Article
Chemistry, Multidisciplinary
Bin Chen, Xue-Peng Wang, Fangying Jiao, Long Ning, Jiaen Huang, Jiatao Xie, Shengbai Zhang, Xian-Bin Li, Feng Rao
Summary: This study shows that by simplifying the composition and miniaturizing the dimensions of traditional GeSbTe-like phase-change materials (PCMs), resistance drift in PCRAM devices can be effectively suppressed. It is demonstrated that a thin Sb film, with an optimal thickness of only 4 nm, enables precise multilevel programming with ultralow resistance drift coefficients. This advancement is achieved through the slightly changed Peierls distortion in Sb and the less-distorted atomic configurations across the Sb/SiO2 interfaces.
Article
Multidisciplinary Sciences
Asir Intisar Khan, Alwin Daus, Raisul Islam, Kathryn M. Neilson, Hye Ryoung Lee, H-S Philip Wong, Eric Pop
Summary: This study demonstrates a reduced switching current density and multi-level operation in flexible superlattice PCM, with excellent performance even after repeated bending and cycling, paving the way for low-power memory in flexible electronics. It also provides key insights for PCM optimization on conventional silicon substrates.
Article
Chemistry, Physical
Xiaotian Zeng, Xiaoqin Zhu, Yifeng Hu
Summary: We report a [C(5 nm)/Sb2Te3(12 nm)]5 phase-change heterostructure film with multilevel phase change properties and low resistance drift. The addition of carbon broke the Sb2Te3 bond and formed new C-Sb and C-Te bonds. The reliability of the phase-change heterostructure was confirmed using transmission electron microscopy. The amorphous resistance drift index of the C/Sb2Te3 film was effectively reduced, possibly due to structure relaxation. PCM devices based on C/Sb2Te3 phase-change heterostructure films also demonstrated the potential for multilevel storage. This research provides a new method for achieving multilevel phase-change memories with low resistance drift.
JOURNAL OF ALLOYS AND COMPOUNDS
(2023)
Article
Chemistry, Multidisciplinary
Benedikt Kersting, Syed Ghazi Sarwat, Manuel Le Gallo, Kevin Brew, Sebastian Walfort, Nicole Saulnier, Martin Salinga, Abu Sebastian
Summary: Chalcogenide phase change materials are utilized for non-volatile, low-latency storage-class memory and new forms of computing, but face challenges with temporal drift in electrical resistance. Research shows that the efficacy of observation is influenced by the observable timescale, and experimental measurements of drift onset can be conducted using threshold-switching voltage. This additional feature of structural relaxation dynamics serves as a new benchmark for evaluating classical models explaining drift.
ADVANCED FUNCTIONAL MATERIALS
(2021)
Article
Engineering, Electrical & Electronic
Soobin Hwang, Jin-Su Oh, Taek Sun Jung, Dasol Kim, Hyeonwook Lim, Changwoo Lee, Cheol-Woong Yang, Jae Hoon Kim, Mann-Ho Cho
Summary: This study reports a significant increase in the thermal stability of C-incorporated Sb2Te3, making it a suitable material for PCM devices with high thermal stability. The thermal stability can be further enhanced by adjusting the C content, although some of the device operation characteristics are slightly degraded.
ACS APPLIED ELECTRONIC MATERIALS
(2021)
Article
Chemistry, Analytical
Yi Lv, Qian Wang, Houpeng Chen, Chenchen Xie, Shenglan Ni, Xi Li, Zhitang Song
Summary: By designing a novel 2T2R structure circuit and a compressing encoding scheme, it is possible to increase the storage density of phase change memory and improve data reliability.
Article
Automation & Control Systems
Shao-Xiang Go, Qiang Wang, Kejie Huang, Tae Hoon Lee, Natasa Bajalovic, Desmond K. Loke
Summary: Modern innovations rely on computers, and in-memory operation offers a promising solution to reduce energy consumption in data centers. However, designing suitable material platforms for in-memory operation is challenging. This study demonstrates a combined M state-based model framework using phase-change materials, which enables nonvolatile and reprogrammable shift register operations.
