Reset Variability in Phase Change Memory for Hardware Security Applications
Authors:
- Nafisa NoorNNNafisa Noor
- Sadid MuneerSMSadid Muneer
- Raihan Sayeed KhanRKRaihan Sayeed Khan
- Anna GorbenkoAGAnna Gorbenko
- L'HAcene AdnaneLAL'HAcene Adnane
- Tashfiq Bin KashemTKTashfiq Bin Kashem
- Jake ScogginJSJake Scoggin
- Faruk DirisaglikFDFaruk Dirisaglik
- Adam CywarACAdam Cywar
- Ali GokirmakAGAli Gokirmak
- Helena SilvaHSHelena Silva
- 30 November 2020
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nanotechnology
- Vol. 20 (1536125X) , 75-82
Abstract
The rapid quench from melt associated with the reset operation in phase change memory is a suitable physical process to leverage programming variability for hardware security. We have experimentally characterized cell-to-cell programming variability in Ge2Sb2Te5 (GST) line cells using a single reset voltage pulse with moderate amplitude. We have observed that the cell dimensions and programming pulse parameters play a critical role on the programming variability. Cells with larger length-to-width ratios (L/W 1.8) show greater programming variability when moderate reset voltage amplitudes (Vappl) are applied. This variability is further increased with shorter pulse edges (tedge) of the applied pulses. We have performed a study for a reset-variability-based physical unclonable function (PUF) with 85 cells (with L/W = 4) by applying identical single pulses with shorter tedge and a pre-calibrated Vappl. Cells are randomly programmed to the full reset (high-resistance-state or HRS), or the partial reset or set (low-resistance-state or LRS) states with an equal probability and the two states are separated by ~1.5 order of magnitude. No bit flipping is observed after 9 months of programming and instead, the separation increases to ~3 orders of magnitude because of the upward and downward resistance drift of the HRS and LRS states. The underlying physical mechanisms behind the enhanced variability are identified as thermal runaway at the onset of GST melting and the additional current overshoot due to the parasitic capacitances, based on the results obtained from COMSOL finite element simulations and SPICE circuit modeling of the experimental setup.Keywords
Funding Information
- Multi-University Research Initiative
- Air Force Office of Scientific Research (FA9550-14-1-0351Z)
- National Science Foundation
- Research Experience for Undergraduate (ECCS 1150960)
This publication has 18 references indexed in Scilit:
- Self-Heating Phase-Change Memory-Array Demonstrator for True Random Number GenerationIEEE Transactions on Electron Devices, 2017
- Emerging Physical Unclonable Functions With NanotechnologyIEEE Access, 2016
- Experimental Characterization of Physical Unclonable Function Based on 1 kb Resistive Random Access Memory ArraysIEEE Electron Device Letters, 2015
- High speed, high temperature electrical characterization of phase change materials: metastable phases, crystallization dynamics, and resistance driftNanoscale, 2015
- STT MRAM-Based PUFs2015
- Exploiting Process Variations and Programming Sensitivity of Phase Change Memory for Reconfigurable Physical Unclonable FunctionsIEEE Transactions on Information Forensics and Security, 2014
- PCKGen: A Phase Change Memory based cryptographic key generator2013
- Physically Unclonable Functions: A Study on the State of the Art and Future Research Directions2010
- Phase change memory technologyJournal of Vacuum Science & Technology B, 2010
- Impact of crystallization statistics on data retention for phase change memories2006