Neuromorphic, Quantum, and Emerging Electronic Devices
Beyond the CMOS Paradigm
Why new devices are needed
As classical CMOS approaches energy, interconnect, and variability limits, alternative devices are being explored for memory, inference, sensing, and quantum control. The key idea is not merely smaller transistors, but new physical state variables and new computational primitives.
Emerging direction
Crossbar arrays, analog in-memory computing, and device-level nonvolatile states are being investigated to reduce the data-movement bottleneck in machine learning workloads.
Memristive and Neuromorphic Devices
Stateful computation
Memristive devices can store conductance states that emulate synaptic weights, enabling compute-in-memory architectures. However, device drift, stochastic switching, endurance limits, and nonlinearity complicate training and inference.
Which issue is especially important in memristive compute-in-memory systems?
Practical memristive devices often exhibit drift, variability, and stochastic switching that must be managed in hardware and algorithms.
Correct answer: State drift and stochasticity
What property of a memristive device makes it useful for in-memory computing?
A persistent resistance or conductance state can represent synaptic weights or stored analog values.
Correct answer: Nonvolatile conductance state
Frontier device families
Spintronic devices
- Use electron spin and magnetic order
- Potentially nonvolatile and fast
- Relevant for memory and logic
Superconducting circuits
- Ultra-low dissipation
- Require cryogenic operation
- Important for quantum control and ultrafast logic