ADVANCED INTELLIGENT SYSTEMS
(2023)
Article
Materials Science, Multidisciplinary
Ruben Jeronimo Freitas, Koichi Shimakawa
Summary: Resistance drift (R-drift) in phase-change random-access memory (PCRAM) is a phenomenon where the electrical resistance of the amorphous phase gradually increases over time after switching from the crystalline state. While various interpretations have been proposed, it is believed to be an intrinsic phenomenon in nonequilibrium systems, where mechanical stress induced in the amorphous state relaxes into a more stable nonequilibrium state, rather than being solely attributed to electronic processes.
PHYSICA STATUS SOLIDI-RAPID RESEARCH LETTERS
(2022)
Article
Physics, Applied
Y. Vorobyov, A. Ermachikhin, A. Yakubov, E. Trusov, M. Fedyanina, P. Lazarenko, S. Kozyukhin
Summary: The non-linear temperature dependence of the Fermi level and extended state conduction by free holes are responsible for the non-Arrhenius behavior of Ge2Sb2Te5 conductivity. By assuming a parabolic function for the temperature-dependent Fermi level, two out of three model parameters can be determined, with the third parameter requiring an assumption about the value of the prefactor for conductivity. This analysis scheme yields reasonable values for the Fermi level position in Ge2Sb2Te5 and demonstrates the change of Fermi level due to resistance drift phenomenon.
JOURNAL OF PHYSICS D-APPLIED PHYSICS
(2021)
Article
Chemistry, Multidisciplinary
Xing Yang, Liangjun Lu, Yu Li, Yue Wu, Ziquan Li, Jianping Chen, Linjie Zhou
Summary: Integrated Mach-Zehnder interferometers (MZIs) with phase-change materials offer low power consumption and compact size for reconfigurable photonic processors. However, they suffer from low optical extinction ratio and limited switching cycles due to material loss and poor reversible repeatability. A non-volatile electrically reconfigurable MZI with a low-loss phase-change material (Sb2Se3) encapsulated in Al2O3 layers is demonstrated. By dividing the Sb2Se3 patch into small sub-cells to restrict material reflow, more than 10,000 reversible phase-change cycles and 6-bit multilevel switching states are achieved.
ADVANCED FUNCTIONAL MATERIALS
(2023)
Article
Chemistry, Multidisciplinary
Christian Stenz, Julian Pries, T. Wesley Surta, Michael W. Gaultois, Matthias Wuttig
Summary: Glasses undergo structural relaxation that affects their physical properties. The combination of TEM electron diffraction and RMC simulations provides information on atomic arrangement. The study reveals structural changes in GeTe, with an increase in bond angle and a decrease in tetrahedrally coordinated Ge atoms. This finding sheds light on the atomic processes involved in structural relaxation in GeTe and other PCMs.
Article
Engineering, Electrical & Electronic
Cheng Chen, Xi Li, Chenchen Xie, Houpeng Chen, Siqiu Xu, Zhitang Song
Summary: This article presents a read optimization method for PCM cells to improve their data reading reliability by studying the effect of different read voltages on the resistance stability. The average drift coefficient decreased significantly, promising higher data reliability.
IEEE TRANSACTIONS ON ELECTRON DEVICES
(2022)
Article
Chemistry, Multidisciplinary
Jiaen Huang, Bin Chen, Gang Sha, Huang Gong, Tao Song, Keyuan Ding, Feng Rao
Summary: Phase-change random access memory (PCRAM) is a promising technique for universal memory and neuromorphic computing. This study demonstrates thickness-independent conductance evolution in ScxSb2Te3 PCRAM material films, leading to an unprecedentedly low resistance-drift coefficient compared to Ge2Sb2Te5. The nanoscale chemical inhomogeneity and constrained Peierls distortion contribute to the ultralow resistance drift of ScxSb2Te3 films, making it an appropriate candidate for high-accuracy cache-type computing chips.
