III IV 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62

Nov_EDFA_Digital

A RESOURCE FOR TECHNICAL INFORMATION AND INDUSTRY DEVELOPMENTS NOVEMBER 2023 | VOLUME 25 | ISSUE 4 ELECTRONIC DEVICE FAILURE ANALYSIS edfas.org SUPERCONDUCTING X-RAY SENSORS FOR TOMOGRAPHY MULTI-PROBE TOMOGRAPHIC AFM FOR RESEARCH AND FA VOLTAGE CONTRAST WITHIN ELECTRON MICROSCOPY LASER-BASED COPPER DEPOSITION FOR SEMICONDUCTOR DEBUG 4 20 12 28 2022 WINNERS − PAGE 56

A RESOURCE FOR TECHNICAL INFORMATION AND INDUSTRY DEVELOPMENTS NOVEMBER 2023 | VOLUME 25 | ISSUE 4 ELECTRONIC DEVICE FAILURE ANALYSIS edfas.org SUPERCONDUCTING X-RAY SENSORS FOR TOMOGRAPHY MULTI-PROBE TOMOGRAPHIC AFM FOR RESEARCH AND FA VOLTAGE CONTRAST WITHIN ELECTRON MICROSCOPY LASER-BASED COPPER DEPOSITION FOR SEMICONDUCTOR DEBUG 4 20 12 28 2022 WINNERS − PAGE 56

edfas.org 1 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 DEPARTMENTS Laser-based Copper Deposition for Semiconductor Debug Applications Michael DiBattista, Scott Silverman, and Matthew M. Mulholland Laser-assisted copper deposition provides a key technology for analyzing complex packaging and integrated circuit challenges. Author Guidelines Author guidelines and a sample article are available at edfas.org. Potential authors should consult the guidelines for useful information prior to manuscript preparation. 4 12 A RESOURCE FOR TECHNICAL INFORMATION AND INDUSTRY DEVELOPMENTS NOVEMBER 2023 | VOLUME 25 | ISSUE 4 edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS 2 GUEST EDITORIAL Edward I. Cole, Jr. 36 EDFAS AWARDS 36 ASM AWARDS 37 CALL FOR PAPERS 38 ISTFA 2023 EXHIBITORS LIST 40 ISTFA 2023 EXHIBITOR SHOWCASE 42 BOARD CANDIDATE PROFILES James Demarest 46 UNIVERSITY HIGHLIGHT Ted Kolasa 48 DIRECTORY OF FA PROVIDERS Rosalinda Ring 50 TRAINING CALENDAR Rosalinda Ring 52 LITERATURE REVIEW Michael R. Bruce 54 PRODUCT NEWS Ted Kolasa 57 GUEST COLUMN Lesley Endrinal and Szu Huat Goh 60 ADVERTISERS INDEX An Innovative Multi-probe Tomographic Atomic Force Microscope for Materials Research and Failure Analysis D. Sharma, M. Tedaldi, L. Wouters, T. Hantschel, Patrick Hole, A.D.L. Humphris, and U. Celano This article describes recent advancements in multi-probe sensing schemes and development of a tomographic atomic force microscopy tool for materials research and failure analysis. 20 For the digital edition, log in to edfas.org, click on the “News/Magazines” tab, and select “EDFA Magazine.” Superconducting X-ray Sensors for Tomography of Microelectronics Joseph W. Fowler, Zachary H. Levine, Paul Szypryt, and Daniel S. Swetz The high energy-resolving power of superconducting x-ray detectors reduces unwanted x-ray backgrounds, uses x-ray photons efficiently, and allows for discrimination among multiple chemical elements in a sample. 12 4 Voltage Contrast within Electron Microscopy: From a Curious Effect to Debugging Modern ICs James Vickers, Blake Freeman, and Neel Leslie A scanning electron microscope system measures voltage contrast on device-under-test surfaces. 28 ABOUT THE COVER See page 56 for a description of the contest images collage on the cover. 20 28

