Ultrafast in-memory computing and highly efficient deep neural networks driven by phase-change memory materials with partially amorphous state transitions

Abstract

Memristive materials (MMs) with adjustable conductance states are becoming the fundamental building blocks of artificial neural networks. However, challenges such as nonlinear weight conductance updates and the trade-off between increasing melting uniformity and reducing solid-to-liquid transition time have impeded the demonstration of in situ learning on a large-scale multiple-layer memristive network. Here, we modulate the dynamic interactions between MMs and multiple-pulse excitations in the melting-based voltage-pulse length diagram and temperature dependent mean-square displacement. We utilize all four melting states (FMSs) to create an integrated framework for attaining rapid in-memory computing and in situ deep neural network applications. We achieved near-ideal R² value uniformity and a partial melting time below random-access memory (RAM) device's switching times. Reversible switching using below-RAM switching time pulses was also demonstrated. The FMS network discloses in situ learning capability and competitive classification accuracy on a conventional machine learning dataset. Simulations suggest that increasing the number of hidden neurons would further improve classification accuracy. Ab initio molecular-dynamics simulations provide a clear insight into the melting kinetics in MMs and the structural origins of the prior complete melting-facilitated decrease in output conductance. The memristive neural network represents a potential hardware platform for artificial intelligence, offering high speed and energy efficiency.

Keywords

ab initio molecular dynamics simulations / amorphous materials / deep neural networks / in-memory computing / melting / phase-change memory materials

References

[1] Krauhausen I, Coen CT, Spolaor S, Gkoupidenis P, van de Burgt Y. Brain-inspired organic electronics: merging neuromorphic computing and bioelectronics using conductive polymers. Adv Funct Mater. 2024;34(15):2307729.

[2] Lu C, Meng J, Song J, et al. Self-rectifying all-optical modulated optoelectronic multistates memristor crossbar array for neuromorphic computing. Nano Lett. 2024;24(5):1667-1672.

[3] Zhao L, Fang H, Wang J, et al. Ferroelectric artificial synapses for high-performance neuromorphic computing: status, prospects, and challenges. Appl Phys Lett. 2024;124(3):030501.

[4] Yuan J, Wu C, Wang S, Wu F, Tan CK, Guo D. Enhancing plasticity in optoelectronic artificial synapses: a pathway to efficient neuromorphic computing. Appl Phys Lett. 2024;124(2):021101.

[5] Cao T, Liu C, Wang W, et al. A non-idealities aware software–hardware co-design framework for edge-AI deep neural network implemented on memristive crossbar. IEEE J Emerg Sel Top Circuits Syst. 2022;12(4):934-943.

[6] Dong B, Li P, Fu Y, et al. Study on memristor-based 3-D convolutional neural network adapting to non-ideal effects. IEEE Trans Electron Dev. 2023;70(3):993-1000.

[7] Chen N-K, Liu Y-T, Wang X-P, Li X-B. Nucleation dynamics of phase-change memory materials: atomic motion and property evolution. Phys Status Solidi Rapid Res Lett. 2021;15(3):2000441.

[8] Yamamoto T, Hatayama S, Song Y-H, Sutou Y. Influence of Thomson effect on amorphization in phase-change memory: dimensional analysis based on Buckingham’s П theorem for Ge₂Sb₂Te₅. Mater Res Express. 2021;8(11):115902.

[9] Attariani H, Wang W, Gałek R. A thermodynamically-consistent multi-physics framework for crystallization of phase-change material. J Cryst Growth. 2020;542:125687.

[10] Alam MS, Laing R, Bolatbek Z, et al. Fast cycling speed with multimillion cycling endurance of ultra-low loss phase change material (Sb₂Se₃) by engineered laser pulse irradiation. Adv Funct Mater. 2023;34(14):2310306.

[11] Mohseni M, Alkhayyat A, Srikaanth PB, et al. Analyzing characteristics for two-step SET operation scheme for improving write time in nanoscale phase-change memory (PCM). J Nanomater. 2022;2022(1):6822884.

