Various NVM technologies are being explored for neuromorphic system realization, including resistive RAM, ferroelectric RAM, phase change RAM, spin transfer torque RAM, and NAND flash. This article discusses the potential of RRAM for such applications and evaluates key performance and reliability metrics in the context of neural network image classification. The authors conclude that the accuracy-power tradeoff may be further improved using alternative material stacks and multi-layer dielectrics so as to achieve better control of the oxygen vacancy or metallic filamentation process that governs RRAM switching characteristics.

Four-dimensional scanning transmission electron microscopy (4D-STEM) is a spatially resolved electron diffraction technique that records the electron scattering distribution at each point of the electron beam raster, thereby producing a four-dimensional dataset. The final article in this series covers ptychography, a form of computational imaging that recovers the phase information imparted to an electron beam as it interacts with a specimen.

This article presents a comprehensive study of physical inspection and attack methods, describing the approaches typically used by counterfeiters and adversaries as well as the risks and threats created. It also explains how physical inspection methods can serve as trust verification tools and provides practical guidelines for making hardware more secure.

X-ray ptychography, as recent studies show, has the potential to bridge the gap that currently exists between conventional X-ray imaging and electron microscopy. This article covers the evolution of the technology from basic 2D imaging to computed tomography to 3D ptychographic X-ray laminography (PyXL) with zoom. To demonstrate the capabilities of PyXL, a 16-nm FinFET logic IC was mechanically polished to a thickness of 20 µm and several regions were imaged at various levels of resolution.

Deprocessing of ICs is often the final step for defect validation in FA cases with limited fault-isolation information. This article presents a workflow for deprocessing ICs from the backside using automated thinning and large-area plasma FIB delayering. Advantages to this approach include a reduction in manual planarization and depackaging and a higher degree of precision and repeatability.

This article discusses the current state of large area integrated circuit deprocessing, the latest achievements in the development of automated deprocessing equipment, and the potential impact of advancements in gas-assisted etching, ion source alternatives, compact spectroscopy, and high-speed lasers.

This article proposes a design for a real-time Trojan detection system and explores possible solutions to the challenge of large-scale SEM image acquisition. One such solution, a deep-learning approach that generates synthetic micrographs from layout images, shows significant promise. Learning-based approaches are also used to both synthesize and classify cells. The classification outcome is matched with the design exchange format file entry to ensure the purity of the underlying IC.

Integrated circuits embedded in everyday devices face an increased risk of tampering and intrusion. In this article, the authors explain how reverse engineering techniques, including automated image analysis, can be employed to provide trust and assurance when dealing with commercial off-the-shelf chips.

This article describes a novel method for improving image resolution achieved using time-resolved photon emission techniques. Instead of directly generating images from photon counting, all detected photons are displayed as a point cloud in 3D space and a new higher-resolution image is generated based on probability density functions associated with photon distributions. Unsupervised learning algorithms identify photon distribution patterns as well as fainter emission sources.

This column assesses the current state and outlook for superconducting device technology and its application in exascale computing.

X-ray computed tomography is a noninvasive technique that can reveal the internal structure of objects in three dimensions with spatial resolution down to 50 nm. This article discusses the basic principles of this increasingly important imaging technology and presents examples of its use on various types of defects in semiconductor packages.

This article discusses some of the early uses of emission microscopy in semiconductor device failure analysis and the challenges that were overcome to make it the invaluable tool it is today. One of the impediments early on was a misconception that silicon cannot emit light when, in fact, it has several light emission mechanisms that have proven useful in electron microscopy. One such mechanism, avalanche luminescence, occurs in junctions during reverse breakdown and is useful for resolving low breakdown voltage and problems with ESD protection circuits. Other light emission mechanisms discussed in the article include forward bias emission, MOS transistor saturation, and dielectric luminescence, which is used to examine oxide test structures and detect oxide defects.

Transmission electron microscopy (TEM) plays an important role semiconductor process development, defect identification, yield improvement, and root-cause failure analysis. At the same time, however, certain artifacts of specimen preparation and imaging present barriers for linear scaling of TEM techniques. This article assesses these challenges and explains how electron tomography is being used to overcome them.

This article describes a proposed novel metric to furnish chip designers with a prognostic tool for x-ray imaging in the pre-silicon stage. This metric is fashioned to provide designers with a concrete measure of how visible the fine-pitched features of their designs are under x-ray inspection. It utilizes a combination of x-ray image data collection, analysis, and simulations to evaluate different design elements.

This is the second article in a two-part series on deep ultraviolet (DUV) microscopy. The first part, published in the February 2000 issue of EDFA, discusses the working principles and capabilities of the method and the types of applications for which it is suited. In this issue, the author describes the basic design of DUV microscopes, the role of major components, and their effect on imaging quality.

Trapped ion systems are one of the leading technology platforms for quantum computing. This article describes the construction and operation of ion trap devices and the various modes of failure that have been observed. Examples of failure in either the rendering or use of packaged trap chips are presented, including electrode shorts and opens, detached bond wires, and RF breakdown damage.

This article proposes the MicroStructural Hierarchy Descriptor (µSHD) as a systematic and quantitative approach to spectra and image data in microelectronics failure analysis. It discusses concrete routes for employing µSHD directly as the quantitative descriptor for supervised and unsupervised machine learning. The authors propose that µSHD tools can be used to automate and improve characterization techniques and image processing and analysis protocols.

This article provides a systematic overview of knowledge-based and machine-learning AI methods and their potential for use in automated testing, defect identification, fault prediction, root cause analysis, and equipment scheduling. It also discusses the role of decision-making rules, image annotations, and ontologies in automated workflows, data sharing, and interoperability.

Engineers at IBM’s Watson Research Center are contending with one of the most fundamental limitations of imaging technology: the tradeoff between spatial resolution and field of view. In this article, they explain how they created tool interfaces, control and automation software, and image analysis and stitching algorithms, enabling photon emission and laser scanning microscopes to produce high-resolution mosaic images of advanced processor cores and other large-area ICs. They describe some of the challenges they faced and explain how their technology can be used to create images based on reflected light, induced voltage, photon emission, and laser stimulation signatures. In one of the latest demonstrations, the technology was used to land and focus a SIL more than 4000 times, acquiring some 16,000 images that were composed into stitched mosaics of several hundred images each.

A deep learning-based nondestructive approach for void segmentation in BGA solder balls using 3D x-ray microscopy is presented.