WORKSHOPS

14:00 - 14:30

Switching Mechanisms in 2D Materials for Neuromorphic Computing Devices

Max C. Lemme (AMO GmbH and RWTH Aachen University, DE)

Two-dimensional (2D) Materials are considered for resistive switching devices and may be potential game-changers for neuromorphic computing hardware [1]. Such volatile or non-volatile ”memristive” devices may be used in computing-in-memory architectures [2], cross-bar arrays [3], or as electronic synapses [4]. Several different microscopic switching mechanisms are proposed in 2D materials, which are different from 3D materials because of the van der Waals layer stacking. Here, I will discuss different MoS2 device configurations that depend on the material growth conditions and lead to different resistive switching behavior and mechanisms, some of which I will show in detail [5–10].
[1] M. C. Lemme et al., Nat Comm 2022, 13, 1392. [2] D. Ielmini, H.-S. P. Wong, Nat Electron 2018, 1, 333. [3] S. Chen et al., Nat Electron  2020, 3, 638. [4] G. Zhou et al., Adv Electron Mat 2022, 8, 2101127. [5] D. Braun et al., in 2022 Device Research Conference (DRC), Columbus, Ohio, USA, 2022. [6] L. Völkel et al., Adv Fun Mat 2024, 34, 2300428. [7] S. Cruces et al., Small Sci n.d., 2400523. [8] K. Ran et al., 2025, DOI 10.48550/arXiv.2504.07980. [9] S. Cruces et al., 2025, DOI 10.48550/arXiv.2504.07979. [10] L. Völkel et al., 2025, DOI 10.48550/arXiv.2501.16359.

​

14:30 - 15:00​​​

Energy-efficient Hardware-Software Co-Optimization for Edge AI Devices

Sungju Ryu (Sogang University, KR)

Recently, generative AI has gained significant attention thanks to its powerful capabilities, but it requires substantially higher computational resources compared to conventional AI models. So, there are challenges in achieving real-time processing, which demands high-performance GPUs and expensive hardware resources. This presentation introduces efficient model compression techniques to operate generative AI on resource-constrained hardware, along with hardware design techniques optimized for such compressed AI models.

​​​​

15:00 - 15:30

Future of Energy Efficient Computing: Towards Ferromorphic Cognitive Chips

Adrian Ionescu (EPFL, CH)

As AI and IoT scale exponentially, hardware must evolve toward energy-efficient, sustainable designs. This talk highlights cutting-edge trends in low-power AI hardware for Edge applications—vital for billions of autonomous systems facing strict energy limits. These technologies will power breakthroughs in Industry 4.0, robotics, personalized healthcare, and environmental sensing. We present a brain-inspired approach using ferromorphic chips—a term coined at EPFL—featuring novel memristive devices that combine ferroelectricity in doped high-k dielectrics and 2D semiconductors to enable biologically plausible synapses, multi-bit memory, and on-chip energy storage.

​

15:30 - 16:00

Coffee break

​​

16:00 - 16:30

Reconfigurable and Ultrafast Non-Volatile Floating-gate Memory Based on van der Waals Heterostructures for Multifunctional Computing

Minh Chien Nguyen (Sungkyunkwan University, KR)

Floating gate memory dominates the non-volatile memory market due to its high density, stability, and low fabrication cost. However, its limited program speed and the physical separation between processing and memory units in the von Neumann architecture pose significant challenges for next-generation neuromorphic computing, which demands high-speed, low-power data processing. Two-dimensional (2D) materials offer promising solutions owing to their atomic thickness, high carrier mobility, and excellent electrostatic control, enabling ultra-scaled devices with minimized memory-logic distances. However, conventional 2D memory devices are typically limited by fixed carrier types, reducing their suitability for complementary in-memory computing architectures. In this work, we present a reconfigurable floating-gate memory (R-FGM) device based on WSe₂/h-BN/graphene van der Waals heterostructures. The device enables dynamic switching between n-type and p-type operation and demonstrates outstanding performance metrics, including a 30 ns program/erase speed, data retention exceeding 10⁴ s, endurance beyond 10⁵ cycles, and multi-level storage with up to 7-bit resolution. These results highlight its potential for multifunctional, high-performance neuromorphic computing systems.

​​

16:30 - 17:00

The Role of Point and Extended Defects in 2D-Materials Based Memristors for Neuromorphic Computing: An Atomistic Study

Mohit D. Ganeriwala (University of Granada, ES)

2D materials-based memristive devices are rapidly emerging as a groundbreaking solution for the next generation of memristive architectures for neuromorphic computing, driven by their exceptional properties—low power consumption, ultra-fast switching, compatibility with flexible substrates, and remarkable scalability through atomically thin layers. Despite these compelling advantages, the full potential of 2D materials-based memristors remains largely untapped, primarily due to performance variability arising from the stochastic nature of resistance switching and a lack of comprehensive understanding of the underlying physical processes. In this talk, I will share results from atomistic simulations that investigate the critical role of intrinsic point defects and grain boundaries in the formation of conductive channels. These findings offer new insights that challenge existing assumptions about memristive mechanisms, prompting a rethinking of the conventional models.2D materials-based memristive devices are rapidly emerging as a groundbreaking solution for the next generation of memristive architectures for neuromorphic computing, driven by their exceptional properties—low power consumption, ultra-fast switching, compatibility with flexible substrates, and remarkable scalability through atomically thin layers. Despite these compelling advantages, the full potential of 2D materials-based memristors remains largely untapped, primarily due to performance variability arising from the stochastic nature of resistance switching and a lack of comprehensive understanding of the underlying physical processes. In this talk, I will share results from atomistic simulations that investigate the critical role of intrinsic point defects and grain boundaries in the formation of conductive channels. These findings offer new insights that challenge existing assumptions about memristive mechanisms, prompting a rethinking of the conventional models.

​​​​​​​​​​

17:00 - 17:30

Multi-scale simulations of 2D material-based memristors

Marian Damiano (University of Pisa, IT)

Two-dimensional materials are promising candidates for memristive devices. In this talk, we present recent advancements on multi-scale simulations of 2D material-based memristors, combining ab-initio (first-principles) calculations of the materials' properties with quantum transport simulations. This approach provides insights into the physical mechanisms that govern the behavior of these devices.​​