[12] Gong N, Chien W, Chou Y, et al. A no-verification multi-level-cell (MLC) operation in cross-point OTS-PCM. In: Presented at 2020 IEEE Symposium on VLSI Technology. 2020.

[13] Skelton J, Loke D, Lee T, Elliott S. Structural insights into the formation and evolution of amorphous phase-change materials. Phys Status Solidi. 2013;250(5):968-975.

[14] Skelton J, Loke D, Lee T, Elliott S. Understanding the multistate SET process in Ge-Sb-Te-based phase-change memory. J Appl Phys. 2012;112(6):064901.

[15] Ren Y, Sun R, Chen SHY, Du C, Han S-T, Zhou Y. Exploring phase-change memory: from material systems to device physics. Phys Status Solidi Rapid Res Lett. 2021;15(3):2000394.

[16] Go S-X, Wang Q, Huang K, Lee TH, Bajalovic N, Loke DK. Toward single-cell multiple-strategy processing shift register powered by phase-change memory materials. Ad Intell Syst. 2023;5(3):2200353.

[17] Go S-X, Lim K-G, Lee T-H, Loke DK. Nonvolatile memristive materials and physical modeling for in-memory and in-sensor computing. Small Sci. 2024;4(3):2300139.

[18] Chen X, Ding F, Huang X, et al. A robust and efficient compact model for phase-change memory circuit simulations. IEEE Trans Electron Dev. 2021;68(9):4404-4410.

[19] Muñoz-Martin I, Bianchi S, Hashemkhani S, Pedretti G, Melnic O, Ielmini D. A brain-inspired homeostatic neuron based on phase-change memories for efficient neuromorphic computing. Front Neurosci. 2021;15:709053.

[20] Paolino C, Antolini A, Zavalloni F, et al. Decoding algorithms and HW strategies to mitigate uncertainties in a PCM-based analog encoder for compressed sensing. J Low Power Electron Appl. 2023;13(1):17.

[21] Kang S, Lee J, Kang M, Song Y. Achievement of gradual conductance characteristics based on interfacial phase-change memory for artificial synapse applications. Electronics. 2020;9(8):1268.

[22] Kang S-Y, Jin S-M, Lee J-Y, et al. Bidirectional electric-induced conductance based on GeTe/Sb₂Te₃ interfacial phase change memory for neuro-inspired computing. Electronics. 2021;10(21):2692.

[23] Narayanan P, Ambrogio S, Okazaki A, et al. Fully on-chip MAC at 14 nm enabled by accurate row-wise programming of PCM-based weights and parallel vector-transport in duration-format. IEEE Trans Electron Dev. 2021;68(12):6629-6636.

[24] Xie H, Zhang W, Zhou P, et al. A novel 1T-DRAM fabricated with 22 nm FD-SOI technology. IEEE Electron Device Lett. 2024;45(4):558-561.

[25] Xiao K, Liu J, Wu W, Xu Z, Wan J. Design of a novel one transistor-dram based on bulk silicon substrate. In: Presented at 2020 China Semiconductor Technology International Conference (CSTIC). 2020.

[26] Koyanagi M. History of MOS memory evolution on DRAM and SRAM. In: 75th Anniversary of the Transistor, Vol 187. 2023:187-213.

[27] Pan Y, Foster P, Serb A, Prodromakis T. A RRAM-based associative memory cell. In: Presented at 2021 IEEE International Symposium on Circuits and Systems (ISCAS). 2021.

[28] Xu W, Cho H, Kim Y-H, et al. Organometal halide perovskite artificial synapses. Adv Mater. 2016;28:5916-5922.

[29] Zhu X, Lu WD. Optogenetics-inspired tunable synaptic functions in memristors. ACS Nano. 2018;12(2):1242-1249.

[30] Bessonov AA, Kirikova MN, Petukhov DI, Allen M, Ryhänen T, Bailey MJ. Layered memristive and memcapacitive switches for printable electronics. Nat Mater. 2015;14(2):199-204.