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 2 For many years, I’ve attended ISTFA and other microelectronics meetings and there is always talk of how we have new challenges that we’ve never encountered before, and the existing infrastructure of capabilities and expertise will not meet the current and upcoming needs. Some of the presentations address the challenges and other gaps are left to future events. With each device technology advancement, failure analysis (FA) “turns another chapter,” again. So, what else is new? FA has and will pretty much always reinvent itself to serve the needs of the microelectronics community. Much of the current microsystems’ trend toward heterogeneous integration to achieve “more than Moore” performance leads the changing FA needs. Silicon ICs, photonic devices, radio frequency (RF) chiplets, and all the interconnections between them can be daunting, previously unknown problems for FA. But we’ve been here before with FA reinventing itself via clever new techniques on existing tools, new tools to employ that haven’t been seen or used before, and new combinations of tools and innovative applications to get to defect localization and root cause corrective action. At the 1994 International Reliability Physics Symposium (IRPS), Intel’s Craig Barrett suggested that in a decade reliability engineers and failure analysts would essentially be quantum mechanics (people, not the field of physics) due to the future nature of the microelectronics industry. While part of this was to make for an interesting and provocative keynote address and he may have been a little early, it is hard to argue that quantum and “modern” physics isn’t part of the state-of-the-art (SOTA) devices we produce and the tools we use to examine them. Even the concept of a scanning electron microscope being resolution limited by the electron wavelength seemed totally foreign 5-10 years ago, but now it is in the conversation. I have enough fingers to count the atoms across a SOTA transistor channel and before too long I may only need one hand. Yes, things are changing. As before, we don’t “throw in the towel,” we get proactive in responding to the challenges. The EDFAS FA Technology Roadmap and IEEE Heterogeneous Integration Roadmap are examples of leaning into what will be coming and how to address it. Identifying the gaps is a first step to solving problems. And the FA engineer is and has not been alone. Manufacturers are interested in having reliable, high-performing products for consumers and tool providers, and interested in supplying the infrastructure to meet those needs. In addition to localization capabilities, we should not forget what the “T” in ISTFA stands for. Testing will continue to be an integral part of the FA process NOVEMBER 2023 | VOLUME 25 | ISSUE 4 A RESOURCE FOR TECHNICAL INFORMATION AND INDUSTRY DEVELOPMENTS ELECTRONIC DEVICE FAILURE ANALYSIS GUEST EDITORIAL FAILURE ANALYSIS TURNS ANOTHER CHAPTER, AGAIN Edward I. Cole, Jr., Sandia National Laboratories coleei@sandia.gov edfas.org Cole (continued on page 44) PURPOSE: To provide a technical condensation of information of interest to electronic device failure analysis technicians, engineers, and managers. Nicholas Antoniou Editor/PrimeNano nicholas@primenanoinc.com Mary Anne Fleming Director, Journals, Magazines & Digital Media Joanne Miller Senior Editor Victoria Burt Managing Editor Allison Freeman Production Supervisor ASSOCIATE EDITORS Navid Asadi University of Florida Guillaume Bascoul CNES France Felix Beaudoin GlobalFoundries Michael R. Bruce Consultant David L. Burgess Accelerated Analysis Jiann Min Chin Advanced Micro Devices Singapore Edward I. Cole, Jr. Sandia National Labs Michael DiBattista Varioscale Inc. Rosine Coq Germanicus Universitié de Caen Normandie Szu Huat Goh Qualcomm Ted Kolasa Northrop Grumman Space Systems Rosalinda M. Ring Thermo Fisher Scientific Tom Schamp Materials Analytical Services LLC David Su Yi-Xiang Investment Co. Martin Versen University of Applied Sciences Rosenheim, Germany FOUNDING EDITORS Edward I. Cole, Jr. Sandia National Labs Lawrence C. Wagner LWSN Consulting Inc. GRAPHIC DESIGN Jan Nejedlik, jan@designbyj.com PRESS RELEASE SUBMISSIONS magazines@asminternational.org Electronic Device Failure Analysis™ (ISSN 1537-0755) is published quarterly by ASM International®, 9639 Kinsman Road, Materials Park, OH 44073; tel: 800.336.5152; website: edfas. org. Copyright © 2023 by ASM International. Receive Electronic Device Failure Analysis as part of your EDFAS membership. Non-member subscription rate is $175 U.S. per year. Authorization to photocopy items for internal or personal use, or the internal or personal use of specific clients, is granted by ASM International for libraries and other users registered with the Copyright Clearance Center (CCC) Transactional Reporting Service, provided that the base fee of $19 per article is paid directly to CCC, 222 Rosewood Drive, Danvers, MA 01923, USA. Electronic Device Failure Analysis is indexed or abstracted by Compendex, EBSCO, Gale, and ProQuest.

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 4 EDFAAO (2023) 4:4-11 1537-0755/$19.00 ©ASM International® SUPERCONDUCTING X-RAY SENSORS FOR TOMOGRAPHY OF MICROELECTRONICS Joseph W. Fowler1,2, Zachary H. Levine3, Paul Szypryt1,2, and Daniel S. Swetz1 1Quantum Electromagnetics Division, National Institute of Standards and Technology Boulder, Colorado 2Department of Physics, University of Colorado, Boulder 3Quantum Measurement Division, National Institute of Standards and Technology Gaithersburg, Maryland joe.fowler@nist.gov INTRODUCTION Integrated circuits (ICs) are highly complex manufactured devices. They contain billions of 3D structures covering a wide range of size scales and are composed of multiple metallic, semiconducting, and dielectric materials. Individual transistor gates and their wiring can be as small as a few nanometers, while a complete die can exceed a centimeter across. To determine the internal structure and composition of an IC after it is manufactured is an important and extremely challenging problem. Many kinds of users could benefit from tomographic analysis of microelectronic devices. Manufacturers with an eye on process control and improvement could use 3D imaging, especially for new processes still in the development or research stage. Imaging could also help researchers to connect functional failures of ICs to the physical defects that caused them. ICs are often designed and manufactured by different organizations in different countries, which raises questions of hardware security assurance. A tomographic imaging system could help to detect disabled features, hardware trojans, counterfeit designs, or other deliberate design changes introduced just before the fabrication stage. Several technical challenges hinder tomographic x-ray imaging at the sub-micrometer length scales required to analyze an IC. An intense x-ray source must be confined to a spot not much larger than the desired resolution, while its position relative to an IC sample must be both measured and controlled with similar precision. Tomographic imaging at fine spatial resolutions also demands the detection of a very large number of x-ray photons, a requirement that grows rapidly with improving resolution. Specifically, a 3D measurement of x-ray attenuation through an optically thin sample of fixed total volume requires detection of total photons scaling as the inverse fourth power of the voxels’ linear dimensions. Two powers come from the increased number of narrower voxels across any fixed cross-section, and two more arise from the reduced attenuation signal in each, thinner voxel. At the same time, making x-ray sources smaller tends to decrease the photon production rate by at least the square of the resolution scale unless the intensity can be made to grow as the source shrinks. In short, even modest improvements in Fig. 1 Design of the Tomcat instrument, shown from the outside, left, and inside the vacuum chamber, right. Right-hand figure reprinted from Ref. 3 with permission.

