Why in news?
A team from the University of Cambridge has developed a memristor that operates at extremely low currents and could reduce the energy consumption of artificial intelligence hardware by up to 70 percent. The innovation paves the way for more efficient neuromorphic computing and on‑device learning.
Background
A memristor is a fundamental circuit component whose resistance depends on the history of current that has passed through it. Proposed in 1971 and realised in the 2000s, memristors can store data and perform logic functions in the same device. They offer promise for brain‑like computing systems because their behaviour mimics synaptic plasticity – the way connections between neurons strengthen or weaken through use.
What’s new about this device?
- P‑n junction design: The researchers used a thin film of hafnium oxide doped with strontium and titanium to create microscopic regions resembling p‑type and n‑type semiconductors. Switching occurs by shifting the energy barrier between these regions rather than forming metal filaments, enabling smoother and more reliable operation.
- Ultra‑low power: The memristor can switch and hold multiple conductance states using currents a million times lower than those required by conventional devices, which dramatically cuts energy consumption.
- Multiple states: It supports dozens of stable conductance levels, allowing it to encode analogue information and implement synaptic learning rules such as long‑term potentiation and depression.
- Challenges: Fabricating the device involves high temperatures (around 900 °C), which may limit compatibility with existing silicon processes. Researchers are exploring ways to reduce the thermal budget.
Significance
- Energy‑efficient AI: Integrating such memristors into processors could lower the power demands of machine‑learning models used in smartphones, wearables and edge computing.
- Neuromorphic computing: By closely emulating synaptic behaviour, the devices could enable hardware that learns and adapts like the human brain, opening avenues for real‑time pattern recognition and adaptive control.
- Advances in materials science: The work shows how carefully engineered oxides can deliver new functionality in electronic devices, inspiring further research into non‑volatile memory elements.
Source: The Hindu
