speed performances and low power consumption. A recent technological breakthrough making a good compromise between these two antagonist conditions has been proposed: the 28-14nm CMOS transistor generations based on a fully-depleted Silicon-on-Insulator (FD-SOI) performed on a thin Si film of 5-6nm. In this paper, we propose to review the TEM characterization challenges that are essential for the development of extremely power-efficient System on Chip (SoC).

While microelectronic devices are frequently characterized with surface-sensitive techniques having nanometer resolution, interconnections used in 3D integration require 3D imaging with high penetration depth and deep sub-micrometer spatial resolution. X-ray tomography is well adapted to this situation. In this context, the purpose of this study is to assess a versatile and turn-key tomographic system allowing for 3D x-ray nanotomography of copper pillars. The tomography tool uses the thin electron beam of a scanning electron microscope (SEM) to provoke x-ray emission from specific metallic targets. Then, radiographs are recorded while the sample rotates in a conventional cone beam tomography scheme that ends up with 3D reconstructions of the pillar. Starting from copper pillars data, collected at the European Synchrotron Radiation Facility, we build a 3D numerical model of a copper pillar, paying particular attention to intermetallics. This model is then used to simulate physical radiographs of the pillar using the geometry of the SEM-hosted x-ray tomography system. Eventually, data are reconstructed and it is shown that the system makes it possible the quantification of 3D intermetallics volume in copper pillars. The paper also includes a prospective discussion about resolution issues.

The robust development of exotic devices strongly relies on physical characterization to better understand the stacked layers architecture and the resulting morphology. In such a context, a silicon nanowire, 200nm long and of around 15nm of diameter, encapsulated by a high-k metal gate stack (TiN/HfSiON) of few nanometers per layer, is characterized by 2D and 3D X-ray energy dispersive spectroscopy in a STEM to evaluate the process flow methodology.

Three-dimensional focused ion beam/scanning electron microscopy (FIB/SEM tomography) is currently an important technique to characterize in 3D a complex semiconductor device or a specific failure. However, the industrial context demands low turnaround time making the technique less useful. To make it more attractive, the following study focuses on a specific methodology going from sample preparation to the final volume reconstruction to reduce the global time analysis while keeping reliable results. The FIB/SEM parameters available will be first analyzed to acquire a relevant dataset in a reasonable time (few hours). Then, a new alignment strategy based on specific alignment marks [using tetraethoxylisane (TEOS) and Pt deposition] is proposed to improve the volume reconstruction speed. These points combined represent a considerable improvement regarding the reliability of the results and the time consumption (gain of factor 3). This method is then applied to various case studies illustrating the benefits of the FIB/SEM tomography technique such as the precise identification of the origin of 3D defects, or the capability to perform a virtual top-down deprocessing on soft material not possible by any mechanical solution.