[31] Zhang F, Zhang H, Krylyuk S, et al. Electric-field induced structural transition in vertical MoTe₂-and Mo_1–xW_xTe₂-based resistive memories. Nat Mater. 2019;18(1):55-61.

[32] Abdollahramezani S, Hemmatyar O, Taghinejad M, et al. Electrically driven reprogrammable phase-change metasurface reaching 80% efficiency. Nat Commun. 2022;13(1):1696.

[33] Boyn S, Grollier J, Lecerf G, et al. Learning through ferroelectric domain dynamics in solid-state synapses. Nat Commun. 2017;8(1):14736.

[34] Ryu H, Kim S. Implementation of a reservoir computing system using the short-term effects of Pt/HfO₂/TaO_x/TiN memristors with self-rectification. Chaos Solitons Fractals. 2021;150:111223.

[35] Yi W, Tsang KK, Lam SK, Bai X, Crowell JA, Flores EA. Biological plausibility and stochasticity in scalable VO₂ active memristor neurons. Nat Commun. 2018;9(1):4661.

[36] Oh S, Shi Y, Del Valle J, et al. Energy-efficient Mott activation neuron for full-hardware implementation of neural networks. Nat Nanotechnol. 2021;16(6):680-687.

[37] Rathi N, Chakraborty I, Kosta A, et al. Exploring neuromorphic computing based on spiking neural networks: algorithms to hardware. ACM Comput Surv. 2023;55(12):1-49.

[38] Go S-X, Jiang Y, Loke DK. A phase-change memristive reinforcement learning for rapidly outperforming champion street-fighter players. Adv Intell Syst. 2023;5(11):2300335.

[39] Go SX, Wang Q, Lim KG, Lee TH, Bajalovic N, Loke DK. Ultrafast near-ideal phase-change memristive physical unclonable functions driven by amorphous state variations. Adv Sci. 2022;9(36):2204453.

[40] Liu D, Yu H, Chai Y. Low-power computing with neuromorphic engineering. Adv Intell Syst. 2021;3(2):2000150.

[41] Vasilopoulou M, Mohd Yusoff ARB, Chai Y, et al. Neuromorphic computing based on halide perovskites. Nat Electron. 2023;6(12):949-962.

[42] Pei Y, Li Z, Li B, et al. A multifunctional and efficient artificial visual perception nervous system with Sb₂Se₃/CdS-core/shell (SC) nanorod arrays optoelectronic memristor. Adv Funct Mater. 2022;32(29):2203454.

[43] Roy S, Sridharan S, Jain S, Raghunathan A. TxSim: modeling training of deep neural networks on resistive crossbar systems. IEEE Trans Very Large Scale Integr Syst. 2021;29(4):730-738.

[44] Dong S, Chen Y, Fan Z, et al. A backpropagation with gradient accumulation algorithm capable of tolerating memristor non-idealities for training memristive neural networks. Neurocomputing. 2022;494:89-103.

[45] Park S-O, Park T, Jeong H, et al. Linear conductance update improvement of CMOS-compatible second-order memristors for fast and energy-efficient training of a neural network using a memristor crossbar array. Nanoscale Horiz. 2023;8(10):1366-1376.

[46] Joshi V, Le Gallo M, Haefeli S, et al. Accurate deep neural network inference using computational phase-change memory. Nat Commun. 2020;11(1):2473.

[47] Liu Z-C, Wang L. Applications of phase change materials in electrical regime from conventional storage memory to novel neuromorphic computing. IEEE Access. 2020;8:76471-76499.

[48] Loke D, Skelton JM, Law LT, et al. Guest-cage atomic interactions in a clathrate-based phase-change material. Adv Mater. 2014;26(11):1725-1730.

[49] Wang W, Shi L, Zhao R, et al. Nanoscaling of phase change memory cells for high speed memory applications. Jpn J Appl Phys. 2009;48(4S):04C060.