edfas.org 5 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 spatial resolution require producing and detecting vastly more x-ray photons and confining them to an ever-smaller, resolution-sized spot. The extreme demands for x-ray intensity mean that published 3D analysis of ICs at nanoscales has generally been the province of research based at synchrotron beamlines. Since the first IC imaging at a synchrotron by absorption contrast in 1999,[1] the technique has been further refined. The complementary technique of x-ray ptychography further exploits both the high x-ray intensity available at synchrotrons and the coherent nature of their radiation. Valuable as they are, synchrotrons are scarce resources that cannot be moved into industrial or other typical research settings. There remains a strong interest in the development of laboratory instruments that can also tomographically analyze semiconductor devices. Some promising work applies a focused-ion beam to delaminate layers of a microelectronic device, alternating with acquisition of 2D surface images from a scanning-electron microscope (SEM).[2] The accurate synthesis of many thousands of such images into the full 3D reconstruction is a challenging mathematical problem, however. The process also destroys the IC, preventing follow-up measurements by spectroscopic or other analytic techniques. We have lately been exploring whether nondestructive tomography of ICs can be accomplished by harnessing the extremely high energy-resolving power of superconducting x-ray detectors. This power can dramatically reduce unwanted x-ray backgrounds to achieve the most efficient possible use of the limited supply of x-ray photons. It also opens the door for discrimination among multiple chemical elements in a sample. In our approach (Fig. 1), a small x-ray spot is achieved by focusing a SEM beam onto a platinum film only 100 nm thick. The sub-micrometer features of an IC are magnified onto an imaging x-ray spectrometer by generating the x-rays in the thin-film target mere micrometers from the layers of interest in the IC. By placing the electron-tophoton converter so close to the attenuating sample, we can perform the measurement in a compact instrument only a few meters across. It is the size not of a synchrotron facility, but of a typical laboratory instrument based on a vacuum chamber. Discrimination of the signal x-rays from a large background is aided by good energy resolution. This is where superconducting sensors enter the picture. TRANSITION-EDGE SENSORS: SUPERCONDUCTING X-RAY MICROCALORIMETERS Microcalorimeters are energy-resolving sensors that detect x-ray photons, or in fact, photons all the way from the optical band to gamma-ray energies. Like any calorimeter detector, they convert photon energy to heat, then take advantage of some temperature-dependent property to measure that thermal energy electrically. In order to make precision measurements of each x-ray photon’s energy, the sensors generally must be made small and kept cryogenically cold. Many designs harness the unusual properties of cryogenic materials: the metallic magnetic calorimeter uses the temperature-dependent magnetization of a metallic paramagnet, while the superconducting tunnel junction uses the quantum tunneling of electrons freed by the absorbed energy. One of the most fully developed and widely used types of microcalorimeter is the superconducting transitionedge sensor (TES),[4] such as those shown in Fig. 2. In a TES, the temperature is sensed by a material held in the transition between its superconducting and normal states. At temperatures in the narrow range of this superconducting phase transition, the material’s electrical resistivity is extremely sensitive to tiny temperature changes. The desired temperature can be reached by placing a constant electrical potential (typically, a few microvolts) across a Fig. 2 Left: A TES array “snout.” The sensors (top), the multiplexing SQUID amplifiers (4 of 8 are shown below the flexible wiring runs), and other support electronics are cryogenically cooled in this assembly. Right: An array approximately 9 mm in diameter containing 240 TES detectors (bottom), similar to the array used in the first version of Tomcat. Courtesy of Daniel Schmidt.

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 6 TES. Negative electrical-thermal feedback then balances the device at the desired point on the superconducting transition. A heating event such as the absorption of an x-ray photon causes a brief reduction in the electrical current through the device, lasting typically a few milliseconds or less. Figure 3 shows the operating concept of a TES. Pulses in the TES current are the signatures of a photon detection; the amplitude of the pulse indicates the photon’s energy. TESs with energy-resolving powers of 2000 and higher have been demonstrated at x-ray energies from near 1 keV[5] up to at least 12 keV,[6] as well as for gamma rays in the range of 100 keV[7] and even for alpha particles with energies of several MeV.[8] (An energy resolving power of 2000 implies a relative uncertainty of 0.05 %.) Once fabricated, a TES operates only at a specific temperature, that of its superconducting phase transition. Thanks to the properties of metallic thin films, it is possible to select the transition temperature of materials that combine a superconductor with a normal metal. Thin-film bilayers of molybdenum with either gold or copper can be produced having a range of transition temperatures up to 1K (the transition temperature of bulk Mo), with the temperature engineered to match the available cryogenic cooling system. A layer of high-Z material such as gold or bismuth is often placed near (and thermally connected to) the TES to improve the calorimeter’s efficiency for absorption of x-ray photons. Thermal fluctuations are the dominant source of noise in a well-designed TES, so the energy resolution is optimized by designing them for the coldest achievable operating temperature. We typically work with detectors in the range of 20 to 140 mK, temperatures accessible to some commercial refrigerators that offer automated operation without using liquid cryogens. The fundamental advantage of a microcalorimeter such as the TES over other technologies for x-ray detection is its extremely high energy resolution. The resolution is as good as that of any but the best diffraction-based wavelength-dispersive spectrometers. In addition, the ability to measure the entire spectral band at once means that measurement throughput exceeds that of most highresolution analyzers. TESs must be made small, however, to minimize their heat capacity and maximize their resolving power. To accelerate measurements with TESs, we employ arrays of hundreds of sensors (Fig. 2, right). Considerations of both thermal loading and system complexity mean we cannot realisti- cally connect thousands of wires from room temperature directly to a sensor array at subKelvin temperatures. Instead, multiplexing readout systems must be used. Many TES applications have used time-division multiplexing: an amplifier chain based on superconducting quantum interference devices (SQUIDs),[9] Fig. 2, left, cycles through dozens of sensors in a time-shared manner. In a more recent development, a frequency-domain multiplexing system built upon microwave SQUIDs delivers even higher multiplexing factors and/or wider readout bandwidth for each detector. Such systems have been demonstrated to multiplex 250 TESs onto a single radio-frequency transmission line.[10] Arrays of TES microcalori- meters have been developed Fig. 3 The TES detection concept. Top left: Photons are absorbed on a microcalorimeter, a small island of low heat capacity containing the TES thermometer. Top right: In the transition, the TES resistance is a very steep function of its temperature. Bottom left: Absorption of a photon causes a transient pulse in the temperature and a corresponding reduction in the bias current. Bottom right: This dip in current is the signal; the size of the pulse indicates the x-ray energy.

