Dr. Stefan Wiefels , Forschungszentrum Jülich

From short- to long-term stability in VCM based RRAM

28.08.2026 (Friday) , 14:00 - 15:00

Abstract

Valence change mechanism (VCM) based resistive random-access memory (RRAM) exhibits intrinsic stochasticity arising from the motion of oxygen vacancies, which directly impacts short-term stability. In this work, short-term fluctuations are investigated across both industrial-scale and laboratory-scale devices. Mbit-scale VCM arrays integrated into 28 nm CMOS reveal characteristic read noise originating from a dynamic equilibrium of oxygen vacancy diffusion, which manifests as short-term instability but remains statistically stable across large device populations [1]. Complementary laboratory-scale studies demonstrate that the amplitude and spectral characteristics of read noise strongly depend on the underlying conduction mechanism of the switching oxide. Devices with shallow defect states exhibit high signal-to-noise ratios, while oxides with deep defect states show pronounced ionic noise driven by vacancy relocation [2]. Process parameters and read conditions are identified as key levers to influence short-term variability [3]. At the same time, this stochastic behavior can be purposefully exploited: VCM devices with deep defect states provide high-quality entropy sources for true random number generation, enabling single-step random bit extraction and passing NIST randomness tests [4].
Despite the pronounced instability on short time scales, the long-term stability is surprisingly high. It is found via accelerated life testing that long-term degradation predominantly affects the high-resistance state and manifests as a gradual shift and broadening of its resistance distribution [1]. This behavior cannot be reliably captured by conventional retention metrics and necessitates a statistical evaluation approach. By tracking the evolution of the distribution width, an activation energy for the degradation process is extracted, enabling lifetime prediction under operating conditions [5].
To consistently explain both long-term degradation and its link to short-term fluctuations, a three-dimensional kinetic Monte Carlo simulation model is introduced. By incorporating diffusion-limiting domains, the model reproduces experimentally observed retention loss and connects the extracted activation energies to oxygen vacancy migration barriers, providing a unified physical picture of stability in VCM-based RRAM [6].
[1] S. Wiefels et al., “Reliability Aspects of 28nm BEOL-Integrated Resistive Switching Random Access Memory,” physica status solidi (a), p. 2300401, 2023, doi: 10.1002/pssa.202300401.
[2] K. Schnieders et al., “Effect of electron conduction on the read noise characteristics in ReRAM devices,” APL Materials, vol. 10, no. 10, p. 101114, 2022, doi: 10.1063/5.0109787.
[3] K. Schnieders et al., “Read noise variability in thermally oxidized TaOx-based ReRAM devices”, 2026, submitted to IEEE Access.
[4] K. Schnieders et al., “Exploiting Read Noise of Filamentary VCM ReRAM for Robust TRNG,” IEEE Transactions on Electron Devices, vol. 72, no. 11, pp. 5988–5994, Nov. 2025, doi: 10.1109/TED.2025.3611916.
[5] N. Kopperberg, D. J. Wouters, R. Waser, S. Menzel, and S. Wiefels, “Accurate Evaluation Method for HRS Retention of VCM ReRAM,” APL Materials, vol. 12, no. 3, p. 031112, 2024, doi: 10.1063/5.0188573.
[6] N. Kopperberg, S. Wiefels, S. Liberda, R. Waser, and S. Menzel, “A Consistent Model for Short-Term Instability and Long-Term Retention in Filamentary Oxide-Based Memristive Devices,” ACS Applied Materials & Interfaces, vol. 13, no. 48, pp. 58066–58075, 2021, doi: 10.1021/acsami.1c14667.