MRS Meetings and Events

 

EL20.07.02 2023 MRS Fall Meeting

Characterization of Crystallization Kinetics in Phase-Change Materials using Indirectly Heated Phase-Change Memory and its Application to One-Shot Weight Transfer

When and Where

Nov 29, 2023
4:00pm - 4:15pm

Hynes, Level 3, Room 301

Presenter

Co-Author(s)

Inhyuk Choi1,Minseung Kang1,Wonseok Choi1,Sangbum Kim1

Seoul National University1

Abstract

Inhyuk Choi1,Minseung Kang1,Wonseok Choi1,Sangbum Kim1

Seoul National University1
Phase-change memory (PCM) exhibits notable advantages such as non-volatility, fast operating speed, scalability, and high reliability, making it a promising candidate for analog synaptic devices in neuromorphic computing. However, challenges related to nonlinearity and asymmetry in weight modulation need to be addressed [1]. Consequently, weight modulation predominantly relies on the gradual crystallization process, as amorphization through a melt-quench process induces abrupt changes in weight values [2]. In the context of off-chip training, concerns regarding nonlinearity and asymmetry can be alleviated since the weights pre-trained by the software can be accurately transferred to the PCM synaptic array using an iterative program and verify scheme [3]. However, it is noteworthy that this iterative scheme is not energy-efficient and can potentially accelerate the endurance failure of PCM devices.<br/>In contrast to conventional PCM structures that employ self-heating of phase-change materials, indirectly heated PCM structures utilizing externally generated heat have been suggested for high-reliability memory devices [4] and RF switches [5]. Here, for the first time, we propose the application of indirectly heated PCM as an energy-efficient and highly reliable one-shot weight transfer device through a single programming pulse with a simultaneous verify process. To ensure precise weight transfer, a comprehensive understanding of the crystallization process occurring within the device is crucial. Indirectly heated PCM itself can be used to characterize the crystallization kinetics [6], enabling faster heating rates close to device operation and broader temperature ranges encompassing melting points compared to previous characterization studies using differential scanning calorimetry (DSC), a laser with transmission electron microscope (TEM) analysis, and other device-based approaches [7-9]. After optimization through finite element method (FEM) simulation, we fabricated an indirectly heated PCM device designed to achieve a uniform temperature distribution in the phase-change material. Crystallization properties, including crystallization temperature, time, and activation energy, were measured over a wide temperature range for various phase-change materials exhibiting either nucleation-dominated or growth-dominated crystallization. For growth-dominant materials, growth rates spanning several orders of magnitude were estimated in conjunction with calibrated FEM simulations. Additionally, we will discuss the crystallization characteristics and device structures that need to be considered to enhance the energy efficiency and accuracy of one-shot weight transfer in indirectly heated PCM devices.<br/><br/>References<br/>[1] K. Byun et al., <i>Advanced Materials Technologies</i>, 2022, DOI: 10.1002/admt.202200884<br/>[2] M. Suri et al., <i>In 2011 International Electron Devices Meeting</i>, 2011, DOI: 10.1109/IEDM.2011.6131488<br/>[3] S. Yu, <i>Proceedings of the IEEE</i>, 2018, DOI: 10.1109/JPROC.2018.2790840<br/>[4] I. Choi et al., <i>Materials Science in Semiconductor Processing</i>, 2021, DOI: 10.1016/j.mssp.2021.105987<br/>[5] N. El-Hinnawy et al., <i>IEEE Electron Device Letters</i>, 2013, DOI: 10.1109/LED.2013.2278816<br/>[6] N. Wainstein et al., <i>IEEE Transactions on Electron Devices</i>, 2021, DOI: 10.1109/TED.2020.3048100<br/>[7] J. Orava et al., <i>Nature Materials</i>, 2012, DOI: 10.1038/nmat3275<br/>[8] M. Salinga et al., <i>Nature Communications</i>, 2013, DOI: 10.1038/ncomms3371<br/>[9] R. Jeyasingh et al., <i>Nano letters</i>, 2014, DOI: 10.1021/nl500940z<br/><br/>Acknowledgements<br/>This work was supported by SK Hynix Co., Ltd (0417-20200185).<br/><br/>*Corresponding Author : Sangbum Kim; E-mail: [email protected]

Keywords

crystallization | phase transformation

Symposium Organizers

Gina Adam, George Washington University
Sayani Majumdar, Tampere University
Radu Sporea, University of Surrey
Yiyang Li, University of Michigan

Symposium Support

Bronze
APL Machine Learning | AIP Publishing

Publishing Alliance

MRS publishes with Springer Nature