edfas.org 7 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 for many analytic purposes. Several have been installed at synchrotron beamlines for diverse energy bands and scientific goals.[11] They are also used for analysis of radioactive materials in the nuclear fuel cycle[12] and to study the decay-energy spectrum of isotopes such as 163Ho that are sensitive to the mass of the electron neutrino. TES spectrometers are planned for multiple orbiting x-ray telescopes and will analyze samples returned from an asteroid. TES arrays are also excellent tools for the analysis of more usual, earth-bound materials. In one recent example, we used the emission spectra of several lanthanide-series elements to improve our understanding of certain x-ray fundamental parameters, including the energies and line shapes[13] and relative intensities[14] of the elements’ L-series emission lines (Fig. 4). While TESs do not directly register photon energies, energy calibration to one part in 104 is possible if enough reference materials—such as the 3d transition metals—are measured along with the unknown samples. The sensitivity and high resolution of the TES also enables the discrimination of distinct chemical states, such as Ka and Kβ emission of titanium (Fig. 5).[15] X-RAY NANO-CT DEMONSTRATED WITH TES ARRAYS The energy resolution of a TES spectrometer can also serve as a powerful tool to distinguish signal photons from backgrounds. A new research instrument called the tomographic circuit analysis tool (Tomcat) is the first to use this property of TESs in an x-ray computed tomography (CT) measurement. It recently imaged a small region of an IC and resolved wiring features as small as 160 nm.[16] In future research, the spectrometer will also be able to analyze the elemental composition of the sample, taking advantage of the element-dependent nature of x-ray transmission as a function of energy. The Tomcat instrument uses the tiny electron beam of a SEM to generate a concentrated source of x-rays for a measurement of x-ray transmission through the IC sample (Fig. 1). The SEM focuses electrons to a spot approximately 100 nm in diameter while accelerating them with a 25 kV potential. A thin-film target, 100 nm of platinum, converts many of the electrons to x-ray photons. The conversion happens in two ways: through broadband Fig. 4 Emission spectra of two lanthanide-series metals, a demonstration of the energy resolution and the wide range of energies and intensities the TES can measure.[13] For clarity, the holmium spectrum is scaled up by a factor of 1000. Fig. 5 TES-measured Kɑ and Kβ emission spectra of titanium in various oxidation states.

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 8 bremsstrahlung radiation and through inelastic scattering that excites atomic electrons, which then relax by emitting characteristic fluorescence radiation. Any x-rays emitted in the forward direction pass through a narrow spacer layer of silicon and then through the IC of interest, to be imaged by the TES array. Two-dimensional images over larger areas are generated by moving the sample laterally relative to the source and detector array, while the third (depth) dimension is explored by rotating the sample so that x-rays cross it at a variety of angles. The Tomcat CT instrument requires specially prepared—but reusable—IC samples. For the first demonstration, we removed the circuit’s largest, back-end-of-line wiring layers by spin-milling the IC in a plasma focused ion beam. Tomcat would still work with the larger wiring layers intact, but their presence would have slowed down imaging of the smallest features that were of primary interest in the initial demonstration. After the thinning step, three wiring and three dielectric layers of silicon dioxide remained in the sample circuit. A carbon wafer transparent to x-rays was then epoxied to the sample to stiffen it for the remaining preparation work. One critical design challenge that Tomcat faces is that of achieving high resolution in a compact, laboratory-sized system. The smallest features of interest, only 160 nm wide, must be magnified onto the surface of the imaging detector so that they are larger than the typical spacing of the pixels, which is 500 μm in the TES array. The magnification requirement could have been met by placing the detectors very far from the sample, but this choice would strain the limits of a “compact” system and also reduce the all-important photon yield. To achieve a lab-scale and high-efficiency instrument, we instead chose to locate the conversion target very close to the sample. IC wafers are far too thick to meet this requirement without further processing— the platinum thin-film electron-conversion target must be placed within micrometers of the transistor layers, in the middle of the wafer itself. Thus, the next step in preparation was to thin the carrier wafer. The great majority of the wafer was removed by lapping and polishing, leaving a spacer layer of silicon only 8.5 μm thick behind the transistor region of the circuit. The 100 nm film of platinum, the conversion target, was then deposited on the remaining spacer. This step fixed the system geometry with a high and, importantly, a constant optical magnification. The spacer thickness was chosen because the imaging array could be no closer than 75 mm from the sample, and because of the minimum feature sizes of this specific IC sample. The choice of platinum as a target material involved several factors; most critically, platinum efficiently emits fluorescence lines at energies that maximize the x-ray absorption contrast between the copper wiring and dielectric in the IC. For CT measurements, the prepared sample is held in a complex stack of positioning instruments that enable 3D placement with 10 nm precision, as well as rotation about a vertical axis. The IC region of interest (ROI) is measured in a raster-scan pattern across a rectangular area, with discrete steps no more than a fraction of the roughly 1 μm viewable by the TES array at any one instant. The ROI is then rotated about the vertical axis and scanned again to access information about the third, depth, dimension. The two 3D-reconstruction algorithms used are based on Bayesian and maximum-entropy methods. We adopt a Bayesian prior that penalizes absolute gradients in the reconstructed image, favoring smoother reconstructions. Maximum-entropy methods are well suited to a problem where we have a set of measurements covering only a limited range of angles. In contrast, filtered backprojection, which reigned in medical tomography for four decades,[17] requires that the data be collected in a complete and regular array of angles, then Fourier transformed. Relying on fast Fourier transforms, filtered backprojection is indeed very fast. The speeds of algorithms are becoming less of a concern, however, and the focus today is on obtaining the best reconstructed images for any given data collection. Fig. 6 Comparison of reconstructions (top) and the design file (bottom). The left images show a single slice; the right images show a 3D view. The finest lines are 160 nm wide and the scale bar is 2 µm. Figure reprinted from Ref. 16 with permission.

