Percolating nanoparticle networks

In light of the global energy crisis, there is an urgent need for sustainable computing alternatives. Inspired by the efficiency of the human brain, neuromorphic computing has emerged as a promising paradigm. Within this context, Julien van der Ree explores percolating nanoparticle networks as potential building blocks for low-power, adaptive computation.
Van der Ree begins with the development of precise measurement setups, enabling controlled creation of networks at the percolation threshold and reliable in-operando electrical analysis. Using these methods, Van der Ree fabricated molybdenum nanoparticle networks that exhibit unprecedentedly low-voltage resistive switching in the millivolt range—over a thousand-fold improvement compared to previous reports. Structural analysis revealed stellate “porcupine-like” nanoparticles, whose sharp features concentrate local fields and enable stable switching, establishing molybdenum as a prime candidate material.
Building on this, scanning electron microscopy uncovered intensity contrasts arising not from morphology or composition but from electrically driven changes in connectivity. In situ correlative measurements demonstrated that these intensity fluctuations coincided with resistive switching events, providing new insight into network dynamics.
Van der Ree also investigated copper networks to benchmark against established materials such as tin, silver, and gold. However, copper required oxide stabilization and exhibited poor long-term stability, with irreversible melting and reconfiguration under bias. These findings clarified material requirements for functional networks and highlighted stability as a central challenge.
Finally, Van der Ree proposes new directions, including semiconducting, magnetic, and phase-change nanoparticles for multifunctional neuromorphic architectures. Together, the results establish molybdenum nanoparticle networks and novel characterization approaches as important advances, paving the way for energy-efficient neuromorphic computing.