[50] Cao Z, Sun B, Zhou G, et al. Memristor-based neural networks: a bridge from device to artificial intelligence. Nanoscale Horiz. 2023;8(6):716-745.

[51] Feng J, Lotnyk A, Bryja H, et al. “Stickier”-surface Sb₂Te₃ templates enable fast memory switching of phase change material GeSb₂Te₄ with growth-dominated crystallization. ACS Appl Mater Interfaces. 2020;12(29):33397-33407.

[52] Zheng L, Song Z, Song W, Zhu X, Song S. Fabrication of stable multi-level resistance states in a Nb-doped Ge₂Sb₂Te₅ device. J Mater Chem C. 2023;11:3770-3777.

[53] Zhou Y-X, Zhang H-Y, Deringer VL, Zhang W. Structure and dynamics of supercooled liquid Ge₂Sb₂Te₅ from machine-learning-driven simulations. Phys Status Solidi Rapid Res Lett. 2021;15(3):2000403.

[54] Zhou Y, Zhang W, Ma E, Deringer VL. Unlocking device-scale atomistic modelling of phase-change memory materials. arXiv preprint arXiv:2207.14228. 2022.

[55] Li K, Liu B, Zhou J, Sun Z. Revealing the crystallization dynamics of Sb-Te phase change materials by large-scale simulations. J Mater Chem C. 2024;12(11):3897-3906.

[56] Abou El Kheir O, Bonati L, Parrinello M, Bernasconi M. Unraveling the crystallization kinetics of the Ge₂Sb₂Te₅ phase change compound with a machine-learned interatomic potential. npj Comput Mater. 2024;10(1):33.

[57] Yang W, Hur N, Lim D-H, Jeong H, Suh J. Heterogeneously structured phase-change materials and memory. J Appl Phys. 2021;129(5):050903.

[58] Xu Y, Zhou Y, Wang XD, et al. Unraveling crystallization mechanisms and electronic structure of phase-change materials by large-scale ab initio simulations. Adv Mater. 2022;34(11):2109139.

[59] Duong T-Q, Bouzid A, Massobrio C, Ori G, Boero M, Martin E. First-principles thermal transport in amorphous Ge₂Sb₂Te₅ at the nanoscale. RSC Adv. 2021;11(18):10747-10752.

[60] Daoudi O, Nolot E, Mazel Y, et al. The effects of Sb/Te ratio on crystallization kinetics in Ge-rich GeSbTe phase-change materials. J Appl Phys. 2024;136(15):155105.

[61] Konstantinou K, Elliott SR. Atomistic modeling of charge-trapping defects in amorphous Ge-Sb-Te phase-change memory materials. Phys Status Solidi Rapid Res Lett. 2023;17(8):2200496.

[62] Yakubov A, Sherchenkov A, Babich A, Lazarenko P, Sagunova I, Kirilenko E. Influence of the adjacent layers on the crystallization kinetics of Ge₂Sb₂Te₅ thin films. J Therm Anal Calorim. 2020;142(2):1019-1029.

[63] Wu X, Khan AI, Ramesh P, et al. Understanding interface-controlled resistance drift in superlattice phase change memory. IEEE Electron Device Lett. 2022;43(10):1669-1672.

[64] Zhang W, Ma E. Unveiling the structural origin to control resistance drift in phase-change memory materials. Mater Today. 2020;41:156-176.

[65] Boybat I, Kersting B, Sarwat SG, et al. Temperature sensitivity of analog in-memory computing using phase-change memory. In: Presented at 2021 IEEE International Electron Devices Meeting (IEDM). 2021.

[66] Gyanathan A, Yeo Y-C. Phase-change random access memory with multilevel resistances implemented using a dual phase-change material stack. IEEE Trans Electron Dev. 2012;59(11):2910-2916.

[67] Yan Z, Chen J, Hu R, Huang T, Chen Y, Wen S. Training memristor-based multilayer neuromorphic networks with SGD, momentum and adaptive learning rates. Neural Netw. 2020;128:142-149.