edfas.org 9 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 We have found it essential for the reconstruction to model the x-ray source accurately. During the research, we improved the CT spatial resolution by a factor of two simply by starting from a better estimate of the mean transmission through the sample. Detailed radiationtransport modeling of the 3D shape of the x-ray source (and its changes as the source was tilted through various angles) improved the resolution by another factor of two. Both analytical steps were needed to produce the highquality reconstruction results we obtained with a mere 100 photons detected per voxel. In addition to the shape of the source, the x-ray energy spectrum is also incorporated into the analysis. Again, we use radiation-transport modeling to learn the form of the spectrum. Our TES detectors report the energy of each transmitted photon, allowing our reconstruction algorithm to weight each photon according to its measured energy. This weighting eliminates “beam-hardening artifacts,” a class of bias often present when imaging with energy-integrating detectors. Maximum-likelihood and Bayesian methods have an additional advantage: users can obtain confidence intervals for the reconstruction. It is possible in principle to give a probability that two adjacent pieces of metal are connected or not, or a confidence interval for the width of a given metal line. The first full IC measurement with Tomcat was made in early 2022 over the course of 300 hours. X-ray transmission scanning at a single angle was repeated for twelve to sixteen hours in a day, then the sample was rotated about the vertical axis to a new angle for the next day, until the sample was imaged many times each over a full range of ±45° in 12 steps. This sample IC was designed and fabricated by colleagues who shared the GDS circuit-design files. The data were used to reconstruct some 100 μm2 of the sample IC, an IC known to have wiring features as narrow as 160 nm. Comparison of the reconstructed data to the GDS file shows that Tomcat successfully identified and resolved all features in field of view (Fig. 6). We are not aware of any similar, published, structure-by-structure verification of a laboratory CT reconstruction based on the circuit’s design file. FUTURE DIRECTIONS We are now pursuing improvements to the existing Tomcat instrument. The first is to install TES arrays with smaller supporting electronics, which allows us to operate many more cryogenic detectors. This change will improve the system’s imaging speed, which has been limited so far by the small x-ray detection area. Figure 7 shows one of the “microsnouts” now being installed in Tomcat. These small structures will allow us to pack 3000 TESs in 12 arrays close to the IC sample.[18] Coupled with higher-transmission x-ray vacuum windows, the new design should measure x-ray photons at twenty times the rate possible with the single array of 240 TESs used in the first design (Fig. 2). Tomcat uses the energy resolution of the TES to distinguish signal photons from background. Many electrons from the SEM beam penetrate the platinum target layer, and the bremsstrahlung they create beyond the target could come from practically anywhere, smearing the images badly. The Pt fluorescence photons, on the other hand, are guaranteed to be emitted in the target and thus confined to a 100 nm spot. This was the initial reason to use high-resolution spectroscopy in the instrument. With our experience and modeling to date, and with the platinum target film roughly 100 nm thick, we have found that even the bremsstrahlung can be mostly confined to a small region—small enough to be a useful signal for tomography at the desired spatial resolution. In this limit, the energy-resolving power of a TES is less critical. Recent Fig. 7 The “microsnout” now replacing its much larger predecessor.

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 10 developments in pixelated, hybrid photon-counting x-ray cameras enable an interesting possibility. We will replace the TES detectors with a commercial detector of this type. This camera is in some ways the opposite of the cryogenic microcalorimeter camera: it gives up the energy resolution of the TESs in favor of large collecting area. It should increase the overall photon-counting rate by a factor of 1000 for signal photons, though the exact penalty that comes with the higher background rate remains difficult to assess through modeling. We expect to find that the optimal approach would blend the virtues of these two contrasting detection techniques. The best instrument might combine them; it might employ both a large-area camera for fast imaging and a camera based on TESs with excellent energy resolution that could exclude the more diffuse bremsstrahlung emission, allowing accurate measurement of the smallest features in a sample. Another reason to use the energyresolving TES camera along with a large-area camera is to enable pan-chromatic imaging. This possibility would mean using the energy spectrum of transmitted x-rays to determine the elemental composition of wiring and other features inside an IC sample. The TES spectrometer would disentangle elements by their characteristic absorption edges, identifying multiple materials and determining their distinctive distributions across an IC. Peering inside the complex 3D structures that make up a modern IC with x-ray tomography is not easy. It requires exquisite control over positioning, intense and tiny x-ray sources, and efficient detection of transmitted photons over large areas. Making the measurements in a laboratory rather than a synchrotron only amplifies the challenges. Superconducting detectors can play an important role with their ability to identify fluorescence emission and to distinguish multiple materials in a sample according to their x-ray transmission. We expect the continued refinement of sources, detectors, and positioning equipment to bring practical tomographic imaging of nanoscale structures to the laboratory in the very near future. ACKNOWLEDGMENTS Tomcat is the result of many years of effort by numerous scientists and engineers who designed, operated, and redesigned many aspects of the Tomcat tool. We are grateful to all the members of the Tomcat team at NIST Boulder Labs and Sandia National Laboratory, i.e., the authors of Ref. 19, as well as an earlier generation of collaborators at BAE Systems.[3] We thank Tom Gurrieri for the generous contribution of a sample IC and detailed information about its design. We also thank Kelsey Morgan for helpful discussions, Nate Ortiz for instrument drawings, and Dan Schmidt for the instrument photographs. The information, data, or work presented herein was funded in part by the Office of the Director of National Intelligence (ODNI), Intelligence Advanced Research Projects Activity (IARPA), via agreements D2019-1908080004, D2019-1906200003, D2021-2106170004, and FA8702-15-D-0001. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the ODNI, IARPA, or the U.S. Government. REFERENCES 1. Z.H. Levine, et al.: “Tomographic Reconstruction of an Integrated Circuit Interconnect,” Appl. Phys. Lett., 1999, 74(1), p. 150-152, doi. org/10.1063/1.123135. 2. D. Zhang, et al.: “Fast, Full Chip Image Stitching of Nanoscale Integrated Circuits,” Technical report, 2019, SRI International, Princeton. 3. P.B. Weichman and E.M. Lavely: “Fluorescent X-ray Scan Image Quality Prediction,” J. Hardw. Syst. Secur., 2020, 4, p. 13-23, doi. org/10.1007/s41635-019-00084-8. 4. K.M. Morgan: “Hot Science with Cool Sensors,” Physics Today, 2018, 71, p. 28-34, doi.org/10.1063/PT.3.3995. 5. S.J. Lee, et al.: “Fine Pitch Transition-edge Sensor X-ray Microcalorimeters with Sub-eV Energy Resolution at 1.5 keV,” Appl. Phys. Letters, 2015, 107, p. 223503, doi.org/10.1063/1.4936793. 6. S.J. Smith, et al.: “Performance of a Broad-Band, High-Resolution, Transition-Edge Sensor Spectrometer for X-ray Astrophysics,” IEEE Trans. Appl. Supercond., 2021, 31, p. 1-6, doi.org/10.1109/ TASC.2021.3061918. 7. D. Bennett, et al.: “A High Resolution Gamma-ray Spectrometer Based on Superconducting Microcalorimeters,” Review of Scientific Instruments, 2012, 83, p. 093113, doi.org/10.1063/1.4754630. 8. A.S. Hoover, et al.: “Measurement of the 240Pu/239Pu Mass Ratio Using a Transition-Edge-Sensor Microcalorimeter for Total Decay Energy Spectroscopy,” Analytical Chem., 2015, doi.org/10.1021/acs. analchem.5b00195. 9. W. Doriese, et al.: “A Practical Superconducting-microcalorimeter X-ray Spectrometer for Beamline and Laboratory Science,” Review of Scientific Instruments, 2017, 88, p. 053108, doi.org/ 10.1063/1.4983316. 10. J.A.B. Mates, et al.: “Simultaneous Readout of 128 X-ray and Gammaray Transition-edge Microcalorimeters using Microwave SQUID Multiplexing,” Appl. Phys. Letters, 2017, 111, p. 062601, doi.org/ 10.1063/1.4986222. 11. J.N. Ullom, et al.: “Transition-Edge Sensor Microcalorimeters for X-ray Beamline Science,” Synchrotron Radiation News, 2014, 27, p. 24-27, doi.org/10.1080/08940886.2014.930806. 12. J.N. Ullom and D.A. Bennett: “Review of Superconducting Transition-edge Sensors for X-ray and Gamma-ray Spectroscopy,” Superconductor Science and Technology, 2015, 28, p. 84003, doi.org/ 10.1088/0953-2048/28/8/084003. 13. J.W. Fowler, et al.: “Absolute Energies and Emission Line Shapes of the L X-ray Transitions of Lanthanide Metals,” Metrologia, 2021, 58, p. 015016, doi.org/10.1088/1681-7575/abd28a. 14. J.W. Fowler, et al.: “The Potential of Microcalorimeter X-ray Spectrometers for Measurement of Relative Fluorescence-line Intensities,” Radiation Physics and Chemistry, 2023, 202, p. 110487, doi.org/ 10.1016/j.radphyschem.2022.110487. 15. L. Miaja-Avila, et al.: “Valence-to-core X-ray Emission Spectroscopy of Titanium Compounds using Energy Dispersive Detectors,” X-Ray

