Scientists built a memory chip that breaks the rules of miniaturization

2026-05-03 · Source: Robotics Research News -- ScienceDaily · Field: Technology & Digital — Emerging Technologies & Innovation, Artificial Intelligence & Machine Learning, Internet of Things (IoT) & Connected Devices · Depth: Novice, short

Summary

Scientists at the Institute of Science Tokyo (Science Tokyo) have developed a novel memory chip that improves performance as it shrinks, challenging a long-standing miniaturization barrier in electronics. This ferroelectric tunnel junction (FTJ) memory, measuring just 25 nanometers, utilizes hafnium oxide, a material known since 2011 to retain electric polarization even when extremely thin. The team, led by Professor Yutaka Majima, addressed nanoscale current leakage by further reducing device size and employing a new fabrication method that heats electrodes to form a semicircular, single-crystal-like structure. This design minimizes crystal boundaries, where leakage typically occurs, resulting in a highly efficient memory unit that could significantly reduce energy consumption and heat generation in future electronics, including smartphones, wearables, and AI systems.

Key takeaway

For AI Scientists designing next-generation hardware, this breakthrough in memory miniaturization suggests a viable path to significantly lower power consumption and enhanced processing speed. Your focus should shift towards integrating hafnium oxide-based ferroelectric tunnel junctions into existing semiconductor manufacturing processes, as this technology offers a direct route to more energy-efficient AI systems and extended device battery life without compromising performance.

Key insights

Extreme miniaturization combined with structural innovation can enhance memory performance, defying conventional scaling limits.

Principles

• Ferroelectric tunnel junctions offer low-power memory.
• Hafnium oxide maintains polarization at extreme thinness.
• Smaller can be better for specific nanoscale designs.

Method

The researchers fabricated 25nm ferroelectric tunnel junctions using hafnium oxide, employing a novel electrode heating method to create semicircular, single-crystal-like structures that reduce current leakage at crystal boundaries.

In practice

• Integrate hafnium oxide FTJs into semiconductor manufacturing.
• Develop ultra-efficient AI hardware with reduced energy demands.
• Extend battery life for smartwatches and IoT sensors.

Topics

• Ferroelectric Tunnel Junction
• Hafnium Oxide
• Nanoscale Memory
• Energy Efficiency
• Miniaturization Limits

Best for: AI Scientist, AI Hardware Engineer, Research Scientist, General Interest

Related on AIssential

• This new brain-like chip could slash AI energy use by 70% — A new hafnium oxide memristor design significantly cuts AI energy use by mimicking brain function.
• A better way to model the behavior of metal alloys — MIT researchers developed a motif-based sampling method using information theory to create diverse training datasets for accurate ML material simulations.
• Engineered van der Waals crystal mimics neuronal cells with light-driven learning — Engineered vdW crystals enable light-driven optoelectronic synaptic devices for brain-inspired computing.
• New magnetic chip design could outperform today’s AI accelerators — A new magnetic switching device offers 1,000x faster operation and lower energy than current AI accelerators.
• AI-powered spectrometer chip shrinks lab technology to the size of a grain of sand — A grain-of-sand-sized AI chip replaces bulky spectrometers by computationally reconstructing light spectra from specialized silicon detectors.
• New light-powered chip could accelerate AI and quantum computing — A new integrated photonic valleytronic chip processes light-based information at room temperature, enabling faster, energy-efficient computing.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Robotics Research News -- ScienceDaily.