Two-dimensional MXene-based advanced sensors for neuromorphic computing intelligent application

As artificial-intelligence workloads explode, the energy cost and latency of shuttling data between discrete sensors, memory and processors have become critical bottlenecks. Now, researchers from the School of Integrated Circuits at Shandong University, led by Professor Jialin Meng and Professor Tianyu Wang, have published a forward-looking review on two-dimensional MXene materials that act simultaneously as ultra-sensitive sensors and neuromorphic synapses. This work charts a direct route toward self-powered, edge-intelligent systems that see, feel and smell their environment while learning on the spot.

Why MXene Sensors Matter
   • Energy Efficiency: Metallic MXene flakes operate at sub-volt biases and deliver femto-joule per spike consumption, eliminating the “sensor-to-processor” energy tax.
   • In-Sensor Computing: Tunable surface terminations enable real-time conductance modulation, allowing the sensor itself to perform weighted summation and activation—no external ADC or DRAM required.
   • Neuromorphic Applications: From retinomorphic vision chips to electronic skin and olfactory neurons, MXene devices emulate biological plasticity, enabling on-device training with 10 000× lower power than CMOS macros.

Innovative Design and Features
   • MXene Types: The review covers Ti₃C₂T_x, Mo₂TiC₂T_x, Nb₂CT_x and ordered i-MXenes, detailing how M-site chemistry controls bandgap (0–2.3 eV) and how T_x (–O, –OH, –F) tailors work-function alignment for photonic, piezo- and chemo-sensing.
   • Functional Materials: Ion-intercalated MXene/polymer aerogels, MXene/CNF textiles and MXene/quantum-dot heterojunctions are highlighted as plug-and-play layers that unify mechanical durability, breathability and sub-µs carrier dynamics.
   • Device Structures: 1T1R crossbars, 3D porous foams and transparent TFT arrays are introduced as scalable back-ends that convert light, pressure and gas stimuli directly into programmable synaptic currents.

Applications and Future Outlook
   • Multi-Level Storage: 21-bit analog states achieved in MXene memristors enable single-device synaptic weights, slashing array area and write energy versus 6-bit oxide RRAM.
   • Digital Logic Gates: MXene hetero-memristors implement IMPLY and NAND operations at <0.5 V, providing a path to fuse sensing, memory and logic in the same physical layer.
   • Artificial Synapses & Neurons: From vision (405 nm photodetectors, 0.5 µs response) to touch (46 730 kPa^-1 sensitivity, 20 Pa limit) and smell (415 % response to 0.2 ppm NO₂), MXene spiking nodes have already driven 784-pixel SNNs to 93 % MNIST accuracy at 0.8 pJ per inference.
   • Challenges & Opportunities: The review pinpoints oxidation under 80 % RH, CVD-scale growth below 400 °C and CMOS backend compatibility as urgent issues. Future work will target ALD Al₂O₃ encapsulation, roll-to-roll molten-salt exfoliation and hybrid CMOS-MXene 3-D stacks to vault MXene neuromorphics from lab demos to wearable, city-scale AI skins.

This comprehensive roadmap provides materials scientists, circuit designers and algorithm engineers with a common language for co-optimizing MXene sensors, synapses and systems. Stay tuned for more breakthroughs from Professor Jialin Meng and Professor Tianyu Wang at Shandong University!