edfas.org 11 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 Spectrometry, 2021, 50, p. 9-20, doi.org/10.1002/xrs.3183. 16. Z.H. Levine, et al.: “A Tabletop X-ray Tomography Instrument for Nanometer-scale Imaging: Reconstructions,” Microsystems & Nanoengineering, 2023, 9, p. 47, doi.org/10.1038/s41378-023-00510-6. 17. M.J. Willemink and P.B. Noël: “The Evolution of Image Reconstruction for CT—From Filtered Back Projection to Artificial Intelligence,” European Radiology, 2019, 29, p. 2185-2195, doi.org/10.1007/ s00330-018-5810-7. 18. P. Szypryt, et al.: “A Tabletop X-ray Tomography Instrument for Nanometer-scale Imaging: Demonstration of the 1000-element Transition-edge Sensor Subarray,” IEEE Trans. Appl. Supercond., 2023, 33, p. 2100705, doi.org/10.1109/TASC.2023.3256343. 19. N. Nakamura, et al.: “A Tabletop x-ray Tomography Instrument for Nanometer-scale Imaging: Integration of a Scanning Electron Microscope with a Transition-edge Sensor Spectrometer,” 2022, preprint arXiv:2212.10591, doi.org/10.48550/arXiv.2212.10591. ABOUT THE AUTHORS Joseph Fowler is a research scientist at the University of Colorado and the NIST Boulder Labs. He completed a B.A. in physics at Rice University in 1993 and his Ph.D. in physics at the University of Chicago in 2000. He was a researcher and later a physics professor at Princeton University from 2000 to 2010, when he moved to Colorado and NIST. His work at NIST focuses on the operation of TES arrays and analysis of their data to achieve optimal energy resolution, as well as the use of TESs for x-ray metrology research and computed tomography. Zachary Levine is a physicist at NIST in Gaithersburg, Maryland. He obtained undergraduate degrees in math and physics from MIT in 1976 and a Ph.D. in physics from the University of Pennsylvania in 1983. A computational physicist by training, he has been doing tomography on and off since about 1997 when he led a team which made the first tomographic reconstruction of an integrated circuit interconnect. Recently, he led a joint experimental-theoretical team which demonstrated an efficient and accurate treatment of diffraction in microtomography with application to an optical fiber. Paul Szypryt is a research associate and NASA Nancy Grace Roman Technology Fellow (RTF) at the University of Colorado and the NIST Boulder Labs. He completed his Ph.D. in physics at the University of California, Santa Barbara in 2017 and B.S. in applied and engineering physics at Cornell University in 2011. As a graduate student and NASA Space Technology Research Fellow (NSTRF), he helped to develop the first superconducting microwave kinetic inductance detector (MKID) instruments for optical astronomy with a focus on the direct imaging of exoplanets. Daniel Swetz is a research physicist and leader of the Quantum Calorimeters Group at NIST. He received his B.S. from the University of Wisconsin in 2003 and his Ph.D. in physics from the University of Pennsylvania in 2009. He joined NIST as a National Research Council Postdoctoral Fellow in 2010. His research at NIST focuses on developing superconducting sensors into arrays of x-ray detectors capable of measuring the energy of individual photons and particle decays in ways that are difficult or impossible with conventional detector techniques. He has authored over 100 publications, and has received several awards, including Department of Commerce Gold, Silver, and Bronze medals, and R&D100 award. Advertise in Electronic Device Failure Analysis magazine! For information about advertising in Electronic Device Failure Analysis: Kelly Johanns, Business Development Manager 440.671.3851, kelly.johanns@asminternational.org Current rate card may be viewed online at asminternational.org/mediakit.