Article
Materials Science, Multidisciplinary
Timothy M. Philip, Kevin W. Brew, Ning Li, Andrew Simon, Zuoguang Liu, Injo Ok, Praneet Adusumilli, Iqbal Saraf, Richard Conti, Odunayo Ogundipe, Robert R. Robison, Nicole Saulnier, A. Sebastian, Vijay Narayanan
Summary: This study investigates the performance of phase-change memory cells, proposes a new type of PCM cell, and verifies it through theoretical and experimental approaches. The experimental results are consistent with the theoretical predictions.
Article
Multidisciplinary Sciences
M. Pourmand, P. K. Choudhury, Mohd Ambri Mohamed
Summary: The study investigates the optical response of metal-dielectric stacks-based cavity structures embedded with graphene microheaters for perfect absorption. By manipulating the refractive indices and crystallinity ratio of GST layers through external electrical pulses, >99% perfect absorption is achieved in the visible and infrared spectrum. The proposed configuration allows for independent transition states for each GST layer, extending absorption from visible to infrared and broadening/red-shifting perfect absorption peaks.
SCIENTIFIC REPORTS
(2021)
Article
Multidisciplinary Sciences
J. Feldmann, N. Youngblood, M. Karpov, H. Gehring, X. Li, M. Stappers, M. Le Gallo, X. Fu, A. Lukashchuk, A. S. Raja, J. Liu, C. D. Wright, A. Sebastian, T. J. Kippenberg, W. H. P. Pernice, H. Bhaskaran
Summary: With the advancement of technology, the demand for fast processing of large amounts of data is increasing, making highly parallelized, fast, and scalable hardware crucial. The integration of photonics can serve as the optical analogue of an application-specific integrated circuit, enabling photonic in-memory computing and efficient computational hardware.
Correction
Multidisciplinary Sciences
J. Feldmann, N. Youngblood, M. Karpov, H. Gehring, X. Li, M. Stappers, M. Le Gallo, X. Fu, A. Lukashchuk, A. S. Raja, J. Liu, C. D. Wright, A. Sebastian, T. J. Kippenberg, W. H. P. Pernice, H. Bhaskaran
Article
Engineering, Electrical & Electronic
Geethan Karunaratne, Manuel Le Gallo, Michael Hersche, Giovanni Cherubini, Luca Benini, Abu Sebastian, Abbas Rahimi
Summary: The emerging brain-inspired computing paradigm, hyperdimensional computing (HDC), offers a lightweight learning framework for various cognitive tasks compared to traditional deep learning methods. This study proposes an architecture for processing spatio-temporal (ST) signals within the HDC framework using in-memory compute arrays, achieving significant energy efficiency, area, and throughput gains while maintaining peak classification accuracy.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS
(2021)
Article
Multidisciplinary Sciences
Geethan Karunaratne, Manuel Schmuck, Manuel Le Gallo, Giovanni Cherubini, Luca Benini, Abu Sebastian, Abbas Rahimi
Summary: The paper proposes a novel architecture that utilizes computational memory units to perform analog in-memory computation on high-dimensional vectors, enhancing neural networks with explicit memory and achieving accuracy matching 32-bit software equivalent.
NATURE COMMUNICATIONS
(2021)
Article
Engineering, Electrical & Electronic
Riduan Khaddam-Aljameh, Michele Martemucci, Benedikt Kersting, Manuel Le Gallo, Robert L. Bruce, Matthew BrightSky, Abu Sebastian
Summary: By designing unit-cell arrays and implementing diagonal connections, we have successfully addressed challenges such as parallel writing and computational precision in memristive crossbar arrays.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS
(2021)
Article
Chemistry, Multidisciplinary
Samarth Aggarwal, Tara Milne, Nikolaos Farmakidis, Johannes Feldmann, Xuan Li, Yu Shu, Zengguang Cheng, Martin Salinga, Wolfram H. P. Pernice, Harish Bhaskaran
Summary: The use of nonlinear elements with memory in photonic computing has gained significant interest due to the rise of artificial intelligence and machine learning. Phase change materials are commonly used for demonstrating the feasibility of such computing, but they suffer from slow switching speeds and phase segregation issues. In this study, we demonstrate reversible, ultrafast switching using sub-5 nm antimony thin films on an integrated photonic platform, with a retention time of tens of seconds. By programming seven distinct memory levels using subpicosecond pulses, this research suggests the potential use of these elements in ultrafast nanophotonic applications.