[68] Boquet G, Macias E, Morell A, Serrano J, Miranda E, Vicario JL. Offline training for memristor-based neural networks. In: Presented at 2020 28th European Signal Processing Conference (EUSIPCO). 2021.

[69] Kim M, Yoo K, Jeon S-P, Park SK, Kim Y-H. The effect of multi-layer stacking sequence of TiO_x active layers on the resistive-switching characteristics of memristor devices. Micromachines. 2020;11(2):154.

[70] Chen J, Li J, Li Y, Miao X. Multiply accumulate operations in memristor crossbar arrays for analog computing. J Semiconduct. 2021;42(1):013104.

[71] Amirsoleimani A, Alibart F, Yon V, et al. In-memory vector-matrix multiplication in monolithic complementary metal-–oxide–semiconductor-memristor integrated circuits: design choices, challenges, and perspectives. Adv Intell Syst. 2020;2(11):2000115.

[72] Yang X, Taylor B, Wu A, Chen Y, Chua LO. Research progress on memristor: from synapses to computing systems. IEEE Trans Circuits Syst I Regul Pap. 2022;69(5):1845-1857.

[73] Hasan R. Robust Ex-situ training of memristor crossbar-based neural network with limited precision weights. In: Presented at Proceedings of the 18th ACM International Symposium on Nanoscale Architectures. 2023.

[74] Bala A, Yang X, Adedotun A, Khandelwal S, Jabir A. Nanoscale Memristor Device and Circuits Design. Elsevier; 2024.

[75] Huang H-M, Wang Z, Wang T, Xiao Y, Guo X. Artificial neural networks based on memristive devices: from device to system. Adv Intell Syst. 2020;2(12):2000149.

[76] Shen J, Li T, Chen X, et al. Dynamic evolution of thermally induced element distribution in nitrogen modified phase change materials. J Appl Phys. 2020;128(7):075701.

[77] Hu C, Yang Z, Bi C, et al. “All-crystalline” phase transition in nonmetal doped germanium–antimony–tellurium films for high-temperature non-volatile photonic applications. Acta Mater. 2020;188:121-130.

[78] Loke DK, Skelton JM, Lee TH, Zhao R, Chong T-C, Elliott SR. Ultrafast nanoscale phase-change memory enabled by single-pulse conditioning. ACS Appl Mater Interfaces. 2018;10(49):41855-41860.

[79] Wang W, Loke D, Shi L, et al. Enabling universal memory by overcoming the contradictory speed and stability nature of phase-change materials. Sci Rep. 2012;2(1):360.

[80] Yang F, Tao Y, Zhang L, et al. Ab initio study on the fast reversible phase transitions of Ge₂Sb₂Te₅. J Appl Phys. 2021;130(2):025106.

[81] Choi Y-J, Ghim M, Jhi S-H. Study of vacancy ordering and the boson peak in metastable cubic Ge-Sb-Te using machine learning potentials. Phys Rev Mater. 2024;8(1):013606.

[82] Yang L, Cao B-Y. Thermal transport of amorphous phase change memory materials using population-coherence theory: a first-principles study. J Phys Appl Phys. 2021;54(50):505302.

[83] Cobelli M, Dragoni D, Caravati S, Bernasconi M. Metal-semiconductor transition in the supercooled liquid phase of the Ge₂Sb₂Te₅ and GeTe compounds. Phys Rev Mater. 2021;5(4):045004.

[84] Zhang J, Rong N, Xu P, et al. The effect of carbon doping on the crystal structure and electrical properties of Sb₂Te₃. Nanomaterials. 2023;13(4):671.

[85] Zheng L, Song W, Zhang S, Song Z, Zhu X, Song S. Designing artificial carbon clusters using Ge₂Sb₂Te₅/C superlattice-like structure for phase change applications. J Alloys Compd. 2021;882:160695.

RIGHTS & PERMISSIONS

2026 The Author(s). InfoScience published by UESTC and John Wiley & Sons Australia, Ltd.