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 12 LASER-BASED COPPER DEPOSITION FOR SEMICONDUCTOR DEBUG APPLICATIONS Michael DiBattista1, Scott Silverman1, and Matthew M. Mulholland2 1Varioscale Inc., San Marcos, California 2Intel Corp., Santa Clara, California miked@varioscale.com EDFAAO (2023) 4:12-16 1537-0755/$19.00 ©ASM International® INTRODUCTION The rapid development of advanced integrated circuits (IC) with increased performance and expanded features now requires more than shrinking the transistor geometry during fabrication. Designs for 2.5D[1] and 3D[2,3] packaging, highly integrated chiplets,[4,5] backside power delivery,[6] and other technologies have emerged. Figure 1 shows an example of new products with complexity that extends beyond monolithic IC fabrication and package assembly. Debug technologies that enable the rapid diagnosis and modification of high-performance circuits in package play a key role to the product introduction success. Focused ion beam (FIB) systems have traditionally had a large role in the study and prototyping of integrated circuits during the debug stage.[8] Circuit edit FIB tools can edit the design in silicon by establishing new signal paths with ion beam-based deposition and cut signal traces by employing well established techniques.[9] Measuring electrical characteristics such as current, voltage, and timing by direct access pico probing can play a critical role in design debug.[10,11] Accessing the targeted signals with a FIB access point is a key capability of these instruments, but depositing material for multiple probe pads with FIB can be time consuming due to deposition rates and cleanup activities. The continuous precursor exposure over long time periods can result in the degradation of key components of the tool, and there are also significant material property limitations of the FIB deposited tungsten (W) or platinum (Pt).[12] The carbon contamination from the selected metal organic precursors results in a significant increase in electrical resistance. This often results in material with a resistance 10 to 100 times greater than the bulk conductivity value of 5.6 µohm-cm for W and 10.6 µohm-cm for Pt. BACKGROUND Copper is an attractive candidate for rerouting signals on integrated circuits and packages due to its low electrical resistivity (1.67 µohm-cm), high thermal conductivity, and compatibility with the existing copper backend interconnect and package routing layers. With the development and introduction of copper(hfac)tetramethylvinylsilane, (Cu(hfac)TMVS), commercially available as CupraSelect, copper deposition with consistently low resistivity can be achieved at temperatures approaching 120°C.[13] Using a pyrolytic laser chemical deposition (LCVD) process, the laser energy acts as a localized heat source when focused on the sample and drives the chemical reaction of the adsorbed precursor.[14] The precursor reaction is composed of a copper (II) hexafluoroacetylacetonate (CuHfac) molecule bonded to a tetramethylvinylsilane (TMVS). This structure provides the copper component with an improved vapor pressure over earlier copper precursors such as: copper (II) acetylacetonate (Cu(acac)2), copper hexafluoracetylacetonate (Cu(hfac)2), or 1,5 cyclooctadience copper (I) hexafluoraceylacetonate (COD-Cu-hfac).[15-17] Fig. 1 2.5D vs. 3D IC designs.[7]

edfas.org 13 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 The CupraSelect chemical reaction is a disproportionation reaction as shown in Fig. 2. This reaction mechanism is well supported in the literature and through experimental results.[18-20] The main component of the reaction involves a rearrangement of the precursor when two precursor molecules are adsorbed on the surface. With the addition of heat, the attached CuHfac molecule slides over the adjacent adsorbed precursor molecule leaving a copper atom on the target surface. The resulting conductivity measurements support pure copper deposition with the lack of silicon and carbon incorporation into the deposited material support. Interestingly, the reaction has been found to be completely reversible. To date FIB based copper deposition has been accomplished with improved resistivity values, but with limited success in achieving bulk resistivity.[21-23] Laser-based systems have also been used in conjunction with FIBs to deposit material that lowers the resistance of the FIB connections deposited on the IC.[24,25] The challenges to FIB implementation include the limited precursor volatility (vapor pressure), which limits the amount of copper precursor that can be delivered to the target surface location and the room temperature deposition ion beam process. Remes has demonstrated this by improving the resistance of FIB deposited lines.[25] This work included deposition for expanding the debug capabilities for integrated circuits. Copper deposition can be performed with an existing VarioEdit Platform configured with a selection of optical objectives 5x, 10x, and 20x. These different objectives focus the laser spot on the sample surface and control the lateral dimensions of deposition by confining the thermal spreading of the heat. The system can be equipped with three lasers: a 532 nm continuous wave (CW) laser, a pulsed 532 nm laser, and 355 nm UV pulsed laser. These lasers enable a range of laser power delivery typically from 30 milliwatt (mW) for glass and ~1 W for thermally conductive sample surfaces such as silicon and metals. The laser system is equipped with a dedicated vacuum system capable of reaching a base pressure of 50 millitorr (mTorr). In general, a predominant component molecule in vacuum chambers at the base pressure is residual water. Previous work has indicated that the water can have beneficial effects on surface adhesion of the copper film, removing requirements to thermally bake out the chamber before deposition work is attempted. Pulsed valves for chemical delivery in the vacuum chamber take advantage of the use of excess TMVS to increase vapor pressure, stabilize the precursor, and prevent clogging of valves.[26] ANALYSIS OF LASER-BASED COPPER DEPOSITION The morphology, growth rate, and chemical composition of laser deposited thin films have been widely studied.[14] The deposited material is composed of 50 to 100 nm grain sizes and has been widely reported in the published literature.[14,18-19] The majority of previous copper laser chemical vapor deposition (LCVD) based results used carrier gases such as hydrogen,[14,19,25] helium,[25] or nitrogen[19] with a bubbler to entrain the liquid precursor and deliver it to the work chamber with flow rates of 1-2 standard cubic centimeter (sccm). Carrier gas methodologies are not efficient in their use of the precursor gas since the reaction area is minimal in comparison to the chamber and sample geometries. Hydrogen carrier gas can also present safety concerns, and helium can be difficult to resource. CupraSelect in an excess of TMVS and a direct delivery provides an efficient chemical delivery mechanism while reducing practical and reliability challenges such as clogged valves and precursor storage. Figure 3 shows the deposition of a 20 µm x 20 µm copper pad with a 20x objective using a 532 nm laser set to 0.3 W power. The initial deposition is shown in Fig. 3a, the copper films start thin and optically bright and shiny with direct illumination but become dark due to light scattering as the thickness increases due to additional Fig. 2 The CupraSelect chemical reaction is a disproportionation reaction.