Review
Multidisciplinary Sciences
Mario Lanza, Abu Sebastian, Wei D. Lu, Manuel Le Gallo, Meng-Fan Chang, Deji Akinwande, Francesco M. Puglisi, Husam N. Alshareef, Ming Liu, Juan B. Roldan
Summary: Memristive devices, which can change their resistance and memory state, have potential applications in various fields. However, there are still challenges to be addressed, including performance and reliability issues.
Article
Computer Science, Hardware & Architecture
Chuteng Zhou, Fernando Garcia Redondo, Julian Buchel, Irem Boybat, Xavier Timoneda Comas, S. R. Nandakumar, Shidhartha Das, Abu Sebastian, Manuel Le Gallo, Paul N. Whatmough
Summary: This article discusses the importance of high energy efficiency in always-on TinyML perception tasks in IoT applications and proposes the use of analog compute-in-memory (CiM) with nonvolatile memory (NVM) to achieve this goal. The authors introduce AnalogNets model architectures and a comprehensive training methodology to maintain accuracy in the presence of analog nonidealities and low-precision data converters. They also present AON-CiM, a programmable phase-change memory (PCM) analog CiM accelerator, designed to reduce the complexity and cost of interconnects. Evaluation results show promising accuracy and efficiency for KWS and VWW tasks.
Article
Multidisciplinary Sciences
Malte J. Rasch, Charles Mackin, Manuel Le Gallo, An Chen, Andrea Fasoli, Frederic Odermatt, Ning Li, S. R. Nandakumar, Pritish Narayanan, Hsinyu Tsai, Geoffrey W. Burr, Abu Sebastian, Vijay Narayanan
Summary: In this study, a hardware-aware retraining approach is developed to examine the accuracy of analog in-memory computing across multiple network topologies and investigate sensitivity and robustness to a broad set of nonidealities. By introducing a realistic crossbar model, significant improvement is achieved compared to earlier retraining approaches. The results show that many larger-scale deep neural networks can be successfully retrained to show iso-accuracy with the floating point implementation, and nonidealities that add noise to the inputs or outputs have the largest impact on accuracy.
NATURE COMMUNICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Manuel Le Gallo, Riduan Khaddam-Aljameh, Milos Stanisavljevic, Athanasios Vasilopoulos, Benedikt Kersting, Martino Dazzi, Geethan Karunaratne, Matthias Brandli, Abhairaj Singh, Silvia M. Mueller, Julian Buchel, Xavier Timoneda, Vinay Joshi, Malte J. Rasch, Urs Egger, Angelo Garofalo, Anastasios Petropoulos, Theodore Antonakopoulos, Kevin Brew, Samuel Choi, Injo Ok, Timothy Philip, Victor Chan, Claire Silvestre, Ishtiaq Ahsan, Nicole Saulnier, Vijay Narayanan, Pier Andrea Francese, Evangelos Eleftheriou, Abu Sebastian
Summary: A multicore analogue in-memory computing chip designed and fabricated in 14 nm complementary metal-oxide-semiconductor technology with backend-integrated phase-change memory is reported. It can be used to reduce the latency and energy consumption of deep neural network inference tasks by performing computations within memory. The chip features interconnection of 64 AIMC cores via an on-chip communication network and implements digital activation functions and additional processing involved in individual convolutional layers and long short-term memory units.
