Self-Assembly of Tunable Molecular Memristors with Long Range Order for Resilient and Energy-Efficient Neuromorphic Computing

Vision

Inspired by the power of the human brain and the extraordinary way neurons transmit information through synapses, the Pacific Northwest National Laboratory team, with support from partners at Lawrence Berkeley National Laboratory, the University of Washington, the University of Michigan, and Michigan State University, is working to develop a predictive understanding of polyoxometalate-peptoid-surface interactions needed to enable the co-design of transformative molecular memristors with controlled structure, chemistry, and long-range order for energy-efficient and resilient computing.

Approach

Natural biomaterials, such as proteins, carry out remarkable functions in energy-efficient and directional electron and ion transport that are necessary for neuromorphic computing. Developing the ability to design and synthesize resilient and energy-efficient artificial synapses based on crystalline hierarchical biomaterials with controlled structure, chemistry, and molecular patterns will offer a new platform for energy-efficient microelectronics and provide fundamental knowledge to advance technologies for neuromorphic memory and computing. AMMEC has the following objectives:

1.     Understand the electronic structures and energy levels of novel hybrid polyoxometalate-peptoid materials and establish structure-property relationships between peptoid sequences and the memristive states of polyoxometalates using advanced imaging techniques combined with multiscale theoretical simulations and machine learning methods.

2.     Synthesize crystalline assemblies of polyoxometalate-peptoid hybrids at interfaces with nanoconfined channels and long-range order, offering tunable local environments and electronic structures for memristive materials with multiple programmable states.

3.     Investigate how long-range order improves ion mobility, increases the signal-to-noise ratio of memristive states, and decreases energy consumption.

4.     Develop design principles to enhance energy efficiency, guide electron and ion transport, and control filament formation by assembling polyoxometalate-peptoid crystalline materials into memristor devices.

5.     Evaluate energy consumption and effectiveness for network implementation by analyzing the switching behavior of molecular memristors under various benchmarking programming conditions for neuromorphic computing.

Why it matters

By mimicking the energy-efficient and resilient processes found in nature, this research will lay the foundation for a new generation of neuromorphic memory and computing technologies. Artificial synapses based on crystalline biomaterials can drastically reduce energy consumption, improve performance, and increase resilience, creating transformative opportunities in microelectronics, computing, and beyond.