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 14 surface structure and film roughness.[13,14] Figure 3b demonstrates that with additional laser passes over the same pad, the film thickness increases and results in additional surface structure that causes the film to appear dark under direct illumination. When deposited on transparent material, such as glass, the copper films appear bright when observed through the glass due to the strong adhesion on the glass substrate. A FIB cross-section examination of the copper deposition morphology is shown in Fig. 4. This copper deposition was grown on a 250 nm silicon oxide (SiO2)/p-doped silicon substrate by spotting the CW laser for 60 seconds at 600 mW. This image has a bright FIB platinum protective capping layer on top of the copper deposition, supported by the silicon oxide and silicon substrate. This copper deposition shows there are two distinct growth regions: a center region with well-defined copper grain structure and two outside edge boundary regions with a porous structure. The copper grain structure in the center corresponds to the location of the direct laser energy exposure, the growth on the boundaries is due to the substrate heat transfer at the surface. The deposition rate under these conditions can be approximated to be greater than 25 µm3/s. The deposition shows the roughness of the copper is similar to previously published literature results. This has been widely attributed to the introduction or presence of water in the deposition chamber. In addition to surface roughness, the presence of water also is reported (a) Fig. 3 A scanning electron microscope (SEM) image showing the complex grain structure and surface roughness of the copper line. Fig. 4 A FIB cross section and SEM image showing the complex grain structure and surface roughness of the copper deposition. Fig. 5 An SEM image showing the interface of the copper deposition on silicon oxide thin film. (b)

edfas.org 15 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 to have excellent adhesion to the surface. Figure 5 shows the copper and SiO2 interface from the cross section of the 60 second dwell time deposition. There appears to be conformal adhesion to the surface of the SiO2 layer. A significant amount of IC backside circuit edit work is performed with ultrathin silicon thicknesses on the order of 3-10 µm from the active circuitry.[8-10] Solutions for IC debug and diagnostics can include thermally confined depositions on a thin layer of SiO2 on top of ~3 µm of silicon above the circuit layer. The low temperature deposition reaction allows for implementation on ultra thinned backside silicon. In the case shown in Fig. 6, the 20x objective and the 532 nm laser set to a power of 100 mW and 0.5 µm/sec scan speed was used to deposit the two 20 µm x 20 µm pads. Two signals have been exposed and connected through the node access holes and brought out to the surface using FIB tungsten and connected to the adjacent LCVD pads. These pads and connections are used to physically probe the signals for interpretation and measurements during test. The surface roughness of the copper deposition is also an enabler for landing the probe pads. This work duplicates previously published work in 2006 from Janne Remes’ Ph.D. thesis[25] with the combination of FIB and laser assisted copper deposition, and demonstrates the electrical testing viability of this solution. These techniques take advantage of the low resistivity, high volumetric deposition rate, and improved timing constant for sensitive performance measurements of circuit modifications as shown in Table 1. CONCLUSIONS The introduction of complex IC packaging with 2.5D and full 3D integration requires the development of new techniques to assist with rapid diagnosis and modification of high-performance circuits. Laser-assisted copper deposition using CupraSelect with a disproportionate reaction at low temperature provides an efficient method to deposit new copper traces for IC design debug applications. The use of excess TMVS to deliver CupraSelect overcomes delivery challenges and the presence of a high background of water pressure also assists with the adhesion to the surface. Laser-based copper deposition techniques have been shown to be useful in combination with traditional FIB techniques with improved resistivity, deposition rate, and timing improvements. REFERENCES 1. D. Nuez, et al.: “Failure Localization Techniques for 7 nm and 16 nm Process Nodes in Monolithic and 2.5D SSIT Packages using OBIRCH, LVP, and Advanced Die Thinning,” Proc. Int. Symp. Test. Fail. Anal. (ISTFA), 2021, p. 73-79. 2. E. Beyne, et al.: “3D SoC Integration, Beyond 2.5D Chiplets,” 2021 IEEE International Electron Devices Meeting (IEDM), doi.org/10.1109/ IEDM19574.2021.9720614. 3. T. Li, et al.: “Chiplet Heterogeneous Integration Technology – Status and Challenges,” Electronics, 9, 2020, p. 670, doi.org/10.3390/electronics9040670. 4. J.H. Lau: “Chiplet Heterogeneous Integration,” Semiconductor Advanced Packaging, Springer, Singapore. 2021, p. 413-469, doi.org/10.1007/978-981-16-1376-0_9. 5. Y.P. Chiang, et al.: “InFO_oS (Integrated Fan-Out on Substrate) Technology for Advanced Chiplet Integration,” 2021 IEEE 71st Electronic Components and Technology Conference (ECTC), San Diego, CA, USA, 2021, p. 130-135, doi.org/10.1109/ECTC32696.2021.00033. 6. H. Lin, et al.: “Efficient Backside Power Delivery for HighPerformance Computing Systems,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 30(11), Nov. 2022, p. 1748-1756, doi.org/10.1109/TVLSI.2022.3183904. 7. D. Medhat: “2.5/3D IC Reliability Verification Has Come A Long Way,” Semiconductor Engineering, 2022. Table 1 Deposition Resistivity, Volumetric Deposition Rate, and RC Timing Constant between FIB W, Pt, and LCVD CupraSelect Key metric Deposition material Ga+ FIB tungsten Ga+ FIB platinum LCVDCupraSelect Resistivity (µohm-cm) 5.6 10.6 1.67 Volumetric deposition rate (µm3/min) 5 25 1500 100 µm connection RC timing constant (tau, ps) 12.13 9.05 8.63 Fig. 6 SEM image showing two signal node access holes with pico-probe pad structures made from LCVD, connected and strapped to FIB deposited tungsten (W).

RkJQdWJsaXNoZXIy MTMyMzg5NA==