NATURE ELECTRONICS
(2023)
Article
Engineering, Electrical & Electronic
Dennis Christensen, Regina Dittmann, Bernabe Linares-Barranco, Abu Sebastian, Manuel Le Gallo, Andrea Redaelli, Stefan Slesazeck, Thomas Mikolajick, Sabina Spiga, Stephan Menzel, Ilia Valov, Gianluca Milano, Carlo Ricciardi, Shi-Jun Liang, Feng Miao, Mario Lanza, Tyler J. Quill, Scott T. Keene, Alberto Salleo, Julie Grollier, Danijela Markovic, Alice Mizrahi, Peng Yao, J. Joshua Yang, Giacomo Indiveri, John Paul Strachan, Suman Datta, Elisa Vianello, Alexandre Valentian, Johannes Feldmann, Xuan Li, Wolfram H. P. Pernice, Harish Bhaskaran, Steve Furber, Emre Neftci, Franz Scherr, Wolfgang Maass, Srikanth Ramaswamy, Jonathan Tapson, Priyadarshini Panda, Youngeun Kim, Gouhei Tanaka, Simon Thorpe, Chiara Bartolozzi, Thomas A. Cleland, Christoph Posch, Shihchii Liu, Gabriella Panuccio, Mufti Mahmud, Arnab Neelim Mazumder, Morteza Hosseini, Tinoosh Mohsenin, Elisa Donati, Silvia Tolu, Roberto Galeazzi, Martin Ejsing Christensen, Sune Holm, Daniele Ielmini, N. Pryds
Summary: This article introduces the characteristics and advantages of von Neumann architecture and neuromorphic computing systems. While traditional von Neumann architecture is powerful, it has high power consumption and cannot handle complex data. Neuromorphic computing systems, inspired by biological concepts, can achieve lower power consumption for storing and processing large amounts of digital information. The aim of this article is to provide perspectives on the current state and future challenges in the field of neuromorphic technology, and to provide a concise yet comprehensive introduction and future outlook for readers.
NEUROMORPHIC COMPUTING AND ENGINEERING
(2022)
Article
Engineering, Electrical & Electronic
Manuel Le Gallo, S. R. Nandakumar, Lazar Ciric, Irem Boybat, Riduan Khaddam-Aljameh, Charles Mackin, Abu Sebastian
Summary: In-memory computing is an efficient approach that utilizes the physical attributes of memory devices to perform computational tasks. However, the computational accuracy of this approach is currently limited due to inter-device variability, inhomogeneity, and randomness in analog memory devices. Bit slicing, a technique for constructing high precision processors, shows promise in overcoming this limitation. This study assesses the computational error in in-memory matrix-vector multiplications using bit slicing, and emphasizes the need to minimize analog matrix representation error through averaging within a given dynamic range.
NEUROMORPHIC COMPUTING AND ENGINEERING
(2022)
Proceedings Paper
Engineering, Multidisciplinary
R. L. Bruce, S. Ghazi Sarwat, I Boybat, C-W Cheng, W. Kim, S. R. Nandakumar, C. Mackin, T. Philip, Z. Liu, K. Brew, N. Gong, I Ok, P. Adusumilli, K. Spoon, S. Ambrogio, B. Kersting, T. Bohnstingl, M. Le Gallo, A. Simon, N. Li, I Saraf, J-P Han, L. Gignac, J. M. Papalia, T. Yamashita, N. Saulnier, G. W. Burr, H. Tsai, A. Sebastian, V Narayanan, M. BrightSky
Summary: Phase change memory (PCM) is being considered for non-von Neumann accelerators for deep neural networks based on in-memory computing. Conductance drift and noise are key challenges for reliable storage of synaptic weights in such accelerators. The integration of a projection liner into multilevel mushroom-type PCM devices demonstrates mitigation of conductance drift and noise, with further improvement shown by combining with a low-drift phase-change material. Large-scale experiments confirm lower drift and device-to-device drift variability for devices with projection liner, crucial for in-memory computing accelerators.
2021 IEEE INTERNATIONAL RELIABILITY PHYSICS SYMPOSIUM (IRPS)
(2021)
